237 50 37MB
English Pages 363 [369] Year 2023
Pseudo-Noun Incorporation and Differential Object Marking
O X F O R D ST U D I E S I N T H E O R ET I C A L L I N GU IST I C S General Editors David Adger and Hagit Borer, Queen Mary University of London
Advisory Editors Stephen Anderson, Yale University; Daniel Bu¨ring, University of Vienna; Nomi Erteschik-Shir, Ben-Gurion University; Donka Farkas, University of California, Santa Cruz; Angelika Kratzer, University of Massachusetts, Amherst; Andrew Nevins, University College London; Christopher Potts, Stanford University; Barry Schein, University of Southern California; Peter Svenonius, University of Tromsø; Moira Yip, University College London
RECENT TITLES 68 The Structure of Words at the Interfaces edited by Heather Newell, Máire Noonan, Glyne Piggott, and Lisa de Mena Travis 69 Pragmatic Aspects of Scalar Modifiers The Semantics-Pragmatics Interface by Osamu Sawada 70 Encoding Events Functional Structure and Variation by Xuhui Hu 71 Gender and Noun Classification edited by Éric Mathieu, Myriam Dali, and Gita Zareikar 72 The Grammar of Expressivity by Daniel Gutzmann 73 The Grammar of Copulas Across Language edited by Marı´a J. Arche, Antonio Fábregas, and Rafael Marı´n 74 The Roots of Verbal Meaning by John Beavers and Andrew Koontz-Garboden 75 Contrast and Representations in Syntax edited by Bronwyn M. Bjorkman and Daniel Currie Hall 76 Nominalization 50 Years on from Chomsky’s Remarks edited by Artemis Alexiadou and Hagit Borer 77 Majority Quantification and Quantity Superlatives A Crosslinguistic Analysis of Most by Carmen Dobrovie-Sorin and Ion Giurgea 78 The Grammar of the Utterance How to Do Things with Ibero-Romance by Alice Corr 79 The Derivational Timing of Ellipsis edited by Gu¨liz Gu¨nes¸ and Anikó Lipták 80 Negation and Negative Dependencies by Hedde Zeijlstra 81 Angles of Object Agreement edited by Andrew Nevins, Anita Peti-Stantic´, Mark de Vos, and Jana Willer-Gold 82 Pseudo-Noun Incorporation and Differential Object Marking by Imke Driemel
For a complete list of titles published and in preparation for the series, see pp. 349–351.
Pseudo-Noun Incorporation and Differential Object Marking I M K E D R I E M EL
Great Clarendon Street, Oxford, OX2 6DP, United Kingdom Oxford University Press is a department of the University of Oxford. It furthers the University’s objective of excellence in research, scholarship, and education by publishing worldwide. Oxford is a registered trade mark of Oxford University Press in the UK and in certain other countries © Imke Driemel 2023 The moral rights of the author have been asserted All rights reserved. No part of this publication may be reproduced, stored in a retrieval system, or transmitted, in any form or by any means, without the prior permission in writing of Oxford University Press, or as expressly permitted by law, by licence or under terms agreed with the appropriate reprographics rights organization. Enquiries concerning reproduction outside the scope of the above should be sent to the Rights Department, Oxford University Press, at the address above You must not circulate this work in any other form and you must impose this same condition on any acquirer Published in the United States of America by Oxford University Press 198 Madison Avenue, New York, NY 10016, United States of America British Library Cataloguing in Publication Data Data available Library of Congress Control Number: 2023930766 ISBN 978–0–19–286640–0 DOI: 10.1093/oso/9780192866400.001.0001 Printed and bound by CPI Group (UK) Ltd, Croydon, CR0 4YY Links to third party websites are provided by Oxford in good faith and for information only. Oxford disclaims any responsibility for the materials contained in any third party website referenced in this work.
Contents General preface Preface List of symbols and abbreviations
vii ix x
1. Introduction
1
2. Methodology and main results 2.1 Diagnostics 2.2 Elicitation methods, consultants 2.3 Main results
5 5 8 9
3. Pseudo-incorporation as a category change phenomenon 3.1 Sequential hybrids 3.2 Theoretical assumptions 3.3 Implementation 3.4 Sequential hybrids vs layered projections
13 13 15 26 36
4. Pseudo-incorporation vs differential object marking 4.1 Case loss is post-syntactic 4.2 PNI within post-syntactic DOM accounts 4.3 Case studies
39 39 50 57
4.3.1 4.3.2 4.3.3 4.3.4 4.3.5
Tamil Mongolian Turkish Korean German
5. PNI-property I: Restriction to low scope 5.1 Evidence for scopal inertness of verbal categories 5.2 PNI-ed arguments are restricted to the event domain 5.3 PNI-ed arguments reconstruct 5.4 Case studies 5.4.1 5.4.2 5.4.3 5.4.4 5.4.5
Tamil Mongolian Turkish Korean German
59 67 75 83 107
110 110 115 123 129 130 144 154 166 179
vi
CONTENTS
6. PNI-property II: Lack of binding and control 6.1 Tamil 6.2 Mongolian 6.3 Turkish 6.4 Korean 6.5 German
189 198 199 202 212 218
7. PNI-property III: Movement patterns 7.1 Tamil 7.2 Mongolian 7.3 Turkish 7.4 Korean 7.5 German
222 227 239 249 261 271
8. Differential object marking 8.1 Tamil 8.2 Korean
278 279 286
9. Previous approaches 9.1 Head movement accounts 9.2 DP/NP accounts 9.3 Raising accounts
292 292 303 316
10. Summary
321
References Language Index Subject Index
324 346 347
General preface The theoretical focus of this series is on the interfaces between subcomponents of the human grammatical system and the closely related area of the interfaces between the different subdisciplines of linguistics. The notion of ‘interface’ has become central in grammatical theory (for instance, in Chomsky’s Minimalist Program) and in linguistic practice: work on the interfaces between syntax and semantics, syntax and morphology, phonology and phonetics, etc. has led to a deeper understanding of particular linguistic phenomena and of the architecture of the linguistic component of the mind/brain. The series covers interfaces between core components of grammar, including syntax/morphology, syntax/semantics, syntax/phonology, syntax/pragmatics, morphology/phonology, phonology/phonetics, phonetics/speech processing, semantics/pragmatics, and intonation/discourse structure, as well as issues in the way that the systems of grammar involving these interface areas are acquired and deployed in use (including language acquisition, language dysfunction, and language processing). It demonstrates, we hope, that proper understandings of particular linguistic phenomena, languages, language groups, or inter-language variations all require reference to interfaces. The series is open to work by linguists of all theoretical persuasions and schools of thought. A main requirement is that authors should write so as to be understood by colleagues in related subfields of linguistics and by scholars in cognate disciplines. In recent years, investigations of the relationship between a verb and its object have begun to uncover rich patterns of complexity. While some objects maintain a fair amount of independence with respect to the selecting verb, others are morphologically and semantically bonded to it. The closest such bonding is full noun incorporation, where the verb and object form a single word, but there are looser dependencies and one of these, pseudo-noun incorporation, is the focus of this volume. In a careful, cross-linguistic exploration of the construction, Imke Driemel shows that it is distinct from full incorporation, and from the superficially similar Direct Object Marking phenomenon, and argues that it involves a categorial change in the object from a nominal to a verbal category. This categorial change explains patterns of movement, control, and binding in the syntax, but in addition feeds into the systems that interface with syntax to capture the particular semantic and morphological patterns that collocate cross-linguistically with pseudo noun incorporation. Overall, the book not only makes a strong argument
viii
GENERAL PREFACE
for a more complex typology of verb object dependencies in the syntax than is usually assumed, but also shows how to understand that complexity in terms of established syntactic possibilities. David Adger Hagit Borer
Preface This book is a revised version of my dissertation, filed in April 2020 at Leipzig University. The work was made possible to a large extent by the people who shared their language with me. I thank the speakers of the languages of this study, who invested time and energy into answering the many questions I had via skype, facebook, email, phone, online questionnaires, and in person—thank you all so much for your patience! My work also benefited from conversations I had with a number of linguists. Since there is significant overlap between these two sets, I will list them all together: Artemis Alexiadou, Sukhbat Baatar, Rajesh Bhatt, Bronwyn Bjorkman, Emily Clem, Rümeysa Dijle, Erdenekhishig Eldev-Ochir, Daniel Gleim, Dolgor Guntsetseg, Fabian Heck, Johannes Hein, Anke Himmelreich, Adiyasuren Jamiyandagva, Greg Kobele, Hyunjung Lee, Gereon Mu¨ller, Andrew ¨ zdemir, Murphy, Johannes Mursell, Jegan Murugesan, Yining Nie, Bilal & Fatos¸ O Manfred Sailer, Martin Salzmann, Ganzaya Sengee, Rajamathangi Shanmugam, Barbara Stiebels, Aravindhan Sukumar, Sandhya Sundaresan, Sergej Tatevosov, So¨ren Tebay, Gombosuren Tsermed, Philipp Weisser, Joanna Zaleska, and Malte Zimmermann. I have presented parts of this work at a number of conferences and departments, where I received valuable feedback. Thanks go to the audiences of the Workshop on Nominals at the Interfaces (2018), North East Linguistic Society (2019), Jahrestagung der Deutschen Gesellschaft für Sprachwissenschaft (2019), GLOW in Asia (2022), as well as the linguistics departments at Leipzig University, Humboldt-University Berlin, Potsdam University, Goethe University Frankfurt, and University of Pennsylvania. Finally, I would like to thank my family and friends, including Jasper, Kati, Jette, Tabea, Theresa, Siri, Jelena, Luise, Anne, Elton, Paula, and Maria for being supportive and accepting the increasingly workaholic lifestyle I have adopted over the past couple of years, while writing this book.
List of symbols and abbreviations Glosses 1,2,3 ABL ABS ACC ADD ADV ADJ AG AGR AOR ATTR AV CAUS CL CLIT CMPD COMP CONJ COP CV DAT DECL DEF DEM DET DOM DP DUR EMPH ERG EXIST F
1st, 2nd, 3rd person ablative absolutive accusative additive particle adverbial suffix adjectival suffix agent agreement aorist attributive agent voice causative classifier clitic compound complementizer conjunction copula converb dative declarative definite demonstrative determiner differential object marking discourse particle durative emphatic ergative existential feminine
FACT FAM FOC FUT GEN HAB HON INF IMP IMPF/IPFV IND INST INT INTR/INTRANS JOIN L LNK LOC M MIR MOD N NE NEG NMLZ NOM NPREF NSF NT OBJ OBL OV
factual familiar focus future genitive habitual honorific infinitive imperative imperfective indicative instrumental intimate intransitive epenthetic vowel L-suffix linking morpheme locative masculine mirative modality neuter nominal particle negation nominalizer nominative nominal prefix noun suffix neutral object oblique patient voice
LIST OF SYMBOLS AND ABBREVIATIONS PAT PERF/PFV PL POSS PROG PST PRS PUNC REFL REL
patient perfective plural possessive progressive past present punctual reflexive relative
RES S SBJ SG STAT T TR/TRANS TOP UNM VOL
resultative S-suffix subject singular stative tense transitive topic unmarked morphological case volitional
Abbreviations BIER DM DOM EC EF EI FA [•F•] [*F *] [•F•], [*F*], [F] GQ LF NI OT MIR PA PIC PF PM PNI QR SCC
xi
binder index evaluation rule Distributional Morphology differential object marking existential closure edge feature event identification functional application structure-building feature agree feature deactivated features generalized quantifier Logical Form noun incorporation Optimality Theory movement interpretation rule predicate abstraction Phase Impenetrability Condition Phonological Form predicate modification pseudo-noun incorporation quantifier raising Strict Cycle Condition
1 Introduction Pseudo-noun incorporation (PNI) describes a phenomenon in which an argument forms a ‘closer than usual’ relation with the verb. The syntactic consequence most often diagnosed is loss of case marking, potentially along with the lack of other functional material such as number marking and overtly expressed determiners. A correlating interpretive consequence is expressed by scope inertness. Both aspects are exemplarily shown for Hindi in (1), where (1a) shows that objects can be optionally marked for case and (1b) illustrates the correlation of case loss and a obligatory narrow scope reading. (1) Dayal (2011: 127,137)
Hindi
a. Anu bacca/bacce-ko sambhaaltii hai Anu child/child-ACC look.after.IMPF be.PRS ‘Anu looks after (one or more) children/the child.’ b. Anu bacca nahiiN samhaalegii Anu child not look.after.FUT ‘Anu will not look after children.’
¬∃, *∃¬
A common intuition many analyses share is that pseudo-incorporated arguments are somehow reduced in their syntactic as well as their semantic capacity. Together with the observation that pseudo-incorporation seems to be restricted to occur with bare nouns and indefinites, both case loss and scope inertness are often traced back to the size of the argument. Pseudo-incorporated arguments are claimed to be NPs, denoting properties ⟨e, t⟩, which do not require case and cannot take scope (Dayal 2011; Massam 2001; van Geenhoven 1998). Within the recent literature on differential object marking (Aissen 2003; Bossong 1991), this size restriction has also been argued to be the cause for lack of specificity/animacy interpretations (Kalin 2014; Levin 2019; López 2012; van Urk 2019b), shown here for Spanish in (2). One of the questions this study will address is whether pseudoincorporation and differential object marking can be considered as two sides of the same coin.¹
¹ I will use the term ‘differential object marking’ (DOM) in this book to subsume differential object marking and differential subject marking.
Pseudo-Noun Incorporation and Differential Object Marking. Imke Driemel, Oxford University Press. © Imke Driemel (2023). DOI: 10.1093/oso/9780192866400.003.0001
2
INTRODUCTION
(2) López (2012: 16)
Spanish
a. Marı´a busca a una gestora. Marı´a seeks DOM a manager ‘Maria is looking for a (certain) manager.’ b. Marı´a busca una gestora. Marı´a seeks a manager ‘Maria is looking for a (*certain) manager.’ This book pursues the idea that the effects of pseudo-noun incorporation are not related to size but to category. Pseudo-incorporated arguments transform from nouns into verbs during the course of the derivation. The verbal nature is responsible for the case drop and the inability to take wide scope: verbs are commonly taken to be incapable of inducing scope shift (Chomsky 2001; Harley 2004) and are cross-linguistically observed to constitute unsuitable hosts for case morphology (Moravcsik 2012; Nichols 1986). Two additional effects can be made to follow from the verbal character of pseudo-incorporated arguments. As has been observed for Turkish (Öztürk 2009), Hindi (Bhatt 2007), and Spanish (Leonetti 2004; López 2012), pseudoincorporation is not licensed in contexts that require the argument to act as a binder or controller. These properties are illustrated in (3) for Turkish. (3) Öztürk (2009: 343–344) a.
Turkish
Ali [c¸erc¸eve-sin-e1∕2 ] resm-ı1 /*resim1 koy-du. Ali frame-POSS.3SG-DAT picture-ACC/picture put-PST ‘A li picture-put in his frame.’
b. * Ali-yi [PRO1 sorgula-mak ic¸in] kasıtlı olarak polis1 tutukla-dı. Ali-ACC interrogate-INF for intentionally police arrest-PST ‘Police-arresting happened to Ali to interrogate him.’ The incompatibility of binding and control with pseudo-noun incorporation has been attributed to the need to raise into or be merged in a dedicated position where case is assigned and from which control and binding takes place (Bhatt 2007; López 2012; Öztürk 2009). The account put forward in this book argues that pseudo-incorporated arguments are incapable of binding a pronoun since binding, i.e. the ability to introduce an index, is a property essentially tied to nominal categories (Baker 2004; Büring 2005). Furthermore, control relations cannot be established since control is dependent on binding (Chomsky 1981; Manzini 1983). The categorial approach pursued in this book makes an additional prediction about the movement behaviour of pseudo-incorporated arguments, a property that has so far not received much attention, as it is known to vary across pseudonoun incorporation languages. Since pseudo-incorporated arguments turn from
INTRODUCTION
3
nouns into verbs, their movement patterns will mimick the movement pattern of VPs, predicting potential cross-linguistic variation across languages which exhibit pseudo-noun incorporation. The account presented here is, thus, well equipped to explain why languages like Tamil and Sakha do not permit movement of pseudoincorporated arguments (Baker 2014b), in contrast to Turkish (Gračanin-Yüksek and İşsever 2011; Jo and Palaz 2019) and Hindi (Dayal 2011), shown in (4) and (5). (4)
Dayal (2011: 137) Kitaabi anu zaroor __ i becegii. book Anu definitely sell-FUT ‘Anu will definitely sell books.’
Hindi
(5)
Jo and Palaz (2019: 23) Turkish Kitapi ben [Ali-nin __ i oku-dug˘-un]-u du¨¸su¨n-mu¨-yor-um. book I Ali-GEN read-NMLZ-POSS.3SG-ACC think-NEG-PRS-1SG ‘I don’t think that Ali does book-reading.’
The empirical basis of this book is formed by elicitation with speakers of Tamil, Mongolian, Korean, Turkish, and German. Five diagnostics are applied across eleven noun types in each of the languages under consideration, including bare nouns, numerals, weak and strong quantifiers, weak and strong definites, demonstratives, proper names, and pronouns. What emerges is a coherent effect on pseudo-incorporated arguments that maps loss of case marking to obligatory narrow scope, lack of binding and control relations, as well as cross-varying restrictions on movement. While case properties are regularly tied to scope and mobility restrictions by DP/NP accounts of pseudo-incorporation (Dayal 2011; Dobrovie-Sorin et al. 2006; Massam 2001), binding and control properties are often made use of by raising approaches (Bhatt and Anagnostopoulou 1996; Kelepir 2001; López 2012). The study provides a unifying theory that is able to capture all properties with a single assumption: pseudo-incorporation effects result from noun phrases that are made up of a nominal and a verbal category feature. Implemented in a minimalist derivational framework (Chomsky 1995, 2000), the nominal feature is active early in the derivation and in that responsible for c-selection and nominal modification, while the verbal feature is active late and crucially derives the effects we have come to recognize as pseudo-noun incorporation. In contrast to previous accounts, the current theory is able to explain cross-linguistic variation considering the extent to which pseudo-incorporated nouns are restricted in their movement capacities by demonstrating parallels to VP movement in each of the languages under consideration. Beyond identifying the core properties of the phenomenon, the book contributes a number of additional empirical observations, the most important being that optional case marking is not necessarily caused by pseudo-incorporation. Tamil and Korean provide evidence that only a subset of optionally case-marked
4
INTRODUCTION
noun types show a correlation with scope, binding, control, and movement constraints. This insight enforces the conclusion that one and the same language can make use of both pseudo-noun incorporation and differential object marking. Moreover, the broad range of noun types investigated in this study allows for important insights into the nominal domain generally, with a particular focus on bare argument languages. The study shows that Mongolian patterns with Turkish, which has previously been reported to allow optional case marking for weak quantifiers and numerals but not for strong quantifiers (Enc˛ 1991). Tamil, however, patterns with Hindi in that only bare nouns are permitted to undergo pseudonoun incorporation. The cross-linguistic differences support analyses that make use of the potential adjectival status of weak quantifiers, numerals, and indefinite articles (Ionin and Matushansky 2006; Milsark 1977; Partee 1988). Further valuable insights are also gained with respect to weak and strong definiteness (Schwarz 2009, 2019), in that case marking is essential to familiarity-based definites, in contrast to uniqueness-based definites. By shifting the scope to morpho-syntactic properties, the study provides a new direction for a research area that has so far primarily focused on the overt spell out of determiners in bare argument languages (Ahn 2017; Hanink 2018; Jenks 2015, 2018). The book is structured as follows. Chapter 2 justifies the diagnostics used in this study and discusses the elicitation methods, while also summarizing the main results. In Chapter 3, basic theoretical assumptions are laid out and applied to the main idea of this thesis, followed by a brief discussion of the literature on category change. Chapter 4 is dedicated to the most prominent syntactic diagnostic of pseudo-incorporation, that is loss of case marking in correlation with semantic effects. For each language, I will propose two ways to model case loss post-syntactically. Chapter 5 presents the scope facts in more detail and the implementation of the categorial approach in terms of the widely observed low scope restriction. In Chapter 6, I extend the proposal to control and binding facts, while Chapter 7 will present the cross-linguistic variation in the movement restrictions of pseudo-incorporated arguments. The two languages of this study that display optional case marking for noun types without correlating scopal behaviour will be given a detailed look in Chapter 8. Finally, Chapter 9 compares the results of this study to existing accounts of pseudo-noun incorporation and DOM and Chapter 10 concludes.
2 Methodology and main results This chapter presents the main results of the study, describes in detail the elicitation methods which were used, and justifies the diagnostics chosen to identify pseudo-noun incorporation.
2.1 Diagnostics Pseudo-noun incorporation is notoriously difficult to diagnose. Since there is no clear overt morpho-syntactic process of forming a complex noun–verb predicate, as is the case with noun incorporation, an interplay of factors is often taken into consideration when diagnosing PNI. A correlation which has frequently been argued to be indicative of pseudo-noun incorporation is the simultaneous restriction to low scope with the absence of case marking (Baker 2014b; Dayal 2011; Kwon and Zribi-Hertz 2006; López 2012; Massam 2001; Öztürk 2005). An example from Hindi was already provided in the introduction, where scopal properties were tested with respect to negation. In (1), we see another example from Spanish, including a universal quantifier. Case-marked objects often show flexible scope. If case marking is dropped, the existentials cannot take scope over negation (1) or a quantified subject (1). For similar scope interactions, see Kwon and Zribi-Hertz (2006: 118) for Korean, Kelepir (2001: 59) for Turkish, Lyutikova and Pereltsvaig (2013: 129) for Tatar, Lidz (2006: 14) for Kannada, and Testelets and Arkadiev (2014: 7) for Adyghe. (1) López (2012: 13,14) a. Todo hombre amo´ a una mujer. every man loved DOM a woman ‘Every man loved a woman.’ b. Todo hombre amo´ una mujer. every man loved a woman ‘Every man loved a woman.’
Spanish ∀∃, ∃∀
∀∃, *∃∀
While the link between low scope and case drop will be used as a primary diagnostic to identify pseudo-noun incorporation within my data set, we can furthermore observe that PNI-ed arguments lose the ability to act as binders or controllers. These interactions have been noticed previously for Spanish (Leonetti 2004; López Pseudo-Noun Incorporation and Differential Object Marking. Imke Driemel, Oxford University Press. © Imke Driemel (2023). DOI: 10.1093/oso/9780192866400.003.0002
6
METHODOLOGY AND MAIN RESULTS
2012), Turkish (Öztürk 2005, 2009), Tatar (Lyutikova and Pereltsvaig 2013), and Hindi (Bhatt 2007), see (2) and (3) for illustration. (2) Lyutikova and Pereltsvaig (2013: 130) a.
Tatar
Marat [balar-lar-nı]1 dzˇiba¨r-de [PRO1 uk-ırga]. Marat child-PL-ACC send-PST study-INF ‘Marat sent children to study.’
b. * Marat [balar-lar]1 dzˇiba¨r-de [PRO1 uk-ırga]. Marat child-PL send-PST study-INF ‘Marat sent children to study.’ (3) Öztürk (2009: 343)
Turkish
a. Ali c¸erc¸eve-sin-e1∕2 resm-i1 koy-du. Ali frame-POSS.3PS-DAT picture-ACC put-PST ‘A li put the picture in its/his frame.’ b. Ali c¸erc¸eve-sin-e*1∕2 resim1 koy-du. Ali frame-POSS.3PS-DAT picture put-PST ‘A li picture-put in his frame.’ The correlation between loss of case marking on the one hand and restriction to low scope as well as lack of binding and control on the other hand provides a robust package diagnostic to track down the types of arguments which can undergo PNI within the languages investigated in this study. Further support is provided if the class of nouns aligns with the cross-linguistic restriction to bare nouns and indefinites. An additional property PNI-ed arguments display is a peculiar movement pattern, one that is different from the case-marked counterparts. This diagnostic, however, should be considered with caution. Movement properties are certainly not homogeneous across languages. Yet there is a general tendency that PNI-ed arguments are more restricted in their movement capabilities than proper arguments. Consider Spanish again: while caseless arguments cannot be left-dislocated (Leonetti 2004; Melis 1995; Pensado 1995), see (4b), they can still precede the subject, shown in (5). Case-marked arguments, however, can additionally undergo left-dislocation. (4) Leonetti (2004: 86) a. Ya conocı´a (a) muchos estudiantes. already knew.1SG DOM many students ‘I already knew many students.’
Spanish
2.1 DIAGNOSTICS
7
b. *(A) muchos estudiantes ya los conocı´a. DOM many students already them knew.1SG ‘Many students I already knew.’ (5) Leonetti (2004: 97) Bueno, este verano, libros, han leı´do. well this summer books have.3PL read ‘Well, this summer, books they have read.’
Spanish
There are PNI accounts that solely rely on an adjacency requirement with the verb, either because case is not marked morphologically on arguments (Coon 2010; Levin 2015) or because case marking is not influenced by PNI (Frey 2015; Farkas and de Swart 2003). I will aim for a theory of pseudo-noun incorporation that can provide a global explanation for the absence of case, wide scope, binding, and control while also being able to account for cross-linguistic variation with respect to movement restrictions. In this study, we will take a detailed look at five languages which have been argued to show effects of pseudo-noun incorporation: Mongolian (Guntsetseg 2009, 2010, 2016), Tamil (Baker 2014b), Turkish (Enc˛ 1991; Öztürk 2005; von Heusinger and Kornfilt 2005), Korean (Ha. Lee 2006; Kwon and Zribi-Hertz 2006), and German (Frey 2015). In contrast to previous work, we will take a rather large number of different argument types into consideration and run each of them through the diagnostics discussed in this section. Two properties are always paired, for the most part we will investigate case loss on arguments and each of its correlations with scope, binding, control, and movement capabilities. German is the only language in the sample set which does not exhibit case drop. Nevertheless, it is argued to pseudo-incorporate bare nouns as well as indefinites, depending on the position in the clause. For German, therefore, we will take sentence position instead of case drop as a primary diagnostic and pair it with the other three. Besides bare nouns and indefinites, this study considers pronouns, proper names, demonstrative and possessor phrases, numerals, and quantifiers. Since the majority of the languages in the set qualify as bare argument languages, contexts were set up to probe for definiteness, thereby paying attention to the weak–strong distinction (Roberts 2003; Schwarz 2009). The upshot of this procedure is two-fold: (i) we can inspect the size restriction proposed for PNI languages across languages and thereby—if uniform—make a valuable contribution to the set of PNI diagnostics and (ii) we can tease apart case drop as a result of PNI from case drop being unrelated to PNI. The latter often falls into the category of differential object marking, a phenomenon which similarly centres around loss/addition of case marking but is traditionally linked to animacy, specificity, and definiteness constraints (Aissen 2003; Bossong 1991). Chapter 8 provides a detailed comparison and discussion for Korean and Tamil,
8
METHODOLOGY AND MAIN RESULTS
the two languages form the core data set which exhibit DOM as well as PNI effects depending on the argument type. Data from the five languages is primarly based on questionnaires and interviews run with native speakers of the respective languages. Two more languages, Spanish and Hindi, enter the picture for comparison’s sake. Data from those languages are taken exclusively from the literature. The results from Spanish and Hindi appear in the overview tables but are visually separated from the core data set and may contain gaps.
2.2 Elicitation methods, consultants The Mongolian data were elicited with the help of a questionnaire, put together via skype interviews with Dolgor Guntsetseg, a native speaker of Mongolian and a linguist who has done extensive work on the case system of Mongolian. The questionnaire was passed on to five native speakers via an online study. The consultants are speakers of Khalka Mongolian, between the ages of 28–49, and live and work in Ulaanbaatar, the capital of Mongolia. All speakers have lived and studied in Germany for at least three years and are now employed in an environment where the workplace language is German and therefore the contact language was German. The survey asked grammaticality judgements on a scale from 1 to 5. Scope and definiteness data were elicited with the help of written contexts as well as a series of pictures. Mongolian sentences were presented in phonemic transcription using Roman basic letters, adhering to Janhunen (2012) and Guntsetseg (2016).¹ The Tamil questionnaire was put together with the help of Jegan Murugesan, native speaker and linguist. Three additional native speakers served as consultants (two linguists and one non-linguist), ranging between the ages of 26–40. Elicitation was conducted through one-on-one interviews, either in person or via skype. Tamil sentences were provided in Roman script, largely following Lehmann (1993), which is based on the Madras University Tamil Lexicon. Again, scope and definiteness data were elicited with the help of written contexts. The contact language was English. Since the work on Korean is based on a collaboration with Hyunjung Lee, linguist and native speaker of Korean, the number of consultants vary to a large degree. The majority of the data was double-checked with a handful of Korean speakers (5–12). For the effects of subject PNI with respect to scope and definiteness, an online study was conducted with 54 participants, ranging from ages 16–65. Speakers were asked to judge the felicity of sentences on a scale from 1 to 5, given specific contexts. The contact language was Korean, using Hangul, the ¹ Romanization is increasingly used in electronic communication, while the orthography of Cyrillic Khalka often does not correspond to the phonemic reality of spoken Khalka (Janhunen 2012: 23).
2.3 MAIN RESULTS
9
Korean alphabet. Data representation in this book, however, will make use of the standardized Yale romanization system. A subset of the Korean data is also discussed in Driemel and Lee (2022). Turkish is already a well-studied language when it comes to case drop and its semantic effects (Aydemir 2004; Enc˛ 1991; Kelepir 2001; Kornfilt 1997, 2008; Öztürk 2005, 2009). Nevertheless, I consulted 2–3 native speakers, all linguists, for scope, control, binding, and movement effects of non-case marked subjects and objects. The contact language was English. Finally, I used my own native speaker judgements for German, while also consulting with a handful of additional native speakers.
2.3 Main results A summary of the study results is shown in Table 2.1. One question that this book attempts to answer is whether optional case marking of certain argument types always boils down to the same trigger. There seems to be significant overlap by now between the phenomena that have been classified as DOM and the cluster of properties which are often associated with pseudo-noun incorporation. DOMaccounts often tie case marking to specificity/definiteness, whereas PNI-accounts
Table 2.1 Case drop due to
PNI
or DOM
BA = bare argument; IND = indefinite; # = numeral; WQ = weak quantifier; DEMP = demonstrative phrase; WD = weak definite; PRO = pronoun; PN = proper noun; POSSP = possessor phrase; SQ = strong quantifier; SD = strong definite
10
METHODOLOGY AND MAIN RESULTS
try to account for the correlation between absence of case marking and low scope. Take Hindi for example: whereas the DOM strategy has been pursued by Butt (1993), Bhatt and Anagnostopoulou (1996), and Aissen (2003), a PNI account was given by Dayal (2003, 2011). For Spanish, López (2012) builds up a system which enables him to account for both, the pseudo incorporation of bare plurals and the differential object marking of indefinites. With the diagnostics set up in Section 2.1 we are able to draw a distinction between DOM and PNI. Only the latter will show an interaction between case drop on the one hand and scope, binding, control, and movement on the other. Diagnosing PNI for one type of argument, however, does not exclude the possibility that other types of arguments can be left caseless for other reasons. It seems reasonable to assume that another trigger for case loss is DOM, detectable e.g. via the absence of epistemically specific readings. Hence, this book actively pursues the idea that a language can make use of PNI as well as DOM, each correlating with different semantic properties.² Table 2.1 marks every noun type with if lack of case marking is generally an option, while cells which are marked as indicate a correlation of case drop with scope, binding, control, and movement, in other words a PNI pattern. If case-marking is optional, but only interacts with animacy or (epistemic) specificity, the cell is marked with . The current study reveals that Mongolian and Turkish display PNI effects throughout, while Korean and Tamil make use of DOM as well as PNI. A note on the selection of argument types is in order. The majority of the PNI/DOM literature discusses bare nouns, definite and indefinite noun phrases, proper names, and pronouns. I additionally include numerals, quantifiers, and demonstrative phrases in the study. Except for German, all languages from the core data set classify as bare argument languages, i.e. they do not have an overt definite determiner. Contexts that establish familiarity mention the referent in the previous discourse and test for strong definiteness. In contrast, weak definiteness can be established by a number of factors such as visual co-presence or global uniqueness in the cultural context (Heim 1982). The contrast between strong and weak definiteness received some attention within the last years, following the works by Roberts (2003) and Schwarz (2009). Jenks (2015, 2018) concludes, based on his work on Thai and Mandarin Chinese, that bare argument languages tend to use bare nouns for weak definite contexts, whereas demonstratives or overt pronouns take over the function of strong familiar definites. To my knowledge, no interaction has been reported so far between case marking and definiteness in bare argument languages. As Table 2.1 shows, strong anaphoric contexts always require case marking of the argument, while weak definiteness can also be licensed without case marking, e.g. in Korean and Tamil. While the optional case drop in Tamil
² Hence, this book aligns in spirit with the proposal by López (2012).
2.3 MAIN RESULTS
11
is a reflection of a DOM effect, Korean in fact pseudo-incorporates weak definites, next to 1/2 person pronouns as well as proper names, thereby providing an exceptional PNI pattern, within the sample set as well as in a more general context. A weak–strong distinction has also been observed for quantifiers (Kamp and Reyle 1993; Milsark 1977; Partee 1988; Zimmermann 2008), where the strong type refers to generalized quantifiers, whereas the weak type instantiates predicate modifiers which inherit their quantificational force from a covert c-commanding existential quantifier. This adjectival treatment has also frequently been proposed for numerals (Link 1987; Krifka 1999; Ionin and Matushansky 2006). Two languages out of the core data set are sensitive to the weak–strong distinction. For Turkish, the possibility to pseudo-incorporate weak quantifiers, including numerals, has not gone unnoticed in the literature (Enc˛ 1991; Kelepir 2001; Özyıldız 2017). The current study shows that Mongolian patterns with Turkish in this respect. Finally, it should be noted that the category indefinite (IND) can conflate with the category numeral (#), as indefinite determiners are often homophone with numeral one in many languages. Dayal (1999, 2004, 2017) classifies Hindi ek as a numeral, denying it any function as an indefinite determiner. It can be shown that Tamil and Mongolian behave like Hindi in this respect. As for Korean, BA and IND collapse into one category, whereas numerals clearly constitute their one category. As pertains to the size restriction, only bare arguments can be shown to robustly pseudo-incorporate across languages. Whereas Turkish and Mongolian extend the PNI class to indefinites, numerals, and weak quantifiers, German seems to include only indefinites additionally. It is reasonable to tie the possibility of numerals and weak quantifiers undergoing PNI to the adjectival nature of these elements. Languages then vary with respect to whether they draw a categorial distinction between weak and strong quantifiers. Korean, perhaps most surprisingly, pseudoincorporates weak definites, 1/2 person pronouns, and proper names—a selection which, seemingly random at first sight, will receive a uniform treatment in Chapter 4. The conclusions I draw for Hindi and Spanish are preliminary, based on the data available in the literature, hence the dashed line which separates the two languages from the core data set. Although the DOM cells in Table 2.1 do not show a particular size restriction, they do reveal general hierarchy effects which are traditionally associated with expected discourse prominence based on grammatical function (objects are typically low in animacy/definiteness, whereas subjects provide the opposite picture). The order of columns in Table 2.1 can be directly mapped onto an elaborate version of Aissen’s definiteness scale (2003: 437), where strong definites are ranked highest and most likely require to be case marked, while bare arguments constitute the low end of the scale. At this point, it should be noted that the table subsumes the PNI/DOM behaviour of different argument types independent of grammatical
12
METHODOLOGY AND MAIN RESULTS
function. Turkish and Korean make it possible to diagnose subject incorporation, revealing that subjects and objects show an identical restriction in terms of noun types. This behaviour serves as a counterargument against functional approaches to differential argument marking (Aissen 2003; Bossong 1991; Croft 1988; Silverstein 1976) which operate on the assumption that only arguments with definiteness/animacy properties prototypical for their grammatical function can be left unmarked, thus predicting differential subject marking to mirror the effects of differential object marking. This insight, however, is not particularly surprising, considering that similar points have been made for Turkish (Kornfilt 2008) and Hindi (de Hoop and Narasimhan 2008) in the past.³
³ This view does not exclude the possibility that PNI/DOM languages show asymmetric tendencies in terms of how frequent case is dropped on prototypical arguments. Corpus studies by Ha. Lee (2006, 2008) show that while Korean does not differentiate categorially between subjects and objects, it does so with respect to the likelihood of case markers being present and the definiteness/animacy status of the argument.
3 Pseudo-incorporation as a category change phenomenon This chapter presents the main idea of this book, it will be argued that pseudoincorporation constitutes a category change phenomenon. In order to implement this proposal, I motivate the hypothesis in Section 3.1, lay out my main assumptions in Section 3.2, and provide a concrete implementation in Section 3.3. The final section compares my proposal to more traditional mixed projections accounts.
3.1 Sequential hybrids Pseudo-incorporated arguments constitute hybrid categories: they are part nominal part verbal. The two core properties of pseudo-noun incorporation— lack of case marking and restriction to low scope—as well as the additional three diagnostics, inability to bind or control and restrictive movement capacities, can be traced back to its verbal nature. The properties PNI-ed arguments share with proper arguments reflect the nominal status, that is they check a c-selectional feature of the verb, they are assigned a θ-role, and they can come with number morphology as well as adjectival modification. Whereas the nominal properties are uncontroversial, the verbal properties require further explanation. In the following, I will discuss each property separately. One of the main reasons why the syntactic status of head movement is so fiercely debated in the recent past comes from the observation that verb movement never seems to change scopal relations. This has led many scholars, most prominently Chomsky (1995, 2000, 2001) with many following him (Boeckx and Stjepanović 2001; Harley 2004; Merchant 2002; Schoorlemmer and Temmerman 2012; Zwart 2017), to place head movement in general in the post-syntactic component. Others have argued for verb movement specifically to take place in syntax but with obligatory semantic reconstruction (Goldberg 2005; Keine and Bhatt 2016; Matushansky 2006). The scopal properties of PNI-ed arguments thus parallel those of verbs, in that they cannot take scope over another operator in the sentence. Often implicitly assumed amongst many scholars is that case is uniformly expressed on nouns and not on verbs; see, however, Blake (2004) and Pseudo-Noun Incorporation and Differential Object Marking. Imke Driemel, Oxford University Press. © Imke Driemel (2023). DOI: 10.1093/oso/9780192866400.003.0003
14
PSEUDO-INCORPORATION AS A CATEGORY CHANGE PHENOMENON
Moravcsik (2012) for explicit statements of such kind. This opinion finds empirical support in the work by Nichols (1986) who identifies case as the predominant morphology category for dependent marking strategies, whereas person, number, and gender morphology are most commonly expressed in head-marking patterns. An early implementation of this dichotomy can be found in the Principles and Parameters tradition of Generative Grammar where the lexical categories were distinguished by two binary distinctive features [±N] and [±V ], [−N] categories being case assigners and [+N] categories the ones which receive case (Chomsky 1981; Stowell 1981). The most recent instalment of case relations as exclusively tied to nominals is the theory of Dependent Case (Baker 2015; Marantz 1991; Stiebels 2006; Wunderlich 1997) where case marking expresses the licensing of one nominal in the local presence of another nominal. The lack of case marking on PNI-ed arguments is thus a direct consequence of their verbal nature. In a traditional framework such as Heim and Kratzer (1998) successful binding of a (reflexive) pronoun is conditioned by the introduction of a binder index, created by movement of a nominal phrase which is co-indexed with a pronoun. Thus, binding essentially constitutes a relation between nominals in A-positions (Chomsky 1981). Büring (2005) provides a more recent account of binding phenomena avoiding the necessary movement step via a special kind of abstraction rule. Nevertheless, the binder rule is restricted to apply between nominals. Finally, Baker (2004) determines the property to bear a referential index as the one identifying trait that separates nouns from verbs. Taken together, these assumptions made in the literature can provide an explanation why PNI-ed arguments are not capable of acting as binders, as it seems to be a property essentially tied to nominal categories. Moreover, the inability of PNI-ed arguments to enter a control relation is derived once it is assumed that the controller has to enter a binding relation with PRO (Chomsky 1981; Koster 1984; Landau 2015, 2017; Manzini 1983). While I have so far pointed out three properties that make PNI-ed arguments behave unlike nominals (case loss, lack of binding, and control) and one property which positively identifies their verbal character (scope), the movement behaviour is as well of the latter kind. Upon closer inspection, the cross-linguistic variation found with movement capabilities of arguments having undergone PNI parallel those of VPs in the respective languages. The nominal traits seem to be relevant early in the derivation—c-selection and θ-role assignment as well as noun phrase internal morphology and modification are operations which apply before or at the point the argument is first merged with the verb. The verbal properties, however, impact operations that are dependent on other arguments and functional heads in the clause. We will implement this observation by employing a derivational framework which is capable of turning a nominal category into a verbal category in the course of the derivation. Hence the name sequential hybrid.
3.2 THEORETICAL ASSUMPTIONS
15
3.2 Theoretical assumptions The proposal is worked out in a minimalist framework, first introduced in Chomsky (1995). Minimalism is a derivational model of grammar in which the basic operations Merge and Agree apply in sequential order. Syntactic structures are built bottom-up by sequential application of Merge and Agree from a set of lexical items, taken from the numeration. Syntactic operations are driven by two types of features: (i) structure-building features triggering Merge and (ii) probe features triggering Agree. They are written as [•F•] and [*F*], respectively. An empty box on a probe feature [*F:□*] signals that the probe is lacking a value. The notation is taken from Heck and Müller (2007) and Sternefeld (2006). While Merge forms complex structures by taking two lexical items (from the workspace or the numeration) forming a new item, Agree copies values from one element to another, where the former is called the goal, the latter the probe, and the latter c-commands the former (cf. Chomsky 2000, 2001). Movement is defined as internal Merge, where the item with the structure-building feature contains the to-be moved item. Finally, [•F•] and [*F*] must be targeted and discharged during the derivation, thereby restricting the possible outcome of syntactic derivations. The assumptions are summarized in (1)–(4). (1) Feature Condition i. Probes ([*F*]) participate in Agree. ii. Structure-building features ([•F•]) participate in Merge. iii. A feature [•F•] or [*F*] must be discharged by the end of the derivation. (adapted from Heck and Müller (2007: 2) and Müller (2011: 123)) (2)
Merge α can be merged with β, yielding {α,{α, β}}, if α bears a structure-building feature [•F•] and F is the label of β. (Heck and Müller 2007: 2)
(3)
Move Move is Merge, with β internal to α.
(Heck and Müller 2007: 2)
(4) Agree α can agree with β with respect to a feature bundle Δ iff (i) and (ii) hold. i. α bears a probe feature [*F*] in Δ and may thereby provide the α-value for a matching goal feature [F] of β in Δ. ii. α c-commands β. (adapted from Heck and Müller 2007: 2) If a head comes with more than one structure-building or probe feature and for each feature the context to apply is met, the Earliness Principle (Pesetsky 1989) demands that the syntactic operations they trigger either apply simultaneously or in a certain order. Recent analyses have made use of different order
16
PSEUDO-INCORPORATION AS A CATEGORY CHANGE PHENOMENON
relations, deriving feeding and bleeding interactions between Merge and Agree, either implicitly (Anand and Nevins 2006; Asarina 2011; Halpert 2012; Kalin and van Urk 2001) or explicitly by making reference to a feature list or feature stack (Assmann et al. 2015; Georgi 2014; Heck and Himmelreich 2017; Müller 2009, 2010, 2011; Murphy and Puškar 2018). I will follow Müller (2011) and assume that the features of one head constitute a list and will be discharged one after another, beginning with the first feature in the list. Furthermore, there are no syntactic operations which are not feature-triggered. Both conditions are stated in (5). (5) Last Resort i. Every syntactic operation must discharge either [•F•] or [*F*]. ii. Only features on the top of a feature list are accessible. (Müller 2011: 168) The last syntactic conditions I want to introduce are concerned with the time window, in which a syntactic operation is allowed to apply. The Strict Cycle Condition (Chomsky 1973), see (6), requires discharge to happen shortly after features enter the derivation, while the Phase-Impenetrability Condition (Chomsky 2000) restricts operations to apply between elements which are sufficiently close, see (7). I take vP and CP to be phases (Chomsky 2000, 2001). (6)
Strict Cycle Condition (SCC) Within the current XP α, a syntactic operation may not target a position that is included within another XP β that is dominated by α. Müller (2011: 120), based on Chomsky (1973: 235)
(7)
Phase-Impenetrability Condition (PIC) In phase α with head H, the domain of H is not accessible to operations outside of α, only H and its edge are accessible to such operations. (Chomsky 2000: 108)
Every probe and structure-building feature can only be targeted once (Heck and Müller 2007: 2). They get discharged, after they have undergone an operation, in order to make room for the next feature on the stack. The discharge has been argued to result in deletion, either immediately (Müller 2011: 123) or at the interface (Heck and Himmelreich 2017: 66). This assumption, however, creates problems in a realizational framework like Distributed Morphology (DM) (Halle and Marantz 1993, 1994) in which morphology inserts exponents into abstract morpho-syntactic feature structures post-syntactically. Deletion of e.g. valued case and agreement features would take away the feature contexts needed for late insertion to apply, thus resulting in a severe lack of inflectional morphology. This issue extends to LF-related features. If co-indexation is the result of an Agree relation between index features, as is the case e.g. in Heck and Himmelreich (2017) between a floating quantifier and a wh-phrase, feature deletion should be avoided
3.2 THEORETICAL ASSUMPTIONS
17
in order to create an interpretable LF. While a pre-syntactic approach to inflectional morphology, as proposed in Müller (2011: 168, fn. 8), is a viable solution for PF-relevant features, it is unclear how it can handle deleted features needed for LF. We will thus assume that features are not deleted but become inactive after they have taken part in a structure-building or Agree relation; see also Georgi (2014: 109); Müller (2009: 288), Müller (2010: 40).¹ Note that (1iii) does not make reference to goal features, they do not have to be discharged for the derivation to converge. They are, however, nevertheless discharged after they have taken part in an operation. Following a long tradition, I assume that structural case is assigned by the functional heads T and v (Chomsky 1995, 2000, 2001; Legate 2008; Pesetsky and Torrego 2001; Wurmbrand 2001). Since probes are standardly characterized as being unvalued (Chomsky 2001) and DPs arguably enter the derivation with an unvalued case feature, eventually being c-commanded by T and v, case assignment does not lend itself to be modelled by downward Agree, as it is defined in (4). Hence, case assignment is often viewed as a ‘by-product’ of Φ-feature agreement between an unvalued probe feature on v/T and the valued Φ-counterpart on DP (Chomsky 2001; Danon 2011; Georgi and Salzmann 2011; Heck and Richards 2010). Since DPs additionally come with an unvalued case feature, case is valued along the way by the already established Agree relation between the head and the DP. This view of case assignment circumvents the problem of unvalued case features being c-commanded by valued probes. A way to implement case assignment directly via Agree is by simply giving up the idea that probehood is directly linked to unvalued features (Assmann et al. 2014, 2015; Heck and Himmelreich 2017; Heck and Müller 2007). The unvalued case feature on an argument can thus serve as a goal, while v/T acts as a probe with a valued case feature, necessitating an additional diacritic (*F*) to mark it as a probe. Yet another way to avoid the reverse direction of probing is to give up on the notion of valuation altogether and instead assume a checking account of case assignment, in which both probe and goal enter the derivation with valued case features but Agree requires matching of features (Müller 2009, 2011). I adopt the second option, a direct valuation approach, although nothing in the present account hinges on that decision. A sample derivation for (8) is given in (9)–(14), focusing on selection and case assignment and ignoring Φ-feature agreement for the sake of simplicity. (8)
Every boy kissed John.
As a first step, V undergoes Merge with the direct object (9), to check its [•D•] feature, resulting in deactivation of both the structure-building feature [•D•] and the goal feature [D] on the object (deactivated [•N•] is the result of an earlier Merge ¹ Deactivated features are marked in grey: [•F•], [*F*], [F].
18
PSEUDO-INCORPORATION AS A CATEGORY CHANGE PHENOMENON
operation between D and NP). Subsequently, v merges with VP, thereby deactivating its own [•V•] feature and the goal feature [V ] on the VP, followed by valuation of the case feature on the DP (10).
VP
(9)
Vkiss D V
v′
(10)
v V *CASE:ACC* D v
DPJohn N D CASE:ACC
VP Vkiss D V ACC
DPJohn N D CASE:ACC
(11) builds up the complex subject by deactivation of the determiners structurebuilding feature [•N•]. The subject enters the derivation via Merge with vʹ in (12), in turn resulting in deactivation of vʹs other structure-building feature [•D•] as well as the subject’s [D] feature. (11)
DP Devery N D CASE:NOM D
vP
(12)
NPboy N
v′ DPevery boy N v VP D V CASE:NOM *CASE:ACC* DPJohn Vkiss D D N D v D V CASE:ACC
T merges with vP in (13) to check off its [•v•] and vʹs own goal feature, while also valuing the case feature on the subject.
T′
(13)
T−ed v *CASE:NOM* D T:PAST NOM
vP DPevery boy N D CASE:NOM D
v′ v V *CASE:ACC* D v
VP Vkiss D V
DPJohn N D CASE:ACC
Since T is equipped with another structure-building feature [•D•] and there is no DP in the numeration left to select, the subject DP is attracted to the specifier of TP, thereby deactivating [D] on DP and [•D•] on T (14). This step can be taken to realize the EPP requirement in English (Chomsky 1995; Lasnik 2002). Note that
3.2 THEORETICAL ASSUMPTIONS
19
the feature stack of the subject DP contains two [D] features, one for c-selection and one for movement, which is reminiscent of minimalist grammars (Kobele 2006; Stabler 1997, 2011). In principle, there is no limit with respect to the number of categorial goal features on a lexical item. Nothing prevents the feature stack of the object DP from containing more than one [D] feature. This possibility, however, will not enable the object to move to spec, TP. Object movement is blocked by the PIC, given in (7), and the assumption that vP constitutes a phase. Note also that DP-movement creates copies in the launching and landing site, where only the highest copy is spelled out (Chomsky 1993, 1995; Nunes 2004). (14)
TP DP1every boy N T−ed D v CASE:NOM *CASE:NOM* D D T:PAST
T′ vP ⟨DP⟩1
v′ v V CASE:ACC * * D v
VP Vkiss D V
DPJohn N D CASE:ACC
I adopt a post-syntactic view of inflectional morphology. Abstract feature stacks combine in the syntactic component and serve as the input for vocabulary insertion and other morphological operations on the way to the final Phonological Form (PF) which serves as an interface with the Articulatory-Perceptual system. I remain agnostic as to whether morphological operations interact based on rule ordering, as is the case in DM (Halle and Marantz 1993, 1994), or constraint ranking, employed in optimality-based systems (Kiparsky 2001; Trommer 2001). While the contexts for insertion are fully specified, vocabulary items for f-morphemes—that is pairs of phonological and morphosyntactic information—are often underspecified, leading to competition of forms. Vocabulary insertion is regulated by the Subset Principle (Halle 1997; Kiparsky 1973) preferring more specific compatible vocabulary items over less specific ones, where specificity increases with the number of compatible features, including contextual restrictions. Contra Marantz (1995); Harley and Noyer (1999, 2000) and with Harley (2015); Pfau (2000, 2009); Siddiqi (2006, 2009), I assume that roots (l-morphemes) are contentful, i.e. they specify the concept they refer to. Examples for roots and f-morphemes are given in (15) and (16), respectively. While there is only one vocabulary item for the subject John in (8), competition arises with respect to the insertion of [PAST]-VIs. Of the irregular tense forms, given in (16a) and (16b), which have access to the context
20
PSEUDO-INCORPORATION AS A CATEGORY CHANGE PHENOMENON
of adjacent l-morphemes, only (16b) is compatible. The elsewhere [PAST] marker (16c) is also compatible but less specific. √ (15) /ʤɑn/ ↔ John √ ⎧ hit ⎫ √ (16) a. /-∅/ ↔ [PAST] / Harley (2012: 2157) ⎨ run ⎬ ... ⎭ ⎩ √ ⎧ kiss ⎫ √ b. /-t/ ↔ [PAST] / ⎨ mean ⎬ ... ⎩ ⎭ c. /-d/ ↔ [PAST] Aside from the PF-branch, there is the Logical Form (LF), an interface to the Conceptual-Intentional system and the semantic representation of a syntactic structure. LFs consist of tree-structures but with logical types as categories. I employ standard model-theoretic compositional mechanisms for the interpretation of syntactic structures, using the typed λ-calculus along the lines of Heim and Kratzer (1998). Syntactic operations correspond to particular semantic operations: Merge is standardly interpreted as Functional Application (FA), whereas Move, i.e. internal Merge, results in a combination of FA and λ-abstraction. One (often covert) instance of internal Merge is Quantifier Raising (QR), originally motivated to derive scope ambiguities and to resolve a type clash created by quantifiers in object position (May 1977, 1985). The current proposal will make extensive use of QR since it is implemented in Neo-Davidsonian event semantics (Davidson 1967; Higginbotham 1985; Parsons 1995) and, following Landman (2000), quantifiers need to raise out of the event domain (vP) in order to be interpreted. Since quantifiers, like every other argument, are first merged inside vP, they obligatorily undergo Move to avoid a type clash. The relevant LF-structure and the necessary denotations for (8) are given in (17)–(23).² I follow Kratzer (1996, 2000), assuming that a verb denotes a relation between an event and their internal argument while external arguments, introduced by separate functional heads, are related to the verb indirectly by their θ-roles, see (18). Verbal projections denote predicates of events until an existential quantifier binds the event variable, a definition is given in (20). Sentences are thus propositions based on existential claims about events. In line with Parsons (1990: 6), existential binding results in scope which is ‘as narrow as possible’. Concretely, existential closure happens after all core arguments are merged in their base positions, in the spirit of Carlson (2003); Chung and Ladusaw (2004); Diesing (1992); Kratzer (1995). The VP meaning is obtained by applying the meaning of V to the meaning of the direct object via functional application 5 . ² Tense, aspect, and intensional semantics are not provided and, in fact, will be ignored throughout this book.
3.2 THEORETICAL ASSUMPTIONS
21
Combining VP and v requires an additional compositional rule Kratzer (1996) coins Event Identification (EI), defined in (21). Step 6 thus conjoins two functions, identifying two separate events as one, thereby chaining together different conditions on the same event. I consider v to be the functional head which introduces the external argument as well as a potential host for verbalizing and causative morphology (Chomsky 1995; Coon and Preminger 2013). However, I leave open the possibility that this head is split into v and voice, see Alexiadou et al. (2006); Harley (2013); Legate (2014); Pylkkänen (2008) for discussion. The result of EI is then applied to the external argument 7 : the lower copy of a generalized quantifier, replaced with a variable of type ⟨e⟩ whose referent is determined via an assignment function g : ℕ ↦ De , see (19). The variable is further up bound via the compositional rule traditionally referred to as Predicate Abstraction (Beck 1996; Heim and Kratzer 1998; Sauerland 1998). I will use Bu¨ring’s μ-operator (2001, 2004, 2005) to implement this rule, which is defined in (22) and executed in step 9 . Following (22a), the μ-operator is inserted immediately below the quantifier’s landing site, triggering MIR (22b), a movement interpretation rule which states that the sister of the μ-operator is to be interpreted relative to the modified assignment function g[tn ↦x] which is distinct from g only in that the movement index n is mapped to x.³ Before MIR happens, however, the event variable is existentially bound at the vP level via (20) in 8 , thereby marking the end of the event domain. Finally, the universal quantifier is interpreted in its landing site by functional application with the predicate created by MIR, see step 10 . 10
(17)
⟨t⟩ DPevery boy ⟨⟨e, t⟩, t⟩
9
⟨e, t⟩ μ1
vP ⟨t⟩ ⟨⟨v, t⟩, t⟩
8
vP ⟨v, t⟩
4
⟨DP⟩1 ⟨e⟩
7
v′ 6 ⟨e,⟨v, t⟩⟩
3
v 2 ⟨e,⟨v, t⟩⟩
VP ⟨v, t⟩ Vkiss 1 ⟨e,⟨v, t⟩⟩
5
DPJohn ⟨e⟩
³ I follow Büring (2001, 2004) in assuming that the domain of the assignment function g is sorted, i.e. indices found on lower copies are treated differently from indices found on pronouns. Bu¨ring uses this distinction to derive Weak Crossover effects, i.e. constellation where a constituent Ā-moves over a possessive pronoun without creating a bound reading.
22
PSEUDO-INCORPORATION AS A CATEGORY CHANGE PHENOMENON g
(18) ⟦
1
⟧ = λxλe[KISS(x)(e)]
⟦
2
⟧ = λxλe[AG(x)(e)]
(19) ⟦
3
⟧ = g(1)⁴
g
g
(20) Existential Closure EC := λP ∈ D⟨v,t⟩ .∃x ∈ Dv [P(x)]
4
(21)
Event Identification If α is of type ⟨e, ⟨v, t⟩⟩ and β is of type ⟨v, t⟩, then ⟦αβ⟧ = λxλe.⟦α⟧(x)(e) ∧ ⟦β⟧(e). (Kratzer 1996: 122)
(22) a. μ-insertion: (Büring 2005: 164,168) Move a wh-NP or QNP α to the clause initial position, leave a trace indexed n, and adjoin a trace binder prefix μn to the sister of α’s landing site:
⁴ This notation is a shorthand for the anaphoric definite description the lower copy of a quantifier turns into via Trace Conversion (Fox 1999, 2002; Sauerland 2004) to render it interpretable. The LF rule inserts a variable coindexed with the variable binder into the lower DP copy. Both index function and restrictor NP combine via Predicate Modification, while the quantifying determiner is replaced by a definite determiner. This determiner introduces a uniqueness presupposition and returns the maximal individual for which the restrictor predicate holds and which is identical to the individual identified by the index. ⟨DP ′ ⟩1 λ P. y[ BOY (y) → P(y)]
(i)
D every λ Q λ P. y[Q(y) → P(y)]
NP boy λ x[ BOY ( x)]
!z[ BOY (z)
LF
λP
⟨DP ′ ⟩1 z = g(1)] .ιz[ BOY (z)
D the !z[P(z)] .ιz[P(z)]
z = g(1)]
NP boy λ x[ BOY ( x) x = g(1)]
NP boy λ x[ BOY ( x)]
g(1) λ y[y = g(1)]
The movement interpretation rule maps the index to the λ-bound variable. The presupposition introduced by the definite determiner restricts the domain of the quantifier’s argument to boys, compare ’ to in (23). This presupposition is vacuously satisfied by the quantifier every boy, see ’. 9
9
10
g
(ii) ⟦ ’ ⟧ = λx : ∃!z[BOY(z) ∧ z = x].∃e[AG(ιz[BOY(z) ∧ z = x])(e) ∧ KISS( john)(e)] 9
g
⟦ ’ ⟧ = ∀y[BOY(y) → ∃e[AG(ιz[BOY(z) ∧ z = y])(e) ∧ KISS( john)(e)] Trace conversion explains how scope shifting movement does not bleed Condition C, since the restrictor NP is maintained in the launching site of the DP, which thus qualifies as an R-expression; see Chomsky (1993) and Fox (1999: 183). The LF rule furthermore accounts for interpretive effects with Antecedent-contained deletion (Sauerland 1998, 2004). Since this book does not focus on the interpretation of movement chains of generalized quantifiers, we will use the shorthand given in (19). 10
3.2 THEORETICAL ASSUMPTIONS
α
S ... α ... ...
LF
23
S* S
µn ...tn
... ...
b. Movement interpretation rule (MIR): g
⟦μn X⟧ = λx.⟦X⟧
(23) ⟦ ⟦ ⟦ ⟦ ⟦ ⟦
5
6
7
8 9
10
g[tn ↦x]
g
⟧ = λxλe[KISS(x)(e)](john) = λe[KISS(john)(e)]
FA
g
⟧ = λxλe[AG(x)(e)], λe[KISS(john)(e)] = λxλe[AG(x)(e) ∧ KISS(john)(e)]
EI
g
⟧ = λxλe[AG(x)(e) ∧ KISS(john)(e)](g(1)) = λe[AG(g(1))(e) ∧ KISS(john)(e)] g
⟧ = ∃e[AG(g(1))(e) ∧ KISS(john)(e)]
FA EC
g
⟧ = λx.∃e[AG(gt1 ↦x (1))(e) ∧ KISS(john)(e)] = λx.∃e[AG(x)(e) ∧ KISS(john)(e)]
MIR
g
⟧ = λP⟨e,t⟩ .∀y[BOY(y) → P(y)](λx.∃e[AG(x)(e) ∧ KISS(john)(e)]) = ∀y[BOY(y) → ∃e[AG(y)(e) ∧ KISS(john)(e)]
FA
I adopt a traditional approach to indefinites by analysing them as existential quantifiers (Frege 1892; Heim and Kratzer 1998; Russell 1905). There are at least two ways recently proposed in the literature to derive specific indefinites; one argues in favour of choice functions (Kratzer 1998; von Heusinger 2002), the other is based on indefinites constituting existential quantifiers (Onea and Geist 2011; Schwarzschild 2002). Specificity only plays a marginal role in this book since it does not seem to be a stable diagnostic for PNI-ed arguments, at least not for the languages discussed in this study.⁵ I consider specificity to be a relevant property for differential object marking. I follow Onea and Geist (2011) in analysing indefinite determiners as existential quantifiers where specific readings result from a domain narrowing mechanism in the sense of Schwarzschild (2002). This can be done by introducing a salient function into the restrictor of the existential quantifier, either by a pragmatic operator or encoded lexically. (24) provides the denotation for an indefinite determiner pragmatically enriched ⁵ I provide an overview at the beginning of Section 4.3.
24
PSEUDO-INCORPORATION AS A CATEGORY CHANGE PHENOMENON
with the domain restriction f(y) = x, where f is an anchoring function from individuals to individuals and y the anchor variable. (24) Indefinite determiner: ⟦a⟧ = λPλQ∃x[P(x) ∧ f(y) = x ∧ Q(x)]] a
f : λz.ιy[f(y, z)]
In (25), we see an example of relative specificity in which the speaker uses the indefinite with a specific reading while still being bound within the scope of another quantifier. It is not any woman every boy loves, rather what the speaker has in mind is that every boy loves his mother. (25) Every boy loves a woman. [Speaker’s intention: Every boy loves his own mother.] ⟦a woman⟧ = λQ∃x[WOMAN(x) ∧ f(y) = x ∧ Q(x)]] a
f : λz.ιy[MOTHER(y, z)]
Epistemic specificity occurs if the function variable is replaced by a constant, the speaker of the utterance, and intended reference is expressed explicitly in f, see (26). (26) Every boy loves a woman. [Speaker has one referent in mind] ⟦a woman⟧ = λQ∃x[WOMAN(x) ∧ f(y) = x ∧ Q(x)]] a
f : λz.ιy[z intends to refer to y](speaker)
Although specificity is not one of my diagnostics, it plays a role within the German PNI pattern. I therefore present a detailed analysis of specific indefinites for German in Section 5.4.5. As for definiteness, I adopt simplified versions of Schwarz’s (2009) weak-strong distinction between uniqueness and familiarity. Weak definites require referents to be unique, whereas strong definites have to refer to the (unique) familiar/salient individual in the context. Following Hawkins (1978), Schwarz (2009) identifies three relevant contexts, two of them license weak definites and one strong definites. German shows a morphological distinction, in that the weak determiner contracts with a preceding preposition, in contrast to the strong determiner. Anaphoric uses are shown in (27), where only the non-contracted version is acceptable. Immediate and global uses license only the contracted definite determiner, see (28) and (29).
3.2 THEORETICAL ASSUMPTIONS
(27)
25
Schwarz (2009: 30) German Hans hat einen Schriftsteller und einen Politiker interviewt. Er hat Hans has a writer and a politician interviewed He has #
vom / von dem Politiker keine interessanten Antworten from-theweak / from thestrong politician no interesting answers bekommen. gotten ‘Hans interviewed a writer and a politician. He didn’t get any interesting answers from the politician.’
(28)
Schwarz (2009: 39) Das Buch, das du suchst, steht im / the book that you look.for stands in-theweak /
#
German in dem in thestrong
Glasschrank. glass-cabinet ‘The book that you are looking for is in the glass-cabinet.’
(29)
Schwarz (2009: 40) German Armstrong flog als erster zum / # zu dem Mond. Armstrong flew as first.one to-theweak / to thestrong moon ‘Armstrong was the first one to fly to the moon.’
Schwarz (2009: 148,260) derives the anaphoric use of the strong determiner by making it require a second argument besides the NP, a pronoun which is equated with the referent. Both weak and strong determiners contain a uniqueness presupposition. I assume simplified versions of Schwarz’s strong and weak determiner. Denotations and structures are given in (30) and (31).⁶ (30) Definite determiners: a. ⟦wDι ⟧ = λP⟨e,t⟩ : ∃!xP(x).ιx[(P)(x)] b. ⟦sDι ⟧ = λP⟨e,t⟩ λy : ∃!x[P(x) ∧ x = y].ιx[P(x) ∧ x = y]
weak definite strong definite
⁶ Schwarz (2009) places his analysis in situation semantics, where situations are defined as parts of possible worlds (Kratzer 1989, 2019). In the original version, both determiners receive an additional domain restriction via Merge with a situation pronoun which either refers to a contextually salient situation, is identified with the topic situation via a syntactically present topic operator, or is bound by a quantifier over situations. Uniqueness is, thus, defined for individuals relative to situations. As far as I can see, the current proposal can be extended to such a system. I will, however, restrict the current proposal and make use of extensional semantics only.
26 (31)
PSEUDO-INCORPORATION AS A CATEGORY CHANGE PHENOMENON Definite determiners:
DP wDι NP ⟨⟨e, t⟩, e⟩
DP g(i)
D′ sDι NP ⟨⟨e, t⟩, ⟨e, e⟩⟩
The semantic set up, introduced in this section, forces quantifiers to QR out of the event domain. A trademark property of pseudo-incorporated arguments, however, is that they can only be interpreted in their base positions. The next section lays out the major components of the analysis, tying among other things the scope properties to the verbal nature of PNI-ed arguments.
3.3 Implementation In Section 3.1, I argued that pseudo-incorporation effects can be traced back to the the categorial status of PNI-ed arguments. They enter the derivation as nominals but they transform into a verbal category over the course of the derivation. A feature-driven framework, as it was introduced in Section 3.2, is capable of providing a concrete theoretical proposal of this idea. PNI-ed arguments are the result of a special D head, a feature list that contains a categorial [D] feature as well as a categorial [V] feature. The features are ordered so that [D] is relevant for processes early in the derivation, i.e. c-selection and θ-role assignment, while [V] is at work for later operations such as case marking, binding, control, and movement. I call this determiner PNI-D for the lack of a better term. The syntactic structures in (32)–(34) provide an exemplary illustration of a subject undergoing pseudo-incorporation. The verbal domain is head-final since this is the head directionality predominantly present in the data set under investigation. (32) presents the internal feature structure of a PNI-ed argument, while (33) sets up the rest of the vP so that the PNI-ed subject can enter the derivation. Nothing in (32) prevents the NP from receiving number morphology or adjectival modification. Note that the feature lists comprising the functional heads v and V do not change, whether the arguments are PNI-ed or not (compare the sample derivation from Section 3.2). The crucial difference lies in the hybrid set-up of PNI-D. Since [D] is ordered higher on the feature stack than [V], PNIDPs are c-selected like DPs, shown in (34). PNI-DPs and DPs are, however, of different categorial status once they are first-merged with vʹ to form vP. This difference now accounts for many of the PNI effects pointed out in Section 3.1.
3.3 IMPLEMENTATION
PNI-DP N
(32)
v′
(33)
v V *CASE:ACC* D v
VP
D
DP N D CASE:ACC D
CASE:NOM V NP
PNI-DP N
27
V D V ACC
N
D CASE:NOM V
vP
(34)
v′
PNI-DP N D CASE:NOM V
v V
VP DP N
V D
D CASE:ACC
V
*CASE:ACC* D v
D
Let us start with binding and control by considering a concrete example. Constellations like (34) in which the object DP constitutes a reflexive pronoun are ungrammatical. An example for (34) is provided in (35b).
(35) a.
Koyangi-ka1 [ku casin-ul]1 halth-ass-e. cat-NOM 3SG self-ACC lick-PST-INT ‘A cat washed itself.’
b. *Koyangi1 [ku casin-ul]1 halth-ass-e. cat 3SG self-ACC lick-PST-INT ‘A cat washed itself.’
Korean
28
PSEUDO-INCORPORATION AS A CATEGORY CHANGE PHENOMENON
Following Büring (2001, 2004, 2005), a binding relation is established between a pronoun and a DP⁷ by insertion of a binder prefix at LF, shown in (36a). The DP minimally c-commanding the binder prefix acts as the semantic binder for the pronoun via the Binder Index Evaluation rule, see (36b). This rule is very similar to MIR, the only difference being that the variable to which the index is mapped is additionally also the individual argument to which the sister of the binder prefix is applied. Whereas MIR does not saturate an argument position, BIER does. (36) a. β-insertion:
DP XP
(Büring 2001: 56, Büring 2004: 25) LF
DP
XP
βn XP where n is an index, and DP occupies an A-position b. Binder Index Evaluation rule (BIER): g g[n↦x] ⟦βn XP⟧ = λx.[⟦XP⟧ (x)] Büring (2005: 1) explicitly states that binding theory is ‘relevant for nominal categories only’. I take this to imply that insertion of a binder prefix is blocked if the antecedent is not nominal. This is exactly the scenario we find in (34). Since the anaphor needs to be bound and there is no other antecedent available, (35b) is ungrammatical. (35a) in contrast allows for insertion of a binder prefix since the subject is nominal. Not only does PNI block binding, it also prevents any kind of control relation between a PNI-ed argument and PRO. We can extend the argumentation made for binding straightforwardly to control structures, under the assumption that the controller needs to establish a binding relation with PRO, see Chomsky (1981); Koster (1984); Landau (2015, 2017); Manzini (1983) for such proposals. Further details and language specific properties are worked out in Chapter 6. In Section 3.1, I reported on two different routes that have been taken to derive the fact that verbs never take wide scope, one is to situate head-movement on PF, the other is to place head movement in syntax proper but assume obligatory reconstruction for lexical verbs. My proposal will be developed along the lines of the latter accounts. The hybrid determiner PNI-D denotes an existential quantifier, deriving the indefinite reading PNI-ed arguments cross-linguistically receive. A promising way to restrict the scope of PNI-ed arguments is to enable interpretation only in the event domain. Under the assumption that syntactic categories are related to semantic types in a meaningful way, I propose that verbal features are ⁷ Büring (2005: 130) reformulates the binder rule with NPs instead of DPs. He does, however, point out that choosing NPs over DPs is nothing more than a preference (Büring 2005: 1).
3.3 IMPLEMENTATION
29
intrinsically linked to events, whereas e.g., nominals are linked to individuals, CPs are linked to worlds etc. (see also von Stechow 2012). The denotation for PNI-D, given in (37), ensures that PNI-ed arguments can be interpreted in their base positions due to the inclusion of an event variable in the quantifier’s scope argument, which enables PNI-DP to combine with V or v directly. (37)
⟦PNI-D⟧ = λP⟨e,t⟩ λQ⟨e,⟨v,t⟩⟩ λe∃z[P(z) ∧ Q(z)(e)] ⏟⏟⏟ ⏟⎴⏟⎴⏟ V∕v
NP
Let us use Korean again and illustrate with an example. (39) and (40) provide the semantic computation for (38b), where the object undergoes PNI, visible via case drop. The hybrid determiner PNI-D first combines with its restrictor NP, similar to a generalized quantifier 2 . The second argument of the quantifier is V, again combining via functional application 4 . VP and v undergo event identification in 6 , followed by functional application to the external argument 7 . Finally, existential closure applies in 8 since all core arguments have been introduced. (38) a. Yusu-ka kkoch-ul sasse. Yusu-NOM flower-ACC bought. ‘Yusu bought a flower.’
Korean
b. Yusu-ka kkoch sasse. Yusu-NOM flower bought. ‘Yusu bought a flower.’ (39)
vP⟨t⟩
8
vP⟨v,t⟩
7
DP⟨e⟩ Yusu
v'⟨e,⟨v,t⟩⟩ VP⟨v,t⟩
PNI-DP ⟨⟨e,⟨v,t⟩⟩, ⟨v,t⟩⟩
v⟨e,⟨v,t⟩⟩
4
V ⟨e,⟨v,t⟩⟩ buy
2
NP⟨e,t⟩ PNI-D ⟨⟨e,t⟩,⟨⟨e, ⟨v,t⟩⟩, ⟨v,t⟩⟩⟩ flower
6
1
3
5
30
PSEUDO-INCORPORATION AS A CATEGORY CHANGE PHENOMENON
(40) ⟦
1
⟧ = λy[FLOWER(y)]
⟦
2
⟧ = λQ⟨e,⟨v,t⟩⟩ λe∃z[λy[FLOWER(y)](z) ∧ Q(z)(e)] = λQ⟨e,⟨v,t⟩⟩ λe∃z[FLOWER(z) ∧ Q(z)(e)]
⟦
3
⟧ = λxλe[BUY(x)(e)]
⟦
4
⟧ = λe∃z[FLOWER(z) ∧ λxλe[BUY(x)(e)](z)(e)] = λe∃z[FLOWER(z) ∧ BUY(z)(e)]
⟦
5
⟧ = λxλe[AG(x)(e)]
⟦
6
⟧ = λxλe∃z[FLOWER(z) ∧ BUY(z)(e) ∧ AG(x)(e)]
⟦
7
⟧ = λxλe∃z[FLOWER(z) ∧ BUY(z)(e) ∧ AG(x)(e)](yusu) = λe∃z[FLOWER(z) ∧ BUY(z)(e) ∧ AG(yusu)(e)]
⟦
8
⟧ = ∃e∃z[FLOWER(z) ∧ BUY(z)(e) ∧ AG(yusu)(e)]
(37), FA LR FA LR EI:(21) FA LR EC:(20)
The semantic type of the PNI-DP prevents it from being interpreted outside the event domain, i.e. above existential closure in 8 . If the PNI-ed argument scrambles out of vP, MIR will not produce the right type that can serve as input to the quantifier in its landing site. Now compare this to the treatment of generalized quantifiers, as they were discussed with regard to (17). Whereas generalized quantifiers can only be interpreted outside the event domain, PNIDPs can only be interpreted inside the event domain. The low scope restriction of PNI-ed arguments thus follows from the denotation given in (37). We can extend this explanation to scope scenarios involving negation by assuming with many others (Chung and Ladusaw 2004; Herburger 2002; Penka 2010; von Stechow 1993, 2012; Swart 2016; Zeijlstra 2004; Zimmermann 2007) that negation applies not lower than existential closure in 8 . This account naturally extends to subject PNI, as v and V are of the same type. More details will be given in Chapter 5.⁸ Both verbal heads and PNI-ed arguments are alike in that they introduce an event variable. With the implementation given above, we additionally treat PNIed arguments similarly to verbs in that they have to reconstruct into their base positions (Keine and Bhatt 2016; Matushansky 2006). While this is true for verbs in general, PNI-ed arguments have to do so only if they move out of vP. This requirement, however, suffices to make them scope under other quantified arguments and negation, given the system set up above. Although PNI-ed arguments ⁸ The result of the computation in (40) is very similar to lexical accounts of incorporation (van Geenhoven 1998) and pseudo-incorporation (Dayal 2011; Dobrovie-Sorin et al. 2006; Espinal and McNally 2011) that situate existential closure of the argument slot within the lexical denotation of verbs. While such accounts have to assume separate incorporation denotations for each verb which can undergo PNI, the present account locates the low scope restriction in the presence of one lexical item only. See Chapter 9, especially Section 9.2, for more discussion.
3.3 IMPLEMENTATION
31
cannot be interpreted outside of vP, they can still move to vP-external positions. All five languages of the core data set allow for scrambling of proper arguments, while there is cross-linguistic variation with respect to the restrictions on scrambling of PNI-ed arguments. Chapter 7 will establish that movement patterns of PNI-ed arguments parallel those of VP-movement in the respective languages. We will be concerned with scrambling mainly (and topicalization for German), where movement is triggered by categorial features on designated functional heads. DPmovement is often argued to be derived by a categorial feature [•D•], be it for scrambling generally or for EPP-movement and object shift specifically (Alexiadou and Anagnostopoulou 1998; Bailyn 2003; Chomsky 1995; Epstein et al. 1998; Kishimoto 2000; Kitahara 1997; Müller 2010). Similar categorial triggers have been proposed for VP/vP/PredP-movement (Collins 2017; Mahajan 2003; Massam 2001; Massam and Smallwood 1997; Müller 2004b; van Urk 2019a). Since scrambling of VPs and PNI-ed arguments behaves completely parallel, both movement operations are triggered by [•V•]. Scrambling of proper arguments on the other hand is triggered by [•D•]. I pick out intermediate scrambling here for illustration purposes, i.e. clause-internal movement across a subject, triggered in the present system by [•V•] (or [•D•]) on T. PNI-ed arguments can undergo intermediate scrambling in Turkish and Korean, but not in Tamil, Mongolian, and German. Consequently, T will be equipped with [•V•] only in Turkish and Korean. Examples for each language category are given in (41) and (42), respectively. Like Korean, Tamil indefinite direct objects are unmarked for case if pseudoincorporated, which is additionally shown by the minimal pair formed by (42a) and (42b). (41) a. Minho-ka chayk ilk-nun-ta. Minho-NOM book read-PRS-DECL ‘Minho is book-reading.’
Korean
b. Chayk1 Minho-ka ⟨chayk1 ⟩ ilk-nun-ta. book Minho-NOM read-PRS-DECL ‘Minho is book-reading.’ (42) Baker (2014b: 7,37) a. Maala veegamaa anda pustagatt-e paɖi-cc-aa. Mala.NOM quickly that¹⁰ book-ACC read-PST-3SG.F ‘Mala read that book quickly.’
Tamil
b. Maala kaŋɖippa pustagam vi-tt-aa. Mala.NOM definitely book sell-PST-3SG.F ‘Mala definitely sold books.’ ¹⁰ Whereas Baker glosses anda as a definite determiner, I follow both reference grammars (Lehmann 1993: 149, Schiffman 1999: 125) and my informants’ intuitions in classifying anda as a demonstrative.
32
PSEUDO-INCORPORATION AS A CATEGORY CHANGE PHENOMENON c. ??Pustagam1 Maala kaŋɖippa ⟨pustagam1 ⟩ vi-tt-aa. book Mala.NOM definitely sell-PST-3SG.F ‘Mala definitely sold books.’
Korean allows for intermediate scrambling of PNI-ed objects, the derivation for (41b) is given in (43). Since I assume that vP constitutes a phase, the movement-inducing feature on T triggers edge feature insertion on v, enabling the direct object to escape phasal spell-out by moving to spec, vP. PNI-DP is thus in an accessible position to be attracted to spec,TP via [•V•] on T. The features responsible for movement are boxed . TP
(43) PNI-DP1 N D CASE:ACC V
Tʹ vP DP N D CASE:NOM D
T v *CASE:NOM* V T
v′ PNI-DP1 N D *CASE:ACC* V
v′ VP ⟨DP1⟩
V D V
v V *CASE:ACC* EF D v
While Tamil is a scrambling language, it does not permit intermediate scrambling of PNI-ed objects. This language specific difference follows if the T head in Tamil cannot come with [•V•]. Neither of the two possible scenarios—T without a movement-inducing feature (44a) or T with [•D•] (44b)—can derive the word order in (42c). TP
(44) a. vP DP N D CASE:NOM D
v′ VP PNI-DP N D *CASE:ACC* V
V D V
v V *CASE:ACC* D v
T v *CASE:NOM* T
3.3 IMPLEMENTATION
33
TP
b. DP1 N D CASE:NOM D
T′ vP DP1 N D CASE:NOM D
v′ VP PNI-DP N D *CASE:ACC* V
V D V
v V *CASE:ACC* D v
T v *CASE:NOM* D T
The verbal nature of PNI-ed arguments provides an immediate explanation, not only for the general restriction on PNI movement patterns compared to proper arguments, but also for the cross-linguistic variation of movement patterns amongst PNI languages. The movement patterns in each of the languages under investigation are analysed in Chapter 7. The final property we want to tie to the verbal status of PNI-DPs is the loss of case marking. In contrast to proper noun incorporation where detransitivization and lack of agreement is robustly attested, these effects are not found with pseudonoun incorporation.¹¹ The disconnection of case marking and agreement comes as a surprise especially for the view that case assignment is simply a by-product of (Φfeature) agreement (Chomsky 2001). Instead, these observations provide strong pointers to analyse lack of case marking as a post-syntactic effect. Under the current account, no interactions between agreement and case marking are predicted, since features are entering syntactic dependencies independent of each other. A categorial feature on an argument does not influence the valuation/checking of a case or Φ feature. Neither do categorial features on distinct arguments interact with one another with respect to case. While some post-syntactic DOM/PNI accounts focus on adjacency and word order change effects (Levin 2015; van Urk 2019b; Wojdak 2008) by making use of M-Merger (Embick and Noyer 2001), there are also plenty of approaches employing impoverishment (Keine 2007; Keine and Müller 2008, 2011, 2015; Weisser 2017, 2018) or special spell-out rules (Lidz 2006; López 2012; Nuger 2010) to account for case loss. I will show how postsyntactic case drop can be modelled in each language of the core data set with DM-style impoverishment rules as well as in an OT-based framework. Either system is able to provide a uniform treatment for PNI and DOM, which will be important for languages such as Tamil and Korean (possibly Hindi too) that show a mixed pattern. Under the current proposal, case loss can be modelled straightforwardly by reference to the categorial [V ]-feature of PNI-DP. We take ¹¹ A detailed description will be given in Section 4.1.
34
PSEUDO-INCORPORATION AS A CATEGORY CHANGE PHENOMENON
Korean again as a showcase for a language which pseudo-incorporates subjects and objects. Impoverishment rules (Bonet 1991; Halle and Marantz 1993, 1994) reduce morpho-syntactic feature bundles/lists by deleting (sub)features and thus retreating to the general case which is often an elsewhere marker that is spelled out as /∅/. Post-syntactic case decomposition allows us to capture case drop on either subject or object with the same rule. They each constitute core arguments of the verb forming the natural class [-OBLIQUE] (Bierwisch 1967; Wiese 1999). Ignoring DOM-marked arguments for now, the single impoverishment rule we need for Korean is given in (45). (45)
[−OBL] → ∅ / [V ]
Korean
An Optimality-based system such as the one proposed by Aissen (1999, 2003) can describe the case drop equally well by integrating PNI-DP on the definiteness scale—one of the prominence scales assumed operative as primitives in grammar. Translated into constraint hierarchies where each scale mate is locally conjoined with a faithfulness constraint that ensures case marking, interspersed with an additional markedness constraint that avoids case marking, Aissen accounts elegantly for cross-linguistic variation among DOM languages. Keine and Müller (2008, 2011, 2015) develop a post-syntactic version of Aissen’s OT-approach. Glossing over details for the moment, a traditional definiteness scale is shown in (46), whereas an Aissen-style definiteness scale including PNI is given in (47). Similar extensions to the definiteness scale have already been proposed by von Heusinger and Kaiser (2007) and von Heusinger (2008). The details are worked out in Chapter 4. (46)
Definiteness scale according to Aissen (2003: 437): PRO ≻ PN ≻ DEF ≻ INDEF SPEC ≻ INDEF NON-SPEC
(47)
Elaborate definiteness scale including PNI: PRO ≻ PN ≻ DEF ≻ SQ ≻ INDEF SPEC ≻ INDEF NON-SPEC ≻ PNI-D ⏟⎴⎴⎴⎴⎴⎴⎴⎴⎴⏟⎴⎴⎴⎴⎴⎴⎴⎴⎴⏟ numerals∕wQ∕∃
This chapter has so far provided an overview over the trademark properties of pseudo-noun incorporation and how they can be tied to the sequential hybrid nature of the PNI determiner. Since the nominal character is transparently present in terms of c-selection and θ-roles, the focus in this section lies in pointing out the verbal traits of PNI-ed arguments. It turns out that each property that distinguishes PNI-ed from proper arguments can be traced back to a categorial [V ]-feature that has to be present in the feature list. Restriction to low scope, lack of binding and control, loss of case marking, and the peculiar restrictions on mobility can all be
3.3 IMPLEMENTATION
35
accounted for, if it is assumed that the PNI determiner changes the categorial status of the argument, once it has been merged in its base position. Before concluding this section, let me point out that there is potential positive support for the existence of a PNI determiner, coming from languages like St’a´t’imcets and Maori. These languages display a dedicated determiner that creates low scope indefinites. In addition, each language provides another indefinite determiner which allows flexible scope in Maori (Chung and Ladusaw 2004; Polinsky 1992) and obligatory wide scope in St’a´t’imcets (Matthewson 1998, 1999). The scopal properties of the two kinds of indefinite determiners in Maori are presented in (48) and (49). The indefinite determiner tētahi forms indefinites which take scope above or below negation. In contrast, the presence of the indefinite determiner he results in obligatory low scope of the indefinite under negation. (48) Chung and Ladusaw (2004: 36,40), cited after Ngata (1994)
Maori
a. Kāore tētahi tangata e mahi mā-na. T.not a person T work T.of-him ‘No one would work for him.’
¬∃
b. Kāore tētahi tangata i waiata mai. T.not a person T sing to.here ‘A (particular) person didn’t sing.’
∃¬
(49) Chung and Ladusaw (2004: 41), Polinsky (1992: 237)
Maori
a. Kāore he tangata i waiata mai. T.not a person T sing to.here ‘No one at all sang.’ (Not: A particular person didn’t sing)
¬∃, *∃¬
b. Kāore he tangata i te mahi. T.not a person T work ‘Nobody is at work.’ (Not: Some men are not at work)
¬∃, *∃¬
Scope contrasts in St’a´t’imcets are illustrated in (50), again with respect to negation. The determiner ti creates wide scope readings, whereas the determiner ku only allows for a reading where negation scopes above the indefinite. (50) Matthewson (1999: 91) a. Cw7aoz kw-s a´z’en-as ti sts’u´qwaz’-a NEG DET-NOM buy-TR-3.ERG DET fish-DET
St’a´t’imcets *¬∃, ∃¬
kw-s Sophie. DET-NOM Sophie ‘Sophie didn’t buy a fish.’ (= There is a fish which Sophie didn’t buy)
36
PSEUDO-INCORPORATION AS A CATEGORY CHANGE PHENOMENON b. Cw7aoz kw-s a´z’en-as ku sts’u´qwaz’ NEG DET-NOM buy-TR-3.ERG DET fish
¬∃, *∃¬
kw-s Sophie. DET-NOM Sophie ‘Sophie didn’t buy any fish.’ For Maori, we can additionally observe a morpho-syntactic interaction with the low scope indefinite determiner. Chung and Ladusaw (2004: 28–29) report that, except for subjects, all arguments are introduced via a preposition. Interestingly, in the Ngāti Porou dialect of Maori, the presence of low scope determiner he is only licensed in the absence of a preposition. Hence, I conclude that Maori and St’a´t’imcets provide promising evidence for overt realizations of a determiner that is responsible for the effects of pseudo-incorporation investigated in this study. The subsequent Chapters 4–7 will develop and execute each diagnostic in every language with respect to the noun types from Table 2.1, followed by an analysis of the noun types which were identified as PNI-ed. We will look at DOM-marked noun types, specifically detected in Korean and Tamil, in Chapter 8. Before we move on to the empirical part, however, I will address why this book does not pursue a more traditional layering account of categorizers, and in which sense pseudo-incorporation is different from other category change phenomena.
3.4 Sequential hybrids vs layered projections The mixed projections approach to be discussed in this section is often based on the hypothesis that roots are syntactically deficient and must exist independently of their categorizers, where the latter are either argued to instantiate syntactic heads (Embick and Marantz 2008; Harley and Noyer 2000; Marantz 2000 et seq.) or are reflected in the functional environment dominating the root (Alexiadou 2001; Borer 2003, 2005a,b). Consequently, mixed syntactic categories are often modelled with a layering approach, most prominently pursued for nominalizations resulting in deverbal nouns, where for example a verbal and a nominal categorizer are stacked in some way (Alexiadou 2017, 2020; Harley 2009, among many others). Alternatively, Panagiotidis (2015) proposed a functional projection to act as a switch to a different lexical category. This account of nominalization, that is a nominal shell above a verbal projection, has, in fact, been around for a long time and is already at play in the analyses by Abney (1987), Baker (1985), and Borsley and Kornfilt (2000). For illustration, let us look at the analysis in (51) provided by Panagiotidis (2015) for the English POSS-ing gerund, a case of nominalization, which comes closest in its implementation to the account presented in this book. Categorizers are encoded as features in this system, similar to the
3.4 SEQUENTIAL HYBRIDS VS L AYERED PROJECTIONS
37
present approach. To ensure that functional heads merge in the projection line of the correct lexical head, Panagiotidis employs Agree and uninterpretable categorial features. Some functional heads can switch the category, as it is the case with Ger in (51). The switch to N on the Ger head ensures that the possessor can take the gerund as an argument, thus deriving a nominal property. (51) POSS-ing gerunds in English
(Panagiotidis 2015: 145)
a. Albert’s eating herring DP
b. Albert
D′ ′s uN
GerP Ger uV, N
AspP Asp uV
VP eating herring V
It is easy to think of verbalization as the mirror image of the structure in (51), and, thus, as an alternative approach to PNI. The switch from nominal to verbal category would take place on a functional head, possibly even D, with the feature bundle [uN, V ]. In (52), I provide an example from Turkish with a proposal of how it would presumably be analysed under Panagiotidis’ mixed projections account. I stay agnostic about the exact nature of the switch. (52) PNI as a layering approach a. Ali eks¸i elma yedi. Ali sour apple ate ‘A li did sour apple eating.’
Turkish (Öztürk 2009: 339)
VP
b.
SWITCHP SWITCH uN, V
AdjP sour uN
V eat
NP apple N
This analysis predicts that every projection above the switch would treat the constituent as a verbal category, analogous to how the DP projection in (51) can select
38
PSEUDO-INCORPORATION AS A CATEGORY CHANGE PHENOMENON
for GerP since it is categorized as a noun. We know, however, from the discussion in the previous sections that PNI-ed arguments are selected by verbs as nominal arguments, not verbal arguments. PNI-languages generally do not vary the form of the verb, depending on whether it selects for a proper argument or a PNIed argument, nor do we detect detransitivization or agreement effects. So for all intents and purposes, the PNI-ed constituent is selected as a run-of-the-mill nominal argument. This indicates that the immediately local context to the functional head which encodes the category switch, e.g. the verb eat in (52), is not in fact sensitive to the categorial switch. Rather, higher probes triggering movement, and later stages of derivation where the PNI-ed argument is in a binding or control figuration, reveal the verbal character of the PNI-ed argument. The account proposed in this book achieves the derivation of the PNI-related verbal properties with a feature list, which in a sense hard-wires later steps of the derivation into the constituent without adding another categorial shell. In contrast, the proposal in (51)–(52) assumes a feature bundle on the functional head, which makes it impossible to distinguish local context from non-local context to the switch. The other phrasal layering accounts mentioned at the beginning of this section face similar issues, as they all focus on properties that are concerned with the immediately local functional heads merged in the projection line of the new categorizer. After all, this is often how we identify, e.g., nominalization, by the possible presence of a determiner or a possessor. Layering approaches, however, also wouldn’t deny that the nominalized constituent behaves like a noun for all later stages of the derivation. Hence, they would presumably predict that Albert’s eating herring undergoes DPmovement and is assigned case. For PNI, it is only the latter kind of diagnostics which can reveal that category change has taken place. The fact that PNI-ed arguments are c-selected based on their nominal status— i.e. pre categorial change—also makes the right prediction with respect to a set of verbal properties that we never seem to find with PNI-ed arguments. Pseudoincorporation never leads to tense or aspect morphology on the noun, nor do overt nominalizers or adjectivizers occur with a PNI-ed noun. Given that the PNI-ed constituent is still a nominal category upon selection, functional heads such as Asp and T are blocked from merging in the projection line even after the switch to a verbal category. The same logic applies to additional nominalizers and the like.¹²
¹² Note that mixed categories have been observed cross-linguistically to always show external nominal behaviour (Borsley and Kornfilt 2000). The observations made in this book can be seen as a counter-example to this generalization, though only with respect to the larger syntactic context. It is true that PNI does not create an outer verbal shell. Interestingly, Alexiadou (2020) explains the asymmetry between the possibility of nominalization and the lack of verbalization by the fact that nouns must be case licensed, the assumption being that verbalization would bleed case licensing. While this assumption is not compatible with the present approach, it is worth noting that lack of case marking is one of the key characteristics of pseudo-incorporation.
4 Pseudo-incorporation vs differential object marking This chapter is concerned with the most prominently discussed property for pseudo-incorporated arguments—the absence of case morphology. Lack of case marking is also a primary diagnostic of differential object marking. Whereas the PNI literature focuses on a correlation of case drop with scope inertness, DOMlanguages often display a correlation of case drop with animacy and/or specificity restrictions. I investigate the case marking properties of eleven noun types for each of the four PNI languages from the core data set. The primary aim of this chapter is to develop a clear distinction between PNI and DOM effects. As it turns out, languages select different classes of noun types which can either be pseudoincorporated or DOM-marked. The theory in this book ties case loss due to PNI to the verbal nature of the hybrid PNI determiner, proposed in the previous chapter. This account of PNI can be neatly integrated into existing accounts of DOMmarking. Section 4.1 provides arguments as well as the tools to model case drop as a post-syntactic phenomenon. Section 4.2 introduces and motivates the choice of the noun types considered in this book, exemplarily shown for Hindi. In Section 4.3, I take each noun type within the respective PNI languages and motivate why case drop signals either PNI or DOM.
4.1 Case loss is post-syntactic There are by now a number of syntactic PNI/DOM-accounts which model the lack of case marking as a consequence of a missing syntactic dependency to a potential case assigner (Kalin 2014, 2018; Levin 2019; van Urk 2019b) or the result of being bled by the application of syntactic head movement (Baker et al. 2005; Baker 2014b). These approaches predict an interaction between agreement and case marking. Often Φ-agreement and case assignment are seen as two sides of the same coin (Chomsky 2001). Hence, if case is blocked from being assigned syntactically, there is a good chance that this affects Φ-agreement or any other kind of agreement as well. Since PNI/DOM is frequently diagnosed with objects, as they are more often marked for case, while agreement happens predominantly with subjects, many PNI/DOM languages do not allow us to investigate an interaction between agreement and case marking. Fortunately, three out of the five Pseudo-Noun Incorporation and Differential Object Marking. Imke Driemel, Oxford University Press. © Imke Driemel (2023). DOI: 10.1093/oso/9780192866400.003.0004
40
PSEUDO-INCORPORATION VS DIFFERENTIAL OBJECT MARKING
languages of our core data set make it possible to see the effects of case marking and agreement simultaneously. None of the three languages shows an interaction between case loss and agreement. I begin with Turkish subject pseudo-incorporation, diagnosed in embedded clauses where the subject is overtly marked for genitive case. The embedded verb inflects for person and number co-varying with the Φ-features of the embedded subject.¹ If we compare (1a) to (1b), we see that the agreement morphology changes from -im to -i depending on the Φ-values of the embedded subject. Crucially, agreement morphology cannot be dropped in (1c) where the subject doktor is not marked for genitive case. Assuming that case drop signals pseudoincorporation, we can conclude from the paradigm in (1) that PNI does not affect subject–verb agreement.² (1) No Φ-agreement interaction with case drop a.
Ko¨y-e (ben-im) gel-dig˘-im-i duy-du-n. village-DAT 1SG-GEN come-NLMZ-1SG-ACC hear-PFV-2SG ‘You heard that I came to the village.’
b.
Ko¨y-e doktor-un gel-dig˘-i-ni duy-du-m. village-DAT doctor-GEN come-NLMZ-3SG-ACC hear-PFV-1SG ‘I heard the doctor came to the village.’
c.
Ko¨y-e doktor gel-dig˘*(-i)-ni duy-du-m. village-DAT doctor come-NLMZ-3SG-ACC hear-PFV-1SG ‘I heard a doctor came to the village.’
Turkish
¹ Both genitive case marking and agreement morphology taken from the nominal paradigm suggest a nominal licenser which has been proposed to come in the form of a nominal Agr projection (Kornfilt 2003b) within the clause or a DP/NP shell (Aygen 2007b; Lees 1965) wrapped around the subordinate clause. ² Öztürk (2005) claims that there is a noticeable interaction between the plural agreement marker and the scope properties of the subject. The judgements for the minimal pair given below are controversial. İşsever (2007) points out that he has not been able to verify the contrast with any of his informants. Similar data is discussed by Kornfilt (1997: 385–386), arguing for an effect of plural agreement triggering a definite reading of subjects (not shown here), which I could not verify with my informants. (i) Öztürk (2005: 170)
Turkish
a. Bu¨tu¨n c¸ocuk-lar Allahtan o test-e gir-me-di-ler. all child-PL luckily that test-DAT take-NEG-PFV-3PL ‘A ll the children luckily didn’t take that test.’
∀¬,*¬∀
b. Bu¨tu¨n c¸ocuk-lar Allahtan o test-e gir-me-di-∅. all child-PL luckily that test-DAT take-NEG-PFV-3SG ‘A ll the children luckily didn’t take that test.’
*∀¬,¬∀
The data is relevant insofar as that Turkish subject PNI has also been discussed for matrix clauses (Öztürk 2005, 2009), with the complication that nominative case is unmarked in Turkish, therefore making it more difficult to diagnose PNI. Given that strong quantifiers do not allow for case drop in embedded clauses (see Section 4.3.3) the contrast in the example is unlikely due to a PNI effect.
4.1 CASE LOSS IS POST-SYNTACTIC
41
Korean is another PNI language which marks case on subjects overtly, even in matrix clauses. Although Korean does not display Φ-agreement between subject and verb, it shows agreement in honorific features (Ahn and Yoon 1989; Chung 2009; Koopman 2005); see also Khym (1998) and Yoon (1995b) for discussions of honorific agreement as an instance of Agree. The configurations in (2) show that honorific case drop has no consequence for honorific agreement morphology. Bare nouns (2a), numeral classifier phrases (2b), and demonstrative phrases (2c) can each drop the honorific case marker without affecting the verbal morphology. As we will see in Section 4.3.4, caseless bare nouns are pseudo-incorporated, whereas numerals and demonstratives lose case as a sign of DOM. Thus, neither DOM nor PNI show an interaction between agreement and case drop. This is one of the reasons why I will propose an integrated system where case loss is modelled in the same way for DOM as well as PNI. (2) No honorific agreement interaction with case drop
Korean
a.
Halapeci(-kkeyse) cenyek-ul capswu*(-si)-n-ta. grandfather-HON.NOM dinner-ACC eat-HON-PRS-DECL ‘Grandfather is having a dinner.’
b.
Halapeci(-kkeyse) twu-pwun cenyek-ul capswu*(-si)-n-ta. grandfather-HON.NOM two-CL dinner-ACC eat-HON-PRS-DECL ‘Two grandfathers are having a dinner.’
c.
I/ce halapeci(-kkeyse) cenyek-ul capswu*(-si)-n-ta. DEM grandfather-HON.NOM dinner-ACC eat-HON-PRS-DECL ‘This/that grandfather is having a dinner.’
It is important to point out that there is an asymmetry between the honorific case marker and the honorific verbal affix. While the former can be left out, the latter always has to be pronounced (Yoon 1995a); see (3). This fact supports an analysis of the data in (2) as an instance of Agree with optional case deletion, in contrast to simply considering honorific case and agreement markers as co-occurrences of discourse-related particles. (3)
Yoon (1995a: 335) *Apenim-kkeyse chayk-ul ilk-ess-ta. Father-HON.NOM book-ACC read-PST-DECL ‘Father read a book.’
Korean
Tamil is the third PNI language which enables us to observe a possible interaction between case drop and lack of agreement, albeit an indirect one. Although Tamil only marks objects overtly with case morphology and mostly displays subject– verb agreement, there is a closed class of verbs which select for dative subjects
42
PSEUDO-INCORPORATION VS DIFFERENTIAL OBJECT MARKING
and nominative objects and agree in Φ-features with the direct object. As will be shown in detail in Sections 5.4.1 and 8.1, Tamil shows a correlation of low scope and non-specific readings with case drop on direct objects—key characteristics of PNI/DOM. We thus predict the same interaction between agreement and scope/specificity for DAT-NOM verbs. Tamil has the option of default N.SG agreement (Baker 2015: 69; Murugesan 2019: 19; Sundaresan 2012: 211); thus we would expect N.SG agreement to be an option e.g. in contexts which make nonspecific object readings very likely. This prediction is not borne out, as (4) shows. While full agreement with the nominative object is grammatical (see (4b)), neuter singular agreement is not an option, as shown in (4a), even within the provided non-specific context.³ (4) Context [−spec]: Mala is producing a play and she got a (random) boy to take part in the play. a. *Mala-kku (oru) paiyan keɖai-cc-itu. Mala-DAT a boy get-PST-3SG.N ‘Mala got a boy.’ b.
Tamil
Mala-kku (oru) paiyan keɖa-cc-aan. Mala-DAT a boy.NOM get-PST-3SG.M ‘Mala got a boy.’
We can conclude, at least with respect to the current data set, that whatever mechanism triggers the lack of case marking on arguments in PNI/DOM contexts cannot additionally influence the agreement relation between the argument and the (potential) case assigner. These observations can be accounted for in at least two different ways. Either case assignment and agreement are decoupled syntactic operations which each operate on their own and thus can also be blocked for different reasons, or case assignment and verbal agreement succeed successfully, possibly also by the same operation, but there is an additional deletion operation preventing case morphology to occur on arguments. The next piece of evidence supports the latter approach, as it reinforces the intuition that the lack of case morphology does not come about by a syntactic operation, since such operation does not have any impact on the grammatical function of the other non-incorporated arguments within the sentence. PNI languages do not display detransitivization effects. Both Turkish and Korean have been shown to display PNI of the subject. In such cases, however, neither Turkish nor Korean reanalyse the non-incorporated object as the single syntactically active argument of the clause, that is the object retains accusative case; see (5)–(6). A similar observation can be made for Adyghe, a language with ³ Baker (2014b) uses this test to argue for the opposite conclusion; see Section 9.1 in Chapter 9 for a detailed discussion of his theory.
4.1 CASE LOSS IS POST-SYNTACTIC
43
an ergative-absolutive alignment, which retains ergative marking of the subject in the absence of absolutive marking on the object, shown in (7).⁴ (5) Kornfilt (2008: 84) Ko¨y-u¨ bir haydut(-un) bas-tıg˘-ın-ı duy-du-m. village-ACC a robber-GEN raid-NMLZ-3SG-ACC hear-PST-1SG ‘I heard that a (certain) robber raided the village.’ (6) Kwon and Zribi-Hertz (2008: 265), Ha. Lee (2006: 329–330)
Turkish
Korean
a. Beoseu(-ga) o-goiss-da. bus-NOM come-PROG-DECL ‘There’s a/the bus coming.’ b. Etten haksayng sihem-ul an machiko naysseyo. some student exam-ACC not complete submitted. ‘Some student submitted the exam without completing it.’ (7) Testelets and Arkadiev (2014: 6) Pŝaŝe-m ̂ʒane(-r) ә-dә-ʁ. girl-ERG dress-ABS 3SG-sew-PST ‘The girl made the dress.’
Adyghe
Öztürk (2005, 2009) provides further evidence from Turkish causative scenarios which show that PNI-ed objects still act as if they carry case morphology. Turkish exhibits a double-case constraint which bans the presence of more than one argument with the same case morphology within the same clause. Thus, the subject of an unergative verb is marked with accusative case if the verb is causativized to prevent the occurrence of two nominative arguments (8a). The subject of a transitive verb occurs with dative case if causativized (8b) to avoid the double occurrence of accusative arguments. Interestingly, if the object of the causativized transitive verb is PNI-ed, the subject occurs with dative, not accusative case morphology (8c), suggesting that the embedded verb behaves like a transitive verb, no matter whether PNI has taken place or not. (8) Öztürk (2009: 347–348)
Turkish
a. Ays¸e Ali-yi kos¸-tur-du. Ays¸e Ali-ACC run-CAUS-PST ‘Ays¸e made Ali run.’
CAUS + UNERG
b. Ays¸e Ali-ye/*-yi balıg˘-ı tut-tur-du. Ays¸e Ali-DAT/-ACC fish-ACC catch-CAUS-PST ‘Ays¸e made Ali catch the fish.’
CAUS + TRANS
⁴ The same point was recently made by Kidwai (2020) for Hindi and its split-ergative case system.
44
PSEUDO-INCORPORATION VS DIFFERENTIAL OBJECT MARKING c. Ays¸e Ali-ye/*-yi balık tut-tur-du. Ays¸e Ali-DAT/-ACC fish catch-CAUS-PST ‘Ays¸e made Ali go fishing.’
CAUS + PNI
Guntsetseg (2016) shows that Mongolian behaves exactly like Turkish with respect to causative scenarios, the only difference being that subjects of causativized transitive verbs are assigned instrumental instead of dative case. The paradigm is given in (9). (9) Guntsetseg (2016: 65) a. Dorzˇ Tujaa-g jav-uul-san. Dorzˇ Tujaa-ACC go-CAUS-PST ‘Dorzˇ let Tujaa go.’
Mongolian CAUS + UNERG
b. Dorzˇ Tujaag-aar/*Tujaa-g ene nom-yg unsˇ-uul-san. CAUS + TRANS Dorzˇ Tujaa-INST/Tujaa-ACC this book-ACC read-CAUS-PST ‘Dorzˇ let Tujaa read this book.’ c. Dorzˇ Tujaag-aar/*Tujaa-g nom unsˇ-uul-san. Dorzˇ Tujaa-INST/Tujaa-ACC book read-CAUS-PST ‘Dorzˇ let Tujaa read a book.’
CAUS + PNI
Since the case loss of pseudo-incorporated arguments seems to have no effect either on co-variance relations with the verb or on the case morphology of other non-incorporated arguments in the clause,⁵ I tentatively conclude that case loss happens in a post-syntactic module, counter-feeding/-bleeding any syntactic operations which could potentially be sensitive to caseless arguments. Note that this decision is not driven by theory-internal requirements. One might imagine the hybrid category present with PNI-ed arguments to have an influence on (Φ)-agreement, e.g. probes on verbal heads might not target goal features on other verbal categories. Note furthermore that the missing detransitivization effects are especially surprising within a theory of Dependent Case (Baker 2014a, 2015; Marantz 1991)—a case assignment model which computes case based on c-command relations between nominal arguments in the clause or phase. One might predict that PNI-ed arguments cannot act as case competitors anymore since they constitute verbal categories after they have been selected.⁶ The empirical facts, presented in this section, speak against such an approach. Rather, ⁵ Niuean, arguably the first language to ever be diagnosed as a PNI language (Massam 2001), can be considered an outlier here. See Chapter 9 for more discussion. ⁶ The problem with Dependent Case and the lack of detransitivization effects is in fact more general. DP/NP accounts of pseudo-noun incorporation operate under the assumption that PNI-ed arguments enter the derivation as NPs which do not require case marking (Dayal 2011; Massam 2001). If this is true, then DP/NP accounts are incompatible with Dependent Case models, in light of the fact that PNI does not lead to detransitivization. It is difficult to maintain the idea that an argument is of the
4.1 CASE LOSS IS POST-SYNTACTIC
45
whatever causes PNI-ed arguments to appear without case morphology must be implemented very late in the derivation, e.g. in post-syntax. It is, thus, no coincidence that the current proposal considers case assignment to be a checking relation between a case probe on a functional head and a case goal feature on a c-commanded argument, where hybrid as well as proper D categories can come with case features. There is a variety of tools post-syntax offers to prevent case morphology from being realized. One possibility is to restrict vocabulary insertion of the case feature to certain contexts, e.g. to only a subset of noun types (Lidz 2006; López 2012; Nuger 2010). An example is given in (10), taken from López (2012), who assumes KP shells which are additionally added to DPs. K heads constitute the f-morphemes that serve as the locus for case morphology. Whereas in Spanish, Romanian, Kiswahili, and Hindi only animate nouns are marked for case, Catalan restricts case marking to a certain noun type, i.e. pronouns. (10) López (2012: 60-61) a. /a/ ↔ [K] /
DP[+anim]
b. /a/ ↔ [K] /
DP[+pro]
Spanish, Romanian, Kiswahili, Hindi Catalan
A very similar way to restrict case morphology from being inserted is by making use of DM-style impoverishment rules (Bárány 2017; Keine 2007, 2010; Weisser 2017, 2018). These rules reduce morpho-syntactic feature bundles/lists by deleting sub-features and thus enforce retreating to the general case, which is often an elsewhere marker that is spelled out as /∅/. An analysis along those lines is presented in (11) for Hindi. Overly simplified, we can observe for Hindi that nonoblique arguments are either marked with -ko if they are objects or -ne if they are subjects, hence the decomposition of case features in (11a) and (11b). The impoverishment rule in (11d) applies in the context of non-human, non-specific objects where the object feature [+OBJ] is deleted, making it impossible for /-ko/ to be inserted. The system retreats to the elsewhere marker, given in (11c), since it is the only compatible exponent, after impoverishment has applied. (11) Keine (2010: 60) a. /-ko/ ↔ [-OBL,+OBJ]
Hindi
b. /-ne/ ↔ [-OBL,+SUBJ] wrong category to receive case, yet of the right category to act as a case competitor. The data in (5)– (7) are also problematic for Baker’s DOM account, put forward in Baker (2015: 125–130), where he analyses lack of case marking as the result of being the single argument in the VP spell-out domain, thus receiving unmarked, and in his data set ∅ case marking. While this theory seems to make the right prediction in languages where the unmarked case is realized with a zero exponent, and caseless nouns are ambiguous between being genuinely caseless and being case marked realized as ∅, languages where both dependent and unmarked case is spelled out overtly show us that Baker’s approach is no longer tenable.
46
PSEUDO-INCORPORATION VS DIFFERENTIAL OBJECT MARKING c. /−∅/ ↔ [ ] d. [+OBJ] → ∅ /DP[−human,−spec]
The concept of impoverishment does not necessarily have to be implemented in a rule-based system. Since impoverishment essentially blocks features from being realized due to economy-based markedness considerations (Noyer 1998; Nevins 2011), one can also think of it as a markedness constraint within an OT-based system, forcing feature deletion in minimal violation of faithfulness constraints which require features to be realized. Keine and Müller (2008, 2011, 2015) develop a post-syntactic DOM account, based on the insights from Aissen (1999, 2003), who models DOM effects in a constraint-based OT-framework. Aissen (2003) analyses differential object marking as a means of resolving the tension between two grammatical principles: iconicity, which requires argument types untypical for their grammatical function to be marked, and economy, which forces a system to avoid morphological marking generally. She argues for prominence scales to be taken as core grammatic principles that regulate the distribution of DOM/case markers in DOM languages. The prominence scales which Aissen suggests to be at work are given in (12). (12) Aissen (2003: 437), Aissen (1999: 674) a. Grammatical function: SUBJ ≻ OBJ b. Definiteness: PRO (1st ,2nd ) ≻ PRO (3rd ) ≻ PN ≻ DEF ≻ INDEF SPEC ≻ INDEF NON-SPEC c. Animacy: HUMAN ≻ ANIMATE ≻ INANIMATE One of the main reasons for an OT-system is that Aissen is able to derive markedness reversal effects by using Harmonic Alignment (Prince and Smolensky 2004: 161). Subjects are more prominent than objects and certain languages only (case) mark subjects low and objects high on prominence scales (Comrie 1989; Croft 1988, 1990). Harmonic alignment is a mechanism which operates on pairs of scales where one scale is binary, in this case (12a), and aligns each of its elements with each element on the other scale. This creates subhierarchies that are, by definition, ordered from left to right for the first element of the binary scale, but ordered from right to left for the second element. An illustration for combining (12a) and (12c) is given in (13). The subhierachies are translated into constraint hierarchies in (14) by turning the scale mates into avoid constraints and reversing the order.
4.1 CASE LOSS IS POST-SYNTACTIC
47
(13) Subhierarchies for (12a) and (12c) via Harmonic alignment a. SUBJ/HUM ≻ SUBJ/ANIM ≻ SUBJ/INANIM b. OBJ/INANIM ≻ OBJ/ANIM ≻ OBJ/HUM
(14) Constraint alignments for (13) a. *SUBJ/INANIM ≫ *SUBJ/ANIM ≫ *SUBJ/HUM b. *OBJ/HUM ≫ *OBJ/ANIM ≫ *OBJ/INANIM In order to enforce case marking on subjects and objects, we locally conjoin each scale mate of the subhierarchies with the faithfulness constraint MAX-C⁷ which penalizes case feature deletion; see (15). Local Conjunction produces constraint ties, i.e. constraint rankings in which candidates can be optimal as long as they satisfy one of the constraints of the tie. Hence, a constraint like e.g. *OBJ/HUM & MAX-C in (15a) expresses that objects can refer to human individuals as long as they are case marked. (15) Iconicity a. *OBJ/HUM & MAX-C ≫ *OBJ/ANIM & MAX-C ≫ ... b. ... ≫ *SUBJ/ANIM & MAX-C ≫ *SUBJ/HUM & MAX-C Finally, to enable case deletion we introduce a markedness constraint *[CASE] which is motivated by economy consideration essentially trying to avoid morphological marking in general. The markedness constraint is interspersed with constraint alignments formed by Harmonic Alignment of prominence scales and Local Conjunction with a faithfulness constraint, where the position of *[CASE] can be intuitively understood as the cut-off point on prominence scales where arguments are considered to need case. Cross-linguistic variation is derived by reranking *[CASE]. The language described in (16) case marks objects if they refer to a human individual and subjects if they refer to non-human entities. In other words, everything to the right of *[CASE] is unmarked for case, whereas everything to the left of *[CASE] is marked for case. (16) Iconicity vs. economy a. *OBJ/HUM & MAX-C ≫ *[CASE] ≫ *OBJ/ANIM & MAX-C ≫ ... b. ... ≫ *SUBJ/ANIM & MAX-C ≫ *[CASE] ≫ *SUBJ/HUM & MAX-C
⁷ I follow Keine and Müller (2008, 2011, 2015) in modelling the iconicity requirement as a faithfulness constraint that penalizes case feature deletion from input to output.
48
PSEUDO-INCORPORATION VS DIFFERENTIAL OBJECT MARKING
To give a concrete example, I will present the analysis of Hindi DOM marking from Keine and Müller (2008, 2011, 2015). As the impoverishment rule in (11d) already suggests, non-specific non-human objects are not case marked, thus we have to make use of all three scales in (12). First we build the harmony scales in (17) and derive the constraint hierarchies in (18). Only the relevant parts of the scales are shown. (17) Subhierarchies via Harmonic alignment
Hindi
a. SUBJ/HUM ≻ SUBJ/NHUM b. SUBJ/SPEC ≻ SUBJ/NSPEC c. OBJ/NHUM ≻ OBJ/HUM d. OBJ/NSPEC ≻ OBJ/SPEC
(18) Constraint alignments
Hindi
a. *SUBJ/NHUM ≫ *SUBJ/HUM b. *SUBJ/NSPEC ≫ *SUBJ/SPEC c. *OBJ/HUM ≫ *OBJ/NHUM d. *OBJ/SPEC ≫ *OBJ/NSPEC In order to combine animacy and specificity, Local Conjunction applies between subhierarchies as well, yielding the relevant rankings in (19), where we focus on object markedness. A simplified notation as well as Local Conjunction with MAX-C is shown in (20). The constraints are partially ranked, as is illustrated in (21). Whereas *OBJ/HUM/SPEC & MAX-C dominates all other constraints in (21), *OBJ/HUM/NSPEC and *OBJ/NHUM/SPEC are not ranked with respect to each other.⁸ (19) Combination of prominence scales
Hindi
a. *OBJ/HUM & *OBJ/SPEC ≫ *OBJ/HUM & *OBJ/NSPEC b. *OBJ/NHUM & *OBJ/SPEC ≫ *OBJ/NHUM & *OBJ/NSPEC c. *OBJ/SPEC & *OBJ/HUM ≫ *OBJ/SPEC & *OBJ/NHUM d. *OBJ/NSPEC & *OBJ/HUM ≫ *OBJ/NSPEC & *OBJ/NHUM
⁸ Aissen (2003: 458–459) derives two dimensional DOM patterns by forming the cross product of the two scales and an additional condition which maintains the scale relations: ‘If a ranges over values on one scale, and b over values on the other, then a pair ⟨a1 , b1 ⟩ ≥ ⟨a2 , b2 ⟩ iff a1 ≥ a2 and b1 ≥ b2 .’
4.1 CASE LOSS IS POST-SYNTACTIC
(20) Iconicity
49
Hindi
a. *OBJ/HUM/SPEC & MAX-C ≫ *OBJ/HUM/NSPEC & MAX-C b. *OBJ/NHUM/SPEC & MAX-C ≫ *OBJ/NHUM/NSPEC & MAX-C c. *OBJ/HUM/SPEC & MAX-C ≫ *OBJ/NHUM/SPEC & MAX-C d. *OBJ/HUM/NSPEC & MAX-C ≫ *OBJ/NHUM/NSPEC & MAX-C *OBJ/HUM/SPEC & MAX-C
(21)
*OBJ/HUM/NSPEC & MAX-C
*OBJ/NHUM/SPEC & MAX-C
*OBJ/NHUM/NSPEC & MAX-C Finally, the markedness constraint *[+GOV], a version of *[+CASE] which applies to object case features, is interspersed with the faithfulness constraints exactly at the point where Hindi seems to show the cut off point on the prominence scales; see (22). In parallel with the impoverishment rule in (11d), case marking of non-human, non-specific objects incurs a crucial violation of *[+GOV], as their faithfulness constraint is ranked lower, shown in (23). In (24), we see that the type of objects which satisfy constraints ranked above *[+GOV] require case marking, otherwise MAX-C is violated. (22) Iconicity vs economy (Keine and Müller 2008: 105)
Hindi
⎧ *OBJ/HUM/SPEC & MAX-C ⎫ *OBJ/HUM/NSPEC & MAX-C ≫ *[+GOV] ≫ *OBJ/NHUM/NSPEC & MAX-C ⎨ ⎬ *OBJ/NHUM/SPEC & MAX-C ⎩ ⎭
(23) Non-specific, non-human objects are not case marked
Hindi
50
PSEUDO-INCORPORATION VS DIFFERENTIAL OBJECT MARKING
(24) Specific, non-human objects are case marked
Hindi
As we have seen in this section, there is ample reason to adopt a post-syntactic approach to case loss of PNI-ed/DOM-ed arguments. Case drop does not interact with (Φ)-feature agreement, nor does it cause other arguments in transitive clauses to be reanalysed as the single argument of the sentence. The three postsyntactic strategies I have presented here for the most part originate from the DOM literature. Scope is often either not taken into account at all or treated as a property indistinguishable from specificity. We know, however, that scopal and epistemic properties do not have to align. Recent quantitative studies for Turkish (von Heusinger and Bamyacı 2016; von Heusinger et al. 2019) reveal that epistemic specificity, i.e. readings in transparent contexts where the speaker has a specific person in mind, is not related to case marking. There is, however, a correlation with scope. Our PNI study on Korean did not reveal any sensitivity between case marking and specific readings in transparent contexts. Tamil, in contrast, exhibits a clear correlation where lack of case marking triggers both a low scope restriction and non-specific readings. The next section will discuss yet a different pattern found in Hindi. I take scopal restrictions triggered by case loss to be a direct consequence of the presence of the PNI determiner PNI-D, argued for in sections 3.1 and 3.3. PNI-ed arguments, however, do not necessarily constitute non-specific indefinites. Specificity is a property tied to DOM contrasts which may or may not interact with scope. The task of the next section is to integrate the PNI determiner into the already existing post-syntactic DOM systems. I will take Hindi again as example and show how PNI and DOM effects can be modelled within the same language, where I provide a DM-style solution with impoverishment rules as well as an OT-solution with the help of prominence scales.
4.2 PNI within post-syntactic DOM accounts This section fulfils two purposes: (i) it provides the full data set on Hindi PNI/DOM effects, and (ii) it establishes the need to treat scope and specificity separately. I will show that PNI-ed arguments can be fully integrated into a system based either on case-deletion rules or economy markedness constraints and exist alongside DOM-marked arguments. Since Hindi is not part of the core data set of this study, the conclusions are not as empirically grounded as they are for the other languages, which will be discussed in the next section. Hindi, nevertheless, serves as a good example to bring across the essential points of the analysis for case drop.
4.2 PNI WITHIN POST-SYNTACTIC DOM ACCOUNTS
51
We will start with the high end of the definiteness scale. Butt (1993) shows that pronouns and proper names require case marking, shown in (25). While the minimal pair also lists demonstrative phrases as arguments which must be case marked, there is some disagreement among speakers. Montaut (2018) provides an example where a demonstrative phrase is unmarked for case; see (26). If (non-contrastively) focused, case-marked demonstratives are even the less preferred option; see (27). (25) Butt (1993: 96) Hindi Zainab-ne Ali*(-ko) / us*(-ko) / us mez*(-ko) dek-aa. Zainab-ERG Ali-ACC / 3SG-ACC / that table-ACC see-PFV.M.SG ‘Zainab saw Ali/him/that table.’
(26) Montaut (2018: 290) Yah tālā maı˜ne xud lagāya hai. this lock 1SG.ERG REFL put.PFV 3SG.M ‘This lock, I put it myself.’
Hindi
(27) Dalrymple and Nikolaeva (2011: 161) A: What did Hassan sell? B: ? Hassan-ne voh kalam-ko bec diyaa. Hassan-ERG that pen-ACC sell gave.PFV.3SG.M ‘Hassan sold that pen.’ B’: Hassan-ne voh kalam bec diyaa. Hassan-ERG that pen sell gave.PFV.3SG.M ‘Hassan sold that pen.’
Hindi
I tentatively assume that case marking is optional for demonstratives (Rajesh Bhatt, p.c.), while pronouns and proper names require case marking. Possessive phrases pattern with the latter; see (28). (28) Montaut (2018: 283) Maı˜ne apnī beṭ ī*(-ko) kal nahı˜ dehā. 1SG.ERG REFL daughter-ACC yesterday not see.PFV.M.3SG ‘I did not see my daughter yesterday.’
Hindi
Definiteness is not as easily detectable since, in contrast to demonstrative determiners, definite determiners are non-overt. Weak definites, i.e. noun phrases with a globally known referent, strongly prefer to be case marked, an example is given in (29).
52
PSEUDO-INCORPORATION VS DIFFERENTIAL OBJECT MARKING
(29) Montaut (2018: 285) Ca˜d-ko dekho! moon-ACC look.IMP ‘Look at the moon!’
Hindi
Strong definites, i.e bare nouns which anaphorically relate to previously introduced discourse referents, seem to require case marking as well. I conclude this from a variety of contexts, provided by Dalrymple and Nikolaeva (2011), which show that bare nouns are obligatorily case marked if they constitute topics referring to a salient entity in the context, compare (30) and (31).⁹ (30) Context paraphrase: (Dalrymple and Nikolaeva 2011: 163) The communicative purpose is to inform the addressee about the speaker’s intention. The object referent is construed as part of the event and is not individuated as a pragmatically salient element. a Ham mez pūch dēge. I.HON table wipe give.FUT.1PL ‘I am going to wipe the table.’
Hindi
(31) Context paraphrase: (Dalrymple and Nikolaeva 2011: 163) The event is presented as a plan, and the utterance is construed as an instruction to the addressee to perform a certain action with respect to the table, so the table plays a prominent pragmatic role in the situation. a Mez-ko pūch do. table-ACC wipe give.IMP ‘Wipe the table!’
Hindi
Quantifiers are rarely discussed in the DOM literature. We can find some information in Dayal (2011). One could interpret this information as signalling a natural class consisting of indefinites, numerals > 1, and weak quantifiers which allow for case drop, in contrast to strong quantifiers. A limited set of determiners, namely kai ‘several’ or numerals are possible [without case marking]. Impressionistically speaking, they are the set of weak determiners, those amenable to an adjectival analysis. I should also note here that there are speakers who might even accept strong determiners like har ‘every’ or vo ‘that’ [...] Most people, however, do not accept strong determiners in the absence of accusative marking. (Dayal 2011: 127, fn.5) ⁹ This is a tentative conclusion. Dayal (2011: fn.4,fn.55) points out that definite readings are possible with unmarked nouns, albeit with special intonation.
4.2 PNI WITHIN POST-SYNTACTIC DOM ACCOUNTS
53
Indefinites are more often discussed (Bhatt 2007; Dayal 1999, 2003, 2011), but the majority of the literature focuses on bare nouns (Bhatt and Anagnostopoulou 1996; Butt 1993; Butt and King 2004; Mohanan 1995). With respect to the specificity property, there is a clear one-to-one relation with case marking for bare nouns (32) as well as for indefinites (33). (32) Butt (1993: 97)
Hindi
a. Us-ke xansaame-ne bazaar-se murγii –spec 3SG-GEN.OBL cook.M.OBL-ERG market.M-from chicken.F.NOM xariid-aa. buy-PERF.F.SG ‘His cook bought a chicken from the market.’ b. Us-ke xansaame-ne bazaar-se murγii-ko +spec 3SG-GEN.OBL cook.M.OBL-ERG market.M-from chicken.F-ACC xariid-aa. buy-PERF.F.SG ‘His cook bought a particular chicken from the market.’
(33) Bhatt (2007: 2)
Hindi
a. Mina ek bacca uṭ hhaa rahii hai. Mina a/one child lift PROG.F be.PRS.3SG ‘Mina is picking up a child.’
–spec
b. Mina ek bacce-ko uṭ hhaa rahii hai. Mina a/one child-ACC lift PROG.F be.PRS.3SG ‘Mina is picking up a particular child.’
+spec
The scope properties of bare nouns and indefinites, however, do not pattern with specificity effects. If we move away from transparent contexts to sentences containing an operator such as negation, we can see that caseless bare nouns take obligatory low scope (34b), whereas caseless indefinites can scope above or below other operators in the sentence (35b). In fact, case marking on indefinites does not have any effect on scope, cf. (35a).¹⁰
¹⁰ The ¬∃ reading is not explicitly mentioned in Dayal (2011: 128) although it is implied in prose; see also Dayal (2003: 4). She explicitly mentions low scope readings for case-marked indefinites with respect to quantified subjects. I assume that there is a low scope reading, albeit a scalar one which strengthens the numeral interpretation of ek, in parallel to the example she discusses in Dayal (2004: 418).
54
PSEUDO-INCORPORATION VS DIFFERENTIAL OBJECT MARKING
(34) Dayal (2011: 128) a. Anu bacce-ko nahiiN samhaalegii Anu child-ACC not look.after.FUT ‘Anu will not look after the child.’ b. Anu bacca nahiiN samhaalegii Anu child not look.after.FUT ‘Anu will not look after children.’
DEFINITE
¬∃,*∃¬
(35) Dayal (2011: 128), Dayal (1999: 35) a. Anu ek bacce-ko nahiiN samhaalegii Anu one child-ACC not look.after.FUT ‘Anu will not look after a particular child.’
¬∃,∃¬
b. Anu ek/koii kitaab nahiiN paRhegii Anu one/some book not read.F ‘Anu won’t read any book./There is a book Anu won’t read’
¬∃,∃¬
We can separate specificity from scope by making the latter depend on the type of determiner. The PNI determiner PNI-D triggers a low scope restriction and is responsible for the readings in (34b). Case loss is tied to the categorical V feature which is part of its feature list. Since we know from cross-linguistic observations that case morphology usually does not occur with verbal categories (Blake 2004; Nichols 1986), PNI-ed arguments will not be case marked either, following the intuition that they are simply not nominal enough to show case morphology. Other determiners are proper nominal categories which are in complementary distribution with PNI-D and form arguments of type ⟨e⟩ or ⟨⟨e, t⟩, t⟩. Hindi ek NPs for example constitute generalized quantifiers, which can show flexible scope. Following Dayal (2004), I analyse ek as a numeral. Similarly, weak and strong definites constitute DPs with D heads that receive a zero spell-out just like PNI-D. I provide underlining structures which show feature lists for the relevant determiners we have discussed so far in (36). The semantics and the vocabulary entries are given in (37) and (38), respectively. Neither (36) nor (37) or (38) is an exhaustive list, of course.¹¹ The denotation of the PNI determiner was introduced in Section 3.3, while denotations for the definite determiners were introduced in Section 3.2, adopted from Schwarz (2009).
¹¹ Other quantifiers such as koii (= some) and har (= every) behave like ek. Similarly, personal pronouns and demonstratives have been argued to constitute D heads (Elbourne 2005, 2013; Hinterwimmer 2015; Patel-Grosz and Grosz 2017; Sauerland 2007). Proper names are traditionally analysed as individual constants (Frege 1892; Heim and Kratzer 1998; Kripke 1972), either forming D heads (Coppock and Champollion 2017) or an NP with an expletive D head (Longobardi 1994).
4.2 PNI WITHIN POST-SYNTACTIC DOM ACCOUNTS
(36) Syntactic structures for Hindi determiners:
DP
PNI-DP DPNI NP [D ≻ V]
Done NP [#1, D]
55
DP DP
wDɩ NP [w-ɩ, D]
g(i)
D′ sDɩ NP [s-ɩ, D]
(37) Semantic denotations for Hindi determiners: a. ⟦D ≻ V⟧ = λP⟨e,t⟩ λQ⟨e,⟨v,t⟩⟩ λe∃z[P(z) ∧ Q(z)(e)]
DPNI
b. ⟦#1, D⟧ = λP⟨e,t⟩ λQ⟨e,t⟩ ∃z[P(z) ∧ |z| = 1 ∧ Q(z)]
Done
c. ⟦w-ι, D⟧ = λP⟨e,t⟩ : ∃!x[P(x)].ιx[(P)(x)]
wDι
d. ⟦s-ι, D⟧ = λP⟨e,t⟩ λy : ∃!x[P(x) ∧ x = y].ιx[P(x) ∧ x = y]
sDι
(38) Vocabulary entries for Hindi determiners: a. /ek/ ↔ [#1] b. /∅/ ↔ [ ]
Done DPNI , wDι , sDι
Specificity can be encoded as a separate feature [±SPEC] which can optionally be added to the determiners. A [+SPEC] feature triggers the domain restriction (Onea and Geist 2011; Schwarzschild 2002), as it was discussed in Section 3.2. This way, we can keep scope and specificity properties apart. Thus, the fact that case loss on ek NP correlates with non-specific readings in (33) does not necessarily predict that there is also a scope restriction, as (35) clearly shows. In other words, specificity is a DOM-related property, whereas scope is a PNI effect. I will now take Keine’s (2010: 60) DM-analysis from the previous section in (11) and add the rule necessary to cover the PNI data; see (39f). Since we are taking more noun types into account, a side effect of this study is that we will also add more DOM-related impoverishment rules.¹² In this case, we add a DOM rule for demonstratives (see (39e)), which applies optionally.¹³ (39) Keine (2010: 60) extended to PNI
Hindi
a. /-ko/ ↔ [-OBL,+OBJ] b. /-ne/ ↔ [-OBL,+SUBJ] ¹² I ignore numerals above > 1 and weak quantifiers, as there is not enough information present in the literature generally as well as with regard to their specificity properties. ¹³ One could alternatively conceive of DEM as [±PROX] and [±DIST].
56
PSEUDO-INCORPORATION VS DIFFERENTIAL OBJECT MARKING c. /−∅/ ↔ [ ] d. [+OBJ] → ∅ /[-HUM,-SPEC]
DOM rule
e. [+OBJ] → ∅ /[-HUM,DEM]
added DOM rule
f. [+OBJ] → ∅ /[V,-HUM]
added PNI rule
An OT-style approach can cover the PNI data equally well. All we need to do is extend the definiteness scale so that it also covers hybrid determiners, a proposal in line with von Heusinger and Kaiser (2007) and von Heusinger (2008). In doing so, we have to add other argument types to this scale, again as side effect of this study. The hybrid determiner PNI-D is ranked at the low end of the scale, implying that the less nominal an argument is, the less prominent it will be in discourse. The adjusted scale is given in (40). (40) Elaborate definiteness scale including PNI: PRO ≻ PN ≻ DEF ≻ SQ ≻ INDEF SPEC ≻ DEM ≻ INDEF NON-SPEC ≻ PNI-D ⏟⎴⎴⏟⎴⎴⏟ ⏟⎴⎴⎴⎴⏟⎴⎴⎴⎴⏟ ek NP
ek NP
Finally, we adopt the OT-ranking proposed in Keine and Müller (2008) and include the observations related to PNI; see (41). Along with non-human, non-specific indefinites, faithfulness requirements of PNI-ed arguments are ranked lower than the markedness constraint *[+GOV] which triggers case loss, shown in (43). The markedness constraint itself is not ranked with respect to the faithfulness requirements of non-human demonstrative phrases, which is why they can optionally be case marked, shown in (42). (41) (Keine and Müller 2008: 105) extended to PNI
Hindi
⎧ *OBJ/HUM/SPEC & MAX-C ⎫ *OBJ/HUM/NSPEC & MAX-C ⎨ ⎬ *OBJ/NHUM/SPEC & MAX-C ⎩ ⎭ *[+GOV] ≫{ } *OBJ/NHUM/DEM & MAX-C ≫{
*OBJ/NHUM/NSPEC & MAX-C } *OBJ/NHUM/PNI & MAX-C
(42) Demonstrative are optionally case marked
Hindi
4.3 CASE STUDIES
(43) PNI-ed objects are not case marked
57
Hindi
This section has shown how case loss due to PNI can be integrated into postsyntactic DOM-accounts. The rule-based as well as the constraint-based solution make use of the hybrid nature of the PNI determiner. The rule-based approach adds the [V ]-feature into its contextual restriction, while the constraint-based approach adds PNI-ed arguments to the definiteness scale. The next section will run the two systems for each of the PNI languages of our core data set.
4.3 Case studies We will now take a closer look at the PNI languages of our study. Table 2.1 from Section 2.3 serves as a summary of the individual case studies presented in this section. If an argument type allows for case drop, further tests were run probing for a correlation with low scope, lack of binding and control properties as well as movement restrictions. If argument types allow for case drop but do not show a correlation with either of the other PNI properties listed above, I conclude that case loss is a DOM-effect, while also running additional specificity tests. The language individual analyses all follow the same rationale: Noun types which cannot undergo PNI constitute DPs either of type ⟨e⟩ or ⟨⟨e, t⟩, t⟩, whereas noun types which can undergo PNI will be turned into arguments by the PNI determiner. For Korean, however, we will make use of an additional ident type-shifter (Partee 1986a,b). As discussed in Section 2.3, Table 2.1 roughly confirms the well-known size restriction of PNI-ed arguments, documented at least for Turkish objects (Enc˛ 1991; Özyıldız 2017), Mongolian objects (Guntsetseg 2016), and partially also for Tamil objects (Lehmann 1993). The PNI/DOM-data of Korean subject and objects are partially based on Driemel and Lee (2022). Furthermore, this book extends the Turkish data set to subject/agent PNI. For Mongolian and Tamil, data relevant to control and binding are contributed. While animacy is a decisive factor for case drop in Hindi (Mohanan 1995), Spanish (Bleam 2000; Torrego Salcedo 1999) and Tamil, its effect is only noticeable as subtle tendencies in Mongolian (Guntsetseg 2008) and Korean (Ha. Lee 2006, 2008). As we have seen in the previous section, if an animacy distinction is active in a PNI/DOM language, it will be relevant for both categories, i.e. noun types undergoing PNI as well as DOM will be sensitive to a [±ANIMATE]-feature. This can be seen as an argument for the OT-based model (Aissen 2003; Keine and Müller
58
PSEUDO-INCORPORATION VS DIFFERENTIAL OBJECT MARKING
2008) since Local Conjunction of the animacy scale with the definiteness scale will target each element on the definiteness scale, independent of whether they constitute arguments with the PNI determiner. The DM-model, in contrast, needs to write the animacy feature into DOM- and PNI-related rules, potentially predicting animacy to be relevant to only one of them. As far as this study goes, however, animacy is always relevant to both PNI and DOM categories. What might be construed as an argument against the OT-model is the fact that there do not seem to be scale reversal effects. Turkish and Korean show subject PNI, yet they do not exhibit the mirror image of object PNI. Rather, they show parallel definiteness requirements for subjects and objects. Since scale reversal is one of the main motivations for the use of an OT-system, deriving the reversal by Harmonic Alignment, one might wonder why this part of the OT-system does not seem to be active in the languages of this study (although see also Müller and Thomas (2017) for adjustments to the GF scale). I will, henceforth, ignore the GF scale for the constraint rankings. The grammatical function of the arguments which can be targeted by PNI/DOM will be determined by the decomposed case feature which makes up either the markedness constraint or the context for the deletion rule. Syntactic case features are decomposed into subfeatures. We have already seen case subfeatures such as [±OBLIQUE] and [±OBJECT], used in the analyses for Hindi in Section 4.2, as well as for Korean in Section 3.3. The nature of these features as well as the composition is to some extent arbitrary (Embick and Noyer 2007; Halle 1997), apart from them forming natural classes for which we see syntactic evidence. The case feature [−OBLIQUE] captures core arguments of the verb, i.e. NOM and ACC within our core data set, while [+GOVERNED] covers cases assigned to arguments which are governed by verbs or prepositions, i.e. ACC, DAT (Bierwisch 1967; Wiese 1999). I will make use of these features, together with [±STRUCTURAL] argued for by Embick and Noyer (2007) and Halle (1997) to distinguish between case assigned on semantic grounds vs. case assigned due to syntactic position.¹⁴ The set-up, which will be more or less the same in each language, is given in (44). (44) Case decomposition a. DAT, INST, LOC, COM: [+GOV,+OBL,−STRUC, ...] b. NOM: [−GOV,−OBL,+STRUC] c. ACC: [+GOV,−OBL,+STRUC] d. GEN: [−GOV,+OBL,+STRUC] One final note on specificity. As we have seen in the previous section, specificity and scope properties do not have to lign up. Hindi case-marked nouns obligatorily receive specific readings while still being scopally flexible. Table 4.1 shows that there is considerable variation across languages. ¹⁴ Other frequently used features are [+SUBJECT], referring to the highest argument (Müller 2004a; Wiese 2001; Wunderlich 1997) and [+N] singling out the case assigned by nominals (Bierwisch 1967; Müller 2005).
4.3 CASE STUDIES
59
Table 4.1 Specificity patterns of indefinites for DOM and PNI
Although there is a long tradition that relates case marking to specificity in Turkish (Enc˛ 1991; Kelepir 2001; von Heusinger and Kornfilt 2005), recent studies (von Heusinger and Bamyacı 2016; von Heusinger et al. 2019) have shown that there is at least no clear evidence for a correlation between accusative marking and epistemic specificity. The specificity tests run for Korean align with this impression. Spanish allows case-marked objects to be specific as well as non-specific, while caseless nouns can only be specific. For Mongolian, we find conflicting judgements in the literature. Finally, Tamil shows a clear split when it comes to specificity, similar to Hindi.
4.3.1 Tamil Tamil shows a nominative-accusative case alignment system. Since nominative case is unmarked, the discussion of DOM/PNI effects is restricted to objects. Like Hindi, Tamil PNI/DOM is sensitive to animacy. Only objects referring to non-human entities can occur without case marking; see (45). (45) Lehmann (1993: 29) a. Kumaar oru peṭ ṭ iy(-ai) vaank-in-aan. Kumar a box-ACC buy-PST-3SG.M ‘Kumar bought a box.’ b. Kumaar oru paiyan*(-aip) paar-tt-aan. Kumar a boy-ACC see-PST-3SG.M ‘Kumar saw a boy.’ Lehmann (1993) reports that proper names and possessive phrases require case marking in Tamil. I can confirm the judgements with my consultants and extend the observation to pronouns. The data is given in (46). Note also that the case
60
PSEUDO-INCORPORATION VS DIFFERENTIAL OBJECT MARKING
marking requirement is tied to the noun type and not animacy, as the inanimate arguments in (46b) and (46e) demonstrate. (46) Personal pronouns, proper names, possessives a. Kumaar Kamalaav*(-aik) kaatali-kkir-aan. Kumar.NOM Kamala-ACC love-PRS-3SG.M ‘Kumar loves Kamala.’
Tamil (Lehmann 1993: 28)
b. Kumaar Avengers*(-ai) veru-t-aan. Kumar.NOM Avengers-ACC hate-PST-3SG.M ‘Kumar hated Avengers (the movie).’ c. Kumaar unn*(-aik) kaatali-kkir-aan. Kumar.NOM 2SG-ACC love-PRS-3SG.M ‘Kumar loves you.’ d. Kumaar enn*(-aik) kaatali-kkir-aan. Kumar.NOM 1SG-ACC love-PRS-3SG.M ‘Kumar loves me.’ e. Naan en cavi*(-ya) tolai-t-een. 1SG.NOM 1SG.OBL key-ACC lose-PST-1SG ‘I lost my key.’
(Lehmann 1993: 28)
Lehmann (1993: 28) reports that definite noun phrase are obligatorily case marked. On closer inspection, however, we see that this only applies to strong definites. With an anaphoric context, given in (47), objects are only licensed if they are case-marked.¹⁵
¹⁵ The fact that the use of bare nouns suffices to trigger strong definite readings is interesting with respect to recent findings by Jenks (2015, 2018) who concludes from observations on Thai and Mandarin Chinese that bare argument languages tend to use bare nouns for weak definite contexts, whereas demonstratives or overt pronouns take over the function of strong familiar definites. It seems that languages can additionally make use of case marking to signal a strong reading without the use of demonstratives. The context below shows that bare nouns can also create strong definite readings in subject position. The additional use of a demonstrative determiner is possible but not necessary. (i) Strong definites in subject position Inniki John booku vaŋgi-n-aan. (Inta) booku romba vela jasti. today John.NOM book buy-PST-3SG.M DEM book.NOM very cost high ‘John bought a book today. That/the book was expensive.’
Tamil
4.3 CASE STUDIES
61
(47) Strong definites require case marking Tamil Inniki John pustagam vaŋgi-n-aan, naan pustagath-ai / today John.NOM book buy-PST-3SG.M 1SG.NOM book-ACC / ??
pustagam paar-kale. book see-NEG ‘Today John bought a book, (but) I didn’t see the book.’ Testing for weak definite uses, we can observe that case marking is not obligatory. In (48), weak definiteness is triggered globally by world knowledge.¹⁶ In (49), we set up a context which creates immediate weak definiteness. (48) Weak definites are optionally case marked (global context) Naan nethikku nila.v-ai / nila paar-t-een. 1SG.NOM yesterday moon-ACC / moon see-PST-1SG ‘I saw the moon yesterday.’
Tamil
(49) Context (immediate weak definite): Imagine a scenario in an office room that has a projector for presentations. Inniki projector-ai / projector repair panni-n-een. today projector-ACC / projector repair do-PST-1SG ‘I repaired the projector today.’
Tamil
Like in Hindi, demonstrative phrases are only optionally case marked; see (50).¹⁷ (50) Demonstratives are optionally case marked Maala inta pustagath-ai / pustagam paɖi-cc-aa. Mala.NOM this book-ACC / book read-PST-3SG.F ‘Mala read this book.’
Tamil
Quantifiers can also lose case marking; see (51a) for a strong quantifier¹⁸ and (51b) and (51c) for weak ones.¹⁹ ¹⁶ We see the animacy restriction re-occurring with weak definites. If the referent is unique but human, case marking becomes obligatory. (i) Human weak definites require case marking Naan Rani*(-ya) paar-t-een. 1SG.NOM queen-ACC see-PST-1SG ‘I saw the queen.’
Tamil
¹⁷ One of my four informants disagrees strongly with the caseless version in (50). ¹⁸ Schiffman (1999: 130) offers ‘all’ as well as ‘every’ as translations. ¹⁹ One of my four informants can only drop case on weak quantifiers and not on strong ones such as in (51a).
62
PSEUDO-INCORPORATION VS DIFFERENTIAL OBJECT MARKING
(51) Quantifiers are optionally case marked
Tamil
a. Naan ella biscuit(-ai)-um saapi-tt-een. 1SG.NOM all biscuit-ACC-ADD eat-PST-1SG ‘I ate all biscuits.’ b. Naan konja biscuit(-ai) saapi-tt-een. 1SG.NOM a.little biscuit-ACC eat-PST-1SG ‘I ate some biscuits.’ c. Naan pala biscuit(-ai) saapi-tt-een. 1SG.NOM many biscuit-ACC eat-PST-1SG ‘I ate many biscuits.’ Finally, Lehmann (1993: 28–29) observes that bare nouns and indefinites can lose case. Baker (2014b) discusses bare nouns but mentions in a footnote (Baker 2014b: 9) that indefinites can also be caseless. I extend the observation to numerals > 1. The data is given in (52). (52) Bare nouns, indefinites, numerals are optionally case marked
Tamil
a. Maala pustagath-ai / pustagam paɖi-cc-aa. Mala.NOM book-ACC / book read-PST-3SG.F ‘Mala read a book.’ b. Maala oru pustagath-ai / pustagam paɖi-cc-aa. Mala.NOM a/one book-ACC / book read-PST-3SG.F ‘Mala read a/one book.’ c. Maala rendu pustagath-ai / pustagam paɖi-cc-aa. Mala.NOM two book-ACC / book read-PST-3SG.F ‘Mala read two books.’ Out of the class of noun types which can be caseless, i.e. bare nouns, indefinites, numerals, quantifiers, weak definites and demonstratives, only bare nouns use case drop to signal pseudo-noun incorporation. The other noun types are optionally case marked as the result of differential object marking. Evidence for the difference between bare nouns and indefinites comes from the scope context in (53) that enforces a wide scope reading. PNI-ed arguments are restricted to low scope readings, while no such requirement is present with DOM-type nouns. In (53), caseless bare nouns are not licensed, in contrast to caseless indefinites. Section 5.4.1 will show that numerals and weak quantifiers pattern with indefinites and not with bare nouns. In fact, I will analyse oru as numeral one, akin to what Dayal (2004) proposed for Hindi ek.
4.3 CASE STUDIES
63
(53) Context ∃¬: A friend of mine wants to sell three books and after looking at them I bought two books from him. So there is a book which I did not buy from him. a.
Naan oru pustagam vaŋga-lle. 1SG.NOM one book buy-NEG ‘I didn’t buy a book.’
Tamil
b. # Naan pustagam vaŋga-lle. 1SG.NOM book buy-NEG ‘I didn’t buy a book.’ We can show that caseless weak definites and demonstratives are not PNI-ed by testing their binding capabilities. As (54b) and (54c) suggest, binding is not affected by case drop. Again, bare nouns serve as the baseline, they can only bind a possessive pronoun if case marked; see (54a). Section 6.1 will provide the full picture. (54) Binding for bare nouns, weak definites, and demonstratives
Tamil
a. Kumar book*(-ai)1 ata-ode1 kadasi pakkam varai Kumar.NOM book-ACC that-GEN last page until padi-c-aan. read-PST-3SG.M ‘Kumar read a book until its last page.’ b. Naan Jupiter(-ai)-um1 ate-ode1 nila.v-ai-um paar-t-een. 1SG.NOM Jupiter-ACC-CONJ that-GEN moon-ACC-CONJ see-PST-1SG ‘I saw Jupiter and its moon.’ c. Kumar [inta book(-ai)]1 ata-ode1 kadasi pakkam varai Kumar.NOM DEM book-ACC that-GEN last page until padi-c-aan. read-PST-3SG.M ‘Kumar read this book until its last page.’ Finally, quantifiers pattern with weak definites and demonstratives with respect to binding, as can be observed in (55). (55) Binding for quantifiers
Tamil
a. Kumar [ella book(-ai)-um]1 ata(-t)-ode1 kadasi pakkam Kumar.NOM all book-ACC-ADD that-RED-GEN last page varai padi-c-aan. until read-PST-3SG.M ‘Kumar read all the books until their last pages.’
64
PSEUDO-INCORPORATION VS DIFFERENTIAL OBJECT MARKING b. Kumar [pala book(-ai)]1 ata-ode1 kadasi pakkam varai Kumar.NOM many book-ACC that-GEN last page until padi-c-aan. read-PST-3SG.M ‘Kumar read many books until their last pages.’
Before we move on to the analysis, let us quickly summarize the findings in (56). (56) Interim summary of case marking properties in Tamil a. Noun types obligatorily case marked possessor phrases, strong definites, pronouns, proper names b. Noun types optionally case marked due to DOM indefinites, numerals, demonstratives, strong quantifiers, weak quantifiers, weak definites c. Noun types optionally case marked due to PNI bare nouns Like in Hindi, only bare nouns undergo PNI when they lose case in Tamil. Hence, I assume identical underlying nominal structures, where numerals, including oru, as well as quantifiers, demonstratives and weak definites constitute D heads. In (57), I list the syntactic structures which can undergo case drop, either via PNI or via DOM. While PNI-DPs, i.e. bare nouns, can lose case marking due to [V ], the other determiners are targeted by DOM-rules. (58) lists the vocabulary items for the structures. In contrast to Hindi, Tamil can shift bare nouns to an existential reading, i.e. case-marked bare nouns do not automatically receive a definite reading. Hence, Tamil exhibits an existential quantifier in addition to numeral quantifiers. Evidence for the existential shift is based on discourse new contexts and existential sentences, which are discussed in Section 5.4.1. Denotations of the relevant D heads are given in (59), together with the PNI determiner. (57) Tamil determiners with case drop:
PNI-DP DPNI NP [D ≻ V ]
a
DP DPone NP [ ,#1, D]
DP
DP
D NP [ , D]
D NP [ , D]
DP wDɩ NP [w-ɩ, D] DP DPmany NP [ , |x| > n, D]
DP Ddem NP [DEM, D]
4.3 CASE STUDIES
65
(58) Vocabulary entries for (some) Tamil determiners: a. /oru/ ↔ [#1]
Done
b. /ella/ ↔ [∀]
D∀
c. /pala/ ↔ [|x| > n]
Dmany
d. /inta/ ↔ [DEM] e. /∅/ ↔ [ ]
Ddem DPNI , D∃ , wDι , sDι
(59) Determiner denotations for oru NP and bare nouns a. ⟦D ≻ V⟧ = λP⟨e,t⟩ λQ⟨e,⟨v,t⟩⟩ λe∃z[P(z) ∧ Q(z)(e)]
DPNI
b. ⟦∃, #1, D⟧ = λP⟨e,t⟩ λQ⟨e,t⟩ ∃z[P(z) ∧ |z| = 1 ∧ Q(z)]
Done
c. ⟦∃, D⟧ = λP⟨e,t⟩ λQ⟨e,t⟩ ∃z[P(z) ∧ Q(z)]
D∃
Tamil shows a one-to-one correlation with case marking and specific readings. The context in (60) enforces a specific reading for the object by placing the referent in an unusual location, so that the speaker will a have a concrete referent in mind when uttering the sentence. The caseless version of the indefinite in (60b) is dispreferred in such a context. (60) Context [+spec]: We are gathering chairs for all the family members who have come to sit at the table but there are still some missing. a.
Naan bathroom-le oru chair-ai paar-t-een. 1SG.NOM bathroom-LOC one chair-ACC see-PST-1SG ‘I saw a chair in the bathroom.’
Tamil
b. # Naan bathroom-le oru chair paar-t-een. 1SG.NOM bathroom-LOC one chair see-PST-1SG ‘I saw a chair in the bathroom.’ Following the reference grammars (Lehmann 1993; Schiffman 1999), I propose the case suffixes in Tamil to be decomposed and underspecified as in (61), where I consider nominative, accusative, and genitive to be structural cases. (61) Case vocabulary items in Tamil: a. DAT, INST, LOC, COM ↔ [+GOV,+OBL,−STRUCT,...] b. /-(y)ai/ ↔ [+GOV,+STRUCT]
(ACC)
c. /-ode/ ↔ [+OBL]
(GEN)
d. /-∅/ ↔ [ ]
(NOM)
66
PSEUDO-INCORPORATION VS DIFFERENTIAL OBJECT MARKING
In line with the observations above, we can now formulate the impoverishment rules as in (62), where (62c), (62d), and (62e) apply optionally. Each rule deletes [+GOV], leading to the insertion of the elsewhere marker (61d) in accusative contexts of weak definites (62c), demonstratives (62d), and strong quantifiers (62e), under the assumption that case is decomposed as in (44). The rule in (62b) refers to weak quantifiers, including indefinites and numerals. Although we only saw a specificity difference for oru NP, it is likely that other numerals and weak quantifiers show the same correlation. If this is true, we can make impoverishment apply to ∃ capturing a specificity distinction across categories.²⁰ (62) Impoverishment rules in Tamil: a. [+GOV] → ∅ / [V,−HUM]
PNI
b. [+GOV] → ∅ / [∃,−SPEC,−HUM]
DOM
c. [+GOV] → ∅ / [W-ι,−HUM]
DOM
d. [+GOV] → ∅ / [DEM,−HUM]
DOM
e. [+GOV] → ∅ / [∀,−HUM]
DOM
Turning now to the OT-implementation, I provide a definiteness scale in (63), which includes the noun types we have taken into consideration in this section. Again, I group indefinites, numerals, and weak quantifiers under one category INDEF, reflecting the intuition of (62b).²¹ Weak definites, strong quantifiers, and demonstratives can optionally lose case. This is implemented by placing them between specific and non-specific scale mates where they are not ranked with respect to the markedness constraint *[+GOV]; see the ranking in (64) where only the relevant parts are shown. Non-specific indefinites as well as PNI-ed objects are ranked below the markedness constraint, thus they occur without case marking. (63) Definiteness scale in Tamil: PRO ≻ PN ≻ SDEF ≻ INDEF SPEC ≻ {WDEF, SQ, DEM } ≻ ⏟⎴⎴⏟⎴⎴⏟ oru∕#∕wQ∕∃
INDEF N-SPEC ≻ PNI-D ⏟⎴⎴⎴⏟⎴⎴⎴⏟ oru∕#∕wQ∕∃
²⁰ If this is not true, we have to reformulate (62b), so that it applies to a more specific context. ²¹ Similarly, if it turns out that only oru NPs show specificity sensitivity, we have to group numerals and weak quantifiers with demonstratives, strong quantifiers, and demonstratives.
4.3 CASE STUDIES
67
(64) Constraint-ranking in Tamil: ⎧ *HUM/SPEC & MAX-C ⎫ *HUM/NSPEC & MAX-C ⎨ ⎬ *NHUM/SPEC & MAX-C ⎩ ⎭ ⎧ ⎪ ≫ ⎨ ⎪ ⎩
*[+GOV] *NHUM/DEM & MAX-C *NHUM/W-ι & MAX-C *NHUM/∀ & MAX-C
⎫ ⎪
⎬ ⎪ ⎭ *NHUM/NSPEC & MAX-C ≫{ } *NHUM/PNI & MAX-C
Summing up, this section has provided an explanation why Tamil only pseudoincorporates bare nouns. PNI-D, the PNI determiner, is in complementary distribution with indefinite, numeral, and other D-heads. Noun types other than bare nouns can show case loss but this is an effect related to DOM. I have shown how post-syntactic impoverishment can capture both phenomena within one system, be it rule-based or constraint-based. The next section will discuss Mongolian where case loss uniformly signals PNI.
4.3.2 Mongolian The case system of Mongolian shows a nominative-accusative alignment, where nominative case is unmarked. In contrast to Tamil, animacy does not play a decisive role. Pronouns, proper names, and demonstratives require case marking (see (65a)), the constraint also holds for pronouns shown in (65b). The same goes for possessive phrases, shown in (66). (65) Guntsetseg and Klein (2009), Guntsetseg (2016: 81)
Mongolian
a. Bi tuun*(-ig) / Tuya*(-g) / ene uul*(-ig) har-san. I 3.ACC / Tuya-ACC / this mountain-ACC see-PST ‘I saw him/her/Tuya/this mountain.’ b. Bi cˇamaig / *cˇi xar-san. I you.ACC / you see-PST ‘I saw you.’ (66) Guntsetseg (2016: 61) Tujaa minij nom*(-yg) unsˇ-san. Tujaa 1SG.GEN book-ACC read-PST ‘Tujaa read my book.’
Mongolian
68
PSEUDO-INCORPORATION VS DIFFERENTIAL OBJECT MARKING
Definite noun phrases also require case marking. Weak contexts are provided in (67) and (68), strong contexts in (69) and (70). I was able to confirm (67) in my study. For immediate definiteness, I additionally set up a context in (68), again confirming Guntsetseg’s intuition that weak definites need to be case marked. For strong definiteness, Guntsetseg provides the context in (69) which I used in the study. I also tested bare nouns and case marking with the same context; see (70). The data in (67)–(70) provide evidence that definiteness requires case marking. They are also in line with Jenk’s (2015, 2018) generalization for bare argument languages, in that weak definiteness is expressed with a bare noun, whereas strong definiteness makes use of a demonstrative determiner. (67) Guntsetseg (2016: 83) Delxij nar*(-yg) tojr-dog. earth sun-ACC circle-HAB ‘The earth circles around the sun.’
Mongolian
(68) Context (immediate weak definite): Imagine a school scenario where a student asks another student where the director is. Tujaa zaxirl# (-yg) xar-san. Tujaa.NOM director-ACC see-PST ‘Tujaa saw the director.’ (69) Context (strong definite): My father gave me a very interesting book... a.
Bi ter nom-yg xurdan unsˇ-san. I DEM book-ACC quickly read-PST ‘I read the book quickly.’
Mongolian
(Guntsetseg 2016: 81) Mongolian
b. # Bi ter nom xurdan unsˇ-san. I DEM book quickly read-PST ‘I read the book quickly.’ (70) Context (strong definite): My father gave me a very interesting book... a. #Bi nom xurdan unsˇ-san. 1SG.NOM book quickly read-PST ‘I did book-reading quickly.’ b. # Bi nom-yg xurdan unsˇ-san. 1SG.NOM book-ACC quickly read-PST ‘I read the book quickly.’
Mongolian
4.3 CASE STUDIES
69
Guntsetseg (2016) also discusses quantifiers but focuses on strong quantifiers which require case marking; see (71). Again, these judgements were confirmed in my study. (71) Guntsetseg (2016: 83-84)
Mongolian
a. Zarixal bu¨x ojuutn-yg / *[bu¨x ojuutan] ˇsalga-san. director all student-ACC / all student examine-PST ‘The director examined all students.’ b. Zarixal [ojuutn bu¨xn]-yg / *[ojuutn bu¨xen] ˇsalga-san. director student every-ACC / student every examine-PST ‘The director examined every student.’ c. Zarixal [ojuutn bolgon]-yg / *[ojuutn bolgon] ˇsalga-san. director student every-ACC / student every examine-PST ‘The director examined every student.’ d. Zarixal [ojuutn bu¨r]-ijg / *[ojuutn bu¨r] ˇsalga-san. director student each-ACC / student each examine-PST ‘The director examined each student.’ e. Zarixal [ixenx ojuutn]-yg / *[ixenx ojuutan] ˇsalga-san. director most student-ACC / most student examine-PST ‘The director examined most students.’ Guntsetseg (2016: 89–90) discusses optional case marking for indefinites, mentioning numerals and the quantifier xeden (= some) in passing. I can extend this class to weak quantifiers generally; see (72). (72) Weak quantifiers and numerals are optionally case marked
Mongolian
a. Zarixal [xojor ojuutn]-yg / [xojor ojuutan] ˇsalga-san. director.NOM two student-ACC / two student examine-PST ‘The director examined two students.’ b. Zarixal [olon ojuutn]-yg / [olon ojuutan] ˇsalga-san. director.NOM many student-ACC / many student examine-PST ‘The director examined many students.’ c. Zarixal [co¨o¨xo¨n ojuutn]-yg / [co¨o¨xo¨n ojuutan] ˇsalga-san. director.NOM a.few student-ACC / a.few student examine-PST ‘The director examined a few students.’ d. Zarixal [xeden ojuutn]-yg / [xeden ojuutan] ˇsalga-san. director.NOM some student-ACC / some student examine-PST ‘The director examined some students.’
70
PSEUDO-INCORPORATION VS DIFFERENTIAL OBJECT MARKING
Caseless bare nouns as well as caseless indefinites show a low scope restriction if they are not marked for case, thereby showing a key characteristic of PNI-ed objects. The scope restriction is demonstrated in (73b) and (73c) where the context enforces a wide scope reading. A case-marked indefinite is the optimal choice; see (73a). (73) Context ∃¬: There are two apples on the table of which Bold ate one. So there is one apple on the table Bold did not eat. a.
Bold neg alim-yg idee-gu¨j. Bold.NOM a apple-ACC eat-NEG ‘There is one apple Bold didn’t eat.’
Mongolian
b. # Bold neg alim idee-gu¨j. Bold.NOM a apple eat-NEG ‘Bold did not eat one apple.’ c. #Bold alim idee-gu¨j. Bold.NOM apple eat-NEG ‘Bold did not eat any apple.’ Caseless bare nouns are only felicitous in contexts that require a low scope reading, as in (74b). Compare also the case-marked bare noun in (74a) which receives a specific/wide scope reading and is thus infelicitous. The observations in (73) and (74) can be extended to numerals > 1 and the weak quantifier xeden; see Section 5.4.2. (74) Context ¬∃: There are some apples on the table of which Bold did not eat any. a. # Bold alim-yg idee-gu¨j. Bold.NOM apple-ACC eat-NEG ‘There are apples Bold didn’t eat.’ b.
Bold alim idee-gu¨j. Bold.NOM apple eat-NEG ‘Bold did not eat any apples.’
We have seen so far that, in contrast to Tamil, Mongolian allows for a much smaller class of noun types to be caseless. The nouns that can undergo case drop show a correlating low scope restriction, supporting evidence also comes from binding in Chapter 6, Section 6.2. We can derive PNI effects for weak quantifiers, numerals and indefinites by analysing them as adjectival modifiers (Kamp and Reyle 1993; Link 1987; Milsark 1977; Partee 1988; Zimmermann 2008). Once modified, they
4.3 CASE STUDIES
71
can merge either with PNI-D, the PNI determiner which triggers case drop (75), or with an existential quantifier which enforces case marking (76).²² As in Tamil, there is independent evidence for the existence of D∃ , given in Section 5.4.2. For comparison’s sake, we provide the semantic denotations for neg, as well as the PNI determiner and the existential quantifier in (78). The modifier neg combines with NPs via Predicate Modification (Heim and Kratzer 1998). (75) Mongolian determiners with case drop:
PNI-DP PNI-DP DPNI NP [D ≻ V ]
DPNI [D ≻ V]
PNI-DP DPNI [D ≻ V]
NP DPone NP [ ,#1, D]
NP DPsome NP [ , |x| < n, D]
a (76) Case-marked Mongolian indefinites:
DP DP D NP [ , D] a
D [ , D]
DP NP
DPone NP [ ,#1, D]
D [ , D]
NP
DPsome NP [ ,|x| < n, D]
(77) Vocabulary entries for (some) Mongolian determiners: a. /neg/ ↔ [#1]
Done
b. /bu¨r/ ↔ [∀]
D∀
c. /xeden/ ↔ [|x| < n]
Dsome
d. /ene/ ↔ [DEM]
Ddem
e. /ter/ ↔ [S-ι] f. /∅/ ↔ [ ]
sDι DPNI , D∃ , wDι
(78) Determiner denotations for neg NP and bare nouns a. ⟦D ≻ V⟧ = λP⟨e,t⟩ λQ⟨e,⟨v,t⟩⟩ λe∃z[P(z) ∧ Q(z)(e)]
DPNI
b. ⟦∃, #1, D⟧ = λz.|z| = 1
Done
c. ⟦∃, D⟧ = λP⟨e,t⟩ λQ⟨e,t⟩ ∃z[P(z) ∧ Q(z)]
D∃
²² The majority of the feature bundles is self-explanatory. One note on Dsome : I take n to indicate a contextual standard. Thus, Dsome establishes an amount below the standard, whereas e.g. Dmany , cf. (57), creates an amount above the standard.
72
PSEUDO-INCORPORATION VS DIFFERENTIAL OBJECT MARKING
When it comes to specificity, authors disagree on the judgements. Guntsetseg (2009, 2016) reports that case-marked indefinites are obligatorily interpreted as specific, while the caseless versions can be specific or non-specific. This observation is made clear with the continuations, given in (79). She provides parallel examples with numerals and xeden; see (80) and (81). (79) Guntsetseg (2016: 97-98)
Mongolian
a. Xen neg n minij zugluulgan-aas neg nom-yg xulgajl-zˇee someone my collection-ABL a book-ACC steal-PST ‘Someone stole a book from my collection.’ ... It is the original ‘The secret history of the Mongols’. ... #I’m trying to find out which book it is.
+SPEC
b. Xen neg n minij zugluulgan-aas neg nom xulgajl-zˇee someone my collection-ABL a book steal-PST ‘Someone stole a book from my collection.’ ... It is the original ‘The secret history of the Mongols’. ... I’m trying to find out which book it is.
±SPEC
(80) Guntsetseg (2016: 89)
Mongolian
a. Tujaa gurvan ajag-yg ugaa-san. Tujaa three cup-ACC wash-PST ‘Tujaa washed three certain cups.’
+SPEC
b. Tujaa gurvan ajaga ugaa-san. Tujaa three cup wash-PST ‘Tujaa washed three (certain) cups.’
±SPEC
(81) Guntsetseg (2016: 90)
Mongolian
a. Zaxiral o¨no¨o¨do¨r xeden ojuutn-yg ˇsalga-san. director today some student-ACC control-PST ‘The director controlled today some students.’
+SPEC
b. Zaxiral o¨no¨o¨do¨r xeden ojuutn ˇsalga-san. director today some student control-PST ‘The director controlled today some students.’
±SPEC
Janhunen (2012) presents a different picture, shown in (82). While case-marked bare nouns can be specific or non-specific, caseless bare nouns can only be non-specific.
4.3 CASE STUDIES
(82) Janhunen (2012: 206)
73
Mongolian
a. neg mory-iig ab-sen. one horse-ACC buy-PFV ‘I bought a (certain) horse.’
±SPEC
b. neg mory ab-sen. one horse buy-PFV ‘I bought a horse.’
−SPEC
I included the paradigm in (79) in my online survey, presented as in (83) and (84). None of my five speakers detected a specificity effect, neither for indefinites nor for bare nouns. With the continuations turned into contexts, case-marked as well as caseless nouns seem to be felicitous in [+spec] as well as [−spec] scenarios. (83) Context [+spec]: Someone broke into my apartment and I see that there is a book missing from the library. I immediately realize that it is the original ‘The secret history of the Mongols’. So I tell the police: a. Xen neg n minij zugluulgan-aas neg nom-yg xulgajl-zˇee someone my collection-ABL a book-ACC steal-PST ‘Someone stole a book from my collection.’ b. Xen neg n minij zugluulgan-aas neg nom xulgajl-zˇee someone my collection-ABL a book steal-PST ‘Someone stole a book from my collection.’ c. Xen neg n minij zugluulgan-aas nom xulgajl-zˇee someone my collection-ABL book steal-PST ‘Someone stole a book from my collection.’ (84) Context [−spec]: Someone broke into my apartment and I see that there is a book missing from the library. I don’t know which book it is. So I tell the police: a. Xen neg n minij zugluulgan-aas neg nom-yg xulgajl-zˇee someone my collection-ABL a book-ACC steal-PST ‘Someone stole a book from my collection.’ b. Xen neg n minij zugluulgan-aas neg nom xulgajl-zˇee someone my collection-ABL a book steal-PST ‘Someone stole a book from my collection.’ c. Xen neg n minij zugluulgan-aas nom xulgajl-zˇee someone my collection-ABL book steal-PST ‘Someone stole a book from my collection.’
74
PSEUDO-INCORPORATION VS DIFFERENTIAL OBJECT MARKING
I tentatively conclude from the discussion above that epistemic specificity does not correlate with case marking, as it was recently also argued for in Turkish (von Heusinger and Bamyacı 2016; von Heusinger et al. 2019). I will now show how these observations can be implemented in a post-syntactic framework. I propose the case suffixes to be encoded as in (85), following Janhunen (2012); Guntsetseg (2016). (85) Case vocabulary items in Mongolian: a. DAT, INST, ABL, POSS, PRIV ↔ [+gov,+obl,–struct,...] b. /-(g)iig/ ↔ [+GOV,+STRUCT]
(ACC)
c. /-iin/ ↔ [+OBL]
(GEN)
d. /-∅/ ↔ [ ]
(NOM)
Since Mongolian is a language that shows a consistent correlation between case drop and low scope restrictions, we need only one impoverishment rule, given in (86). [+GOV] is deleted on PNI-D, the PNI determiner, in an accusative context. This leaves the feature bundle proposed for accusative in (44) with an insertion context, only compatible with the elsewhere marker in (85d). As was shown in (75), weak quantifiers, numerals, and indefinites have a modifier status in Mongolian; they do not select for an NP, hence they attach to NP without forming an argument to the verb. This gives PNI-D the opportunity to take modified nouns as arguments and derive their PNI properties. [+GOV] is not deleted on D∃ or any other proper D head, which derives the rest of the paradigm, specifically the possibility for case-marked indefinites, weak quantifiers, and numerals to take wide scope. (86) Impoverishment rule for Mongolian: a [+GOV] → ∅ / [V ]
PNI
The constraint-based system is also simplified, due to the assumption that weak quantifiers, numerals, and indefinite determiners can adjoin to NP. I collapse the categories INDEF SPEC and INDEF N-SPEC on the definiteness scale in (87) since case drop does not seem to interact with specificity. The markedness constraint *[+GOV] is ranked between case faithfulness constraints of indefinites and PNI-ed arguments (88), thereby deriving the size restriction. (87) Definiteness scale in Mongolian: PRO ≻ PN ≻ DEF ≻ SQ ≻ DEM ≻ INDEF ≻ PNI ⏟⏟⏟ ⏟⏟⏟ neg∕#∕wQ
neg∕#∕wQ
4.3 CASE STUDIES
75
(88) Constraint-ranking in Mongolian: ⎧ *PRO & MAX-C ⎪ *PN & MAX-C ⎪ *DEF & MAX-C ⎨ *SQ & MAX-C ⎪ ⎪ *DEM & MAX-C ⎩ *INDEF & MAX-C
⎫ ⎪ ⎪ ⎬ ⎪ ⎪ ⎭
≫
{
*[+GOV]
}
≫
{
*PNI & MAX-C
}
This section presented the case drop pattern of a language which relates case loss exclusively to PNI properties. Based on the data by Janhunen (2012) and my own survey results, I concluded that specificity does not play a decisive role in Mongolian case marking. This assumption simplifies both the rule-based as well as the constraint-based analysis. The PNI language of the next section behaves very much like Mongolian.
4.3.3 Turkish Turkish pronouns, proper names,²³ definites, and demonstratives require case marking; see (89). A case-marked bare noun in Turkish automatically receives a definite reading (von Heusinger and Kornfilt 2005). Turkish behaves like Hindi in this respect and contrasts with Mongolian and Tamil. ²³ Interestingly, for some speakers it is possible to coerce a PNI reading with caseless proper names if the referent is sufficiently known and there is a PNI interpretation available which implies aspects ¨ zdemir, p.c.). beyond the literal meaning (Bilal O (i) Coercion with case drop on proper names a. Du¨n Harrison Ford-u go¨r-du¨-m. yesterday Harrison Ford-ACC see-PFV-1SG ‘I saw Harrison Ford yesterday.’ b. Du¨n Harrison Ford go¨r-du¨-m. yesterday Harrison Ford see-PFV-1SG ‘I did Harrison-Ford-seeing yesterday (I’m special now).’ Recall that in fn. 11, I suggest to analyse proper names traditionally as individual constants (Frege 1892; Heim and Kratzer 1998; Kripke 1972). There are, however, also proposals treating proper names as properties (Burge 1973; Elbourne 2002). If this is true for Turkish, the PNI determiner would be able to merge with a proper name, creating a PNI reading, whereas a definite determiner triggers the standard individual reading. The acceptability of caseless proper names in these rare scenarios could thus be accounted for if proper names are analysed as properties. Since PNI is generally restricted in its application to conventionalized activities, proper names have to refer to an iconic individual in order to be coerced into a PNI reading. Gehrke (2015) makes a similar observation for adjectival passives, which she argues pseudo-incorporate by-phrases in German: (ii) Gehrke (2015: 929) Das Manuskript ist von Chomsky / ?Sandberger zitiert. the manuscript is by Chomsky Sandberger cited ‘The manuscript is cited by Chomsky.’
German
76
PSEUDO-INCORPORATION VS DIFFERENTIAL OBJECT MARKING
(89) Enc˛ (1991: 9) Turkish Zeynep Ali*(-yi) / on*(-u) / adam*(-ı) / o masa*(-yı) go¨rdu. Zeynep Ali-ACC / he-ACC / the.man-ACC / that table-ACC saw ‘Zeynep saw Ali/him/the man/that table.’ (90) presents the pronominal paradigm and its requirement for case marking. (90) Personal pronouns require case marking
Turkish
a. Ali du¨n ben*(-i)/biz*(-i)/sen*(-i)/siz*(-i) go¨r-du¨-∅. Ali.NOM yesterday I-ACC/us-ACC/you.SG-ACC/you.PL-ACC see-PFV-3SG ‘A li saw me/us/you/you all yesterday.’ b. Ben du¨n o*(-nu)/onlar*(-ı) go¨r-du¨-m. 1SG.NOM yesterday 3.SG-ACC/3.PL-ACC see-PFV-1SG ‘I saw {him/her/it}/them yesterday.’ In (91), we see the identical case requirement for subjects. I test subject/agent incorporation in embedded clauses where subjects are overtly marked for genitive case.²⁴ (91) Genitive case requirement with subjects Turkish Sen-i Trump*(-ın) / o*(-nun) / Bas¸kan*(-ın) / o Bas¸kan*(-ın) 2SG-ACC Trump-GEN / (s)he-GEN / president-GEN / that president-GEN ara-dıg˘-ı-nı duy-du-m. call-NMLZ-POSS.3-ACC hear-PFV-1SG ‘I have heard that Trump / (s)he / the President / that President called you.’ The case requirement holds for strong as well as weak definites. I provide a global context in (92) and immediate contexts in (93) and (94), where we see obligatory case marking for objects as well as for subjects if uniqueness needs to be expressed. Anaphoric definiteness is probed for in (95), with parallel results. (92) Weak global definites require case marking:
Turkish
a. S¸ehr-e kralic¸e*(-nin) gel-dig˘-i-ni oku-du-m. city-DAT queen-GEN come-NMLZ-POSS.3SG-ACC read-PFV-1SG ‘I have read that the queen came to town.’ b. Du¨n kralic¸e*(-yi) go¨r-du¨-m. yesterday queen-ACC see-PFV-1SG ‘I saw the queen yesterday.’ ²⁴ The cause for subjects receiving genitive case in selected subordinate environments is often tied to a nominal licenser, either in form of a nominal Agr projection (Kornfilt 2003b) or a DP/NP shell (Aygen 2007b; Lees 1965).
4.3 CASE STUDIES
77
(93) Context (immediate weak definite): Imagine a scenario in an office room that has a projector for presentations. Bugu¨n projekto¨r# (-u¨) tamir et-ti-m. today projector-ACC fixing do-PFV-1SG ‘I fixed the projector today.’
Turkish
(94) Context (immediate weak definite): Imagine a situation in a classroom where students wait for their teacher to come. Bugu¨n hoca# (-nın) gel-eceg˘-i-ni so¨yle-di-ler. today teacher-GEN come-NMLZ-POSS.3SG-ACC say-PFV-3PL ‘They said that the teacher would come today.’ (95) Anaphoric definites are case marked
Turkish
Turkish
a. Bugu¨n internet-ten bir kitap al-dı-m. Yarın kitab# (-ın) today Internet-ABL a book buy-PFV-1SG tomorrow book-GEN gel-eceg˘-i yaz-ıyor-du. come-NMLZ-POSS.3SG write-PROG-PST ‘I bought a book online today. It said (on the website) that the book would come tomorrow.’ b. Bugu¨n internet-ten bir kitap al-dı-m. Yarın kitab# (-ı) today Internet-ABL a book buy-PFV-1SG tomorrow book-ACC teslim ed-ecek-leri yaz-ıyor-du. delivery do-NMLZ-POSS.3PL write-PROG-PST ‘I bought a book online today. It said (on the website) that they would deliver the book tomorrow.’ Like Mongolian, Turkish also displays a strong–weak contrast when it comes to case marking on quantifiers. Enc˛ (1991: 10-18) discusses her (= every) which triggers case marking obligatorily, in contrast to numerals, among them bir (= a/one), and other weak quantifiers. In (96), I extend the data set of strong quantifiers to tu¨m (= all) and ¸cog˘u (= most). (96) Strong quantifiers require case marking in Turkish a. Mehmet her hoca*(-yı) sev-iyor-∅. Mehmet.NOM every professor-ACC like-PROG-3SG ‘Mehmet likes every professor.’ b. Mehmet tu¨m hoca-lar*(-ı) sev-iyor-∅. Mehmet.NOM all professor-PL-ACC like-PROG-3SG ‘Mehmet likes all professors.’
Turkish
78
PSEUDO-INCORPORATION VS DIFFERENTIAL OBJECT MARKING c. Mehmet c¸og˘u hoca*(-yı) sev-iyor-∅. Mehmet.NOM most professor-ACC like-PROG-3SG ‘Mehmet likes most professors.’
In (97), I provide a list of weak determiners which allow for case drop. The status of bazı (= some) is under debate. Enc˛ (1991: 15) reports that case drop is not licensed for bazı, a data point which fits with the observation that bazı is not licensed in existential sentences, either. There is, however, dialectal variation, as Özyıldız (2017: 913–915) points out. Interestingly, speakers who accept case drop on noun phrases modified by bazı also allow bazı occuring in existential clauses. My consultants pattern with the latter group. (97) Optionally case marking for weak quantifiers, numerals a. Ali u¨c¸ kitap / kitab-ı oku-du-∅. Ali.NOM three book / book-ACC read-PFV-3SG ‘A li read three books.’
Turkish
b. Mehmet birc¸ok hoca(-yı) sev-iyor-∅. Mehmet.NOM many professor-ACC like-PROG-3SG ‘Mehmet likes many professors.’ c. Mehmet birkac¸ hoca(-yı) sev-iyor-∅. Mehmet.NOM some professor-ACC like-PROG-3SG ‘Mehmet likes some professors.’ d. Mehmet az hoca(-yı) sev-iyor-∅. Mehmet.NOM a.few professor-ACC like-PROG-3SG ‘Mehmet likes a few professors.’ e. Mehmet bazı hoca-lar% (-ı) sev-iyor-∅. Mehmet.NOM some professor-PL-ACC like-PROG-3SG ‘Mehmet likes some professors.’ Finally, indefinites and bare nouns can undergo case drop, shown in (98). (98) von Heusinger and Kornfilt (2005: 5) a. Ben kitap / kitab-ı oku-du-m. I book / book-ACC read-PST-1SG ‘I was book-reading. / I read the book.’ b. Ben bir kitap / bir kitab-ı oku-du-m. I a book / a book-ACC read-PST-1SG ‘I read a book.’
Turkish
4.3 CASE STUDIES
79
Case loss on bare nouns and indefinites triggers a low scope restriction in Turkish. The inability of wide scope readings for caseless bare nouns is exemplified with (99a). Indefinites, if case marked, can take scope over a universally quantified subject, cf. (99b) to (99c). These judgements are in line with earlier work by Kelepir (2001) and Aygen (2007a). The split between (99b) and (99c) was additionally verified by a recent quantative study (von Heusinger and Bamyacı 2016). Thus, Mongolian and Turkish pattern alike in that bare nouns as well as indefinites undergo PNI. (99) Öztürk (2005: 67-68)
Turkish
a. Her c¸ocuk kitap okudu. every child book read ‘Every child did book-reading.’
∀∃,*∃∀
b. Her c¸ocuk bir kitap okudu. every child one book read ‘Every child read a book.’
∀∃,*∃∀
c. Her c¸ocuk bir kitab-ı okudu. every child one book-ACC read ‘Every child read a book.’
∀∃,∃∀
The same observations can be made for subjects, shown here for bare nouns and negation. While a case-marked subject automatically receives a definite reading (100b), a caseless bare noun does not scope above negation (100a). (100) PNI of bare subjects in embedded clauses
Turkish
a. Hasta-lar-ı doktor go¨r-me-dig˘-i-ni duy-du-m. patient-PL-ACC doctor see-NEG-NMLZ-POSS.3SG-ACC hear-PFV-1SG ‘I have heard that no doctor saw the patients.’ ¬∃,*∃¬ b. Hasta-lar-ı patient-PL-ACC
doktor-un doctor-GEN
go¨r-me-dig˘-i-ni see-NEG-NMLZ-POSS.3SG-ACC
duy-du-m. hear-PFV-1SG ‘I have heard that the doctor didn’t see the patients.’
DEFINITE
We can diagnose the PNI status of weak quantifiers by examining their binding capabilities. In (101a), the possessor either co-refers with an individual salient in the context or it is bound by the quantifier in subject position. Once the subject loses its case marker in (101b), the bound reading is not possible any more.
80
PSEUDO-INCORPORATION VS DIFFERENTIAL OBJECT MARKING
(101) Caseless weak quantifiers do not act as binders in Turkish a. Sahib-i-ni1∕2 [birc¸ok ko¨peg˘-in]1 ısır-dıg˘-ı-nı owner-POSS.3SG-ACC many dog-GEN bite-NMLZ-POSS.3SG-ACC duy-du-m. hear-PFV-1SG ‘I heard that many dogs bit their/someone else’s owner(s).’ b. Sahib-i-ni*1∕2 [birc¸ok ko¨pek]1 ısır-dıg˘-ı-nı owner-POSS.3SG-ACC many dog bite-NMLZ-POSS.3SG-ACC duy-du-m. hear-PFV-1SG ‘I heard that many dogs bit someone else’s owner.’ The evidence given so far should suffice by now to preliminary conclude that bare nouns, indefinites, numerals, and weak quantifiers form a natural class that can undergo PNI. For a complete picture of the scope, binding, and control properties see Sections 5.4.3 and 6.3, respectively. Turkish case drop does not correlate with specific readings. Contra the traditional view from the literature (Enc˛ 1991; von Heusinger and Kornfilt 2005; Kelepir 2001), quantitative studies run by von Heusinger and Bamyacı (2016) and von Heusinger et al. (2019) were not able to detect a sensitivity to epistemic specificity based on the case marking of objects. The minimal pair in (102) provides one of the test items. (102) von Heusinger and Bamyacı (2016: 7)
Turkish
a. Mustafa bir sandalye-yi al-dı. ±SPEC Mustafa a chair-ACC buy-PST-3SG ‘Mustafa bought a chair.’ ... This is a very similar one to the rocking chair I bought last month. ... But I have not yet seen what type of chair this is. b. Mustafa bir sandalye al-dı. ±SPEC Mustafa a chair buy-PST-3SG ‘Mustafa bought a chair.’ ... This is a very similar one to the rocking chair I bought last month. ... But I have not yet seen what type of chair this is. Since Turkish behaves like Mongolian with respect to the class of noun types undergoing PNI as well as the overall lack of any DOM-effects, I will adopt a very similar analysis for post-syntactic case drop. The only necessary adjustment concerns the underlying nominal structures, an aspect which is independent of the PNI pattern and rather relates to the availability of a silent existential quantifier. As the reader might have noticed, case-marked bare nouns automatically shift
4.3 CASE STUDIES
81
to a definite reading. In parallel to what Dayal (2004) proposes for Hindi, I will assume that the silent existential shift is not available in Turkish. Consequently, case-marked indefinites result from numerals and weak quantifiers serving a double function: they can act as modifiers as well as argument forming determiners.²⁵ The structures are sketched in (103) and (104) where I make an explicit distinction between a modifier and a selecting determiner with the presence/absence of [•N•].²⁶ The semantic representations make reference to this distinction in (106). A selection of vocabulary entries is given in (105). Since exponents can be underspecified, determiners like bir and birc¸ok can be inserted in feature lists for non-selecting DP modifiers as well as selecting D heads. (103) Turkish determiners with case drop: PNI-DP PNI-DP NP DPNI [•N• ≻ D ≻ V]
DPNI [•N• ≻ D ≻ V]
PNI-DP NP
DPone NP [#1, D]
DPNI [•N• ≻ D ≻ V]
NP
DPmany NP [|x| > n, D]
(104) Case-marked Turkish indefinites:
DP NP Done [#1, •N• ≻ D]
DP Dmany NP [|x| > n, •N• ≻ D]
a (105) Vocabulary entries for (some) Turkish determiners: a. /bir/ ↔ [#1]
Done
b. /her/ ↔ [∀]
D∀
c. /birc¸ok/ ↔ [|x| > n] d. /∅/ ↔ [ ]
Dmany DPNI , wDι , sDι
(106) Determiner denotations for bir NP and bare nouns a. ⟦•N• ≻ D ≻ V⟧ = λP⟨e,t⟩ λQ⟨e,⟨v,t⟩⟩ λe∃z[P(z) ∧ Q(z)(e)]
DPNI
b. ⟦#1, D⟧ = λz.|z| = 1
Done
c. ⟦#1, •N• ≻ D⟧ = λP⟨e,t⟩ λQ⟨e,t⟩ ∃z[P(z) ∧ Q(z)]
Done
²⁵ See Kornfilt (1997); Özge (2011) for treating bir as an indefinite determiner, and Aygen (2007a); Sağ (2018) for an analysis as a numeral modifier. The need for both functions is also argued for in Öztürk (2005). ²⁶ The selecting feature [•N•] is always present on argument forming determiners, but is not made explicit in the previous sections.
82
PSEUDO-INCORPORATION VS DIFFERENTIAL OBJECT MARKING
I encode case suffixes in Turkish as shown in (107). Since the accusative marker and the genitive marker share the feature [+STRUCT], they will both be affected if the feature is deleted. This way the impoverishment rule in (108) will lead to insertion of the elsewhere marker for both accusative and genitive contexts, as they are defined in (44). Since the rule applies only in the context of [V ], it singles out PNI contexts.²⁷ (107) Vocabulary items: a. DAT, LOC, ABL ↔ [+GOV,+OBL,–STRUCT,...] b. /-(n)In/ ↔ [+OBL,+STRUCT]
(GEN)
c. /-(y)I/ ↔ [+STRUCT]
(ACC)
d. /-∅/ ↔ [ ]
(NOM)
(108) Impoverishment rule for Turkish: a [+STRUCT] → ∅ / [V ]
PNI
The constraint-based system is also very similar to what was proposed for Mongolian. Again, I collapse the categories INDEF SPEC and INDEF N-SPEC on the definiteness scale in (109) since case drop does not seem to interact with specificity. The markedness constraint *[+STRUCT] is ranked between case faithfulness constraints of indefinites and PNI-ed arguments (110), thereby deriving the size restriction. Notice also that it does not matter for the definiteness scale whether numerals and weak quantifiers can act as modifiers and D heads or whether they are restricted to a modifying function. (109) Definiteness scale in Turkish: PRO ≻ PN ≻ DEF ≻ SQ ≻ DEM ≻ INDEF ≻ PNI ⏟⏟⏟ ⏟⏟⏟ bir∕#∕wQ
bir∕#∕wQ
(110) Constraint-ranking in Turkish: ⎧ *PRO & MAX-C ⎪ *PN & MAX-C ⎪ *DEF & MAX-C ⎨ *SQ & MAX-C ⎪ ⎪ *DEM & MAX-C ⎩ *INDEF & MAX-C
⎫ ⎪ ⎪ ⎬ ⎪ ⎪ ⎭
≫
{
*[+STRUCT]
}
≫
{
*PNI & MAX-C
}
²⁷ With nothing else being said, the PNI-determiner is predicted not to apply in the nominal domain, e.g. as forming a possessor in a complex noun phrase. PNI-DP has to combine with a verbal projection, i.e. a function from individuals to predicates of events, which is presumably not available in the nominal domain.
4.3 CASE STUDIES
83
We have seen in this section how a PNI language can pseudo-incorporate objects as well as subjects. The key to making this happen is by letting the impoverishment rule or the markedness constraint apply to a natural class of arguments, which is effortlessly done by adjusting the case subfeature. In all other PNI/DOM aspects, Turkish behaves like Mongolian, in that a smaller class of noun types can lose case marking but no additional DOM effects are diagnosed.
4.3.4 Korean Korean is the second language in our data set which allows us to diagnose subject as well as object pseudo-incorporation since both argument types are overtly casemarked. We start our investigation with bare nouns this time and then work our way up the definiteness scale. Parts of the data set were first presented in Driemel and Lee (2022). An example for optional object case marking is given in (111), optional subject case marking is shown in (112). (111) Ha. Lee (2011: 20) Ecey Minswu-ka chinkwu(-lul) manna-ss-ta. yesterday Minsoo-NOM friend-ACC meet-PST-DECL ‘Minsoo met (his) friend yesterday.’ (112) Kwon and Zribi-Hertz (2008: 265) Beoseu(-ga) o-goiss-da. bus-NOM come-PROG-DECL ‘There’s a/the bus coming.’
Korean
Korean
In contrast to Turkish, Korean bare nouns can get definite as well as indefinite interpretations (Ch. Lee 1989,1992), Section 5.4.4 will provide homogeneity and discourse new contexts which support this claim. Korean is a classifier language (Sohn 1999; Ko 2007), i.e. numerals usually occur with an additional classifier that turns the denotation of the NP into countable units. Post-nominal numeral classifier constructions are optionally case marked, in subject and in object position, examples are given in (113).²⁸ (113) Post-nominal numeral + classifier a. Yusu-ka kkoch(-ul) twu-songi sa-ss-e. Yusu-NOM flower-ACC two-CL buy-PST-INT ‘Yusu bought two flowers.’
Korean
²⁸ Numeral classifiers themselves can be case marked, we exclude them from the discussion and assume with Ko (2014: 104) that they are adverbial and thus generated outside of vP.
84
PSEUDO-INCORPORATION VS DIFFERENTIAL OBJECT MARKING b. Pesu(-ka) twu-tay o-ney. bus- NOM two-CL come-FAM ‘Two buses are coming.’ c. Kay(-ka) han mali cic-nun-ta. dog-NOM one CL bark-PRS-DECL ‘One dog barks. ’
By setting-up comparable scope scenarios, I show that case loss on bare nouns leads to a low scope restriction. In contrast, case loss on numeral classifier constructions has no effect on the scopal properties.²⁹ The use of a caseless bare noun in (114) is infelicitous in a context that excludes a reading where the existential scopes under negation. The caseless numeral in (115), in contrast, is perfectly acceptable in a context which enforces a wide scope reading. Scope properties are discussed in detail in Section 5.4.4, where also parallel scope restriction for caseless subjects are reported. (114) Context ∃¬: Yusu’s friend has only a few flowers left to sell and he wants to sell everything by the end of the day. Yusu decides to buy some of them but not all. So there is at least one flower he did not buy. Kkoch# (-ul)1 Yusu-ka ⟨DP1 ⟩ sa-ci anh-ass-ta. flower-ACC Yusu-NOM buy-CI NEG-PST-DECL ‘Yusu did not buy a specific flower.’ (115) Context 1¬: Yusu’s friend wants to sell three flowers and Yusu bought two from him. So there is one flower Yusu did not buy. [Kkoch(-ul) han-songi]1 Yusu-ka ⟨DP1 ⟩ sa-ci anh-ass-ta. flower-ACC one-CL Yusu-NOM buy-CI NEG-PST-DECL ‘One flower, Yusu did not buy.’ We can conclude from the contrast between (114) and (115) that bare nouns can get pseudo-incorporated, whereas case drop on numeral classifiers signals a DOM effect. Let us now move on to noun types which need to be obligatorily case marked. Korean quantifiers do not allow for case drop, examples are given in (116)–(118).
²⁹ Scope readings are tested with scrambled structures since scrambling is a precondition for many speakers to create inverse readings. We will briefly discuss the influence of scope rigidity in PNI languages in Section 5.4.4.
4.3 CASE STUDIES
85
(116) Strong quantifier all requires case marking a. Yusu-ka motun kkoch?? (-ul) sa-ss-e. Yusu-NOM all flower-ACC buy-PST-INT ‘Yusu bought all the flowers.’
Korean
b. Motun pesu?? (-ka) o-ney. all bus-NOM come-FAM ‘all the buses are coming’ c. Motun haksayng? (-i) swuyengha-yss-e/wus-ess-e. all student-NOM swim-PST-INT/laugh-PST-INT ‘A ll the students swam/laughed.’ (117) Strong quantifier most requires case marking a. Yusu-ka taypwupwun-uy kkoch?? (-ul) sa-ss-e. Yusu-NOM most-GEN flower-ACC buy-PST-INT ‘Yusu bought most of the flowers.’
Korean
b. Taypwupwun-uy pesu?? (-ka) o-ney. most-GEN bus-NOM come-FAM ‘Most of the buses are coming’ c. Taypwupwun-uy haksayng?? (-i) swuyengha-yss-e/wus-ess-e. most-GEN student-NOM swim-PST-INT/laugh-PST-INT ‘Most of the students swam/laughed.’ (118) Weak quantifier a few requires case marking a. Yusu-ka myechmyech chayk?? (-ul) sa-ss-e. Yusu-NOM a.few book-ACC buy-PST-INT ‘Yusu bought a few books.’
Korean
b. Myechmyech pesu?? (-ka) o-ney. a.few bus-NOM come-FAM ‘A few buses are coming’ c. Myechmyech haksayng*? (-i) swuyengha-yss-e/wus-ess-e. a.few student-NOM swim-PST-INT/laugh-PST-INT ‘A few students swam/laughed.’ Possessor phrases similarly cannot be licensed without overt case marking; see (119). (119) Possessor phrases require case marking a. Yusu-ka Neytellantu-uy kkoch? (-ul) kacyewa-ss-e. Yusu-NOM Netherlands-GEN flower-ACC bring-PST-INT ‘Yusu brought Netherlands’ flowers.’
Korean
86
PSEUDO-INCORPORATION VS DIFFERENTIAL OBJECT MARKING b. Leipzig-uy pesu? (-ka) o-ney. Leipzig-GEN bus-ACC come-FAM ‘Leipzig’s bus is coming.’ c. Suzi-uy komo? (-ka) swuyengha-yss-e/wus-ess-e. Suzi-GEN aunt-NOM swim-PST-INT/laugh-PST-INT ‘Suzi’s aunt swam/laughed.’
Like Tamil and Hindi, Korean allows demonstratives phrases to be caseless. This observation was originally made by Kwon and Zribi-Hertz (2006) for objects, but can be extended to subjects. The data is provided in (120). (120) Demonstratives can be optionally case marked a. Yusu-ka i/ce kkoch(-ul) sa-ss-e. Yusu-NOM this/that flower-ACC buy-PST-INT. ‘Yusu bought this/that flower(s).’
Korean
b. I/ce pesu(-ka) pwusecy-ess-e. this/that bus-NOM be.broken-PST-INT ‘This/that bus is broken.’ c. I/ce sonye(-ka) swuyengha-yss-e. this/that girl-NOM swim-PST-INT ‘This/that girl swam.’ What makes Korean stand out from the other languages in the data set is that proper names and pronouns are also optionally case marked (Ha. Lee 2006, 2008; Kwon and Zribi-Hertz 2008). (121) 1st , 2nd person pronouns are optionally case marked a. Yusu-ka na(-lul)/ne(-lul) manna-ss-e. Yusu-NOM I-ACC/you-ACC meet-PST-INT ‘Yusu met me/you.’ b. Na(-ka)/ne(-ka) tochakhay-ss-e. I-NOM/you-NOM arrive-PST-INT ‘I/you arrived.’ c. Na(-ka)/ne(-ka) wus-ess-e. I-NOM/you-NOM laugh-PST-INT ‘I/you laughed.’
Korean
4.3 CASE STUDIES
87
(122) Proper names are optionally case marked a. Yusu-ka nwutheylla(-ul) sa-ss-e. Yusu-NOM Nutella-ACC buy-PST-INT ‘Yusu bought Nutella.’
Korean
b. Daniel(-i) salacy-ess-e. Daniel-NOM disappeare-PST-INT ‘Daniel disappeared.’ c. Suzi(-ka) swuyengha-yss-e/wus-ess-e. Suzi-NOM swim-PST-INT/laugh-PST-INT ‘Suzi swam/laughed.’ The literature only discusses local pronouns, i.e. 1st and 2nd person. Strikingly, 3rd person pronouns behave differently from local pronouns, in that they require case marking, shown in (123). (123) 3rd person pronouns require case marking a. Yusu-ka ku? (-lul)/kunye? (-lul) manna-ss-e. Yusu-NOM he-ACC/she-ACC meet-PST-INT ‘Yusu met her/him.’
Korean
b. Ku*? (-ka)/Kunye*? (-ka) tochakhay-ss-e. he-NOM/she-NOM arrive-PST-INT ‘She/he arrived.’ c. Ku?? (-ka)/Kunye?? (-ka) wus-ess-e. he-NOM/she-NOM laugh-PST-INT ‘She/he laughed.’ Since pronouns, proper names, and demonstratives intrinsically outscope other operators, we turn to binding properties to diagnose their ability to undergo PNI. Caseless bare nouns serve as a baseline here since we have already seen that they additionally show a scope restriction. As predicted, bare nouns cannot bind a reflexive pronoun, once the case is dropped (124a). Crucially, proper names and 1st , 2nd person pronouns pattern with bare nouns, which leads us to conclude that they must be able to pseudo-incorporate as well. Demonstratives, shown in (124d), can still act as a binder, even without case marking. Hence, their case loss seems to be caused by DOM.
88
PSEUDO-INCORPORATION VS DIFFERENTIAL OBJECT MARKING
(124) Binding blocked for caseless bare nouns, 1st /2nd pronouns, and proper names a. Koyangi*(-ka)1 [ku casin-ul]1 halth-ass-e. cat-NOM 3rd self-ACC lick-PST-INT ‘A cat washed itself.’
Korean
b. Nay? *(-ka)1 /Ney? *(-ka)1 [nay/ney casin-ul]1 chingchanhay-ss-e. 1SG-NOM/2SG-NOM 1SG/2SG self-ACC praise-PST-INT ‘I/you praised myself/yourself.’ c. Cinwu? *(-ka)1 [ku casin-ul]1 chingchanhay-ss-e. Cinwu-NOM 3rd self-ACC praise-PST-INT ‘Cinwu praised himself.’ d. I/ce koyangi(-ka)1 [ku casin-ul]1 halth-ass-e. DEM cat-NOM 3rd self-ACC lick-PST-INT ‘This/that cat washed itself.’ Finally, let us address definite noun phrases. Like in Mongolian, there is a demonstrative determiner which seems to develop into a definite determiner. In addition to the proximal demonstrative i and the distal demonstrative ce, Sohn (1999: 210) ascribes demonstrative ku the function to refer to individuals close to the hearer as well as ‘known to both speaker and hearer’. Ch. Lee (1989, 1992) observes that bare nouns function as weak definites but need to be case marked, whereas ku acts as an anaphoric strong determiner. Although they do not consider case drop, both Kang (2015, 2018) and Ahn (2017, 2018) agree with Ch. Lee (1989, 1992) that bare nouns license weak definite contexts. There is disagreement, however, on anaphoric contexts. Whereas Kang (2015, 2018) reports that ku is optional in anaphoric context, Ahn (2017, 2018) claims ku is obligatory. We agree with Kang (2015, 2018) in that ku is optional in strong anaphoric contexts. Case marking, however, is obligatory.³⁰ The contexts in (125) and (126) introduce discourse referents which are anaphorically referred to within the continuations. (125b) and (125d) show that strong definites require case marking in object position, independent of ku. (126b) and (126d) provide the parallel picture in subject position. (125) Context (strong definite, object): Ecey na-nun yeca-lul manna-ss-e... Yesterday I-TOP woman-ACC meet-PST-INT ‘I met a woman yesterday...’
Korean
³⁰ Note also that (Ahn 2019: 17) diverges from her earlier views in providing anaphoric contexts where bare nouns without ku are licensed after all.
4.3 CASE STUDIES
a.
89
... Na-nun ku yeca-lul kuly-ess-e. I-TOP DET woman-ACC paint-PST-INT ‘... I painted the woman.’
b. ? #... Na-nun ku yecal kuly-ess-e. I-TOP DET woman paint-PST-INT ‘... I painted the woman.’ c.
... Na-nun yeca-lul kuly-ess-e. I-TOP woman-ACC paint-PST-INT ‘... I painted the woman.’
d. ? #... Na-nun yeca kuly-ess-e. I-TOP woman paint-PST-INT ‘... I painted the woman.’ (126) Context (strong definite, subject): Ecey na-nun yeca-lul mannas-s-e... Yesterday I-TOP woman-ACC meet-PST-INT ‘I met a woman yesterday...’ a.
Korean
... Ku yeca-ka chwumchu-ess-e. DET woman-NOM dance-PST-INT ‘... The woman danced.’
b. ? #... Ku yeca chwumchu-ess-e. DET woman dance-PST-INT ‘... The woman danced.’ c.
... Yeca-ka chwumchu-ess-e. woman-NOM dance-PST-INT ‘... The woman danced.’
d. ? #... Yeca chwumchu-ess-e. woman dance-PST-INT ‘... The woman danced.’ Weak definite contexts exhibit a more intricate picture. While they show a general tendency to be case marked, the demonstrative ku seems to be able to repair the licensing of a caseless noun phrase in a weak definite context. Note also that our speakers generally find ku felicitous in weak definite contexts, contrasting with the intuitions in Ahn (2017) and Kang (2015). We are not the first ones to point this out. Cho (2017: 372) provides examples where the use of ku is preferred over the use of bare nouns in contexts without an explicit linguistic antecedent but where a referent is implicitly given.
90
PSEUDO-INCORPORATION VS DIFFERENTIAL OBJECT MARKING
The paradigms for global definiteness scenarios are given in (127) and (128). Like in strong contexts, objects and subjects have to be case marked, proven by the unacceptability of (127d) and (128d). The addition of ku in (127b) and (128b), however, seems to facilitate a weak definite reading for a caseless argument. (127) Weak global definiteness, object: a.
Na-nun ku yewang-ul eceyspam manna-ss-e. I-TOP DEM queen-ACC last.night meet-PST-INT ‘I met the queen last night.’
Korean
b. ? Na-nun ku yewang eceyspam manna-ss-e. I-TOP DEM queen last.night meet-PST-INT ‘I met the queen last night.’ c.
Na-nun yewang-ul eceyspam manna-ss-e. I-TOP queen-ACC last.night meet-PST-INT ‘I met the queen last night.’
d. ?* Na-nun yewang eceyspam manna-ss-e. I-TOP queen last.night meet-PST-INT ‘I met the queen last night.’ (128) Weak global definiteness, subject: a.
Ku yewang-i eceyspam ttena-ss-e. DEM queen-NOM last.night leave-PST-INT ‘The queen left last night.’
Korean
b. ? Ku yewang eceyspam ttena-ss-e. DEM queen last.night leave-PST-INT ‘The queen left last night.’ c.
Yewang-i eceyspam ttena-ss-e. queen-NOM last.night leave-PST-INT ‘The queen left last night.’
d. ?*Yewang eceyspam ttena-ss-e. queen last.night leave-PST-INT ‘The queen left last night.’ The weak definite pattern was confirmed with an online study. The study tested, among other things, immediate weak definite contexts such as the one in (129), where the situation is narrowed sufficiently so that the individual can be identified as unique by both speaker and hearer. As (129b) and (129d) show, caseless objects are not licensed, while adding ku seems to make the weak definite reading more readily available.
4.3 CASE STUDIES
91
(129) Context (immediate weak definite): Suzi and Jamin are close friends. They are supposed to hang-out together by going to Karaoke. Everytime they go to the Karaoke place they usually run into a small gray stray cat. On the way to the Karaoke place, they meet the same gray cat approaching them. Suzi says to Jamin, ... a.
... Jamin-a ku koyangi-ka o-ney! Jamin-VOC DEM cat-NOM come-FAM ‘... Jamin, the cat is coming!’
Korean
b. ?... Jamin-a ku koyangi o-ney! Jamin-VOC DEM cat come-FAM ‘... Jamin, the cat is coming!’ c.
... Jamin-a koyangi-ka o-ney! Jamin-VOC cat-NOM come-FAM ‘... Jamin, the cat is coming!’
d. # ... Jamin-a koyangi o-ney! Jamin-VOC cat come-FAM ‘... Jamin, the cat is coming!’ Similar to the previous sections, we test for the PNI status of caseless weak definites by inspecting binding configurations. Case-marked indefinites can bind reflexive pronouns, as (130a) and (130c) show. (130d) is unacceptable independent of the need to bind the reflexive. The crucial sentence is presented in (130b), where even with the addition of ku caseless weak definites do not seem to be able to bind a reflexive pronoun. Weak definites thus pattern with bare nouns, proper names, and 1st /2nd pronouns, in that case drop signals pseudo-incorporation. (130) Binding blocked for caseless weak definites (global uniqueness) a.
[Ku yewang-i]1 [ku casin-ul]1 tachikeyhay-ss-e DEM queen-NOM 3rd self-ACC hurt-PST-INT ‘The queen hurt herself.’
b. ?? [Ku yewang]1 [ku casin-ul]1 tachikeyhay-ss-e DEM queen 3rd self-ACC hurt-PST-INT ‘The queen hurt herself.’ c.
Yewang-i1 [ku casin-ul]1 tachikeyhay-ss-e queen-NOM 3rd self-ACC hurt-PST-INT ‘The queen hurt herself.’
d. *Yewang1 [ku casin-ul]1 tachikeyhay-ss-e queen 3rd self-ACC hurt-PST-INT ‘The queen hurt herself.’
Korean
92
PSEUDO-INCORPORATION VS DIFFERENTIAL OBJECT MARKING
In (131), I provide a short summary of Korean noun types and their capability as well as motivation to lose case marking. (131) Interim summary of case marking properties in Korean a. Noun types obligatorily case marked possessor phrases, strong definites, strong quantifiers, weak quantifiers, 3rd person pronouns b. Noun types optionally case marked due to DOM numeral classifiers, demonstratives c. Noun types optionally case marked due to PNI bare nouns/indefinites, weak definites (with ku), 1st /2nd pronouns, proper names Similar to the PNI languages in the previous sections, I take the argument types in (131a) to contain proper D heads which are in complementary distribution with the PNI determiner, thereby triggering the requirement to be case marked. Like in Tamil and Hindi, Korean demonstratives also require a DP-shell but can optionally lose case, as they can be targeted by impoverishment rules/constraints. I apply the same reasoning to postnominal numeral classifiers. In order to do this, we will take a short detour now and provide a series of arguments why (postnominal) classifiers should be considered DPs. While there are many accounts who consider classifiers to be functional heads (Borer 2005c; Cheng and Sybesma 1999; Cheng et al. 2017; Simpson 2008; Syed and Simpson 2017), Ko (2014) and Bruening et al. (2018) argue for Korean classifiers specifically to constitute adjuncts. I adopt the functional head analysis, based on the arguments provided by Kim and Melchin (2018a,b). In (132), we see the the two analyses juxtaposed for the numeral construction in (113b). (132) Korean post-nominal classifiers Ko (2014: 91): nP
Kim and Melchin (2018b: 12): ClP
nP
ClP
nP
Cl
pesu
#P Cl twu tay
pesu
twu-tay
Kim and Melchin (2018a: 23) argue that the classifier’s obligatory occurence in counting contexts is an argument against an adjunct analysis. This property, however, presumably results from the fact that classifiers are always selected for by numerals (Bruening et al. 2018: 2, fn.3), which would in turn still enable the numeral-classifier unit to adjoin to NP. Classifiers are, furthermore, necessary for
4.3 CASE STUDIES
93
semantic composition, since predicates in classifier languages have been argued to have a mass denotation, requiring classifiers to turn them into countable units before numerals can apply (Borer 2005c; Chierchia 1998). A more convincing argument for the head status of classifiers comes from apparent selectional restrictions (see (133)), which suggest an analysis in which classifier and NP seem to undergo a feature checking relation. (133) Kim and Melchin (2018a: 24) a. sensayngnim twu pwun/? myeng teacher two CL.HON/ CL.person ‘two teachers’
Korean
b. salam ney myeng/*mari person four CL.person/ CL.animal ‘four people’ c. horangi ney mari/*myeng tiger four CL.animal/ CL.person ‘four lions’ I would like to put forward a few other arguments in favour of the functional head status. Saito et al. (2008) compare Japanese classifiers with Chinese classifiers with respect to displacement and NP-ellipsis properties. They settle on the structures in (134). Japanese numeral classifiers do not license NP-ellipsis but can be displaced by fronting. Chinese numeral classifiers, on the other hand, license NP-ellipsis but cannot undergo fronting. The ban on fronting in Chinese follows from the fact the numeral and classifier do not form a constituent, in contrast to the underlining structure for Japanese. The ban on NP-ellipsis in Japanese follows from an assumption Saito et al. (2008) make about NP-ellipsis licensing. Lobeck (1995) shows how NP-ellipsis is only licit if the licencer undergoes strong Agree with the ellipsis site. In non-agreeing languages such as Chinese and Japanese, this requirement amounts to the licenser projecting a specifier, Saito et al. (2008) argue. This is the case for the structure in Chinese where the classifier second-merges with the numeral phrase, but not in Japanese where neither CL nor D project a specifier. (134) Chinese:
Japanese:
(Saito et al. 2008: 262)
DP
DP D
NP
CIP #P
Cl′ Cl NP
NP
ClP #P CL
D
94
PSEUDO-INCORPORATION VS DIFFERENTIAL OBJECT MARKING
Korean patterns with Chinese, and not with Japanese with respect to displacement and NP-ellipsis; see (135) and (136). (135) Kang (2002: 388) ??Sey-cang1 , ai-ka [cepsi-lul ⟨sey-cang1 ⟩] kkayttuliessta. 3-CL child-NOM dish-ACC broke ‘The child broke three dishes.’
Korean
(136) NP-ellipsis in Korean John-un [chayk-ul sey kwen] sass-ko, Mary-nun [chayk-ul John-TOP book-ACC three CL bought-and Mary-TOP book-ACC twu kwen] sassta. two CL bought ‘John bought three books, and Mary bought two.’ The displacement test, however, is not entirely conclusive since numeral classifier can front if they are accompanied by focus adverbs, which is shown in (137). One should also note that if Korean allows for DP remnant movement, displacement properties might not be telling to begin with. (137) Kang (2002: 388) Sey-cang-ina/-man ai-ka cepsi-lul kkayttuliessta. 3-CL-even/-only child-NOM dish-ACC broke ‘The child broke even/only three dishes.’
Korean
Since extraction of classifiers leaves us with inconclusive results, we turn to the possibility of stranding post-nominal classifiers. While an adjunction analysis might predict a classifier adjunct to be movable, moving the adjunction site should be banned, if nothing else is said. A functional head approach, on the other hand, readily allows for extraction out of complement position. The possibility to strand numeral classifiers is shown in (138b). (138) Ko (2007: 50) a. John-i maykcwu-lul sey-pyeng masi-ess-ta. John-NOM beer-ACC 3-CL drink-PST-DECL ‘John drank three bottles of beer.’ b. Maykcwu-lul1 beer-ACC
John-i John-NOM
[⟨maykcwu-lul1 ⟩
masi-ess-ta. drink-PST-DECL ‘John drank three bottles of beer.’
sey-pyeng] 3-CL
Korean
4.3 CASE STUDIES
95
We thus have reason to believe that classifiers constitute functional heads. Furthermore, classifiers are D heads, similar to what Cheng and Sybesma (1999) claim to be the case in Mandarin and Cantonese, as well as Kang (2002) in Korean.³¹ In order to derive NP-ellipsis, we propose the DP-structure on the left in (139), where we follow Merchant (2001) in that ellipsis is triggered by an [E]-feature which is situated on a head and marks the terminal nodes of its complement to not be overtly spelled out. Since NP-ellipsis is only licensed in Korean if there is an additional specifier present, [E] also comes with an EPP feature. An adjunct analysis, as it is shown on the right in (139), would not be able to license NP-ellipsis since Cl does not take NP as a complement, nor does it project a specifier. Placing the [E]-feature with a higher D head would quite likely predict the classifier to be elided with the NP—given that ellipsis is licensed at all since D would have to project a specifier. (139) Korean NP-ellipsis DP-analysis: DP
NP
D′
NP1
D′
#P
Adjunct analysis: NP
Dcl ⟨NP1⟩ [EuCl, EPP] E
ClP #P
Cl *[EuCl, EPP] E
As the DP-structure in (139) already reveals, I assume that NP moves within DP; see also Ahn (2018) and Syed and Simpson (2017) for this kind of approach to post-nominal classifiers.³² The other important analytical step concerns the determiner nature of the classifier. Kang (2002) observes that post-nominal classifiers ³¹ See Gebhardt (2011) and Wu and Bodomo (2009) for a discussion of classifiers as determiners in Chinese. ³² I assume that the nominal domain is head-initial in Korean, based on the observation that demonstratives and quantifiers occur to the left of the noun. One could, of course, also imagine a head-final nominal domain, in line with Korean’s SOV property. Demonstratives and quantifiers would have to be merged in leftward specifiers. It is important to note, however, that even in a head-final system postnominal classifiers require DP-internal movement to ensure semantic composition as well as linear order. (i) Post-nominal classifiers in head-final structures: DP
ClP NP pesu
D#
D# Cl ⟨tay⟩ twu
Cl tay
96
PSEUDO-INCORPORATION VS DIFFERENTIAL OBJECT MARKING
can only create indefinite readings. He compares the constructions to pre-nominal genitive marked classifiers who allow for definite as well as indefinite readings. One piece of evidence comes from strict and sloppy readings in VP-ellipsis. Strict readings are only possible if the numeral classifier phrase refers to a unique referent in the context, which in turn is only possible with a definite determiner. While post-nominal classifiers in (140a) can only trigger sloppy readings, pre-nominal classifiers in (140b) allow for strict and sloppy readings.
(140) Kang (2002: 381)
Korean
a. Cengswu-ka [haksayng-ul sey-myeng] manna-ss-ta. Cengswu-NOM student-ACC 3-CL meet-PST-DECL Myenghuy-to kulay-ss-ta. [sloppy] Myenghuy-also do-PST-DECL ‘Cengswu met three students. Myenghuy did so too.’ b. Cengswu-ka [sey-myeng-uy haksayng-ul] manna-ss-ta. Cengswu-NOM 3-CL-GEN student-ACC meet-PST-DECL Myenghuy-to kulay-ss-ta. [strict/sloppy] Myenghuy-also do-PST-DECL ‘Cengswu met three students. Myenghuy did so too.’
The other piece of evidence comes from an incompatibility with topic marker -nun in post-nominal but not in pre-nominal classifier constructions; see (141). Since topics are commonly assumed to be definite, only pre-nominal classifiers allow for co-occurrence with -nun.
(141) Kang (2002: 382)
Korean
a. *Sonyen-un twu-myeng tochakhay-ss-ta. boy-TOP 2-CL arrive-PST-DECL ‘The two boys arrived.’ b.
Twu-myeng-uy sonyen-un tampay-lul phiko-issessta. 2-CL-GEN boy-TOP cigarette-ACC was-smoking ‘The two students were smoking.’
Since post-nominal classifiers determine the definiteness value of the argument, I analyse post-nominal classifiers as instantiating a [−def ] D head. They select an NP and a numeral phrase which will eventually create indefinite readings, more details are given in Section 5.4.4. I do not provide an analysis for pre-nominal
4.3 CASE STUDIES .
97
classifiers here, but a treatment akin to possessor phrases seems suitable, as they also require case-marked NPs obligatorily.³³ This concludes the discussion of numeral classifiers in Korean. In (142), I summarize and provide structures for the noun types which can lose case marking, ignoring for the moment weak definites, proper names and pronouns. In line with the previous sections, I provide the denotation of the numeral han (= one) and compare it to the PNI determiner and the existential quantifier in (143). The denotation of the numeral is what derives the obligatory indefinite reading of postnominal classifiers. I will come back to the denotation of the classifiers in Section 5.4.4. In (144), I show a selected list of determiner vocabulary items, like we saw in the previous sections. Since Korean provides ample evidence for the existence of a silent existential determiner, it is listed in (144), together with the definite determiners and the PNI determiner. I should also note that this book departs from Ahn (2017, 2018) in assigning the strong determiner a zero spell-out and ³³ There is yet another numeral configuration which triggers definite partitive readings and has been analysed as an underlying partitive construction by Ahn (2018). Interestingly, case drop is blocked in this construction. (i) Pre-nominal numerals require case marking Yusu-ka twu kkoch?? (-ul) sa-ss-e. Yusu-NOM two flower-ACC buy-PST-INT ‘Yusu bought two of the flowers.’
Korean
Ahn (2018) assume that the definite partitive reading comes about by a partitive head taking the numeral classifier phrase as a complement and a definite noun phrase with the identical NP as a specifier. The DP is specifier position is deleted under identity. By assumption, the classifier is spelledout as zero, triggering head-movement of the numeral to the Part head since numerals need a host to their right. Ahn (2018: 31,35)
(ii)
a Yeca sey myeng-i wassta. Twu yeca-nun anc-ass-ta. woman three CL-NOM came two woman-TOP sit-PST-DECL ‘Three women came. Two of the women sat down.’ b
PartP DP yeca
Part ′ Part Num n twu
Part
NumP NPi yeca
Num ′ tn
ClP Cl ti Ø
If this is the right analysis, we can accommodate the case requirement for pre-nominal numerals by letting PNI-D not select for PartP, while proper D heads are able to do so. They would additionally have to be considered high on a prominence scale since partitives cannot be targeted by DOM-related impoverishment either.
98
PSEUDO-INCORPORATION VS DIFFERENTIAL OBJECT MARKING
not the exponent ku since the data present optional occurrence of ku in anaphoric contexts and uniqueness contexts. (142) Korean determiners with case drop a. DOM-caused case drop: DP NP Ddem [DEM, D]
DP NP1
D′ DPone [#1, D]
D′ Dcl ⟨NP1⟩ [CL, D]
b. PNI-caused case drop:
PNI-DP DPNI NP [D ≻ V]
(143) Determiner denotations for numerals and bare nouns a. ⟦D ≻ V⟧ = λP⟨e,t⟩ λQ⟨e,⟨v,t⟩⟩ λe∃z[P(z) ∧ Q(z)(e)]
DPNI
b. ⟦#1, D⟧ = λP⟨e,t⟩ λQ⟨e,t⟩ ∃z[P(z) ∧ |z| = 1 ∧ Q(z)]
Done
c. ⟦∃, D⟧ = λP⟨e,t⟩ λQ⟨e,t⟩ ∃z[P(z) ∧ Q(z)]
D∃
(144) Vocabulary entries for (some) Korean determiners: a. /han/ ↔ [#1]
Done
b. /motun/ ↔ [∀]
D∀
c. /songi/ ↔ [CL]
Dcl
d. /ce/ ↔ [DEM] e. /∅/ ↔ [ ]
Ddem DPNI , D∃ , wDι , sDι
Let us now turn to the class of argument types which can undergo PNI somewhat unexpectedly from a typological perspective, i.e. weak definites and proper names as well as 1st and 2nd person pronouns. In the languages we have discussed so far, these argument types were prevented from pseudo-incorporating since they constitute DPs and denote semantic objects of type ⟨e⟩. In order to resolve this
4.3 CASE STUDIES
99
issue, I propose that Korean can make use of ident, a type-shift operator which maps elements onto their singleton sets (Partee 1986a,b), thus enabling the PNI determiner to apply and trigger case loss as well as binding/control restrictions.³⁴ This type-shifter is restricted in its application, as it does not apply to any semantic object of type ⟨e⟩ per se—recall that 3rd person pronouns and strong definite phrases crucially have to be case marked. Rather, ident applies to e-type entities for which speaker and hearer can universally agree on a referent. This is the one property which is decisive in forming the natural class of weak definites, proper names, and 1st /2nd person pronouns. For each of those argument types reference is established between interlocutors, prior to any immediate linguistic context. In contrast, 3rd person pronouns and strong definite phrases are anaphoric, thus they can only be interpreted with the help of previous context, thereby blocking identtype shift. We implement this restriction as a presupposition³⁵ on ident; see (145a) for the complete denotation. The type shifter can be spelled out optionally as ku, explaining its occurrence in the weak definite paradigm in (129). It is a functional head which selects for a DP; see the feature bundle in (145b). (145) The ident-type shifter a. ⟦ident⟧ = λx⟨e⟩ λy⟨e⟩ [x = y], defined iff speaker and hearer can universally agree on x’s referent b. [ident, •D• ≻ N] The syntactic structures for PNI contexts are provided in (146). The semantic objects formed with ident are, of course, also compatible with the e-type creating D heads but if they apply, the interpretation will not differ from structures without ident. (146) PNI-caused case drop with ident a. Proper names: PNI-DP
DPNI [D ≻ V]
NP⟨e,t⟩ N DP⟨e⟩ [ident, •D• ≻ N]
³⁴ It should be noted that an ident-type shifter is also made use of by Schwarz (2014) to derive a subtype of weak definites, i.e. those that have lost their uniqueness requirement (Bill is in the hospital) as an instance of pseudo-incorporation. In contrast to the Germanic examples Schwarz discusses, Korean weak definites crucially retain their uniqueness requirement. ³⁵ This restriction is reminiscent of the dagger operator discussed in Stalnaker (1978). Stalnaker introduces propositions as propositional concepts, which are not only evaluated in different believe worlds but also in a second dimension where worlds represent utterance contexts. The dagger is a twodimensional modal operator which yields a constant propositional concept, i.e. a proposition which receives the same truth value independent of utterance context.
100
PSEUDO-INCORPORATION VS DIFFERENTIAL OBJECT MARKING b. 1st /2nd person pronouns: PNI-DP
DPNI [D ≻ V]
NP⟨e,t⟩ N DP⟨e⟩ [ident, •D• ≻ N] [+part, D]
c. Weak definites: PNI-DP
DPNI [D ≻ V]
NP⟨e,t⟩ N [ident, •D• ≻ N]
DP⟨e⟩ wDɩ NP [w-ɩ, D]
Supporting evidence for the ident-type shifter comes from the optional cooccurrence of ku with proper names, given in (147). Im (2007) conducts a corpus study on the distribution of adnominal modifiers in ‘The 21st century Sejong Corpora’. Out of 64085 uses of ku, 546 co-occur with a proper name. An example is given in (147). transcribed from Hangul (147) Im (2007: 386) ... akka malhay-ss-t-en ku Uycawang-i natha-nan a.little.while.ago tell-PST-PERF-REL DEM Uycawang-NOM show.up-REL kes-i-ta. NMLZ-COP-DECL ‘Uycawang, who I told you a while ago, showed up.’ Semantic compositions are given in (148), representative for the PNI contexts in (121), (122) and (127)–(129). I leave it open how the proper name receives its e-type denotation, some suggestions were given in footnote 11. Free pronouns denote indices. Following Sauerland (2003, 2008b) and Heim (2008), participant ϕ-features on pronouns denote partial indentity functions of type ⟨e, e⟩. For free pronouns the relevant assignment is given by the utterance context.³⁶ ³⁶ A translation of person features into semantic denotations can be given along the following lines: c
(i) a. ⟦−hearer,+speaker⟧ = λxe : x includes sc [x] c
b. ⟦−speaker,+hearer⟧ = λxe : x includes hc [x] c. ⟦−speaker,−hearer⟧ = λxe [x]
4.3 CASE STUDIES
101
(148) PNI-caused case drop with ident: denotations a. Proper names:
PNI-DP ⟨⟨e,⟨v,t⟩⟩, ⟨v,t⟩⟩ λQλe z[suzi = z Q(z)(e)]
PNI-DP ⟨⟨e,t⟩,⟨⟨e,⟨v,t⟩⟩, ⟨v,t⟩⟩⟩ λPλQλe z[P(z) Q(z)(e)]
NP⟨e,t⟩ λy[suzi = y] N DP⟨e⟩ ⟨e,⟨e,t⟩⟩ Suzi λxλy[x = y]
b. 1st /2nd pronouns: PNI-DP ⟨⟨e,⟨v,t⟩⟩, ⟨v,t⟩⟩ λQλe : x includes sc . z[x = z Q(z)(e)] PNI-D ⟨⟨e,t⟩,⟨⟨e,⟨v,t⟩⟩, ⟨v,t⟩⟩⟩ λPλQλe z[P(z) Q(z)(e)]
NP⟨e,t⟩ λy : x includes sc[x = y] N DP⟨e⟩ ⟨e,⟨e,t⟩⟩ x includes sc[x] λxλy[x = y]
c. Weak definites: PNI-DP ⟨⟨e,⟨v,t⟩⟩, ⟨v,t⟩⟩ λQλe : !x[QUEEN(x) = z]. z[ιx[QUEEN(x) = z] Q(z)(e)] PNI-D ⟨⟨e,t⟩,⟨⟨e,⟨v,t⟩⟩, ⟨v,t⟩⟩⟩ λPλQλe z[P(z) Q(z)(e)]
NP⟨e,t⟩ λy : !x[QUEEN(x) = y]. ιx[QUEEN(x) = y] N ⟨e,⟨e,t⟩⟩ λxλy[x = y]
DP⟨e⟩ !x[QUEEN(x)]. ιx[QUEEN(x)] NP⟨e,t⟩ D λy[QUEEN(y)] ⟨⟨e,t⟩,e⟩ λP: !x[P(x)]. ιx[P(x)]
We are now left with the question why ku not only occurs with weak definites and proper names, but crucially also in strong definite contexts. There is no need for the ident-type shifter to apply, at least not in the original proposal by Schwarz
102
PSEUDO-INCORPORATION VS DIFFERENTIAL OBJECT MARKING
(2009), where anaphoricity was achieved by merging a pronoun in the specifier of sDι . Interestingly, there is a recent modification of Schwarz’ strong DP denotation proposed by Hanink (2018), based on an idea by Matushansky (2010). Instead of letting the D head introduce the index, Hanink proposes that the index projects its own phrase, interspersed between DP and NP. The index is shifted via ident to its singleton set and combines with the NP via predicate modification. Strong and weak D heads end up having the same denotation in this system; the different underlining structures are given in (149). (149) Strong definite noun phrases Schwarz (2009): DP
Hanink (2018): DP
D′
g(i) sDι
Dι NP
⟨⟨e, t⟩, e⟩
iP g(i) NP
⟨⟨e, t⟩, ⟨e, e⟩⟩
ident sDι = λP⟨e,t⟩λy : !x[P(x) x = y]. ιx[P(x) x = y]
Dι = λP⟨e,t⟩λy : !x[P(x)].ιx[P(x)] g(i)
ident
λy[g(i) = y]
I adopt a modified version of the iP-shell approach for Korean, where ident is not a covert type-shifter, but syntactically present and optionally spelled-out by ku. According to the analyses of Matushansky (2010) and Hanink (2018), ku is in this sense semantically identical to English same, as for example used in (150). (150) Abby bought Neverwhere. Beth bought the same book. Note that we cannot make use of ident as it is defined in (145). Nident does not select for a DP, nor is it restricted to e-type objects which need to be universally identified by speaker and hearer. Instead, this version of ident is simply used to shift indices to a property denotation, i.e. objects which require a linguistic context to be interpreted. Thus, Korean seems to exhibit an additional simpler version of ident which is used in strong definite contexts, compare (151) to (145). (151) Another ident-type shifter a. ⟦ident⟧ = λx⟨e⟩ λy⟨e⟩ [x = y] b. [ident, N] Both versions of ident, however, can be spelled out as ku.³⁷ ³⁷ It is worth pointing out again that the addition of (151) does not overgenerate. In particular, one might argue that (151) is able to apply to 3rd person pronouns, judging from its semantics in (151a)
4.3 CASE STUDIES
103
(152) Vocabulary entry for ident-type shifters /ku/ ↔ [ident] I now provide the syntactic structure and the semantic denotation for definite noun phrases in Korean, based on the DP in (125)/(126). In (153a), the syntactic structure draws a distinction between minimal and maximal projections with diacritics (see also Müller 2017). The index selects for the type-shifter ident, in order to subsequently undergo Predicate Modification with NP; see (153b). (153) Strong definite noun phrase in Korean a. Syntactic structure: DP
iP⟨e,t⟩
Dι [ι, •i• ≻ D]
NP⟨e,t⟩
i
N g(i) [ident, N] [•N0•≻•N2• ≻ i]
b. Semantic composition: DP⟨e⟩ !x[g(3) = y WOM(x)]. ιx[g(3) = y WOM(x)]
D
⟨⟨e, t⟩, e⟩
λP : !x[P(x)]. ιx[P(x)]
iP⟨e,t⟩ λy[g(3) = y WOM(y)] i⟨e,t⟩ NP⟨e,t⟩ λy[g(3) = y] λy[WOM(y)] N
⟨e, ⟨e,t⟩⟩
g(3)
λxλy[x = y] The index selects for ident and not vice versa since it is crucial that i projects its own phrase, which serves as the complement to D. This analysis addresses a concern from Schwarz (2019) who points out that an analysis like the one proposed by Hanink (2018) makes it difficult to distinguish between weak and strong D heads and its specific application to indices within strong definites. This scenario, although possible from a semantic perspective, is excluded for syntactic reasons. The feature set-up in (151b) ensures that this type-shifter cannot take DPs as arguments, including free standing pronouns, as it does not select for D.
104
PSEUDO-INCORPORATION VS DIFFERENTIAL OBJECT MARKING
since they receive the same denotation, the only difference being the complexity of the DP, where strong DPs come with an additional iP layer which is missing in weak DPs. Although there is no problem in Korean since both D heads receive a zero spell-out, we have already seen cross-linguistic evidence that requires a distinction between weak and strong D heads. If we want to extend the analysis of definite DPs e.g. to Mongolian, we need to make sure that the strong D head is overtly spelled out, while weak D receives a zero spell-out. Moreover, Tamil shows DOM-motivated impoverishment of case marking for weak definites but not for strong definites. We can, however, accommodate these effects not by a semantic difference but a morpho-syntactic one. A weak D head consists of the feature list [ι, •N• ≻ D], while a strong D head is made up of [ι, •i• ≻ D]. Depending on the level of underspecification of the exponents, these contexts for insertion are sufficiently distinguished to trigger different realizations. Impoverishment rules/constraints can similarly make use of the different selectional features. Another way to account for different spell-outs specifically is by reference to allomorphy of D in the context of iP. In accordance with locality conditions which restrict allomorphy to apply within the same maximal projection (Bobaljik 2012; Bobaljik and Harley 2017), the sisterhood relation between iP and D is sufficient to license different realizations of the same abstract D morpheme. Crucially, these post-syntactic tools are only of help if the index projects its own phrase, which in turn serves as the complement to D.³⁸ We are now in a position to provide the post-syntactic analysis. Case loss in Korean is attributed to DOM as well as PNI, as this section has shown. As in Turkish, Korean does not show specificity effects. A small survey was conducted with 12 Korean speakers where the effect of accusative case drop was tested on continuations signalling either specific or non-specific readings. The majority of speakers (9/12) did not perceive a correlation between case marking and specificity of the bare objects. Hence, both continuations for (154) are felicitous, with or without case marking on the object. (154) Minho-ka chayk(-ul) ilk-nun-ta ... Minho-NOM book-ACC read-PRS-DECL ‘Minho is reading a book.’ .... Na-nun mwusun chayk-in-ci mol-un-ta. I-TOP which book-COP-CI NEG.know-PRS-DECL ‘I don’t know which book it is.’
−SPEC
³⁸ Hanink’s (2018) reply would presumably involve additional D-to-i lowering, as this is her way to derive the preposition contraction facts we observed for the German paradigm in Section 3.2 (D lowers to i which bleeds contraction with P). This lowering operation would be an alternative to create the locality that is needed for allomorphy to apply. Independent evidence for lowering, however, is difficult to come by in a language like Tamil where definite D is spelled out as zero.
4.3 CASE STUDIES
.... Ku chayk-un Twilight-i-ta. that book-TOP Twilight-COP-DECL ‘That book is Twilight.’
105 +SPEC
The judgements do not change for scrambled objects. Neither the caseless nor the case-marked scrambled object in (155) excludes a continuation. (155) Chayk(-ul)1 Minho-ka ⟨chayk(-ul)1 ⟩ ilk-nun-ta ... book-ACC Minho-NOM read-PRS-DECL ‘Minho is reading a book.’ .... Na-nun mwusun chayk-in-ci mol-un-ta. I-TOP which book-COP-CI NEG.know-PRS-DECL ‘I don’t know which book it is.’
−SPEC
.... Ku chayk-un Twilight-i-ta. that book-TOP Twilight-COP-DECL ‘That book is Twilight.’
+SPEC
Let us turn to the post-syntactic analysis. Following the information provided in O’Grady (1991) and Sohn (1999), we propose the case vocabulary items in (156). Both the nominative and the accusative case markers can be captured with the subfeature [−OBL]. The impoverishment rule in (157a) deletes [−OBL] in the context of [V ], triggering insertion of the elsewhere marker for both subjects and objects if they constitute PNI-DPs, i.e. if they are pseudo-incorporated. Since case marking is additionally optional for demonstratives and post-nominal numeral classifiers, we add two DOM rules in (157b) and (157c). (156) Vocabulary items: a. DAT, INST, ABL, COM, ... ↔ [+GOV,+OBL,−STRUCT,...] b. /-kkeyse/ ↔ [−GOV,−OBL] / [+HON] c. { /-ka/, /-i/ } ↔ [−GOV,−OBL]
(HON.NOM) (NOM)
d. /-(l)ul/ ↔ [+GOV,−OBL]
(ACC)
e. /-uy/ ↔ [+OBL]
(GEN)
f. /-∅/ ↔ [ ] (157) Impoverishment rules for Korean: a. [−OBL] → ∅ / [V ]
PNI
b. [−OBL] → ∅ / [DEM]
DOM
c. [−OBL] → ∅ / [CL]
DOM
106
PSEUDO-INCORPORATION VS DIFFERENTIAL OBJECT MARKING
Within the constraint-based OT system, we again provide a slightly modified definiteness scale, where the specific and non-specific indefinite categories are collapsed into one like in Turkish; see (158). Similar to Tamil, there is a class of noun types which is not ranked with respect to each other nor in relation to the markedness constraint *[-OBL] within the constraint ranking given in (159). These are the DOM-marked categories, i.e. demonstratives and numeral classifiers. Case-marked bare nouns constitute the category indefinite which is ranked above the markedness constraint, caseless bare nouns are PNI-DPs which are ranked lower than *[-OBL]. Finally, we can capture the case loss on weak definites, proper names, and 1/2 pronouns by the use of [ident, •D• ≻ N] which ensures Merge and semantic composition with the PNI-determiner (not shown on the scale).³⁹ (158) Definiteness scale in Korean: PRO ≻ PN ≻ DEF ≻ Q ≻ INDEF ≻ {#-CL, DEM } ≻ PNI (159) Constraint-ranking in Korean: ⎧ *PRO & MAX-C ⎪ ⎪ *PN & MAX-C *DEF & MAX-C ⎨ ⎪ *Q & MAX-C ⎪ *INDEF & MAX-C ⎩
⎫ ⎪ ⎪ ⎬ ⎪ ⎪ ⎭
≫
⎧ *[-OBL] ⎫ { } *DEM & MAX-C ≫ *PNI & MAX-C ⎬ ⎨ *CL & MAX-C ⎭ ⎩
This section presented another language which displays PNI as well as DOM effects, in addition to Hindi and Tamil. Moreover, Korean allows for a typologically unexpected class of noun types to undergo PNI, i.e. proper names, 1st/2nd pronouns, and weak definite phrases. This aspect of the Korean PNI pattern was attributed to the presence of an additional ident-type shifter which coerces e-type semantic objects which are universally identifiable into properties, enabling the PNI determiner to apply and eventually trigger case drop by being ranked low on the definiteness scale. Under the current analysis, the fact that 1st/2nd pronouns and proper names allow case loss does not question the universality of prominence scales. Rather, it supports the idea that PNI is a separate phenomenon which can interact with DOM-marking. The next section will look at German, a language which does not show case loss to begin with. ³⁹ With respect to the class of noun types which can lose case due to [ident, •D• ≻ N] and subsequent Merge with the PNI determiner, we might wonder how the case features on the contained DPs can be targeted by impoverishment. One possibility is to unify the sets of features in the nominal domain by post-syntactic Fusion; see e.g. Assmann et al. (2014), Deal (2015), and Müller (2006) for rule-based accounts and Zimmermann (2017) for an application in a constraint-based system. If the feature sets of the nominal heads are unified, impoverishment applies to all case features as soon as one nominal head constitutes the PNI determiner.
4.3 CASE STUDIES
107
4.3.5 German German stands out in that PNI is not reflected in case drop but in terms of syntactic position. Frey (2015) argues that bare plurals and non-specific indefinites pseudoincorporate in German, based on the observation that they can only occur in a position adjacent to V. In (160a), the indefinite occurs next to the verb and can only be non-specific, proven by the infelicity of adding bestimmt (= certain) in (160b). If the indefinite scrambles out of the verb phrase, it receives a specific reading; see (161a), where the underlying structure is given in (161b). The constrast in (162) shows the identical scrambling ban for bare plurals; see (162b) and (162c). Like non-specific indefinites, bare plurals are licensed in V-adjacent position, shown in (162a). (160) Frey (2015: 237-238) a.
German
Otto hat heute wunderbar eine charmante Mozart-Sonate Otto has today beautifully a charming Mozart sonata gespielt. played ‘Today Otto played a charming Mozart sonata beautifully.’
b. ?? Otto hat heute wunderbar eine bestimmte Mozart-Sonate Otto has today beautifully a certain Mozart sonata gespielt. played ‘Today Otto played a certain Mozart sonata beautifully.’ (161) Frey (2015: 239)
SPECIFIC!
a. Gespielt hat Otto heute eine charmante Mozart-Sonate wunderbar. played has Otto today a charming Mozart sonata beautifully ‘Today Otto played a charming Mozart sonata beautifully.’ b. [CP [VP ⟨...Sonate1 ⟩ gespielt]2 [Cʹ hat Otto heute [eine... Sonate]1 wunderbar ⟨VP2 ⟩]] (162) Frey (2015: 228) a.
Max wird heute Karten spielen. Max will today cards play ‘Max will play cards today.’
b. *Spielen wird Max heute Karten. play will Max today cards ‘Max will play cards today.’ c.
[CP [VP ⟨Karten1 ⟩ spielen]2 [Cʹ will Max heute [Karten]1 ⟨VP2 ⟩]]
108
PSEUDO-INCORPORATION VS DIFFERENTIAL OBJECT MARKING
Similar to Turkish, I attribute the German indefinite determiner a double function, in that it can act as a modifier as well as a D head turning NPs into arguments, compare (163) to (164). I encode specificity for German indefinites lexically into the denotation of the D head in (165c), where I follow Onea and Geist (2011) with respect to the implementation of specificity. This assumption derives the movement requirement of specific indefinites: Since existential quantifiers have to move out of the event domain and specific readings are only made possible by the use of (164)/(165c), specific indefinites obligatorily leave the verbal domain and thus precede low manner adverbs. More details will be given in Section 5.4.5. Non-specific indefinites are PNI-DPs, they do not move due to the [V ] feature of PNI-DP. Since German does not scramble VPs (Grewendorf and Sternefeld 1990; Grewendorf 1995; Müller 1998; von Stechow and Sternefeld 1988), PNI-ed arguments do not undergo scrambling either. Tying the immobility to [V ] also predicts that PNI-ed arguments should be able to topicalize, in parallel to verb phrases. This is in fact borne out; see Section 7.5 for more details. (163) German bare plurals and non-specific indefinites:
PNI-DP DPNI [•N• ≻ D ≻ V]
PNI-DP DPNI NP [•N• ≻ D ≻ V]
NP DPone NP [#1, D]
a (164) Specific German indefinites:
DP NP Done [#1, •N• ≻ D] a (165) Determiner denotations for numerals and indefinites a. ⟦D ≻ V⟧ = λP⟨e,t⟩ λQ⟨e,⟨v,t⟩⟩ λe∃z[P(z) ∧ Q(z)(e)]
DPNI
b. ⟦#1, D⟧ = λz.|z| = 1
Done
c. ⟦#1, •N• ≻ D⟧ = λPλQ : ∃f ∀x[x ∈ De → P ( f(x))].
Done
∃x[P(x) ∧ f(y) = x ∧ Q(x)]
4.3 CASE STUDIES
109
(166) Vocabulary entries for (some) German determiners: a. /ein-/ ↔ [#1]
Done
b. /d-/ ↔ [ι]
wDι , sDι
c. /∅/ ↔ [ ]
DPNI
Since German does not signal PNI with case loss, there does not seem to be an impoverishment rule of the kind we saw in the previous chapters. A constraintbased system would model the German pattern by ranking the markedness constraint below all case faithfulness constraints of the elements of the definiteness scale. Such a ranking is shown in (168). (167) Definiteness scale in German: PRO ≻ PN ≻ DEF ≻ Q ≻ DEM ≻ INDEF ≻ PNI (168) Constraint-ranking in German: ⎧ *PRO & MAX-C ⎪ *PN & MAX-C ⎪ ⎪ *DEF & MAX-C *Q & MAX-C ⎨ ⎪ *DEM & MAX-C ⎪ *INDEF & MAX-C ⎪ ⎩ *PNI & MAX-C
⎫ ⎪ ⎪ ⎪ ⎬ ⎪ ⎪ ⎪ ⎭
≫
{
*[+CASE]
}
The last section of this chapter has demonstrated how PNI languages do not necessarily have to show an effect in case marking. Although it makes PNI more difficult to diagnose, German seems to provide the right set of data, if one takes the syntactic positions in which non-specific indefinites and bare plurals are licensed into account. Another potential PNI language to which this pattern can possibly be extended is Hungarian (Farkas and de Swart 2003) where PNI-ed arguments also occupy a certain position in the clause but are still marked for case. Overall this chapter has shown how pseudo-noun incorporation constitutes its own phenomenon and must be separated from differential object marking. The current account offers a way to do so by analysing PNI as the presence of a hybrid determiner which starts off as a nominal category but turns into a verbal category, which is eventually accessed in post-syntax by case impoverishment rules/constraints. The idea can easily be integrated into existing DOM accounts, under the commonly shared assumption that verbal categories are not prominent enough to show case marking. The following chapters are dedicated to the diagnostics we have applied in this chapter to give the complete typological overview.
5 PNI-property I: Restriction to low scope One of the positive properties in which pseudo-incorporated arguments behave like verbs is the inability to take scope above another c-commanding operator in the clause. Section 5.1 will provide evidence for the scopal inertness of verbal categories. Section 5.2 will lay out the gist of the analysis of scope inertness, exemplarily shown for Hindi and Spanish. Section 5.3 shows how PNI-ed arguments obligatorily reconstruct after they have moved from their base positions. Finally, in section 5.4 I will go through the PNI languages of the core data set and show for each of those languages how a restriction to low scope is a natural consequence for PNI-ed arguments.
5.1 Evidence for scopal inertness of verbal categories V-to-T or V-to-C movement never seems to affect the scopal relations within a clause, an observation that has led Chomsky (1995: 368) to ship head movement off to the PF component of grammar, where it counterfeeds any kind of scopal shifts. This assumption resolves other problematic properties of head movement such as the violation of the Extension Condition (Chomsky 1995) and the inability of head-moved constituents to c-command their traces.¹ Moreover, Chomsky (2001: 37) names Baker-style incorporation (Baker 1988, 1995, 1996) as another likely candidate to apply at PF, due to the lack of interpretive effects, again explicitly discussed in Baker (2009: 158). In the following, I will recap three empirical arguments that have been made, specifically in favour of the scopal inertness of lexical verbs. The first relates to the scope properties directly, the second is based on NPI licensing and the third on VP-ellipsis. Harley (2004) demonstrates semantically vacuous V-to-T movement with French verb raising past negation; compare (1a) to (1b).
¹ Placing head movement on PF is of course not the only solution which has been proposed to overcome its syntactic pecularities. One line of approaches analyses verb movement as instances of remnant XP-movement; see Mahajan (2003) for Hindi, Müller (2004b) for German, and Wiklund et al. (2007) for Norwegian and Icelandic. Another cluster of accounts involves the idea of reprojection, where it is the moving head that determines the label of the newly formed root node by taking the root node as its complement (Bayer and Brandner 2013; Fanselow 2002a; Georgi and Müller 2010; Holmberg 1991).
Pseudo-Noun Incorporation and Differential Object Marking. Imke Driemel, Oxford University Press. © Imke Driemel (2023). DOI: 10.1093/oso/9780192866400.003.0005
5.1 EVIDENCE FOR SCOPAL INERTNESS OF VERBAL CATEGORIES
(1) Harley (2004: 6)
111
French
a. Jean ne parlait pas franc¸ais. Jean speak.IMP not French ‘John didn’t speak French.’
¬V
b. Jean n’as pas parle´ franc¸ais. Jean has not spoken French ‘John hasn’t spoken French.’
¬V
Platzack (2013), furthermore, shows how V2 does not interact with quantified subjects in Swedish. In (2a) the modal kan has moved to C, whereas in (2b) the auxiliary forces the modal to be spelled out lower, presumably below the quantified subject. Either clause, however, allows for both readings. (2) Platzack (2013: 33–34)
Swedish
a. Inte alla pojkarna kan ta en plats i laget. not all boys.DEF can take a place in team.DEF ‘Not all the boys have been able to take a place in the team.’
◊∀, ∀◊
b. Inte alla pojkarna har kunnat ta en plats i laget. not all boys.DEF have can.INF take a place in team.DEF ‘Not all the boys have been able to take a place in the team.’
◊∀, ∀◊
While Platzack (2013) and Harley (2004) take the scope facts above to be indicative of verb movement at PF, Keine and Bhatt (2016) discuss scope data from German that must rather be reconciled with syntactic head movement and obligatory semantic reconstruction of the verb in its first merged position. Verb clusters in German long passive structures exhibit compactness properties that suggest complex predicate formation; compare (3a) to (3b). The key characteristic for long passives is the nominative case on the embedded object. (3) Haider (2010: 313)
German
a. Gestern wurde den Traktor zu reparieren (nicht) versucht. yesterday was the tractor.ACC to repair not tried ‘Yesterday it was (not) tried to repair the tractor.’ LOCAL PASS b. Gestern wurde der Traktor zu reparieren (*nicht) versucht. yesterday was the tractor.NOM to repair not tried ‘Yesterday it was (not) tried to repair the tractor.’ LONG PASS In order to capture the case contrast, local passives are assumed to be derived by the presence of a vP projection, assigning accusative case to the object; see (4), a projection which is missing from long passives, given in (5). Keine and Bhatt (2016) propose compactness in long passives to derive from V-to-V movement,
112
PNI-PROPERT Y I: RESTRICTION TO LOW SCOPE
where embedded V and matrix V undergo function composition and the embedded V leaves a higher type trace which forces the verb complex to be interpreted in the base position of embedded V.
VP
(4)
vP v
VP DP den Traktor.ACC
VP
(5)
V zu reparieren
V
VP
V versucht
VP DP
der Traktor
λQ V Q
V V zu reparieren versucht
Evidence for reconstruction comes from scopal interactions of object quantifiers with the matrix verb. In long passives, object quantifiers outscope the matrix verb although they are asymmetrically c-commanded by matrix V. Whereas the local passive in (6) can only receive a forget » only reading, i.e. a scenario where Fritz was supposed to be introduced to only one student but someone forgot and introduced him to all of them, the long passive in (7) can only get an only » forget reading, i.e. a situation in which Fritz was supposed to be introduced to all students but there is one student to whom it was forgotten that Fritz was to be introduced. (6) Keine and Bhatt (2016: 1454) German local passive Erst gestern wieder wurde den Fritz nur einem einzigen just yesterday again was the Fritz.ACC only a single Studenten vorzustellen vergessen student.DAT to.introduce forgotten ‘Just yesterday it was forgotten to introduce Fritz to only one student.’ FORGET » ONLY, *ONLY » FORGET (7) Keine and Bhatt (2016: 1454) German long passive Erst gestern wieder wurde der Fritz nur einem einzigen just yesterday again was the Fritz.NOM only a single Studenten vorzustellen vergessen student.DAT to.introduce forgotten ‘Just yesterday it was forgotten to introduce Fritz to only one student.’ *FORGET » ONLY, ONLY » FORGET After the embedded verb has undergone movement in (5), it composes with the matrix verb. Function composition enables predicates to share their argument structure, abstractly shown in (8). (8) Function Composition ⟦pred1 ⟧ ° ⟦pred2 ⟧ = λx⟦pred1 ⟧(⟦pred2 ⟧(x))
(Steedman 1985: 530)
5.1 EVIDENCE FOR SCOPAL INERTNESS OF VERBAL CATEGORIES
113
Since embedded verb and matrix verb share their arguments, the matrix verb has to be interpreted wherever the embedded verb is interpreted with respect to the arguments they are taking. The embedded verb leaves a higher type trace, forcing it to take scope in its launching site. Together these two assumptions derive the fact that the object quantifier semantically scopes over the matrix verb in (5), although it is still in the matrix verb’s syntactic scope. The details are given in (9). VP
(9) VP VP
nur einem einzigen Studenten
V
2
λQ
1
V′
DP DP
4
V vorzustellen
3
V vergessen
V Q
Fritz
⟦ ⟦
1
2
g
⟧ = ⟦DP⟧(⟦V ʹ⟧) = λR∃!x[STUDENT(x) ∧ ∃e[R(x)(e)]](Q( fritz)) = ∃!x[STUDENT(x) ∧ ∃e[Q( fritz)(x)(e)]] g
⟧ = λQ∃!x[STUDENT(x) ∧ ∃e[Q( fritz)(x)(e)]]
FA LA
g
⟦
3
⟧ = ⟦vergessen⟧ ° ⟦vorzustellen⟧ = λxλy⟦vergessen⟧(⟦vorzustellen⟧(x)(y)) = λxλy(λP λe[FORGET(e) ∧ TH(e) = P]) (λzλuλeʹ[INTRO(eʹ) ∧ TH(eʹ) = z ∧ G(eʹ) = u](x)(y)) FC = λxλy(λP λe[FORGET(e) ∧ TH(e) = P]) (λeʹ[INTRO(eʹ) ∧ TH(eʹ) = x ∧ G(eʹ) = y]) = λxλyλe[FORGET(e) ∧ TH(e) = λeʹ[INTRO(eʹ) ∧ TH(eʹ) = x ∧ G(eʹ) = y]
⟦
4
⟧ = ⟦VP⟧(⟦V⟧) = λQ∃!x[STUDENT(x) ∧ ∃e[Q( fritz)(x)(e)]] (λxλyλe[FORGET(e) ∧ TH(e) = λeʹ[INTRO(eʹ) ∧ TH(eʹ) = x ∧ G(eʹ) = y]) = ∃!x[STUDENT(x) ∧ ∃e[λxλyλe[FORGET(e) ∧ TH(e) = λeʹ[INTRO(eʹ) ∧ TH(eʹ) = x ∧ G(eʹ) = y]( fritz)(x)(e)]] = ∃!x[STUDENT(x) ∧ ∃e[FORGET(e) ∧ TH(e) = λeʹ[INTRO(eʹ) ∧ TH(eʹ) = fritz ∧ G(eʹ) = x]]
g
Keine and Bhatt’s account of German long passives stands out not only in that it represents one of the rare semantic analyses of syntactic verb movement, but it also shows how verb movement can result in semantic effects, albeit of the reverse kind: V-to-V movement in effect pulls down the semantic denotation of the whole verb complex to the base position of the lowest merged V. Moving away from direct scope evidence, we now turn to two other kinds of arguments that have been made in favour of verbs being interpreted in their base positions. Johannessen (2003) argues for head movement at PF based on verbal
114
PNI-PROPERT Y I: RESTRICTION TO LOW SCOPE
Negative Polarity Items (NPI) in Norwegian, which move to C where they are not c-commanded by the negative licenser in the clause. The verbal NPIs raker and enset cannot occur without negation, although they can precede it, shown in (10a) and (10b), respectively. (10) Johannessen (2003: 51) a.
Det raker *(ikke) de utenlandske arbeidsgiverne. it concerns not the foreign employers.DEF ‘It is of no business to the foreign employers.’
b.
Hun enset *(ikke) bråket. she sensed not noise.DEF ‘She didn’t notice the noise.’
Norwegian
Another set of data which signals verb movement but necessary interpretation in base position concerns the verb identity requirement in Verb-stranding VP-ellipsis (Goldberg 2005; Gribanova 2013; McCloskey 2010)—an observation about verbs which escape their VP prior to deletion and which need to be identical to the verb in the antecedent clause. The verb identity requirement is shown with fragment answers in (11) for Hebrew. (11) Goldberg (2005: 174)
Hebrew
A:
(Ha’im) Miryam hevi’a et Dvora la-xanut? Q Miryam bring[PST.3SG.F] ACC Dvora to.the-store ‘(Did) Miryam bring Dvora to the store?’ B: Ken, hi hevi’a. yes she bring[PST.3SG.F] ‘Yes, she brought Dvora to the store.’ B’: #Ken, hi lakxa. yes she take[PST.3SG.F] ‘Yes, she took Dvora to the store.’ B”: #Lo, hi ˇsalxa! no she send[PST.3SG.F] ‘No, she sent Dvora to the store!’ Following Merchant (2001), a VP can be deleted if there is an antecedent VP by which it is entailed and vice versa (modulo ∃-type shifting). Crucially, the verb has to be interpreted in its base position inside the VP, in order to explain the contrasts in (11). Since only (12b) and (13a) mutually entail each other, only (11B) is an appropriate answer to (11A).²
² F-closure, the other component of Merchant’s analysis, is not relevant here since no constituent within the VP is focused.
5.2 PNI-ED ARGUMENTS ARE RESTRICTED TO THE EVENT DOMAIN
115
(12) Antecedent VP in (11A) a. meaning of antecedent VP: [VP λx.x bring Dvora to the store] b. ∃-type shift of antecedent VP: [VP ∃x.x bring Dvora to the store] (13) Elided VPs a. ∃-type shift of elided VP: [VP ∃x.x bring Dvora to the store]
(11B)
b. ∃-type shift of B’-answer: [VP ∃x.x took Dvora to the store]
(11B’)
c. ∃-type shift of B”-answer: [VP ∃x.x sent Dvora to the store]
(11B”)
The verb identity requirement in Verb-stranding VP-ellipsis has been taken as evidence either in favour of verb movement at PF (Schoorlemmer and Temmerman 2012) or obligatory reconstruction of verb movement in syntax proper (Goldberg 2005). To sum up, I have presented three types of empirical arguments that require base interpretation of verbal categories. We will now move on to pseudo-incorporated nouns and show that they seem to show the same restriction.
5.2 PNI-ed arguments are restricted to the event domain For this study, I consider two types of scope contexts, one involves negation and the other universally quantified subjects. The interaction of scope and case drop is illustrated with data from the PNI/DOM literature on Hindi and Spanish. Spanish objects in (14) can only scope above negation in case they are marked with case. (14) López (2012: 13,14)
Spanish
a. Juan no amo´ a una mujer. Juan NEG loved DOM a woman ≠ Juan did not love any woman. = There was a woman Juan did not love.
*¬∃ ∃¬
b. Juan no amo´ una mujer. Juan NEG loved a woman = Juan did not love any woman. ≠ There was a woman Juan did not love.
¬∃ *∃¬
Caseless bare objects in Hindi are equally blocked from taking wide scope over another quantifier in the clause; see (15c). If they are case marked they become definite; see (15b). Hindi objects are not generally excluded from taking wide scope from their base positions, as (15a) shows. With the use of the indefinite determiner ek objects are able to scope below and above the universally quantified subject.
116
PNI-PROPERT Y I: RESTRICTION TO LOW SCOPE
(15) Dayal (2011: 128) a. Har aurat ek bacce-ko samhaal rahii thii. every woman one child-ACC look-after PROG be.PST ‘Every woman was looking after a particular child.’ b. Har aurat bacce-ko samhaal rahii thii. every woman child-ACC look-after PROG be.PST ‘Every woman was looking after the child.’ c. Har aurat bacce samhaal rahii thii. every woman child look-after PROG be.PST ‘Every woman was looking after a child/children.’
Hindi ∀∃,∃∀
DEFINITE
∀∃,*∃∀
As I have argued in Section 3.3, we derive the inability of PNI-ed arguments to take scope over another operator in the clause by assigning them the semantics of an existential quantifier that must be interpreted inside the event domain. This quantifier exists alongside other generalized quantifiers (Barwise and Cooper 1981), which I show for comparison in (16). I also include the definite D heads for completeness sake. (16) Types of D heads a. ⟦DPNI ⟧ = λP⟨e,t⟩ λQ⟨e,⟨v,t⟩⟩ λe∃z[P(z) ∧ Q(z)(e)]
PNI
b. ⟦D∃ ⟧ = λP⟨e,t⟩ λQ⟨e,t⟩ ∃z[P(z) ∧ Q(z)]
existential GQ
c. ⟦D∀ ⟧ = λP⟨e,t⟩ λQ⟨e,t⟩ ∀z[P(z) → Q(z)]
universal GQ
d. ⟦wDι ⟧ = λP⟨e,t⟩ : ∃!x[P(x)].ιx[(P)(x)]
weak definite
e. ⟦sDι ⟧ = λP⟨e,t⟩ λy : ∃!x[P(x) ∧ x = y].ιx[P(x) ∧ x = y]
strong definite
DPNI is different from the other D heads in that it restricts its second argument slot to be of type ⟨e, ⟨v, t⟩⟩, a function from individuals to predicates of events. This, I argue, is the verbal component of pseudo-incorporated arguments which distinguishes them from proper arguments. Similar to the lexical verbs we looked at in Section 5.1, PNI-ed arguments have to undergo reconstruction if they move away from their base positions, the only difference being that verbs do so independently of where they move to, whereas PNI-ed arguments only reconstruct if they leave vP. The limitation to the event domain does not per se predict PNI-ed arguments to necessarily scope below other quantifiers and negative operators in the sentence. This property follows from the treatment of quantifiers in a Neo-Davidsonian event semantics as well as from a commonly shared assumption about the scope of negation. Since verbs within event semantics introduce an additional event variable, they cannot directly undergo semantic composition with quantifiers in
5.2 PNI-ED ARGUMENTS ARE RESTRICTED TO THE EVENT DOMAIN
117
their in-situ positions. Different solutions have been proposed: Schein (1993) e.g. proposes to rewrite lexical entries of quantifiers, while Champollion resolves the type mismatch with silent θ-role heads (2011, 2015) and/or type shifters (2016a, 2016b) which attach to the respective quantified arguments. I take a syntactic approach and assume with Landman (2000) that GQs must move out of the event domain in order to be interpreted. An abstract LF representation is shown in (17). (17) Quantifiers QR out of the event domain
(Landman 2000: 52–53)
vP ⟨t⟩ DP[–def ] ⟨⟨e, t⟩, t⟩
vP ⟨e, t⟩ 1
vP ⟨t⟩ vP ⟨v, t⟩ t1 ⟨e⟩
v ′ ⟨e, ⟨v, t⟩⟩
This move ensures semantic composition while also drawing a distinction between (16a) on the one hand and (16b) and (16c) on the other. Since GQs are forced to move out of the event domain in order to be interpreted, and PNI-ed arguments can only be interpreted inside the event domain, the former will always outscope the latter. Together with the assumption that negation applies not lower than existential closure (Chung and Ladusaw 2004; Penka 2010; Swart 2016; von Stechow 2012; Zeijlstra 2004), we can explain the low scope restriction of PNI-ed arguments. (18) provides an abstract version of the proposal. (18) Overview scope scenarios: ⟨t⟩
⟨t⟩
PNI
GQ
⟨e,t⟩ μ1
μ1
⟨t⟩ ⟨t,t⟩
⟨e,t⟩
⟨t⟩
⟨t⟩ ⟨t,t⟩
⟨t⟩
vP⟨v,t⟩
vP⟨v,t⟩ vP⟨e,⟨v,t⟩⟩
PNI VP
v⟨e,⟨v,t⟩⟩
PNI V⟨e,⟨v,t⟩⟩
vP⟨e,⟨v,t⟩⟩
GQ VP
v⟨e,⟨v,t⟩⟩
GQ V⟨e,⟨v,t⟩⟩
118
PNI-PROPERT Y I: RESTRICTION TO LOW SCOPE
I will now illustrate the proposal for Spanish and Hindi focusing on the data in (14) and (15). Spanish negation applies immediately after existential closure of the event variable. Hence, GQs raise above negation (covertly) whereas PNI-DPs stay inside the event domain. In (14a), the object is case marked and therefore an instance of an existential quantifier. LFs are provided in (19) and (20). (19) ∃ > ¬ reading for (14a) NegP⟨t⟩ x[WOM(x) e[LOVE(x)(e) AG( juan)(e)]] DP2 NegP⟨e,t⟩ ⟨⟨e,t⟩,t⟩ λx e[LOVE(g t2 x (2))(e) AG( juan)(e)] λP x[WOM(x) P(x)] μ2
NegP⟨t⟩ e[LOVE(g(2))(e) AG( juan)(e)] Neg⟨t,t⟩ p[p]
vP⟨t⟩ e[LOVE(g(2))(e) ∧AG( juan)(e)] vP⟨v,t⟩ λe[LOVE(g(2))(e) AG( juan)(e)] DP Juan
v′⟨e,⟨v,t⟩⟩ λxλe[LOVE(g(2))(e) AG( x)(e)] v⟨e,⟨v,t⟩⟩ λxλe[AG(g)(e)]
VP⟨v,t⟩ λe[LOVE(g(2))(e)] V⟨e,⟨v,t⟩⟩ ⟨DP2⟩ λxλe[LOVE(x)(e)]
(20) *¬ > ∃ reading for (14a)
VP⟨???⟩ V⟨e,⟨v,t⟩⟩ λxλe[LOVE(x)(e)]
DP2
⟨⟨e,t⟩, t⟩
λP x[WOM(x) P(x)]
The object in (14b) is PNI-ed since it is not case marked. The low scope reading follows if the direct object is an instance of PNI-DP; see (21). PNI-DP cannot scope over negation since it cannot be interpreted outside the event domain; see (22).
5.2 PNI-ED ARGUMENTS ARE RESTRICTED TO THE EVENT DOMAIN
(21) ¬ > ∃ reading for (14b) NegP⟨t⟩ e z[WOM(z) LOVE(z)(e) AG( juan)(e)] Neg⟨t,t⟩ p[p]
vP⟨t⟩ e z[WOM(z) LOVE(z)(e) AG( juan)(e)] vP⟨v,t⟩ λe z[WOM(z) LOVE(z)(e) AG( juan)(e)] DP Juan
v′⟨e⟨v,t⟩⟩ λxλe z[WOM(z) LOVE( z)(e) AG( x)(e)] v⟨e,⟨v,t⟩⟩ λxλe[AG(x)(e)]
VP⟨v,t⟩ λe z[WOM(z) LOVE(z)(e)] PNI-DP ⟨⟨e,⟨v,t⟩⟩, ⟨v,t⟩⟩ λQλe z[woM(z) Q(z)(e)]
V⟨e,⟨v,t⟩⟩ λxλe[LOVE(x)(e)]
PNI-D NP⟨e,t⟩ λy[woM(y)] ⟨⟨e,t⟩,⟨⟨e,⟨v,t⟩⟩, ⟨v,t⟩⟩⟩ λPλQλe z[P(z) Q(z)(e)]
(22) *∃ > ¬ reading for (14b) NegP⟨???⟩
PNI-DP2 ⟨⟨e,⟨v,t⟩⟩, ⟨v,t⟩⟩ λQλe z[woM(z) Q(z)(e)]
NegP⟨e,t⟩
λx. e[LOVE(g t2 x (2))(e) AG(juan)(e)] μ2
NegP⟨t⟩ e[LOVE(g(2))(e) ∧AG(juan)(e)] Neg⟨t,t⟩ p[p]
vP⟨t⟩ e[LOVE(g(2))(e) AG( juan)(e)] vP⟨v,t⟩ λe[LOVE(g(2))(e) AG(juan)(e)] DP Juan
v′⟨e,⟨v,t⟩⟩ λxλe[LOVE(g(2))(e) AG( x)(e)] v⟨e,⟨v,t⟩⟩ λxλe[AG(x)(e)]
VP⟨v,t⟩ λe[LOVE(g(2))(e)] V⟨e,⟨v,t⟩⟩ ⟨DP2⟩ λxλe[LOVE(x)(e)]
119
120
PNI-PROPERT Y I: RESTRICTION TO LOW SCOPE
Let us turn to the GQ scenarios in Hindi. In (15a) both subject and object constitute GQs, thus they both move covertly out of the event domain.³ While the translation suggests only the ∃∀ reading, Dayal mentions in prose (2011: 128) that the low scope reading is also available. For the surface scope reading ∀∃, the subject scopes over the object at LF; see (23). (23) ∀ > ∃ reading of (15a) vP⟨t⟩ y[WOM(y) → x[CHILD(x) AG(y)(e)]]] DP1 ⟨⟨e, t⟩, t⟩ λQ y[WOM(y) Q(y)]
e[LOOK(x)(e)
vP⟨e,t⟩ λy x[CHILD(x) e[LOOK(x)(e) AG(g t1 y (1))(e)]] μ1
vP⟨t⟩ x[CHILD(x) e[LOOK(x)(e) AG(g(1))(e)]]
DP2 ⟨⟨e, t⟩, t⟩ λP x[CHILD (x) P(x)]
vP⟨e,t⟩ λx e[LOOK(g t2 x(2))(e) AG(g(1))(e)] μ2
vP⟨t⟩ e[LOOK(g(2))(e) AG(g(1))(e)] vP⟨v,t⟩ λe[LOOK(g(2))(e) AG(g(1))(e)] ⟨DP1⟩
v ′⟨e,⟨v,t⟩⟩ λxλe[LOOK(g(2))(e) AG(x))(e)]
VP⟨v,t⟩ λe[LOOK(g(2))(e)] ⟨DP2⟩
v ⟨e,⟨v,t⟩⟩ λxλe[AG(x)(e)]
V⟨e,⟨v,t⟩⟩ λxλe[LOOK(x)(e)]
For the inverse reading ∃∀, the object scopes over the subject at LF, sketched in (24). (24) ∃ > ∀ reading of (15a) [DP2 μ2 ... DP1 μ1 ... ∃ [vP ⟨DP1 ⟩ [VP ⟨DP2 ⟩ look]]] ³ I am glossing over the fact that there is good reason to assume that ek is a numeral determiner, following the arguments in Dayal (2004, 2017). Foreshadowing the discussion in Section 5.4.1, a numeral determiner can be given the denotation below. (i) ⟦Done ⟧ = λP⟨e,t⟩ λQ⟨e,t⟩ ∃z[P(z) ∧ |z| = 1 ∧ Q(z)] Exchanging this denotation with the existential quantifiers in (23) and (24) does not influence the scopal relations for (15a).
5.2 PNI-ED ARGUMENTS ARE RESTRICTED TO THE EVENT DOMAIN
121
The definite reading in (15b) follows from the presence of either (16d) or (16e). The subject quantifier, again, has to raise out of the event domain. Since Dayal (2011: 128) describes a context involving a contextually salient child, we show the LF with the strong definite determiner; see (25).
(25) definite reading of (15b) vP⟨t⟩ !x[CHILD(x) x = g(3)]. y[WOM(y) e[LOOK(ιx[CHILD(x) x = g(3)])(e) AG(y)(e)] DP1 ⟨⟨e, t⟩, t⟩ λQ y[WOM(y) Q(y)]
vP⟨e,t⟩ λ x : !x[CHILD(x) x = g(3)]. e[LOOK(ιx[CHILD(x) x = g(3)])(e) AG(g t1 μ1
x
(1))(e)]
vP⟨t⟩ !x[CHILD(x) x = g(3)]. e[LOOK(ιx[CHILD(x) x = g(3)])(e) AG(g(1))(e)] vP⟨v,t⟩ λ e : !x[CHILD(x) x = g(3)]. LOOK(ιx[CHILD(x) x = g(3)])(e) AG(g(1))(e) ⟨DP1⟩
v ′⟨e,⟨v,t⟩⟩ λ xλ e : !x[CHILD(x) x = g(3)]. LOOK(ιx[CHILD(x) x = g(3)])(e) AG(x)(e)
VP⟨v,t⟩ λ e : !x[CHILD(x) x = g(3)]. LOOK(ιx[CHILD(x) x = g(3)])(e) DP⟨e⟩ !x[CHILD(x) x = g(3)]. ιx[CHILD(x) x = g(3)]
(26) definite noun phrase in (15b) DP⟨e⟩ !x[CHILD(x) x = g(3)]. ɩx[CHILD(x) x = g(3)] D′⟨e,e⟩ g(3) λy : !x[CHILD(x) x = y]. ɩx[CHILD(x) x = y ] NP⟨e,t⟩ D λy[CHILD(y)] ⟨⟨e,t⟩,⟨e,e⟩⟩ λPλy : !x[P(x) x = y]. ɩx[P(x) x = y ]
V⟨e,⟨v,t⟩⟩ λ xλ e[LOOK(x)(e)]
v ⟨e,⟨v,t⟩⟩ λ xλ e[AG(x)(e)]
122
PNI-PROPERT Y I: RESTRICTION TO LOW SCOPE
(15c) presents a PNI context, therefore DPNI forms the object DP, triggering case drop at PF. The surface scope reading ∀∃ is possible since PNI-DP can be interpreted in its base position; see (27). In order for the object DP to take scope above the subject, however, it needs raise out of the event domain. Since this is not an option for PNI-DP, the reading is not licensed. The attempted LF is shown in (28) (27) ∀ > ∃ reading of (15c) vP⟨t⟩ y[WOM(y) → e z[CHILD(z) LOOK(z)(e) AG(y)(e)]] DP1 ⟨⟨e,t⟩,t⟩ λQ y[WOM(y) →Q(y)]
vP⟨e,t⟩ λx e z[CHILD(z) LOOK(z)(e) AG(g t1
μ1
x
(1))(e)]
vP⟨t⟩ e z[CHILD(z) LOOK(z)(e) AG(g(1))(e)]
vP⟨v,t⟩ λ e z[CHILD(z) LOOK(z)(e) AG(g(1))(e)]
⟨DP1⟩
v′⟨e,⟨v,t⟩⟩ λxλe z[CHILD(z) LOOK(z)(e) AG(x)(e)]
VP⟨v,t⟩ λe z[CHILD(z) LOOK(z)(e)]
v⟨e,⟨v,t⟩⟩ λxλe[AG(x)(e)]
V⟨e,⟨v,t⟩⟩ PNI-DP λxλe[LOOK(x)(e)] ⟨⟨e,⟨v,t⟩⟩,⟨v,t⟩⟩ λQλe z[CHILD(z) Q(z)(e)] PNI-D NP⟨e,t⟩ ⟨⟨e,t⟩,⟨⟨e,⟨v,t⟩⟩,⟨v,t⟩⟩⟩ λy[CHILD(y)] λPλQλe z[P(z) Q(z)(e)]
5.3 PNI-ED ARGUMENTS RECONSTRUCT
123
(28) *∃ > ∀ reading of (15c) vP⟨??? ⟩
vP⟨e,t⟩ P NI -DP 2 λ x y[ WOM (y) → e[ LOOK ( g t 2 x (2))(e) ⟨⟨e, ⟨v, t⟩⟩, ⟨v, t⟩⟩ AG (y)(e)]] λ Q λ e z[ CHILD (z) Q(z)(e)] μ2 vP⟨t⟩ y[ WOM (y) → e[ LOOK ( g(2))(e) AG (y)( e)]] DP 1 vP⟨e,t⟩ λ x e[ LOOK ( g(2))(e) ⟨⟨e, t⟩, t⟩ AG ( g t 1 x (1))(e)] λ Q y[ WOM (y) → Q(y)] μ1 vP⟨t⟩ e[ LOOK ( g(2))(e) AG ( g(1))(e)] vP⟨v,t⟩ λ e[ LOOK ( g(2))(e) AG ( g(1))(e)] ⟨DP1 ⟩
v ′⟨e,⟨v,t⟩⟩ λ xλ e[ LOOK ( g(2))(e) AG ( x)(e)]
VP ⟨v,t⟩ λ e[ LOOK ( g(2))(e)] ⟨DP2 ⟩
v ⟨e,⟨v,t⟩⟩ λ xλ e[ AG ( x)(e)]
V ⟨e,⟨v,t⟩⟩ λ xλ e[ LOOK ( x)(e)]
As we have seen in this section, the low scope restriction of PNI-ed arguments can be readily derived by the denotation of the hybrid determiner DPNI and commonly shared assumptions of Neo-Davidsonian event semantics. Before we turn to the case studies, we will briefly discuss the interpretation of PNI-ed arguments which have overtly moved out of the event domain.
5.3 PNI-ed arguments reconstruct Dayal (2011: 137) shows for Hindi that PNI-ed arguments do not have to be linearly adjacent to V. The sentences in (29) show negation occurring in between the verb and the direct object. As can be seen in (29b), this word order is also possible with caseless bare objects. Hence, scrambling of the caseless object over
124
PNI-PROPERT Y I: RESTRICTION TO LOW SCOPE
negation does not affect the scopal relation. That this property is tied to the PNI status becomes obvious if we compare (29b) to (29a), where an existential quantifier presumably undergoes the same movement operation but with the ability to take scope in its launching as well as in its landing site. (29) Dayal (2011: 128,137)
Hindi
a. Anu [ek bacce-ko]1 nahiiN ⟨DP1 ⟩ samhaalegii Anu one child-ACC not look.after.FUT ‘Anu will not look after a particular child.’
¬∃,∃¬
b. Anu bacca1 nahiiN ⟨DP1 ⟩ samhaalegii Anu child not look.after.FUT ‘Anu will not look after children.’
¬∃,*∃¬
Similar observation can be made for Turkish and Korean, although the data is not as suggestive since, in contrast to Hindi, negation in Turkish and Korean is realized as a bound morpheme attaching to the lexical verb, thereby making it difficult to diagnose whether the object has in fact left the vP. Öztürk (2009) provides examples which show that bare PNI-ed objects can be dislocated from their base positions. Intermediate scrambling of the PNI-ed object in (30) does not enable the object to take scope above negation, as the translation suggests and as my informants confirm. (30) Öztürk (2009: 339) C ¸ ay1 ben ⟨DP1 ⟩ ic¸-me-di-m. tea I drink-NEG-PST-1SG ‘I did not do tea-drinking.’
Turkish *∃¬, ¬∃
Similar to Hindi, the PNI counterpart in Turkish must occur with an indefinite determiner. If the object is case marked, it scopes above negation. If the object is caseless, negation scopes over the object. Again, if we assume that the indefinites in (31a) and (31b) as well as the bare noun in (30) each target the same landing site, the contrast is derived by obligatory reconstruction of PNI-DPs into base position in (30) and (31b). (31) Scrambling of PNI-ed object and reconstruction a. [Bir c¸ay-ı]1 ben ⟨DP1 ⟩ ic¸-me-di-m. a tea-ACC 1SG.NOM drink-NEG-PST-1SG ‘There is a tea I didn’t drink.’
Turkish ∃¬, *¬∃
5.3 PNI-ED ARGUMENTS RECONSTRUCT
b. [Bir c¸ay]1 ben ⟨DP1 ⟩ ic¸-me-di-m. a tea 1SG.NOM drink-NEG-PST-1SG ‘I did not do tea-drinking.’
125
*∃¬, ¬∃
Kwon and Zribi-Hertz (2006) provide judgements for Korean caseless objects in their in-situ positions and with respect to negation. In line with other PNI languages, caseless objects are scopally inert, shown in (32). (32) Kwon and Zribi-Hertz (2006: 118) a. Minsu-neun sagwa-leul meog-ji anh-ass-da. Minsu-TOP apple-ACC eat+JI NEG-PST-DECL ‘Minsu did not eat apples/the apple.’ b. Minsu-neun sagwa meog-ji anh-ass-da. Minsu-TOP apple eat+JI NEG-PST-DECL ‘lit. Minsu did not apple-eat.’
Korean ∃¬,¬∃
*∃¬,¬∃
Crucially, the scope relations do not change once the caseless object has undergone scrambling, shown here to a position above the subject in (33b). The comparison to a case-marked object in (33a) suggests that the scopal properties in (33b) are not caused by the type of landing site. (33) Scrambling of PNI-ed object and reconstruction a. Kkoch-ul1 Yusu-ka ⟨DP1 ⟩ sa-ci anh-ass-ta. flower-ACC Yusu-NOM buy-CI NEG-PST-DECL ‘Yusu did not buy a specific/any flower(s)’ b. Kkoch1 Yusu-ka ⟨DP1 ⟩ sa-ci anh-ass-ta. flower Yusu-NOM buy-CI NEG-PST-DECL ‘Yusu did not buy any flower(s)’
Korean ∃¬, ¬∃
*∃¬, ¬∃
What becomes apparent from the scope facts above is that, at least in Hindi, Turkish, and Korean, PNI-ed arguments can nevertheless move out of vP, even though they cannot be interpreted in those positions. Consequently, I will assume that PNI movement obligatorily reconstructs, if it targets a position outside of vP. There are at least two competing theories of reconstruction available. The syntactic approach is a direct reaction to the Copy Theory of Movement (Chomsky 1993, 1995) and assumes that only the lowest copy of a movement chain is interpreted at LF (Fox 1999; Poole 2017; Romero 1997, 1998). Sportiche (2016)
126
PNI-PROPERT Y I: RESTRICTION TO LOW SCOPE
argues that syntactic reconstruction can be made to follow if the Principle of Full Interpretation (Chomsky 1995) applies to syntactic objects in general rather than their occurences. A movement chain is considered a syntactic object, therefore if at least one of the copies is interpreted, Full Interpretation is satisfied. He derives interpretation of the lowest copy by another LF requirement that demands predicate argument relations to be strictly local. Following Sportiche (2016) and Keine and Poole (2018b), we can describe syntactic reconstruction as higher copy neglection. Semantic reconstruction is based on the possibility of leaving higher type traces (Chierchia 1995; Cresti 1995; Rullmann 1995). This enables moved phrases to still be interpreted in their landing sites. An application was already discussed in Section 5.1. Keine and Bhatt (2016) employ semantic reconstruction for German complex predicate formation. The mechanism is best described when compared to scope shifting quantifier movement where lower copies are turned into variables of type ⟨e⟩ and the quantifier takes the λ-abstraction as an argument in its landing position. Semantic reconstruction of a quantifier lets the lower copy be of type ⟨⟨e, t⟩, t⟩ which in turn makes the quantifier in its landing position an argument to the λ-abstraction. This constellation will lead to the quantifier scoping in its launching site, although it is interpreted in its landing site. (34) gives an abstract overview comparing syntactic and semantic reconstruction. (34) Syntactic vs. semantic reconstruction of generalized quantifiers a. DP ... OP ... DP
LF: [ DP ⟨⟨e,t ⟩,t ⟩ ... [ ... OP ... DP⟨⟨e,t ⟩,t ⟩ ]]
syntactic
b. DP ... OP ... DP
LF: DP [ λQ ⟨⟨e,t ⟩,t ⟩ ... [ ... OP ... Q ⟨⟨e,t ⟩,t ⟩ ]]
semantic
The current approach to pseudo-noun incorporation is compatible with syntactic as well as semantic reconstruction, which I will demonstrate with the Hindi data in (29). I provide the LF tree for the syntactically reconstructed reading of (29b) in (35). According to Bhatt (2016) and Keine and Poole (2018b), Hindi intermediate scrambling is ambiguous between A-scrambling and Ā-scrambling. Hence, the word order in (29b) is the result of intermediately scrambling the PNI-ed object out of vP above negation. Due to the PNI-DP status, the object has to reconstruct into its base position. Additionally, the subject moves to Spec,TP (without any semantic consequences).
5.3 PNI-ED ARGUMENTS RECONSTRUCT
127
(35) ¬ > ∃ reading for (29b) via syntactic reconstruction TP⟨t⟩ e z[ CHILD (z) LOOK (z)(e) AG (anu)(e)] DP Anu
λ x.
′ T⟨e,t⟩ e z[ CHILD (z) LOOK (z)(e) AG ( g t 1 x (1))(e)]
µ1
′ T⟨t⟩ e z[ CHILD (z) LOOK (z)(e) AG ( g(1))(e)]
vP⟨t⟩ e z[ CHILD (z) LOOK (z)(e) AG ( g(1))(e)] P NI -DP 2 ⟨⟨e, ⟨v, t⟩⟩, ⟨v, t⟩⟩ λ Q λ e z[ CHILD (z) Q(z)(e)]
T
vP⟨t⟩ e z[ CHILD (z) LOOK (z)( e) AG ( g(1))( e)] NegP ⟨t,t⟩ p[ p]
vP⟨t⟩ e z[ CHILD (z) LOOK (z)(e) AG ( g(1))(e)] vP⟨v,t⟩ λ e z[ CHILD (z) LOOK (z)( e) AG ( g(1))( e)] ⟨DP1 ⟩
′ v ⟨e,⟨v,t⟩⟩ λ xλ e z[ CHILD (z) LOOK (z)(e) AG ( x)(e)]
VP ⟨v,t⟩ λ e z[ CHILD (z) LOOK (z)( e)]
v ⟨e,⟨v,t⟩⟩ λ xλ e[ AG ( x)(e)]
V ⟨e,⟨v,t⟩⟩ ⟨DP2 ⟩ λ xλ e[ LOOK ( x)(e)] ⟨⟨e, ⟨v, t⟩⟩, ⟨v, t⟩⟩ λ Q λ e z[ CHILD (z) Q(z)( e)]
Now let us look at a derivation which employs semantic reconstruction. (36) provides the LF tree for (29b) where PNI-DP leaves behind a copy which is interpreted as variable Q of type ⟨⟨e, ⟨v, t⟩⟩, ⟨v, t⟩⟩. In its landing site, PNI-DP serves as an argument for the λ-abstraction over Q.
128
PNI-PROPERT Y I: RESTRICTION TO LOW SCOPE
(36) ¬ > ∃ reading for (29b) via semantic reconstruction TP⟨t⟩ e[AG(anu)(e) z[CHILD(z) ∧ LOOK(z)(e)]] DP Anu
T′⟨e,t⟩ λx. e[AG(g t1 x (1))(e)] z[CHILD(z) ∧ LOOK(z)(e)]] μ1
T′⟨t⟩ e[AG(g(1))(e)] z[CHILD(z) ∧ LOOK(z)(e)]]
vP⟨t⟩ T e[AG(g(1))(e) λPλe′′ z[CHILD](z) P(z)(e′′)]( λyλe′[LOOK(y)(e′)])(e)]* PNI-DP2 ⟨⟨e, ⟨v,t⟩⟩, ⟨v,t⟩⟩ λPλe′′ z[CHILD(z) P(z)(e′′)]
vP⟨t⟩ λQ. e[AG(g(1))(e) (g 2 Q Q( λyλe′[LOOK(y)(e′)]))(e)] λQ
vP⟨t⟩ e[AG(g(1))(e)
Q( λyλe′[LOOK(y)(e′)])(e)] NegP⟨t,t⟩ p[p]
vP⟨t⟩ e[AG(g(1))(e)
Q( λyλe′[LOOK(y)(e′)])(e)] vP⟨v,t⟩
λe[AG(g(1))(e) Q( λyλe′[LOOK(y)(e′)])(e)] ⟨DP1⟩
v′⟨e,⟨v,t⟩⟩
λxλe[AG(x)(e) Q( λyλe′[LOOK(y)(e′)])(e)]
VP⟨v,t⟩
Q( λyλe′ [LOOK(y)(e′)])
v⟨e,⟨v,t⟩⟩
λxλe[AG(x)(e)]
V⟨e,⟨v,t⟩⟩ ⟨DP2⟩ ⟨⟨e, ⟨v,t⟩⟩, ⟨v,t⟩⟩ λyλe′[LOOK(y)(e′)] Q
(37) * = ¬∃e[AG(g(1))(e) ∧ λeʹʹ∃z[CHILD(z) ∧ λyλeʹ[LOOK(y)(eʹ)](z)(eʹʹ)](e)] = ¬∃e[AG(g(1))(e) ∧ λeʹʹ∃z[CHILD(z) ∧ LOOK(z)(eʹʹ)](e)] = ¬∃e[AG(g(1))(e) ∧ ∃z[CHILD(z) ∧ LOOK(z)(e)]]
5.4 CASE STUDIES
129
This section has shown that the inability to be interpreted outside vP does not prevent PNI-ed arguments from moving to such positions. Either semantic or syntactic reconstruction can provide interpretable LFs, albeit with the effect that PNI-ed arguments scope below operators merged outside vP.
5.4 Case studies We now move on to the case studies and provide contexts as well as LF trees for the relevant noun types in each language, respectively. I apply two scope tests: one involves universal quantification (= ∀∃) and the other negation (= ¬∃). The two scope tests were chosen based on comparability with previous PNI/DOM literature as well as cross-linguistic applicability. Although ∀∃ contexts are sometimes discussed to be unsuitable to test scopal properties, as the ∃∀ reading always entails the ∀∃ reading (Reinhart 1997, 2006), this caveat is widely ignored in the PNI/DOM literature (Dayal 2003, 2011; Frey 2015; Kelepir 2001; López 2012)— with the exception of Chung and Ladusaw (2004: 41) who allude to this problem and provide ¬∃ contexts instead. For this reason, I decided to test both ¬∃ and ∀∃ contexts, and it is crucial for the scope investigations in this study that both scope scenarios align accordingly. I provide an overview in Table 5.1, listing the noun types for which scopal properties can be diagnosed. The first two columns are the PNI contexts where the scopal properties show a clear restriction to low scope. Columns three and four list the scopal potential for case-marked nouns. While my proposal can derive the scope distribution in columns one, two, and four, it will have nothing enlightening to say about the cross-linguistic variation in column three. I assume the variation is due to individual differences, independent of PNI. Table 5.1 Scope restrictions of noun types which can undergo PNI
We already discussed the scope facts for Hindi and Spanish, data was taken from the indicated literature. The Turkish PNI scope facts are prominently discussed in the literature. I based my investigation on the scope judgements from Kelepir
130
PNI-PROPERT Y I: RESTRICTION TO LOW SCOPE
(2001) as well as Öztürk (2005), and extended the data when necessary (mostly for subject PNI). Mongolian PNI-ed objects have received some attention (Guntsetseg 2016), but mostly with respect to adverbial quantification. I contribute scope patterns of PNI-ed arguments in relation to negation and other quantifiers. Tamil and Korean scope interactions have only been addressed sporadically, I provide the first full-fledged study in this respect. A final note on the NP-category is in order. Turkish and Hindi shift case-marked bare nouns automatically to a definite reading. Thus, NP is irrelevant for the third and fourth column with respect to those languages. Since Hindi indefinites (ek NP) do not pseudo-incorporate, they are not listed as a possible noun type. Hence, the checkmarks in the third and fourth column for Hindi relate to the scopal properties of indefinites.
5.4.1 Tamil Recall from Section 4.3.1 that case drop on indefinites and numerals was analysed as an effect of DOM-marking, not PNI. I will repeat and extend the relevant scope contexts in this section to prove this point. Baker (2014b) provides one example where a caseless bare noun scopes under a negative operator; see (38). The illicit continuation ensures that the low scope reading is the only one available. (38) Baker (2014b: 18) Tamil Naan pustagam vanga-lle. (#Adu meese mele iru-kk-itu.) ¬∃,*∃¬ 1SG.NOM book buy-NEG it table on be-PRS-3SG.N ‘I didn’t buy (any) book. (# It is on the table.)’ He does not discuss indefinites but hints at them in a footnote (Baker 2014b: 9), suggesting that they can also undergo PNI. The following context provides counter-evidence against this claim. (39) sets up a scenario which makes a ¬∃ reading impossible. If the friend bought two books, it is not true that he did not buy any books. While it is not surprising that the case-marked object is licensed in an ∃¬ context, the felicity of the caseless object in (39a) is not predicted under a PNI account. Compare (39a) to (39c), a caseless bare noun which, in line with Baker’s observation, does not scope above negation, making its occurence infelicitious in a ∃¬ context. (39) Context ∃¬: A friend of mine wants to sell three books and after looking at them I bought two books from him. So there is a book which I did not buy from him.
5.4 CASE STUDIES
a.
Naan oru pustagam vaŋga-lle. 1SG.NOM one book buy-NEG ‘I didn’t buy a book.’
b.
Naan oru pustagath-ai vaŋga-lle. 1SG.NOM one book-ACC buy-NEG ‘I didn’t buy a book.’
131 Tamil
c. #Naan pustagam vaŋga-lle. 1SG.NOM book buy-NEG ‘I didn’t buy a book.’ Numerals > 1 pattern with oru (see (40)), suggesting that oru is in fact a numeral and not an indefinite determiner. The numeral rendu is licensed in a context which requires wide scope above negation, independent of case marking. Again, (40a) provides counter-evidence against the claim that case drop uniformly correlates with the inability to take wide scope. If the friend bought two books, it is not true that he did not buy two books. Since the ¬2 is excluded in (40), the licensed caseless object in (40a) must be able to take scope above negation. (40) Context 2¬: A friend of mine wants to sell four books and after looking at them I bought two books from him. So there are two books which I did not buy from him. a. Naan rendu pustagam vaŋga-lle. 1SG.NOM two book buy-NEG ‘I didn’t buy two books.’
Tamil
b. Naan rendu pustagath-ai vaŋga-lle. 1SG.NOM two book-ACC buy-NEG ‘I didn’t buy two books.’ The scope pattern is identical for weak quantifiers, shown in (41). (41) Context ∃¬: A friend of mine wants to sell some books and after looking at them I bought a few books from him. There are still some books left which I didn’t buy so you should have a look as well. a. Naan sila pustagam vaŋga-lle. 1SG.NOM some book buy-NEG ‘I didn’t buy some book.’ b. Naan sila pustagath-ai vaŋga-lle. 1SG.NOM some book-ACC buy-NEG ‘I didn’t buy some book.’
Tamil
132
PNI-PROPERT Y I: RESTRICTION TO LOW SCOPE
Recall from Section 4.2 that Dayal makes a similar observation for Hindi. I repeat the example here for convenience in (42). In (42b) the direct object is not case marked, yet it is able to scope above negation. (42) Dayal (2011: 128), Dayal (1999: 35) a. Anu ek bacce-ko nahiiN samhaalegii Anu one child-ACC not look.after.FUT ‘Anu will not look after a particular child.’
¬∃,∃¬
b. Anu ek/koii kitaab nahiiN paRhegii Anu one/some book not read.F ‘Anu won’t read any book./There is a book Anu won’t read’
¬∃,∃¬
The following discussion will focus on the status of the numerals ek/oru and ignore weak quantifiers for the most part. The numeral in (42b) also allows for scope under negation. This reading, however, is not readily available but must be triggered by narrow focus on the numeral. Dayal (2004) provides another context which supports that the low scope reading is not neutral but scalar. The question in (43) cannot be answered negatively with a numeral phrase since ∃¬ contradicts the negative response. Neither is the scalar ∃¬ reading licit since, according to Dayal, the question requires a neutral answer and does not provide a scalar context. (43) Dayal (2004: 418) A: Is there a mouse in the room? B: #NahiiN kamre meN ek cuuhaa nahii hai no room in one mouse not is ‘No, there’s one mouse which is not in the room.’ OR ‘No, not even one mouse is there in the room.’
Hindi ¬∃,∃¬
We can replicate the scope facts in Tamil. The scenario in (44) requires the object to scope under negation which is not possible for numerals per se, shown in (44a) and (44b), but either narrow focus on the numeral (44c) or a focus particle (44d) can trigger the scalar low scope reading (Sandhya Sundaresan, p.c.). Importantly, scope readings are not influenced by case marking for oru NP, i.e. (44a) does not enforce low scope because accusative case is not marked on the object. Contrast this with case drop on a bare noun (44e) which does license the object in a ¬∃ context. (44) Context ¬∃: A friend of mine wants to sell three books and after looking at them I didn’t buy any books from him.
5.4 CASE STUDIES
a. #Naan oru pustagam vaŋga-lle. 1SG.NOM one book buy-NEG ‘I didn’t buy a book.’
133 Tamil
b. #Naan oru pustagath-ai vaŋga-lle. 1SG.NOM one book-ACC buy-NEG ‘I didn’t buy a book.’ c.
Naan [oru]F pustagam vaŋga-lle. 1SG.NOM one book buy-NEG ‘I didn’t buy ONE book.’
d. Naan oru pustagam=um vaŋga-lle. 1SG.NOM one book=FOC buy-NEG ‘I didn’t even buy one book.’ e.
Naan pustagam vaŋga-lle. 1SG.NOM book buy-NEG ‘I didn’t buy any book.’
The scopal behaviour of ek NP/oru NP with respect to negation is not the only aspect in which Tamil and Hindi pattern alike. Another way to the test the potential of indefinite determiners is the acceptability in generic contexts (Chierchia 1998). English indefinite determiners create generic readings, in contrast to numerals and determiners like some, which have received a choice function treatment (Kratzer 1998). (45) Generic readings a. A dog barks.
English GENERIC
b. Some dog barks.
*GENERIC
c. One dog barks.
*GENERIC
Dayal (2004) applies this test to Hindi ek and shows that it does not allow a generic reading. (46) Dayal (2004: 418) Ek kutta bhaunktaa hai. one dog barks ‘One dog barks.’
Hindi *GENERIC
A similar pattern emerges with Tamil oru. The determiner blocks a generic reading, only a bare noun can trigger the generic reading.
134
PNI-PROPERT Y I: RESTRICTION TO LOW SCOPE
(47) Generic readings a. Oru vada-Indian naan-ai saapidu-v-aan. One North-Indian bread-ACC eat-FUT-3SG.M ‘One North-Indian eats bread.’ b. Vada-Indian naan-ai saapidu-v-aan. North-Indian bread-ACC eat-FUT-3SG.M ‘North-Indians eat bread.’
Tamil *GENERIC
GENERIC
We can draw two conclusions from this discussion: (i) Hindi ek and Tamil oru instantiate numerals, not indefinite determiners, and (ii) since Hindi ek and Tamil oru can occur without case marking while taking wide scope, case loss is not due to PNI, but an effect of DOM. Numeral determiners can be given a denotation along the lines of (48a). I repeat the denotation of the PNI-determiner for comparison’s sake. (48) a. ⟦Done ⟧ = λP⟨e,t⟩ λQ⟨e,t⟩ ∃z[P(z) ∧ |z| = 1 ∧ Q(z)] b. ⟦DPNI ⟧ = λP⟨e,t⟩ λQ⟨e,⟨v,t⟩⟩ λe∃z[P(z) ∧ Q(z)(e)]
ek/oru PNI
Since both Done and DPNI turn a property into an argument, they can never cooccur. Hence, bare arguments without case marking will be headed by DPNI , thereby triggering a low scope restriction. Case can also be dropped on arguments headed by ek/oru but this will be due to DOM which does not entail a low scope restriction. Case drop for ek/oru NPs can be related to DOM-aspects such as animacy and specificity, which has been argued to be relevant for ek NPs in Hindi (Bhatt and Anagnostopoulou 1996; Bhatt 2007). As was pointed out in Chapter 4, animacy and specificity also plays a role in Tamil. Before we move on to case-marked bare nouns, let us first derive the wide scope reading of oru NPs in (39a) and (39b). The LF tree in (49)⁴ presents the scopal ⁴ The subject in (39a) and (39b) is a 1PS pronoun, referring to the speaker of the utterance. I take free pronouns to denote indices, i.e. variables which are mapped to a semantic value by an assignment function g. Following (Sauerland 2003, 2008b) and Heim (2008), ϕ-features on pronouns denote partial indentity functions of type ⟨e,e⟩. For free pronouns the relevant assignment is given by the utterance context. Below I provide an overview of the person feature system and how it translates into semantic denotations. Presuppositions are omitted in LF trees which include 1PS/2PS pronouns. (i) a. 1st = [+SPEAKER,−HEARER] c
b. ⟦1st ⟧ = λxe : x includes sc [x] (ii) a. 2nd = [−SPEAKER,+HEARER] c
b. ⟦2nd ⟧ = λxe : x includes hc [x] (iii) a. 3rd = [−SPEAKER,−HEARER] b. ⟦3rd ⟧ = λxe [x]
5.4 CASE STUDIES
135
relations of oru NPs in (39a) as well as (39b), i.e. with and without case marking. The LF tree also predicts that (44a) and (44b) are not licensed since the context enforces a ¬∃ reading. There is an interaction with focus that seems to enable a ¬∃ reading after all, shown in (44c) and (44d), which can be derived by letting the object QR immediately below NegP but above existential closure. The interactions of information structure and scope, however, are not investigated any further in this book and therefore have to be left open for now. (49) ∃ > ¬ reading for (39a) and (39b) ∃x[
λP∃x[
NegP⟨t⟩ (x) ∧ x =1∧ ¬∃e[ ∧ (g(3))(e)]]
(x)(e)
NegP⟨e,t⟩ λ x ¬∃e[ (g t2↦ x(2))(e) ∧ (g(3))(e)]
DP2 ⟨⟨e,t⟩,t⟩ (x)∧ x = 1∧P(x)] μ2
NegP⟨t⟩ ¬∃e[ (g(2))(e) ∧ (g(3))(e)] Neg⟨t,t⟩ ¬p[p]
vP⟨t⟩ ∃e[ (g(2))(e) ∧ (g(3))(e)] ∃
vP⟨v,t⟩ λe[ (g(2))(e) ∧ (g(3))(e)] DP g(3)
v′⟨e,⟨v,t⟩⟩ λxλe[ (g(2))(e) ∧ (x)(e)] VP⟨v,t⟩ λe[ ⟨DP2⟩
(g(2))(e)]
v⟨e,⟨v,t⟩⟩ λxλe[ (x)(e)]
V⟨e,⟨v,t⟩⟩ λxλe[ (x)(e)]
Caseless bare nouns receive their low scope reading from the requirement of PNIDPs to be interpreted within the event domain. The LF structure in (50) shows why (39c) is not licensed in an ∃¬ context. (50) *∃ > ¬ reading for (39c) [PNI-DP2⟨⟨e,⟨v,t⟩⟩,⟨v,t⟩⟩ type clash [μ2 ... ¬ ∃ [vP DP [VP ⟨DP2 ⟩ buy]]]⟨e,t⟩ ]
136
PNI-PROPERT Y I: RESTRICTION TO LOW SCOPE
Caseless bare nouns are licensed in a ¬∃ context such as the one in (44). The LF structure in (51) provides the structure for (44e). (51) ¬ > ∃ reading for (44e) [¬ ∃ [vP DP [VP PNI-DP⟨⟨e,⟨v,t⟩⟩,⟨v,t⟩⟩ buy⟨e,⟨v,t⟩⟩ ]]] Since we now established that oru is a numeral determiner, we might ask ourselves if there is an additional existential D head available—in other words a true indefinite determiner that requires case marking and is able to let the argument scope above (or below) negation. At first glance, this does not seem to be true. Using the same ∃¬ context as above, we see that a case-marked bare noun is also dispreferred. (52) Context ∃¬: A friend of mine wants to sell three books and after looking at them I bought two books from him. So there is one book which I did not buy from him. a. #Naan pustagam vaŋga-lle. 1SG.NOM book buy-NEG ‘I didn’t buy a book.’
Tamil
b. ??Naan pustagath-ai vaŋga-lle. 1SG.NOM book-ACC buy-NEG ‘I didn’t buy a book.’
One way to explain the infelicity of (52b) is to essentially follow Dayal’s (2004; 2017) work on Hindi where she shows that the existential interpretation of bare nouns can only arise via pseudo-noun incorporation or a derived kind interpretation (Chierchia 1998), otherwise bare nouns shift to definites or kinds, one instance of this was already shown in (15b). Translating Dayal’s assumptions into the current system would amount to denying the existence of D∃ , defined in (16b). The infelicity of (52b) would now be explained in the following way. PNI is not an option for (52b) since the bare noun is case marked. A derived kind interpretation can only be created of plural-marked arguments. Thus, the only possibility left for the object in (52b) is to shift to a definite interpretation via wDι or sDι . Neither strongly definite nor weakly definite objects are licensed by the context in (52) since uniqueness is not established. In the following, I will demonstrate that this reasoning cannot be extended to Tamil bare nouns. The first reason to doubt the inexistence of D∃ is given with a slightly modified ∃¬ context in (53). The use of the case-marked bare noun in (53b) is judged slightly better than in (52b). Still the best way to express the scope relations is the use of an overt quantifier; see (53c). Similarly, the use of the numeral oru in a context like (52) is the optimal way to express the scope relations.
5.4 CASE STUDIES
137
(53) Context ∃¬: A friend of mine wants to sell some books and after looking at them I bought a few books from him. There are still some books left which I didn’t buy so you should have a look as well. a.
#Naan pustagam vaŋga-lle. 1SG.NOM book buy-NEG ‘I didn’t buy a book/books.’
Tamil
b. ?% Naan pustagath-ai vaŋga-lle. 1SG.NOM book-ACC buy-NEG ‘I didn’t buy a book/books.’ c.
Naan sila pustagath-ai vaŋga-lle. 1SG.NOM some book-ACC buy-NEG ‘I didn’t buy some book/books.’
It seems like the indefinite reading is available for case-marked bare nouns, but its application is often blocked by an overt exponent, either the numeral oru or another quantifier. The numeral oru is more informative in a context like (52); thus it is preferred. For the context in (53) the situation is less clear. While the overt quantifier is still preferred, it is certainly not the only option, at least for some speakers. We will now look at four environments which do not license definite interpretations of objects but readily allow for indefinite interpretations. Dayal has shown for each of these contexts that Hindi does not allow bare nouns to occur, numeral phrases have to be used instead. In contrast, Tamil allows for bare nouns as well as oru NP in at least three of the four environments. This provides further evidence for the existence of D∃ . The first test concerns homogeneity which is based on the rationale that a unique individual cannot be ascribed contradicting properties, thereby excluding definite readings in such contexts. (54) Homogeneity
English
a. A dog is sleeping and a dog is barking. b. * The dog is sleeping and the dog is barking. Hindi bare nouns are not allowed in those contexts. Dayal concludes that this is due to bare nouns automatically shifting to definite interpretations. (55) Dayal (2017: 88) *Kutta so-rahaa-hai aur kutta bhaunk-rahaa-hai. dog sleep-PROG-PRS and dog bark-PROG-PRS ‘The dog is sleeping and the dog is barking.’
Hindi
138
PNI-PROPERT Y I: RESTRICTION TO LOW SCOPE
Tamil patterns with Hindi in this respect. The use of a bare noun is illicit, numerals have to occur obligatorily.⁵ (56) Context: You enter a room and you see a dog sleeping and a dog barking. Now you tell me what you see. a. #Naai tuuŋgu-tu naai kolaikku-tu. dog.NOM sleep-3SG.N dog.NOM bark-3SG.N ‘A dog is sleeping and a dog is barking.’ b.
Tamil
Oru naai tuuŋgu-tu oru naai kolaikku-tu. one dog.NOM sleep-3SG.N one dog.NOM bark-3SG.N ‘A dog is sleeping and a dog is barking.’
The next environment excludes the use of definites as referring to a part of a previously introduced set. (57) Partitivity
English
a. There were several kids in the room... I talked to a child. b. There were several kids in the room... #I talked to the child. Hindi bare nouns cannot be used in a partitive context such as the one above. (58) Dayal (2017: 88) Hindi kamre meN kaii bacce the. #laRkaa aur laRkii taash khel-rahe-the. room in several kids were boy and girl cards play-PROG-PST ‘There were several children in the room. #The boy and the girl were playing cards.’ Tamil, however, allows for the use of bare nouns, alongside numeral constructions. (59) Context: You entered a room and you saw several kids in the room. One kid is behaving badly so you scolded it. So you say... a. Oru kulantai-ye titi-n-een. one child-ACC scold-PST-1SG ‘I scolded a child.’
Tamil
b. Kulantai-ye titi-n-een. child-ACC scold-PST-1SG ‘I scolded a child.’ ⁵ For some reason, Dayal (2004, 2017) does not provide an example with numerals but judging from her theory their use should be licensed. The same is true for partitive scenarios.
5.4 CASE STUDIES
139
The third environment are contexts which force objects to be discourse new, e.g. at the beginning of a story. Since definites often come with a familiarity requirement, they do not serve the purpose of introducing new discourse referents. (60) Discourse new
English
a. (Beginning of a story) Many years ago... An old dog died in this house. b. (Beginning of a story) Many years ago...#The old dog died in this house. Hindi requires discourse referents to be introduced with the numeral ek. Bare nouns cannot fulfill this function. (61) Dayal (2004: 409) Hindi Bahut saal pahle, yehaaN *(ek) aurat rahtii thii. aurat bahut many years ago here one woman lived woman very brave bhadur thi... was ‘Once upon a time, a woman used to live here. The woman was very brave.’ Again, Tamil provides a different picture. Bare nouns as well as numeral phrases are able to introduce new discourse referents. (62) Context: (Beginning of a story) Many years ago.../Romba varusatukku munadi... a. inta vitt-le oru naai sethu po-cc-u. this house-LOC one dog.NOM die go-PST-3SG.N ‘A dog died in this house.’
Tamil
b. inta vitt-le naai sethu po-cc-u. this house-LOC dog.NOM die go-PST-3SG.N ‘A dog died in this house.’ Finally, the fourth environment concerns existential sentences which do not allow definite phrases to occur (Milsark 1977). (63) Existential constructions a. There was some beer/beer at the party. b. *There was the beer at the party. Hindi does not allow bare nouns to occur in an existential sentence.
English
140
PNI-PROPERT Y I: RESTRICTION TO LOW SCOPE
(64) Dayal (2017: 92) kamre meN ?? (ek) buuDhii aurat hai. room in one old woman is ‘There’s an old woman in the room.’
Hindi
Tamil freely allows for bare nouns as well as numerals in existential sentences. (65) a. Parti-le kulanta iru-nt-atu. party-LOC child.NOM be-PST-3SG.N ‘There was a child at the party.’
Tamil
b. Parti-le oru kulanta iru-nt-atu. party-LOC one child.NOM be-PST-3SG.N ‘There was one child at the party.’ This small survey casts reasonable doubt on the hypothesis that case-marked bare nouns in Tamil obligatorily lead to a definite interpretation. Tamil differs significantly from Hindi in that it allows bare nouns to occur freely in environments that favour an indefinite but exclude a definite interpretation. The existential determiners available in Tamil are summarized again in (66). Whereas (66a) is spelled out as oru, both (66b) and (66c) are realized by the elsewhere zero exponent. (66b) leads to case drop due to its hybrid categorial nature. (66c), however, is a traditional D head which requires case marking. (66) Existential determiners in Tamil a. ⟦Done ⟧ = λP⟨e,t⟩ λQ⟨e,t⟩ ∃z[P(z) ∧ |z| = 1 ∧ Q(z)] b. ⟦DPNI ⟧ = λP⟨e,t⟩ λQ⟨e,⟨v,t⟩⟩ λe∃z[P(z) ∧ Q(z)(e)] c. ⟦D∃ ⟧ = λP⟨e,t⟩ λQ⟨e,t⟩ ∃z[P(z) ∧ Q(z)] The LF tree for (52b)/(53b), shown in (67), is almost identical to the LF tree in (49), the only difference being that D∃ contributes slightly less information, i.e. it is not explicit about the number of elements in the restrictor set. Precisely for this reason, I speculate, (52b) is dispreferred by my informants. An ∃¬ context which explicitly excludes the possibility of the statement not being true of more than one entity strongly prefers the use of the numeral one.
5.4 CASE STUDIES
141
(67) ∃ > ¬ reading for (52b)/(53b) NegP⟨t⟩ ∃x[BOOK(x)∧ ¬∃e[BUY(x)(e) ∧AG(g(3))(e)]] NegP⟨e,t⟩ λ x ¬∃e[BUY( g t2↦x(2))(e) ∧AG(g (3))(e)]
DP2 ⟨⟨e,t⟩,t⟩ λ P∃x[BOOK(x)∧P(x)] µ2
NegP⟨t⟩ ¬∃e[BUY(g(2))(e) ∧AG(g(3))(e)] Neg⟨t,t⟩ ¬p[p]
vP⟨t⟩ ∃e[BUY(g(2))(e) ∧AG(g(3))(e)] ∃
vP⟨v,t⟩ λe[BUY(g(2))(e) ∧AG(g(3))(e)] v′⟨e,⟨v,t⟩⟩ λ xλe[BUY(g(2))(e) ∧AG(x)(e)]
DP g(3)
VP⟨v,t⟩ λe[BUY( g(2))(e)] ⟨DP2⟩
v⟨e,⟨v,t⟩⟩ λxλe[AG(x)(e)]
V⟨e,⟨v,t⟩⟩ λxλe[BUY(x)(e)]
We will now turn to the final piece of evidence for the existence of D∃ in Tamil. It comes from scope contexts including a universally quantified subject. Casemarked bare objects scope above subjects, while caseless objects show the by now familiar low scope restriction. Surface scope is enforced with the context in (68), inverse scope is required in (69). (68) Context ∀∃: Imagine there are three students. The first student read Pride and Prejudice, the second one as well. The third student read Ramayanam. a.
Ella students-um pustagam padi-c-aaŋga. all students.NOM-ADD book read-PST-3PL ‘A ll students read a book.’
Tamil
142
PNI-PROPERT Y I: RESTRICTION TO LOW SCOPE b. #Ella students-um pustagath-ai padi-c-aaŋga. all students.NOM-ADD book-ACC read-PST-3PL ‘A ll students read a book.’
(69) Context ∃∀: Imagine there are three students. There is one book that all three students read. It is Ramayanam. a. #Ella students-um pustagam padi-c-aaŋga. all students.NOM-ADD book read-PST-3PL ‘A ll students read a book.’ b.
Tamil
Ella students-um pustagath-ai padi-c-aaŋga. all students.NOM-ADD book-ACC read-PST-3PL ‘A ll students read a book.’
In contrast to the negation contexts discussed above, (69b) shows that casemarked bare nouns can take wide scope more readily with respect to a universally quantified subject. This is probably due to the fact that the context does not exclude a reading where each student read more than one book. Specificity might also boost the wide scope existential reading in (69b). I give the LF structure for (69b) in (70). Why case-marked objects trigger inverse scope readings but exclude the surface scope reading must be due to some independent factor. (70) ∃ > ∀ reading of (69b) [DP2⟨⟨e,t⟩,t⟩ [μ2 ... DP1⟨⟨e,t⟩,t⟩ [μ1 ... ∃ [vP ⟨DP1 ⟩ [VP ⟨DP2 ⟩ read]]]⟨e,t⟩ ]⟨e,t⟩ ] Caseless bare noun behaves as predicted. The LF structures, given in (71) and (72), show that the PNI determiner leads to an in situ interpretation. (71) ∀ > ∃ reading of (68a) [DP1⟨⟨e,t⟩,t⟩ [μ1 ... ∃ [vP ⟨DP1 ⟩ [VP PNI-DP⟨⟨e,⟨v,t⟩⟩,⟨v,t⟩⟩ read⟨e,⟨v,t⟩⟩ ]]]⟨e,t⟩ ]
(72) *∃ > ∀ reading of (69a) [PNI-DP2⟨⟨e,⟨v,t⟩⟩,⟨v,t⟩⟩ t.cl.[μ2 ..DP1⟨⟨e,t⟩,t⟩ [μ1 ..∃[vP ⟨DP1 ⟩[VP ⟨DP2 ⟩ read]]]⟨e,t⟩ ]⟨e,t⟩ ] If we exchange bare nouns with numeral phrases in the contexts above, no correlation between case marking and scope sensitivity can be detected, shown (73) and (74). This is in line with our assumption that numerals do not undergo PNI, case loss is simply a DOM-effect.⁶ ⁶ One speaker finds (74a) only marginally acceptable. Again, we can make sense of this judgement if the speaker interprets the object as necessarily specific, as the context marginally suggests.
5.4 CASE STUDIES
143
(73) Context ∀∃: Imagine there are three students. The first student read Pride and Prejudice, the second one as well. The third student read Ramayanam. a. Ella students-um oru pustagam padi-c-aaŋga. all students.NOM-ADD one book read-PST-3PL ‘A ll students read one book.’
Tamil
b. Ella students-um oru pustagath-ai padi-c-aaŋga. all students.NOM-ADD one book-ACC read-PST-3PL ‘A ll students read one book.’ (74) Context ∃∀: Imagine there are three students. There is one book that all three students read. It is Ramayanam. a. Ella students-um oru pustagam padi-c-aaŋga. all students.NOM-ADD one book read-PST-3PL ‘A ll students read one book.’
Tamil
b. Ella students-um oru pustagath-ai padi-c-aaŋga. all students.NOM-ADD one book-ACC read-PST-3PL ‘A ll students read one book.’ LF structures for the numeral constructions are given in (75) and (76), where (75) is identical to (70) since numeral quantifiers only differ from existential quantifiers in terms of an additional numerical restriction. In order to derive the ∀ > ∃ reading, quantifiers raise in opposite order in (76). (75) ∃ > ∀ reading of (74a) and (74b) [DP2⟨⟨e,t⟩,t⟩ [μ2 ... DP1⟨⟨e,t⟩,t⟩ [μ1 ... ∃ [vP ⟨DP1 ⟩ [VP ⟨DP2 ⟩ read]]]⟨e,t⟩ ]⟨e,t⟩ ] (76) ∀ > ∃ reading of (73a) and (73b) [DP1⟨⟨e,t⟩,t⟩ [μ1 ... DP2⟨⟨e,t⟩,t⟩ [μ2 ... ∃ [vP ⟨DP1 ⟩ [VP ⟨DP2 ⟩ read]]]⟨e,t⟩ ]⟨e,t⟩ ] This concludes our discussion about the scopal properties of bare nouns and numeral determiners in Tamil (and Hindi). The main conclusion we can draw is that Tamil (and quite possibly also Hindi) make a difference between bare nouns and numerals, in that only the former exhibits a correlation between case marking and scope. Tamil is additionally different from Hindi in that it seems to create indefinite readings with case-marked bare nouns.⁷ We will now move on to Mongolian and Turkish, two languages for which numerals, weak quantifiers as well as bare nouns undergo PNI. ⁷ Weak quantifiers have been given less attention in this section. They pattern with numerals and not with bare nouns.
144
PNI-PROPERT Y I: RESTRICTION TO LOW SCOPE
5.4.2 Mongolian Let us start with quantifier scope relations. Mongolian bare nouns behave identically to Tamil and Hindi bare nouns, they cannot take scope over a universally quantified subject. They are felicitous in a ∀∃ context, given in (77), but not in a ∃∀ context, provided in (78). In contrast, case-marked bare nouns trigger inverse scope since they are licensed in (78) but not in (77). (77) Context ∀∃: There are three students Bold, Tujaa, and Martin. Bold will read Pride and Prejudice, Martin will read Sherlock Holmes, and Tujaa will read The Secret History of the Mongols. a.
Ojuutan bu¨r nom unshi-na. student.NOM each book read-FUT ‘Each student will do book-reading.’
Mongolian
b. #Ojuutan bu¨r nom-yg unshi-na. student.NOM each book-ACC read-FUT ‘Each student will read a book.’ (78) Context ∃∀: There are three students Bold, Tujaa, and Martin. There is one book every student will read: The Secret History of the Mongols. a. #Ojuutan bu¨r nom unshi-na. student.NOM each book read-FUT ‘Each student will do book-reading.’ b.
Mongolian
Ojuutan bu¨r nom-yg unshi-na. student.NOM each book-ACC read-FUT ‘Each student will read a book.’
I sketch the relevant structures for the possible and impossible reading in (79) and (80), respectively. Case drop signals the presence of PNI-DP, hence caseless bare nouns can only be interpreted in the event domain. These LF structures are completely parallel to the GQ scenarios in Tamil. (79) ∀ > ∃ reading of (77a) [DP1⟨⟨e,t⟩,t⟩ [μ1 ... ∃ [vP ⟨DP1 ⟩ [VP PNI-DP⟨⟨e,⟨v,t⟩⟩,⟨v,t⟩⟩ read⟨e,⟨v,t⟩⟩ ]]]⟨e,t⟩ ]
(80) *∃ > ∀ reading of (78a) [PNI-DP2⟨⟨e,⟨v,t⟩⟩,⟨v,t⟩⟩ t.cl.[μ2 ..DP1⟨⟨e,t⟩,t⟩ [μ1 ..∃[vP ⟨DP1⟩[VP ⟨DP2 ⟩read]]]⟨e,t⟩ ]⟨e,t⟩ ]
5.4 CASE STUDIES
145
The scope data are in line with what Guntsetseg (2016) reports for caseless bare nouns and their relation to the scope of universally quantified adjectives; see (81). (81) Guntsetseg (2016: 63) Tujaa o¨do¨r bolgon nom unsˇ-dag. Tujaa day every book red-HAB ‘Tujaa does book-reading every day.’
Mongolian ∀ > ∃,*∃ > ∀
The wide scope reading for case-marked bare nouns in (78b) is given in (82), which presuppose the existence of a zero D∃ head so that both the existential and the universal quantifier must leave the event domain in order to be interpreted. We provide evidence for D∃ further below. Again, as in Tamil, the low scope reading is not licensed, which must be caused by an independent factor. (82) ∃ > ∀ reading of (78b) [DP2⟨⟨e,t⟩,t⟩ [μ2 ... DP1⟨⟨e,t⟩,t⟩ [μ1 ... ∃ [vP ⟨DP1 ⟩ [VP ⟨DP2 ⟩ read]]]⟨e,t⟩ ]⟨e,t⟩ ] Moving on to indefinite objects, we can now observe a clear contrast between Mongolian on the one hand and Hindi/Tamil on the other. Indefinites pattern with bare nouns: Case drop leads to scope below the universally quantified subject. Thus, the caseless indefinite is infelicitous in an ∃∀ context (see (84b)), but perfectly acceptable in a ∀∃ context; see (83b). The case-marked indefinite, in contrast, needs an ∃∀ context, in order to be licensed, compare (83a) to (84b). (83) Context ∀∃: There are three students Bold, Tujaa, and Martin. Bold will criticize his philosophy professor, Martin will criticize his English professor, and Tujaa will criticize her history professor. a. #Ojuutan bu¨r neg professor-yg ˇsu¨u¨mzˇle-ne. student.NOM each a professor-ACC critisize-FUT ‘Each student will criticize a professor.’ b.
Mongolian
Ojuutan bu¨r neg professor ˇsu¨u¨mzˇle-ne. student.NOM each a professor critisize-FUT ‘Each student will criticize a professor.’
(84) Context ∃∀: There are three students Bold, Tujaa, and Martin. There is one professor every student will criticize since he is very demanding. a.
Ojuutan bu¨r neg professor-yg ˇsu¨u¨mzˇle-ne. student.NOM each a professor-ACC critisize-FUT ‘Each student will criticize a professor.’
Mongolian
146
PNI-PROPERT Y I: RESTRICTION TO LOW SCOPE b. #Ojuutan bu¨r neg professor ˇsu¨u¨mzˇle-ne. student.NOM each a professor critisize-FUT ‘Each student will criticize a professor.’
Upon closer inspection, neg NP seems to behave like a numeral rather than an indefinite. Similar to Tamil and Hindi, generic readings require the use of of a bare noun and block the occurence of neg. This can be tested easily with the use of habitual mood, marked overtly on the verb, which enforces generic readings, shown in (85). Following Chierchia (1998) and Dayal (2004), we can tentatively conclude that neg should be treated like a numeral. (85) Generic readings a. Neg noxoj xuc-dag. one dog.NOM bark-HAB ‘One dog barks.’ b. Noxoj xuc-dag. dog.NOM bark-HAB ‘A dog barks.’
Mongolian *GENERIC
GENERIC
In contrast to Tamil oru and Hindi ek, however, Mongolian neg does not act as a D head, as it is not in complementary distribution with PNI-D, the hybrid determiner. We have to ensure that neg NPs can serve as arguments to PNID, i.e. we need to make sure that neg NPs denote properties. I assume that numerals (as well as weak quantifiers) denote one-place predicates which modify the noun via Predicate Modification (Heim and Kratzer 1998: 65). (86b) provides the denotation, I also repeat the denotation of numeral determiners in (86a) for comparison’s sake. The adjectival nature has been pointed before, for numerals specifically (Link 1987; Krifka 1999; Ionin and Matushansky 2006) as well as for weak quantifers in general (Kamp and Reyle 1993; Milsark 1977; Partee 1988; Zimmermann 2008). The cross-linguistic comparison between Tamil and Hindi on the one hand and Mongolian (as well as Turkish) on the other reveals that languages can choose whether they encode the strong weak contrast syntactically and semantically or whether they treat quantifiers uniformly. (86) a. ⟦ek/oru⟧ = λP⟨e,t⟩ λQ⟨e,t⟩ ∃z[P(z) ∧ |z| = 1 ∧ Q(z)] b. ⟦neg⟧ = λz.|z| = 1
Hindi/Tamil Mongolian
Adjectival numerals attach to NP which in turn can serve as an argument to either DPNI or D∃ , where the former leads to case drop but the latter does not. The nominal-internal derivations are given in (87) and (88).
5.4 CASE STUDIES
147
(87) neg professor (83b)/(84b): PNI-DP ⟨⟨e,⟨v,t⟩⟩,⟨v,t⟩⟩ λQλe∃z[PROF(z)∧ z =1∧Q(z)(e)] NP ⟨e,t⟩ λx[PROF(x)∧ x =1]
PNI-D ⟨⟨e,t⟩,⟨⟨e,⟨v,t⟩⟩,⟨v,t⟩⟩⟩ λPλQλe∃z[P(z)∧Q(z)(e)]
NP⟨e,t⟩ DP⟨e,t⟩ = neg λz. z =1 λy[PROF(y)]
(88) neg professor-yg (83a)/(84a): DP ⟨⟨e,t⟩,t⟩ λQ∃z[PROF(z)∧ z =1∧Q(z)] D∃ ⟨⟨e,t⟩,⟨⟨e,t⟩,t⟩⟩ λPλQ∃z[P(z)∧Q(z)]
NP ⟨e,t⟩ λx[PROF(x)∧ x =1] NP⟨e,t⟩ DP⟨e,t⟩ = neg λy[PROF(y)] λz. z =1
We need D∃ to exist independently since this D head predicts case-marked bare nouns to be felicitous in indefinite contexts, recall (78b). Further evidence is provided by the occurrence of bare nouns in existential sentences; see (89). (89) Janhunen (2012: 233) En-d nom bai-n’. here-LOC book be-DUR ‘There is a book here.’
Mongolian
Existential sentences can also be formed with numeral phrases, shown in (90). (90) Janhunen (2012: 232) Mongolian Baroon our-d tal-d neg-e.n xun couloon doursgel bii. right front-DAT side-DAT one-ATTR man stone monument EXIST ‘On the southwest side there is an ancient stone statue.’ Note that the numeral is realized as a nasal variant here—an interesting morphosyntactic side effect which supports the assumption that numerals are adjectival. Janhunen (2012: 111–112) identifies two stems in Mongolian, the plain one and the (instable) nasal stem. While the nasal stem is unattested in accusative case, in
148
PNI-PROPERT Y I: RESTRICTION TO LOW SCOPE
nominative case the nasal is used to signal an attributive function, along the lines of (91a). The nasal variant can also be lexicalized, deriving a semantically related adjective; see (91b). Since the subject of the existential clause in (90) bears nominative case (realized as a zero exponent), I take the nasal ending of the numeral as overt evidence for its adjectival position. (91) Janhunen (2012: 111) a. mory / mory-e.n tereg / mory-e.n jil horse horse-ATTR cart horse-ATTR year ‘horse / horse cart / year of the horse’
Mongolian
b. nogao / nogaon vegetable green ‘vegetable / green’ The stem alternation can also be observed in another prominently discussed indefinite context. Guntsetseg (2016) reports that numeral constructions can be used to introduce new discourse referents. (92) Guntsetseg (2016: 44) Ertee ur’d-yun cag-t neg xo¨vgu¨u¨n baj-zˇee. earlier previous-GEN time-DAT a boy be-PRF ‘Once upon a time, there was a boy.’
Mongolian
The best way to introduce a new discourse referent, however, is to make use of the nasal variant of the numeral (Dolgor Guntsetseg, p.c.). This provides further evidence for the adjectival status of numerals.⁸ (93) Dolgor Guntsetseg, p.c. Ertee ur’d-yun cag-t neg-e.n xo¨vgu¨u¨n baj-zˇee. Mongolian earlier previous-GEN time-DAT one-ATTR boy be-PERF ‘Once upon a time, there was a boy.’ I will now provide the LF trees for the numeral constructions in (83) and (84). Since numerals are treated as adjectives in Mongolian, case loss signals the presence of the hybrid determiner PNI-D, whereas case marking implies that numerals phrases are headed by D∃ . The derivations for the caseless variants are given in (94) and (95). ⁸ Although there is disagreement amongst my informants, there seems to be a tendency towards bare nouns being dispreferred in discourse new contexts. Here Mongolian diverges from Tamil, in that bare nouns do not as easily take over the function of an indefinite. (i) ?Ertee ur’d-yun cag-t xo¨vgu¨u¨n baj-zˇee earlier previous-GEN time-DAT boy be-PERF ‘Once upon a time, there was a boy.’
Mongolian
5.4 CASE STUDIES
149
(94) ∀ > ∃ reading of (83b) vP⟨t⟩ y[ STUD (y) → e z[ PROF (z) AG (y)(e)]] DP 1 ⟨⟨e, t⟩, t⟩ λ Q y[ STUD (y) → Q(y)]
CRIT (z)(e)
vP⟨e,t⟩ λ x e z[ PROF (z) CRIT (z)(e) AG ( g t 1 x (1))(e)] μ1
vP⟨t⟩ e z[ PROF (z) CRIT (z)(e) AG ( g(1))(e)] vP⟨v,t⟩ λ e z[ PROF (z) CRIT (z)(e) AG ( g(1))(e)] v ′⟨e,⟨v,t⟩⟩ λ xλ e z[ PROF (z) CRIT (z)(e) AG (x)(e)]
⟨DP1 ⟩
VP ⟨v,t⟩ λ e z[ PROF (z) CRIT (z)(e)] P NI -DP ⟨⟨e, ⟨v, t⟩⟩, ⟨v, t⟩⟩ λ Q λ e z[ PROF (z) z = 1 Q(z)(e)] P NI -D ⟨⟨e, t⟩, ⟨⟨e, ⟨v, t⟩⟩, ⟨v, t⟩⟩⟩ λ Pλ Q λ e z[P(z) Q(z)(e)]
v ⟨e,⟨v,t⟩⟩ λ xλ e[ AG ( x)(e)]
V ⟨e,⟨v,t⟩⟩ λ xλ e[ CRIT ( x)(e)]
NP ⟨e, t⟩ λ x[ PROF ( x) x = 1] NP ⟨e,t⟩ DP ⟨e,t⟩ λ z. z = 1 λ y[ PROF (y)]
(95) *∃ > ∀ reading of (84b) [PNI-DP2⟨⟨e,⟨v,t⟩⟩,⟨v,t⟩⟩ t.cl. [μ2 ..DP1⟨⟨e,t⟩,t⟩ [μ1 ..∃ [vP ⟨DP1 ⟩[VP ⟨DP2 ⟩ criticize]]]⟨e,t⟩ ]⟨e,t⟩ ] The case-marked variant provides the wide scope reading due to the presence of D∃ , a generalized quantifier which needs to be interpreted outside the event domain, shown in (96). As with the case-marked bare nouns, a low scope reading is excluded with case-marked numerals. (96) ∃ > ∀ reading of (84a) [DP2⟨⟨e,t⟩,t⟩ [μ2 ... DP1⟨⟨e,t⟩,t⟩ [μ1 ... ∃ [vP ⟨DP1 ⟩ [VP ⟨DP2 ⟩ criticize]]]⟨e,t⟩ ]⟨e,t⟩ ]
150
PNI-PROPERT Y I: RESTRICTION TO LOW SCOPE
Before we move on to negation contexts, let me briefly point out an additional scope scenario from Guntsetseg (2016), involving an additional adverbial quantifier. (97) Guntsetseg (2016: 95–96)
Mongolian
a. Ojuutan bu¨r 3 udaa neg professor ˇsu¨u¨mzˇle-ne. student each 3 time a professor critisize-FUT ‘Each student will criticize a professor three times.’
∀ > 3t > ∃
b. Ojuutan bu¨r 3 udaa neg professor-yg ˇsu¨u¨mzˇle-ne. student each 3 time a professor-ACC critisize-FUT ‘Each student will criticize a professor three times.’
∀ > ∃ > 3t
c. Ojuutan bu¨r neg professor-yg 3 udaa ˇsu¨u¨mzˇle-ne. student each a professor-ACC 3 time critisize-FUT ‘Each student will criticize a professor three times.’
{∀, ∃} > 3t
d. Neg professor-yg ojuutan bu¨r 3 udaa ˇsu¨u¨mzˇle-ne. a professor-ACC student each 3 time critisize-FUT ‘Each student will criticize a professor three times.’
∃ > ∀ > 3t
The paradigm in (97) is in line with our current assumptions. In (97a) the caseless object shows the familiar low scope restriction. Once it is case marked, it can interact with the other quantifiers in (97b)–(97d). Following Swart (1993) and Fintel (1994), adverbials like 3 times quantify over sets of events, where times constitutes an abstract restrictor argument of type ⟨v, t⟩, possibly the spell-out of a contextual domain variable. The adverbial attaches to the vP, overtaking the function of existential closure of the event variable, the scope scenarios are abstractly shown in (98). (98) Another scope scenario: ⟨t⟩
⟨t⟩
PNI
GQ
⟨e,t⟩
µ1
µ1
⟨t⟩
Adv ⟨⟨v,t⟩,t⟩
⟨e,t⟩
vP⟨v,t⟩
Adv vP⟨e,⟨v,t⟩⟩
PNI VP
⟨t⟩
v ⟨e,⟨v,t⟩⟩
PNI V ⟨e,⟨v,t⟩⟩
⟨⟨v,t⟩,t⟩
vP⟨v,t⟩
GQ
vP⟨e,⟨v,t⟩⟩ VP
v ⟨e,⟨v,t⟩⟩
GQ V ⟨e,⟨v,t⟩⟩
5.4 CASE STUDIES
151
While case marking alone seems to enable scope only over the adverbial quantifier (97b), the additional overt scrambling can trigger inverse scope scenarios with the universally quantified subject (97c). Finally, fronting the case-marked object leads to obligatory wide scope (97d). Since it is not my intention here to provide a detailed analysis of quantifier scope relations in SOV scrambling languages, I will leave the details of the analysis to future research.⁹ Negation contexts align with quantifier contexts. In contrast to Tamil and Hindi, Mongolian numeral phrases without case marking scope below negation, in line with caseless bare nouns; see (99b) and (99c). (99) Context ∃¬: There are two apples on the table of which Bold ate one. So there is one apple on the table Bold did not eat. a.
Bold neg alim-yg idee-gu¨j. Bold.NOM a apple-ACC eat-NEG ‘There is one apple Bold didn’t eat.’
Mongolian
b. #Bold neg alim idee-gu¨j. Bold.NOM a apple eat-NEG ‘Bold did not eat one apple.’ c. #Bold alim idee-gu¨j. Bold.NOM apple eat-NEG ‘Bold did not eat any apple.’ Numerals > 1 and weak quantifiers show the same split with respect to scope above negation, in that only case-marked objects are able to take wide scope, whereas caseless objects obligatorily scope below negation, proven be the infelicity of (100b) and (101b). If we compare again the acceptability of a case-marked bare noun with a case-marked overt quantifier phrase, cf. (101c) to (101b), we see that the overt quantifier is preferred by some speakers, similar to what we saw in Tamil in the previous section.
⁹ The analysis in (98) is preliminary at best. Guntsetseg (2016) provides scope contexts where a caseless object is able to outscope a temporal quantifier. She verifies her native speaker intuitions with a quantitive study. (i) Guntsetseg (2016: 118–120) Bold o¨do¨r bolgon neg kino u¨ze-x-ijg xu¨se-zˇ baj-na. Bold day every a movie watch-FUT-ACC want-CV be-PRS ‘Bold wants to see a movie every day.’
Mongolian ∃ > ∀t,∀t > ∃
The ∃ > ∀t reading might seem to contradict the general low scope restriction of caseless objects. There is, however, no a priori reason why adverbial and nominal scope patterns should align. It might even be expected that adverbial quantification lends itself to more interaction with PNI-ed arguments since their scope readings are intrinsically more closely connected to the event domain. Further investigation is needed in this respect.
152
PNI-PROPERT Y I: RESTRICTION TO LOW SCOPE
(100) Context 2¬: There are four apples on the table of which Bold ate two. So there are two apples on the table Bold did not eat. a.
Bold xodor alim-yg idee-gu¨j. Bold.NOM two apple-ACC eat-NEG ‘There are two apples Bold didn’t eat.’
Mongolian
b. #Bold xodor alim idee-gu¨j. Bold.NOM two apple eat-NEG ‘Bold did not two apples.’ (101) Context ∃¬: There are some apples on the table of which Bold ate a few. So there are still some apples on that table Bold did not eat. a.
Bold xeden alim-yg idee-gu¨j. Bold.NOM some apple-ACC eat-NEG ‘There are some apples Bold didn’t eat.’
Mongolian
b. #Bold xeden alim idee-gu¨j. Bold.NOM some apple eat-NEG ‘There are some apples Bold didn’t eat.’ c. %Bold alim-yg idee-gu¨j. Bold.NOM apple-ACC eat-NEG ‘There are apples Bold didn’t eat.’ A low scope context is provided in (102). As expected, the optimal way to express low scope below negation is with a caseless bare noun; see (102c). A case-marked quantifier phrase or bare noun are not licensed, shown in (102a) and (102b). Negation thus parallels the quantifier contexts above, in that Mongolian shows a wide scope preference for case-marked constituents. The predicted felicity of the caseless quantifier phrase (102d), however, suffers from the presence of a much more preferred alternative, the caseless bare noun (102c). (102) Context ¬∃: There are some apples on the table of which Bold did not eat any. a. #Bold xeden alim-yg idee-gu¨j. Bold.NOM some apple-ACC eat-NEG ‘There are some apples Bold didn’t eat.’ b. #Bold alim-yg idee-gu¨j. Bold.NOM apple-ACC eat-NEG ‘There are apples Bold didn’t eat.’
Mongolian
5.4 CASE STUDIES
c.
153
Bold alim idee-gu¨j. Bold.NOM apple eat-NEG ‘Bold did not eat any apples.’
d. ?Bold xeden alim idee-gu¨j. Bold.NOM some apple eat-NEG ‘Bold did not eat some apple(s).’ Finally, let us complete the picture with a low scope context, specific to neg NPs. As we can see in (103a), the case-marked numeral is dispreferred. Again, speakers did not agree on the felicity of (103b). In contrast to Tamil, where the low scope reading of the numeral depends on the availability of a scalar focus context, the dispreference for the Mongolian equivalent in (103b) is likely due to the tendency to parse the sentence with constituent negation (Dolgor Guntsetseg, p.c.). If (103b) is understood as Bold eating not one but more than one apple, it is not felicitous in the context given in (103). (103) Context ¬∃: There are two apples on the table of which Bold did not eat any. a. #Bold neg alim-yg idee-gu¨j. Bold.NOM a apple-ACC eat-NEG ‘There is one apple Bold didn’t eat.’
Mongolian
b. %Bold neg alim idee-gu¨j. Bold.NOM a apple eat-NEG ‘Bold did not eat one apple.’ Let us now turn to the LF derivations. We begin with the distribution of caseless bare nouns. They receive their low scope reading from the requirement of PNIDPs to be interpreted within the event domain. In (104), we see why (99c) is not licensed in a ∃¬ context. (104) *∃ > ¬ reading for (99c) [PNI-DP2⟨⟨e,⟨v,t⟩⟩,⟨v,t⟩⟩ type clash [μ2 ... ¬ ∃ [vP DP [VP ⟨DP2 ⟩ eat]]]⟨e,t⟩ ] Caseless bare nouns are licensed in a ¬∃ context such as the one in (102). The LF representation in (105) provides the structure for (102c). (105) ¬ > ∃ reading for (102c) [¬ ∃ [vP DP [VP PNI-DP⟨⟨e,⟨v,t⟩⟩,⟨v,t⟩⟩ eat⟨e,⟨v,t⟩⟩ ]]] Caseless quantifier phrases can be treated identically. Recall that weak quantifiers such as xeden also adjoin to NP, like it is shown in (87). Thus, xeden alim is not
154
PNI-PROPERT Y I: RESTRICTION TO LOW SCOPE
licensed in (101b) because PNI-DP needs to be interpreted outside vP which is impossible, as shown in (104). As discussed above, xeden alim is not preferred in (102d) due to the less marked alternative in (102c). Case-marked bare nouns scope outside the event domain, they are headed by D∃ . Hence, they are licensed in (101c); see (106) for a sketched LF representation. Just like in Tamil they resist a reading below negation; see (102b). Case-marked quantifiers receive the same treatment, as they are also licensed in (101a). Recall that D∃ and DPNI can both take weak quantifiers adjoined to NP as arguments. (106) ∃ > ¬ reading for (101c) [DP1⟨⟨e,t⟩,t⟩ [μ1 ... ¬ ∃ [vP DP [VP ⟨DP1 ⟩ eat]]]⟨e,t⟩ ] The numeral constructions can be analysed in exactly the same way, the only difference is an additional constraint on the restrictor argument of DPNI and D∃ . Caseless numerals constitute PNI-DPs in Mongolian, therefore low scope is obligatory, which is provided in (103) but not in (99); see the structures below. (107) *∃ > ¬ reading for (99b) [PNI-DP2⟨⟨e,⟨v,t⟩⟩,⟨v,t⟩⟩ type clash [μ2 ... ¬ ∃ [vP DP [VP ⟨DP2 ⟩ eat]]]⟨e,t⟩ ] (108) ¬ > ∃ reading for (103b) [¬ ∃ [vP DP [VP PNI-DP⟨⟨e,⟨v,t⟩⟩,⟨v,t⟩⟩ eat⟨e,⟨v,t⟩⟩ ]]] Finally, the wide scope reading of case-marked numerals is made possible by the presence of D∃ , a generalized quantifier which needs to be interpreted outside the event domain. The LF structure is given in (109). (109) ∃ > ¬ reading for (99a) [DP1⟨⟨e,t⟩,t⟩ [μ1 ... ¬ ∃ [vP DP [VP ⟨DP1 ⟩ eat]]]⟨e,t⟩ ] As I have shown in this section, Mongolian displays a much broader class of noun types which can undergo PNI. We can diagnose a consistent correlation of case drop with low scope for bare nouns, numerals, and weak quantifiers. As in Hindi and Tamil, the ‘indefinite’ determiner should be treated like a numeral. In contrast to Hindi and Tamil, numerals as well as weak quantifiers receive an adjectival treatment, thereby making it possible for the hybrid PNI determiner to take them as arguments. We will now turn to Turkish, a PNI language which resembles Mongolian in many ways.
5.4.3 Turkish The interaction of case marking with scope/specificity properties has received much attention in the literature on Turkish (Aygen 2007a; Enc˛ 1991; Kelepir 2001;
5.4 CASE STUDIES
155
Kornfilt 2008; Özge 2011; Öztürk 2005, 2009; von Heusinger and Bamyacı 2016; von Heusinger and Kornfilt 2005). The purpose of this section is therefore mainly to show how the current account can cover PNI scope interactions, reported for Turkish. I will, however, extend the data set with respect to the scopal properties of PNI-ed subjects in embeddeded clauses. Turkish behaves like Hindi, in that case-marked bare nouns receive a definite reading. The caseless counterpart is often translated along the lines of a PNI-meaning. (110) von Heusinger and Kornfilt (2005: 5) a. (Ben) kitap oku-du-m. I book read-PST-1SG ‘I was book-reading.’ b. (Ben) kitab-ı oku-du-m. I book-ACC read-PST-1SG ‘I read the book.’
Turkish PNI
DEFINITE
Öztürk (2005, 2009) reports on the lack of wide scope readings for caseless bare nouns (see (111a)), in contrast to indefinites which, if case marked, can take scope over a universally quantified subject, cf. (111b) to (111c). These judgements are in line with earlier work by Kelepir (2001) and Aygen (2007a). The split between (111b) and (111c) was additionally verified by a recent quantative study (von Heusinger and Bamyacı 2016). Thus, Mongolian and Turkish pattern alike in that bare nouns as well as indefinites undergo PNI. (111) Öztürk (2005: 67–68)
Turkish
a. Her c¸ocuk kitap okudu. every child book read ‘Every child did book-reading.’
∀∃,*∃∀
b. Her c¸ocuk bir kitap okudu. every child one book read ‘Every child read a book.’
∀∃,*∃∀
c. Her c¸ocuk bir kitab-ı okudu. every child one book-ACC read ‘Every child read a book.’
∀∃,∃∀
In contrast to Tamil, Hindi, and Mongolian, bir NPs can occur in generic contexts. The Turkish aorist marking enforces genericity (Kornfilt 1997: 280). As (112b) and (112c) show, a bare noun as well as bir NP are licensed and create generic readings.
156
PNI-PROPERT Y I: RESTRICTION TO LOW SCOPE
(112) Generic readings
Turkish
a. Bir fil ¸seker kamıs¸-ı y-iyor. a elephant.NOM sugar cane-CMPD eat-PROG ‘An elephant is eating sugar cane.’
*GENERIC
b. Bir fil ¸seker kamıs¸-ı ye-r. a elephant.NOM sugar cane-CMPD eat-AOR ‘An elephant eats / would eat sugar cane.’
GENERIC
c. Fil ¸seker kamıs¸-ı ye-r. elephant.NOM sugar cane-CMPD eat-AOR ‘The elephant eats sugar cane.’
GENERIC
I take the felicity of bir NPs in generic contexts to indicate that they can act as indefinites, in addition to their numeral function; see (113). The nature of bir has been widely discussed in the literature on Turkish indefiniteness, with analyses treating bir NPs as indefinites (Kornfilt 1997; Özge 2011), numeral modifiers (Aygen 2007a; Sağ 2018) or both (Öztürk 2005). We make use of the latter assumption and derive the case alternation with bir NPs by the choice of adjectival bir vs. determiner bir. If bir NPs are case marked, they are formed with bir instantiating an existential determiner, as shown in (115). Caseless bir NPs are formed with adjectival bir where the PNI determiner PNI-D turns them into arguments; see (114). It is the indefinite interpretation in (115) which allows bir NPs to occur in generic contexts. (113) Indefinite/numeral bir in Turkish a. ⟦bir1 ⟧ = λP⟨e,t⟩ λQ⟨e,t⟩ ∃z[P(z) ∧ Q(z)] b. ⟦bir2 ⟧ = λz.|z| = 1 (114) bir kitap (111b): PNI-DP ⟨⟨e,⟨v,t⟩⟩,⟨v,t⟩⟩ λ Q λ e z[ BOOK (z) z = 1 Q(z)(e)] P NI -D ⟨⟨e, t⟩, ⟨⟨e, ⟨v, t⟩⟩, ⟨v, t⟩⟩⟩ λ Pλ Q λ e z[P(z) Q(z)(e)]
NP ⟨e, t⟩ λ x[ BOOK ( x)
x = 1]
DP ⟨e,t⟩ NP ⟨e,t⟩ = bir λ y[ BOOK (y)] λ z. z = 1
5.4 CASE STUDIES
157
(115) bir kitab-ı (111c):
DP ⟨⟨e, t⟩, t⟩ λ Q z[ BOOK (z) Q(z)] NP ⟨e,t⟩ D ⟨⟨e, t⟩, ⟨⟨e, t⟩, t⟩⟩ = bir λ y[ BOOK (y)] λ Pλ Q z[P(z) Q(z)] The fact that D∃ is spelled-out as bir explains why case-marked bare nouns automatically receive a definite reading, recall (110b). Like in Hindi, there is no zero exponent for D∃ . The only D head which is spelled out as zero and requires marking is the definite one. In contrast to Hindi, however, caseless bir NPs show a low scope restriction. The additional adjectival use of bir enables the PNI determiner to combine with bir NPs, thereby triggering case loss post-syntactically. It is worthwhile to compare Turkish bir to Mongolian, Tamil, and Hindi indefinites/numerals. Whereas in Mongolian neg receives a numeral treatment exclusively, Turkish bir can also function as the indefinite determiner. In order to turn neg NPs into arguments in Mongolian, we had to assume an additional silent D∃ head for case-marked neg NPs. We saw independent evidence for this D head, e.g. with respect to existential sentences where bare nouns as well as neg NPs were licensed to occur, identical to the situation in Tamil. Unfortunately, Turkish existential sentences do not provide a good test for indefiniteness, as even pronouns and proper names, i.e. prototypical definites, are licensed as pivots (Özyıldız 2017: 913). Another test for indefiniteness are discourse new contexts. While Tamil allows for bare nouns, Hindi does not. Dayal thus argues that there is no silent D∃ in Hindi. Tamil, in contrast, provides robust evidence for the presence of such a determiner. The discourse new context in (116) shows that Turkish patterns with Hindi, and not with Tamil and Mongolian, as bare nouns are not allowed to occur. (116) Context: (Beginning of a story) Many years ago... a.
Turkish
bu ev-de yas¸lı bir ko¨peg˘-in o¨l-du¨g˘-u¨-nu¨ this house-LOC old a dog-GEN die-NMLZ-POSS.3SG-ACC so¨yl-er-ler. say-IPFV-PL ‘They say that an old (specific) dog died in this house.’
b. #bu ev-de yas¸lı ko¨peg˘-in o¨l-du¨g˘-u¨-nu¨ so¨yl-er-ler. this house-LOC old dog-GEN die-NMLZ-POSS.3SG-ACC say-IPFV-PL ‘They say that the old dog died in this house.’
158
PNI-PROPERT Y I: RESTRICTION TO LOW SCOPE
To sum up, the indefinite determiner is spelled-out as bir, triggering the occurence of case-marked indefinites in generic as well as discourse new contexts. Furthermore, bir can act as an adjectival modifier, enabling semantic composition with the definite determiner or PNI-D. It should also be noted that we must extend this analysis to numerals generally, as well as weak quantifiers, since they show the same case-scope sensitivity. The LF structure for the licit low scope reading of the PNI-ed bare noun in (111a) is given in (117). An identical structure can be given for bir NP in (111b), except for the additional numerical restriction bir2 on the domain of PNI-DP. (117) ∀ > ∃ reading of (111a)/(111b) [DP1⟨⟨e,t⟩,t⟩ [μ1 ... ∃ [vP ⟨DP1 ⟩ [VP PNI-DP⟨⟨e,⟨v,t⟩⟩,⟨v,t⟩⟩ read⟨e,⟨v,t⟩⟩ ]]]⟨e,t⟩ ] Wide scope readings are not possible due to PNI-DP’s requirement to be interpreted inside the event domain, while the universally quantified subject raises out of the vP, shown in (118). Again, the bir NP equivalent in (111b) receive an identical treatment but with an additional numerical restriction bir2 on the domain of PNI-DP. (118) *∃ > ∀ reading of (111a)/(111b) [PNI-DP2⟨⟨e,⟨v,t⟩⟩,⟨v,t⟩⟩ t.cl. [μ2 ..DP1⟨⟨e,t⟩,t⟩ [μ1 ..∃ [vP ⟨DP1 ⟩ [VP ⟨DP2 ⟩ read]]]⟨e,t⟩ ]⟨e,t⟩ ] The case marking on bir NP in (111c) results from the presence of the existential determiner bir1 . Hence, the two generalized quantifiers both need to raise out of the event domain, where surface scope is shown in (119) and inverse scope in (120). (119) ∀ > ∃ reading of (111c) [DP1⟨⟨e,t⟩,t⟩ [μ1 ... DP2⟨⟨e,t⟩,t⟩ [μ2 ... ∃ [vP ⟨DP1 ⟩ [VP ⟨DP2 ⟩ read]]]⟨e,t⟩ ]⟨e,t⟩ ] (120) ∃ > ∀ reading of (111c) [DP2⟨⟨e,t⟩,t⟩ [μ2 ... DP1⟨⟨e,t⟩,t⟩ [μ1 ... ∃ [vP ⟨DP1 ⟩ [VP ⟨DP2 ⟩ read]]]⟨e,t⟩ ]⟨e,t⟩ ] The definite reading in (110b) follows from the presence of either the strong or the weak definite article. I show the LF with the strong form; see (121).
5.4 CASE STUDIES
159
(121) definite reading of (110b) vP⟨t⟩ !x[ BOOK ( x) x = g(3)] . e[ READ (ιx[ BOOK ( x) x = g(3)])(e) AG ( g(1))(e)] vP⟨v,t⟩ ! x[ BOOK ( x) x = g(3)] . READ (ιx[ BOOK ( x) x = g(3)])(e) AG ( g(1))(e) λe
DP g(1)
v ′⟨e,⟨v,t⟩⟩ λ xλ e !x[ BOOK ( x) x = g(3)] . READ (ιx[ BOOK ( x) x = g(3)])(e) AG ( x)(e) VP ⟨v,t⟩ !x[ BOOK ( x) x = g(3)] . READ (ιx[ BOOK ( x) x = g(3)])(e) λe
DP ⟨e⟩ !x[ BOOK ( x) x = g(3)] . ιx[ BOOK ( x) x = g(3)] g(3) λy
v ⟨e,⟨v,t⟩⟩ λ xλ e[ AG ( x)(e)]
V⟨e,⟨v,t⟩⟩ λ xλ e[ READ ( x)(e)]
′ D⟨e,e⟩ !x[ BOOK ( x) x = y]. ιx[ BOOK ( x) x = y]
NP ⟨e,t⟩ D λ y[ BOOK (y)] ⟨⟨e, t⟩, ⟨e, e⟩⟩ λ Pλ y !x[P( x) x = y]. ιx[P( x) x = y]
We can make the same observations for scope contexts including negation. In (122), we see a correlation between case drop and a restriction to low scope for bir NPs. The pair in (123) extends this observation to numerals generally. (122) Kelepir (2001: 85) a. Hasan Ali-ye bir hediye al-ma-dı. Hasan Ali-DAT a present buy-NEG-PST ‘Hasan didn’t buy Ali a present.’
Turkish ¬∃,*∃¬
b. Hasan bir o¨dev-i yap-ma-dı. ¬∃,∃¬ Hasan a homework-ACC do-NEG-PST ‘Hasan didn’t do a homework./Hasan didn’t do one of the homeworks.’
160
PNI-PROPERT Y I: RESTRICTION TO LOW SCOPE
(123) Kelepir (2001: 85)
Turkish
a. Hasan iki kitap oku-ma-dı. Hasan two book read-NEG-PST ‘Hasan didn’t read two books.’ (He read only one book.)
¬2,*2¬
b. Hasan iki kapı-yı cilala-ma-dı. ¬2,2¬ Hasan two door-ACC polish-NEG-PST ‘Hasan didn’t polish two doors./Hasan didn’t polish two of the doors.’ The low scope restriction of the caseless bir NP in (122a) is a consequence of PNIDP, bir acts as an adjective, as was discussed above. The LF trees are given in (124) and (125) for the respective readings. We will model the additional benefactive argument in (122a) with an applicative projection, following the proposed syntax and semantics in Pylkkänen (2008). The applicative head is very similar to v in that it simply adds another participant to the event. (124) ¬ > ∃ reading for (122a) NegP⟨t⟩ ¬ e z[ PRES (z) z = 1 BUY (z)(e) BEN ( ali)( e) AG (hasan)(e)] Neg ⟨t,t⟩ ¬p[ p]
vP⟨t⟩ e z[ PRES (z) z = 1 BUY (z)(e) BEN ( ali)( e) AG (hasan)(e)]
vP⟨v,t⟩ λ e z[ PRES (z) z = 1 BUY (z)(e) BEN ( ali)( e) AG (hasan)(e)] DP Hasan
v ′⟨e,⟨v,t⟩⟩ λ xλ e z[ PRES (z) z = 1 BUY (z)( e) BEN ( ali)( e) AG ( x)( e)] ApplP⟨v,t⟩ λ e z[ PRES (z) z = 1 BUY (z)( e) BEN ( ali)( e)]
DP Ali
v ⟨e,⟨v,t⟩⟩ λ xλ e[ AG ( x)(e)]
Appl ′⟨e,⟨v,t⟩⟩ λ xλ e z[ PRES (z) z = 1 BUY (z)( e) BEN ( x)(e)] VP⟨v,t⟩ λ e z[ PRES (z) z = 1 BUY (z)(e)]
P NI -DP ⟨⟨e, ⟨v, t⟩⟩, ⟨v, t⟩⟩ λ Q λ e z[ PRES (z) z = 1 Q(z)(e)]
V ⟨e,⟨v,t⟩⟩ λ xλ e[ BUY ( x)(e)]
Appl ⟨e,⟨v,t⟩⟩ λ xλ e[ BEN ( x)(e)]
5.4 CASE STUDIES
161
(125) *∃ > ¬ reading for (122a) [PNI-DP2⟨⟨e,⟨v,t⟩⟩,⟨v,t⟩⟩ type clash [μ2 ... ¬ ∃ [vP DP [ApplP DP [VP ⟨DP2 ⟩ buy]]]]⟨e,t⟩ ] Case-marked bir NPs, as in (122b), instantiate existential quantifiers and thus need to move outside of the event domain, in order to be interpreted. Since they show flexible scope with respect to negation, they must be able to QR right above existential closure but below negation (126) or above negation (127). (126) ¬ > ∃ reading for (122b) [¬ DP1⟨⟨e,t⟩,t⟩ [μ1 ... ∃ [vP DP [VP ⟨DP1 ⟩ do]]]⟨e,t⟩ ]
(127) ∃ > ¬ reading for (122b) [DP1⟨⟨e,t⟩,t⟩ [μ1 ... ¬ ∃ [vP DP [VP ⟨DP1 ⟩ do]]]⟨e,t⟩ ] Numeral phrases which are not case marked constitute PNI-DPs. Hence they show a low scope restriction, as can be seen in (123a). The LF structures are given in (128) and (129), respectively. Just like bir, numerals > 1 are able to act as adjectival modifiers. (128) *2 > ¬ reading for (123a) [PNI-DP2⟨⟨e,⟨v,t⟩⟩,⟨v,t⟩⟩ type clash [μ2 ... ¬ ∃ [vP DP [VP ⟨DP2 ⟩ read]]]⟨e,t⟩ ]
(129) ¬ > 2 reading for (123a) [¬ ∃ [vP DP [VP PNI-DP⟨⟨e,⟨v,t⟩⟩,⟨v,t⟩⟩ read⟨e,⟨v,t⟩⟩ ]]] Case-marked numeral phrases, as in (123b), are derived with numeral determiners. Since they constitute DPs, they are case marked and show flexible scope with respect to other operators, shown here for negation in (130) and (131). As with bir NPs, we assume that DPs can QR above or below negation. (130) ¬ > 2 reading for (123b) [¬ DP1⟨⟨e,t⟩,t⟩ [μ1 ... ∃ [vP DP [VP ⟨DP1 ⟩ polish]]]⟨e,t⟩ ]
(131) 2 > ¬ reading for (123b) [DP1⟨⟨e,t⟩,t⟩ [μ1 ... ¬ ∃ [vP DP [VP ⟨DP1 ⟩ polish]]]⟨e,t⟩ ] Caseless bare nouns show the same scope restrictions as caseless bir NPs; see (132a). Case marking triggers a definite interpretation, shown in (132b).
162
PNI-PROPERT Y I: RESTRICTION TO LOW SCOPE
(132) PNI reading vs. definite reading a. Hasan hediye al-ma-dı. Hasan.NOM present buy-NEG-PFV ‘Hasan has not bought presents.’ b. Hasan hediye-yi al-ma-dı. Hasan.NOM present-ACC buy-NEG-PFV ‘Hasan has not bought the present.’
Turkish ¬∃,*∃¬
DEFINITE
The LF structures for the PNI configuration in (132a) are given in (133) and (134) below. Bare nouns are not marked for case due to the presence of the hybrid PNI determiner, hence the restriction to low scope. The definite reading for a casemarked bare noun is shown in (135). As discussed above, Turkish D∃ is spelled out overtly by bir. Thus, the only other silent D head available is the definite one. (133) ¬ > ∃ reading for (132a) [¬ ∃ [vP DP [VP PNI-DP⟨⟨e,⟨v,t⟩⟩,⟨v,t⟩⟩ buy⟨e,⟨v,t⟩⟩ ]]]
(134) *∃ > ¬ reading for (132a) [PNI-DP2⟨⟨e,⟨v,t⟩⟩,⟨v,t⟩⟩ type clash [μ2 ... ¬ ∃ [vP DP [VP ⟨DP2 ⟩ buy]]]⟨e,t⟩ ]
(135) definite reading of (132b) [¬ ∃ [vP DP [VP DP⟨e⟩ buy⟨e,⟨v,t⟩⟩ ]]] Up until now, we have focused on direct objects since they are overtly marked for structural case. As for subjects, the PNI languages investigated so far mark them non-overtly with nominative case. Turkish, however, marks subjects overtly for genitive case in selected subordinate environments, in line with other Altaic languages such as Japanese (Miyagawa 1993, 2011; Ochi 2001). The sensitivity of genitive case drop to scope/specificity has not gone unnoticed in the literature (Aygen 2007b; Jo and Palaz 2018; Kornfilt 2003a,b, 2008). I will now link the case-scope interaction to the PNI account argued for in this book. Case as well as agreement morphology signals a nominal licenser, which has been suggested to come in form of a nominal Agr projection (Kornfilt 2003b) or a DP/NP shell (Aygen 2007b; Lees 1965). For present purposes, I will gloss over analytical details and simply assume that genitive case is assigned to the subject by some nominal head within a nominalized complement clauses, which will also result in nominal agreement morphology on the embedded verb. The minimal contrast in (136) demonstrates how subject PNI parallels object PNI: case drop on bare arguments
5.4 CASE STUDIES
163
triggers a low scope restriction, whereas case-marked bare arguments receive a definite interpretation.¹⁰ (136) PNI of bare subjects in embedded clauses
Turkish
a. Ko¨y-e doktor gel-me-dig˘-i-ni duy-du-m. village-DAT doctor come-NEG-NMLZ-POSS.3SG-ACC hear-PFV-1SG ‘I have heard that no doctor came to the village.’ ¬∃,*∃¬ b. Ko¨y-e doktor-un gel-me-dig˘-i-ni duy-du-m. village-DAT doctor-GEN come-NEG-NMLZ-POSS.3SG-ACC hear-PFV-1SG ‘I have heard that the doctor didn’t come to the village.’ DEFINITE Similarly, bir NPs show a low scope restrictions if they are not case marked; see (137a). If they are marked for genitive case, they can scope above or below negation; see (137b). (137) PNI of bir subjects in embedded clauses
Turkish
a. Ko¨y-e bir doktor gel-me-dig˘-i-ni duy-du-m. village-DAT a doctor come-NEG-NMLZ-POSS.3SG-ACC hear-PFV-1SG ‘I have heard that no doctor came to the village.’ ¬∃,*∃¬ b. Ko¨y-e bir doktor-un gel-me-dig˘-i-ni village-DAT a doctor-GEN come-NEG-NMLZ-POSS.3SG-ACC duy-du-m. hear-PFV-1SG ‘I have heard that not even one doctor came to the village./I have heard that there is a doctor who did not come to the village.’ ¬∃,∃¬ As with objects, subjects constituting numeral phrases > 1 are able to undergo PNI, i.e. case drop on numeral subjects correlates with obligatory low scope, as shown in (138a), while case-marked numeral subjects are scopally flexible; see (138b).
¹⁰ Agent pseudo-incorporation is preferred with intransitives but not impossible with transitive verbs. An example for agent PNI with transitive verbs is given below with parallel readings; see also Öztürk (2005, 2009) for more examples with transitive verbs. (i)
PNI of bare subjects in embedded clauses a. Hasta-lar-ı doktor go¨r-me-dig˘-i-ni duy-du-m. patient-PL-ACC doctor see-NEG-NMLZ-POSS.3SG-ACC hear-PFV-1SG ‘I have heard that no doctor saw the patients.’ b. Hasta-lar-ı doktor-un go¨r-me-dig˘-i-ni duy-du-m. patient-PL-ACC doctor-GEN see-NEG-NMLZ-POSS.3SG-ACC hear-PFV-1SG ‘I have heard that the doctor didn’t see the patients.’
Turkish ¬∃,*∃¬
DEFINITE
164
PNI-PROPERT Y I: RESTRICTION TO LOW SCOPE
(138) PNI of numeral subjects in embedded clauses
Turkish
a. Ko¨y-e iki doktor gel-me-dig˘-i-ni duy-du-m. village-DAT two doctor come-NEG-NMLZ-POSS.3SG-ACC hear-PFV-1SG ‘I have heard that two doctors didn’t come to the village.’ (but one) ¬2,*2¬ b. Ko¨y-e iki doktor-un gel-me-dig˘-i-ni village-DAT two doctor-GEN come-NEG-NMLZ-POSS.3SG-ACC duy-du-m. hear-PFV-1SG ‘I have heard that there are two doctors that didn’t come to the village./I have heard that not even two doctors came to the village.’ 2¬,¬2 We begin with the LF trees for the bare nouns in (136).¹¹ The embedded subject in (136a) is not case marked and therefore has to be interpreted within the event domain, cf. (139) and (140). Case-marked bare subjects are definite noun phrases; see (141). (139) ¬ > ∃ reading for (136a) NegP⟨t⟩ ¬ e z[ DOCT (z) COME (e) AG (z)(e)] Neg ⟨t,t⟩ ¬ p[ p]
vP⟨t⟩ e z[ DOCT (z) COME (e) AG (z)(e)] vP⟨v,t⟩ λe z[ DOCT (z) COME (e) AG (z)(e)] P NI -DP ⟨⟨e, ⟨v, t⟩⟩, ⟨v, t⟩⟩ λ Q λe z[ DOCT (z) Q(z)(e)]
v ′⟨e,⟨v,t⟩⟩ λ xλ e[ COME (e) AG (x)(e)] VP ⟨v,t⟩ λ e[ COME (e)]
v ⟨e,⟨v,t⟩⟩ λ xλ e[ AG (x)(e)]
(140) *∃ > ¬ reading for (136a) [PNI-DP2⟨⟨e,⟨v,t⟩⟩,⟨v,t⟩⟩ type clash [μ2 ... ¬ ∃ [vP ⟨DP2 ⟩ [VP come]]]⟨e,t⟩ ] ¹¹ I will ignore the locative adjunct ko¨y-e for the subject PNI cases in the interest of space.
5.4 CASE STUDIES
165
(141) definite reading for (136b) [¬ ∃ [vP DP⟨e⟩ [vʹ [VP come ] v ]⟨e,⟨v,t⟩⟩ ]] Moving on to bir NPs, (137) equally shows how subjects parallel objects in their scopal distribution. In (137a), bir functions as a numeral, adjoining to NP and combining with PNI-D to form a PNI-ed subject. As always, low scope follows from the requirement of PNI-DPs to be interpreted inside the event domain; see (142) and (143). (142) *∃ > ¬ reading for (137a) [PNI-DP2⟨⟨e,⟨v,t⟩⟩,⟨v,t⟩⟩ type clash [μ2 ... ¬ ∃ [vP ⟨DP2 ⟩ [VP come]]]⟨e,t⟩ ]
(143) ¬ > ∃ reading for (137a) [¬ ∃ [vP PNI-DP⟨⟨e,⟨v,t⟩⟩,⟨v,t⟩⟩ [vʹ [VP come ] v ]⟨e,⟨v,t⟩⟩ ]] For bir NPs to get case marked, bir has to act as the existential determiner, i.e. as bir1 . D∃ forms a generalized quantifier, which can only be interpreted outside the event domain. This enables case-marked bir NPs to show flexible scope behaviour with respect to negation in (137b), shown in (144) and (145). (144) ¬ > ∃ reading for (137b) [¬ DP1⟨⟨e,t⟩,t⟩ [μ1 ... ∃ [vP ⟨DP1 ⟩ [VP come]]]⟨e,t⟩ ] (145) ∃ > ¬ reading for (137b) [DP1⟨⟨e,t⟩,t⟩ [μ1 ... ¬ ∃ [vP ⟨DP1 ⟩ [VP come]]]⟨e,t⟩ ] Recall that numerals > 1 also have a double function in Turkish, therefore we will treat them the same way. They can serve as modifiers as well as determiners, where the former leads to PNI (or definite readings) in (138a) and the latter leads to flexible scope above or below negation in (138b). The possible and impossible readings for PNI-ed subjects in (138a) are shown in (146) and (147), respectively. (146) *2 > ¬ reading for (138a) [PNI-DP2⟨⟨e,⟨v,t⟩⟩,⟨v,t⟩⟩ type clash [μ2 ... ¬ ∃ [vP ⟨DP2 ⟩ [VP come]]]⟨e,t⟩ ]
(147) ¬ > 2 reading for (138a) [¬ ∃ [vP PNI-DP⟨⟨e,⟨v,t⟩⟩,⟨v,t⟩⟩ [vʹ [VP come ] v ]⟨e,⟨v,t⟩⟩ ]]
166
PNI-PROPERT Y I: RESTRICTION TO LOW SCOPE
Finally, case-marked numeral phrases in (138b) result from numerals acting as determiners, thereby enforcing QR out of the event domain. Hence, case-marked numeral subjects can interact with the negative operator in (138b). The respective readings are given in (148) and (149). (148) ¬ > 2 reading for (138b) [¬ DP1⟨⟨e,t⟩,t⟩ [μ1 ... ∃ [vP ⟨DP1 ⟩ [VP come]]]⟨e,t⟩ ]
(149) 2 > ¬ reading for (138b) [DP1⟨⟨e,t⟩,t⟩ [μ1 ... ¬ ∃ [vP ⟨DP1 ⟩ [VP come]]]⟨e,t⟩ ] This section has shown that PNI-DPs can enter the derivation as objects as well as subjects. In principle, the current account allows for both argument types to undergo PNI. Since we need visible case markers to diagnose a correlation with low scope restrictions, not every PNI language will make a good candidate to show subject PNI transparently. Fortunately, nominalized complement structures in Turkish provide an optimal testing ground for subject pseudo-incorporation, as they mark subjects overtly for genitive case. Korean is another language which marks subjects overtly for case. As in Turkish, we can detect a correlation between case drop and obligatory low scope readings, as the next section will demonstrate. The other important empirical contribution this section has provided relates to the nature of D∃ . As in Hindi, Turkish case-marked bare nouns receive a definite reading. For an indefinite reading bir has to be used. Whereas Turkish bir can take over the function of an indefinite, proven by the licensed occurence in generic contexts, the Hindi/Tamil/Mongolian equivalents ek/oru/neg cannot. As we have seen in Section 5.4.1, Hindi/Tamil ek/oru instantiate D heads which are in complementary distribution with the PNI-determiner PNI-D, hence ek/oru NPs do not undergo PNI. This is not true for Turkish bir, which leads us to the conclusion that bir (as well as other numerals and a set of weak quantifiers) must be able to function as an adjectival modifier which enables bir NPs to form PNI-DPs. Turkish is in this sense different from Mongolian since neg exclusively serves as an adjectival modifier, recall section 5.4.2. In the next section, we will investigate Korean, a classifier language in which bare nouns can undergo PNI but where case drop on numerals constructions signals DOM rather than PNI.
5.4.4 Korean A restriction to low scope readings of caseless objects was originally observed by Kwon and Zribi-Hertz (2006) with respect to contexts including negation.
5.4 CASE STUDIES
(150) Kwon and Zribi-Hertz (2006: 118) a. Minsu-neun sagwa meog-ji anh-ass-da. Minsu-TOP apple eat+JI NEG-PST-DECL ‘lit. Minsu did not apple-eat.’ b. Minsu-neun sagwa-leul meog-ji anh-ass-da. Minsu-TOP apple-ACC eat+JI NEG-PST-DECL ‘Minsu did not eat apples/the apple.’
167
Korean *∃¬,¬∃
∃¬,¬∃
The translation in (150b) suggests that case-marked bare nouns do not automatically shift to a definite interpretation, they also allow for an indefinite interpretation. This impression seems to be shared amongst Korean speakers; see e.g. Ch. Lee (1989: 7) and more recently Ahn (2018: 27). The intuition is supported by the licit occurrence of bare nouns in the various indefiniteness environments we investigated so far throughout this study. Recall from Section 5.4.1 that Hindi bare nouns were not licensed in homogeneity contexts. Korean bare nouns, in contrast are allowed to occur; see (151a), besides e.g. demonstrative uses (151b). (151) Context: You enter a room and you see a dog sleeping and a dog barking. Now you tell me what you see. a. Kay-ka ca-ko kay-ka cic-ko iss-ta. dog-NOM sleep-CONJ dog-NOM bark-CONJ exist-DECL ‘A dog is sleeping and a dog is barking.’
Korean
b. Ku kay-ka ca-ko ku kay-ka cic-ko iss-ta. DEM dog-NOM sleep-CONJ DEM dog-NOM bark-CONJ exist-DECL ‘That dog is sleeping and that (other) dog is barking.’ Bare nouns can also occur in discourse new contexts, shown in (152a). Compare this to (152b) which employs demonstratives, an argument type which is usually banned in contexts that introduce a new discourse referent. (152) Context: (Beginning of a story) Many years ago... a. Nulkun kay-ka i cip-eyse cwuk-ess-e old dog-NOM DEM house-at die-PST-INT ‘An old dog died in this house.’ b. #Ku nulkun kay-ka i cip-eyse cwuk-ess-e DEM old dog-NOM DEM house-at die-PST-INT ‘That old dog died in this house.’
Korean
168
PNI-PROPERT Y I: RESTRICTION TO LOW SCOPE
Further evidence for the indefinite potential of bare nouns can be found in Ch. Lee (1989) who discusses occurrences in existential sentences. We can conclude from these tests that there exists solid evidence for the presence of a silent D∃ head in Korean which is responsible for indefinite readings of case-marked bare nouns, beside the definite ones. Korean, thus, patterns with Tamil and Mongolian in this respect. Although Kwon and Zribi-Hertz (2006) provide in-situ examples for the scope readings in (150), the sentences in my study involve scrambled objects. Most of the speakers only allowed for an alternation of surface and inverse readings if the object has overtly moved over another operator. Korean contrasts in this regard with every other PNI language we have looked at so far. For Spanish, Hindi, Tamil, Mongolian, and Turkish, case marking of bare nouns and indefinites on its own enabled objects to scope over quantified subjects or negation, i.e. there was no additional need for overt scrambling. One possibility is to attribute this requirement to the fact that Korean is often discussed as a scope rigid language; see e.g. Y. Lee (1999: 28) and Eu. Lee (2019: 154) but also Ahn (1990); Sohn (1995). Scope rigidity, however, seems to be a general property of scrambling languages and has been proven to prevail in at least three of the five languages, alongside PNI; see Kelepir (2001) and Bošković and Şener (2014) for Turkish, López (2012) for Spanish, and Keine and Poole (2018a) for Hindi.¹² Thus it seems that scope rigidity can be overwritten by the effects of PNI in some languages, such as Turkish, Hindi, and Spanish, but not in others. In other words, Korean requires case marking as well as overt scrambling for inverse readings. The fact that scopal relations between different operators can be determined not only on the basis of one grammatical mechanism is not completely unheard of. German e.g. makes use of scrambling and/or marked intonation patterns to trigger inverse or reconstructed readings (Jacobs 1982; Frey 1993; Féry 1993; Krifka 1998).¹³ Kelepir (2001) and López (2012) take the simultaneous presence of PNI insitu effects and scope rigidity as evidence to analyse case-marked indefinites as variables ranging over choice functions (Kratzer 1998; Reinhart 1997; Winter ¹² Scope rigidity can be tested independently of PNI by making use of strong quantifiers instead of bare nouns or indefinites. ¹³ I can confirm scope rigidity for Tamil, based on the judgements of two native speakers. Interestingly, overt movement does not suffice to create the inverse reading. (i)
Scope rigidity and scrambling
Tamil
a. Nethiki oru doctor ella patient-ai-um paar-t-aar. yesterday a doctor.NOM all patient-ACC-ADD see-PST-3SG.HON ‘Yesterday, a doctor saw all patients.’
∃∀,*∀∃
b. Nethiki [ella patient-ai-um]1 oru doctor ⟨DP1 ⟩ paar-t-aar. yesterday all patient-ACC-ADD a doctor.NOM see-PST-3SG.HON ‘Yesterday, a doctor saw all patients.’
∃∀,*∀∃
One speaker used a strong distributive quantifier in subject position in addition to overt scrambling of the object to trigger the ∀∃ reading, whereas the other speaker used a passive; see Sundaresan and McFadden (2017) for a detailed analysis of passive voice in Tamil. This, furthermore, confirms that there is a multitude of strategies languages employ to create inverse readings.
5.4 CASE STUDIES
169
1997). The scope of a choice function is determined by the position of a dedicated closure operator binding the choice function. Since this operator is silent, indefinites derived by choice functions do not need to overtly scramble to their scope positions, nor is there any other overt evidence for the scope taking position. This account is in principle compatible with the current system, given that Kratzer (1998) and Matthewson (1999) assume choice functions to have scope at the sentence level where they are bound by the existential closure operator. The position of this operator quite possibly mirrors the effects created by the assumption that GQs have to be interpreted outside the event domain. Further tests, however, would have to be run for Mongolian and Tamil investigating the potential of case-marked indefinites to scope out of scope islands, which is another hallmark property of choice functions. Kelepir (2001) and López (2012) provide evidence along those lines. Korean would still require another solution since case-marked indefinites do need to move in order to take wide scope, which is not predicted if they were analysed as choice functions. The availability of choice functions and their interaction with scope rigidity will have to be left open for the purpose of this book. We now turn to the scope interactions of caseless bare nouns. As was already shown above, case marking and scrambling over the subject enables the bare noun to scope above negation. Although Korean allows for intermediate scrambling of caseless objects, they have to reconstruct for interpretation. The context in (153) enforces low scope of the object, both case-marked and caseless objects are licensed. There is, in fact, no need to scramble the object, but we want to keep the contrast between (153) and (154) as minimal as possible. For a context that enforces wide scope of the object, as in (154), only the case-marked object is licensed. (153) Context ¬∃: Yusu’s friend is selling flowers. Yusu looked at all of them but decided not to buy any. a. Kkoch-ul1 Yusu-ka ⟨kkoch-ul1 ⟩ sa-ci anh-ass-ta. flower-ACC Yusu-NOM buy-CI NEG-PST-DECL ‘Yusu did not buy any flowers.’
Korean
b. Kkoch1 Yusu-ka ⟨kkoch1 ⟩ sa-ci anh-ass-ta. flower Yusu-NOM buy-CI NEG-PST-DECL ‘Yusu did not buy any flowers.’ (ii)
∀ > ∃ via passive or scrambling+ovoru a. Nethiki [ella patient-um]1 oru doctor-al ⟨DP1 ⟩ paar-ka-pat-atu. yesterday all patient.NOM-ADD a doctor-INST see-INF-PASS-3SG.N ‘Yesterday, every patient was seen by a doctor.’
Tamil ∃∀,∀∃
b. Nethiki [ella patient-ai-um]1 ovoru doctor-um ⟨DP1 ⟩ paar-t-aar. ∃∀,∀∃ yesterday all patient-ACC-ADD each doctor.NOM-ADD see-PST-3SG.HON ‘Yesterday, each doctor saw all patients.’
170
PNI-PROPERT Y I: RESTRICTION TO LOW SCOPE
(154) Context ∃¬: Yusu’s friend has only a few flowers left to sell and he wants to sell everything by the end of the day. Yusu decides to buy some of them but not all. So there is at least one flower he did not buy. a. Kkoch-ul1 Yusu-ka ⟨kkoch-ul1 ⟩ sa-ci anh-ass-ta. flower-ACC Yusu-NOM buy-CI NEG-PST-DECL ‘Yusu did not buy a specific flower.’
Korean
b. #Kkoch1 Yusu-ka ⟨kkoch1 ⟩ sa-ci anh-ass-ta. flower Yusu-NOM buy-CI NEG-PST-DECL ‘Yusu did not buy any flowers.’ I provide the LF trees for the caseless nouns in (155) and (156). The lack of casemarking signals the presence of PNI-DP. I use syntactic reconstruction, i.e. higher copy neglection, to derive the low scope reading in (153b), recall Section 5.3 for an overview over reconstruction theories. The wide scope reading is blocked in (154b) since PNI-DP cannot be interpreted outside the event domain.¹⁴ (155) ¬ > ∃ reading for (153b) NegP⟨t⟩ ¬ e z[ FLOW (z) BUY (z)(e) AG (yusu)(e)] P NI -DP 2 ⟨⟨e, ⟨v, t⟩⟩, ⟨v, t⟩⟩ λ Q λ e z[ FLOW (z) Q(z)( e)]
NegP⟨t⟩ ¬ e z[ FLOW (z) BUY(z)(e) AG (yusu )(e)] Neg ⟨t,t⟩ ¬p[ p]
vP⟨t⟩ e z[ FLOW (z) BUY (z)(e) AG (yusu)(e)]
vP⟨v,t⟩ λ e z[ FLOW (z) BUY (z)(e) AG (yusu)(e)] DP Yusu
v ′⟨e,⟨v,t⟩⟩ λ xλ e z[ FLOW(z) BUY (z)(e) AG (x)(e)] VP⟨v,t⟩ λ e z[FLOW(z) BUY (z)(e)]
⟨DP2 ⟩ ⟨⟨e, ⟨v, t⟩⟩, ⟨v, t⟩⟩ λ Q λ e z[ FLOW (z) Q(z)(e)]
v ⟨e,⟨v,t⟩⟩ λ xλ e[ AG (x)(e)]
V ⟨e,⟨v,t⟩⟩ λ xλ e[ BUY (x)(e)]
¹⁴ The scope contexts contain the long form of negation which heads its own projection, following Han et al. (2007).
5.4 CASE STUDIES
171
(156) *∃ > ¬ reading for (154b) [PNI-DP2⟨⟨e,⟨v,t⟩⟩,⟨v,t⟩⟩ type clash [μ2 ... ¬ ∃ [vP DP [VP ⟨DP2 ⟩ buy]]]⟨e,t⟩ ] The case-marked bare nouns constitute existential quantifiers, derived with D∃ . As the discourse new and homogeneity contexts have shown above, there is an independent need for a silent D∃ head. Like any other D head, this D head requires case. As a generalized quantifier, the case-marked bare noun has to QR out of the event domain, it can do so targeting positions above (158) or below (157) negation. In contrast to the other PNI languages we have discussed in the previous sections, this movement has to happen overtly. (157) ¬ > ∃ reading for (153a) [¬ DP1⟨⟨e,t⟩,t⟩ [μ1 ... ∃ [vP DP [VP ⟨DP1 ⟩ buy]]]⟨e,t⟩ ]
(158) ∃ > ¬ reading for (154a) [DP1⟨⟨e,t⟩,t⟩ [μ1 ... ¬ ∃ [vP DP [VP ⟨DP1 ⟩ buy]]]⟨e,t⟩ ] We can replicate the scope data with universal quantifiers. This time we use continuations to detect possible scope readings. The first continuation in (159) and (160) supports the distributive reading, i.e. the existential takes scope under the universal quantifier. Similar to the negation contexts above, bare nouns allow for such a continuation whether they are case marked or not. The second continuation is licensed if the existential scopes above the universal quantifier. This continuation is allowed for case-marked but blocked caseless bare nouns. In other words, PNI-ed arguments obey the low scope restriction. (159) Haksayng-ul1 student-ACC
motun all
kyoswunim-i professor-NOM
⟨haksayng-ul1 ⟩
∀∃,∃∀
mann-ass-e. meet-PST-INT ‘A ll professors met a student.’ ... Gereon-un Imke-lul, Barbara-nun Luise-lul, kuliko Jochen-un Gereon-TOP Imke-ACC Barbara-TOP Luise-ACC CONJ Jochen-TOP Jelena-lul mann-ass-e. Jelena-ACC meet-PST-INT ‘Gereon met Imke, Barbara met Luise, and Jochen met Jelena.’ ... Ku haksayng-un Imke-i-ya. That student-TOP Imke-COP-INT ‘That student is Imke.’
172
PNI-PROPERT Y I: RESTRICTION TO LOW SCOPE
(160) Haksayng1 motun kyoswunim-i ⟨haksayng1 ⟩ mann-ass-e. student all professor-NOM meet-PST-INT ‘A ll professors met a student.’
∀∃,*∃∀
... Gereon-un Imke-lul, Barbara-nun Luise-lul, kuliko Jochen-un Gereon-TOP Imke-ACC Barbara-TOP Luise-ACC CONJ Jochen-TOP Jelena-lul mann-ass-e. Jelena-ACC meet-PST-INT ‘Gereon met Imke, Barbara met Luise, and Jochen met Jelena.’ ... # Ku haksayng-un Imke-i-ya. That student-TOP Imke-COP-INT ‘That student is Imke.’ LF structures for readings in (160) are given in (161) and (162). Again, the inverse linear order of subject and object is not necessary to create ∀∃ readings, we simply use the scrambled version to maintain the minimal pair. Syntactic reconstruction is employed to trigger the reconstructed reading in (160); see (161). The wide scope reading is blocked due to PNI-DP’s inability to be interpreted outside the event domain, shown in (162). (161) ∀ > ∃ reading of (160) vP⟨t⟩ y[ PROF (y) → e z[ STUD (z) AG (y)( e)]] P NI -DP 2 ⟨⟨e, ⟨v, t⟩⟩, ⟨v, t⟩⟩ λ Q λ e z[ STUD (z) Q(z)(e)]
MEET (z)( e)
vP⟨t⟩ y[ PROF (y) → e z[ STUD (z) MEET (z)(e) AG (y)(e)]] vP⟨e,t⟩ DP 1 λ x e z[ STUD (z) MEET (z)(e) ⟨⟨e, t⟩, t⟩ AG ( g t 1 x (1))(e)] λ Q y[ PROF (y) → Q(y)] μ1
vP⟨t⟩ e z[ STUD (z) MEET (z)(e) AG ( g(1))(e)] vP⟨v,t⟩ λ e z[ STUD (z) MEET (z)(e) AG ( g(1))(e)] ⟨DP1 ⟩
v ′⟨e,⟨v,t⟩⟩ λ xλ e z[ STUD (z) MEET (z)(e) AG (x)(e)]
VP ⟨v,t⟩ λ e z[ STUD (z) MEET (z)(e)] ⟨DP2 ⟩ ⟨⟨e, ⟨v, t⟩⟩, ⟨v, t⟩⟩ λ Q λ e z[ STUD (z) Q(z)(e)]
v ⟨e,⟨v,t⟩⟩ λ xλ e[ AG (x)(e)]
V ⟨e,⟨v,t⟩⟩ λ xλ e[ MEET (x)(e)]
5.4 CASE STUDIES
173
(162) *∃ > ∀ reading of (160) [PNI-DP2⟨⟨e,⟨v,t⟩⟩,⟨v,t⟩⟩ t.cl. [μ2 ..DP1⟨⟨e,t⟩,t⟩ [μ1 ..∃ [vP ⟨DP1 ⟩ [VP ⟨DP2 ⟩ meet]]]⟨e,t⟩ ]⟨e,t⟩ ] The case-marked bare nouns can interact scopally with universally quantified subjects, which can be observed in (159). We predict this interaction since both arguments constitute generalized quantifiers, which forces them to undergo typedriven movement outside of the event domain. They can do so in different orders, thereby creating different scope relations. Korean differs here from the other PNI languages we have looked at so far since object movement has to happen overtly. Under both readings, given in (163) and (164), DP2 moves overtly out of vP. The subject DP1 can only move covertly, where subject wide scope is presented in (163) and subject narrow scope in (164). (163) ∀ > ∃ reading of (159) [DP1⟨⟨e,t⟩,t⟩ [μ1 ... DP2⟨⟨e,t⟩,t⟩ [μ2 ... ∃ [vP ⟨DP1 ⟩ [VP ⟨DP2 ⟩ meet]]]⟨e,t⟩ ]⟨e,t⟩ ] (164) ∃ > ∀ reading of (159) [DP2⟨⟨e,t⟩,t⟩ [μ2 ... DP1⟨⟨e,t⟩,t⟩ [μ1 ... ∃ [vP ⟨DP1 ⟩ [VP ⟨DP2 ⟩ meet]]]⟨e,t⟩ ]⟨e,t⟩ ] We can additionally verify scope sensitivity for caseless matrix subjects in Korean. The judgements for the contexts in (165) and (166) were tested with an online study with 54 Korean native speakers, ranging from ages 16 to 65. Whereas casemarked subjects were accepted in both ¬∃ and ∃¬ contexts, caseless subjects were licensed only in the former. The ∃¬ in (166) context was carefully set-up, so not to create a weak definite reading for the subject. This reading would not be licensed for independent reasons since, as the reader might recall from Section 4.3.4, weak definites only allow for case drop in the presence of ku. The context ensures that knowledge about the subject’s referent is not shared by speaker and hearer. It is also implied that the domain of the existential quantifier is not necessarily reduced to a singleton set, as there might have been more buses which did not come.¹⁵ (165) Context ¬∃: Kangwu lives very close to the bus stop where five buses are running. Since yesterday, the bus companies have gone on strike and all the buses stopped running. Kangwu’s younger sister, Jiswu, does not know about this. She wonders why today it is more silent than other days. Kangwu says to Jiswu... ... Pesu-ka an tany-e bus-NOM NEG run-INT ‘No bus is runnning.’
Korean
¹⁵ Nouns unmarked for number are number neutral in Korean (independent of case marking). Moreover, it is not unusual for Korean speakers to assume that buses can run every 2–3 minutes, thereby enabling a plural reading in (166).
174
PNI-PROPERT Y I: RESTRICTION TO LOW SCOPE ... Pesu an tany-e bus NEG run-INT ‘No bus is running.’
(166) Context ∃¬: Kangwu and Hannah are close classmates and live next door to each other. They usually take the same bus number 107 at the nearby bus stop to go to University. One day in the morning, Kangwu is waiting for bus 107, but for some reason buses/a bus did not come. Hannah woke up late today and comes to the bus stop 20 minutes later, and wonders why Kangwu is still waiting at this stop. Kangwu says to Hannah... ... Pesu-ka an wa-ss-e. bus-NOM NEG come-PST-INT ‘A bus has not come.’
Korean
... # Pesu an wa-ss-e. bus NEG come-PST-INT ‘A bus has not come.’ The readings for caseless bare nouns are provided in (167) and (168). Since it is unclear whether the subject has in fact moved overtly above negation, we show the in-situ derivation in (167). Consistent with Turkish caseless bare subjects, case drop is triggered by the presence of PNI-DP which can be interpreted in situ but not outside the event domain, shown in (168). (167) ¬ > ∃ reading in (165)
NegP⟨t⟩ ¬ e z[ BUS (z) COME (e) AG (z)(e)] Neg ⟨t,t⟩ ¬p[ p]
vP⟨t⟩ e z[ BUS (z) COME (e) AG (z)(e)] vP⟨v,t⟩ λ e z[ BUS (z) COME (e) AG (z)(e)] P NI -DP ⟨⟨e, ⟨v, t⟩⟩, ⟨v, t⟩⟩ λ Q λ e z[ BUS (z) Q(z)(e)]
v ′⟨e,⟨v,t⟩⟩ λ xλ e[ COME (e) AG (x)(e)] VP⟨v,t⟩ λ e[ COME (e)]
v ⟨e,⟨v,t⟩⟩ λ xλ e[ AG (x)(e)]
5.4 CASE STUDIES
175
(168) *∃ > ¬ reading in (166) [PNI-DP2⟨⟨e,⟨v,t⟩⟩,⟨v,t⟩⟩ type clash [μ2 ... ¬ ∃ [vP ⟨DP2 ⟩ [VP come]]]⟨e,t⟩ ] In contrast to Turkish, case-marked bare subjects in Korean can get an indefinite interpretation. The presence of D∃ forces the subject to QR out of the event domain, either above or below negation, shown in (169) and (170). (169) ¬ > ∃ reading in (165) [¬ DP1⟨⟨e,t⟩,t⟩ [μ1 ... ∃ [vP ⟨DP1 ⟩ [VP come]]]⟨e,t⟩ ] (170) ∃ > ¬ reading in (166) [DP1⟨⟨e,t⟩,t⟩ [μ1 ... ¬ ∃ [vP ⟨DP1 ⟩ [VP come]]]⟨e,t⟩ ] As was established in Section 4.3.4, (post-nominal) numeral classifiers act as D heads and are in complementary distribution with PNI-D. We will now provide the semantic composition of the numeral classifier construction. Arguments in classifier languages have been argued to have a mass noun interpretation, due to the fact that they often lack a number distinction (Borer 2005c; Chierchia 1998). If we take (171) to be the basic structure of a property containing the individuals a, b, c, we can form sums such as a ⊕ c out of the atomic parts. Whereas properties in number-marking languages can hold for pluralities and singularities, properties in classifier languages contain sums as well as atomic parts, very much like mass nouns in number-marking languages. Since counting requires isolated units, mass nouns in number-marking languages need additional measure units in order to be counted, in contrast to count nouns. Consequently, these measure units—in other words classifiers—have to occur obligatorily in counting contexts within classifier languages. They take a mass property and return a property which can be counted, often in the form of reducing the property to the atomic domain. (171) Parthood relations between individuals contained in a set
a⊕b⊕c sums
atoms
a⊕b
a⊕c
b⊕c
a
b
c
mass
While there is the plural marker -tul in Korean, Kim and Melchin (2018a) convincingly argue for its status as an nP modifier, based on (i) its optional occurence in plural contexts (ii) its non-obligatory participation in agreement processes and (iii) the number-neutral interpretation of arguments in the absense of the pluralmarker; see (172). Similar to what Wiltschko (2008) concludes for Halkomelem
176
PNI-PROPERT Y I: RESTRICTION TO LOW SCOPE
and Kramer (2015) for Amharic, Kim and Melchin (2018a) analyse Korean -tul as a modifier and not a functional head. (172) Kim and Melchin (2018a: 3–10) a. salam(-tul) ney myeng human-PL four CL ‘four people’ b. i-tul/ce-tul/ku-tul sakwa(-tul) this-PL/that-PL/the-PL apple-PL ‘these/those/the apple(s)’ c. sakwa apple ‘an apple/apples’ d.
Korean OPTIONAL IN PL CONTEXT
OPTIONAL CONCORD
NUMBER NEUTRAL
nP [plural]
nP √sakwa n
STRUCTURE
We take the number neutral interpretation in (172c) as indicative of a missing number distinction in Korean which is the cause for the obligatory occurence of classifiers in counting contexts. Along with Ahn (2018), we analyse classifiers as function which make properties denote in the atomic domain; see (173).¹⁶ (173) ⟦Dcl ⟧ = λP⟨e,t⟩ λz[P(z) ∧ AT(z)] This is a precondition for the semantic composition with numerals. Following the arguments presented in Section 4.3.4, numerals are merged as specifiers of the classifier head and eventually turn the NP into an existential quantifier. Hence, we predict post-nominal numeral classifiers to create indefinite readings only, which is in line with Kang’s (2002) observations, considering the restriction to sloppy readings in VP-ellipsis contexts and the impossibility to co-occur with topic marker -nun. I give a sample derivation in (174).¹⁷ ¹⁶ Ahn (2018) additionally attributes classifiers the function to turn kinds into sets of object level predicates. The full entry for classifiers is given below. (i) Ahn (2018: 32) ⟦Cl⟧ = λk⟨s,e⟩ λx[AT(∪ k(x))] This analysis presupposes that every Korean NP is first shifted into a kind interpretation via ∩ . The result needs to be lowered via ∪ which returns the set of singular and plural instantiations of the kind. We will remain agnostics about the application of ∪ and ∩ and henceforth use the simpler notation in (173), akin to Dayal’s (2011) treatment for NumPs in Hindi. ¹⁷ We might wonder at this point why the other PNI languages we have discussed so far do not require classifiers in the context of numerals. An obvious way to account for this is that those languages let NPs
5.4 CASE STUDIES
177
(174) salam ney myeng (172a):
DP ⟨⟨e, t⟩, t⟩ λ Q z[ HUMAN(z) AT (z) z = 4 Q(z)]
#P ⟨⟨e, t⟩, ⟨⟨e, t⟩, t⟩⟩ λ Pλ Q z[P(z) z = 4 Q(z)]
D′ ⟨e, t⟩ λ z[ HUMAN(z) D cl ⟨⟨e, t⟩, ⟨e, t⟩⟩ λ Pλ z[P(z) AT(z)]
AT(z)]
NP ⟨e, t⟩ λ y[ HUMAN(y)]
Since numeral classifier constructions seem to function exclusively as numerals, it is no surprise that they are not licensed in generic contexts; see (175). A bare noun has to be used instead.¹⁸ (175) Generic readings a. Kay-ka cic-nun-ta. dog-NOM bark-PRS-DECL ‘Dogs barks.’ b. Kay-ka han mali cic-nun-ta. dog-NOM one CL bark-PRS-DECL ‘One dog barks.’
Korean GENERIC
*GENERIC
Turning now to the scope contexts, we can show that, although numeral classifiers can be unmarked for case, they do not undergo PNI, i.e. there is no correlation between case drop and a restriction to low scope. This scope behaviour is predicted by the underlying structure we established above. Since the numeral in the numeral classifier construction turns the NP into an argument and is syntactically selected by the classifier, there is no room for the PNI determiner to enter the derivation. The context in (176) requires the numeral to scope below negation, which is possible for both caseless and case-marked arguments. Similar to what we have noticed for the other PNI languages above, numerals can only scope below negation if they are narrowly focused. The crucial context is given in (177) where case drop on the numeral construction does not restrict it from taking wide denote in the atomic domain without the addition of a classifier. This is assumed by Dayal (2011) for Hindi and has recently been proposed by Sağ (2018) for Turkish. ¹⁸ While Ch. Lee (1989) agrees to use bare nouns in generic contexts, he prefers the subject to be marked with topic marker -nun.
178
PNI-PROPERT Y I: RESTRICTION TO LOW SCOPE
scope. The data in (178) show that the scope behaviour we see in (177) applies to numerals generally. (176) Context ¬1: Yusu’s friend has only three flowers left to sell and he wants to sell everything by the end of the day. But Yusu looked at all three of them and decided they were too rotten to buy any of them. [Kkoch(-ul) han-songi]1 Yusu-ka ⟨DP1 ⟩ sa-ci anh-ass-ta. flower-ACC one-CL Yusu-NOM buy-CI NEG-PST-DECL ‘Yusu did not buy [one]F flower.’ (177) Context 1¬: Yusu’s friend wants to sell three flowers and Yusu bought two from him. So there is one flower Yusu did not buy. [Kkoch(-ul) han-songi]1 Yusu-ka ⟨DP1 ⟩ sa-ci anh-ass-ta. flower-ACC one-CL Yusu-NOM buy-CI NEG-PST-DECL ‘One flower, Yusu did not buy.’ (178) Context 2¬: A friend of Yusu wants to sell four flowers and after looking at them Yusu bought two flowers from him. So there are two flowers which Yusu did not buy from him. [Kkoch(-ul) twu-songi]1 Yusu-ka ⟨DP1 ⟩ sa-ci anh-ass-ta. flower-ACC two-CL Yusu-NOM buy-CI NEG-PST-DECL ‘Two flowers, Yusu did not buy.’ The LF structures for the two readings a numeral classifier construction can provide are given in (179) and (180). (179) ¬ > ∃ reading for (176) [¬ DP1⟨⟨e,t⟩,t⟩ [μ1 ... ∃ [vP ⟨DP1 ⟩ [VP come]]]⟨e,t⟩ ] (180) ∃ > ¬ reading for (177) and (178) [DP1⟨⟨e,t⟩,t⟩ [μ1 ... ¬ ∃ [vP ⟨DP1 ⟩ [VP come]]]⟨e,t⟩ ] In this section, we have looked at Korean bare nouns and numeral classifier phrases. While both noun types can lose case marking, only the former triggers a low scope restriction with case drop. Similar to Hindi and Tamil, numeral contructions are in complementary distribution with PNI-DPs in Korean. Whereas we analysed numerals as instantiating D# in Hindi and Tamil, it is a classifier which instantiates the D head in Korean with numerals being merged in specifier position. Classifiers are needed on independent grounds since Korean does not show
5.4 CASE STUDIES
179
a number distinction. The next section will consider a language which does not allow arguments to lose case at all, but nevertheless shows scope restrictions for bare plurals and indefinite phrases.
5.4.5 German So far we have limited our investigation to PNI languages which show a correlation between case loss and scope inertness. German, as argued by Frey (2015), displays PNI effects for bare plurals and non-specific indefinites, albeit without an effect in case marking. Frey’s diagnostics consist of scope inertness of certain noun types in combination with positions these noun types can occupy in the clause. He observes that there is only a small class of arguments, consisting of nonspecific indefinites and bare plurals, that can follow manner adverbs and negation, shown for the adverb wunderbar in (181a) and (181b). In contrast, specific indefinites and universal quantifiers (see (181c) and (181d)), must precede the manner adverb. (181) Frey (2015: 237–238)
German
a. Otto hat heute wunderbar eine charmante Mozart-Sonate Otto has today beautifully a charming Mozart sonata gespielt. played ‘Today Otto played a charming Mozart sonata beautifully.’ b. Otto hat im letzten Jahr wunderbar Mozart-Sonaten gespielt. Otto has in.the last year beautifully Mozart sonata played ‘Last year Otto played Mozart sonata beautifully.’ c.??Otto hat heute wunderbar eine bestimmte Mozart-Sonate Otto has today beautifully a certain Mozart sonata gespielt. played ‘Today Otto played a certain Mozart sonata beautifully.’ d.??Otto hat im letzten Jahr wunderbar jede Mozart-Sonate Otto has in.the last year beautifully every Mozart sonata gespielt. played ‘Last year Otto played every Mozart sonata beautifully.’ Moreover, non-specific indefinite objects are unable to scope above a universally quantified subject, if they follow a low manner adverb, shown in (182).
180
PNI-PROPERT Y I: RESTRICTION TO LOW SCOPE
(182) Frey (2015: 238) German Jeder hat heute wunderbar eine charmante Mozart-Sonate everyone has today beautifully a charming Mozart sonata gespielt. played ‘Today everybody played a charming Mozart sonata beautifully.’
∀∃,*∃∀
If manner adverbs are vP/VP adjuncts and quantifiers cannot move out of the event domain covertly but must do so overtly, we can explain why (181c) and (181d) are unacceptable. Consequently, the objects in (181a), (181b), and (182) must be pseudo-incorporated.¹⁹ Frey (2015) does not provide a scope context for bare plurals, presumably because scopal restrictions of existentially interpreted bare plurals have been known to exist independently of any morpho-syntactic PNI effect (Carlson 1977; Chierchia 1998; Dayal 2004). This can be shown for English; see e.g. (183), where the bare plural object cannot take scope above negation, in contrast to the singular indefinite. (183) Chierchia (1998: 368) a. I didn’t see a spot on the floor.
¬∃,∃¬
b. I didn’t see spots on the floor.
¬∃,*∃¬
¹⁹ Frey has a different analysis. He uses wh-indefinites to identify the position of low manner adverbs and negation. Wh-indefinites are often assumed to be immobile in German, due to data like (i) where two wh-indefinites have to maintain their base orders. Since manner adverbs and negation can linearly follow wh-indefinites (see (ii)), Frey concludes that they must be able to enter the derivation before any of the wh-arguments is merged in their base positions. (i) Frey (2015: 230) dass (*was) wer (was) lesen will that something someone something to-read wants ‘that someone wants to read something’
German
(ii) Frey (2015: 230,235) a. Maria schaut sich gerade was ganz genau an. Maria looks.at RFL currently something very carefully PRT ‘Maria is currently looking at something very carefully.’
German
b. weil hier wer was nicht beachtet hat since here someone something not noticed has ‘since someone has not noticed something here’ Since bare plurals and non-specific indefinites can follow manner adverbs and negation, they are analysed as parts of a complex predicate. (iii) Frey (2015: 226,229) dass Max heute [V [VP [NP Karten] spielen] wird] that Max today cards play will ‘that Max will play cards today’
German
5.4 CASE STUDIES
181
Hence, scope tests with bare plurals, PNI-ed or not, will very likely return the same result, i.e. they will reveal the general scopelessness of bare plurals. Instead, Frey (2015) provides a different test for PNI, one that has received little attention in this book so far. Pseudo-noun incorporation is generally productive only with noun-verb combinations which are in some sense conventionalized or frequent (Asudeh and Mikkelsen 2000; Barrie and Li 2015; Dayal 2011; Mithun 1984). Frey shows this restriction for bare plurals with minimal contrasts like the ones in (184) and (185). Peeling potatoes and watering flowers are considered frequent activities. Bare nouns in immediate pre-verbal positions are thus quick to receive PNI-readings. Peeling pears and watching flowers, however, are less common which makes their occurence in those positions slightly odd. (184) Frey (2015: 241) a.
Er will heute nicht Kartoffeln scha¨len. he wants today not potatoes to-peel ‘He does not want to peel potatoes today.’
German
b. ?Er will heute nicht Birnen scha¨len. he wants today not pears to-peel ‘He does not want to peel pears today.’ (185) Frey (2015: 241) a.
Maria muss heute nicht Blumen gießen. Maria has today not flowers to-water ‘Maria does not have to water any flowers today.’
German
b. ?Maria muss heute nicht Blumen beobachten. Maria has today not flowers to-watch ‘Maria does not have to watch any flowers today.’ Dayal (2011: 164) proposes that the conventional aspect of PNI can be encoded in the form of a presupposition, tied to the specialized PNI-denotations which are stored for some verbs in the lexicon.²⁰ Dayal speculates that the presupposition has to be formulated along the lines of ‘P is/are often v’d’, where P is the property argument PNI verbs select for. This idea can be translated into the current approach by adding a presupposition to PNI-DP, the PNI-determiner, in the form of ‘e must be conventionalized/frequent’. Tying the presupposition to the quantifier might even be considered more intuitive since quantifiers express relations ²⁰ Another route taken to encode conventionalized meanings is to treat verbs which allow for PNI readings as event kinds. Since kinds have to be established, the requirement follows automatically. This analysis is proposed by Schwarz (2014) for weak definites and by Gehrke (2015) for event modifiers of adjectival passives in German.
182
PNI-PROPERT Y I: RESTRICTION TO LOW SCOPE
between sets. In other words, the nominal denotation is just as relevant to successful PNI readings as the verbal denotation, as (184) and (185) clearly show. With this presupposition in place, we are able to explain the contrasts in (184) and (185). The natural kind reading is filtered out through the presence of objectlevel predicates. Existential readings are made possible via pseudo-incorporation, albeit with a restriction to low scope. This scopal behaviour is predicted by the denotation of the PNI determiner. The oddness of (184b) and (185b) is an effect of the presupposition, encoded in the lexical entry of the PNI determiner, which requires the event expressed to be a conventionalized or at least a frequent activity. Bare plurals and non-specific indefinites cannot undergo scrambling, which is shown with remnant VP-fronting in (186) and (187). The availability of full VP-fronting is shown in (186a), while the unacceptability of (186b) is due to the bare plural having to vacate the VP first, in order for the remnant VP to undergo movement to spec,CP. The underlying structure for (186b) is given in (186c). An indefinite equivalent to (186b) can be provided with (187a), where scrambling of the object leads to a specific interpretation. I will interpret this effect as a ban on scrambling of non-specific indefinite objects.²¹ (186) Frey (2015: 228) a.
Karten spielen wird Max heute. cards play will Max today ‘Max will play cards today.’
b. *Spielen wird Max heute Karten. play will Max today cards ‘Max will play cards today.’ c.
[CP [VP ⟨Karten1 ⟩ spielen]2 [C ʹ wird Max heute [Karten]1 ⟨VP2 ⟩]]
²¹ Frey takes a different path to explain the data in (186) and (187). He operates under the assumption that verbs on their own constitute VPs and can move to spec,CP, again based on observations with respect to wh-indefinites (Frey 2015: 233), but see also Frey and Tappe (1991) and Fanselow (1992). While this assumption might explain the unacceptability of (186b), since it would require excorporation out of the complex predicate the PNI-ed argument has formed with the verb, it leads him to a rather ad hoc stipulation when it comes to a movement asymmetry with respect to topicalization. Interestingly, while bare plurals do not allow for scrambling, they can nevertheless undergo movement to spec,CP. (i) Frey (2015: 228) [Karten]1 wird Max heute [VP ⟨Karten1 ⟩ spielen] cards will Max today play ‘Max will play cards today.’ Frey must assume that PNI-ed arguments can excorporate out of the complex predicate, whereas verbs cannot. I will provide a different explanation for the movement patterns of PNI-ed arguments, one that ties them to VP-movement patterns in German; see Chapter 7 for more details.
5.4 CASE STUDIES
(187) Frey (2015: 239)
183
SPECIFIC!
a. Gespielt hat Otto heute eine charmante Mozart-Sonate wunderbar. played has Otto today a charming Mozart sonata beautifully ‘Today Otto played a charming Mozart sonata beautifully.’ b. [CP [VP ⟨...Sonate1 ⟩ gespielt]2 [C ʹ hat Otto heute [eine... Sonate]1 wunderbar ⟨VP2 ⟩]] The movement restriction is the second part of Frey’s diagnostic, which leads him to conclude that bare plurals and non-specific indefinites are pseudo-incorporated in German. We can capture the movement behaviour by making both noun types serve as arguments to the PNI determiner.²² Since PNI-DPs are partially verbal, they will exhibit the movement patterns of VPs, which have been observed to resist scrambling in German (Grewendorf 1995; Grewendorf and Sternefeld 1990; Müller 1998; von Stechow and Sternefeld 1988). Thus, the effects in (186) and (187) do not result from a semantic incompatibility. Recall from Section 5.3 that PNI-ed arguments can reconstruct into their base positions for which we have already seen evidence in Hindi and Korean. Rather, there is a syntactic reason why PNI-ed argument do not scramble in German. The scope restriction in (182) is predicted by PNI-DPs inability to be interpreted outside the event domain. I follow Kobele and Zimmermann (2012) in assuming that German ein can functions as an indefinite determiner as well as a numeral expression. Like Turkish bir, German ein can create generic readings; see (188a). (188) Gerstner-Link (1995: 79) a. Ein Hund ist ein Haustier. a dog is a pet ‘A dog is a pet.’
GENERIC
b. Hunde sind Haustiere. dogs are pets ‘Dogs are pets.’
GENERIC
²² One disadvantage of extending our PNI analysis to the obligatory low scope of bare plurals is that it does not prevent bare singulars from undergoing PNI. While there are some German PNI structures where the object is singular, e.g. Fahrrad fahren (= to ride a bike) or Flugzeug fliegen (= to ride a plane), these seem to be lexicalized and highly restricted, much more so than bare plurals. The fact that low scope existential readings require pluralized bare nouns (or mass nouns) in non-bare argument languages is one of the main motivations for Chierchia (1998: 366) to develop an approach based on derived kind readings. He argues that bare nouns in English and German are type-shifted to kinds if they serve as arguments to kind-level predicates but they can only do so if they are pluralized since singular individuals do not form natural kinds. The low scope existential reading is created by the application of a specialized rule called Derived Kind Predication which ensures low scope existential readings with object-level predicates. This seems to be a clear advantage of the kind based approach over the PNI approach. The kind based approach, on the other hand, is not able to predict the parallel movement patterns of VPs and PNI-ed arguments.
184
PNI-PROPERT Y I: RESTRICTION TO LOW SCOPE
Hence, I will treat the German indefinite/numeral like Turkish bir in Section 5.4.3, i.e. it can function as an indefinite determiner as well as a numeral adjective. In the latter case, the PNI determiner turns the numeral expression into an argument. (189) Indefinite/numeral ein in German a. ⟦ein1 ⟧ = λP⟨e,t⟩ λQ⟨e,t⟩ ∃z[P(z) ∧ Q(z)]
(preliminary)²³
b. ⟦ein2 ⟧ = λz.|z| = 1 I provide LF trees for the readings in (182) with (191) and (192), where I include the relevant manner adverb wunderbar for the semantic composition. With Ernst (2002) and Katz (2003) I assume that adverb adjunction sites are restricted by semantic compatibility. Manner adverbs are thus event modifiers, the denotation for the adverb in question is given in (190). Since the object constitutes a PNI-DP, an additional numeral restriction shows up in the denotation of the PNI-DP in (191) and (192), triggered by the contribution of (189b). (190) ⟦wunderbar⟧ = λP⟨v,t⟩ λe⟨v⟩ [P(e) ∧ BEAUTIFUL(e)] (191) ∀ > ∃ reading of (182) vP⟨t⟩ y[ PERS (y) → e z[M O -S O (z) z = 1 PLAY (z)(e) AG (y)(e) BEAUT (e)]] DP 1 ⟨⟨e, t⟩, t⟩ λ Q y[ PERS (y) → Q(y)]
PLAY
vP⟨e,t⟩ λ x e z[M O -S O (z) z = 1 (z)(e) AG ( g t 1 x (1))(e) BEAUT(e)] µ1
vP⟨t⟩ e z[MO -SO (z) z = 1 PLAY (z)(e) AG ( g(1))(e) BEAUT (e)]
PLAY
vP⟨v,t⟩ λ e z[MO -SO (z) z = 1 (z)(e) AG ( g(1))(e) BEAUT (e)]
Adv ⟨⟨v, t⟩, ⟨v, t⟩⟩ λ Pλ e[P(e) BEAUT (e)] λ e z[MO -SO (z)
vP⟨v,t⟩
z =1
PLAY (z)(e)
AG ( g(1))(e)]
(192) *∃ > ∀ reading of (182) [PNI-DP2⟨⟨e,⟨v,t⟩⟩,⟨v,t⟩⟩ t.cl. [μ2 ..DP1⟨⟨e,t⟩,t⟩ [μ1 ..∃ [vP ⟨DP1 ⟩ [VP ⟨DP2 ⟩ play]]]⟨e,t⟩ ]⟨e,t⟩ ] ²³ We will refine this entry later on to include the specificity contrast.
5.4 CASE STUDIES
185
I will now explain the distribution in (181) and how it can be made to follow within the current system. Since the manner adverb wunderbar serves as either a VP or a vP adverb, we do not expect PNI-ed arguments to cause ungrammaticality if they are spelled-out in a position following the manner adverb, that is inside the event domain. Thus, (181a), (181b), and (182) constitute acceptable word orders. Universal quantifiers and specific indefinites, however, are not licensed in a position following the manner adverb; see (181c) and (181d). Like Korean, German shows scope rigidity effects (Jacobs 1982; Frey 1993). If this means that quantifiers do not raise covertly but must raise overtly, we have an explanation for the unacceptability of (181c) and (181d). The universally quantified object in (181d) is supposed to raise out of the event domain in order to be interpreted and must do so overtly, thereby requiring the object to precede the manner adverb. Since the object follows the adverb in (181d), the sentence is not acceptable. The same explanation can be given for (181c), if we tie the ability of receiving specific interpretations to the nature of a generalized quantifier. All that is left to do is find an implementation for specific readings. Since German ein can also act as an indefinite determiner, I take the indefinite in (181c) to constitute an existential quantifier, as it is preliminarily defined in (189a). There are at least two ways, recently proposed in the literature to derive specific indefinites, one argues in favour of choice functions (Kratzer 1998; von Heusinger 2002), the other is based on indefinites constituting existential quantifiers (Onea and Geist 2011; Schwarzschild 2002). López (2012) argues for Spanish case-marked indefinites that their [+spec] readings depend on their status as a choice function. Following von Heusinger (2002), he proposes that specificity is caused by a referential variable which comes with choice functions and must be anchored to another discourse referent, in most cases the speaker of the utterance. Indexed choice functions express that the referent is known to whoever discourse referent is co-indexed with it. This analysis makes sense for Spanish since case-marked indefinites can stay in-situ and take wide scope, implying that their syntactic position is irrelevant since scope is determined by a covert existential binder over choice functions. German indefinites, on the other hand, show word order effects if they are interpreted as specific—an empirical fact we know since Diesing (1992). Thus, I would like to argue that German is more suitable for accounts which predict indefinites to occur in certain syntactic positions correlating with their specificity. Since generalized quantifiers undergo type-driven movement out of the event domain and indefinites preceding manner adverbs are obligatorily interpreted as specific, I will hard-wire a specificity constraint into the denotation of the indefinite determiner. The PNI determiner, in contrast, does not have such a requirement and thus produces non-specific readings. I follow Onea and Geist (2011) in analysing indefinite determiners as existential quantifiers where specific readings result from a domain narrowing mechanism in the sense of Schwarzschild (2002). This can be done by introducing a salient
186
PNI-PROPERT Y I: RESTRICTION TO LOW SCOPE
function into the restrictor of the existential quantifier. Often functional readings are subcases of plain narrow scope readings which are enforced by modifiers like certain. In (193) for example, only the first continuation represents a natural function, while the second continuation assembles different values and is thus infelicitous. (193) Each husband had forgotten a certain date. ... his wife’s birthday. ...#John had forgotten Mary’s birthday, Michael his wedding anniversary etc. Onea and Geist (2011) give (193) the denotation in (194), where f(y) = x introduces the domain restriction, an anchoring function f from individuals to individuals and an anchor variable y. This referential anchor is added pragmatically, if indefinites do not per se trigger functional readings. The presence of certain, however, enforces functional readings and thus encodes the anchor lexically. The contextually specified function for (194) is λz.ιy[WIFE-BDAY(y, z)], representing the licit continuation in (193). Onea and Geist (2011) additionally encode a presupposition for referential anchors since the use of a certain date requires that (i) there are dates and (ii) f refers to something which is a date. The denotation in (195) provides the lexical entry for a specific indefinite determiner. (194) ⟦193⟧ = ∀y[HUSBAND(y) → ∃x[DATE(x) ∧ f( y) = x ∧ FORGET(y, x)]] a f : λz.ιy[WIFE-BDAY(y, z)] (195) Specific indefinite determiner: (Onea and Geist 2011: 209) ⟦a certain⟧ = λPλQ : ∃ f ∀x[ x ∈ De → P( f(x))]. ∃x[P(x) ∧ f(y) = x ∧ Q(x)] While the anchor variable is bound by the universal quantifier in (194), it can also be a constant, referring to the speaker of the utterance. This is the key to understanding epistemic specificity, which is at work in (181c). It requires both anchor variable and anchor function to be constants along the lines of f in (196), reflecting the intuition that the speaker wants to refer to a particular individual. (196) Epistemically specific indefinite determiner: (Onea and Geist 2011: 210) ⟦a certain⟧ = λPλQ : ∃f ∀x[x ∈ De → P( f(x))].∃x[P(x) ∧ f(y) = x ∧ Q(x)] a f : λz.ιy[z intends to refer to y](speaker) In contrast to English, where the referential anchor is lexically encoded only if the indefinite determiner co-occurs with certain, the German indefinite determiner contains a referential anchor, even without spelling out the German equivalent bestimmt. The final lexical entries for the indefinite/numeral ein are given in (197).
5.4 CASE STUDIES
(197) Indefinite/numeral ein in German
187
( final)
a. ⟦ein1 ⟧ = λPλQ : ∃f ∀x[x ∈ De → P( f(x))].∃x[P(x) ∧ f(y) = x ∧ Q(x)] b. ⟦ein2 ⟧ = λz.|z| = 1 Since the use of ein1 instantiates a generalized quantifier, it must move out of the event domain and thus cannot follow the manner adverb in (181c). This is shown with the sketched LF structure in (198). I also provide the grammatical version in (199) with the LF in (200) where the GQ is scrambled out of vP, thereby preceding the manner adverb. Anchor function f and anchor variable y are realized by constants, as it is made explicit in (188), creating epistemic specificity. In contrast, the numeral variant ein2 is made use of in (181a) and (182), so that together with the PNI determiner the indefinite object receives non-specific, low scope readings.²⁴ (198) *specific reading for (181c) [∃ [vP Adv [vP DP [VP DP⟨⟨e,t⟩,t⟩
type clash
play⟨e,⟨v,t⟩⟩ ]]]]
(199) Scrambling of specific indefinites German Otto hat heute [eine bestimmte Mozart-Sonate]2 wunderbar ⟨DP2 ⟩ Otto has today a certain Mozart sonata beautifully gespielt. played ‘Today Otto played a certain Mozart sonata beautifully.’
²⁴ One might wonder at this point why the use of ein2 plus PNI-D does not trigger ambiguous readings, as PNI-ed arguments are simply not lexicalized for either +spec or −spec. Since referential anchors can also be added pragmatically, we could imagine that PNI-DPs receive functional readings by a pragmatic enrichment operator, as it is sketched by Onea and Geist (2011: 209), triggering specific readings after all in (181a) and (182). The crucial difference between specific readings created by pragmatic operations and lexically encoded specificity as in (197a) is that the latter additionally presupposes the existence of anchor functions. This makes (197a) stronger than a pragmatically created specific reading for PNI-ed arguments, blocking the application of the latter. One is inclined to adhere this blocking effect to the pragmatic principle of Maximize Presupposition (Heim 1991; Percus 2006; Sauerland 2008a). The use of a universal quantifier e.g. can be blocked by a stronger presuppositional determiner: (i) a. *All of John’s eyes are open. b. Both of John’s eyes are open. Notice, however, that in contrast to the minimal pair ⟨all,both⟩ where the competition lies between two determiners which differ overtly, two homophonous exponents compete in the discussion above. The difference between the two indefinites only becomes apparent by the positions they can occupy in the clause.
188
PNI-PROPERT Y I: RESTRICTION TO LOW SCOPE
(200) specific reading for (199)
vP⟨t⟩ f x[x De → MO -SO ( f ( x))] . x[MO -SO ( x) f (y) = x e[ PLAY ( x)( e) AG (otto)(e) BEAUT (e)]]
DP 2 ⟨⟨e, t⟩, t⟩ λ Q f x[x D e → M O -S O ( f ( x))] . x[M O -S O ( x) f (y) = x Q( x)]
vP⟨e,t⟩ λ x e[ PLAY ( g t 2 x (2))(e) AG (otto)(e) BEAUT (e)] μ2
vP⟨t⟩ e[PLAY ( g(2))(e) A G (otto)(e) BEAUT (e)] vP⟨v,t⟩ λe[ PLAY ( g(2))(e) AG (otto)( e) BEAUT (e)] Adv ⟨⟨v, t⟩, ⟨v, t⟩⟩ λ Pλ e[P(e) BEAUT (e)]
vP⟨v,t⟩
λe[ PLAY( g(2))(e)
AG (otto)(e)]
This concludes our discussion of the scopal properties PNI-ed arguments show across languages. We have seen that there is a robust correlation between lack of case marking and restriction to low scope when it comes to Spanish, Hindi, Turkish, Tamil, and Korean. Both case drop and low scope can be attributed to the verbal nature of the PNI determiner. Languages vary, however, as to the types of nouns which can undergo PNI as well as to the scopal properties of the corresponding case-marked nouns. For Tamil, Hindi, and Korean, we have seen that certain noun types, although able to occur without case, do not show a correlated scopal restriction. These cases are analysed as DOM effects, more details are given in Chapter 8. Noun types which do not show PNI effects are analysed as DPs where the D head is in complementary distribution with PNI-D, the PNI determiner. Finally, this section has shown that a PNI language does not necessarily have to exhibit a case marking effect at all. In German, we find a correlation between clausal positions and scopal restrictions for noun types which typically undergo PNI cross-linguistically. Just like the lack of case marking and the scope inertness, we can identify the movement patterns of PNI-ed arguments as a verbal property: German VPs do not undergo scrambling, neither do German PNI-ed arguments. In the next chapter, I will consider two additional diagnostics for PNI-ed arguments, the inability to participate in binding and control relations.
6 PNI-property II: Lack of binding and control This chapter will give a detailed description of pseudo-incorported arguments and their binding and control properties. The current study shows that binding and control relations are consistently blocked for those argument types we have identified in Chapter 4 as being able to undergo PNI. The lack of binding and control is another aspect in which PNI-ed arguments exhibit their non-nominal character. Binding as well as the insertion of binder indices have been described as processes essentially tied to nominal categories (Baker 2004; Büring 2005). Since PNI-ed arguments are headed by a determiner which starts off as a nominal category but turns into a verbal category after it has been c-selected, and constituents can only act as binders once they have entered the derivation, binding applies too late to access the nominal feature of PNI-ed arguments. At the point in the derivation when PNI-ed arguments need to insert a binder prefix, be it via movement (Heim and Kratzer 1998) or in-situ (Büring 2001, 2004, 2005), they constitute verbal categories. Furthermore, if a control relation relies on the controller being able to bind PRO (Chomsky 1981; Koster 1984; Landau 2015, 2017; Manzini 1983), PNI will be incompatible with control predicates. The interaction of binding/control with optional case marking was, to my knowledge, first observed for Spanish.¹ Leonetti (2004) observes that case-marked indefinite arguments are prominent binders for possessive pronouns, whereas without case marking possessives are bound by another antecedent, shown in (1).
(1) Leonetti (2004: 102)
Spanish
a. Devolvieron [a un prisionero]1 a su1 tribu. they.returned DOM a prisoner to his tribe ‘They returned a prisoner to his tribe.’
¹ Spanish shows the following size restriction when it comes to case loss. Pronouns and proper names are obligatorily marked for case (Aissen 2003; Fábregas 2013; Pensado 1995; Torrego Salcedo 1999), the same is true for possessives (Galdeano 2000: 2) as well as demonstratives (von Heusinger 2018: 320). Indefinites are optionally marked, which correlates with a specificity effect (López 2012; García García 2018; von Heusinger and Kaiser 2007). Leonetti (2004) and Bleam (2000) also discuss weak quantifiers including numerals, which pattern with indefinites.
Pseudo-Noun Incorporation and Differential Object Marking. Imke Driemel, Oxford University Press. © Imke Driemel (2023). DOI: 10.1093/oso/9780192866400.003.0006
190
PNI-PROPERT Y II: L ACK OF BINDING AND CONTROL b. Devolvieron [un prisionero]1 a su*1∕2 tribu. they.returned a prisoner to his tribe ‘They returned a prisoner to his tribe.’
López (2012) extends this observation to binding of reflexive pronouns (2) and object control structures (3). In both configurations, objects have to be case marked. (2) López (2012: 41) a.
Spanish
María le entrego´ [a sí mismo]1 [a un hombre]1 . Maria CL.DAT delivered DAT himself DOM a man ‘Maria delivered a man to himself.’
b. * María le entrego´ [a sí mismo]1 [un hombre]1 . Maria CL.DAT delivered DAT himself a man ‘Maria delivered a man to himself.’
(3) López (2012: 53,58) a.
Spanish
Juan forzo´ [a un nin˜o]1 [PRO1 a hacer los deberes]. Juan forced DOM a boy to do.INF the homework ‘Juan forced a boy to do his homework.’
b. * Juan forzo´ [un nin˜o]1 [PRO1 a hacer los deberes]. Juan forced a boy to do.INF the homework ‘Juan forced a boy to do his homework.’ López (2012) furthermore provides a parallel control example for Hindi, shown in (4), while Bhatt (2007) shows the same interaction for adjunct control clauses; see (5).² This last observation is important, as it excludes an explanation for obligatory case marking in control structures that makes reference to the possibility that control predicates simply do not make good PNI predicates. Under this view, control is blocked not because PNI-ed arguments cannot control PRO but rather because control predicates are not able to form a meaning with their objects that would qualify as a conventionalized or frequent activity. Adjunct control offers an opportunity to test control abilities in isolation. ² The alert reader might be puzzled at this point since we so far considered indefinites to not undergo PNI in Hindi. We will come back to the observation made in (5) when we discuss the control behaviour of DOM-marked noun types in Tamil in Section 8.1.
PNI-PROPERT Y II: L ACK OF BINDING AND CONTROL (4) López (2012: 127) a.
191 Hindi
Radhaa-ne ciRiyaa-ko1 [PRO1 kitaab parh-ne-par] majbuur Radhaa-ERG bird-ACC book be-INF-LOC force ki-yaa. do-PERF ‘Radhaad forced a bird to read a book.’
b. * Radhaa-ne ciRiyaa1 [PRO1 kitaab parh-ne-par] majbuur ki-yaa. Radhaa-ERG bird book be-INF-LOC force do-PERF ‘Radhaad forced a bird to read a book.’ (5) Bhatt (2007: 17)
Hindi
a. Mina-ne1 bazaar-me [ek sailaani-ko]2 [PRO1∕2 naacte-hue] Mina-ERG market-in a/one tourist-ACC dancing-while dekh-aa. see-PFV ‘Mina saw a tourist while dancing.’ b. Mina-ne1 bazaar-me [ek sailaani]2 [PRO1∕??2 naacte-hue] Mina-ERG market-in a/one tourist dancing-while dekh-aa. see-PFV ‘Mina saw a tourist while dancing.’ The implementation of the ban on binding and control relations is fairly simple and straightforward. I follow Büring (2001, 2004, 2005) in that a binding relation is established by insertion of a binder prefix β at LF; see (6a). The DP minimally c-commanding the binder prefix acts as the semantic binder via the Binder Index Evaluation rule. The rule was already given in Section 3.3 and is repeated here for convenience; see (6b). (6) Assumptions for Binding a. β-insertion:
DP XP
(Büring 2001: 56, Büring 2004: 25) LF
DP
XP
βn XP where n is an index, and DP occupies an A-position b. Binder Index Evaluation rule (BIER): g g[n↦x] ⟦βn XP⟧ = λx.[⟦XP⟧ (x)]
192
PNI-PROPERT Y II: L ACK OF BINDING AND CONTROL
Since binding is tied to nominal categories Büring (2005: 1), I assume that insertion of the binder prefix β is blocked for verbal categories. I show the two different binding scenarios schematically in (7). Although both DPs and PNI-DPs start off as nominal categories, only the latter turn into verbal categories, after they have entered the derivation. (7) Binding for proper arguments and PNI-ed arguments a. β-insertion for DP
DP D CASE:VAL
... LF
D
DP D CASE:VAL
... XP
βn
D
b. No β-insertion for PNI-DP
PNI-DP D CASE:VAL V
... LF
PNI-DP D CASE:VAL
... βn
XP
V
I use possessive as well as reflexive pronouns to diagnose binding. Possessive pronouns are mostly tested for object PNI since the reflexive pronouns in Tamil and Mongolian are subject-oriented. I take the Spanish example in (1) and demonstrate the successful binding relation for case-marked direct objects in (1a) with the LFs in (8). With López (2012: 40), I assume that the direct object A-moves over the indirect object from which it can bind into it. I take this position to be an inner specifier of vP. Crucial for the analysis is that the assignment function g has a sorted domain, i.e. g(tn ) ≠ g(pronn ), and that the μ-operator only maps movement indices to variables within the semantic object of its sister; see also footnote 3. Pronoun indices can only be mapped to variables by the binder prefix β. The distinction between the types of indices is made explicit in this chapter (t2 vs 2). Since case-marked indefinites constitute pure nominal arguments, the direct object can insert a binder prefix, after it has moved to Spec, vP. This is the crucial step that enables the possessive pronoun contained in the indirect object to be bound.
PNI-PROPERT Y II: L ACK OF BINDING AND CONTROL
193
(8) Successful binding in (1a) a. Step 1: Semantic composition of ApplP ApplP⟨v,t⟩ λe[DEL(g(t2))(e) REC(ιy.TRIBE(y)(g(2)))(e)] PP⟨e⟩ ιy.TRIBE(y)(g(2))
Appl′⟨e,⟨v,t⟩⟩ λxλe[DEL(g(t2))(e) REC(x)(e)] Appl⟨e,⟨v,t⟩⟩ λxλe[REC(x)(e)]
VP⟨v,t⟩ λe[DEL(g(t2))(e)] V⟨e,⟨v,t⟩⟩ λxλe[DEL(x)(e)]
⟨DP2⟩
b. Step 2: Binding within vP vP⟨v,t⟩ λ e[ DEL ( g(t 2 ))(e) REC (ιy. TRIBE (y)( g(t 2 )))(e) AG ( g(3))(e)] DP g(3)
v ′⟨v,t⟩ λ uλ e[ DEL ( g(t 2 ))(e) REC (ιy. TRIBE (y)( g(t 2 ))(e) AG (u)(e)] ⟨DP2 ⟩
v ′⟨e,⟨v,t⟩⟩ λ z[λ xλ uλ e[ DEL ( x)(e) REC (ιy. TRIBE (y)( g 2 z ( g(2)))(e) AG (u)(e)]( z)] = λ z[λ uλ e. DEL (z)( e) REC (ιy. TRIBE (y)( z)(e) AG (u)(e)] β2
v ⟨e,⟨v,t⟩⟩ ′ λ xλ uλ e[ DEL ( g t 2 x (t 2 ))(e) REC (ιy. TRIBE (y)( g(2)))(e) AG (u)(e)] μ2
v ′⟨e,⟨v,t⟩⟩ λ uλ e[ DEL ( g(t 2 ))(e) REC (ιy. TRIBE (y)( g(t 2 )))(e) AG (u)(e)] v ⟨e,⟨v,t⟩⟩ λ uλ e[ AG (u)(e)]
ApplP ⟨v,t⟩ λ e[ DEL ( g(t 2 ))(e) REC (ιy. TRIBE (y)( g(2)))(e)]
194
PNI-PROPERT Y II: L ACK OF BINDING AND CONTROL c. Step 3: Movement out of the event domain vP⟨t⟩ z[ MAN (z) e[ DEL (z)(e) REC (ιy. TRIBE (y)( z))(e) AG ( g(3))(e)]] DP 2 ⟨⟨e, t⟩, t⟩ λ Q z[ MAN (z) Q(z)]
vP⟨e,t⟩ λ x. e[ DEL ( g t 2 x (t 2 ))(e) REC (ιy. TRIBE (y)( g t 2 x (t 2 )))(e) AG ( g(3))(e)] μ2
vP⟨t⟩ e[ DEL ( g(t 2 ))(e) REC (ιy. TRIBE (y)( g(t 2 )))(e) AG ( g(3))(e)] vP⟨v,t⟩ λe[ DEL ( g(t 2 ))(e) REC (ιy. TRIBE (y)( g(t 2 )))(e) AG ( g(3))(e)]
Let us now turn to the derivation of the PNI case. I interpreted the lack of case marking on the direct object in (1b) as an indication that PNI-DP is present, which blocks insertion of the binder prefix after the object has moved to Spec, vP. Since the μ-operator only sees movement indices, it will fail to bind the possessive pronoun. I provide the relevant part of the LF in (9). There is no need for a second movement step, as RESP is interpreted in the event domain.³ Co-indexation in (1b) is impossible due to the hybrid categorial nature of the PNI determiner that blocks variable binding. Possessive pronouns, however, can also act as pronouns. Hence, the only way to interpret the possessive pronoun in (1b) is by co-reference with an antecedent in the preceding discourse.
³ The analysis for short scrambling of PNI-ed arguments in Chapter 7 will in fact require the direct object as well as the subject to move out of vP. These additional steps are ignored here since they are irrelevant for the binding configurations under consideration.
PNI-PROPERT Y II: L ACK OF BINDING AND CONTROL
195
(9) No binding within vP in (1b) v ′⟨e,⟨v,t⟩⟩ λ uλ e z[ MAN(z) DEL (z)(e) REC (ιy. TRIBE (y)( g(2)))(e) AG (u)(e)] P NI -DP 2 ⟨⟨e, ⟨v, t⟩⟩, ⟨v, t⟩⟩ λ Q λ e z[ MAN (z) Q(z)(e)]
v ′⟨e,⟨e,⟨v,t⟩⟩⟩ λ xλ uλ e[ DEL ( g t 2 x (t 2 ))(e) REC (ιy. TRIBE (y)( g(2)))(e) AG (u)(e)] μ2
v ′⟨e,⟨v,t⟩⟩ λ uλ e[ DEL ( g(t 2 ))(e) REC (ιy. TRIBE (y)( g(t 2 )))(e) AG (u)(e)] v ⟨e,⟨v,t⟩⟩ λ uλ e[ AG (u)(e)]
ApplP⟨v,t⟩ λ e[ DEL ( g(t 2 ))(e) REC (ιy. TRIBE (y)( g(2)))(e)]
PP⟨e⟩ ιy. TRIBE (y)( g(2))
Appl′ ⟨e,⟨v,t⟩⟩ λ xλ e[ DEL ( g(t 2 ))(e) REC ( x)(e)]
Appl ⟨e,⟨v,t⟩⟩ λ xλ e[ REC ( x)(e)]
VP⟨v,t⟩ λ e[ DEL ( g(t 2 ))(e)] V⟨e,⟨v,t⟩⟩ λ xλ e[ DEL ( x)(e)]
⟨DP2 ⟩
I additionally demonstrate binding of a reflexive pronoun, as this is the scenario we used for subject PNI in Korean. The minimal pair, introduced in Section 3.3, is repeated here in (10). (10) Binding of reflexives by PNI-ed subjects a.
Koyangi-ka1 [ku casin-ul]1 halth-ass-e. cat-NOM 3SG self-ACC lick-PST-DECL ‘A cat washed itself.’
Korean
b. *Koyangi1 [ku casin-ul]1 halth-ass-e. cat 3SG self-ACC lick-PST-DECL ‘A cat washed itself.’ The case-marked subject introduces the binder prefix in situ, shown in (11). The DP, being a quantifier, must move out of the event domain. Hence the pronoun is in fact bound by the lower copy of the subject, which in turn is bound by the μ-operator, inserted immediately below the quantifiers, landing site.
196
PNI-PROPERT Y II: L ACK OF BINDING AND CONTROL
(11)
Successful binding in (10a) x[ CAT ( x)
vP⟨t⟩ e[ LICK ( x)(e)
AG ( x)(e)]
vP⟨e,t⟩ DP 2 λ x e[ LICK ( g t 2 x (t 2 ))(e) ⟨⟨e, t⟩, t⟩ AG ( g t 2 x (t 2 ))(e)] λ P x[ CAT ( x) P( x)] μ2
vP⟨t⟩ e[ LICK ( g(t 2 ))(e) AG ( g(t 2 ))(e)] vP⟨v,t⟩ λ e[ LICK ( g(t 2 ))(e) AG ( g(t 2 ))(e)] ⟨DP2 ⟩⟨e⟩
′ v ⟨e,⟨v,t⟩⟩ λ y[λ xλ e[ LICK ( g 2 y (2))( e) AG ( x)(e)]( y)] = λ y[λ e. LICK (y)(e) AG (y)(e)]
v ′⟨e,⟨v,t⟩⟩ λ xλ e[ LICK ( g(2))( e) AG ( x)( e)]
β2
VP⟨v,t⟩ λ e[ LICK ( g(2))(e)]
v ⟨e,⟨v,t⟩⟩ λ xλ e[ AG ( x)(e)]
DP ⟨e⟩ V ⟨e,⟨v,t⟩⟩ g(2) λ xλ e[ LICK ( x)(e)]
Lack of case marking on the subject in (10b) signals presence of PNI-DP, hence insertion of a binder prefix is not possible, the LF is given in (12). Note that moving the subject within vP does not enable binding, as was demonstrated above with (9). According to Binding Principle A, reflexive pronouns constitute variables and have to be bound. Since the PNI-ed subject is not able to do so, (10b) is unacceptable. (12)
Binding blocked in (10b) vP⟨v,t⟩ λ e∃z[ CAT (z) ∧ LICK ( g(2))( e) ∧ AG (z)(e)] P NI -DP 2 ⟨⟨e, ⟨v, t⟩⟩, ⟨v, t⟩⟩ λ Q λ e∃z[ CAT (z) ∧ Q(z)(e)]
v ′⟨e,⟨v,t⟩⟩ λ xλ e[ LICK ( g(2))(e) ∧ AG ( x)(e)] VP ⟨v,t⟩ λ e[ LICK ( g(2))( e)]
v ⟨e,⟨v,t⟩⟩ λ xλ e[ AG ( x)(e)]
V ⟨e,⟨v,t⟩⟩ DP ⟨e⟩ g(2) λ xλ e[ LICK ( x)(e)]
The binding argument transfers to control structures, under the assumption that embedded PRO is directly bound by the controller in the matrix clause. This is a traditional syntactic take on obligatory control, in which the control clause
PNI-PROPERT Y II: L ACK OF BINDING AND CONTROL
197
constitutes a proposition with an anaphor-like element, i.e. PRO, which established the co-indexed reading with the controller via variable binding (Borer 1989; Chomsky 1980, 1981; Koster 1984; Manzini 1983; Partee 1975; Sag and Pollard 1991). The syntactic approach makes it possible to explain agreement connectivity effects between the embedded verb and the matrix controller, as there is a direct connection between PRO and the controller. Simplified control structures under the current account are given in (13). (13) Syntactic control blocked by PNI a. DPcontroller β 1 ... control predicate [CP PRO1 ... ]
b. P NI -DP controller β1 ... control predicate [ CP
PRO1
... ]
The competing semantic account of obligatory control goes back to the work by Chierchia (1984, 1989) and takes control clauses to constitute nominalized one-place properties which are selected for by the control predicate. Thus, the co-indexed reading is a consequence of a predication relation between the control clause and the controller. The major advantage of the semantic account is the ability to restrict the modal accessibility relation to pick out those individuals that the author of the utterance context identifies as its doxastic counterpart, i.e. it can create the commonly observed obligatory de se readings for control structures (Chierchia 1990). A recent revival of the syntactic account, however, has been given by Landau (2015, 2017) who develops a theory of control which captures agreement effects between the controller and PRO as well as the obligatory de se readings. We will come back to Landau’s theory when we discuss the one exception in the data set considering the lack of control readings, found in Korean. The current approach to PNI is compatible with the syntactic account of obligatory control and can thus be seen as an argument in favour of it and against the semantic account of control. All in all, the control diagnostic should be taken with caution. Since obligatory control predicates form a closed class and do not necessarily overlap with the set of PNI predicates, we cannot safely conclude that the ban on PNI-ed controllers is in fact due to the inability of inserting a binder prefix that would establish a relation with PRO. What prevents PNI in control scenarios might simply be an incompatibility of control predicates with conventionalized activities. As already pointed out above, adjunct control structures make it possible to test for the possibility of PNI readings, separately from control readings. Since adjunct control overwhelming allows only for subject controllers (Green 2018; Landau 2013), the diagnostic was applied in Turkish and Korean which allow for subject PNI. It is, however, unclear how far adjunct control yields obligatory control readings or non-obligatory control readings with additional pragmatic restrictions. This diagnostic needs to be developed further in future research.
198
PNI-PROPERT Y II: L ACK OF BINDING AND CONTROL
6.1 Tamil Since Tamil shows overt case marking for direct objects and does not allow PNIed objects to scramble, we test for the binding properties of in situ objects into low adjuncts. The baseline without a potential bindee is given in (14). A case-marked bare noun with the unmarked word order is given in (14a). The bare noun can also be caseless, and it can precede (14b) or follow (14c) the adjunct. Two out of four informants think that (14b) is slightly marked, although nevertheless possible with a stressed object. (14) Bare nouns and low adjuncts a.
Kumar book-ai kadasi pakkam varai padi-c-aan. Kumar.NOM book-ACC last page until read-PST-3SG.M ‘Kumar read a book until the last page.’
b.
Kumar book kadasi pakkam varai padi-c-aan. Kumar.NOM book last page until read-PST-3SG.M ‘Kumar read a book until the last page.’
c.
Kumar kadasi pakkam varai book padi-c-aan. Kumar.NOM last page until book read-PST-3SG.M ‘Kumar read a book until the last page.’
Tamil
If the adjunct contains a possessive pronoun which needs to be bound by the direct object, the argument cannot be caseless, independent of word order, shown in (15c) and (15d). Case-marked objects are able to establish a binding relation; see (15b) and (15a). I assume that the PP originates in a position c-commanded by the direct object where the possessive pronoun can be bound by the binder prefix introduced by the direct object. The structures in which variable binding is blocked are derived with PNI-DP which does not introduce a binder. (15) Binding for bare nouns a.
Tamil
Kumar book-ai1 [ata-ode1 kadasi pakkam varai] Kumar.NOM book-ACC that-GEN last page until padi-c-aan. read-PST-3SG.M ‘Kumar read a book until its last page.’
b.
Kumar [ata-ode1 kadasi pakkam varai]2 book-ai1 ⟨PP2 ⟩ Kumar.NOM that-GEN last page until book-ACC padi-c-aan. read-PST-3SG.M ‘Kumar read a book until its last page.’
6.2 MONGOLIAN
199
c. *Kumar book1 [ata-ode1 kadasi pakkam varai] padi-c-aan. Kumar.NOM book that-GEN last page until read-PST-3SG.M ‘Kumar read a book until its last page.’ d. *Kumar [ata-ode1 kadasi pakkam varai]2 book1 ⟨PP2 ⟩ Kumar.NOM that-GEN last page until book padi-c-aan. read-PST-3SG.M ‘Kumar read a book until its last page.’ Caseless bare nouns are also incapable of binding into a control clause. This is exemplified with the contrast in (16) involving an object control predicate. The control verb is taken from Krishnamurti (2003: 442–443) who discusses infinitival clauses and their functions in Dravidian.⁴ (16) Object control for bare nouns a.
Tamil
Raja naai-ye1 [PRO1 kutikk-a] kattaya-paduthi-n-aan. Raja.NOM dog-ACC drink-INF compel-make-PST-3SG.M ‘Raja forced a dog to drink.’
b. *Raja naai1 [PRO1 kutikk-a] kattaya-paduthi-n-aan. Raja.NOM dog drink-INF compel-make-PST-3SG.M ‘Raja forced a dog to drink.’ Other noun types for which we have shown that case marking is optional will be discussed in Chapter 8, Section 8.1.
6.2 Mongolian Mongolian is restricted like Tamil in that PNI can only be diagnosed for direct objects, and once they are PNI-ed, they are immobile. As in Tamil, reflexive pronouns are subject-oriented, the binding test therefore employs possessive particles contained within the indirect argument of the positional/resultative verb xii (= put); see (17) where I provide a baseline with definite noun phrases acting as binders. Possessor relations in Mongolian can be expressed with prenominal possessive pronouns and/or postnominal possessor particles. Since the use of a prenominal possessor triggers a strong preference for a co-reference reading with a salient individual in the context (Guntsetseg 2016: 55) (see (17a)), we will make ⁴ One speaker mentions that (16b) can be coerced into a reading in which the caseless object serves as a PNI-ed object to the embedded verb, along the lines of Raja forced someone (pro) to do dog-drinking where dog would refer to a soda brand. This alternative reading is not a control reading.
200
PNI-PROPERT Y II: L ACK OF BINDING AND CONTROL
use of structures which contain the postnominal possessive particle only, as shown in (17b).⁵ The absence of prenominal possessors creates a strong tendency to interpret possessive pronouns as bound to a local antecedent. As (17c) shows, scrambling of the bindee does not affect binding relations. (17) Binding for definite noun phrases (Dolgor Guntsetseg, p.c.)
Mongolian
a. Ali [ene zurg-iig]1 tu¨u¨nii raamen-d n’? 1∕2 xii-sen. Ali.NOM this picture-ACC 3.POSS frame-DAT 3.POSS put-PST ‘A li put this picture in ? its/her frame.’ b. Ali [ene zurg-iig]1 raamen-d n’1∕? 2 xii-sen. Ali.NOM this picture-ACC frame-DAT 3.POSS put-PST ‘A li put this picture in its/? her frame.’ c. Ali [raamen-d n’1∕? 2 ]3 [ene zurg-iig]1 ⟨DP3 ⟩ xii-sen. Ali.NOM frame-DAT 3.POSS this picture-ACC put-PST ? ‘A li put this picture in its/ her frame.’ Crucially, the bound reading is impossible if the direct object lacks case marking, shown in (18a) for bare nouns and in (18b) for indefinites. Compare also the casemarked indefinite in (18c) which, as expected, allows for bound readings. (18) Binding for bare nouns and indefinites (Dolgor Guntsetseg, p.c.) a. Ali [raamen-d n’*1∕2 ]3 zurag1 ⟨DP3 ⟩ xii-sen. Ali.NOM frame-DAT 3.POSS picture put-PST ‘A li put a picture in her frame.’
Mongolian
b. Ali [raamen-d n’*1∕2 ]3 [neg zurag]1 ⟨DP3 ⟩ xii-sen. Ali.NOM frame-DAT 3.POSS a picture put-PST ‘A li put a picture in her frame.’ c. Ali [raamen-d n’1∕? 2 ]3 [neg zurg-iig]1 ⟨DP3 ⟩ xii-sen. Ali.NOM frame-DAT 3.POSS a picture-ACC put-PST ‘A li put a picture in its/? her frame.’ The observation can be extended to numerals and weak quantifiers. If they constitute caseless direct objects, they cannot bind the possessive particle in the indirect object. Minimal pairs are given in (19) and (20), respectively. ⁵ The morphological status of these particles is unclear, they can be treated either as enclitic pronouns or possessive suffixes (Janhunen 2012: 137).
6.2 MONGOLIAN
(19) Binding for numerals (Dolgor Guntsetseg, p.c.)
201
Mongolian
a. Ali [raamen-uud-ad n’*1∕2 ]3 [xojor zurag]1 ⟨DP3 ⟩ xii-sen. Ali.NOM frame-PL-DAT 3.POSS two picture put-PST ‘A li put two pictures in her frames.’ b. Ali [raamen-uud-ad n’1∕? 2 ]3 [xojor zurg-iig]1 ⟨DP3 ⟩ xii-sen. Ali.NOM frame-PL-DAT 3.POSS two picture-ACC put-PST ‘A li put two pictures in their/? her frames.’ (20) Binding for weak quantifiers (Dolgor Guntsetseg, p.c.)
Mongolian
a. Ali [raamen-d n’*1∕2 ]3 [olon zurag]1 ⟨DP3 ⟩ xii-sen. Ali.NOM frame-DAT 3.POSS many picture put-PST ‘A li put many pictures in her frames.’ b. Ali [raamen-d n’1∕? 2 ]3 [olon zurg-iig]1 ⟨DP3 ⟩ xii-sen. Ali.NOM frame-DAT 3.POSS many picture-ACC put-PST ? ‘A li put many pictures in their/ her frames.’ As the reader might have noticed, I treat the positional ditransitive xii (= put) differently from e.g. the Spanish ditransitive structure above, where it is the direct object which moves into an A-position to bind a possessive pronoun contained in the indirect object. The Mongolian speakers in my study, however, perceive the DO-before-IO structure for ditransitive xii as the unmarked word order. Hence, I assume that ditransitives of the positional/resultative type such as xii select for the indirect object before they merge with the direct object, akin to what has been proposed for prepositional datives in English (Bruening 2010, 2018b; Larson 1988; Marantz 1993). This approach is still compatible with object PNI since the verb is of type ⟨e, ⟨v, t⟩⟩, once it has merged with the indirect object. The underlying structure as well as a denotation for xii is given in (21). Although the PNI-DP is in principle able to compose with the verb in this position, it cannot introduce a binder. Hence, caseless direct objects cannot trigger a bound reading of the possessive pronoun in (18)–(20). The case-marked equivalents in (18)–(20) insert a binder prefix in situ and furthermore undergo type-driven movement out of the event domain. (21) Syntax and semantics of positional/resultative xii a. ⟦xii⟧ = λyλxλe[PUT(x)(e) ∧ LOC(y)(e)]
202
PNI-PROPERT Y II: L ACK OF BINDING AND CONTROL VP
b. P NI -DP neg zurag
V ′⟨e,⟨v,t⟩⟩ DP raamend n’
V ⟨e,⟨e,⟨v,t⟩⟩⟩ xii
As in Tamil, we see the same case-marking sensibility for object control structures. Following Monahan (2006), I investigate control clauses in which the embedded verb inflects with -x, a marker that can be analysed as a nominalizer or infinitive marker (Janhunen 2012: 159). The data in (22) show that controllers must be casemarked. (22) No control for caseless bare nouns, indefinites, weak quantifiers a. Ali suragch*(-yg)1 [PRO1 ger-t-ee xari-x]-yg Ali.NOM student-ACC home-DAT-REFL.POSS go-INF-ACC shaard-san. force-PST ‘A li forced a student to go home.’ b. Ali [neg suragch*(-yg)]1 [PRO1 ger-t-ee xari-x]-yg Ali.NOM a student-ACC home-DAT-REFL.POSS go-INF-ACC shaard-san. force-PST ‘A li forced a student to go home.’ c. Ali [xeden suragch*(-yg)]1 [PRO1 ger-t-ee Ali.NOM some student-ACC home-DAT-REFL.POSS xari-x]-yg shaard-san. go-INF-ACC force-PST ‘A li forced some students to go home.’ As this section has shown, each noun type which can potentially drop case marking in Mongolian simultaneously loses the ability to bind and control.
6.3 Turkish The interaction of binding and case drop was originally observed for bare nouns by Öztürk (2005, 2009). As in Mongolian and Tamil, scrambling does not affect the binding relations in Turkish; compare (23a) to (23b). If the direct object, however, is not case marked, binding into the indirect object becomes impossible; see (23c).
6.3 TURKISH
203
¨ ztu¨rk does not make explicit what the alternative readings for the possessive proO noun can be. According to my consultants, the pronoun can either be bound by the subject of the clause or can co-refer with a salient antecedent in the discourse. I subsume these two possibilities with index 2 in (23), and I will keep this notation for the rest of the section in order to reduce the number of indices for each example. (23) Öztürk (2009: 343)
Turkish
a. Ali resm-i1 c¸erc¸eve-sin-e1∕2 koy-du. Ali picture-ACC frame-POSS.3SG-DAT put-PST ‘A li put the picture in its/his frame.’ b. Ali [c¸erc¸eve-sin-e1∕2 ]3 resm-i1 ⟨DP3 ⟩ koy-du. Ali frame-POSS.3SG-DAT picture-ACC put-PST ‘A li put the picture in its/his frame.’ c. Ali [c¸erc¸eve-sin-e*1∕2 ]3 resim1 ⟨DP3 ⟩ koy-du. Ali frame-POSS.3SG-DAT picture put-PST ‘A li picture-put in his frame.’ Binding relations are not only blocked for bare nouns in PNI scenarios. I extend the observation to indefinites (24), numerals > 1 (25), and weak quantifiers (26). Essentially any noun type I have identified in Section 4.3.3 as being able to drop case cannot act as a binder without case. (24) Binding for indefinites
Turkish
a. Ali [c¸erc¸eve-si-ne1∕2 ]3 [bir resm-i]1 ⟨DP3 ⟩ koy-du-∅. Ali.NOM frame-POSS.3SG-DAT a picture-ACC put-PFV-3SG ‘A li put a picture in its/his frame.’ b. Ali [c¸erc¸eve-si-ne*1∕2 ]3 [bir resim]1 ⟨DP3 ⟩ koy-du-∅. Ali.NOM frame-POSS.3SG-DAT a picture put-PFV-3SG ‘A li put a picture in his frame.’ (25) Binding for numerals > 1
Turkish
a. Ali [c¸erc¸eve-si-ne1∕2 ]3 [u¨c¸ resm-i]1 ⟨DP3 ⟩ koy-du-∅. Ali.NOM frame-POSS.3SG-DAT three picture-ACC put-PFV-3SG ‘A li put three pictures in their/his frames.’ resim]1 ⟨DP3 ⟩ koy-du-∅. b. Ali [c¸erc¸eve-si-ne*1∕2 ]3 [u¨c¸ put-PFV-3SG Ali.NOM frame-POSS.3SG-DAT three picture ‘A li put three pictures in his frames.’
204
PNI-PROPERT Y II: L ACK OF BINDING AND CONTROL
(26) Binding for weak quantifiers
Turkish
a. Ali [c¸erc¸eve-si-ne1∕2 ]3 [bazı resim-ler-i]1 ⟨DP3 ⟩ koy-du-∅. Ali frame-POSS.3SG-DAT some picture-ACC put-PFV-3SG ‘A li put some pictures in their/his frames.’ b. Ali [c¸erc¸eve-si-ne*1∕2 ]3 [bazı resim-ler]1 ⟨DP3 ⟩ koy-du-∅. Ali.NOM frame-POSS.3SG-DAT some picture put-PFV-3SG ‘A li put some pictures in his frames.’ We treat Turkish koy (= put) exactly like Mongolian xii (= put); see (27). The positional ditransitive verb first merges with the indirect object which determines the endpoint of the motion, followed by Merge with the direct object. The PNI-ed object will be able to semantically combine with the verb in that position. It will not, however, be able to bind into the indirect object since it cannot introduce a binder prefix. The case-marked objects in (23)–(26) introduce the binder in situ and furthermore undergo type-driven movement out of the event domain. (27) Syntax and semantics of positional/resultative koy a. ⟦koy⟧ = λyλxλe[PUT(x)(e) ∧ LOC(y)(e)] VP b. V ′⟨e,⟨v,t⟩⟩
P NI -DP bir resim
DP çerçeve-si-ne
V ⟨e,⟨e,⟨v,t⟩⟩⟩ koy
Lack of binding has also been attested for subject PNI, albeit only in matrix clauses. Kornfilt (2003a) provides a minimal contrast with respect to reflexive pronouns, where PNI is not signalled by case drop but rather by the non-specific reading of the subject; see (28). (28) Kornfilt (2003a: 150–151) a.
Turkish
Bir c¸ocuk1 kendin-i1 yarala-mıs¸. a child.NOM REFL-ACC wound-REP.PST ‘A child [+specific] is said to have wounded himself.’
b. *Kendin-i1 bir c¸ocuk1 yarala-mıs¸. REFL-ACC a child.NOM wound-REP.PST ‘A child [−specific] is said to have wounded himself.’ Öztürk (2005, 2009) provides binding scenarios for subject PNI involving a possessive pronoun, the paradigm is given in (29), parallel to object PNI in (23).
6.3 TURKISH
205
¨ ztu¨rk makes Indicative of a PNI scenario in (29c) is here the PNI reading which O explicit in the translation. (29) Öztürk (2009: 343)
Turkish
a. Doktor1 hasta-sın-ı1∕2 go¨r-du¨. doctor.NOM patient-POSS.3SG-DAT see-PST ‘The doctor examined his patient.’ b. [Hasta-sın-ı1∕2 ]3 doktor1 ⟨DP3 ⟩ go¨r-du¨. patient-POSS.3SG-DAT doctor.NOM see-PST ‘The doctor examined his patient.’ c. [Hasta-sın-ı*1∕2 ]3 doktor1 ⟨DP3 ⟩ go¨r-du¨. patient-POSS.3SG-DAT doctor see-PST ‘His patient underwent doctor-examination.’ Since PNI readings and specificity properties are not robustly attested in Turkish (see Section 4.3.3), I test binding relations in embedded clauses where the subject is overtly marked for genitive case. The following data show a clear correlation between genitive case drop on the subject and lack of bound readings for possessive pronouns contained in the direct object. Minimal contrasts are given for bare nouns (30), indefinites (31), numerals > 1 (32), and weak quantifiers (33). (30) Binding for bare subjects
Turkish
a. [Sahib-i-ni1∕2 ]3 ko¨peg˘-in1 ⟨DP3 ⟩ ısır-dıg˘-ı-nı owner-POSS.3SG-ACC dog-GEN bite-NMLZ-POSS.3SG-ACC duy-du-m. hear-PFV-1SG ‘I heard that a dog bit its/someone’s owner.’ b. [Sahib-i-ni*1∕2 ]3 ko¨pek1 ⟨DP3 ⟩ ısır-dıg˘-ı-nı owner-POSS.3SG-ACC dog bite-NMLZ-POSS.3SG-ACC duy-du-m. hear-PFV-1SG ‘I heard that a dog bit someone’s owner.’
(31) Binding for indefinite subjects
Turkish
a. [Sahib-i-ni1∕2 ]3 [bir ko¨peg˘-in]1 ⟨DP3 ⟩ ısır-dıg˘-ı-nı owner-POSS.3SG-ACC a dog-GEN bite-NMLZ-POSS.3SG-ACC duy-du-m. hear-PFV-1SG ‘I heard that a dog bit its/someone’s owner.’
206
PNI-PROPERT Y II: L ACK OF BINDING AND CONTROL b. [Sahib-i-ni*1∕2 ]3 [bir ko¨pek]1 ⟨DP3 ⟩ ısır-dıg˘-ı-nı owner-POSS.3SG-ACC a dog bite-NMLZ-POSS.3SG-ACC duy-du-m. hear-PFV-1SG ‘I heard that a dog bit someone’s owner.’
(32) Binding for numeral subjects a. [Sahib-i-ni1∕2 ]3 owner-POSS.3SG-ACC
[u¨c¸ three
Turkish ko¨peg˘-in]1 dog-GEN
⟨DP3 ⟩
ısır-dıg˘-ı-nı duy-du-m. bite-NMLZ-POSS.3SG-ACC hear-PFV-1SG ‘I heard that three dogs bit their/someone’s owners.’ b. [Sahib-i-ni*1∕2 ]3 [u¨c¸ ko¨pek]1 ⟨DP3 ⟩ ısır-dıg˘-ı-nı owner-POSS.3SG-ACC three dog bite-NMLZ-POSS.3SG-ACC duy-du-m. hear-PFV-1SG ‘I heard that three dogs bit someone’s owners.’ (33) Binding for weak quantifier subjects a. [Sahib-i-ni1∕2 ]3 owner-POSS.3SG-ACC
[birc¸ok many
Turkish ko¨peg˘-in]1 dog-GEN
⟨DP3 ⟩
ısır-dıg˘-ı-nı duy-du-m. bite-NMLZ-POSS.3SG-ACC hear-PFV-1SG ‘I heard that many dogs bit their/someone’s owners.’ b. [Sahib-i-ni*1∕2 ]3 [birc¸ok ko¨pek]1 ⟨DP3 ⟩ ısır-dıg˘-ı-nı owner-POSS.3SG-ACC many dog bite-NMLZ-POSS.3SG-ACC duy-du-m. hear-PFV-1SG ‘I heard that many dogs bit someone’s owners.’ As with objects, caseless subjects signal the presence of PNI-DP. Hence, a binder prefix necessary for establishing variable binding cannot be inserted by the subject. The case-marked subjects introduce a binder in situ, parallel to the derivation shown in (11). Case drop in Turkish also blocks control readings. According to Özsoy (2001) and Słodowicz (2007), clauses without agreement morphology qualify as control clauses. In contrast to Mongolian and Tamil, Turkish control clauses can be
6.3 TURKISH
207
scrambled over the controller without losing the obligatory control reading. The following object control paradigms each provide the baseline in a., the scrambled version in b., and the additional case drop on the controller in c. Control tests are run for bare nouns (34), indefinites (35), and weak quantifiers (36). (34) Object control for bare nouns
Turkish
a.
Sen suc¸lu-yu1 [PRO1 silah-ı bırak-ma]-ya zorla-dı-n. 2SG criminal-ACC weapon-ACC drop-NMLZ-DAT force-PFV-2SG ‘You forced the criminal to drop the weapon.’
b.
Sen [PRO1 silah-ı bırak-ma]-ya3 suc¸lu-yu1 ⟨TP3 ⟩ 2SG weapon-ACC drop-NMLZ-DAT criminal-ACC zorla-dı-n. force-PFV-2SG ‘You forced the criminal to drop the weapon.’
c. *Sen [PRO1 silah-ı bırak-ma]-ya3 suc¸lu1 ⟨TP3 ⟩ 2SG weapon-ACC drop-NMLZ-DAT criminal zorla-dı-n. force-PFV-2SG ‘You forced a criminal to drop the weapon.’
(35) Object control for indefinites a.
Turkish
Sen [bir suc¸lu-yu]1 [PRO1 silah-ı bırak-ma]-ya 2SG a criminal-ACC weapon-ACC drop-NMLZ-DAT zorla-dı-n. force-PFV-2SG ‘You forced a criminal to drop the weapon.’
b.
Sen [PRO1 silah-ı bırak-ma]-ya3 [bir suc¸lu-yu]1 ⟨TP3 ⟩ 2SG weapon-ACC drop-NMLZ-DAT a criminal-ACC zorla-dı-n. force-PFV-2SG ‘You forced a criminal to drop the weapon.’
c. *Sen [PRO1 silah-ı bırak-ma]-ya3 [bir suc¸lu]1 ⟨TP3 ⟩ 2SG weapon-ACC drop-NMLZ-DAT a criminal zorla-dı-n. force-PFV-2SG ‘You forced a criminal to drop the weapon.’
208
PNI-PROPERT Y II: L ACK OF BINDING AND CONTROL
(36) Object control for weak quantifiers a.
Turkish
Sen [bazı suc¸lu-lar-ı]1 [PRO1 silah-ı bırak-ma]-ya 2SG some criminal-PL-ACC weapon-ACC drop-NMLZ-DAT zorla-dı-n. force-PFV-2SG ‘You forced some criminals to drop the weapon.’
b.
Sen [PRO1 silah-ı bırak-ma]-ya3 [bazı suc¸lu-lar-ı]1 2SG weapon-ACC drop-NMLZ-DAT some criminal-PL-ACC ⟨TP3 ⟩ zorla-dı-n. force-PFV-2SG ‘You forced some criminals to drop the weapon.’
c. * Sen [PRO1 silah-ı bırak-ma]-ya3 [bazı suc¸lu-lar]1 ⟨TP3 ⟩ 2SG weapon-ACC drop-NMLZ-DAT some criminal-PL zorla-dı-n. force-PFV-2SG ‘You forced some criminals to drop the weapon.’ The identical picture can be provided for subject control structures. Tested, again, were bare nouns (37), indefinites (38), and weak quantifiers (39). (37) Subject control for bare nouns a.
Turkish
Polis-in1 [PRO1 Ali-yi tutukla-ma]-ya c¸alıs¸-tıg˘-ı-nı police-GEN Ali-ACC arrest-NMLZ-DAT try-NMLZ-POSS.3SG-ACC duy-du-m. hear-PFV-1SG ‘I have heard that the police tried to arrest Ali.’
b.
[PRO1 Ali-yi tutukla-ma]-ya3 polis-in1 ⟨TP3 ⟩ Ali-ACC arrest-NMLZ-DAT police-GEN c¸alıs¸-tıg˘-ı-nı duy-du-m. try-NMLZ-POSS.3SG-ACC hear-PFV-1SG ‘I have heard that the police tried to arrest Ali.’
c. *[PRO1
Ali-yi Ali-ACC
tutukla-ma]-ya3 arrest-NMLZ-DAT
polis1 police
⟨TP3 ⟩
duy-du-m. c¸alıs¸-tıg˘-ı-nı try-NMLZ-POSS.3SG-ACC hear-PFV-1SG ‘I have heard that there was police-trying to arrest Ali.’
6.3 TURKISH
(38) Subject control for indefinites a.
209
Turkish
[Bir polis-in]1 [PRO1 Ali-yi tutukla-ma]-ya a police-GEN Ali-ACC arrest-NMLZ-DAT c¸alıs¸-tıg˘-ı-nı duy-du-m. try-NMLZ-POSS.3SG-ACC hear-PFV-1SG ‘I have heard that a police man tried to arrest Ali.’
b.
[PRO1 Ali-yi tutukla-ma]-ya3 [bir polis-in]1 ⟨TP3 ⟩ Ali-ACC arrest-NMLZ-DAT a police-GEN c¸alıs¸-tıg˘-ı-nı duy-du-m. try-NMLZ-POSS.3SG-ACC hear-PFV-1SG ‘I have heard that a police man tried to arrest Ali.’
c. *[PRO1 Ali-yi tutukla-ma]-ya3 [bir polis]1 ⟨TP3 ⟩ Ali-ACC arrest-NMLZ-DAT a police c¸alıs¸-tıg˘-ı-nı duy-du-m. try-NMLZ-POSS.3SG-ACC hear-PFV-1SG ‘I have heard that there was police-trying to arrest Ali.’
(39) Subject control for weak quantifiers a.
Turkish
[Bazı polis-ler-i]1 [PRO1 Ali-yi tutukla-ma]-ya some police-PL-GEN Ali-ACC arrest-NMLZ-DAT c¸alıs¸-tıg˘-ı-nı duy-du-m. try-NMLZ-POSS.3SG-ACC hear-PFV-1SG ‘I have heard that some police men tried to arrest Ali.’
b.
[PRO1 Ali-yi tutukla-ma]-ya3 [bazı polis-ler-i]1 ⟨TP3 ⟩ Ali-ACC arrest-NMLZ-DAT some police-PL-GEN c¸alıs¸-tıg˘-ı-nı duy-du-m. try-NMLZ-POSS.3SG-ACC hear-PFV-1SG ‘I have heard that some police men tried to arrest Ali.’
c. *[PRO1 Ali-yi tutukla-ma]-ya3 [bazı polis-ler]1 ⟨TP3 ⟩ Ali-ACC arrest-NMLZ-DAT some police-PL c¸alıs¸-tıg˘-ı-nı duy-du-m. try-NMLZ-POSS.3SG-ACC hear-PFV-1SG ‘I have heard that there was police-trying to arrest Ali.’ Additionally, Turkish allows us to test for adjunct control by PNI-ed subjects. This is important since lack of subject or object control, as it was shown above, can always be related to the possibility that control predicates do not constitute suitable
210
PNI-PROPERT Y II: L ACK OF BINDING AND CONTROL
PNI predicates, as they might not express conventionalized or frequent activities. Öztürk (2005, 2009) provides adjunct control data for subject PNI in matrix clauses, and shows that a PNI reading is incompatible with a control reading; see (40). (40) Öztürk (2009: 344) a.
Turkish
Polis1 Ali-yi kasıtlı olarak [PRO1 sorgula-mak ic¸in] police.NOM Ali-ACC intentionally interrogate-INF for tutukla-dı. arrest-PST ‘The police arrested Ali to interrogate him.’
b. *Ali-yi [PRO1 sorgula-mak ic¸in]3 kasıtlı olarak polis1 ⟨TP3 ⟩ Ali-ACC interrogate-INF for intentionally police tutukla-dı. arrest-PST ‘Police-arresting happened to Ali to interrogate him.’ The unacceptability of (40b) is per se not expected with adjunct control since Turkish is a pro-drop language. Hence, a reading in which the non-overt subject in the control clause refers to a salient individual in the context should be readily available. One reason why this reading might be difficult to create is that the context is set-up in a way that no referent other than the matrix subject is likely to exist. I therefore choose a context for our adjunct control diagnostic that allows other discourse referents in the context to co-refer with the non-overt subject in the control clause. Moreover, I test for adjunct control in subordinate clauses where genitive marked subject controllers show a clear morpho-syntactic reflex of PNI. The paradigms are given in (41) for bare nouns, in (42) for indefinites, and in ¨ ztu¨rk’s observation for matrix (43) for weak quantifiers. They demonstrate that O clauses can be replicated for embedded clauses. In other words, PNI prevents subjects from controlling into adjunct clauses. (41) Adjunct subject control for bare nouns
Turkish
¨ g˘renci-nin1 [PRO1∕2 sınav-dan c¸ık-tık-tan a. O sonra] student-GEN exam-ABL exit-NMLZ-ABL after gel-dig˘-i-ni duy-du-m. come-NMLZ-POSS.3SG-ACC hear-PFV-1SG ‘I have heard that the student came after leaving the exam.’ b. [PRO1∕2 sınav-dan c¸ık-tık-tan sonra]3 o¨g˘renci-nin1 ⟨TP3 ⟩ student-GEN exam-ABL exit-NMLZ-ABL after gel-dig˘-i-ni duy-du-m. come-NMLZ-POSS.3SG-ACC hear-PFV-1SG ‘I have heard that the student came after leaving the exam.’
6.3 TURKISH
211
c. [PRO*1∕2 sınav-dan c¸ık-tık-tan sonra]3 o¨g˘renci1 ⟨TP3 ⟩ exam-ABL exit-NMLZ-ABL after student gel-dig˘-i-ni duy-du-m. come-NMLZ-POSS.3SG-ACC hear-PFV-1SG ‘I have heard that a student came after leaving the exam.’ (42) Adjunct subject control for indefinites
Turkish
a. [Bir o¨g˘renci-nin]1 [PRO1∕2 sınav-dan c¸ık-tık-tan sonra] a student-GEN exam-ABL exit-NMLZ-ABL after gel-dig˘-i-ni duy-du-m. come-NMLZ-POSS.3SG-ACC hear-PFV-1SG ‘I have heard that a student came after leaving the exam.’ b. [PRO1∕2 sınav-dan c¸ık-tık-tan sonra]3 [bir o¨g˘renci-nin]1 ⟨TP3 ⟩ exam-ABL exit-NMLZ-ABL after a student-GEN gel-dig˘-i-ni duy-du-m. come-NMLZ-POSS.3SG-ACC hear-PFV-1SG ‘I have heard that a student came after leaving the exam.’ c. [PRO*1∕2 sınav-dan c¸ık-tık-tan sonra]3 [bir o¨g˘renci]1 ⟨TP3 ⟩ exam-ABL exit-NMLZ-ABL after a student gel-dig˘-i-ni duy-du-m. come-NMLZ-POSS.3SG-ACC hear-PFV-1SG ‘I have heard that a student came after leaving the exam.’ (43) Adjunct subject control for weak quantifiers
Turkish
a. [Bazı o¨g˘renci-ler-in]1 [PRO1∕2 sınav-dan c¸ık-tık-tan sonra] some student-PL-GEN exam-ABL exit-NMLZ-ABL after gel-dig˘-i-ni duy-du-m. come-NMLZ-POSS.3SG-ACC hear-PFV-1SG ‘I have heard that some students came after leaving the exam.’ b. [PRO1∕2 sınav-dan c¸ık-tık-tan sonra]3 [bazı o¨g˘renci-ler-in]1 exam-ABL exit-NMLZ-ABL after some student-PL-GEN ⟨TP3 ⟩ gel-dig˘-i-ni duy-du-m. come-NMLZ-POSS.3SG-ACC hear-PFV-1SG ‘I have heard that some students came after leaving the exam.’ c. [PRO*1∕2 sınav-dan c¸ık-tık-tan sonra]3 [bazı o¨g˘renci-ler]1 ⟨TP3 ⟩ exam-ABL exit-NMLZ-ABL after some student-PL gel-dig˘-i-ni duy-du-m. come-NMLZ-POSS.3SG-ACC hear-PFV-1SG ‘I have heard that some students came after leaving the exam.’
212
PNI-PROPERT Y II: L ACK OF BINDING AND CONTROL
We have seen in this section that the class of noun types which allow for optional case marking in Turkish additionally shows a consistent correlation between absence of case marking and an inability to enter binding or control relations. No difference is detectable between object and subject PNI. The next section will present a very similar picture in Korean.
6.4 Korean Pseudo-incorporation blocks arguments from triggering variable binding. This was already shown in Section 4.3.4 and is repeated here in more detail. The test sentences show subjects binding reflexive pronouns in object position. While this might not be unexpected for bare nouns (44), Korean is unique in allowing proper names (45), weak definites (46), and 1st /2nd pronouns (47) to exhibit the same correlation between case drop and lack of binding. (44) Binding for bare nouns a.
Koyangi-ka1 [ku casin-ul]1 halth-ass-e. cat-NOM 3rd self-ACC lick-PST-INT ‘A cat washed itself.’
Korean
b. *Koyangi1 [ku casin-ul]1 halth-ass-e. cat 3rd self-ACC lick-PST-INT ‘A cat washed itself.’
(45) Binding for proper names a.
Cinwu-ka1 [ku casin-ul]1 chingchanhay-ss-e. Cinwu-NOM 3rd self-ACC praise-PST-INT ‘Cinwu praised himself.’
Korean
b. ?*Cinwu1 [ku casin-ul]1 chingchanhay-ss-e. Cinwu 3rd self-ACC praise-PST-INT ‘Cinwu praised himself.’
(46) Binding for weak definites (global uniqueness) a.
[Ku yewang-i]1 [ku casin-ul]1 tachikeyhay-ss-e DEM queen-NOM 3rd self-ACC hurt-PST-INT ‘The queen hurt herself.’
Korean
6.4 KORE AN
213
b. ?? [Ku yewang]1 [ku casin-ul]1 tachikeyhay-ss-e DEM queen 3rd self-ACC hurt-PST-INT ‘The queen hurt herself.’ c.
Yewang-i1 [ku casin-ul]1 tachikeyhay-ss-e queen-NOM 3rd self-ACC hurt-PST-INT ‘The queen hurt herself.’
d. *Yewang1 [ku casin-ul]1 tachikeyhay-ss-e queen 3rd self-ACC hurt-PST-INT ‘The queen hurt herself.’
(47) Binding for 1st /2nd pronouns a.
Nay-ka1 [nay casin-ul]1 chingchanhay-ss-e. 1SG-NOM 1SG self-ACC praise-PST-INT ‘I praised myself.’
Korean
b. ?*Nay1 [nay casin-ul]1 chingchanhay-ss-e. 1SG 1SG self-ACC praise-PST-INT ‘I praised myself.’ c.
Ney-ka1 [ney casin-ul]1 chingchanhay-ss-e. 2SG-NOM 2SG self-ACC praise-PST-INT ‘You praised yourself.’
d. ?*Ney1 [ney casin-ul]1 chingchanhay-ss-e. 2SG 2SG self-ACC praise-PST-INT ‘You praised yourself.’ In Section 4.3.4, we provided an explanation for the unusual class of PNI noun types by assuming that there is an independent ident-type shifter available, developed from the demonstrative ku, which takes over the function of turning uniquely identifiable e-type arguments into singleton properties. These properties in turn can serve as an argument for the PNI determiner. Hence, the caseless subjects in (44)–(47) constitute PNI-DPs which are not nominal enough to introduce binder prefix β. The compositions for (44) were given in (11) and (12) at the beginning of this chapter. We present binding scenarios for proper names here briefly in (48) and (49), representative of the class of noun types which undergo PNI via additional ident-type shift.
214
PNI-PROPERT Y II: L ACK OF BINDING AND CONTROL
(48) Successful binding in (45a) vP⟨v,t⟩ λ e[ PRAISE (cinwu )( e) DP⟨e⟩ Cinwu
AG (cinwu )(e)]
v ′⟨e,⟨v,t⟩⟩ 2 λ y[λ xλ e[ PRAISE ( g y (2))(e) = λ y[λ e. PRAISE (y)(e)
AG ( x)(e)]( y)] AG (y)(e)]
v ′⟨e,⟨v,t⟩⟩ λ xλ e[ PRAISE ( g(2))(e) AG ( x)(e)]
β2
v ⟨e,⟨v,t⟩⟩ λ xλ e[ AG ( x)(e)]
VP ⟨v,t⟩ λ e[ PRAISE ( g(2))(e)] V ⟨e,⟨v,t⟩⟩ DP ⟨e⟩ g(2) λ xλ e[ PRAISE ( x)(e)]
(49) Binding blocked in (45b) vP⟨v,t⟩ λ e z[cinwu = z P NI -DP 2 ⟨⟨e, ⟨v, t⟩⟩, ⟨v, t⟩⟩ λ Q λ e z[cinwu = z Q(z)(e)]
PRAISE ( g(2))(e)
AG (z)(e)]
v ′⟨e,⟨v,t⟩⟩ λ xλ e[ PRAISE ( g(2))(e) VP ⟨v,t⟩ λ e[ PRAISE ( g(2))(e)]
AG ( x)(e)]
v ⟨e,⟨v,t⟩⟩ λ xλ e[ AG ( x)(e)]
V ⟨e,⟨v,t⟩⟩ DP⟨e⟩ g(2) λ xλ e[ PRAISE ( x)(e)]
The results of the control diagnostic are in line with the binding data above, with one exception, which we address at the end of the section. Each noun type showing a correlation between case drop and lack of binding is tested for object control, subject control, and adjunct subject control. In Korean, subject control predicates require the addition of the volitional modal -keyss, while object control predicates are formed with the imperative marker -la (K-Y. Lee 2009; Park 2018; Yang 1985). The paradigms for bare nouns are given in (50). If bare subjects are caseless, PRO is not bound, thus the sentence becomes unacceptable.
6.4 KORE AN
(50) No object, subject, and adjunct control for caseless bare nouns
215
Korean
a. Yusu-ka haksayng*(-ul)1 [PRO1 ttena-la-ko] seltukhay-ss-e. Yusu-NOM student-ACC leave-IMP-COMP persuade-PST-INT ‘Yusu persuaded a student to leave.’ b. Haksayng*(-i)1 [PRO1 ttena-keyss-ta-ko] kyelsimhay-ss-e student-NOM leave-VOL-DECL-COMP decide-PST-INT ‘A student decided to leave.’ c. [Pilok PRO1 olay-toyess-ciman] khemphyuthe? (-ka)1 cal although old-became-but computer-NOM well tolak-a. work-INT ‘[Although PRO1 is old] a computer1 works well.’ Control properties for proper names are shown in (51). Again, if case is not marked on the controller, the sentences become unacceptable. (51) No object, subject, and adjunct control for caseless proper name
Korean
a. Yusu-ka Sohyung*(-ul)1 [PRO1 ttena-la-ko] seltukhay-ss-e. Yusu-NOM Sohyung-ACC leave-IMP-COMP persuade-PST-INT ‘Yusu persuaded Sohyung to leave.’ b. Yusu*(-ka)1 [PRO1 ttena-keyss-ta-ko] kyelsimhay-ss-e Yusu-NOM leave-VOL-DECL-COMP decide-PST-INT ‘Yusu decided to leave.’ c. [PRO1 samchon-cip-ul tanyeon hwuey] Minswu? (-ka)1 woori-eykey uncle-house-ACC come.by after Minsu-NOM we-DAT cenhwa-lul hay-ss-e. call-ACC do-PST-INT ‘Minsu1 called us [after PRO1 visiting his uncle]. The control behaviour of weak definites is presented in (52)–(54). As was shown in Section 4.3.4, case drop is possible if ku is present. If, however, a weak definite controller lacks case marking, sentences become unacceptable. (52) No object control for caseless weak global definites a. Yusu-ka Yusu-NOM
[ku DEM
??
yewang (-ul)]1 queen-ACC
[PRO1
pemcoyca-lul criminal-ACC
yongseha-la-ko] seltukhay-ss-e. forgive-IMP-COMP persuade-PST-INT ‘Yusu persuaded the queen to forgive the criminal.’
Korean
216
PNI-PROPERT Y II: L ACK OF BINDING AND CONTROL b. Yusu-ka yewang*(-ul)1 [PRO1 pemcoyca-lul yongseha-la-ko] Yusu-NOM queen-ACC criminal-ACC forgive-IMP-COMP seltukhay-ss-e. persuade-PST-INT ‘Yusu persuaded the queen to forgive the criminal.’
(53) No subject control for caseless weak global definites
Korean
a. [Ku yewang? (-i)]1 [PRO1 ttena-keyss-ta-ko] kyelsimhay-ss-e DEM queen-NOM leave-VOL-DECL-COMP decide-PST-INT ‘The queen decided to leave.’ b. Yewang*(-i)1 [PRO1 ttena-keyss-ta-ko] kyelsimhay-ss-e queen-NOM leave-VOL-DECL-COMP decide-PST-INT ‘The queen decided to leave.’
(54) No adjunct subject control for caseless weak global definites
Korean
?
a. [Pilok PRO1 pappu-ciman] [ku yewang (-i)]1 phathi-ey although busy-but DEM queen-NOM party-DAT wa-ss-e. come-PST-INT ‘[Although PRO1 is busy] the queen1 came to the party.’ b. [Pilok PRO1 pappu-ciman] yewang*(-i)1 phathi-ey wa-ss-e. although busy-but queen-NOM party-DAT come-PST-INT ‘[Although PRO1 is busy] the queen1 came to the party.’ The observations made so far for control cannot be extended to 1st /2nd person pronouns. The paradigms in (55), (56), and (57) show that a coreferential reading between the controller and PRO is completely unaffected by the absence of case marking. (55) Object control for 1st /2nd pronouns a. Yusu-ka na(-lul)1 [PRO1 ttena-la-ko] seltukhay-ss-e. Yusu-NOM 1SG-ACC leave-IMP-COMP persuade-PST-INT ‘Yusu persuaded me to leave.’ b. Yusu-ka ne(-lul)1 [PRO1 ttena-la-ko] seltukhay-ss-e. Yusu-NOM 2SG-ACC leave-IMP-COMP persuade-PST-INT ‘Yusu persuaded you to leave.’
Korean
6.4 KORE AN
(56) Subject control for 1st /2nd pronouns
217
Korean
a. Nay(-ka)1 [PRO1 ttena-keyss-ta-ko] kyelsimhay-ss-canha 1SG-NOM leave-VOL-DECL-COMP decide-PST-DP ‘I decided to leave.’ b. Ney(-ka)1 [PRO1 ttena-keyss-ta-ko] kyelsimhay-ss-canha 2SG-NOM leave-VOL-DECL-COMP decide-PST-DP ‘You decided to leave.’
(57) Adjunct subject control for 1st /2nd pronouns
Korean
a. [Pilok PRO1 pappu-ciman] nay(-ka)1 phathi-ey ka-ss-e. although busy-but 1SG-NOM party-DAT go-PST-INT ‘[Although PRO1 is busy] I1 went to the party.’ b. [Pilok PRO1 pappu-ciman] ney(-ka)1 phathi-ey ka-ss-e. although busy-but 2SG-NOM party-DAT go-PST-INT ‘[Although PRO1 is busy] you1 went to the party.’ We can make sense of this exception by taking a closer look at Landau’s (2015, 2017) recent proposal for control under attitude verbs. Landau takes attitude control predicates to quantify over sets of contexts. The control complementizer takes the control clause as a complement and a context tuple as its specifier. The latter consists of coordinates that make up the logophoric centre, i.e. the context of speech/though of the matrix event, shown in (58). (58) Attitude complementizers introduce logophoric centres (Landau 2017: 12) a. Context tuple: < prox , proy , time(i), world(i) > b. Presuppositions: prox = AUTHOR(i); proy = ADDRESSEE(i) Within the control clause, PRO undergoes operator movement, creating a λ-abstract which is then predicated of a projected coordinate of the context tuple. For object and subject control this coordinate is prox or proy , a minimal pronoun which is bound by the respective controller in the matrix clause. To ensure de se readings, Landau embeds pro in a Concept Generator (Percus and Sauerland 2003) which establishes the aquaintance relation between the attitude holder of the attitude predicate and pro. An additional presupposition on the control complementizer makes this aquaintance relation reflexive, i.e. one of self-identification. We take subject control in (51a) as an example and show simplified LFs in (59), adjusted to the current system and leaving out the Concept Generator. In (59a), the controller is case marked, thus can introduce a binder prefix which binds the author coordinate prox , thereby ensuring that the controller will be the author of
218
PNI-PROPERT Y II: L ACK OF BINDING AND CONTROL
the attitude context of the matrix event. The PNI variant in (59b) cannot introduce a binder wich leaves prox unbound, thereby not fulfilling the presupposition in (58b). (59) Subject control with proper names a. [Yusu-ka β x ... [ CP prox [C ′ [FinP PR O 1 [Fin ′ ⟨ PRO 1 ⟩ ...] Fin] [ C -ko]] ...] binding
predication
b. *[Yusu β x [CP prox [C ′ [FinP PR O 1 [Fin ′ ⟨ PRO 1 ⟩ ...] Fin] [ C -ko]] ...] predication
As we have seen above, if case marking is missing on 1st /2nd person controllers, control clauses are still entirely acceptable. Hence, there seems to be an alternative way to derive the control relation, specifically available to 1st /2nd person pronouns. One possible way is to let the prox coordinate of the matrix clause also project as a specifier of matrix CP. Matrix prox can then bind prox of the control clause, which is shown in (60c). (60) Subject control with 1st /2nd pronouns a. [nay-ka β x ... [ CP prox [C ′ [FinP PR O 1 [Fin ′ ⟨ PRO 1 ⟩ ...] Fin] [ C -ko]] ...] binding
predication
b. *[nay β x [CP prox [C ′ [FinP PR O 1 [Fin ′ ⟨ PRO 1 ⟩ ...] Fin] [ C -ko]] ...] predication
c.
[ CP pro x β x [TP nay β x [CP prox [C ′ [FinP PR O 1 ... ⟨ PRO 1 ⟩ ...] [ C -ko]] ...] C] binding
predication
While Landau (2015: 39) assumes that context coordinates are normally only implicitly given outside of logophoric control scenarios, the control data for PNI-ed 1st /2nd person controllers showcase the need for projection of context coordinates in matrix clauses. Note also that this way of overcoming blocked control readings by PNI-ed controllers can only be available to 1st /2nd person pronouns, as only referents of participant pronouns can be coreferential with speaker and addressee of the utterance context.
6.5 German German PNI does not effect case marking, which makes it increasingly more difficult to diagnose. Recall from Section 5.4.5 that Frey (2015) takes bare plurals and non-specific indefinites to be pseudo-incorporated, if they are able to denote a conventionalized activity together with the verb. Although Frey does not
6.5 GERMAN
219
discuss binding or control structures, the following contrasts suggest that parallel restrictions can be observed for PNI-ed objects in German. In (61), I provide the paradigm for bare plurals, where I test binding of a possessive pronoun into a locative adjunct. The possessive both in situ (61c) as well as in topicalized position (61d) seems to disallow a covaried reading with the bare plural and prefers coreference with a salient antecedent in the discourse. (61a) and (61b) provide baselines without binding. Since planting flowers is a likely PNI context, we can attribute the ban on bound readings to the inability of PNI-DPs to introduce a binder prefix.
(61) Binding for bare plurals
German
a. Frank wird Blumen in die Ka¨sten pflanzen. Frank.NOM will flowers.ACC in the boxes.ACC plant ‘Frank will plant flowers in the boxes.’ b. [In die Ka¨sten]3 wird Frank Blumen ⟨PP3 ⟩ pflanzen. in the boxes.ACC will Frank.NOM flowers.ACC plant ‘Frank will plant flowers in the boxes.’ c. Frank wird Blumen1 in ihre?1∕2 Ka¨sten pflanzen. Frank.NOM will flowers.ACC in POSS.ACC boxes.ACC plant ‘Frank will plant flowers in ? their/her boxes.’ d. [In ihre??1∕2 Ka¨sten]3 wird Frank Blumen1 ⟨PP3 ⟩ in POSS.ACC boxes.ACC will Frank.NOM flowers.ACC pflanzen. plant ‘Frank will plant flowers in ?? their/her boxes.’
If we exchange bare plurals with indefinite objects, as in (62a) and (62b), we see that they can act as local binders for the possessive pronoun only if they are interpreted as specific. Non-specific readings, shown in (63a) and (63b), do not seem to be possible under co-indexation with the possessive. The possessives receive a co-reference reading with a local discourse antecedent. Following the arguments in Section 5.4.5, there are good reasons to treat specific indefinites in German as generalized quantifiers which are in complementary distribution with the PNI determiner. Consequently, only specific indefinites can establish a binding relation with the possessive.
220
PNI-PROPERT Y II: L ACK OF BINDING AND CONTROL
(62) Binding for specific indefinites
German
a. Frank wird [eine[+spec] Blume]1 in ihren1∕2 Kasten Frank.NOM will a.ACC flower.ACC in POSS.ACC box.ACC pflanzen. plant ‘Frank will plant a (certain) flower in its/her box.’ b. [In ihren1∕2 Kasten]3 wird Frank [eine[+spec] Blume]1 in POSS.ACC boxes.ACC will Frank.NOM a.ACC flower.ACC ⟨PP3 ⟩ pflanzen. plant ‘Frank will plant a (certain) flower in its/her box.’ (63) Binding for non-specific indefinites
German
a. Frank wird [eine[−spec] Blume]1 in ihren?1∕2 Kasten Frank.NOM will a.ACC flower.ACC in POSS.ACC box.ACC pflanzen. plant ‘Frank will plant a (non-specific) flower in ? its/her box.’ b. [In ihren??1∕2 Kasten]3 wird Frank [eine[−spec] Blume]1 in POSS.ACC boxes.ACC will Frank.NOM a.ACC flower.ACC ⟨PP3 ⟩ pflanzen. plant ‘Frank will plant a (non-specific) flower in ?? its/her box.’ We can construct object control structures that show similar restrictions. Although object controllers can take the form of bare plurals and indefinites, only specific readings seem to be possible; see (64) and (65). As with the other control data, we cannot be entirely sure that the control predicates themselves qualify as PNI predicates. This is particularly an issue in German due to the conventionalization/frequency presupposition, discussed in Section 5.4.5, as well as the fact there is no case drop which could provide independent evidence for PNI.⁶ ⁶ It should also be noted that German bare plurals have been argued to undergo PNI while still being able to act as controllers. Müller (2018) provides the example in (1), suggesting that the controller receives a non-specific reading. As discussed above, it is not clear whether Passanten auffordern constitutes a PNI context. If not, the non-specific existential reading can still arise via the DKP rule (Chierchia 1998); recall footnote 22. (i) Müller (2018: 3) German Sie hat umsta¨ndlich [Passanten]1 aufgefordert [PRO1 nach Hause zu gehen]. she has circuitously passersby.ACC asked to home to go.INF ‘She circuitously asked passersby to go home.’
6.5 GERMAN
(64) Object control for bare plurals
221
German
a. Frank hat Kinder1 gebeten [PRO1 nicht auf der Straße zu Frank.NOM has kids.ACC asked not on the street to spielen]. play.INF ‘Frank asks kids not to play on the street.’
[+SPEC]
b. Frank hat Lehrer1 u¨berredet [PRO1 keine Hausaufgaben Frank.NOM has teachers.ACC persuades no homework aufzugeben]. assign.INF ‘Frank persuades teachers not to assign homework.’ (65) Object control for indefinites
[+SPEC] German
a. Frank bittet [ein Kind]1 [PRO1 nicht auf der Straße zu Frank.NOM asks a.ACC kid.ACC not on the street to spielen]. play.INF ‘Frank asked a kid not to play on the street.’
[+SPEC]
b. Frank u¨berredet [einen Lehrer]1 [PRO1 keine Hausaufgaben Frank.NOM persuaded a.ACC teacher.ACC no homework aufzugeben]. assign.INF ‘Frank persuaded a teacher not to assign homework.’
[+SPEC]
This chapter has shown how binding and control data mostly align with the observations we have made in Chapters 4 and 5. Noun types identified as pseudoincorporated, based on a correlation between scope inertness and case drop (or positional restrictions in German), are furthermore shown to be unable to establish binding and control relations. The next chapter is concerned with a third diagnostic that takes the movement characteristics into account.
7 PNI-property III: Movement patterns While most PNI languages restrict the movement properties of pseudoincorporated arguments, they vary in terms of how far PNI-ed arguments are able to dislocate from the verb. Tamil and Sakha, e.g., have been described as PNI languages that require surface adjacency between the caseless object and the verb (Baker 2014b, 2015; Baker and Vinokurova 2010; Levin 2015). Hindi has been shown to allow PNI-ed objects to scramble over subjects; see (1) and (2).¹ Short scrambling, however, is not an option; see (3). (1) Context: Today the decision ought to be made. Gambhir (1981: 303) Lekin, faislaa1 ham log roz hii ⟨DP1 ⟩ karte haı˜. Hindi but decision we people daily EMPH do are ‘But, we make decisions every day.’ (2) Dayal (2011: 137)
Hindi
a. Kitaab1 anu zaroor ⟨DP1 ⟩ becegii. book Anu definitely sell-FUT ‘Anu will definitely sell books.’ b. Kitaab1 anu ⟨DP1 ⟩ becegii, akhbaar nahiiN. book Anu sell-FUT newspaper not ‘Anu will sell books, not newspapers.’ (3)
Bhatt and Anagnostopoulou (1996: 137)
Hindi
a. Ram-ne Anita-ko chitthii bhej-ii. Ram-ERG Anita-ACC letter.F send-PFV.F ‘Ram sent the letter to Anita.’ ¹ PNI-ed objects might even scramble out of finite clauses as the structure below suggests. The translation, however, suggests a definite reading; see also footnote 9 for definite readings with unmarked objects in Hindi. (i) (Gambhir 1981: 216) Hindi Kitaab1 , meraa khayaal hai [ki Dev-ne us-ko aaj ⟨DP1 ⟩ de dii hai. book my opinion is that Dev-ERG 3SG-DAT today give.PST PFV has ‘I think Dev has given the book to him today.’
Pseudo-Noun Incorporation and Differential Object Marking. Imke Driemel, Oxford University Press. © Imke Driemel (2023). DOI: 10.1093/oso/9780192866400.003.0007
PNI-PROPERT Y III: MOVEMENT PATTERNS
223
b. Ram-ne chitthii*(-ko)1 Anita-ko ⟨DP1 ⟩ bhej-ii. Ram-ERG letter.F-ACC Anita-ACC send-PFV.F ‘Ram sent the letter to Anita.’ Spanish bare plurals cannot be case marked and, according to López (2012: 48), should not be able to move. They normally occur post-verbally, as in (4), although structures like in (5) suggest that bare plurals are not completely immobile. Left dislocation, however, is not permitted for caseless objects (Leonetti 2004; Melis 1995; Pensado 1995), shown in (6) for weak quantifiers and in (7) for numerals. (4)
López (2012: 20) Yo contracto (*a) traductores. 1SG hire DOM translators ‘I hire translators.’
Spanish
(5)
Leonetti (2004: 97) Bueno, este verano, [libros]1 , han leı´do ⟨DP1 ⟩. well this summer books have.3PL read ‘Well, this summer, books they have read.’
Spanish
(6) Leonetti (2004: 86)
Spanish
a. Ya conocı´a (a) muchos estudiantes. already knew.1SG DOM many students ‘I already knew many students.’ b. [*(A) muchos estudiantes]1 ya los conocı´a ⟨DP1 ⟩. DOM many students already them knew.1SG ‘Many students I already knew.’ (7) Leonetti (2004: 86)
Spanish
a. Habı´an incluido (a) dos catedra´ticos en la lista. had.3PL included DOM two professors in the list ‘They included two professors in the list.’ b. [*(A) dos catedra´ticos]1 , los habı´an incluido ⟨DP1 ⟩ en la lista. DOM two professors them had.3PL included in the list ‘Two professors they included in the list.’ The previous chapters established that Tamil and Hindi pseudo-incorporate bare nouns, while Turkish and Mongolian additionally pseudo-incorporate indefinites, numerals, and weak quantifiers. German seems to let bare plurals and indefinites undergo PNI, whereas Korean singles out bare nouns, proper names, weak definites, and 1st /2nd person pronouns. Table 7.1 below summarizes the movement
224
PNI-PROPERT Y III: MOVEMENT PATTERNS
qualities of PNI-ed objects from the core data set of this study and the data points taken from the literature. Table 7.1 Movement restrictions of noun types which can undergo PNI
The languages listed in Table 7.1 are all documented to display scrambling of casemarked nominal arguments, which will be supported by the data given in this chapter. Movement properties of PPs are also investigated for comparison; they pattern with case marked nominals.² The data from Hindi and Spanish above as well as the summary of the core data set in Table 7.1 signal that movement of PNI-ed arguments is generally restricted, in contrast to case-marked nouns and PPs. Yet not every language shows the same restrictions. In order to account for the cross-linguistic variation, one would ideally adhere to individual properties of the languages under consideration. The present approach is able to do so in a straightforward manner. Since PNI-ed arguments constitute hybrid categories which start off as nominals but turn into verbal arguments, once they have been c-selected, we predict their movement patterns to parallel VP movement in the respective languages. Since VP-movement is often more restricted than DP-movement cross-linguistically, the PNI-movement patterns are not surprising. I assume that scrambling and topicalization is driven by categorial features. While this is a commonly shared assumption for nominal arguments undergoing scrambling, EPP-movement, or object shift (Alexiadou and Anagnostopoulou 1998; Bailyn 2003; Chomsky 1995; Epstein et al. 1998; Kishimoto 2000; Kitahara 1997; Müller 2010), it has also been proposed for VP/vP/PredP-movement (Collins 2017; Mahajan 2003; Massam 2001; Massam and Smallwood 1997; Müller 2004b; van Urk 2019a). Hence, DP-movement is triggered by [•D•], whereas VP-movement as well as PNI-movement results from the presence of [•V•] on the respective functional heads in the clause (T for intermediate scrambling and ² I thank an anonymous reviewer for suggesting this additional test.
PNI-PROPERT Y III: MOVEMENT PATTERNS
225
C for long scrambling and/or topicalization). As was pointed out in Section 3.2, movement obeys the PIC, where vP and CP are assumed to be phases (Chomsky 2000, 2001).³ Phrases can escape phases by the optional insertion of an edge feature (EF) on phase heads triggering movement into phase head specifiers. Edge features are category-neutral, they are inserted if the current phase of the derivation is unbalanced—an idea I adopt from Heck and Müller (2003). In order to avoid the notorious look-ahead problem EF insertion creates, we take the numeration into account and demand for each structure-building feature a potentially available matching feature, either in the numeration or at edge of the current phase. If this is not the case, the phase is unbalanced, thereby necessitating the insertion of an edge feature on the phase head. EF insertion is in principle optional, but if it does not happen in cases where a phase is unbalanced, the derivation will crash eventually, since the movement-inducing feature will not be checked off. The definitions are given in (8)–(10). (8)
Balanced Phase A phase is balanced iff, for every movement-inducing feature [•F•] in the numeration, there is a distinct potentially available feature [F]. (Müller 2011: 128)
(9) Potential availability A feature [F] is potentially available at the XP level if (i) or (ii) holds: i. [F] is on X or edgeX of the present root of the derivation. ii. [F] is in the workspace of the derivation. (Müller 2011: 128) (10) Edge Feature Condition An edge feature [•X•] can be assigned to the head γ of a phase only if i., ii., and iii. hold: i. The phase headed by γ is otherwise not balanced. ii. γ has not yet discharged all its structure-building or probe features. iii. [•X•] ends up on top of γ’s list of structure-building features. (Müller 2011: 169) Derivations which contain two potential goals for a movement-inducing feature, e.g. a DP subject and DP object, will target the goal which is accessible, i.e. not contained in the current spell-out domain. Usually, this will be the subject. A scrambling language is able to make more than one goal available by a movementinducing feature on v which allows an object to escape the vP phase. If there is no such feature, vP phases will always be balanced since any [•D•] on a higher functional head will find its matching feature on the subject DP which is still in the numeration at the point when v is merged. This way, DP subjects can act as ³ For an account of the movement patterns without phase theory; see Driemel (2020).
226
PNI-PROPERT Y III: MOVEMENT PATTERNS
interveners for movement of object DPs without making reference to the Minimal Link Condition (Fanselow 1990; Chomsky 1995). For PNI contexts, intervention will not play a role, at least not for short and intermediate scrambling. Due to the fact that PNI readings are mostly tied to conventionalized activities, often only detectable for objects, scenarios in which two PNI-DPs enter the derivation are hardly attested. The only other potential intervener is VP which is, by definition, already inaccessable at the point when phase head v undergoes Merge. The situation is different for long scrambling, where matrix V is still in the numeration at the point when phase balance is evaluated for embedded CP. Matrix V acts as an intervener for any movement triggered by [•V•] on a functional head within the matrix clause, and thus blocks EF insertion. Thus, scrambling languages which allow PNI-ed arguments to undergo long scrambling must be able to let C introduce [•V•], otherwise PNI-DPs will not be able to escape embedded CPs. In a phase-based approach like the one proposed here, PNI-ed objects scramble above the subject in a cyclic manner, that is an edge feature attracts them to spec,vP to escape phasal spell-out, followed by movement to the criterial position spec,TP. Topicalization proceeds in a similar fashion, the only difference being the final landing site spec,CP. Long scrambling requires [•V•] on embedded C to ensure that PNI-ed objects can escape the embedded clause, with an additional [•V•] feature on matrix T, so that they reach the criterial position in which they end up preceding the matrix subject. The presence/absence of [•V•] on C will be the crucial assumption to derive the difference between Korean and Turkish with respect to long scrambling. Korean PNI-ed arguments are prevented from escaping the embedded CP since it lacks [•V•] on C, while Turkish C can come with [•V•], thereby enabling long scrambling of PNI-ed arguments. Finally, I implement short scrambling by a combination of movement-inducing features on T, thus T[•V• ≻ •D•] . This triggers EF insertion on v, moving a PNI-ed object to spec,vP. Both subject and direct object are subsequently attracted by T, where PNI-DP is moved first, followed by subject movement. Indirect arguments are introduced by an Appl head (Marantz 1993; Pylkkänen 2008) which takes VP as a complement.⁴ Table 7.2 provides an overview over the featural set-up of each PNI language, deriving the different PNI movement patterns. The following sections will provide evidence that PNI-movement patterns parallel those of VP-movement in each of the PNI languages in the study. Based on those observations, I will provide the relevant derivations, following the assumptions in Table 7.2.
⁴ One could imagine a simpler implementation of short scrambling, e.g. by introducing [•V•] on v which moves the direct object to spec,vP from where it will c-command the indirect object introduced by an Appl head. This option, however, overgenerates with respect to long scrambling, as the discussions for Turkish and Korean will make clear.
7.1 TAMIL
227
Table 7.2 Feature set-up for movement of PNI-ed arguments
Mongolian Tamil Turkish Korean German
SHORT SCR
INTERMED SCR
TOP
LONG SCR
– –
– – T[•V•] T[•V•] –
– – – – C[•V•]
– – T[•V•] , C[•V•] – –
T[•V• ≻ •D•] T[•V• ≻ •D•] –
7.1 Tamil Tamil has been described as a free word order language (Annamalai and Steever 1998; Lehmann 1993; Sarma 1999, 2003). Generally, we can observe that casemarked arguments are able to undergo short, intermediate, and long scrambling. I provide a paradigm in (11) where I also indicate the underlying assumptions about the movement paths of the arguments under consideration, anticipating the analysis which will be given in the latter part of this section.⁵ (11) Scrambling of case marked DPs in Tamil a. Naan pustagath-ai1 anda ponnu-kiʈʈe ⟨DP1 ⟩ kuɖu-tt-een. 1SG.NOM book-ACC DEM girl-LOC give-PST-1SG ‘I gave a book to that girl.’ [SHORT SCR] b. Pustagath-ai1 naan anda ponnu-kiʈʈe ⟨DP1 ⟩ kuɖu-tt-een. book-ACC 1SG.NOM DEM girl-LOC give-PST-1SG ‘I gave a book to that girl.’ [INTERMED SCR] c. Book-ai1 Mani [Banu ⟨DP1 ⟩ padi-ch-aal-nnu] book-ACC Mani.NOM Banu.NOM read-PST-3SG.F-COMP so-n-aan. say-PST-3SG.M ‘Mani said that Banu read a book.’
[LONG SCR]
PPs can also move freely, which is demonstrated with the postposition -kiʈʈe (Lehmann 1993; Sundaresan 2006) in (12). ⁵ I should point out that the analysis for short scrambling and the underlying ditransitive structures assumed here in this study are not in line with the proposal in Sundaresan (2006) who adopts a Marantz-style (1993) structure for ditransitive kuɖu involving prepositional objects marked with -kiʈʈe; see below. (i) [VP DPtheme [V ʹ PPgoal V ] ]
(Sundaresan 2006: 397)
228
PNI-PROPERT Y III: MOVEMENT PATTERNS
(12) Scrambling of PPs in Tamil a. [anda ponnu-kiʈʈe]2 naan ⟨PP2 ⟩ pustagath-ai kuɖu-tt-een. DEM girl-LOC 1SG.NOM book-ACC give-PST-1SG ‘I gave a book to that girl.’ [INTERMED SCR] b. [anda ponnu-kiʈʈe]2 Mani [naan ⟨PP2 ⟩ pustagath-ai DEM girl-LOC Mani.NOM 1SG.NOM book-ACC kuɖu-tt-een-nnu] so-n-aan. give-PST-1SG-COMP say-PST-3SG.M ‘Mani said that I gave a book to that girl.’
[LONG SCR]
PNI-ed arguments, however, seem to be immobile in Tamil. In Section 4.3.1, we have identified bare nouns as the only noun type which can undergo PNI. Baker (2014b) observes that bare nouns cannot move over an indirect object; see (13). (13) Baker (2014b: 8–9) a.
Naan anda pombale-kiʈʈe pustagam kuɖu-tt-een 1SG.NOM the woman-LOC book give-PST-1SG.SBJ ‘I gave a book to the woman.’
b. *Naan pustagam anda pombale-kiʈʈe kuɖu-tt-een 1SG.NOM book the woman-LOC give-PST-1SG.SBJ ‘I gave a book to the woman.’ This observation was confirmed by my consultants.⁶ Furthermore, intermediate and long scrambling is also banned. The complete paradigm for the movement patterns of caseless bare objects is given in (14). (14) Scrambling of PNI-ed objects in Tamil a.
Naan anda ponnu-kiʈʈe pustagam kuɖu-tt-een. 1SG.NOM DEM girl-LOC book give-PST-1SG ‘I gave a book to the girl.’
[BASELINE]
b. ?? Naan pustagam1 anda ponnu-kiʈʈe ⟨DP1 ⟩ kuɖu-tt-een. 1SG.NOM book DEM girl-LOC give-PST-1SG ‘I gave a book to the girl.’ [SHORT SCR] c. *Pustagam1 naan anda ponnu-kiʈʈe ⟨DP1 ⟩ kuɖu-tt-een. book 1SG.NOM DEM girl-LOC give-PST-1SG ‘I gave a book to the girl.’ [INTERMED SCR] ⁶ What I could not confirm is that PNI-ed objects in Tamil require surface adjacency, as it is claimed in the literature (Baker 2014b; Levin 2015). The adjacency data are discussed in Section 9.1.
7.1 TAMIL
229
d. *Pustagam1 Mani [Banu ⟨DP1 ⟩ padi-ch-aal-nnu] book Mani.NOM Banu.NOM read-PST-3SG.F-COMP so-n-aan. say-PST-3SG.M ‘Mani said that Banu read a book.’
[LONG SCR]
Based on the following discussion, we can tentatively conclude that VPs show the same movement restrictions. Long scrambling of VPs seems to be unacceptable, judging by the contrasts in (15) and (16).⁷ Note that tense and agreement morphology is affixed to the verb, suggesting potential V-to-T movement which would provide an alternative explanation for the unacceptability of (15b) and (16b): VPs that contain inflected verbs would not be able to occur in a dislocated position because head movement will bleed VP-movement. Unfortunately, and in contrast to the PNI languages discussed in the following sections, Tamil does not provide viable tests for VP-remnant movement. Hence, the VP-fronting data with respect to long scrambling are not entirely conclusive. (15) Long scrambling of VPs? a.
Tamil
Mani [Banu book-ai padi-ch-aal enru] Mani.NOM Banu.NOM book-ACC read-PST-3SG.F that so-n-aan. say-PST-3SG.M ‘Mani said that Banu read a book.’
b. *[Book-ai padi-ch-aal]1 Mani [Banu ⟨VP1 ⟩ enru] book-ACC read-PST-3SG.F Mani.NOM Banu.NOM that so-n-aan. say-PST-3SG.M ‘Mani said that Banu read a book.’ (16) Long scrambling of VPs? a.
Tamil
Mani [Banu book-ai padi-ch-aal] en-r-aan. Mani.NOM Banu.NOM book-ACC read-PST-3SG.F say-PST-3SG.M ‘Mani said that Banu read a book.’
b. *[Book-ai padi-ch-aal]1 Mani [Banu ⟨VP1 ⟩] book-ACC read-PST-3SG.F Mani.NOM Banu.NOM en-r-aan. say-PST-3SG.M ‘Mani said that Banu read a book.’ ⁷ The complementizer enru is derived from the verb en (=say) (Lehmann 1993; Schiffman 1999). Since Annamalai and Steever (1998: 172) claim that enru cliticize to the embedded verb, I provide structures with and without the complementizer, so that there is no independent reason why VP-movement should be blocked.
230
PNI-PROPERT Y III: MOVEMENT PATTERNS
More can be said about VP-movement within clauses. In the following, we will take the data set given in (17) and compare a VP-movement analysis with an alternative proposal that takes post-verbal constituents to come about via rightward movement. The translations in (17) signal a very particular information structure; they are often translated as clefts or pseudo-clefts in which the post-verbal constituent constitutes the pivot, or alternatively as mono-clausal structures with narrow focus on the post-verbal phrase. (17) Short and intermediate scrambling of VPs?
Tamil
a. Naan [book-ai kuɖu-tt-een] anda ponnu-kiʈʈe. 1SG.NOM book-ACC give-PST-1SG DEM girl-LOC ‘The one who I gave a book to is that girl.’ b. [Book-ai padi-ch-aal] Banu. book-ACC read-PST-3SG.F Banu.NOM ‘The one who read a book is Banu.’ Postverbal constituents in Dravidian and Tibeto-Burman languages are overwhelming classified as pivots of underlining clefts, not only because of their characteristic information structure but also due to the presence of an overtly spelled-out copula, often co-occurring to the right of the cleft clause (Annamalai and Steever 1998; Bhattacharya and Devi 2004; Krishnamurti 1998, 2003; Lehmann 1998); see for example comparable structures in Malayalam and Meiteilon in (18) and (19). Hence, bi-clausal analyses are prevalent (Bhattacharya and Devi 2004; Jayaseelan 2001; Jayaseelan and Amritavalli 2005, 2017; Madhavan 1987; Selvanathan 2017). (18) Jayaseelan (2001: 64) n˜aan innale kaND-atә Mary-(y)e aaNә. I yesterday saw-NMLZ Mary-ACC is ‘It is Mary that I saw yesterday.’
(19) Bhattacharya and Devi (2004: 5) Hui-nә cakhi-bә (pot) әdu sem-ni. dog-NOM ate-NMLZ/INF thing DET apple-COP ‘It is an apple that the dog ate.’
Malayalam
Meiteilon
7.1 TAMIL
231
Tamil, however, does not spell-out the copula, thereby making it more difficult to diagnose a cleft. The examples in (20) prove that neither a predicative nor an identificational copula sentence requires an overt copula. (20) Sarma (1999: 90), Jegan Murugesan (p.c.) Tamil Dharmaa suudaaTakkaaran. / Inta suudaaTakkaaran Dharmaa. Dharma.NOM gambler.NOM / DEM gambler.NOM Dharma.NOM ‘Dharma is a gambler. / This gambler is Dharma.’ Sarma (1999, 2003) discusses two variants of Tamil post-verbal constituent structures, one in which the verb shows full agreement (see (21a) and (21b)) and one with default agreement – or alternatively analysed as a nominalizer (the two are homophonous in Tamil). The second variant is shown in (21c). (21) Sarma (1999: 95,60,90) a. Dharma toT-r-aan bhiimaav-ai. Dharma.NOM lose-PST-3SG.M Bhima-ACC ‘Dharma lost BHIMA.’
Tamil
b. Shakuni dharmaa-kku kodu-tt-aan daayatt-ai. Shakuni.NOM Dharma-DAT give-PST-3SG.M dice-ACC ‘It is the dice that Shakuni gave to Dharrna.’ c. Dharma toT-r-adu draupadi-ai. Dharma.NOM lose-PST-NMLZ/3SG.N Draupadi-ACC ‘It was Draupadi that Dharma lost to.’ Let us first consider the default agreement/nominalizer version. These structures have received a bi-clausal treatment by Sarma (1999, 2003), and more recently Selvanathan (2017), in which the clefted phrase originates in the cleft clause but moves to a designated focus position. The accounts are presented in (22) and (23), applied to (21c). While Sarma takes -adu to be the spell-out of N, a nominal layer above the cleft clause, Selvanathan analyses -adu as the result of anti-agreement (Ouhalla 1993; Schneider-Zioga 2007) between an N-layer and the clefted phrase, i.e. ϕ-agreement is disrupted because the clefted phrase is involved in another A’-dependency, here the Foc head.⁸ ⁸ Note that Sarma’s analysis requires sideward movement, an operation which forms a copy and remerges this copy with an unconnected phrase marker, assembled independently in the workspace (Hornstein and Nunes 2002; Nunes 2001). Sideward movement is different from ordinary movement, in that (i) there is no c-commanding trigger and (ii) the moved phrase is merged in a tree from which it was not taken.
232
PNI-PROPERT Y III: MOVEMENT PATTERNS
(22) Sarma (1999: 89):
CopP NP
Cop ′ N Draupadi-ai i -adu
TP
CopØ
FocP
TP T′
Dharma j
⟨Draupadi-ai⟩i FocØ T -r-
vP v′
⟨Dharma⟩ j
v
VP ⟨Draupadi-ai⟩i
V toT-
(23) Selvanathan (2017: 12,18):
IP
NPj
I′ N′
⟨Draupadi-ai⟩i
N -adu
AspP Asp -r-
vP
VP
⟨Draupadi-ai⟩i V toT-
Asp
FocP
Draupadi-ai i
Foc′
v′
Dharma
Pred′ v
I
AspP
⟨NP⟩j
FocØ PredØ
⟨Draupadi-ai⟩i ...
The motivation for such analyses comes from case connectivity and reflexive binding of the clefted phrase by an antecedent in the cleft clause; see (24). This property was first observed by Sarma (1999) and then later on also discussed by Selvanathan (2017). (24) Reconstruction for Principle A with -adu (Sarma 1999: 87) Dharma1 daayatt-il toT-r-adu tan1 manaivi-ai. Dharma.NOM dice-LOC lose-PST-NMLZ REFL wife-ACC ‘It was his1 wife that Dharma1 lost to in the game of dice.’ Sarma also provides an analysis for the full agreement versions in (21a) and (21b) which is identical to the structure in (22), minus the nominal layer and the copula phrase. The structure in (25) models the sentence in (21a).
7.1 TAMIL
(25) Sarma (1999: 95):
233
TP FocP
TP
Tʹ
Dharma j
Bhima-aii
FocØ
T -r-aan
vP
vʹ
⟨Dharma⟩ j
v
VP
⟨Bhima-ai⟩i
V toT-
Case connectivity also holds for full agreement structures, while (26) shows that reconstruction for reflexive binding is similarly required. (26)
Reconstruction for Principle A with full agreement Dharma1 daayatt-il toT-r-aan tan1 manaivi-ai. Dharma.NOM dice-LOC lose-PST-3SG.M REFL wife-ACC ‘It was his1 wife that Dharma1 lost to in the game of dice.’
Movement of the clefted phrase additionally triggers weak crossover effects, shown by Sarma for the full agreement versions; see (27). Again, we can make a parallel observation for the default agreement/nominalizer versions, shown in (28). (27)
Weak crossover effect with full agreement (Sarma 2003: 244) *Avan-uDaia1 aNNaa daayatt-il toT-r-aan ellaar-ai-um1 . 3SG-GEN brother.NOM dice-LOC lose-PST-3SG.M everyone-ACC ‘His1 brother lost to everyone1 in the game of dice.’
(28)
Weak crossover effect with -adu *Avan-uDaia1 aNNaa daayatt-il toT-r-atu ellaar-ai-um1 . 3SG-GEN brother.NOM dice-LOC lose-PST-3SG.N everyone-ACC ‘His1 brother lost to everyone1 in the game of dice.’
While case connectivity and reconstruction for Principle A are arguments that the clefted phrase undergoes movement out of the cleft clause, weak crossover effects indicate that it must be Ā-movement, which fits well with the information structural nature that is proposed under those accounts.
234
PNI-PROPERT Y III: MOVEMENT PATTERNS
Note that neither of the analyses makes use of VP-movement, which seems to support the hypothesis that the data in (17) are derived by movement of the postverbal constituent, in line with the assumption that VPs are immobile in Tamil. It is worth noting, however, that the observations made above can potentially be also made to follow from a VP-movement analysis. Let us assume for the moment the analysis in (29) where the focused phrase vacates the VP into a leftward branching focus position, with subsequent leftward VP-movement across the focused phrase. This type of analysis has in fact been proposed for equivalent postverbal structures in Malayalam (Jayaseelan 2001, 2004). Case can be assigned in the objects first-merged position, the same position into which the object reconstructs for reflexive binding. Weak crossover is not ameliorated since the object movement is triggered by a focus feature, arguably an Ā-type movement. We can understand the analysis in (29) as the full agreement counterpart version of (23). (29) An alternative VP-movement analysis? a.
FocP Foc′
DPF vP
Foc[uF] v′
DP VP ⟨DPF⟩
b.
v V
TP
T′
DP
T[EPP]
FocP FocP
VP ⟨DPF⟩
Foc′
V DPF vP ⟨DP⟩
Foc v′
⟨VP⟩
v
7.1 TAMIL
235
There is, however, one piece of evidence showing that rightward movement for post-verbal constituents has to be at least an option. Tamil bans subextraction, i.e. extraction out of phrase which has already been moved. This is shown with the set of examples in (30)–(32). An embedded CP can in principle move over the matrix subject, compare (30a) to (30b). (31) shows that arguments can also dislocate to the right periphery across clause boundaries. Both movement operations, however, cannot co-occur, as the contrast in (32) exemplifies. Sarma attributes the unacceptability of (32b) to a Freezing effect (Ross 1967; Culicover and Wexler 1977). Like in (30b) the embedded CP has been topicalized. Crucially, further subextraction by rightward movement of the clefted phrase is not licensed. (30) Topicalization of embedded CPs a.
(Sarma 1999: 101)
[TP Dharma [CP2 Shakuni daayatt-il jei-tt-aan Dharma.NOM Shakuni.NOM dice-LOC win-PST-3SG.M enru] son-n-aan]] that say-PST-3SG.M ‘Dharma said that Shakuni won in (the game of ) dice.’
b.
[TP [CP2 Shakuni daayatt-il jei-tt-aan enru]-naa2 Shakuni.NOM dice-LOC win-PST-3SG.M that-TOP Dharma ⟨CP2 ⟩ son-n-aan] Dharma.NOM say-PST-3SG.M ‘That Shakuni won (in the game of ) dice, Dharma said.’
(31)
Movement out of embedded CP (Sarma 1999: 86) [TP Shakuni [CP2 Dharma ⟨DP3 ⟩ toT-r-aan enru] Shakuni.NOM Dharma.NOM lose-PST-3SG.M that ninai-tt-adu draupadi-ai3 ] say-PST-NMLZ Draupadi-ACC ‘It was Draupadi that Shakuni thought that Dharma lost to.’
(32) Tamil exhibits freezing effects a.
(Sarma 1999: 81)
[TP Shakuni [CP2 Dharmaa daayatt-il raajyatt-ai Shakuni.NOM Dharma.NOM dice-LOC kingdom-ACC izha-pp-aan enru] son-n-aan] lose-FUT-3SG.M that say-PST-3SG.M ‘Shakuni said that Dharma will lose (his) kingdom in (the game of ) dice.’
236
PNI-PROPERT Y III: MOVEMENT PATTERNS b. ??[TP [CP2 Dharmaa daayatt-il ⟨DP3 ⟩ izha-pp-aan enru]-naa2 Dharma.NOM dice-LOC lose-FUT-3SG.M that-TOP Shakuni ⟨CP2 ⟩ son-n-aan] raajyatt-ai3 ] Shakuni.NOM say-PST-3SG.M kingdom-ACC ‘Shakuni said that Dharma will lose (his) kingdom in (the game of ) dice.’
If Tamil exhibits freezing effects, we should be able to create similar configurations in mono-clausal structures. In light of the discussion above and under the assumptions in (29), the missing contrast in (33) is unexpected. If postverbal constituents were to come about via remnant VP-movement, it should not be possible to subextract the wh-phrase from the VP in its derived position in (33b), nevertheless this seems possible. (33) Sarma (1999: 66) a. Draupadi avan-uDaia maamaa-kku yaar-ai anupp-in-aaL? Draupadi.NOM 3SG-GEN uncle-DAT who-ACC send-PST-3SG.F ‘Who did Draupadi send to his uncle?’ b. Yaar-ai Draupadi anupp-in-aaL avan-uDaia maamaa-kku? who-ACC Draupadi.NOM send-PST-3SG.F 3SG-GEN uncle-DAT ‘Who did Draupadi send to his uncle?’ The derivation under the theory in (29) is sketched in (34).
(34)
[CP who2 [TP Draupadi [VP ⟨DP1⟩ ⟨DP2⟩ send]3 [Focʹ[to his uncle]1 [vP ⟨VP3⟩ ]]]
An analysis that derives post-verbal constituents directly via rightward movement makes the right prediction for (33) since the arguments undergo movement to the right and to the left periphery independent of each other. The derivation is given in (35). (35)
[CP who2 ... [TP [vPDraupadi ⟨DP1⟩ ⟨DP2⟩ send] [to his uncle]1] ]
7.1 TAMIL
237
This argument, of course, does not exclude the possibility of VP-movement altogether, but it does make it unlikely that VP-movement is responsible for post-verbal constituent structures in Tamil.⁹ We now come back to the analysis of PNI-movement patterns in Tamil. The rest of this section will account for the paradigm given in (14). As is the case with VPs, PNI-ed arguments cannot undergo short, intermediate, or long scrambling. The ban on short and intermediate scrambling is illustrated with the sequence of derivational steps in (36). Recall from Table 7.2 that Tamil is a language which cannot equip the T head with a [•V•]-feature, neither alone nor in combination with another feature. The numeration, as shown in (36a), will always lead to a balanced phase since it contains no movement-inducing feature. Hence, no EF is inserted, PNI-DP is thus trapped inside VP. T enters the derivation in (36d) and assigns nominative case to the subject. Finally, (36e) shows why PNI-DP is immobile: neither is there a movement-inducing feature available nor is PNI-DP accessible due to the PIC. (36) No short or intermediate scrambling in Tamil a. Numeration: {C, T, D,...} → phase balanced b. No EF insertion in ditransitive contexts: v′ P NI -DP 1
v′
N
v
ApplP
D CASE : ACC
A PPL
Appl′
DP
V
N
VP
D CASE : DAT
D
* CASE : ACC *
P NI -DP 1
Appl V
N
D
D
V
V
D
v
* CASE : DAT *
D
[ EF ]
CASE : ACC
V
⁹ One might also doubt the VP-movement analysis in (29) against the background of the Mu¨llerTakano Generalization (Müller 1996, 1998; Takano 1992), which makes reference to the fact that extraction and remnant movement cannot involve the same type of movement. In German e.g. DPscrambling out of infinitival clauses feeds topicalization but bleeds scrambling of remnant infinitives. (i)
Müller (1996: 357–358) a.
[t1 zu lesen]2 hat keiner [das Buch]1 t2 versucht. to read has nobody the book tried ‘No one has tried to read the book.’
b. * dass [t1 zu lesen]2 keiner [das Buch]1 t2 versucht hat. that to read nobody the book tried has ‘that no one has tried to read the book’ The analysis in (29) makes both the focused phrase and the remnant VP target spec,FocP positions, presumably involving the same type of movement. This makes the acceptability of post-verbal constituents in Tamil stand in clear contrast with the observation made for German above.
238
PNI-PROPERT Y III: MOVEMENT PATTERNS c. No EF insertion in transitive contexts: vP P NI -DP 1
v′
N CASE : ACC
V
v
VP
D
V
V
P NI -DP 1 N
D
D
V
EF
* CASE : ACC *
D
v
CASE : ACC
V
d. Case assignment: T′ NOM
vP N
* CASE : NOM *
T
v
VP
D D
v
v′
DP
CASE :NOM
T
V
P NI -DP
V
D
V
N
* CASE : ACC *
D
D
v
* CASE : ACC *
V
e. No trigger for PNI-DP-movement: TP T′
P NI -DP 1 N CASE : ACC
V
T
vP
D
v
DP
v′
N
CASE :NOM
D
v
VP
D P NI -DP 1
* CASE :NOM *
T
V
V
N
D
D
V
* CASE : ACC *
D
v
* CASE : ACC *
V
phase domain
Long scrambling of PNI-ed objects is impossible in Tamil since C does not introduce a [•V•]-feature. Again, the phase is balanced since the numeration does not contain a movement-inducing feature. No EF is inserted, thus PNI-DP does not move to the phase edge in (37b). The PNI-ed argument cannot move to spec,CP in (37c) since C cannot come with [•V•]. Moreover, PNI-DP is inaccessible upon Merge with C due to the PIC. For the same reasons, PNI-DP cannot be attracted by matrix T (not shown here).
7.2 MONGOLIAN
239
(37) No long scrambling in Tamil a. Numeration: {C,T, v, V[...V...] , D, C, T, D,...} → phase balanced b. No EF insertion: vP P NI -DP 1
v′
N * CASE : ACC *
V
v
VP
D P NI -DP 1
V
V
N
D
D
V
EF
* CASE : ACC *
D
v
CASE : ACC
V
c. No trigger for PNI-DP-movement: CP C′
P NI -DP 1 N
C
TP
D CASE : ACC
T
T
vP
V DP N
CASE : NOM
D
...
v′ v
VP
D P NI -DP 1
V
V
N
D
D
V
* CASE : ACC *
V
C
* CASE : ACC *
D
v
phase domain
In this section, I have shown that PNI-ed objects in Tamil are not allowed to undergo any type of scrambling operation. Moreover, I argued against potential positive evidence of VP-scrambling. The analysis is based on the assumption that the unifying movement-inducing feature [•V•] is not available, hence both VPs as well as PNI-ed objects can only be realized in their respective base positions.
7.2 Mongolian Like Tamil, Mongolian shows verb-final base order but with the option to scramble (Janhunen 2012; Guntsetseg 2016; Fong 2019). Mongolian PNI-movement patterns are identical to the one found in Tamil, i.e. pseudo-incorporated objects
240
PNI-PROPERT Y III: MOVEMENT PATTERNS
are not allowed to dislocate from their base positions.¹⁰ Section 4.3.2 has identified bare nouns, indefinites, and weak quantifiers as noun types which can undergo PNI. Guntsetseg (2016) reports that caseless indefinites are not able to perform short (38), intermediate (39) or long scrambling (40), in contrast to the case-marked counterparts. My consultants overall share her judgements.¹¹ (38) Short scrambling of indefinites in Mongolian (Guntsetseg 2016: 107) a. Tujaa Dorzˇi-d neg nom(-yg) o¨g-so¨n. Tujaa.NOM Dorz-DAT a book-ACC give-PST ‘Tujaa gave Dorz a book.’
[BASELINE]
b. Tujaa [neg nom*(-yg)]1 Dorzˇi-d ⟨DP1 ⟩ o¨g-so¨n. [SHORT SCR] Tujaa.NOM a book-ACC Dorz-DAT give-PST ‘Tujaa gave Dorz a book.’ (39) Intermediate scrambling of indefinites in Mongolian (Guntsetseg 2016: 106) a. Zaxiral [neg ojuutn-yg] / [neg ojuutan] ˇsalga-san. director.NOM a student-ACC / a student examine-PST ‘The director examined a student.’ [BASELINE] b. [Neg ojuutn-yg]1 / *[neg ojuutan]1 zaxiral ⟨DP1 ⟩ a student-ACC / a student director.NOM ˇsalga-san. examine-PST ‘The director examined a student.’
[INTERMEDIATE SCR]
¹⁰ It is difficult to test for the movement pattern of PPs, as most postpositions behave like nominals (Janhunen 2012: 203). To the extend that there are some, Guntsetseg (2016: 25) reports in passing that they often occur preverbally but they can also precede the subject. ¹¹ There is some disagreement about the general existence of long scrambling in Mongolian. Fong (2019) argues against movement across finite clause boundaries with data like the one shown below. In fact, only one of my three additional consultants accepts long scrambling of case-marked objects generally. I did not provide contexts, however, which presumably has an impact on the acceptability, especially of long scrambling configurations. The general availability of long scrambling, thus, remains an open questions and needs to be investigated further in future research. (i)
Long scrambling in Mongolian a.
(Fong 2019: 22)
Navchaa chang-aar [Odgerel deeremchn-iig bari-san gej] khel-sen. Navchaa loud-INSTR Odgerel.NOM thief -ACC catch-PST COMP say-PST ‘Navchaa said loudly that Odgerel caught a (certain) thief.’ [BASELINE]
b. * (deeremchn-iig1 ) Navchaa (deeremchn-iig1 ) chang-aar [Odgerel ⟨DP1 ⟩ bari-san thief -ACC Navchaa thief -ACC loud-INSTR Odgerel.NOM catch-PST gej] khel-sen. COMP say-PST ‘Navchaa said loudly that Odgerel caught a (certain) thief.’ [LONG SCR]
7.2 MONGOLIAN
(40) Long scrambling of indefinites in Mongolian
241
(Guntsetseg 2016: 155)
a. Tujaa Bat neg noxoj(-g) surga-san gezˇ med-sen. Tujaa.NOM Bat.NOM a dog-ACC train-PST that know-PST ‘Tujaa knew that Bat trained a dog.’ [BASELINE] b. [Neg noxoj*(-g)]1 Tujaa [Bat ⟨DP1 ⟩ surga-san gezˇ] a dog-ACC Tujaa.NOM Bat.NOM train-PST that med-sen. know-PST ‘Tujaa knew that Bat trained a dog.’
[LONG SCR]
Furthermore, I can confirm that bare nouns and the weak quantifier xeden show identical movement restrictions. The data is provided in (41) to (45). (41) Short scrambling of bare nouns in Mongolian a. Tujaa Dorzˇi-d nom(-yg) o¨g-so¨n. Tujaa.NOM Dorz-DAT book-ACC give-PST ‘Tujaa did book-giving to Dorz.’ b. Tujaa nom*(-yg)1 Dorzˇi-d ⟨DP1 ⟩ o¨g-so¨n. Tujaa.NOM book-ACC Dorz-DAT give-PST ‘Tujaa did book-giving to Dorz.’
[BASELINE]
[SHORT SCR]
(42) Short scrambling of xeden NP in Mongolian a.
Tujaa Dorzˇi-d xeden nom o¨g-so¨n. Tujaa.NOM Dorz-DAT some book give-PST ‘Tujaa gave some books to Dorz.’
b. ??Tujaa [xeden nom]1 Dorzˇi-d ⟨DP1 ⟩ o¨g-so¨n. Tujaa.NOM some book Dorz-DAT give-PST ‘Tujaa gave some books to Dorz.’
[BASELINE]
[SHORT SCR]
(43) Intermediate scrambling of bare nouns in Mongolian a. Zaxiral ojuutan / ojuutn-yg ˇsalga-san. director.NOM student / student-ACC examine-PST ‘The director examined a student.’
[BASELINE]
b. Ojuutan1 / *ojuutn-yg1 zaxiral ⟨DP1 ⟩ ˇsalga-san. student / student-ACC director.NOM examine-PST ‘The director examined a student.’ [INTERMEDIATE SCR]
242
PNI-PROPERT Y III: MOVEMENT PATTERNS
(44) Intermediate scrambling of xeden NP in Mongolian a. Zaxiral [xeden ojuutn-yg] / [xeden ojuutan] ˇsalga-san. director.NOM some student-ACC / some student examine-PST ‘The director examined some student.’ [BASELINE] b. [Xeden ojuutn-yg]1 / *[xeden ojuutan]1 zaxiral ⟨DP1 ⟩ some student-ACC / some student director.NOM ˇsalga-san. examine-PST ‘The director examined some student.’
[INTERMEDIATE SCR]
(45) Long scrambling of xeden NP in Mongolian a. Tujaa Bat xeden noxoj(-g) surga-san gezˇ med-sen. Tujaa.NOM Bat.NOM some dog-ACC train-PST that know-PST ‘Tujaa knew that Bat trained some dog.’ [BASELINE] b. [Xeden noxoj*(-g)]1 Tujaa [Bat ⟨DP1 ⟩ surga-san gezˇ] some dog-ACC Tujaa.NOM Bat.NOM train-PST that med-sen. know-PST ‘Tujaa knew that Bat trained some dog.’
[LONG SCR]
In order to investigate the possibility for VP-movement, we test for the acceptability of post-verbal constituents. The informants of my study rejected any kind of postverbal constituent structure, ranging from adverbials over indirect objects to subjects; see (46). (46) No post-verbal constituents in Mongolian a. *Bi [ene nom-yg unsh-san] xurdan. 1SG.NOM DEM book-ACC read-PST quickly ‘I quickly read this book.’
S[OV ]Adv
b. *Bi [ene nom-yg unsh-san] o¨chigdo¨r. 1SG.NOM DEM book-ACC read-PST today ‘I read this book today.’
S[OV ]Adv
c. *Tujaa [ene nom-yg o¨g-so¨n] Dorzˇi-d. Tujaa.NOM DEM book-ACC give-PST Dorz-DAT ‘Tujaa gave Dorz this book.’
S[OV ]IO
zaxiral. d. *[Ene ojuutn-yg ˇsalga-san] DEM student-ACC examine-PST director.NOM ‘The director examined this student.’
[OV ]S
7.2 MONGOLIAN
243
These results are in line with the observations made in the literature. Guntsetseg (2016: 25) reports that postverbal elements are generally disallowed, except when they are separated by a pause and denote ‘additional information’; see also Poppe (1951: 112) and Binnick (1979: 122). The pragmatic function of these structures have also been described as an afterthought by Janhunen (2012: 228) and Öztürk (2013: 192). The following contrast in (47) highlights the information structural aspect of the construction where | signals a pause. (47) Guntsetseg (2016: 25) a. *Murat o¨chigdo¨r ir-sen Ankarag-aas. Murat today come-PST Ankara-ABL ‘Murat returned from Ankara yesterday.’ b.
Tujaa o¨chigdo¨r ir-sen | German-aas. Tujaa today come-PST Germany-ABL ‘Yesterday, Tujaa came – from Germany.’
Öztürk (2013) offers a bi-clausal analysis in which postverbal constituents in Mongolian are part of a separate clause, adjoined to the preceding one. The second clause is partially deleted under phonological identity with the antecedent clause. In the first clause, the counterpart of the post-verbal constituent is realized as pro. Similar analyses have been carried out for Japanese (Tanaka 2001) as well as Dutch and German (Ott and de Vries 2016). (48) Öztürk (2013: 190) a. [ TP [TP pro1 ... O ... V] [ TP S 1 ... O ... V]] → structure for [OV]S b. [ TP [TP S ... pro1 ... V] [ TP O 1 ... S ... ⟨O 1 ⟩ ... V]] → structure for [SV]O
Apart from the intonational break and the afterthought profile, Öztürk (2013) provides a variety of syntactic tests that argue against a mono-clausal movement analysis. Postverbal constituents are e.g. incompatible with idiom formation; see (49b), where only the literal reading is available. (49) Idiom formation blocked with postverbal constituent (Öztürk 2013: 188) a. Bulgan narma-ig nee-sen. Bulgan nose-ACC burst-PST Literal meaning: ‘Bulgan burst a nose.’ Idiomatic meaning: ‘Bulgan beat someone.’
S[OV ]
244
PNI-PROPERT Y III: MOVEMENT PATTERNS b. [Bulgan nee-sen] narma-ig. Bulgan burst-PST nose-ACC Literal meaning: ‘Bulgan burst a nose.’ Idiomatic meaning: ‘Bulgan beat someone.’
[SV ]O
Moreover, corresponding gaps of the postverbal phrases in the antecedent clause can be pronominalized (50a), in clear contrast to leftward scrambling in Mongolian, for which this strategy is not available (50b). (50) Pronoun formation with postverbal constituent
(Öztürk 2013: 185)
a. [Bulgan pro1 / ter-ig1 unsh-san] nom-ig1 . Bulgan / DEM-ACC read-PST book-ACC ‘Bulgan read this, the book.’
[SV ]O
b. Nom-ig1 Bulgan ⟨nom-ig1 ⟩ / *ter-ig1 unsh-san. book-ACC Bulgan / DEM-ACC read-PST ‘Bulgan read the book.’
O[SV ]
Given the discussion above, we can conclude that postverbal constituents do not result from VP-movement. Rather, postverbal constituents are derived from a complex bi-clausal structure with subsequent deletion of all clausal material in the second clause except for the postverbal constituent. We can, furthermore, observe that VPs do not undergo long scrambling, i.e. movement across a clause boundary; see (51b). (51) No long scrambling of VPs in Mongolian a.
Tujaa [Bat neg noxoj-g surga-san gezˇ] med-sen. Tujaa.NOM Bat.NOM a dog-ACC train-PST that know-PST ‘Tujaa knew that Bat trained a dog.’ [BASELINE]
b. *[neg noxoj-g surga-san]1 Tujaa [Bat ⟨VP1 ⟩ gezˇ] a dog-ACC train-PST Tujaa.NOM Bat.NOM that med-sen. know-PST ‘Tujaa knew that Bat trained a dog.’
[LONG SCR]
Similar to the situation in Tamil, we are faced with the problem of isolating the VP-constituent from the affixed tense morphology. Again, a likely explanation for the fact that we do not find evidence for VP-movement might be the existence of obligatory V-to-T movement which bleeds any dislocation of headed VPs from their base positions. Fortunately, there is a class of low manner adverbs in Mongolian with which we can test remnant VP-movement. Low manner adverbs like dandaa (=always) (Guntsetseg 2016: 25) and xurdan (=quickly) (Fong 2019: 16) cannot occur before the subject, shown in (52). Adverb movement is apparently not an option in Mongolian.
7.2 MONGOLIAN
(52)
245
Fong (2019: 16) Dorj [(*khurdan) Nara (khurdan) baishin (khurdan) bari-san gej] Dorj quickly Nara quickly house quickly build-PST COMP khel-sen. say-PST ‘Dorj said that Nara built a house quickly.’
Such adverbs that show positional restriction can be used to test remnant VPscrambling. Let us assume for now that V raises to T, thus vacating the VP, and low manner adverbs adjoin to VP. By dislocating the adverb together with the direct object, we can probe for VP-movement. The following set of examples tests remnant VP-scrambling within a transitive clause. (53a) proves again the positional restrictions of xurdan. Crucially, the unacceptability of (53b) provides evidence that VPs do not scramble across a subject. The adverb cannot precede the subject, not even if it is contained in VP. (53) No intermediate scrambling of remnant VPs in Mongolian a.
(*xurdan) zaxiral (xurdan) neg ojuutn-yg ˇsalga-san. quickly director.NOM quickly a student-ACC examine-PST ‘The director quickly examined a student.’ [BASELINE]
b. *[VP xurdan neg ojuutn-yg]1 zaxiral ⟨VP1 ⟩ ˇsalga-san. quickly a student-ACC director.NOM examine-PST ‘The director quickly examined a student.’ [INTERMEDIATE SCR] The same pattern emerges in ditransitive scenarios, where (54b) and (54c) show that short as well as intermediate scrambling is not an option for remnant VPs. (54) No short or intermediate scrambling of remnant VPs in Mongolian a.
Tujaa Dorzˇi-d xurdan neg nom-yg butsaa-j Tujaa.NOM Dorz-DAT quickly a book-ACC give.back-CV.IMPF o¨g-so¨n. give-PST ‘Tujaa quickly gave Dorz a book back.’
[BASELINE]
b. *Tujaa [VP xurdan neg nom-yg]1 Dorzˇi-d ⟨VP1 ⟩ Tujaa.NOM quickly a book-ACC Dorz-DAT butsaa-j o¨g-so¨n. give.back-CV.IMPF give-PST ‘Tujaa quickly gave Dorz a book back.’
[SHORT SCR]
246
PNI-PROPERT Y III: MOVEMENT PATTERNS c. *[VP xurdan neg nom-yg]1 Tujaa.NOM Dorzˇi-d ⟨VP1 ⟩ quickly a book-ACC Tujaa Dorz-DAT butsaa-j o¨g-so¨n. give.back-CV.IMPF give-PST ‘Tujaa quickly gave Dorz a book back.’
[INTERMEDIATE SCR]
Neither is long scrambling permitted with remnant VPs, exemplified with the paradigm in (55). The adverb sain (=well) is another low manner adverb which cannot precede the subject; see (55a). Scrambling the remnant VP across a sentence boundary is banned, i.e. neither (55b) nor (55c) produces acceptable results. (55) No long scrambling of remnant VPs in Mongolian a.
Tujaa [(*sain) Bat (sain) neg noxoj-g (sain) surga-san Tujaa.NOM well Bat.NOM well a dog-ACC well train-PST gezˇ] med-sen. that know-PST ‘Tujaa knew that Bat trained a dog well.’
[BASELINE]
b. *[VP neg noxoj-g sain] Tujaa [Bat ⟨VP1 ⟩ surga-san gezˇ] a dog-ACC well Tujaa.NOM Bat.NOM train-PST that med-sen. know-PST ‘Tujaa knew that Bat trained a dog well.’
[LONG SCR]
c. *[VP sain neg noxoj-g] Tujaa [Bat ⟨VP1 ⟩ surga-san gezˇ] well a dog-ACC Tujaa.NOM Bat.NOM train-PST that med-sen. know-PST ‘Tujaa knew that Bat trained a dog well.’
[LONG SCR]
Since Mongolian behaves exactly like Tamil, in that it cannot move PNI-ed arguments or VPs, we will run exactly the same analysis. The ban on short scrambling is illustrated with the steps in (56). Like Tamil, Mongolian is language which does not equip the T head with a [•V•]-feature, neither alone nor in combination with another feature. The numeration, as shown in (56), will always lead to a balanced phase since it contains no movement-inducing feature. Since no EF is inserted, PNI-DP stays inside VP and becomes inaccessible for further operations. T enters the derivation in (56d) and assigns nominative case to the subject. Since PNI-DP is trapped inside VP and T does not come with [•V•], it cannot be attracted to spec,TP, shown in (56e).
7.2 MONGOLIAN
247
(56) No short or intermediate scrambling in Mongolian a. Numeration: {C, T, D,...} → phase balanced b. No EF insertion in ditransitive contexts: v′ P NI -DP 1
v′
N
v
ApplP
D CASE : ACC
V
A PPL
Appl ′
DP N
* CASE : ACC *
VP
D CASE : DAT
Appl
D
V
V
P NI -DP 1 N
D
D
V
D
v
* CASE : DAT *
[ EF ]
D
CASE : ACC
V
c. No EF insertion in transitive contexts: vP P NI -DP 1
v′
N
v
VP
D * CASE : ACC *
V
P NI -DP 1
V
V
N
D
D
V
[ EF ]
* CASE : ACC *
D
v
CASE : ACC
V
d. Case assignment: T′ NOM
vP DP
v′
N
CASE : NOM
D
V
P NI -DP
V
D
V
N
* CASE : ACC *
V
D
v
* CASE : NOM *
v
VP
D
T
* CASE : ACC *
D
v
T
248
PNI-PROPERT Y III: MOVEMENT PATTERNS e. No trigger for PNI-DP-movement: TP T′
P NI -DP 1 N
T
vP
D CASE : ACC
V
v
DP
v′
N
* CASE : NOM *
CASE : NOM
V
V
P NI -DP 1
D
T
v
VP
D N
D
D
V
* CASE : ACC *
D
v
* CASE : ACC *
phase domain
V
Long scrambling of PNI-ed objects is impossible in Mongolian since C does not introduce a [•V•]-feature. Similar to intermediate scrambling, the phase is balanced since the numeration does not contain a movement-inducing feature. No EF is inserted, thus PNI-DP does not move to the phase edge in (57b). The PNI-ed argument cannot move to spec,CP in (57c) since C cannot come with [•V•]. Moreover, PNI-DP is inaccessible upon Merge with C. For the same reasons, PNI-DP cannot be attracted by matrix T (not shown here). (57) No long scrambling in Mongolian a. Numeration: {C, T, v, V[...V...] , D, C, T, D,...} → phase balanced b. No EF insertion:
vP
P NI -DP 1
v′
N
v
VP
D * CASE : ACC *
V
V
P NI -DP 1
V
N
D
D
V
[ EF ]
* CASE : ACC *
D
v
* CASE : ACC *
V
c. No trigger for PNI-DP-movement: CP C′
P NI -DP 1 N
C
TP
D CASE : ACC
T
vP
V
T
DP
D
v
VP
D CASE : NOM
...
v′
N
P NI -DP 1
V
V
N
D
D
V
* CASE : ACC *
V
* CASE : ACC *
D
v
phase domain
C
7.3 TURKISH
249
This section demonstrated the parallelism between PNI-movement and VPmovement restrictions in Mongolian. Since both types of constituents are triggered by [•V•] which is, by assumption, missing on functional heads such as T and C, they remain immobile throughout the derivations.
7.3 Turkish While Turkish displays short, intermediate, and long scrambling properties (Akan 2009; İşsever 2007, 2008; Kornfilt 1997; Öztürk 2005; Termucu 2005; JiménezFernández and İşsever 2012), caseless bare nouns in Turkish are traditionally assumed to be restricted to occur in their base positions (Aygen 2007a; Dede 1986; Enc˛ 1991; Erguvanlı 1984; Kornfilt 2003a). Recent studies, however, provide data which question this generalization. Kornfilt (2003a: 152) mentions in a footnote that caseless bare nouns can be non-adjacent to V in colloquial speech. There is by now ample evidence that caseless bare nouns can scramble and extrapose (Gračanin-Yüksek and İşsever 2011; İşsever 2003; Kamali 2015; Kan 2010; Öztürk 2009; Uygun 2006), for example as shown in (58) for intermediate scrambling. (58) Uygun (2006: 5)
Turkish
A: Would you like to have some cake? B: Pasta1 ben daha du¨n aks¸am ⟨DP1 ⟩ ye-di-m. cake I already last.night eat-PST-1SG ‘I already ate cake last night.’ Below, I extend the data set and provide movement patterns for ditransitive structures, realizing short (59b) and intermediate scrambling (59c) of PNI-ed bare objects. As expected, case marked objects are also possible in these configurations. (59) Optional case marking and clause-internal scrambling in Turkish ¨ g˘retmen o¨g˘renci-ler-e o¨dev(-i) a. O ver-di-∅. teacher.NOM student-PL-DAT homework-ACC give-PFV-3 ‘The teacher did homework-giving to the students.’
[BASELINE]
¨ g˘retmen o¨dev(-i)1 b. O o¨g˘renci-ler-e ⟨DP1 ⟩ ver-di-∅. teacher.NOM homework-ACC student-PL-DAT give-PFV-3 ‘A s for homework the teacher gave it to the students.’ [SHORT SCR] ¨ dev(-i)1 c. O o¨g˘retmen o¨g˘renci-ler-e ⟨DP1 ⟩ ver-di-∅. homework-ACC teacher.NOM student-PL-DAT give-PFV-3 ‘A s for homework the teacher gave it to the students.’ [INTERMED SCR] For long scrambling of caseless bare objects, we find different judgements in the literature. Gračanin-Yüksek and İşsever (2011) claim that long scrambling is blocked for caseless nouns, based on the data in (60). Jo and Palaz (2018, 2019), on the other hand, provide data where long scrambling is clearly acceptable; see (61).
250
PNI-PROPERT Y III: MOVEMENT PATTERNS
Note that the authors make use of different matrix verbs. My informants agree with the judgements in (61) and (60a), but they disagree on the acceptability of (60b). It seems that Turkish long scrambling is sensitive to whether the matrix verb qualifies as a bridging verb. Verbs of saying often allow long distance extraction more readily than e.g. low frequency factives like regret, while some languages also block highly frequent factives like know (Hawkins 1999; Kluender 1992). The acceptability of long scrambling with a typical bridging verb like so¨ye (=say) in (62) confirms the explanation for the different judgements in (60) and (61). Crosslinguistic observations concerning long distance extraction ascribe so¨ye and du¨¸su¨n a better chance for enabling long scrambling since they qualify as prototypical bridging verbs. Hence, I conclude that the unacceptability of the structures in (60) is not tied to pseudo-incorporation. (60) Gračanin-Yüksek and İşsever (2011: 10)
Turkish
a. *Kitap1 Ali [Ays¸e-nin ⟨DP1 ⟩ oku-dug˘-un]-u book Ali.NOM Ayse-GEN read-NMLZ-POSS.3SG-ACC biliyor. know.PRES.PROG.3SG ‘A li knows that Ayse does book-reading.’
[LONG SCR]
[Ays¸e-nin ⟨DP1 ⟩ oku-mas-ın]-ı b. *Kitap1 Ali book Ali.NOM Ayse-GEN read-NMLZ-POSS.3SG-ACC istedi. want.PST.3SG ‘A li wants Ayse to do book-reading.’
(61)
[LONG SCR]
Jo and Palaz (2019: 23) Turkish Kitap1 ben [Ali-nin ⟨DP1 ⟩ oku-dug˘-un]-u du¨¸su¨n-mu¨-yor-um. book I Ali-GEN read-NMLZ-POSS.3SG-ACC think-NEG-PRS-1SG ‘I don’t think that Ali does book-reading.’ [LONG SCR]
(62) Long scrambling of PNI-ed bare objects in Turkish a.
Ays¸e [Ali-nin kitap oku-dug˘-un]-u so¨yle-m-iyor-∅. Ays¸e.NOM Ali-GEN book read-NMLZ-POSS.3SG-ACC say-NEG-IPFV-3 ‘Ays¸e doesn’t say that Ali does book-reading.’ [BASELINE]
b.
Kitap1 Ays¸e [Ali-nin ⟨DP1 ⟩ oku-dug˘-un]-u book Ays¸e.NOM Ali-GEN read-NMLZ-POSS.3SG-ACC so¨yle-m-iyor-∅. say-NEG-IPFV-3 ‘Ays¸e doesn’t say that Ali does book-reading.’
[LONG SCR]
7.3 TURKISH
251
We can observe the same movement flexibility for PPs in Turkish, shown here with the postposition ic¸in ‘for’ (Göksel and Kerslake 2005: 215). (63) PP scrambling in Turkish a.
¨ g˘retmen [o¨g˘renci-ler ic¸in] not bırak-tı-∅. O teacher.NOM student-PL for note leave-PFV-3 ‘The teacher left a note for the students.’
b.
[o¨g˘renci-ler ic¸in]2 o¨g˘retmen ⟨PP2 ⟩ not bırak-tı-∅. student-PL for teacher.NOM note leave-PFV-3 ‘A s for the students, the teacher left a note.’ [INTERMED SCR]
c.
[o¨g˘renci-ler student-PL
ic¸in]2 for
ben I
[o¨g˘retmen teacher.NOM
⟨PP2 ⟩
[BASELINE]
not note
bırak-tıg˘-ın]-ı du¨¸su¨n-mu¨-yor-um. leave-NMLZ-POSS.3SG-ACC think-NEG-IPFV-1SG ‘A s for the students, I don’t think that the teacher left a note.’ [LONG SCR] Excluded from the discussion are indefinites. Kamali (2015) argues that bir NPs do not undergo leftward scrambling since this type of movement is associated with topic readings (Erguvanlı 1984; İşsever 2003; Vallduví and Engdahl 1996) and indefinites do not make good topics. She creates contexts which enforce a topic reading and observes that fronting an indefinite is infelicitous, certainly if the object lacks case marking but also if it is case marked, compare (64) and (65). A caseless bare noun, however, is perfectively acceptable; see (66). (64) Kamali (2015: 116) A:
Bir aslanın boyu ne kadardır acaba?
‘I wonder how tall a lion is.’ B: ??Bir aslan1 ben ⟨DP1 ⟩ go¨r-du¨-m. 2 metre var. one lion I see-PST-1SG 2 meter exist ‘A lion, I’ve seen one. It’s about 2 meters.’ (65) Kamali (2015: 118) A: Hayvanat bahc¸esine yeni gelen hayvanların boyu ne kadardır acaba? ‘I wonder how tall the new zoo animals are.’ B: ?Bir aslan-ı1 ben ⟨DP1 ⟩ go¨r-du¨-m. 2 metre var. one lion-ACC I see-PST-1SG 2 meter exist ‘Of the lions, I’ve seen one. It’s about 2 meters.’
252
PNI-PROPERT Y III: MOVEMENT PATTERNS
(66) Kamali (2015: 116) A: Aslanların n boyu ne kadardır acaba? ‘I wonder how tall lions are.’ B: Aslan1 ben ⟨DP1 ⟩ go¨r-du¨-m. 2 metre var. lion I see-PST-1SG 2 meter exist ‘Lions, I’ve seen some. They’re about 2 meters.’ It seems that if movement is clearly connected to an information structural function, it fails to serve as a straightforward PNI diagnostic. Interactions between topicality and PNI are expected, as both phenomena relate to definiteness and specificity: Whereas PNI-ed arguments are often indefinite and non-specific, topics are definite and specific. Hence, if a movement operation requires a property that is incompatible with PNI to begin with, unacceptability will not tell us anything about the movement pattern of PNI-ed arguments. A more detailed explanation for the variation must be left to future research.¹² We now turn to VP-movement. Following the argumentation of the previous chapters, we expect VP-movement to be freely available. This prediction seems to be borne out. The paradigm given in (67) shows that post-verbal constituents ¹² The picture is equally unclear with weak quantifiers and numerals. While there is speaker variation with respect to the acceptance of short scrambling a caseless numeral phrase, intermediate scrambling seems to be only marginally acceptable. The weak quantifier birkac¸ requires the insertion of focus particle dA in order to undergo short scrambling, whereas this strategy does not seem to be available for intermediate scrambling. (i) Short and intermediate scrambling of numerals in Turkish ¨ g˘retmen a. O o¨g˘renci-ler-e u¨c¸ o¨dev ver-di-∅. teacher.NOM student-PL-DAT three homework give-PFV-3 ‘The teacher gave three homeworks to the students.’
[BASELINE]
¨ g˘retmen b. %O [u¨c¸ o¨dev]1 o¨g˘renci-ler-e ⟨DP1 ⟩ ver-di-∅. teacher.NOM three homework student-PL-DAT give-PFV-3 ‘A s for three homeworks, the teacher gave it to the students.’
[SHORT]
c. ?[U ¨ c¸ o¨dev]1 o¨g˘retmen o¨g˘renci-ler-e ⟨DP1 ⟩ ver-di-∅. three homework teacher.NOM student-PL-DAT give-PFV-3 ‘A s for three homeworks, the teacher gave it to the students.’
[INTER]
(ii) Short and intermediate scrambling of birkac¸ NP in Turkish ¨ g˘retmen a. O o¨g˘renci-ler-e birkac¸ o¨dev ver-di-∅. teacher.NOM student-PL-DAT some homework give-PFV-3 ‘The teacher gave some homework to the students.’ b.
c.
¨ g˘retmen O [birkac¸ o¨dev]1 *(de) o¨g˘renci-ler-e ⟨DP1 ⟩ ver-di-∅. teacher.NOM some homework FOC student-PL-DAT give-PFV-3 ‘A s for some homework, the teacher gave it to the students.’ *[Birkac¸ o¨dev]1 o¨g˘retmen o¨g˘renci-ler-e ⟨DP1 ⟩ ver-di-∅. some homework teacher.NOM student-PL-DAT give-PFV-3 ‘A s for some homework, the teacher gave it to the students.’
[BASELINE]
[SHORT]
[INTER]
7.3 TURKISH
253
are acceptable in Turkish. Note, however, that these structures might as well be created via rightward movement of the post-verbal constituent since extraposition has been argued for independently in Turkish (Kornfilt 2005; Kural 1997; Termucu 2005). Again, verbs are inflected for tense/aspect, suggestive of V-to-T movement. For long scrambling in (68), the verbs form a phonological word with morphology signalling a nominalizer and case of the clausal complement. (67) Short and intermediate scrambling of VPs? ¨ g˘retmen o¨g˘renci-ler-e o¨dev a. O ver-di-∅. teacher student-PL-DAT homework give-PFV-3 ‘The teacher did homework-giving to the students.’
Turkish [BASELINE]
¨ g˘retmen [o¨dev b. O ver-di-∅]1 o¨g˘renci-ler-e ⟨VP1 ⟩. teacher homework give-PFV-3 student-PL-DAT ‘What the teacher did to the students was homework-giving.’
[SHORT]
¨ dev c. [O ver-di-∅]1 o¨g˘retmen o¨g˘renci-ler-e ⟨VP1 ⟩. homework give-PFV-3 teacher student-PL-DAT ‘What the teacher did to the students was homework-giving.’
[INTER]
(68) Long scrambling of VPs?
Turkish
a. [Kitap oku-dug˘-un]-u1 ben [Ali-nin ⟨VP1 ⟩] book read-NMLZ-POSS.3SG-ACC 1SG.NOM Ali-GEN du¨¸su¨n-mu¨-yor-um. think-NEG-IPFV-1SG ‘I don’t think that Ali does book-reading.’
[LONG SCR]
b. [Kitap oku-dug˘-u]-nu1 ben [Ali-nin ⟨VP1 ⟩] book read-NMLZ-POSS.3SG-ACC 1SG.NOM Ali-GEN so¨yle-m-iyor-um. say-NEG-IPFV-1SG ‘I don’t say that Ali does book-reading.’
[LONG SCR]
Since the VP-status of the moved phrases in (67) and (68) is not entirely conclusive,¹³ we test for remnant VP-movement, that is movement of the direct object together with a low manner adverb which is assumed to be adjoined to VP. The examples in (69) and (70) are in line with the previous observations, i.e. VPs are allowed to undergo long scrambling in Turkish.
¹³ Although see Gračanin-Yüksek and İşsever (2011) who take examples like (68) as evidence for long scrambling of VPs.
254
PNI-PROPERT Y III: MOVEMENT PATTERNS
(69) Remnant VP-movement Turkish [VP Hızlıca kitab-ı]1 ben [Ali-nin ⟨VP1 ⟩ oku-dug˘-u]-nu quickly book-ACC 1SG.NOM Ali-GEN read-NMLZ-POSS.3SG-ACC du¨¸su¨n-m-u¨yor-um. think-NEG-IPFV-1SG ‘I don’t think that Ali read(s) the book rapidly.’
[LONG SCR]
(70) Context: There was a bad anonymous review in the papers which influenced the book sale. Ali is one of three potential reviewers. [VP Acımasızca kitab-ı]1 ben [Ali-nin ⟨VP1 ⟩ ruthlessly book-ACC 1SG.NOM Ali-GEN eles¸tir-dig˘-i]-ni du¨¸su¨n-m-u¨yor-um. criticize-NMLZ-POSS.3SG-ACC think-NEG-IPFV-1SG ‘I don’t think that Ali criticized the book ruthlessly.’
[LONG SCR]
An additional argument for VP-movement comes from discontinuous possessor phrases. Turkish can extract possessors to the left and to the right periphery, shown in (71); see also Bošković and Şener (2014). (71) Possessor displacement in Turkish a. Ali-nin ben [anne-si-nin du¨n kitap Ali-GEN 1SG.NOM mother-POSS.3SG-GEN yesterday book oku-dug˘-u]-nu duy-du-m. read-NMLZ-POSS.3SG-ACC hear-PFV-3SG ‘I have heard that Ali’s mother read books yesterday.’ b. Anne-si du¨n kitap oku-du-∅ Ali-nin. mother-POSS.3SG yesterday book read-PFV-3 Ali-GEN ‘A li’s mother read books yesterday.’ Displacement of the possessum, however, seems to be restricted, in that the possessum can only show up right peripherally; see (72). (72) Possessum displacement in Turkish a.
Ali-nin du¨n kitap oku-du-∅ anne-si. Ali-GEN yesterday book read-PFV-3 mother-POSS.3SG ‘A li’s mother read books yesterday.’
7.3 TURKISH
255
b. * Anne-si-nin ben [Ali-nin du¨n kitap mother-POSS.3SG-GEN 1SG.NOM Ali-GEN yesterday book oku-dug˘-u]-nu duy-du-m. read-NMLZ-POSS.3SG-ACC hear-PFV-3SG ‘I have heard that Ali’s mother read books yesterday.’ Let us assume for now that the possessor is merged in the specifier position of a nominal projection that takes the possessum as a complement. We predict that the possessum can only move as a remnant DP, after the possessor has moved out. The structures in (72), thus, require two movement operations, one creating the remnant constituent (possessor extraction) and one which constitutes the actual remnant movement operation. If (72a) were derived via (i) rightward movement of the entire possessor phrase and (ii) subsequent subextraction of the possessor, sketched in (73), a freezing effect should be triggered, contrary to fact. (73) [CP Ali-nin1 ... [ TP [TP ⟨DP2 ⟩ kitap okudu]
[DP ⟨Ali-nin1 ⟩ anne-si]2 ]]
We would, however, make the right prediction for (72b) since possessor extraction targets a position not accessible any more, after the possessor phrase has undergone long scrambling. This derivation is shown in (74). (74) [TP-2 [DP ⟨Ali-nin 1 ⟩ anne-si]2 ... [ vP-2 ben [ TP-1 Ali-nin1 ... ⟨DP 2 ⟩ ...]]] A promising way to derive the acceptability of (72a), is to reverse the order of operations, i.e. (i) leftward extraction of the possessor and (ii) rightward remnant movement, possibly as adjunction to CP. This derivation is shown in (75). (75) a. [ TP Ali-nin1 [vP [DP ⟨Ali-nin1 ⟩ anne-si] [VP kitap okudu]]] b. [CP [TP Ali-nin1 [vP ⟨DP 2 ⟩ [VP kitap okudu]]] ... [DP ⟨Ali-nin 1 ⟩ anne-si ]2 ] The explanation for (72a), however, does not extend to (72b), which would be minimally different from (75) in that remnant movement occurs to the left (see (76)), leaving us with no explanation for its unacceptability. (76) a. [ TP-1 Ali-nin1 ... [DP ⟨Ali-nin1 ⟩ anne-sin] ...] b. [ TP-2 [DP ⟨Ali-nin1 ⟩ anne-si]2 ... [ vP-2 ben [ TP-1 Ali-nin1 ... ⟨DP 2 ⟩ ...]]]
256
PNI-PROPERT Y III: MOVEMENT PATTERNS
We have, thus, arrived at an impasse. No possible order of operations is able to derive the contrast in (72). There is one option, however, we have not considered so far. I propose that there is a general restriction against remnant possessor phrase movement in Turkish, which derives the ungrammaticality of (72b). The displacement of the possessum in (72a) is a consequence of VP-movement. The derivation is sketched in (77). Hence, in order to derive the acceptability of (72a), we crucially need VP-movement to be an option in Turkish. (77)
[TP Ali-nin1 [VP kitap okudu]3 [vP [DP ⟨Ali-nin 1 ⟩ anne-si ] ⟨VP3 ⟩]] We now turn to the analysis of PNI-movement.¹⁴ This section has demonstrated so far that Turkish permits PNI-ed arguments to undergo short, intermediate, and long scrambling, as well as the general need for VP-scrambling. Short scrambling is triggered by T[...•V• ≻ •D•...] , following the assumptions in Table 7.2. If T comes with such a feature combination, as it is shown in (78b), EF insertion follows since [•V•] does not have a matching feature in the numeration or workspace. The order of movement features on T derives the surface order we find with short scrambling; see (78d) and (78e). The features responsible for PNI-DP movement are always boxed in all following derivations. (78) Short scrambling in Turkish a. Numeration-I: {C, T, D,...} → phase balanced (not shown) b. Numeration-II: {C, T[...•V• ≻ •D•...] , D[...D...] ,...} → phase not balanced c. EF insertion:
v′
P NI -DP 1
v′
N
V
v
ApplP
D CASE : ACC
APPL
Appl ′
DP N
VP
D CASE : DAT
D
P NI -DP 1
* CASE : ACC *
Appl V
N
D
D
V
V
D
v
* CASE : DAT *
D
[ EF ]
CASE : ACC
V
¹⁴ It is worth pointing out that the analysis pursued in this book is in a sense similar to the proposal by Gračanin-Yüksek and İşsever (2011) who account for the movement properties of caseless bare objects in Turkish by assuming that they are generally immobile and can only move via remnant VP-movement.
7.3 TURKISH
257
T′
d.
T′
P NI -DP 1 N
V
T
vP
D CASE : ACC
v
DP
v′
N
* CASE : NOM *
P NI -DP 1
D
N
CASE : NOM
D
V
v′ v
VP
D * CASE : ACC * V
T
V
V
⟨DP1 ⟩
D
* CASE : ACC *
D
EF
V
D
v
e.
TP T′
DP N
D CASE : NOM
D
T′
P NI -DP 1 N
V
T ...
vP
D CASE : ACC
DP
v′
N
D
V D
v′
⟨DP1 ⟩
CASE : NOM
D ⟨DP1 ⟩
T v
VP
V
V
D
V
* CASE : ACC * EF D
v
If T comes with [•V•], intermediate scrambling is triggered. This is the case with the numeration in (79b) which contains a movement-inducing feature that does not have a matching feature potentially available, neither in the numeration nor on the phase head v or its edge. Thus, EF is inserted, so that the PNI-DP moves to the edge of the phase. T enters the derivation in (79d), assigning case to the subject. The next feature on the T’s stack is [•V•], triggering PNI-DP movement to spec,TP, shown in (79e). (79) Intermediate scrambling in Turkish a. Numeration-I: {C, T, D,...} → phase balanced (not shown) b. Numeration-II: {C, T[...•V•...] , D,...} → phase not balanced
258
PNI-PROPERT Y III: MOVEMENT PATTERNS c. EF insertion:
v′
P NI -DP 1
v′
N
v
VP
D * CASE : ACC * V
V
V
P NI -DP 1 N
D
D
V
[ EF ]
* CASE : ACC * D
v 1
CASE : ACC
2
V T′
d. vP DP
v
v′
N
* CASE : NOM *
P NI -DP 1
D
V
v′
N
CASE : NOM
D
T
NOM
T v
VP
D * CASE : ACC * V
⟨DP1 ⟩
V
V
* CASE : ACC *
D
EF
V
D
v TP
e.
T′
P NI -DP 1 N
V
T
vP
D CASE : ACC
DP N
D CASE : NOM
D
v
v′
* CASE : NOM *
P NI -DP 1
⟨DP1 ⟩
T v
VP
D * CASE : ACC * V
V
v′
N
V
V
D
V
* CASE : ACC * EF D
v
Let us now turn to long scrambling. As always, a numeration without a movementinducing feature will create balanced phases throughout; see N-I in (80a). Interestingly, if matrix T is equiped with [•V•], matrix VP will provide the matching [V ]-feature, thereby creating a balanced phase; see N-II in (80b). Hence, an additional scrambling feature is necessary, in order to derive long scrambling of PNI-ed arguments in Turkish. I propose that C can come with [•V•] in Turkish; see N-III in (80c). This feature constellation will create an unbalanced phase, that necessitates EF insertion (80d). Eventually, [•V•] on C will trigger PNI-DP-movement to the phase edge of the embedded CP (80e). At the point
7.3 TURKISH
259
when matrix v is merged we check for phase balance with the current numeration. As (80f) shows, the vP phase is not balanced which leads to EF insertion in (80g). Finally, the PNI-ed object moves to spec,TP to check off the movement-inducing feature on matrix T (80h).¹⁵ (80) Long scrambling in Turkish a. N-I: {C, T, D, V[...V...] , C, T, D, ...} → phase balanced (not shown) b. N-II: {C, T[...•V•...] , D, V[...V...] , C, T, D, ...} → phase balanced (not shown) c. N-III: {C, T[...•V•...] , D, V[...V...] , C[...•V•...] , T, D, ...} → phase not balanced d. EF insertion: v′ P NI -DP 1
v′
N
v
VP
D * CASE : ACC *
V V
V
V
P NI -DP 1 N
D
D
V
[ EF ]
* CASE : ACC * D
v 1
CASE : ACC
V V
2
e. Movement to phase edge of embedded CP: CP C′
P NI -DP 1 N
V V
C
TP
D CASE : ACC
T
vP DP
v′
N
D CASE : NOM
D
P NI -DP 1 N
D CASE : ACC
V V
T
V
...
C
v′ v
VP ⟨DP1 ⟩
V
...
...
f. N-III’: {C, T[...•V•...] , D, ...} → phase not balanced ¹⁵ Note that the PNI-DP comes with two categorial [V ]-features. As was pointed out in Section 3.2, the number of categorial features is in principle free since they do not have be deactivated by the end of the derivation like probe features.
260
PNI-PROPERT Y III: MOVEMENT PATTERNS g. EF insertion: v′
v′
P NI -DP 1 N
D V V
v ...
CP
CASE : ACC
D
v
C′
P NI -DP 1 N
D C
TP
CASE : ACC
V V
T
[ EF ]
V
C
T
vP
... DP
v′
... v′
⟨DP1 ⟩
v
VP
... V
⟨DP1 ⟩
phase domain
...
h. Movement to the final landing site: TP
T′
P NI -DP 1 N
D V V
T ...
v′
CASE : ACC
V
DP
T
v′
... P NI -DP 1
v′
N
D CASE : ACC
V V
CP ...
v ... EF D
v
Since Turkish shows PNI-effects also for agents, we expect the same flexibility for PNI-ed subjects. This prediction, however, is not borne out. The contrast in (81) demonstrates that PNI-ed agents cannot move across a clause boundary.
7.4 KORE AN
261
(81) No long scrambling for PNI-ed subjects in Turkish a. [Ali-yi arı(-nın) sok-tug˘-u-nu] duy-du-m. Ali-ACC bee-GEN sting-NMLZ-POSS.3SG-ACC hear-PFV-1SG ‘I heard that a bee stung Ali.’ [BASELINE] b. Arı*(-nın)1 ben [Ali-yi ⟨DP1 ⟩ sok-tug˘-u-nu] bee 1SG.NOM Ali-ACC sting-NMLZ-POSS.3SG-ACC duy-du-m. hear-PFV-1SG ‘I heard that Ali got bee-stung.’
[LONG SCR]
Moreover manner adverbs cannot intervene between a PNI-ed subject and the embedded verb, as (82) exemplifies. A similar observation is made by Kornfilt (2008) with respect to locative adjuncts. (82) No adverbs between PNI-ed subjects and the embedded verb a. Ko¨y-u¨ haydut*(-un) acımasızca bas-tıg˘-ı-nı village-ACC bandit-GEN ruthlessly raid-NMLZ-POSS.3SG-ACC duy-du-m. hear-PFV-1SG ‘I heard the bandit raided the village ruthlessly.’ b. Ko¨y-u¨ bir haydut*(-un) acımasızca bas-tıg˘-ı-nı village-ACC a bandit-GEN ruthlessly raid-NMLZ-POSS.3SG-ACC duy-du-m. hear-PFV-1SG ‘I heard a bandit raided the village ruthlessly.’ There seems to be an additional restriction, present with subject PNI, that I will not be able to investigate further in this book, but see Jo and Palaz (2019) for a proposal.
7.4 Korean Korean displays verb-final base order and case marked arguments can undergo short, intermediate, and long scrambling (Cho 1994; Jackson 2008; Ko 2007, 2014; Levin 2016; Müller 1995; Yoon 1991). The distribution of case marked nouns can be observed in (83). In Section 4.3.4, we identified four types of arguments which can show the effects of pseudo-incorporation: bare nouns, proper names, 1st /2nd person pronouns, and weak definites co-occurring with ku. All four types can be scrambled within clause boundaries, i.e. short and intermediate, but not long. I show the movement patterns for each noun type in (83)–(86).
262
PNI-PROPERT Y III: MOVEMENT PATTERNS
(83) Scrambling of bare nouns in Korean a. Yusu-ka chayk(-ul)1 Suzi-eykey ⟨DP1 ⟩ cwu-ess-e. Yusu-NOM book-ACC Suzi-DAT give-PST-INT ‘Yusu gave Suzi books.’
[SHORT SCR]
b. Chayk(-ul)1 Yusu-ka Suzi-eykey ⟨DP1 ⟩ cwu-ess-e. book Yusu-NOM Suzi-DAT give-PST-INT ‘Yusu gave Suzi books.’ [INTERMED SCR] c. Chayk* (-ul)1 Suzi-nun [Minho-ka ⟨DP1 ⟩ ilk-nun-ta-ko] book-ACC Suzi-TOP Minho-NOM read-PRS-DECL-COMP saynkakha-y. think-INT ‘Suzi thinks that Minho is book-reading.’
[LONG SCR]
(84) Scrambling of proper names in Korean a. Suzi-ka Hannah(-lul)1 Minsu-eykey ⟨DP1 ⟩ sokay-hay-ess-e. Suzi-NOM Hannah-ACC Minsu-DAT introduce-v-PST-INT ‘Suzi introduced Hannah to Minsu.’ [SHORT SCR] b. Hannah(-lul)1 Suzi-ka Minsu-eykey ⟨DP1 ⟩ sokay-hay-ess-e. Hannah-ACC Suzi-NOM Minsu-DAT introduce-v-PST-INT ‘Suzi introduced Hannah to Minsu.’ [INTERMED SCR] c. Hannah* (-lul)1 na-nun [Suzi-ka Minsu-eykey ⟨DP1 ⟩ Hannah-ACC 1SG-TOP Suzi-NOM Minsu-DAT sokay-ss-ta-ko] saynkakha-y. introduce-v-PST-DECL-COMP think-INT ‘I think Suzi introduced Hannah to Minsu.’
[LONG SCR]
(85) Scrambling of 1st /2nd person pronouns in Korean a. Suzi-ka na(-lul)1 /ne(-lul)1 Minsu-eykey ⟨DP1 ⟩ sokay-hay-ess-e. Suzi-NOM 1SG-ACC/2SG-ACC Minsu-DAT introduce-v-PST-INT ‘Suzi introduced me/you to Minsu.’ [SHORT SCR] b. Na(-lul)1 /ne(-lul)1 Suzi-ka Minsu-eykey ⟨DP1 ⟩ sokay-hay-ess-e. 1SG-ACC/2SG-ACC Suzi-NOM Minsu-DAT introduce-v-PST-INT ‘Suzi introduced me/you to Minsu.’ [INTERMED SCR] c. Na*(-lul)1 /ne*(-lul)1 emma-nun [Suzi-ka Minsu-eykey ⟨DP1 ⟩ 1SG-ACC/2SG-ACC mom-TOP Suzi-NOM Minsu-DAT sokay-ss-ta-ko] saynkakha-y. introduce-v-PST-DECL-COMP think-INT ‘Mom thinks Suzi introduced me/you to Minsu.’
[LONG SCR]
7.4 KORE AN
263
(86) Scrambling of ku + weak definites in Korean a. Suzi-ka [ku yewang(-ul)]1 saylowun yolisa-eykey ⟨DP1 ⟩ Suzi-NOM DEM queen-ACC new cook-DAT sokay-hay-ess-e. introduce-v-PST-INT ‘Suzi introduced the queen to a new cook.’
[SHORT SCR]
b. [Ku yewang(-ul)]1 Suzi-ka saylowun yolisa-eykey ⟨DP1 ⟩ DEM queen-ACC Suzi-NOM new cook-DAT sokay-hay-ess-e. introduce-v-PST-INT ‘Suzi introduced the queen to a new cook.’
[INTERMED SCR]
c. [Ku yewang* (-ul)]1 na-nun [Suzi-ka saylowun yolisa-eykey DEM queen-ACC 1SG-TOP Suzi-NOM new cook-DAT ⟨DP1 ⟩ sokay-ss-ta-ko] saynkakha-y. introduce-v-PST-DECL-COMP think-INT ‘I thinks Suzi introduced the queen to a new cook.’
[LONG SCR]
VP-movement displays patterns exactly parallel to the movement restrictions of PNI-ed arguments, presented in (87). (87) Scrambling of VPs in Korean? a.
Suzi-ka [Hannah-lul sokay-hay-ess-e]1 Minsu-eykey ⟨VP1 ⟩. Suzi-NOM Hannah-ACC introduce-v-PST-INT Minsu-DAT ‘Suzi introduced Hannah to Minsu.’ [SHORT SCR]
b.
[Hannah-lul sokay-hay-ess-e]1 Suzi-ka Minsu-eykey ⟨VP1 ⟩. Hannah-ACC introduce-v-PST-INT Suzi-NOM Minsu-DAT ‘Suzi introduced Hannah to Minsu.’ [INTERMED SCR]
c. *[Chayk-ul ilk-nun-ta-ko]1 Suzi-nun [Minho-ka ⟨VP1 ⟩] book-ACC read-PRS-DECL-COMP Suzi-TOP Minho-NOM saynkakha-y. think-INT ‘Suzi thinks that Minho is book-reading.’
[LONG SCR]
Again, we cannot safely conclude that the moved phrase in (87) does in fact constitute a VP. Hence, we test for remnant VP-movement by inserting a manner adverb and displacing it together with the direct object. As (88) and (89) show, the patterns are consistent with VP-movement patterns.
264
PNI-PROPERT Y III: MOVEMENT PATTERNS
(88) Scrambling of remnant VPs in Korean a.
b.
[Ppalli chayk-ul]1 Suzi-ka ⟨VP1 ⟩ ilk-ess-e. quickly book-ACC Suzi-NOM read-PST-INT ‘Suzi was quickly book-reading.’
[SHORT SCR]
Suzi-ka [ppalli Hannah-lul]1 Minsu-eykey ⟨VP1 ⟩ Suzi-NOM quickly Hannah-ACC Minsu-DAT sokay-hay-ess-e. introduce-v-PST-INT ‘Suzi quickly introduced Hannah to Minsu.’
c. *[Ppalli quickly
chayk-ul]1 book-ACC
Minsu-nun Minsu-TOP
[Suzi-ka Suzi-NOM
ilk-ess-ta-ko] saynkakha-y. read-PST-DECL-COMP thinks-INT ‘Minsu thinks that Suzi was quickly book-reading.’
[INTERMED SCR] ⟨VP1 ⟩
[LONG SCR]
(89) Scrambling of remnant VPs in Korean a.
[Cosimsulepkey chayk-ul]1 Suzi-ka ⟨VP1 ⟩ ilk-ess-e. carefully book-ACC Suzi-NOM read-PST-INT ‘Suzi was carefully book-reading.’ [SHORT SCR]
b.
Suzi-ka [cosimsulepkey Hannah-lul]1 Minsu-eykey ⟨VP1 ⟩ Suzi-NOM carefully Hannah-ACC Minsu-DAT sokay-hay-ess-e. introduce-v-PST-INT ‘Suzi carefully introduced Hannah to Minsu.’
[INTERMED SCR]
c. *[Cosimsulepkey chayk-ul]1 Minsu-nun [Suzi-ka ⟨VP1 ⟩ carefully book-ACC Minsu-TOP Suzi-NOM ilk-ess-ta-ko] saynkakha-y. read-PST-DECL-COMP thinks-INT ‘Minsu thinks that Suzi was carefully book-reading.’
[LONG SCR]
We could imagine an alternative explanation for the unacceptable long scrambling examples in (88) and (89), which is based on the assumption that the adverb and the object each scramble separately and there is an independent restriction on recursive scrambling. Recursive scrambling, however, is licensed in Korean; see (90). Hence, we conclude that PNI-ed arguments as well as VPs cannot undergo long scrambling, as predicted by the current theory. Scrambling within clauses, however, is permitted.
7.4 KORE AN
265
(90) Multiple scrambling in Korean Minsu-eykey1 chayk-ul2 na-nun [Suzi-ka ⟨DP1 ⟩ ⟨DP2 ⟩ Minsu-DAT book-ACC 1SG-TOP Suzi-NOM sokayhay-ess-ta-ko] saynkakha-y. introduce-PST-DECL-COMP think-INT ‘I think that Suzi introduced a book to Minsu.’ I now extend the analysis to the Korean movement paradigm. Short scrambling results from the presence of T[...•V• ≻ •D•...] ; recall Table 7.2. This T head triggers EF insertion since [•V•] does not have a matching feature in the numeration or workspace; see (91b). The order of movement features on T derives the surface order which is identified as short scrambling; see (91d) and (91e). (91) Short scrambling in Korean a. Numeration-I: {C, T, D,...} → phase balanced (not shown) b. Numeration-II: {C, T[...•V• ≻ •D•...] , D[...D...] ,...} → phase not balanced c. EF insertion: v′ P NI -DP 1
v′
N CASE : ACC
V
v
ApplP
D
A PPL
Appl ′
DP N
VP
D CASE : DAT
D
P NI -DP 1
* CASE : ACC * D
Appl
N
D
D
V
v
V
V
* CASE : DAT * D
[ EF ]
CASE : ACC
V T′
d.
T′
P NI -DP 1 N
V
T
vP
D CASE : ACC
DP N
D CASE : NOM
D
v
v′
* CASE : NOM *
P NI -DP 1 N
V
⟨DP1 ⟩
D
v
VP
D * CASE : ACC *
V
v′ V
V
D
V
* CASE : ACC * EF D
v
T
266
PNI-PROPERT Y III: MOVEMENT PATTERNS e.
TP T′
DP N
T′
P NI -DP 1
D
N
CASE : NOM
T ...
vP
D
D
CASE : ACC
DP
V
v′
N
V
v′
⟨DP1 ⟩
D
D
CASE : NOM
T v
VP
D
V
V
⟨DP1 ⟩
* CASE : ACC *
D
EF
V
D
v
If T comes with [•V•], intermediate scrambling is triggered. This is the case with the numeration in (92b), which contains a movement-inducing feature for which there is no matching feature available. Thus, EF is inserted, so that PNI-DP moves to the edge of the phase; see (92c). T enters the derivation in (92d), assigning case to the subject. The next feature on the T’s stack is [•V•], triggering PNI-DP movement to spec,TP, shown in (92d). (92) Intermediate scrambling in Korean a. Numeration-I: {C, T, D,...} → phase balanced (not shown) b. Numeration-II: {C, T[...•V•...] , D,...} → phase not balanced c. EF insertion:
v′
P NI -DP 1
v′
N
v
VP
D CASE : ACC
V
2
V
P NI -DP 1
V
N
D
D
V
[ EF ]
* CASE : ACC * D
v
CASE : ACC
1
V
T′
d. vP DP N
v
v′
D
P NI -DP 1
CASE : NOM
N
D
T
NOM
V
V
v′
⟨DP1 ⟩
T v
VP
D CASE : ACC
* CASE : NOM *
V
V D
V
* CASE : ACC * EF D
v
7.4 KORE AN
267
TP
e.
T′
P NI -DP 1 N
V
T
vP
D CASE : ACC
DP N
D
* CASE : NOM *
P NI -DP 1
⟨DP1 ⟩
T v
VP
D * CASE : ACC * V
V
v′
N
CASE : NOM
D
v
v′
V
V
D
V
* CASE : ACC * EF D
v
As we have seen in this section, long scrambling of PNI-ed objects is not licensed. We derive this observation by the assumption that Korean C heads do not come with [•V•]. Attempts to derive long scrambling with other feature combinations are presented in (93a) and (93b). N-I and N-II present numerations which produce balanced phases, where movement-inducing features are either not present or are matched with other features. N-III, however, creates an unbalanced phase, thereby triggering EF insertion in (93d). The [•V•]-feature on embedded T attracts PNI-DP to its specifier (93e). Crucially, when the embedded CP is checked for phase balance, the current numeration in (93f) will only have one movementinducing feature left which is matched by matrix V. Hence, EF insertion is blocked (93g) and any further operation will not be able to access PNI-DP, shown in (93h). (93) No long scrambling in Korean a. N-I: {C, T, D, V[...V...] , C, T, D, ...} → phase balanced (not shown) b. N-II: {C, T[...•V•...] , D, V[...V...] , C, T, D, ...} → phase balanced (not shown) c. N-III: {C, T[...•V•...] , D, V[...V...] , C, T[...•V•...] , D, ...} → phase not balanced
268
PNI-PROPERT Y III: MOVEMENT PATTERNS d. EF insertion: v′ P NI -DP 1
v′
N
v
VP
D * CASE : ACC * V V
V
V
P NI -DP 1 N
D
D
V
EF
* CASE : ACC * D
v 1
CASE : ACC
V V
2
TP
e.
T′
P NI -DP 1 N
T
vP
D CASE : ACC
v
DP
V V
v′
N
* CASE : NOM *
P NI -DP 1
D
N
CASE : NOM
CASE : ACC
V V
...
V
⟨DP1 ⟩
T v
VP
D
V
v′
...
f. N-III’: {C, T[...•V•...] , D, V[...V...] , ...} → phase balanced g. No EF insertion:
CP C′
P NI -DP i N
V V
C
TP
D CASE : ACC
P NI -DP i N
V V
C T ...
vP
D CASE : ACC
T
T′
DP
v′
V
N
D CASE : NOM
T
v′
⟨DP1 ⟩
v
VP ⟨DP1 ⟩
V ...
...
EF
7.4 KORE AN
269
h. PNI-DP inaccessible for further movement: TP
T′
P NI -DP 1 N
D T ...
CP
CASE : ACC
V V
V
C
TP
T
T
C T′
P NI -DP 1 N
D V V
T ...
vP
CASE : ACC
V
DP
T
v′
N
D CASE : NOM
v′
⟨DP1 ⟩
v
VP
... ⟨DP1 ⟩
phase domain
V ...
Like Turkish, Korean is a PNI-language which pseudo-incorporates subjects. Scrambling PNI-ed subjects across a temporal adverb is licensed for all four noun types; see (94)–(97). Assuming that the adverb’s adjunction site is sufficiently high enough, subjects preceding the adjunct signals a movement step. (94) Scrambling of bare nouns in subject position a. Yecenhi etten yuksangsenswu(-ka) cip-ey iss-ey. still certain athlete-NOM home-at exist-FAM ‘An athlete is still at home.’ b. [Etten yuksangsenswu(-ka)]1 yecenhi ⟨DP1 ⟩ cip-ey iss-ey. certain athlete-NOM still home-at exist-FAM ‘An athlete is still at home.’
(95) Scrambling of proper names in subject position a. Yecenhi Yumi(-ka) cip-ey iss-ey. still Yumi-NOM home-at exist-FAM ‘Yumi is still at home.’ b. Yumi(-ka)1 yecenhi ⟨DP1 ⟩ cip-ey iss-ey. Yumi-NOM still home-at exist-FAM ‘Yumi is still at home.’
270
PNI-PROPERT Y III: MOVEMENT PATTERNS
(96) Scrambling of 1st /2nd person pronouns in subject position a. Yecenhi nay(-ka)/ney(-ka) cip-ey iss-ey. still 1SG-NOM/2SG-NOM home-at exist-FAM ‘I am still at home. / You are still at home.’ b. Nay(-ka)1 /Ney(-ka)1 yecenhi ⟨DP1 ⟩ cip-ey iss-ey. 1SG-NOM/1SG-NOM still home-at exist-FAM ‘I am still at home / You are still at home.’
(97) Scrambling of ku + weak definites in subject position a. Yecenhi ku yewang(-i) kwungcen-ey iss-ney. still DEM queen-NOM palace-at exist-FAM ‘The queen is still at the palace.’ b. [Ku yewang(-i)]1 yecenhi ⟨DP1 ⟩ kwungcen-ey iss-ney. DEM queen-NOM still palace-at exist-FAM ‘The queen is still at the palace.’ Unfortunately, long scrambling cannot be tested since there is a general ban on long scrambling of embedded subjects in Korean. Intended long scrambling structures are reinterpreted as base orders, the sentence in (98) can only have the reading where Minho thinks that Suzie is reading a book. This restriction is not process-based since disambiguating with embedded honorific subjects leads to unacceptability; see (99).¹⁶
¹⁶ This curious restriction can also be observed with PP-movement and is further evidence against a processing effect. Let me first note that the investigation of PP-movement in general is limited by the fact that the status of postpositions is controversial (Choi-Jonin 2008). The reference grammar, for example, lists postpositions under case particles together with the other case markers (Sohn 1999: 213–214). One of the more likely particles argued by Choi-Jonin (2008: 140) to be a postposition is -haku ‘with’. As shown below, PPs formed with -haku can scramble across the subject but not out of a clause. (i) a.
b.
Swunhi-ka Inswu-hako sanchaykha-n-ta. Swunhi-NOM Inswu-COM go.for.a.walk-PST-DEC ‘Sunhi goes for a walk with Insoo.’ Inswu-hako1 Swunhi-ka ⟨PP1 ⟩ sanchaykha-n-ta. Inswu-COM Swunhi-NOM go.for.a.walk-PST-DEC ‘Sunhi goes for a walk with Insoo.’
[BASELINE]
[INTERMED SCR]
c. * Inswu-hako1 Suzi-nun [Swunhi-ka ⟨PP1 ⟩ sanchaykha-n-ta] saynkakha-ta. Inswu-COM Suzi-TOP Swunhi-NOM go.for.a.walk-PST-DEC think-DEC ‘Intended: Suzi thinks that Sunhi goes for a walk with Insoo.’ [LONG SCR]
7.5 GERMAN
271
(98) No long scrambling of subjects in Korean * Minho-ka1 Suzi-nun [⟨DP1 ⟩ chayk-ul ilk-nun-ta-ko] Minho-NOM Suzi-TOP book-ACC read-PRS-DECL-COMP saynkakha-y. think-INT ‘Intended: Suzie thinks that Minho is book-reading.’
(99) No long scrambling of subjects in Korean *Halmeni-kkeyse1 Suzi-nun [⟨DP1 ⟩ grandmother-NOM.HON Suzi-TOP
chayk-ul book-ACC
ilke-si-n-ta-ko] saynkakha-y. read-HON-PRS-DECL-COMP think-INT ‘Suzie thinks that grandmother is book-reading.’ This concludes the discussion of Korean PNI-ed arguments and their movement restrictions. While PNI-ed arguments in Korean are permitted to scramble clauseinternally, long scrambling is not allowed, identical to the movement properties of (remnant) VPs. Long scrambling is blocked since Korean C heads are unable to come with a movement-inducing [•V•]-feature which prevents VPs as well as PNI-ed arguments to escape the CP phase.
7.5 German German matrix clauses constitute CPs, the finite verb moves to C. In general, word order is relatively free, that is nominal arguments and prepositional arguments can scramble or topicalize to spec,CP (Fanselow 2002b; Frey 2006; Schweikert 2005). Frey (2015), however, has shown that PNI-ed arguments can topicalize but not scramble. Section 5.4.5 provided a PNI analysis of bare plurals and non-specific indefinites, based on the positions these noun types can appear in, while Section 4.3.5 demonstrated that German PNI-ed arguments retain case marking when pseudo-incorporated. I repeat the relevant data sets in (100) and (102), accompanied by sketched underlying structures in (101) and (103). Let me first address bare plurals. Topicalization of a VP and a bare plural is exemplified in (100b) and (100c), respectively. The structure in (100d) is not acceptable since the bare plural needs to scramble out of the VP first, before the remnant VP can topicalize. (100) Frey (2015: 226-228) a.
dass Max heute Karten spielen wird that Max today cards play will ‘that Max will play cards today’
German
272
PNI-PROPERT Y III: MOVEMENT PATTERNS b.
Karten spielen wird Max heute. cards play will Max today ‘Max will play cards today.’
c.
Karten wird Max heute spielen. cards will Max today play ‘Max will play cards today.’
d. *Spielen wird Max heute Karten. play will Max today cards ‘Max will play cards today.’
(101) Syntactic structures for (100) a. [CP [C dass] [TP Max heute Karten spielen wird ]] b. [ CP [VP Karten spielen]2 [C ′ wird Max heute ⟨V P2 ⟩ ]] c. [ CP [Karten] 1 [C ′ wird Max heute [VP ⟨Karten 1 ⟩ spielen] ]] d. [CP [VP ⟨Karten 1 ⟩ spielen] 2 [C ′ wird Max heute [Karten] 1 ⟨V P2 ⟩ ]]
We now turn to indefinite objects. Recall from the discussion in Section 5.4.5 that the adverb wunderbar marks the VP/vP boundary. The unacceptability of (102b) provides evidence that objects following the adverb have not vacated their base positions. If they had done so, remnant VP-topicalization would proceed successfully. The indefinite in (102c) precedes the adverb, thus it must have left the VP, and presumably also the entire event domain. Since we assumed in Section 5.4.5 that quantifiers QR overtly in German and specificity is hard-wired into the denotation of the existential quantifier that makes up the indefinite in (102c), an obligatorily specific reading is triggered. The indefinite in (102a) on the other hand is derived with the PNI determiner which is why it follows the manner adverb and receives a non-specific reading. Crucial for the obligatory specific reading of the indefinite in (102c) is that PNI-ed arguments do not scramble, otherwise they would be able to immediately precede the adverb, hence creating ambiguity between specific and non-specific readings.
7.5 GERMAN
(102) Frey (2015: 237,239) a.
273
German
Otto hat heute wunderbar eine charmante Mozart-Sonate Otto has today beautifully a charming Mozart sonata gespielt. played ‘Today Otto played a charming Mozart sonata beautifully.’
b. *Gespielt hat Otto heute wunderbar eine charmante played has Otto today beautifully a charming Mozart-Sonate. Mozart sonata ‘Today Otto played a charming Mozart sonata beautifully.’ c.
Gespielt hat Otto heute eine charmante Mozart-Sonate played has Otto today a charming Mozart sonata wunderbar. beautifully ‘Today Otto played a charming Mozart sonata beautifully.’
[+SPEC]
(103) Syntactic structures for (102) a.
[CP Otto2 [C hat1 ] [TP heute ⟨DP2 ⟩ wunderbar [VP M-S gespielt] ⟨hat1 ⟩ ]]
b.
[ CP [V gespielt]2 [C ′ hat Otto heute wunderbar ⟨V P2 ⟩ ]] +spec!
c.
[CP [VP ⟨M-S 1 ⟩ gespielt] 2 [C ′ hat Otto heute [M-S] 1 wunderbar ⟨V P2 ⟩ ]]
As predicted by the current theory, VPs exhibit exactly the same movement restrictions as PNI-ed arguments. In contrast to the other PNI languages of the core data set, German does not necessarily fuse tense/agreement or C morphology with V. Hence, VP-movement can be diagnosed in a straightforward manner. It is widely acknowledged that German VPs can undergo topicalization, but not scrambling (Grewendorf 1995; Grewendorf and Sternefeld 1990; Müller 1998; von Stechow and Sternefeld 1988). A minimal pair is given in (104) with the underlining structures in (105) and (106).
274
PNI-PROPERT Y III: MOVEMENT PATTERNS
(104) Grewendorf (1995: 1306) a.
German
Das Buch gegeben hat Peter dem Jungen. the book given has Peter the boy ‘Peter gave the book to the boy.’
b. *weil das Buch gegeben Peter dem Jungen hat because the book given Peter the boy has ‘because Peter gave the book to the boy’
(105)
Syntactic structure for (104a) [CP [VP das Buch gegeben] 2 [C ′ hat Peter dem Jungen ⟨V P2 ⟩ ]]
(106)
Syntactic structure for (104b) [CP [C weil] [TP [VP das Buch gegeben]2 [TP Peter dem Jungen ⟨V P2 ⟩ hat]]]
Let us now turn to the implementation of the scrambling-topicalization contrast, observed for VPs and PNI-ed objects in this section. In terms of scrambling, German patterns with Tamil and Mongolian and therefore receives the same analysis. The necessary steps are shown in (107). Like Tamil and Mongolian, German is a language which cannot equip the T head with a [•V•] feature, neither alone nor in combination with another feature. The numeration, as shown in (107a), will always lead to a balanced phase when it comes to VP-movement or PNI-movement since there is no [•V•] feature to be matched. Hence, no EF is inserted, PNI-DP is thus trapped inside VP. T enters the derivation in (107d) and assigns nominative case to the subject. Finally, (107e) shows why PNI-DP is immobile. Neither is there a movement-inducing feature available, nor is PNI-DP accessible. (107) No short or intermediate scrambling in German a.
Numeration: {C, T, D,...} → phase balanced
b.
No EF insertion in ditransitive contexts: v′ P NI -DP 1
v′
N CASE : ACC
V
v
ApplP
D
A PPL
DP
Appl ′
N
VP
D CASE : DAT
D
P NI -DP 1
Appl V
N
D
D
V
CASE : ACC
V
V * CASE : DAT * D
* CASE : ACC * D
v
EF
7.5 GERMAN
c.
275
No EF insertion in transitive contexts: vP P NI -DP 1
v′
N
v
VP
D * CASE : ACC * V
V
V
P NI -DP 1 N
D
D
V
EF
* CASE : ACC * D
v
CASE : ACC
V
d. Case assignment: T′ NOM
vP DP
T v
v′
N
* CASE : NOM *
CASE : NOM
D
T
v
VP
D
V
P NI -DP
V
D
V
N
* CASE : ACC *
D
D
v
* CASE : ACC * V
e.
No trigger for PNI-DP-movement: TP
T′
P NI -DP 1 N
V
T
vP
D CASE : ACC
v
DP
v′
N
CASE : NOM
D
v
VP
D P NI -DP 1
V
V
N
D
D
V
* CASE : ACC * V
* CASE : NOM * T
* CASE : ACC * D
v
phase domain
Topicalization is a possible movement operation for PNI-ed objects and VPs. This observation receives a straightforward implementation by assuming that C can occur with movement-inducing [•V•]. If the numeration contains this feature, the CP phase is unbalanced, leading to EF insertion, shown in (108b). T assigns case but does not attract an argument to its specifier (108c). Finally, C triggers topicalization for PNI-DP in (108d). (108) Topicalization in German a.
Numeration: {C[...•V•...] , T, D,...} → phase not balanced
276
PNI-PROPERT Y III: MOVEMENT PATTERNS b.
EF insertion:
vP v′
P NI -DP 1 N
V
* CASE : ACC * V
P NI -DP 1
V
* CASE : ACC *
N
D
D
D
V
v
EF
* CASE : ACC * V
2
c.
v
VP
D
1
Case assignment:
T′ vP
T
NOM
DP
v
v′
N
* CASE : NOM *
P NI -DP 1
D
N
CASE : NOM
T
v′ v
VP
D * CASE : ACC * V
V
V
⟨DP1 ⟩
* CASE : ACC *
D
EF
V
D
v
d. Movement to the final landing site: CP
P NI -DP 1
C′
N
C
T′
D CASE : ACC
T
vP
V DP
v′
N
D CASE : NOM
P NI -DP 1
...
C
v
VP
D V
V
v′
N
* CASE : ACC *
T
⟨DP1 ⟩
V
V
D
V
* CASE : ACC * EF D
v
phase domain
This section provided clear evidence for the parallel movement properties of VPs and PNI-ed objects in German. In contrast to the preceding sections, VPmovement can be more readily identified, as there is no fused tense/agreement morphology to consider as an intervening factor. Hence, German constitutes an important member of the language set investigated in this study. While German might not at first site be considered a PNI language since there is no effect in case
7.5 GERMAN
277
marking, it shows all the more convincing evidence for the verbal nature of PNI-ed arguments in terms of its movement capacities. Overall this chapter showcased the parallelism we find across languages between the movement patterns of PNI-ed arguments and those of VPs. Since the current proposal ties movement restrictions to the verbal nature of PNI-ed arguments, a straightforward implementation suggests itself which makes reference to categorially triggered movement operations. Since the composition of feature bundles/lists for functional heads can be prone to cross-linguistic variation, the diversity of PNI-movement patterns is not surprising. The analysis is, in this sense, superior to traditional DP/NP approaches which essentially derive PNI properties from the lack of a DP-shell, thus predicting complete immobility (Barrie and Li 2015; Dayal 2011; Frey 2015; López 2012; Massam 2001) or surface adjacency (Levin 2015).¹⁷ Finally, an additional comparison to the movement patterns of PPs in the respective languages (where this was possible) excluded an alternative analysis, by which the movement restrictions are simply attributed to the non-nominal character of PNI-ed arguments. Since PPs pattern with case marked DPs, but crucially not with PNI-ed arguments, in terms of their movement possiblities, the cut does not appear to be one between nominal and non-nominal categories. This furthermore justifies an analysis which identifies PNI as a process of category change rather than of size.
¹⁷ A notable exception is Müller (2018) where the DP-shell of the PNI-ed argument is removed in the course of the derivation, thereby enabling [•D•]-triggered movement operation prior to DP-shell removal.
8 Differential object marking This chapter is dedicated to the noun types of this study for which we have identified optional case marking as an effect of differential object marking. If nothing else is said, these noun types are predicted to be unaffected in their scope, binding, control, and movement properties, with or without case marking. While this prediction is by and large borne out, there are also some exceptions. An overview for the two PNI languages which use DOM-marking for a subset of arguments is given in Tables 8.1 and 8.2. The most noticeable exception is the control ability in Tamil which correlates with DOM. The ability to act as a controller seems to be tied to case marking generally, independent of whether it is caused by PNI or DOM. In this chapter, this requirement will be linked to the role specificity plays in Tamil. The scope data was for the most part already discussed in Chapter 5 and will briefly be repeated for convenience. Some of the binding data was used in Chapter 4 to motivate the impoverishment rules/constraints. I will provide a complete picture in this chapter and extend the observations to scrambling and control properties. Table 8.1 DOM-marked noun types in Tamil Numeral Quantifier Demonstrative Weak definite
Table 8.2 DOM-marked noun types in Korean Numeral-classifier
Demonstrative
Pseudo-Noun Incorporation and Differential Object Marking. Imke Driemel, Oxford University Press. © Imke Driemel (2023). DOI: 10.1093/oso/9780192866400.003.0008
8.1 TAMIL
279
8.1 Tamil We established in Section 5.4.1 that only bare nouns show a low scope restriction with case drop. The crucial wide scope contexts are repeated here for oru NP (1), numeral #2 (2), and the quantifier sila (3). (1)
Context ∃¬: A friend of mine wants to sell three books and after looking at them I bought two books from him. So there is a book which I did not buy from him. Naan oru pustagam / oru pustagath-ai vaŋga-lle. 1SG.NOM one book / one book-ACC buy-NEG ‘I didn’t buy a book.’
(2)
Context 2¬: A friend of mine wants to sell four books and after looking at them I bought two books from him. So there are two books which I did not buy from him. rendu pustagam / rendu pustagath-ai vaŋga-lle. Naan 1SG.NOM two book / two book-ACC buy-NEG ‘I didn’t buy two books.’
(3)
Tamil
Tamil
Context ∃¬: A friend of mine wants to sell some books and after looking at them I bought a few books from him. There are still some books left which I didn’t buy so you should have a look as well. Naan sila pustagam / sila pustagath-ai vaŋga-lle. 1SG.NOM some book / some book-ACC buy-NEG ‘I didn’t buy some book.’
Tamil
Recall from Section 7.1 that caseless bare nouns are immobile. In contrast, caseless numerals as well as other caseless quantifiers can freely scramble without case-marking, shown in (4) for oru NP, in (5) for the weak quantifier pala, and in (6) for the strong quantifier ella. The paradigms provide the noun types in question preceding a temporal adverb, an indirect object (short scrambling), a subject (intermediate scrambling), and a matrix subject (long scrambling). Note that scrambling out of a finite clause is slightly marked for caseless objects since it likely causes processing problems to parse three linearly adjacent arguments without overt case marking (Jegan Murugesan, p.c.).
280
DIFFERENTIAL OBJECT MARKING
(4) Scrambling of caseless oru NP
Tamil
a.
Naan inniki oru pustagam paɖi-cc-een. 1SG.NOM today one book read-PST-1SG ‘I read one book today.’
[BASELINE]
b.
Naan oru pustagam inniki paɖi-cc-een. 1SG.NOM one book today read-PST-1SG ‘I read one book today.’
[ADVERB]
c.
Naan [oru pustagam]1 anda ponnu-kiʈʈe ⟨DP1 ⟩ kuɖu-tt-een. 1SG.NOM one book that woman-LOC give-PST-1SG ‘I gave one book to that woman.’ [SHORT SCR]
d. [Oru pustagam]1 naan inniki ⟨DP1 ⟩ paɖi-cc-een. one book 1SG.NOM today read-PST-1SG ‘I read one book today.’ [INTERMED SCR] e.
[Oru pustagam]1 naan anda ponnu-kiʈʈe ⟨DP1 ⟩ kuɖu-tt-een. one book 1SG.NOM that woman-LOC give-PST-1SG ‘I gave one book to that woman.’ [INTERMED SCR]
f. ?[Oru pustagam]1 Mani [Kumaar ⟨DP1 ⟩ one book Mani.NOM Kumaar.NOM paɖi-cc-aan-nnu] so-n-aan. read-PST-3SG.M-COMP say-PST-3SG.M ‘Mani said Kumaar read one book.’ (5) Scrambling of caseless weak quantifier
[LONG SCR] Tamil
a.
Naan inniki pala pustagam paɖi-cc-een. 1SG.NOM today many book read-PST-1SG ‘I read many books today.’
[BASELINE]
b.
Naan pala pustagam inniki paɖi-cc-een. 1SG.NOM many book today read-PST-1SG ‘I read many books today.’
[ADVERB]
c.
Naan [pala pustagam]1 anda ponnu-kiʈʈe ⟨DP1 ⟩ kuɖu-tt-een. 1SG.NOM many book that woman-LOC give-PST-1SG ‘I gave many books to that woman.’ [SHORT SCR]
d. [Pala pustagam]1 naan inniki ⟨DP1 ⟩ paɖi-cc-een. many book 1SG.NOM today read-PST-1SG ‘I read many books today.’ [INTERMED SCR] e.
[Pala pustagam]1 naan anda ponnu-kiʈʈe ⟨DP1 ⟩ kuɖu-tt-een. many book 1SG.NOM that woman-LOC give-PST-1SG ‘I gave many books to that woman.’ [INTERMED SCR]
8.1 TAMIL
281
f. ? [Pala pustagam]1 Mani [Kumaar ⟨DP1 ⟩ many book Mani.NOM Kumaar.NOM paɖi-cc-aan-nnu] so-n-aan. read-PST-3SG.M-COMP say-PST-3SG.M ‘Mani said Kumaar read many books.’
[LONG SCR]
(6) Scrambling of caseless strong quantifier
Tamil
a. Naan inniki ella pustagam-um paɖi-cc-een. 1SG.NOM today all book-ADD read-PST-1SG ‘I read all the books today.’
[BASELINE]
b. Naan ella pustagam-um inniki paɖi-cc-een. 1SG.NOM all book-ADD today read-PST-1SG ‘I read all the books today.’
[ADVERB]
c. Naan [ella pustagam-um]1 anda ponnu-kiʈʈe ⟨DP1 ⟩ kuɖu-tt-een. 1SG.NOM all book-ADD that woman-LOC give-PST-1SG ‘I gave all the books to that woman.’ [SHORT SCR] d. [Ella pustagam-um]1 naan inniki ⟨DP1 ⟩ paɖi-cc-een. all book-ADD 1SG.NOM today read-PST-1SG ‘I read all the books today.’ [INTERMED SCR] e. [Ella pustagam-um]1 naan anda ponnu-kiʈʈe ⟨DP1 ⟩ kuɖu-tt-een. all book-ADD 1SG.NOM that woman-LOC give-PST-1SG ‘I gave all the books to that woman.’ [INTERMED SCR] f. [Ella pustagam-um]1 Mani [Kumaar ⟨DP1 ⟩ all book-ADD Mani.NOM Kumaar.NOM paɖi-cc-aan-nnu] so-n-aan. read-PST-3SG.M-COMP say-PST-3SG.M ‘Mani said Kumaar read all the books.’
[LONG SCR]
The pattern can be replicated for demonstrative phrases, shown in (7). (7) Scrambling of caseless demonstrative
Tamil
a.
Naan inniki inta pustagam paɖi-cc-een. 1SG.NOM today DEM book read-PST-1SG ‘I read this book today.’
[BASELINE]
b.
Naan inta pustagam inniki paɖi-cc-een. 1SG.NOM DEM book today read-PST-1SG ‘I read this book today.’
[ADVERB]
c.
Naan [inta pustagam]1 anda ponnu-kiʈʈe ⟨DP1 ⟩ kuɖu-tt-een. 1SG.NOM DEM book that woman-LOC give-PST-1SG ‘I gave this book to that woman.’ [SHORT SCR]
282
DIFFERENTIAL OBJECT MARKING d. [Inta pustagam]1 naan inniki ⟨DP1 ⟩ paɖi-cc-een. DEM book 1SG.NOM today read-PST-1SG ‘I read this book today.’ [INTERMED SCR] e.
[Inta pustagam]1 naan anda ponnu-kiʈʈe ⟨DP1 ⟩ kuɖu-tt-een. DEM book 1SG.NOM that woman-LOC give-PST-1SG ‘I gave this book to that woman.’ [INTERMED SCR]
f. ?[Inta pustagam]1 Mani [Kumaar ⟨DP1 ⟩ DEM book Mani.NOM Kumaar.NOM paɖi-cc-aan-nnu] so-n-aan. read-PST-3SG.M-COMP say-PST-3SG.M ‘Mani said Kumaar read this book.’
[LONG SCR]
Table 8.1 highlights that weak definites are an exception to the overall empirical picture we have provided so far. Caseless weak definites seem to become immobile when they lose their case marker. This is true for global weak definites as well as immediate ones; see (8) and (9). I leave it to future research to provide an explanation for what singles out weak definite phrases from the rest of the DOM-marked noun types in terms of movement. (8) Scrambling of caseless global weak definites a.
Tamil
Naan nethikku nila paar-t-een. 1SG.NOM yesterday moon see-PST-1SG ‘I saw the moon yesterday.’
[BASELINE]
b. %Naan nila nethikku paar-t-een. 1SG.NOM moon yesterday see-PST-1SG ‘I saw the moon yesterday.’
[ADVERB]
c.
*Nila1 naan nethikku ⟨DP1 ⟩ paar-t-een. moon 1SG.NOM yesterday see-PST-1SG ‘I saw the moon yesterday.’
[INTERMED SCR]
d. *Nila1 Mani [naan nethikku ⟨DP1 ⟩ paar-t-een-nnu] moon Mani.NOM 1SG.NOM yesterday see-PST-1SG-COMP so-n-aan. say-PST-3SG.M ‘Mani said I saw the moon yesterday.’
[LONG SCR]
(9) Context (immediate weak definite): Imagine a scenario in an office room that has a projector for presentations. a.
Naan inniki projector repair panni-n-een. 1SG.NOM today projector repair do-PST-1SG ‘I repaired the projector today.’
[BASELINE]
8.1 TAMIL
b. %Naan projector inniki repair panni-n-een. 1SG.NOM projector today repair do-PST-1SG ‘I repaired the projector today.’
283
[ADVERB]
c. % Naan projector1 anda ponnu-kiʈʈe ⟨DP1 ⟩ kuɖu-tt-een. 1SG.NOM projector that woman-LOC give-PST-1SG ‘I gave the projector to that woman.’ [SHORT SCR] d. *?Projector1 naan inniki ⟨DP1 ⟩ repair panni-n-een. projector 1SG.NOM today repair do-PST-1SG ‘I repaired the projector today.’ [INTERMED SCR] e. *? Projector1 naan anda ponnu-kiʈʈe ⟨DP1 ⟩ kuɖu-tt-een. projector 1SG.NOM that woman-LOC give-PST-1SG ‘I gave the projector to that woman.’ [INTERMED SCR] f.
* Projector1 Mani [naan ⟨DP1 ⟩ repair panni-n-een] projector Mani.NOM 1SG.NOM repair do-PST-1SG so-n-aan. say-PST-3SG.M ‘Mani said I repaired the projector.’
[LONG SCR]
Moving on to binding, I have motivated the DOM-status of weak definites, demonstratives, and quantifiers in Section 4.3.1 with the observation that these noun types can still act as binders, even without case marking. The data is repeated in (10).¹ (10) Binding for caseless weak definites, demonstratives, quantifiers
Tamil
a.
Naan Jupiter(-ai)-um1 ate-ode1 nila.v-ai-um paar-t-een. I Jupiter-ACC-CONJ that-GEN moon-ACC-CONJ see-PST-1SG ‘I saw Jupiter and its moon.’
b.
Kumar [inta book(-ai)]1 ata-ode1 kadasi pakkam varai Kumar.NOM DEM book-ACC that-GEN last page until padi-c-aan. read-PST-3SG.M ‘Kumar read this book until its last page.’
c.
Kumar [ella book(-ai)-um]1 ata(-t)-ode1 kadasi pakkam Kumar.NOM all book-ACC-ADD that-RED-GEN last page varai padi-c-aan. until read-PST-3SG.M ‘Kumar read all the books until their last pages.’
¹ One speaker prefers the weak quantifier pala in (10c) to co-occur with plural-marked nouns and an additional particle, i.e. booku-gal-ai-um (= book-PL-ACC-ADD).
284
DIFFERENTIAL OBJECT MARKING d. Kumar [pala book(-ai)]1 ata-ode1 kadasi pakkam varai Kumar.NOM many book-ACC that-GEN last page until padi-c-aan. read-PST-3SG.M ‘Kumar read many books until their last pages.’
I also provide an immediate weak definite contexts in (11) which confirms that weak definites, be it in a global or an immediate scenario, do not depend on case marking to bind a possessive pronoun. (11)
Context (immediate weak definite): Imagine a scenario in an office room that has a projector for presentations. Inniki projector(-ai)1 ata-ode1 idath-le mati-n-een. today projector-ACC that-GEN place-LOC hang-PST-1SG ‘I hung the projector in its place.’
Tamil
We can extend the data set to oru NPs. The case-marked numeral in (12a) can bind the possessive pronoun. Dropping the case marker in this configuration, as in (12b), is acceptable to two of my four consultants. All speakers accept the structure in (12c), where the bindee additionally scrambles over the binder. (12) Binding for caseless oru NP a.
Tamil
Kumar [oru book-ai]1 ata-ode1 kadasi pakkam varai Kumar.NOM one book-ACC that-GEN last page until padi-c-aan. read-PST-3SG.M ‘Kumar read a book until its last page.’
b. %Kumar [oru book]1 ata-ode1 kadasi pakkam varai Kumar.NOM one book that-GEN last page until padi-c-aan. read-PST-3SG.M ‘Kumar read a book until its last page.’ c.
Kumar [ata-ode1 kadasi pakkam varai]2 [oru book]1 ⟨PP2 ⟩ Kumar.NOM that-GEN last page until one book padi-c-aan. read-PST-3SG.M ‘Kumar read a book until its last page.’
Across all argument types investigated in Tamil, case marking is required for the ability to control PRO. This includes the argument types in Table 8.1. The data is given in (13) to (14).
8.1 TAMIL
(13) No control with caseless numerals, weak definites, demonstratives a. Raja Raja.NOM
[oru/rendu one/two
naai*(-ye)]i dog-ACC
[PROi
kutikk-a] drink-INF
veliya out
vara] come.INF
285 Tamil
kattaya-paduthi-n-aan. compel-make-PST-3SG.M ‘Raja forced one/two dog(s) to drink.’ b. Indiran Indiran.NOM
suriyan*(-ai)i sun-ACC
[PROi
kataya-paduthi-n-aan. compel-make-PST-3SG.M ‘Indiran forced the sun to come out.’ c. Raja [inta naai*(-ye)]i [PROi kutikk-a] kattaya-paduthi-n-aan. Raja.NOM DEM dog-ACC drink-INF compel-make-PST-3SG.M ‘Raja forced this dog to drink.’ (14) No control with caseless quantifiers a. Raja Raja.NOM
[ella all
naai*(-ye)-yum]i dog-ACC-ADD
Tamil [PROi
kutikk-a] drink-INF
kattaya-paduthi-n-aan. compel-make-PST-3SG.M ‘Raja forced all the dogs to drink.’ b. Raja Raja.NOM
[pala many
naai*(-ye)]i dog-ACC
[PROi
kutikk-a] drink-INF
kattaya-paduthi-n-aan. compel-make-PST-3SG.M ‘Raja forced many dogs to drink.’ One way to link the inability to control PRO to case loss in general is by characterizing this interaction as a specificity effect. Landau (2013: 12–13) provides the following contrast between a raising predicate (15) and a control predicate (16). Since the subject in (15) can receive a specific and a non-specific reading, i.e. take scope above and below the intensional operator, it must have raised from a position in the embedded clause to the matrix clause. The subject of the control predicate in (16) can only get a specific reading, thus it must be first-merged in the matrix clause. (15) [Seven civilians]1 are likely [⟨seven civilians1 ⟩ to starve to death this weekend] ⇝ It is likely that seven civilians (whoever they might be) will starve to death this weekend.
286
DIFFERENTIAL OBJECT MARKING ⇝ Seven specific civilians, namely John, Peter, Nancy etc., are likely to starve to death this weekend.
(16) [Seven civilians]1 are afraid [PRO1 to starve to death this weekend] ̸⇝ There is a general fear that seven civilians (whoever they might be) will starve to death this weekend. ⇝ Seven specific civilians, namely John, Peter, Nancy, etc., are afraid to starve to death this weekend. While Landau analyses the contrast above as arising from scopal specificity, one might just as well take the unavailability of a non-specific reading for the control structure in (16) as evidence for a requirement of controllers to be interpreted as epistemically specific. As the reader might recall from Chapter 4, Tamil is the only language of the core data set which ties case loss to non-specific readings and case-marking to specific readings. Hence, I take the fact that controllers have to be case-marked as a probable consequence of the need for controllers to receive a specific interpretation. Support for this hypothesis comes from Hindi and Korean. Like Tamil, Hindi indefinites are equally incapable to establish a control relation if they are not case-marked, recall footnote 2 in Chapter 6. As was discussed at large in Chapter 4, Hindi DOM is sensitive to specificity. Hence, Tamil and Hindi are completely parallel in terms of their interaction between specificity and control. Korean, in contrast, allows DOM-triggered case drop even if the argument controls PRO, as the next section will show. Interestingly, Korean does not show a specificity correlation with case marking. Thus, if case marking is not required for specific readings, it will not be required in control scenarios either. We now turn to Korean and provide the full picture.
8.2 Korean We begin again with the scope facts which are repeated from Section 5.4.4 in (17) and (18) for numeral classifiers. Both contexts enforce wide scope readings in which numeral classifier constructions are licensed, with or without case. (17)
Context 1¬: Yusu’s friend wants to sell three flowers and Yusu bought two from him. So there is one flower Yusu did not buy. [Kkoch(-ul) han-songi]1 Yusu-ka ⟨DP1 ⟩ sa-ci anh-ass-ta. flower-ACC one-CL Yusu-NOM buy-CI NEG-PST-DECL ‘One flower, Yusu did not buy.’
8.2 KORE AN
(18)
287
Context 2¬: A friend of Yusu wants to sell four flowers and after looking at them Yusu bought two flowers from him. So there are two flowers which Yusu did not buy from him. [Kkoch(-ul) twu-songi]1 Yusu-ka ⟨DP1 ⟩ sa-ci anh-ass-ta. flower-ACC two-CL Yusu-NOM buy-CI NEG-PST-DECL ‘Two flowers, Yusu did not buy.’
Recall from Section 7.4 that PNI-ed arguments cannot undergo long scrambling, but they can be scrambled clause-internally. We provide the full scrambling paradigm for the numeral classifiers #1 and #2 in (19) and (20), as well as demonstrative phrases in (21). Long scrambling is also slightly marked for caseless numeral classifiers. They differ, however, clearly from the completely unacceptable argument types whose caseloss is due to PNI, discussed in Section 7.4. (19) Scrambling of caseless NP han-Cl in object position a. Suzi-ka Suzi-NOM
emma-eykey mom-DAT
chinkwu(-lul) friend-ACC
Korean
han-myeng one-CL
sokay-hay-ess-e. introduce-v-PST-INT ‘Suzi introduced one friend to her mom.’
[BASELINE]
b. Suzi-ka [chinkwu(-lul) han-myeng]1 emma-eykey ⟨DP1 ⟩ Suzi-NOM friend-ACC one-CL mom-DAT sokay-hay-ess-e. introduce-v-PST-INT ‘Suzi introduced one friend to her mom.’
[SHORT SCR]
c. [Chinkwu(-lul) han-myeng]1 Suzi-ka emma-eykey ⟨DP1 ⟩ friend-ACC one-CL Suzi-NOM mom-DAT sokay-hay-ess-e. introduce-v-PST-INT ‘Suzi introduced one friend to her mom.’
[INTERMED SCR]
d. [Chinkwu? (-lul) han-myeng]1 na-nun [Suzi-ka emma-eykey ⟨DP1 ⟩ friend-ACC one-CL I-TOP Suzi-NOM mom-DAT sokay-ss-ta-ko] saynkakha-y. introduce-v-PST-DECL-COMP ‘I think Suzi introduced one friend to her mom.’
[LONG SCR]
288
DIFFERENTIAL OBJECT MARKING
(20) Scrambling of caseless NP #-Cl in object position a. Suzi-ka Suzi-NOM
emma-eykey mom-DAT
chinkwu(-lul) friend-ACC
Korean twu-myeng two-CL
sokay-hay-ess-e. introduce-v-PST-INT ‘Suzi introduced two friends to her mom.’
[BASELINE]
b. Suzi-ka [chinkwu(-lul) twu-myeng]1 emma-eykey ⟨DP1 ⟩ Suzi-NOM friend-ACC two-CL mom-DAT sokay-hay-ess-e. introduce-v-PST-INT ‘Suzi introduced two friends to her mom.’
[SHORT SCR]
c. [Chinkwu(-lul) twu-myeng]1 Suzi-ka emma-eykey ⟨DP1 ⟩ friend-ACC two-CL Suzi-NOM mom-DAT sokay-hay-ess-e. introduce-v-PST-INT ‘Suzi introduced two friends to her mom.’
[INTERMED SCR]
d. [Chinkwu? (-lul) twu-myeng]1 na-nun [Suzi-ka emma-eykey ⟨DP1 ⟩ friend-ACC two-CL I-TOP Suzi-NOM mom-DAT sokay-ss-ta-ko] saynkakha-y. introduce-v-PST-DECL-COMP think-INT ‘I think Suzi introduced two friends to her mom.’
[LONG SCR]
(21) Scrambling of caseless demonstratives in object position
Korean
a. Suzi-ka Minsu-eykey i/ce chikwauysa(-lul) sokay-hay-ess-e. Suzi-NOM Minsu-DAT DEM dentist-ACC introduce-v-PST-INT ‘Suzi introduced this/that dentist to Minsu.’ [BASELINE] b. Suzi-ka [i/ce chikwauysa(-lul)]1 Minsu-eykey ⟨DP1 ⟩ Suzi-NOM DEM dentist-ACC Minsu-DAT sokay-hay-ess-e. introduce-v-PST-INT ‘Suzi introduced this/that dentist to Minsu.’
[SHORT SCR]
c. [I/ce chikwauysa(-lul)]1 Suzi-ka Minsu-eykey ⟨DP1 ⟩ DEM dentist-ACC Suzi-NOM Minsu-DAT sokay-hay-ess-e. introduce-v-PST-INT ‘Suzi introduced this/that dentist to Minsu.’
[INTERMED SCR]
8.2 KORE AN
289
d. [I/ce chikwauysa? (-lul)]1 na-nun [Suzi-ka Minsu-eykey ⟨DP1 ⟩ DEM dentist-ACC I-TOP Suzi-NOM Minsu-DAT sokay-ss-ta-ko] saynkakha-y. introduce-v-PST-DECL-COMP think-INT ‘I think Suzi introduced this/that dentist to Minsu.’
[LONG SCR]
As was pointed out in Section 7.4, Korean subjects generally do not allow for long scrambling. Hence, we cannot test whether DOM-marked subjects undergo long scrambling. What can be demonstrated, however, is that caseless numeral classifiers and caseless demonstratives are in principle mobile. Using the same temporal adverb as in Section 7.4, we see in (22) and (23) that numeral classifiers and demonstratives can follow and precede the adverb. (22) Scrambling of caseless NP #-Cl in subject position
Korean
a. Yecenhi yuksangsenswu(-ka) sey-myeng cip-ey iss-e. still athlete-NOM three-CL home-at exist-INT ‘Three athletes are still at home.’ [BASELINE] b. [Yuksangsenswu(-ka) sey-myeng]1 yecenhi ⟨DP1 ⟩ cip-ey iss-e. athlete-NOM three-CL still home-at exist-INT ‘Three athletes are still at home.’ [ADVERB] (23) Scrambling of caseless demonstratives in subject position a. Yecenhi i/ce yuksangsenswu(-ka) cip-ey iss-e. still DEM athlete-NOM home-at exist-INT ‘This/that athlete is still at home.’
Korean
[BASELINE]
b. [I/ce yuksangsenswu(-ka)]1 yecenhi ⟨DP1 ⟩ cip-ey iss-e. DEM athlete-NOM still home-at exist-INT ‘This/that athlete is still at home.’ [ADVERB] Section 4.3.4 showed that demonstratives can still act as binders for reflexive pronouns, even without case. We repeat this observation in (24) and extend it to numeral classifiers. (24) Caseless demonstratives and NP #-Cl can bind a. [I/ce koyangi(-ka)]1 [ku casin-ul]1 halth-ass-e. DEM cat-NOM 3SG self-ACC lick-PST-INT ‘This/that cati washed itselfi .’ b. [Koyangi(-ka) han-mali]1 [ku casin-ul]1 halth-ass-e. cat-NOM one-CL 3SG self-ACC lick-PST-INT ‘One cati washed itselfi .’
Korean
290
DIFFERENTIAL OBJECT MARKING c. [Koyangi(-ka) twu-mali]1 [ku casin-ul]1 halth-ass-e. cat-NOM two-CL 3SG self-ACC lick-PST-INT ‘Two catsi washed themselvesi .’
We will now provide evidence that case loss due to DOM does not affect the ability to control PRO. This observation holds for object control (25), subject control (26), and subject adjunct control (27). Again, as with long scrambled caseless numerals, there is some markedness noticeable with caseless numerals as controllers in object control clauses as well as in adjunct control clauses. The judgements, however, clearly contrast with the unacceptability of PNI-ed arguments in such positions. (25) Object control for caseless demonstratives and NP #-Cl a.
Korean
?
Yusu-ka [i/ce haksayng (-ul)]1 [PRO1 ttena-la-ko] Yusu-NOM DEM student-ACC leave-IMP-COMP seltukhay-ss-e. persuade-PST-INT ‘Yusu persuaded this/that student to leave.’
b.
Yusu-ka [haksayng(-ul) han-myeng]1 [PRO1 ttena-la-ko] Yusu-NOM student-ACC one-CL leave-IMP-COMP seltukhay-ss-e. persuade-PST-INT ‘Yusu persuaded one student to leave.’
c.
Yusu-ka [haksayng(-ul) twu-myeng]1 [PRO1 ttena-la-ko] Yusu-NOM student-ACC two-CL leave-IMP-COMP seltukhay-ss-e. persuade-PST-INT ‘Yusu persuaded two students to leave.’
(26) Subject control for caseless demonstratives and NP #-Cl
Korean
a. [I/ce haksayng(-i)]1 [PRO1 ttena-keyss-ta-ko] kyelsimhay-ss-e DEM student-NOM leave-VOL-DECL-COMP decide-PST-INT ‘This student decided to leave.’ b. [Haksayng(-i) han-myeng]1 [PRO1 ttena-keyss-ta-ko] student-NOM one-CL leave-VOL-DECL-COMP kyelsimhay-ss-e decide-PST-INT ‘One student decided to leave.’
8.2 KORE AN
291
c. [Haksayng(-i) twu-myeng]1 [PRO1 ttena-keyss-ta-ko] student-NOM two-CL leave-VOL-DECL-COMP kyelsimhay-ss-e decide-PST-INT ‘Two students decided to leave.’
(27) Adjunct subject control for caseless demonstratives and NP #-Cl
Korean
a. [Pilok PRO1 olay-toyess-ciman] [i/ce khemphyuthe(-ka)]1 cal although old-became-but DEM computer-NOM well tolak-a. work-INT ‘[Although PRO1 is old] [this/that computer]1 works well.’ b. [PRO1 kangto-lul mokkyekhan hwuey] [namca? (-ka) han-myeng]1 robber-ACC witness after man-NOM one-CL kyengchal-ul pwull-ess-e. police-ACC call-PST-INT ‘[One man]1 called the police [after PRO1 witnessing a robber]. c. [PRO1 kangto-lul mokkyekhan hwuey] [namca? (-ka) twu-myeng]1 robber-ACC witness after man-NOM two-CL kyengchal-ul pwull-ess-e. police-ACC call-PST-INT ‘[Two men]1 called the police [after PRO1 witnessing a robber]. This concludes our discussion of DOM-marked noun types in Korean and Tamil. In contrast to PNI-ed argument types, caseless nouns triggered by differential object marking are much less restricted in terms of their scopal, binding, control, and movement capabilities. The contrast between the noun types discussed in this section and the noun types discussed in the rest of the book necessitate a distinction between DOM and PNI. Pseudo-noun incorporation is triggered by the presence of a hybrid determiner which is responsible not only for lack of case marking but also for other semantic and syntactic properties we have employed as diagnostics in this study, all going back to the verbal nature of PNI-ed arguments. differential object marking is a separate phenomenon which does not necessarily trigger the additional restrictions we see with PNI. But it does not exclude them either. As we have seen for Tamil, control relations require case-marked controllers generally—an observation we have explained by adherence to a general specificity requirement for controllers.
9 Previous approaches This chapter reviews previous approaches to pseudo-noun incorporation and differential object marking. Section 9.1 discusses head movement accounts of PNI languages. Section 9.2 gives an overview over the most widely adopted account within the PNI literature, the DP/NP approach. In Section 9.3, we will look at raising accounts. Each section will summarize the main idea and point out shortcomings, either conceptually or with respect to the empirical consequences concerning the data set of this study.
9.1 Head movement accounts Head movement accounts are predominantly proposed for noun incorporation (Baker 1988, 1995, 2009; Baker et al. 2005; Chung and Ladusaw 2004; Sadock 1980). Baker (2014b), however, extends his analysis to pseudo-noun incorporation (1b), taking Sakha and Tamil as case studies. PNI is different from noun incorporation in that it has to be string vacuous, i.e. noun and verb have to be linearly adjacent. (1) Baker (2014b: 9) a. Naan nalla paʐam tee-r-een. I good fruit seek-PRS-1SG.SBJ ‘I am looking for (some/a) good fruit(s).’ b.
Tamil
VP V
NP AP
N′
nalla
N
N
V
paᶼam i tee
paᶼam i
Baker adopts the copy theory of movement and follows Nunes (2004) in that lower copies are deleted due to the presence of their uninterpretable features. Crucial Pseudo-Noun Incorporation and Differential Object Marking. Imke Driemel, Oxford University Press. © Imke Driemel (2023). DOI: 10.1093/oso/9780192866400.003.0009
9.1 HE AD MOVEMENT ACCOUNTS
293
for Baker’s analysis is that head movement is not enforced by feature checking, thereby making its application completely optional. PNI semantics, however, can only be generated if N and V form a complex head, i.e. N denotes a property only if it is dominated by a V node (Baker 2014b: 20–21). Since the type of head movement shown in (1b) is not feature triggered, either one of the copies can be spelled-out, leading to potential linearization issues if the movement is not string vacuous. Thus, Baker derives the requirement for surface adjacency between N and V through the avoidance of an ordering paradox on PF. Evidence for his proposal comes from the immobility of caseless objects in Tamil and Sakha, which is in line with our observations for Tamil in Chapter 7. To show that linear adjacency is needed, Baker provides data which show that low manner adverbs cannot intervene between the verb and the object; see (2) for Tamil. (2) Baker (2014b: 8–9) Maala (veegamaa) pustagam (*veegamaa) paɖi-cc-aa. Mala quickly book quickly read-PST-3.F.SBJ ‘Mala read a book/books quickly.’
Tamil
None of my informants, however, could verify (2). All four informants accepted sentences of the form in (3a) with an intervening manner adverb, while two even accepted intervening temporal adverbs; see (3b). The variance in speaker judgements is possibly due to dialectal variation, Baker points out that his informant for Tamil speaks a Singaporean dialect. (3) a.
Maala pustagam veegamaa paɖi-cc-aa. Mala.NOM book quickly read-PST-3SG.F ‘Mala read a book/books quickly.’
Tamil
b. % Maala pustagam inniki paɖi-cc-aa. Mala.NOM book today read-PST-3SG.F ‘Mala read a book/books today.’ It is, however, worth pointing out that string adjacency is also not required with respect to focus adverbs/particles which occur to the right of nouns, in between N and V. Lehmann (1993) provides an example for the Tamil equivalent of only; see (4). This data point was accepted by all my consultants. (4) Lehmann (1993: 112) Kumaar oru iṭ li maṭ ṭ um caappiṭ ṭ -aan. Kumar a Idli only eat-PST-3SG.M ‘Kumar ate one Idli only.’
Tamil
The focus particle -kuut˛a behaves similarly. The sentence in (5) shows how it can be right attached to a PNI-ed object.
294
PREVIOUS APPROACHES
(5) Maala pustagam-kuut˛a paɖi-cc-aa. Mala.NOM book-MIR read-PST-3SG.F ‘Mala even read a book/books.’
Tamil
Another argument Baker provides comes from resultative structures. He uses the word order contrast in (6a) and (6b) to illustrate this point, proposing the underlying structure in (6c). PNI is blocked in (6a) since the PP intervenes between the direct object and the verb. As soon as the PP moves out of the way, PNI can successfully be applied. (6) Baker (2014b: 12–14)
Tamil
a. Balaa pustagatt-e/*pustagam mesai-kku kiiɭ va-kkir-aan. Baala book-ACC/book table-DAT under put-PRS-3SG.M ‘Balaa puts book(s) under the table.’ b. Balaa peʈʈi-kku uɭɭe paʐam va-kkir-avan. Baala box-DAT in fruit put-PRS-3SG.M ‘Bala is the one who puts fruit(s) in (the) box(es).’ c. vP v′
NP Balaa
VP
v V′
NP pustagam
PP mesai-kku kii
V va
One potential problem that arises with this argument is the fact that the verb does not c-command the direct object, neither in (6a) nor in (6b), making it difficult for head movement to apply in the first place, if nothing else is said. We could potentially strengthen Baker’s account by the assumption of a small clause (Hoekstra 1988; Kayne 2017; Stowell 1983) that forms a predication relation between the PP and the direct object (see (7)), ensuring that N can head-move to V. The structure can be linearized only if the PP has moved away from its base position which accounts for the contrast between (6a) and (6b).
9.1 HE AD MOVEMENT ACCOUNTS
(7)
295
vP v′
NP Balaa
VP V
SC NP pustagam
v
PP
va
mesai-kku kii
There is at least one reason to assume that a small clause analysis might be dispreferred. Small clauses have been noted to constitute opaque domains for anaphora binding (Bruening 2010, 2018a; Pesetsky 2010). Tamil, however, allows for anaphora binding into the PP by the subject. (8) a. *Maxwell1 considers [SC her proud of himself1 ]. (Bruening 2018a: 554) b.
Balaa1 [SC pustagatt-e tana-kku1 pakkat-le] va-kkir-aan. Tamil Baala.NOM book-ACC REFL-DAT near-LOC put-PRS-3SG.M ‘Balaa puts book(s) next to herself.’
In light of the problems with both Baker’s structure in (6c) as well as the small clause structure in (7), let us investigate how the current account is able to explain the contrast between (6a) and (6b). If we follow Pylkkänen (2008) in analysing directive objects of locative structures in Tamil as arguments introduced by an applicative head which takes VP as its complement, then PP would be merged higher than the direct object, thus making (6b) to represent the underlying structure and (6a) the derived one. As was established in Chapter 7, Tamil, together with Mongolian, Hindi, and German, does not allow for scrambling of VPs and PNI-ed arguments. We implemented this observation through the distribution of categorial scrambling features. Tamil v can only come with [•D•] but not [•V•], allowing short scrambling for proper arguments but not PNI-ed ones. Taken together, these assumptions result in the structure in (9), which provides a new take on the contrast in (6). Since (6a) requires the direct object to move over the PP object, only proper DP arguments can occur. PNI is permitted in (6b) because the object can stay in its base position.
296
PREVIOUS APPROACHES v′
(9) P NI -DP 1
v′
N CASE : ACC
V
v
ApplP
D
V
PP VP
P CASE : DAT
P
* CASE : ACC *
Appl ′
N
P NI -DP 1
Appl [loc] V
N
D
D
V
V * CASE : DAT *
D
( D ) v
P
CASE : ACC
V
Within the current account, derived adjacency effects in PP-locative structures constitute a subcase of the broader generalization that verbal categories, be it VPs or PNI-ed arguments, are simply immobile. The reanalysis of resultatives fits well with the independently provided counter examples to the surface adjacency requirement at the beginning of this section. In order to explain the case loss, Baker follows Baker et al. (2005) by making use of a parameterized deletion rules which removes Φ-features on traces of head movement. Along with Φ-features, Baker assumes case information can also be lost. He provides data from DAT-NOM verbs which select a dative subject and a nominative object and agree with the object. These verbs can show neuter singular (default) agreement, but only with non-specific objects. He concludes that pseudoincorporation blocks agreement. (10) Baker (2014b: 33–34)
Tamil
a. En-akku anda ponnu teve-ppaɖ-r-aa/*teve-ppaɖ-itu. I-DAT that girl.NOM need-suffer-PRS-3SG.F/need-suffer-3SG.N ‘I need that girl.’ (one out of an established group) b. En-akku ponnu teve-ppaɖ-itu. I-DAT girl need-suffer-3SG.N ‘I need a girl (a bride).’ (no specific one in mind) Baker does not provide contexts for (10). With the help of appropriate contexts, I was able to diagnose no correlation between agreement and specificity with my informants. With the non-specific context in (11), either agreement form is available. The same is true for the specific context in (12).
9.1 HE AD MOVEMENT ACCOUNTS
297
(11) Context [−spec]: Mala is producing a play and she got a (random) girl to take part in the play. a. Mala-kku ponnu keɖai-cc-itu. Mala-DAT girl get-PST-3SG.N ‘Mala got a girl.’
Tamil
b. Mala-kku ponnu keɖa-cc-aal. Mala-DAT girl.NOM get-PST-3SG.F ‘Mala got a girl.’ (12) Context [+spec]: Mala is producing a play. Initially, she managed to convince a couple of talented girls to take part but then they decided that it was too much work. One of the girls, however, came back. a. Mala-kku (anda) ponnu keɖai-cc-itu tirumba. Mala-DAT that girl get-PST-3SG.N again ‘Mala got a/that girl back.’
Tamil
b. Mala-kku (anda) ponnu keɖa-cc-aal tirumba. Mala-DAT that girl.NOM get-PST-3SG.F again ‘Mala got a/that girl back.’ Moreover, it seems that ponnu (= girl) is Φ-variant, possibly due to the availability of both formal and semantic Φ-features, akin to German Ma¨dchen (Wurmbrand 2017). The sentence in (13a) shows variable Φ-agreement with ponnu in subject position, in contrast to paiyan in (13b), which is not Φ-variant and thus can only agree with one form. (13) a. Ponnu idli saap-ida.v-aal/saap-idu-tu. girl.NOM idli eat-PFV-3SG.F/eat-PFV-3SG.N ‘The girl ate idlis.’
Tamil
b. Paiyan idli saap-ida.v-aan/*saap-idu-tu. boy.NOM idli eat-PFV-3SG.M/eat-PFV-3SG.N ‘The boy ate idlis.’ Since the availability of singular neuter agreement seems to be tied to the lexical properties of the agreement controller, it is presumably not default agreement that occurs in the non-specific context in (11). Setting up the same context with paiyan instead of ponnu should therefore make singular neuter agreement unavailable. This prediction is borne out; see (14).
298
PREVIOUS APPROACHES
(14) Context [−spec]: Mala is producing a play and she got a (random) boy to take part in the play. a. *Mala-kku paiyan keɖai-cc-itu. Mala-DAT boy get-PST-3SG.N ‘Mala got a boy.’ b.
Tamil
Mala-kku paiyan keɖa-cc-aan. Mala-DAT boy.NOM get-PST-3SG.M ‘Mala got a boy.’
The discussion of the agreement facts above reveals that Tamil patterns with Turkish and Korean, in that (Φ-)agreement and case marking are independent of each other. This behaviour is predicted by the current proposal that implements case loss as a post-syntactic deletion of case markers. At least for Tamil, we also saw empirical evidence against a requirement of surface adjacency. This does not exclude that languages like Sakha or even other varieties of Tamil might exhibit PNI that is dependent on surface adjacency. There is, however, one additional conceptual concern with the head movement account that is worth pointing out. Baker derives surface adjacency from obligatory string vacuous movement—the only type of movement that does not create linearization issues if chain reduction is blocked due to identical feature bundles. This leaves him with the problem of multiple spell-out: if chain reduction is blocked we expect spell-out of adjacent copies. Baker (2014b: 26) adds another assumption to avoid this problem: copy chains whose head is part of a complex morphological object (the complex head in PNI) delete all lower copies of the chain. With this rule in place, however, we never run into linearization issues for head movement to begin with. It does not matter how many constituents intervene between the launching site and the landing site, as long as the highest copy is part of a complex head, all lower copies are deleted. More generally and especially within the scope of this study, Baker’s string vacuous head movement fails to account for languages such as Turkish and Korean which allow for subjects to undergo PNI, a configuration in which head movement is not allowed to occur. Finally, Baker must allow for excorporation out of complex heads in order to account for languages such as Hindi, Turkish, and Korean which permit scrambling of PNI-ed arguments. Another PNI language which has received a head movement analysis is Turkish. While many accounts assume some kind of complex predicate formation (Aydemir 2004; Knecht 1986; Nilsson 1985), either lexically or syntactically, Kornfilt (2003a) proposes that the compactness between object and verb comes about via head movement of N to V. Under her analysis, Turkish nominal arguments are KPs where the K head spells out the case marker. In a PNI context, K is not filled overtly and head movement precedes as in (15). In non-PNI contexts, the K head is overtly filled which blocks head movement to V since the overt realization of
9.1 HE AD MOVEMENT ACCOUNTS
299
the K head is suggested to act as a barrier for antecedent and lexical government of N’s trace, thereby linking lack of case marking to compactness via the Empty Category Principle (Chomsky 1981, 1986). Just like an overt complementizer is responsible for creating that-trace effects, so can an overt case marker intervene for incorporation. (15)
Kornfilt (2003a: 143) VP V
KP K Ni
DP
V
NP D ti N′
ti
ti
As pointed out by Öztürk (2005, 2009), the exact nature of head movement in (15) is unclear. Kornfilt explicitly states that a functional head such as K does not incorporate into the lexical V head (see also Li 1990). Yet, the ramifications for head movement in (15) are not discussed. Furthermore, the analysis is at odds with more recent empirical generalizations, in line with the observations made in Section 7.3, which provide overwhelming evidence that caseless bare nouns in Turkish can freely dislocate from their base positions and thus be separated from verbs. As with Baker’s analysis, an additional issue for the head movement account lies in the fact that subject/agent PNI is also an option in Turkish (Öztürk 2005, 2009). The last approach I want to discuss in this section is the head movement account put forward by Chung and Ladusaw (2004) for Chamorro, an Austronesian language that exhibits obligatory incorporation for verbs of possession. They define a semantic composition rule which they call Restrict that is able to combine the denotations of verbs that select for an argument of type ⟨e⟩ with property-type indefinite arguments. The argument does not saturate the predicate but rather narrows down the denotation of the verb. Existential closure at the vP level existentially closes off the ⟨e⟩-type variable left by the argument, along with the event variable (Chung and Ladusaw 2004: 5). The account ensures obligatory narrow scope of the incorporated object since, by assumption, negation applies above existential closure and quantifiers QR out of the event domain (Chung and Ladusaw 2004: 13-14). Although the account was developed for noun incorporation, it shares some assumptions with the current theory which is why it is worth comparing. The two verbs that require incorporation are given in (16) with the underlining structures in (16c).
300
PREVIOUS APPROACHES
(16) Chung and Ladusaw (2004: 107–109) a. Ga¨i-kareta si Antonio. AGR.have-car UNM Antonio ‘Antonio has a car.’
Chamorro
b. Ta¨i-kareta si Antonio. AGR.not.have-car UNM Antonio ‘Antonio doesn’t have a car.’ c. Structures: (VP 3 ) (¬)
VP 2 VP 1 V′
DP Antonio
NP
V V
N
have
car i
ti
A step-by-step derivation is provided in (17). In (17a), Restrict applies between a verb and a property argument, not saturating the predicate but modifying it. The object argument slot is therefore left open and will only be closed off at the event level. Since the predicate applies to the subject next, Chung and Ladusaw (2004: 10) propose that the λ-prefix for the object can be demoted to a position right above the event variable. This enables functional application to the subject in (17b). Finally, existential closure applies to the individual variable as well as to the event variable in (17c). The negative possession verb in (16b) includes a negative operator which applies above existential closure in (17d). (17) a. Restrict: ⟦V⟧ = λyλxλe[HAVE(y)(x)(e)] (⟦N⟧) = λyλxλe[HAVE(y)(x)(e)∧CAR(y)] b. λ-demotion & functional application: ⟦VP1 ⟧ = λyλxλe[HAVE(y)(x)(e)∧CAR(y)](⟦DP⟧) = λxλyλe[HAVE(y)(x)(e)∧CAR(y)](⟦DP⟧) = λyλe[HAVE(y)(antonio)(e)∧CAR(y)] c. Existential closure: ⟦VP2 ⟧ = λyλe[HAVE(y)(antonio)(e)∧CAR(y)] = ∃e∃y[HAVE(y)(antonio)(e)∧CAR(y)]
9.1 HE AD MOVEMENT ACCOUNTS
301
d. Negation: ⟦VP3 ⟧ = ∃e∃y[HAVE(y)(antonio)(e)∧CAR(y)] = ¬∃e∃y[HAVE(y)(antonio)(e)∧CAR(y)] Chung and Ladusaw (2004: 87-88) show that possession verbs cannot incorporate any argument bigger than an NP, i.e. demonstrative, quantifiers, pronouns, and possessor phrases. Morpho-syntactic evidence for noun incorporation can be given in comparison to non-incorporating verbs. Possession verbs show adjacency effects (Chung and Ladusaw 2004: 143) and furthermore inflect as intransitive predicates. Chamorro indicates valency with inflectional morphology on the verb; see (18). In (19), we see another incorporation example where the inflectional morphology patterns with intransitive verbs, rather than transitive verbs, thereby providing evidence for detransitivization, which is a key characteristic of noun incorporation. (18) Chung and Ladusaw (2004: 84) a. Man-malagu. AGR.INTRANS-run ‘They ran.’
Chamorro
b. Ma-ba¨ba i etta. AGR.TRANS-open the door ‘They opened the door.’ (19) Chung and Ladusaw (2004: 90) Chamorro Man-ga¨i-ga’ ha¨m nuskuantus ga’lagu yan in-pe´peksai AGR.INTRANS-have-pet we several dog and AGR-raise.PROG siha. them ‘We have several dogs and are raising them.’ The structure in (19) highlights another aspect of Chamorro incorporation which in fact is one of the main motivation for the operation Restrict. The incorporated object ga’ can be doubled in the original argument position by the full DP nuskuantus ga’lagu. Since Restrict does not saturate an argument slot, the additional DP argument can saturate the predicate after Restrict has applied. Existential closure will in this case only apply to the event variable. Crucial for the current PNI study is the parallel Chung and Ladusaw (2004) draw to Maori, a Polynesian language which exhibits two types of indefinites, one that creates obligatory low scope readings (see (21)), and one that enables flexible scope; see (20). The former is accompanied by he, the latter by tētahi. Chung
302
PREVIOUS APPROACHES
and Ladusaw (2004) tie these exponent to the spell-outs of two different semantic composition rules, tētahi signals the type shift to a choice function which can be existentially closed above or below negation, whereas he is the morpho-syntactic reflex of Restrict, the operation that interprets the following NP as a restrictive modifier without saturating the argument slot of the verb, thus necessitating existential closure at the vP-level. (20) Chung and Ladusaw (2004: 36,40), cited after Ngata (1994)
Maori
a. Kāore tētahi tangata e mahi mā-na. T.not a person T work T.of-him ‘No one would work for him.’
¬∃
b. Kāore tētahi tangata i waiata mai. T.not a person T sing to.here ‘A (particular) person didn’t sing.’
∃¬
(21) Chung and Ladusaw (2004: 41), Polinsky (1992: 237)
Maori
a. Kāore he tangata i waiata mai. T.not a person T sing to.here ‘No one at all sang.’ (Not: A particular person didn’t sing)
¬∃, *∃¬
b. Kāore he tangata i te mahi. T.not a person T work ‘Nobody is at work.’ (Not: Some men are not at work)
¬∃, *∃¬
There is an additional morpho-syntactic difference between tētahi and he (Chung and Ladusaw 2004: 28–29). Maori employs prepositions to indicate grammatical relations, except for the subject. In Standard Maori, indefinites marked with he cannot co-occur with prepositions, hence he is restricted to occur in subject position. In the Ngāti Porou dialect of Maori, he can appear in non-subject position but the respective preposition is not pronounced if it does so. Although Chung and Ladusaw (2004) do not analyse the low scope indefinites in Maori as pseudo-incorporation, the similarities with the PNI account proposed in this book are striking. It seems that Maori he spells out the PNI determiner overtly, resulting in unacceptable co-occurence with prepositions and an obligatory low scope reading. Notice that the operation Restrict together with existential closure of the individual variable deliver the same result as applying the PNI determiner, a quantifier essentially conjoining two sets and existentially binding an individual variable in their intersection. Instead of taking he to spell-out a compositional mode, I propose that he constitutes overt evidence for the PNI determiner.
9.2 DP/NP ACCOUNTS
303
9.2 DP/NP accounts DP/NP approaches operate under the assumption that arguments are free to enter the derivation as DPs or NPs. The latter will be licensed in a PNI scenario, that is in a context where the verb and the object denote a conventionalized or at least frequently occurring event. The size of the noun phrase correlates with meaning, mobility, and case. Case marking is often tied to a [+D]-feature and case assignment has been proposed to be category-sensitive (Barrie and Li 2015; Dayal 2011; Dobrovie-Sorin et al. 2006; Massam 2001; Müller 2018). Less structure has been assumed to restrict incorporated arguments to semantic objects of the property type. As a consequence, verb denotations have to be adjusted to ensure composition with a property type argument (Dayal 2003, 2011; Dobrovie-Sorin et al. 2006; van Geenhoven 1998). Another way to solve the type mismatch is by stating a separate compositional mode to combine NPs and verbs (Chung and Ladusaw 2004; Espinal and McNally 2011; McNally 2004). Mobility restrictions follow if reduced nominal structure is taken to imply no phasal status (Clemens 2019; López 2012) or no need to move into a case position (Massam 2001) or the possibility to form a complex predicate with the verb (Aydemir 2004; Frey 2015). In the following, I will present some proposals in more detail and point out for each account how it compares to the categorial theory proposed in this book, before discussing some general issues DP/NP accounts face in light of this study.¹ Frey (2015: 243) proposes that PNI-ed objects in German constitute NPs and denote properties, which he argues requires complex predicate formation with PNI verbs to ensure semantic composition. This in turn leads to a general compactness requirement, shown at length in Section 7.5. Frey (2015) presumably refers to the incorporation semantics van Geenhoven (1998) proposed for noun
¹ Another set of recently developed DOM/PNI accounts operates under the assumption that certain types of arguments have to be exceptionally licensed by establishing a syntactic (Kalin 2014, 2018; Levin 2019; Tyler 2019) or post-syntactic (Levin 2015; van Urk 2019b) dependency with another licenser in the clause. A prerequisite for such accounts is the possibility for arguments of different sizes to enter the derivation. Hence, licensing approaches can be seen as elaborate versions of DP/NP approaches. Kalin (2018: 137) e.g. proposes a split DP structure by encoding DOM prominence scales directly as separate projections. Languages vary as to which functional head introduces an uninterpretable case feature: e.g. if a language needs to mark animates, than Animateo will have an unvalued case feature feature. Since projections are privative, inanimate arguments will not need case licensing because their size will be restricted to a nominal phrase minus AnimateP. I will not engage with licensing accounts for two reasons. First, the works cited above do not investigate the languages based on scope, binding, or control properties, they rather tie case loss to definiteness/specificity. Thus, licensing approaches cannot account for the core PNI properties developed in this study. Second, there are a number of reasons to doubt an underlying syntactic dependency to be the cause for PNI/DOM effects. As was shown in Section 4.1, the missing interaction between Φ-agreement and case marking in Turkish, Korean, and Tamil as well as the lack of detransitivization effects in Mongolian, Turkish, Korean, and Adyghe are puzzling under this view; see also Driemel and Lee (2022) for discussion. As for the accounts that argue for a post-syntactic dependency, mostly driven by a requirement for surface adjacency (Levin 2015; van Urk 2019b), see Driemel and Tebay (2022) for an alternative account of the data, which is unrelated to PNI/DOM.
304
PREVIOUS APPROACHES
incorporation in Greenlandic where incorporation verbs come with a build-in existential quantifier which closes off the variable, imported by the NP property denotation. Compare the denotations below where the PNI verb in (22a) is enriched with an existential quantifier. (22) van Geenhoven (1998) a. ⟦seekinc ⟧ = λPλx∃y[SEEK(x,y)∧P(y)] b. ⟦seek⟧ = λyλx[SEEK(x,y)] One of the minimal pairs from section 7.5 is repeated here in (23). The analysis is given in (23c), suggesting that the verb cannot extract from the complex predicate, thereby violating the compactness requirement. (23) Frey (2015: 228) a.
Karten spielen wird Max heute. cards play will Max today ‘Max will play cards today.’
German
b. *Spielen wird Max heute Karten. play will Max today cards ‘Max will play cards today.’ c.
[V [VP [NP Karten] spielen] wird]
Recall that Frey is not able to explain why PNI-ed objects can excorporate, shown again in (24a). It is not possible to tie the contrast to categorial differences since verbs can also excorporate in V2 constructions; see (24b). (24) Against compactness in German PNI a. Karten wird Max heute spielen. cards will Max today play ‘Max will play cards today.’ b. Max spielt heute Karten. Max plays today cards ‘Max is playing cards today.’
(Frey 2015: 228)
V2 in PNI scenario
The contrast between (24a) and (23b) follows under the current account since mobility effects result from the verbal characteristics of PNI-ed arguments. Hence, Karten displays movement patterns like VPs in German, thus allowing for topicalization in (24a). (23b) is unacceptable because VPs do not scramble but the object has to scramble out of VP so that VP can undergo remnant VP fronting. Since this analysis does not affect verbal movement patterns, V2 is predicted to be perfectly acceptable.
9.2 DP/NP ACCOUNTS
305
Massam (2001) provides a very different account for PNI in Niuean, an Oceanic language with obligatory verb-initial word order. Consider the minimal pair in (25), where (25b) constitutes the PNI scenario. Both takafaga and ika have to be adjacent, while the object ika lost its number and case marking. Note also that the subject’s case marking switches from ergative (25a) to absolutive in (25b). (25) Massam (2001: 157) a. Takafaga tūmau nī e ia e tau ika. hunt always EMPH ERG he ABS PL fish ‘He is always fishing.’
Niuean
b. Takafaga ika tūmau nī a ia. hunt fish always EMPH ABS he ‘He is always fishing.’ Massam (2001) proposes that pseudo-incorporated arguments are NPs and therefore do not require case. Hence, they do not move into the case position which is projected in a dedicated functional layer outside of the verbal domain. The analysis is demonstrated in (26b), based on the PNI context in (26a). Since the object does not move for case, the subject merges in the specifier of AbsP instead, where it is assigned absolutive case. VP-fronting is triggered by the feature [PRED] on I. PNI-ed objects are obligatorily interpreted as non-specific indefinite, which Massam takes to be a consequence of the NP status. (26) Massam (2001: 158) a. Ne inu kofe kono a Mele. PST drink coffee bitter ABS Mele ‘Mele drank bitter coffee.’ b.
Niuean
IP I′
VPi V drink
NP
I[PRED]
AbsP
NP
AdjP
DP abs
Abs ′
coffee
bitter
Mele
K abs ⟨VP i ⟩
Note that the case switch on the subject requires not only movement into spec,AbsP but also the lack of a vP projection, i.e. no functional head that assigns ergative in non-PNI contexts. Such a scenario is sketched in (27), based on (25a).
306
PREVIOUS APPROACHES
(27)
Massam (2001: 163) IP I′
VPi V
DP
I[PRED]
hunt
⟨fish j ⟩
vP v′
DPerg he
v
AbsP DPabs
Abs ′
fish j
K abs ⟨VP i ⟩
Hence, PNI happens in Niuean as the result of a conspiracy: V merges with an NP and there is no v in the numeration that assigns ergative case. Clemens (2019) points out another version of this conspiracy with respect to instrumental arguments, which can also undergo PNI in Niuean (28a). Massam proposes that in such cases the absolutive object merges as the specifier of VP, which puts it in the right position to move to the case assignment position for absolutive case. The structure is shown in (28b). This argumentation entails that PNI of instrumental objects is possible only in case direct objects are not merged as sisters to V. (28) Clemens (2019: 9) a. Ne hele titipi e Fiti e fua vine. PST cut knife ERG Fiti ABS fruit passionfruit ‘Fiti cut the passionfruit with a knife/by knife.’ b.
Niuean
IP I′
VP i V′
DP ⟨fruit j ⟩
V
I[PRED] NP
cut knife
vP v′
DPerg Fiti
v
AbsP DPabs
Abs ′
fruit j
K abs ⟨VPi ⟩
9.2 DP/NP ACCOUNTS
307
In light of the PNI study put forth in this thesis, we cannot safely categorize Niuean as a PNI language. Case switch and detransitivization effects are a key feature of noun incorporation (Baker 1988; Chung 1978), whereas PNI languages do not provide evidence of such effects, as was shown in Section 4.1. Massam (2001: 173) argues for the NP status based on interpretive effects, i.e. PNI-ed objects are necessarily non-referential and non-specific. The obligatory presence of such properties, however, was recently questioned by Clemens (2019: 5). It is also worth pointing out that Massam (2001: 175) describes a class of PNI predicates for which PNI-ed objects can be referential after all. Given that the semantic properties are not stable enough to count as evidence for PNI and the fact that valency changes have taken place, it is reasonable to re-analyse PNI in Niuean as proper noun incorporation. The main motivation for Massam, to consider caseless nouns to be phrasal and thus pseudo-incorporated is the fact that they can be modified, shown in (26a). Modifiers, however, also occur with noun incorporation, though they are usually stranded (Allen et al. 1984). Upon closer inspection, the PNI data in Niuean which include modifiers, e.g. (26a), are entirely compatible with a stranding analysis. All in all, there is no strong motivation to consider Niuean a core example of PNI languages. Another frequently cited DP/NP account is the one put forward by Dayal (2011) for Hindi. The main characteristics are summarized in (29). Hindi objects can optionally lack case marking (29a). If they do, they receive obligatory low scope, even if they precede the scope taking operator (29b). (29) Dayal (2011: 127,137) a. Anu bacca/bacce-ko sambhaaltii hai Anu child/child-ACC look.after.IMPF be.PRS ‘Anu looks after (one or more) children/the child.’ b. Anu bacca nahiiN samhaalegii Anu child not look.after.FUT ‘Anu will not look after children.’
Hindi
¬∃, *∃¬
Dayal’s account is primarily concerned with number neutrality, a PNI property which was not taken into consideration in this book.² The analysis is sketched in (30), where PNI-ed arguments are reduced to NumPs which are excluded from case licensing, as Dayal argues. They are, however, able to scramble, e.g. across ² Except for German, every language of the core data set has been reported to show general number (Corbett 2000), i.e. a number system where case-marked arguments unmarked for number can receive a singular or a plural interpretation and overt plural marking triggers a strict plural reading; see Lehmann (1993: 20) for Tamil, Janhunen (2012: 99) for Mongolian, Kang (1994) for Korean, and Bale et al. (2011) for Turkish. Since number neutrality seems to exist independent of case marking, this PNI diagnostic was excluded from the current study. Moreover, Tamil, Turkish, and Korean allow for plural marking on PNI-ed arguments, in line with Hindi.
308
PREVIOUS APPROACHES
negation as is shown in (29b), but need to reconstruct for interpretation. The adverb in (30a) ensures an atelic interpretation, where Dayal assumes the presence of a pluractional operator (PA), which creates iterative events. NumPs are properties which, although denoting in the atomic domain, can receive a plural interpretation under the scope of PA. Singular interpretation arises if it is the same individual picked out by the predicate in each event, plural if different individuals are picked out. Number neutrality, thus, does not arise because there is no number interpretation but rather because the PA operator can produce different interpretations. (30) Dayal (2011: 145) a. Anu puure din cuuhaa pakaRtii rahii Anu whole day mouse catch.IMPF PROG ‘Anu kept catching mice (different ones) the whole day.’ b.
Hindi
IP VP
DP Anu
NumP ⟨e,t⟩ Num sg
V ⟨⟨e,t⟩,⟨e,⟨S,t⟩⟩⟩
NP
NP mouse
PA
V ⟨⟨e,t⟩,⟨e,⟨s,t⟩⟩⟩ caught
c. ⟦NumP⟧ = λx[MOUSE(x) & AT(x)] d. PA = λVλPλyλE[card(e) ≥ 2 & ∀e∀eʹ ∈ E[V(e)(y)(P) & ¬τ(e) ° τ(eʹ) ]] ⏟⎴⎴⏟⎴⎴⏟ ⏟ ⎴⎴⎴⎴⎴ ⏟ ⎴⎴⎴⎴⎴ ⏟ ⏟⎴⎴⏟⎴⎴⏟ plurality
event type
no overlap
Evidence for NumP comes from (i) the fact that PNI-ed objects in Hindi can be marked for plural (see (31)), and (ii) the observation that sentences with accomplishment readings exclude plural readings for singular marked objects which are pseudo-incorporated. The latter is tested with telic adverbs and collective predicates, i.e. predicates which necessitate plural readings; see (32). Telic readings are created by the absence of the PA operator which predicates the number interpretation should be transparent, as is proven by (32). (31)
Dayal (2011: 141) Anu botal/botaleN ikaTThaa kartii hai Anu bottle/bottles collect do.IMPF be.PRS ‘Anu collects bottles.’
Hindi
9.2 DP/NP ACCOUNTS
(32)
309
Dayal (2011: 142) Anu-ne tiin ghanTe meN *kitaab/kitaabeN ikaTThaa kar lii. Anu-ERG three hours in book/books collect do COMP.PFV ‘Anu read a book in three hours.’
Important for our discussion is that Dayal assumes PNI-ed objects to denote properties, thus requiring PNI variants for the verb denotations in PNI structures, given in (33). The denotations are similar to the ones proposed by van Geenhoven (1998), recall (22), with the difference that the PNI-ed objects serves as the modifier of the event, rather than ascribing a property to the existentially bound variable.³ (33) Dayal (2011: 146) a. ⟦catch⟧ = λxλyλe[CATCH(e) & AG(e) = y & TH(e) = x] b. ⟦catchinc ⟧ = λPλyλe[P-CATCH(e) & AG(e) = y], where ∃e[P-CATCH(e)] = 1 iff ∃eʹ[CATCH(eʹ) & ∃x[P(x) & TH(eʹ) = x] Like every DP/NP account we have discussed so far, Dayal’s theory requires two undesirable general assumptions. First, the approach in (33) enforces a significant enrichment of the lexicon. Every verb which allows for pseudo-incorporation has its usual denotation as well as a separate PNI entry. In languages which allow for subject PNI, yet another PNI entry is needed.⁴ Second, verbs do not consistently c-select for one category any more, they can select for NPs and DPs. The PNI account put forward in this thesis circumvents these issues by locating PNI syntax and semantics in one single entry, the PNI determiner. The discussion so far also highlights another shortcoming of DP/NP accounts. Depending on the proposal, PNI-ed arguments are predicted to show either no or absolute movement flexibility. PNI accounts based on complex predicate formation (Aydemir 2004; Frey 2015) and movement for case assignment (Dobrovie-Sorin et al. 2006; Massam 2001) predict the object and the verb to form a very close bond. Dayal (2011) on the other hand makes no such prediction for Hindi since she assumes case is assigned in base position. The question of whether and in how far PNI-ed arguments are able to move from their base positions seems to be rather orthogonal to the DP/NP contrast. The PNI account pursued in this book makes clear predictions concerning the dislocation properties of PNI-ed arguments in their respective languages: PNI-movement patterns with ³ There is some unclarity as to how P-V is to be understood; see Dobrovie-Sorin et al. (2006: 69) and Espinal and McNally (2011: 109) for modifications. ⁴ This feature also carries over to DP/NP accounts which additionally assume movement of the casemarked argument to a higher position, like Massam (2001) and López (2012), as the verb would still have to compose with the trace or a reduced copy, commonly assumed to be of type ⟨e⟩.
310
PREVIOUS APPROACHES
VP-movement, and not with DP-movement, PP-movement, or the distribution of adverbs. Detailed evidence was given in Chapter 7. Another potential advantage of the current account over DP/NP accounts can be found with languages like Maori (Chung and Ladusaw 2004) and St’a´t’imcets (Matthewson 1999), two languages which each show a dedicated indefinite determiner triggering obligatory low scope readings. The morpho-syntactic link to PNI is additionally reported for low scope determiner he in Maori (Chung and Ladusaw 2004: 28–29), as it is only licensed in the absence of prepositions. It is impossible for DP/NP accounts to capture the behaviour of this class of determiners with a PNI analysis since PNI-ed arguments lack determiners altogether. Finally, let me address one last issue which is often overlooked in the literature on PNI/DOM. A strict DP/NP-dichotomy predicts that case marked arguments can only take wide scope. This is, however, not always the case. Spanish is one prominent example where DOM-marked objects can take narrow and wide scope. López (2012) takes this observation to be one of the motivations to argue that DOM-marking introduces a choice function (Reinhart 1997) that enables flexible scope. We will take a closer look at his proposal in the remainder of this section, but note that Spanish is not the only DOMlanguage which shows this pattern. Flexible scope of case-marked arguments in PNI/DOM languages have been documented for Korean (Kwon and Zribi-Hertz 2006: 118), Turkish (Kelepir 2001: 59), Kannada (Lidz 2006: 14), and Hindi (Dayal 2011: 128).⁵ Let us now turn to the proposal made by López (2012) for the Spanish DOM/PNI system. The basic facts are repeated in (34) for the correlation of case loss and scope inertness and (35) for a specificity effect. (34) López (2012: 13,14) a. Todo hombre amo´ a una mujer. every man loved DOM a woman = For every man, there was a woman that he loved. = There was a woman that every man loved.
Spanish ∀>∃ ∃>∀
⁵ While this observation might simply force DP/NP accounts to adopt choice functions, the consequence for case-licensing accounts (Kalin 2014, 2018; Levin 2019; Tyler 2019) are a little bit more severe. Recall from footnote 1 that such accounts adopt a split projection structure for the nominal domain where each feature relevant for PNI/DOM is introduced by a separate privative nominal layer, e.g. AnimateP or SpecP. If it is true that SpecP is responsible for specific readings and the absence of SpecP for non-specific readings, while also taking SpecP to be the locus of case marking, it seems impossible to model a DOM language where case marking indicates ambiguity between specific and non-specific readings.
9.2 DP/NP ACCOUNTS
b. Todo hombre amo´ una mujer. every man loved a woman = For every man, there was a woman that he loved. ≠ There was a woman that every man loved.
311
∀>∃ *∃ > ∀
(35) López (2012: 16) a. Marı´a busca a una gestora. Marı´a seeks DOM a manager ‘Maria is looking for a manager.’
[±SPEC]
b. Marı´a busca una gestora. Marı´a seeks a manager ‘Maria is looking for a manager.’
[−SPEC]
López (2012: 46–48) seems to operate under the assumption that non-DOM marked indefinites and bare plurals are immobile.⁶ He analyses accusative case as the spell-out of a K head which, if present, introduces a choice function (Reinhart 1997) that enables flexible scope. He explains the movement restrictions for Spanish low-scope indefinites and bare plurals by the assumption that they do not form phases. K heads project KPs which constitute phases and thus are able to undergo scrambling. Any nominal argument smaller than a KP, i.e. DPs, #Ps, and NPs, is not a phase and denotes a property. They must remain in their base position and can only be interpreted via Restrict (Chung and Ladusaw 2004), a semantic operation which combines properties with verb denotations which apply to individual type arguments, recall Section 9.1. I illustrate the approach based on the contrast in (35). The structure for (35a) is given in (36) with the steps for the semantic composition in (37)–(40). KP moves to become a specifier of αP to check case with v, a process that is independent of the presence or absence of a KP shell but needs to happen in a local configuration, i.e. under government. αP can be understood as an equivalent to ApplP (Marantz 1993). K introduces a choice function (38), a function that maps a property onto an entity which has that property. The external argument is introduced via event identification (Kratzer 1996), shown in (39). Finally, the choice function is bound in (40). Following von Heusinger (2002), choice functions can receive an anchor, triggered by the feature [+SPEC], that ties the referent of the choice function to the individual the speaker has in mind, illustrated in (40c) and paraphrased in (40d). The feature [−SPEC] will not create an anchor; see (40a) and (40b). ⁶ He does not provide movement data to back up this empirical claim. Recall from the Spanish examples in (4–7) in Chapter 7 that movement of non-DOM marked arguments seems to be restricted to some extent though not entirely impossible.
312
PREVIOUS APPROACHES
(36)
Structure for (35a) ...
vP⟨t⟩
f2
vP⟨s,t⟩ EA ⟨e⟩
v ′⟨e,⟨s,t⟩⟩ v [acc]
BIN
DIN
αP⟨s,t⟩
⟨e,⟨s,t⟩
α ′⟨e,⟨s,t⟩⟩
KP 1⟨e⟩
G
ACC
DP ⟨e,t⟩
K 2[uC] ⟨⟨e,t⟩,e⟩
D[uC]
α′
λ1 #P
α
VP ⟨s,t⟩ V
⟨KP 1 ⟩
⟨e,⟨s,t⟩⟩
⟨e⟩
#[uC] NP [uC]
(37) Semantic composition of αP a. ⟦VP⟧ = λxλe [LOOK(x)(e)] (t1 ) = λe [LOOK(t1 )(e)] gx=1
FA
= λxλe [LOOK gx=1 (t1 ) (e)] = λxλe [LOOK (x) (e)]
PA
c. ⟦αP⟧ = λe [LOOK (f({x:MANAGER(x)})) (e)], where f = g(2)
FA
b. λx⟦αʹ⟧
(38) Semantic composition of KP a. ⟦DPindef ⟧ = λx [MANAGER(x)] b. ⟦K2 ⟧ = g(2)
if g(2) is a choice function, undefined otherwise
c. ⟦KP⟧ = g(2)({x:MANAGER(x)}) (39) Introduction of the external argument a. ⟦vʹ⟧ = λyλe [AG(y)(e) ∧ LOOK( f({x:MANAGER(x)}))(e)], where f = g(2)
EI
b. ⟦vP⟧ = λe [AG(maria)(e) ∧ LOOK( f({x:MANAGER(x)}))(e)], where f = g(2) FA c. ⟦vP⟧ = ∃e [AG(maria)(e) ∧ LOOK( f({x:MANAGER(x)}))(e)], where f = g(2) EC
9.2 DP/NP ACCOUNTS
313
(40) Binding of the choice function g f=2
g
a. ⟦∃2 vP⟧ = ∃f ⟦vP⟧ EC of CH = ∃f ∃e [AG(maria)(e) ∧ LOOK(f({x:MANAGER(x)}))(e)] b. There is some choice function f such that when you apply it to the set of managers, it gives you an entity that Maria is looking for. c. ∃f ∃e [AG(maria)(e) ∧ LOOK( fspeaker ({x:MANAGER(x)}))(e)] d. There is some choice function f such that when you apply it to the set of managers, it gives you an entity, known to the speaker, that Maria is looking for. Objects without a case (DOM) marker are, by assumption, smaller than KP: Indefinites are DPs, bare plurals #Ps. Since neither DPs nor #Ps are phases, non-case marked indefinites and bare plurals cannot move. They nevertheless have to check case, which proceeds by movement of the highest head in the nominal projection to V (which then head moves to v). In Spanish, the lowest D copy is spelled out. Obligatory low scope is achieved by Restrict in combination with existential closure immediately above VP. The schematic structure for (35b) is provided in (41), with the semantics given in (42) and (43). (41)
Structure for (35b) vP⟨t⟩ vP⟨s,t⟩ ′ v ⟨e,⟨s,t⟩⟩
EA ⟨e⟩ v ⟨e,⟨s,t⟩ V2
VP⟨s,t⟩ VP⟨e,⟨s,t⟩⟩
v [acc]
⟨V 2 ⟩⟨e,⟨s,t⟩⟩
D1[uC] V AC
C
⟨D 1⟩ ⟨V ⟩ ⟨D 1⟩
DP⟨e,t⟩ #P #[uC] NP[uC]
(42) Semantic composition of VP a. ⟦DPindef ⟧ = λx [MANAGER(x)] b. ⟦VP⟧ = λxλe [LOOK(x)(e) ∧ MANAGER(x)]
Restrict
c. ⟦VP⟧ = ∃xλe [LOOK(x)(e) ∧ MANAGER(x)]
EC
314
PREVIOUS APPROACHES
(43) Semantic composition of vP a. ⟦vʹ⟧ = λyλe [AG(y)(e) ∧ ∃x[LOOK(x)(e) ∧ MANAGER(x)]]
EI
b. ⟦vP⟧ = λe [AG(maria)(e) ∧ ∃x[LOOK(x)(e) ∧ MANAGER(x)]]
FA
c. ⟦vP⟧ = ∃e [AG(maria)(e) ∧ ∃x[LOOK(x)(e) ∧ MANAGER(x)]]
EC
Evidence for the analysis comes from binding and control. As was already pointed out in Chapter 6, only DOM-marked objects can enter binding and control relations, the relevant data is repeated below.⁷
(44) López (2012: 41) a.
Spanish
Marı´a le entrego´ [a sı´ mismo]1 [a un hombre]1 . Maria CLIT.DAT delivered DAT himself DOM a man ‘Maria delivered a man to himself.’
b. *Marı´a le entrego´ [a sı´ mismo]1 [un hombre]1 . Maria CLIT.DAT delivered DAT himself a man ‘Maria delivered a man to himself.’
(45) López (2012: 53,58) a.
Spanish
Juan forzo´ [a un nin˜o]1 [PRO1 a hacer los deberes]. Juan forced DOM a boy to do.INF the homework ‘Juan forced a boy to do his homework.’
b. *Juan forzo´ [un nin˜o]1 [PRO1 a hacer los deberes]. Juan forced a boy to do.INF the homework ‘Juan forced a boy to do his homework.’ The binding facts follow if the indirect object is introduced by αP and KP moves to an outer specifier of KP where it can bind into the indirect object in (44a). Non-KPs do not move; thus they cannot bind into an indirect object which c-commands KP in its base position. López (2012: 58) derives the control contrast by the assumption that the PNI-ed object cannot check its case feature due to the Merge requirements of the control predicate, compare (46a) to (46b). Since the object is merged in the specifier of VP, the highest nominal head cannot undergo head movement to V and then further on to v, as it is done in (41). ⁷ In (44), the dative argument presumably scrambled across the direct object, after it was bound.
9.2 DP/NP ACCOUNTS
315
(46) Control structures a. ... for DOM-marked controller in (45a) vP v′
EA
αP
v [case]
α′
CASE
KP1 α
DP
K [uC] D[uC]
VP ⟨KP1 ⟩
#P #[uC] NP [uC]
V′ V CP
b. ... for PNI-ed controller in (45b) vP v′
EA
VP
v [case]
V′
DP D[uC]
#P #[uC]
V CP
NP [uC]
One question that arises with the the control analysis in (46b) is why the D head has to undergo head movement to V and v in order to check case. The PNI-ed object is already in the right local configuration to check case, very much like KP in (36). If the D head does not have to undergo head movement, there is no reason for the structure to crash, thus the PNI-ed argument should be able to control into the CP. A related question concerns the requirement to move KP to Spec,αP. The K head could just as easily undergo head movement and check case, equivalent to how it is proposed for smaller arguments. Here, López (2012: 49) offers an
316
PREVIOUS APPROACHES
answer by assuming that the K head is contextually specified to be spelled-out as a prefix on D. This rules out head movement from K to V and only leaves the possibility for KP to Spec,αP. This step, however, requires Lo´pez to set-up a powerful post-syntactic filter that seems to be able to block syntactic movement. Another potential issue might come up with argument types higher on the prominence scale such as proper names, definites, and pronouns which obligatorily occur with DOM marking. If K introduces a choice function, these argument types would each have to constitute properties, if nothing else is said. In the context of the cross-linguistic study presented in this book, Lopez’ account can presumably not be extended to languages like Turkish, Korean, and German where caseless arguments can undergo XP movement to some extend. It is also unclear how his theory would account for subject/agent PNI, especially given that we know from Korean and Turkish that binding and control is also blocked in Spec,vP position. Since it is possible to interpret PNI-ed subjects via Restrict and the argument is in a local relation to the case assigner T, they should be able to bind and control into their c-command domain according to Lopez’ theory.
9.3 Raising accounts The theory proposed by López (2012) stands out in that it combines a DP/NP approach with a raising account. The latter is a theory that adheres DOM/PNI effects to different syntactic positions arguments occupy, where one is the base position and one is found higher, mostly outside of VP. The raised position has been taken to be the locus of case assignment (Baker 2015; Dobrovie-Sorin et al. 2006; López 2012; Merchant 2009; Rodríguez-Mondoñedo 2007; Torrego Salcedo 1999), the escape of existential closure (Diesing 1992; Kelepir 2001) or both (Bhatt 2007; Bhatt and Anagnostopoulou 1996). In this section, we will take a detailed look at the raising account proposed by ¨ ztu¨rk draws a distinction between NPs Öztürk (2005, 2009) for PNI in Turkish. O that constitute arguments, which are introduced by dedicated θ-heads, and NPs which form a complex predicate as complements to V. The former are assigned strong case under spec-head relation with the respective θ-head, while the latter are assigned weak case via downward Agree (Öztürk 2009).⁸ The discussion in Öztürk (2005: 143–146) implies that weak case does not enter the computation of morphological case assignment, hence PNI-ed arguments end up without ¨ ztu¨rk’s analysis is designed to capture object (47) and subject case marking. O pseudo-incorporation (48). ⁸ Öztürk (2005: 114) proposes that weak case is licensed via head movement.
9.3 RAISING ACCOUNTS
317
(47) Öztürk (2005: 32) a. Ali kitab-ı okudu. Ali book-ACC read ‘A li read the book.’
Turkish
b. Ali kitap okudu. Ali book read ‘A li did book reading.’ c. AgentP NP Ali
Agent ′
NOM
Agent [sCase,+ref]
ThemeP VP
Theme [wCase, −ref]
NP V book read ACC
(48) Öztürk (2005: 42) a. Arı Ali-yi soktu. bee Ali-ACC stung ‘The bee stung Ali.’ b. Ali-yi arı soktu. Ali-ACC bee stung ‘A li got bee stung.’ c.
Turkish
AgentP Agent [wCase, −ref]
ThemeP NP Ali
AC
C
Theme ′
VP V NP bee sting
Theme [sCase,+ref] NO
M
318
PREVIOUS APPROACHES
¨ ztu¨rk is able to derive the case lacking propWith the introduction of weak case, O erty on the one hand, while also predicting that PNI-ed arguments still act as if they were case marked with respect to detransitivization and causative scenarios, recall the discussion from Section 4.1. In this sense, her analysis makes the same predictions as any account that employs post-syntactic case drop. Similar to the binding and control scenarios in Spanish, Turkish PNI-ed arguments do not act as binder or controllers, as was shown in Section 6.3. Öztürk (2009) attributes these effects to the non-argument status of PNI-ed NPs. Scope inertness for PNI-ed NPs in Turkish is argued to follow from the low syntactic position (Öztürk 2009: 345). Another semantic property that she argues to derive is that caseless bare nouns do not set-up a referent which could potentially be picked up by a pronoun in further discourse; see (49a). (49) Öztürk (2009: 337) a. * Ali kitapi okudu. proi reng-i kırmızı-ydı. Ali book read color-3 red-PST ‘A li did book reading. It was red.’ b.
Ali kitap-ıi okudu. proi reng-i kırmızı-ydı. Ali book-ACC read color-3 red-PST ‘A li read the book. It was red.’
She connects non-referentiality to weak case by assuming that proper arguments are assigned strong case and [+REF] in spec,θP, whereas NPs in the complement position of V are assigned weak case and [−REF]. She draws a parallel to the weak-strong distinction made by de Hoop (1996), in that strong case acts like a type-shifter triggering referential readings (Öztürk 2005: 60).⁹ It should be noted here that de Hoop’s original weak-strong distinction is of a different kind: strong case triggers type shift to a generalized quantifier ⟨⟨e, t⟩, t⟩, while weak case is a kind of default case assigned to objects which are integrated into the verb’s denotation, either by denoting a predicate modifier ⟨⟨e, t⟩, ⟨e, t⟩⟩ or an individual ⟨e⟩ (de Hoop 1996: 113). In light of the common association of referential types as ⟨e⟩ (Partee 1986b), the connection between [+REF] and strong case seems rather misleading. Öztürk (2005: 80-81) addresses this problem when extending the analysis to case marking of Turkish indefinites, where the semantic effect is not one of referentiality but one of specificity, shown in (50). (50) Enc˛ (1991: 4–5) a. Ali bir piyano-yu kiralamak istiyor. Ali one piano-ACC to-rent wants ‘A li wants to rent a (certain) piano.’
[+SPEC]
⁹ Öztürk (2005) claims that Turkish arguments are always NPs, independent of interpretation.
9.3 RAISING ACCOUNTS
b. Ali bir piyano kiralamak istiyor. Ali one piano to-rent wants ‘A li wants to rent a piano.’
319
[−SPEC]
¨ ztu¨rk proTo capture both effects of caseless bare arguments and indefinites, O poses that the common cause for case marking is related to uniqueness, thereby making reference to a notion of specific indefinites that is based on existential quantifiers whose restrictor argument is reduced to a singleton domain (Schwarzschild 2002). Thus, strong case on bare arguments as well as indefinites signals uniqueness, leading in the former to referential readings and in the latter to specific readings. Potentially problematic for this view are universal quantifiers who are also marked for case, thus visible case marking per se cannot lead to uniqueness readings. The discourse properties of caseless arguments are currently debated and part of an active research field. Aydemir (2004) points out that the discourse properties of bare nouns and indefinites are not equally affected by lack of case marking. Caseless indefinites seem to license co-referential readings more readily than caseless bare nouns. (51) Aydemir (2004: 468) a. *Du¨n filmi seyret-ti-m, o-nui /on-lar-ıi sen de yesterday film watch-PST-1SG that-ACC/that-PL-ACC you too
b.
seyret-meli-sin. watch-MOD-2SG ‘I watched movies/did movie watching yesterday, you should watch them too.’ Du¨n [bir film]i seyret-ti-m, o-nui sen de yesterday one film watch-PST-1SG that-ACC you too seyret-meli-sin. watch-MOD-2SG ‘I watched a movie yesterday, you should watch it too.’
Öztürk (2005: 75–76) accounts for this difference by postulating conditions on the licensing of pro: it requires a referential antecedent with number specification. A recent study by Seidel (2019), however, reveals that even caseless bare nouns can serve as discourse antecedents in certain environments. We can tentatively conclude that whatever determiners the discourse properties, it is not exclusively dependent on case marking. Considerably less attention is given to the scopal behaviour of caseless bare ¨ ztu¨rk’s study. Since Öztürk (2009: 339) also acknowlnouns and indefinites in O edges that PNI-ed objects can scramble and extrapose, nothing prevents them
320
PREVIOUS APPROACHES
from taking scope in such derived positions. If weak case was tied to property denoting NPs, or predicate modifiers in the sense of de Hoop (1996), scope inertness can be argued to result from the semantic type. Crucially, however, the scope quality does not follow from positional restrictions but would have to be derived by an independent assumption. One also wonders how it is derived that case marked arguments are flexible with respect to scope and do not necessarily take wide scope, as was discussed in the last section for a number of languages, among them Turkish. If strong case necessarily introduces uniqueness/referentiality, it is unclear why case marked arguments can also scope low. Another problem of this raising approach that needs to be pointed out here is its seemingly implicit understanding of the properties that constitute an argument. While binding and control seems to be crucially tied to argumenthood, movement is not. In contrast to Öztürk (2005, 2009), the current approach to PNI ties case, scope, movement, binding, and control properties to one source: the verbal category feature of the PNI determiner. While the account predicts PNI-ed arguments to block binding and control readings, it does make a uniform prediction concerning movement. This aspect allows Turkish to differ from PNI languages such as Tamil Mongolian in exhibiting movement flexibility of PNI-ed objects.
10 Summary This book explored the pseudo-incorporation patterns across five different languages with a detailed investigation taking eleven different noun types into consideration. Tamil, Mongolian, Turkish, Korean, and German are investigated primarily via elicitation with native speaker consultants. The results are compared to the well documented pseudo-noun incorporation effects in Hindi and Spanish. The main diagnostic used to identify pseudo-noun incorporation is a correlation of case drop and a resulting narrow scope restriction of the argument. Crucially, not every noun type exhibits a correlation between lack of case marking and narrow scope. I conclude from this discrepancy that at least for Tamil, Hindi, and Korean case loss does not necessarily imply pseudo-noun incorporation. Instead, they are a reflection of differential object marking. These findings support the view that pseudo-incorporation and differential object marking must be treated as separate phenomena. I propose that the case and scope effects follow from the hybrid categorial nature pseudo-incorporated arguments display. Pseudo-incorporation effects result from a nominal category that turns into a verbal category in the course of the derivation. Case and scope facts follow accordingly: Verbs normally do not present hosts for case morphology across languages (Blake 2004; Nichols 1986), nor do they shift scope (Chomsky 2001; Harley 2004). The hybrid category status is encoded by a hybrid pseudo-noun incorporation determiner. I assume that syntactic elements are made up of ordered lists of features (Müller 2009; Stabler 1997), the set-up for the determiner is given in (1a), the denotation in (1b). (1) pseudo-noun incorporation determiner a. PNI-D: [•N• ≻ D ≻ CASE:VAL ≻ V] b. ⟦PNI-D⟧ = λP⟨e,t⟩ λQ⟨e,⟨v,t⟩⟩ λe∃z[P(z) ∧ Q(z)(e)] A feature which enters a syntactic operation is deactivated, so that the next feature on the stack is accessible. The pseudo-noun incorporation determiner first selects an NP via [•N•] forming a DP. The next feature [D] enables c-selection with the verb, where a subsequent case feature ensures that the argument is assigned case. After pseudo-incorporated arguments are merged in vP, they behave for all subsequent relations like [V]-type categories. The pseudo-noun incorporation determiner essentially denotes an existential quantifier, where the verbal nature is Pseudo-Noun Incorporation and Differential Object Marking. Imke Driemel, Oxford University Press. © Imke Driemel (2023). DOI: 10.1093/oso/9780192866400.003.0010
322
SUMMARY
reflected by the fact that the second argument constitutes a function from individuals to predicates of events ⟨e, ⟨v, t⟩⟩. This will restrict pseudo-incorporated arguments to be interpreted inside vP, the event domain, leading to a narrow scope restriction. I propose that case loss is modelled post-syntactically via impoverishment rules/constraints which target the case information by being contextually triggered via the [V]-feature on pseudo-incorporated arguments. Motivation for the post-syntactic treatment comes from the lack of any interaction of case drop and agreement as well as missing effects of grammatical function change. I integrate the pseudo-noun incorporation account into existing accounts of differential object marking (Aissen 2003; Keine 2010; Keine and Müller 2008). The proposal is supported by three additional pseudo-noun incorporation properties: (i) like verbs pseudo-incorporated arguments do not act as binders for pronouns, (ii) pseudo-incorporated arguments do not participate in control relations as this requires a binding relation with PRO, and (iii) pseudo-incorporated arguments show movement patterns parallel to VP-movement in the respective languages. Pseudo-incorporated arguments are excluded from binding relations since binding is commonly assumed to be tied to nominal categories (Baker 2004; Büring 2005; Chomsky 1981). Lack of control directly follows if control is dependent on a binding relation being established between the controller and PRO. Finally, movement properties follow straightforwardly if movement is assumed to be categorially triggered. Since pseudo-incorporated arguments are of category [V], they are considered VPs with respect to any functional head which comes with a movement inducing categorial feature. Apart from establishing a consistent set of pseudo-noun incorporation diagnostics and a coherent analysis that can capture all pseudo-noun incorporation diagnostics under one roof, I have made various additional observations which are of potential interest for future work on bare argument languages in general and on pseudo-noun incorporation and DOM in particular. I will summarize them in the following. Tamil is one of the pseudo-noun incorporation languages which has recently been argued to require surface adjacency between the noun and the verb (Baker 2014b; Levin 2015). This book provides evidence to the contrary, based on data from grammars as well as via elicitation with Tamil speakers. Where overt case marking can be detected, languages apply pseudo-noun incorporation to structural arguments generally, i.e. to direct objects and subjects. There is, however, no evidence for a reversal of prominence hierarchy effects, which serves as a counterargument against functional approaches to differential object marking (Aissen 2003; Bossong 1991; Croft 1988), i.e. theories that operate on the assumption that only arguments with definiteness/animacy properties prototypical for their grammatical function can be left unmarked. Hence, subject pseudo-noun incorporation/DOM and object pseudo-noun incorporation/DOM targets the same class of arguments.
SUMMARY
323
Pseudo-noun incorporation as well as DOM can be sensitive to the strong– weak definiteness distinction, argued for by Schwarz (2009), where weak implies uniqueness of the referent and strong includes familiarity in addition to uniqueness. Korean can pseudo-incorporate weak definites but not strong ones. Tamil reflects the strong–weak distinction in terms of differential object marking. Furthermore, strong definite heads are not necessarily realized by demonstrative pronouns in bare argument languages, as Jenks’ generalization (2015, 2018) suggests. While this is the case in Mongolian (Guntsetseg 2016), the current study provides counter-evidence from Tamil, Turkish, and Korean where strong and weak definites are realized with bare arguments. Strong definites do, however, require case marking. Contra Ahn (2017), I analyse ku as the spell-out of an ident-type shifter (Partee 1986a) which allows it to occur in strong as well as weak definite contexts, in line with the observations in Kang (2015, 2018) and Ahn (2019). The presence of the type-shifter, furthermore, enables weak definites, proper names and 1st/2nd person pronouns to undergo pseudo-noun incorporation. Moreover, we provide a variety of arguments to consider numeral classifiers in Korean as determiner heads. A strong weak distinction can also be drawn within the class of quantifiers, where numerals pattern with weak quantifiers. While this has already been known for Turkish (Enc˛ 1991), I extend the observation to Mongolian. I explain the split by treating weak quantifiers and numerals as adjectival modifiers (Link 1987; Milsark 1977; Partee 1988; Zimmermann 2008)—for Mongolian exclusively and for German and Turkish as an additional option. Another issue I investigate is the treatment of the indefinite determiner and its possible double function as a numeral for each pseudo-noun incorporation language within the core data set. Dayal (2004, 2017) argues that the Hindi determiner ek is exclusively a numeral; I draw the same conclusion for Tamil oru and Mongolian neg. Dayal (2004, 2017), furthermore, claims that Hindi bare nouns do not receive existential readings; they shift to definite readings automatically. This is not true for Tamil, Mongolian, and Korean, which I demonstrate with a variety of tests.
References Abney, S. P. (1987). The English Noun Phrase in its Sentential Aspect. PhD thesis, MIT. Ahn, D. (2017). Definite and demonstrative descriptions: A micro-typology. In Erlewine, M. Y., editor, Proceedings of GLOW in Asia XI, pages 33–48. MIT Working Papers in Linguistics, Cambridge, MA. Ahn, D. (2018). Korean classifier-less number constructions. In Sauerland, U. and Solt, S., editors, Proceedings of Sinn und Bedeutung 22, pages 23–38. ZASPiL 60, ZAS, Berlin. Ahn, D. (2019). THAT Thesis: A Competition Mechanism for Anaphoric Expressions. PhD thesis, Harvard University. Ahn, H.-D. and Yoon, H.-J. (1989). Functional categories in Korean. In Kuno, S., Whitman, J., Kang, Y.-S., Lee, I.-H., Bak, S.-Y., and Kim, Y.-J., editors, Harvard Studies in Korean Linguistics III, pages 79–88. Hanshin, Seoul. Ahn, S.-H. (1990). Korean Quantification and Universal Grammar. PhD thesis, University of Connecticut, Storrs. Aissen, J. (1999). Markedness and subject choice in optimality theory. Natural Language and Linguistic Theory, 17:673–711. Aissen, J. (2003). Differential object marking: Iconicity vs. economy. Natural Language and Linguistic Theory, 21:435–483. Akan, T. (2009). On Scrambling in Turkish. PhD thesis, Boğazıc¸ı University. Alexiadou, A. (2001). Functional Structure in Nominals. John Benjamins, Philadelphia/Amsterdam. Alexiadou, A. (2017). On the complex relationship between deverbal compounds and argument supporting nominals. In Bloch-Trojnar, M. and Malicka-Kleparska, A., editors, Aspect and Valency in Nominals, pages 53–82. De Gruyter Mouton, Berlin. Alexiadou, A. (2020). D vs. n nominalizations within and across languages. In Alexiadou, A. and Borer, H., editors, Nominalization: 50 Years on from Chomsky’s Remarks, pages 87–109. Oxford University Press, Oxford. Alexiadou, A. and Anagnostopoulou, E. (1998). Parametrizing Agr: Word order, V-movement and EPP-checking. Natural Language and Linguistic Theory, 16:491–539. Alexiadou, A., Anagnostopoulou, E., and Schäfer, F. (2006). The properties of anticausatives crosslinguistically. In Frascarelli, M., editor, Phases of Interpretation, pages 187–211. Mouton de Gruyter, Berlin. Allen, B., Gardiner, D., and Frantz, D. (1984). Noun incorporation in Southern Tiwa. International Journal of American Linguistics, 50:292–311. Anand, P. and Nevins, A. (2006). The locus of ergative case assignment. In Karimi, S., editor, Ergativity: Emerging Issues, Studies in Natural Language and Linguistic Theory, pages 3–26. Springer, Dordrecht. Annamalai, E. and Steever, S. B. (1998). Modern Tamil. In Steever, S. B., editor, The Dravidian Languages, pages 100–128. Routledge, London/New York. Asarina, A. (2011). Case in Uyghur and Beyond. PhD thesis, MIT. Assmann, A., Edygarova, S., Georgi, D., Klein, T., and Weisser, P. (2014). Case stacking below the surface: On the possessor case alternation in Udmurt. The Linguistic Review, 31:447–485.
REFERENCES
325
Assmann, A., Georgi, D., Heck, F., Müller, G., and Weisser, P. (2015). Ergatives move too early. Syntax, 18:343–387. Asudeh, A. and Mikkelsen, L. H. (2000). Incorporation in Danish: Implications for interfaces. In Cann, R., Grover, C., and Miller, P., editors, A Collection of Papers on Headdriven Phrase Structure Grammar, pages 1–15. CSLI Publications, Stanford. Aydemir, Y. (2004). Are Turkish preverbal bare nouns syntactic arguments? Linguistic Inquiry, 35:465–474. Aygen, G. (2007a). Specificity and subject-object positions / scope interactions in Turkish. Dil ve Edebiyat Dergisi/Journal of Linguistics and Literature, 4:11–43. Aygen, G. (2007b). Syntax and semantics of genitive subject-case in Turkic. California Linguistic Notes. Volume XXXII No. 2 Spring. Bailyn, J. (2003). A (purely) derivational account of Russian scrambling. In Browne, W., Kim, J.-Y., Partee, B., and Rothstein, R., editors, Proceedings of Formal Approaches to Slavic Linguistics 11, pages 41–62, Ann Arbor, MI. University of Michigan Press. Baker, M. (1985). The Mirror principle and morphosyntactic explanation. Linguistic Inquiry, 16:373–415. Baker, M. (1988). Incorporation: A Theory of Grammatical Function Changing. University of Chicago Press, Chicago. Baker, M. (1995). Lexical and nonlexical noun incorporation. In Egli, U., Pause, P., Schwarze, C., von Stechow, A., and Wienold, G., editors, Lexical Knowledge in the Organization of Language, page 3–34. John Benjamins, Amsterdam. Baker, M. (1996). The Polysynthesis Parameter. Oxford University Press, Oxford. Baker, M. (2004). Lexical Categories: Verbs, Nouns and Adjectives. Cambridge University Press, Cambridge. Baker, M. (2009). Is head movement still needed for noun incorporation? Lingua, 119: 148–165. Baker, M. (2014a). On dependent ergative case (in Shipibo) and its derivation by phase. Linguistic Inquiry, 45:341–379. Baker, M. (2014b). Pseudo noun incorporation as covert noun incorporation: Linearization and crosslinguistic variation. Language and Linguistics, 15:5–46. Baker, M. (2015). Case. Its Principles and Parameters. Cambridge University Press, Cambridge. Baker, M., Aranovich, R., and Golluscio, L. (2005). Two types of syntactic noun incorporation: Noun incorporation in Mapudungun and itstypological implications. Language, 81:138–176. Baker, M. and Vinokurova, N. (2010). Two modalities of case assignment in Sakha. Natural Language and Linguistic Theory, 28:593–642. Bale, A., Khanjian, H., and Gagnon, M. (2011). Cross-linguistic representations of numerals and number marking. In Li, N. and Lutz, D., editors, Proceedings of SALT 20, page 582–598. LSA and CLC Publications. Bárány, A. (2017). Person, Case, and Agreement: The Morphosyntax of Inverse Agreement and Global Case Splits. Oxford University Press, Oxford. Barrie, M. and Li, A. (2015). The semantics of (pseudo) incorporation and case. In Borik, O. and Gehrke, B., editors, The Syntax and Semantics of Pseudo-Incorporation, pages 159–188. Brill, Leiden/Boston. Barwise, J. and Cooper, R. (1981). Generalized quantifiers and natural language. Linguistics and Philosophy, 4:159–219. Bayer, J. and Brandner, E. (2013). On Wh-head-movement and the doubly filled comp filter. In Chang, C. B. and Haynie, H. J., editors, Proceedings of the 26th West Coast Conference on Formal Linguistics, pages 87–95. Cascadilla Proceedings Project, Somerville, MA.
326
REFERENCES
Beck, S. (1996). Wh-Constructions and Transparent Logical Form. PhD thesis, Universität Tübingen. Bhatt, R. (2007). Unaccusativity and case licensing. Talk presented at McGill University. Bhatt, R. (2016). Minimalist approaches to South Asian syntax. In Hock, H. H. and Bashir, E., editors, The Languages and Linguistics of South Asia, pages 506–529. De Gruyter Mouton, Berlin/Boston. Bhatt, R. and Anagnostopoulou, E. (1996). Object shift and specificity: Evidence from kophrases in Hindi. In Dobrin, L., Singer, K., and McNair, L., editors, Papers from the 32nd Regional Meeting of the Chicago Linguistic Society, pages 11–22. Chicago Linguistic Society, Chicago. Bhattacharya, T. and Devi, T. (2004). Why cleft? Proceedings of SALA XXIII., lingbuzz/000457. Bierwisch, M. (1967). Syntactic Features in Morphology: General Problems of So-Called Pronominal Inflection in German. In To Honour Roman Jakobson, pages 239–270. Mouton, The Hague. Binnick, R. I. (1979). Modern Mongolian. A Transformational Grammar. University of Toronto Press, Toronto/Buffalo/London. Blake, B. (2001/2004). Case. Cambridge University Press, Cambridge. Bleam, T. (2000). Leí sta Spanish and the Syntax of Clitic Doubling. PhD thesis, University of Delaware. Bobaljik, J. (2012). Universals in Comparative Morphology. Suppletion, Superlatives, and the Structure of Words. MIT Press, Cambridge. Bobaljik, J. and Harley, H. (2017). Suppletion is local. Evidence from Hiaki. In Newell, H., Noonan, M., Piggot, G., and Travis, L., editors, The Structure of Words at the Interfaces, pages 141–159. Oxford University Press, Oxford. Boeckx, C. and Stjepanović, S. (2001). Head-ing toward PF. Linguistic Inquiry, 32:345–355. Bonet, E. (1991). Morphology after Syntax. PhD thesis, MIT. Borer, H. (1989). Anaphoric AGR. In Jaeggli, O. and Safir, K. J., editors, The Null Subject Parameter, pages 69–109. Kluwer Academic Publishers, Dordrecht. Borer, H. (2003). Exo-skeletal vs. endo-skeletal explanations: Syntactic projections and the lexicon. In Moore, J. and Polinsky, M., editors, The Nature of Explanation in Linguistic Theory, pages 31–67. CSLI Publications, Stanford. Borer, H. (2005a). Structuring Sense, vol. 1: In Name Only. Oxford University Press, Oxford. Borer, H. (2005b). Structuring Sense, vol. 2: The Normal Course of Events. Oxford University Press, Oxford. Borer, H. (2005c). Structuring Sense, vol. I: In Name Only. Oxford University Press, Oxford. Borsley, R. and Kornfilt, J. (2000). Mixed extended projection. In Borsley, R., editor, The Nature and Function of Syntactic Categories, pages 101–131. Academic Press, San Diego. Bossong, G. (1991). Differential object marking in romance and beyond. In Kibbee, D. A. and Wanner, D., editors, New Analyses in Romance Linguistics, pages 143–170. John Benjamins, Amsterdam. Bošković, Željko. and Şener, S. (2014). The Turkish NP. In Hofherr, P. C. and ZribiHertz, A., editors, Crosslinguistic Studies on Noun Phrase Structure and Reference, pages 102–140. Brill, Leiden/Boston. Bruening, B. (2010). Ditransitive asymmetries and a theory of idiom formation. Linguistic Inquiry, 41:519–562. Bruening, B. (2018a). Depictive secondary predicates and small clause approaches to argument structure. Linguistic Inquiry, 49:537–559. Bruening, B. (2018b). Double object constructions and prepositional dative constructions are distinct: A reply to Ormazabal and Romero 2012. Linguistic Inquiry, 49:123–150.
REFERENCES
327
Bruening, B., Dinh, X., and Kim, L. (2018). Selection, idioms, and the structure of nominal phrases with and without classifiers. Glossa, 3(42):1–46. Burge, T. (1973). Reference and proper names. The Journal of Philosophy, 70:425–439. Büring, D. (2001). A situation semantics for binding out of DP. In Hastings, R., Jackson, B., and Zvolenski, Z., editors, Proceedings from Semantics and Linguistic Theory XI, pages 56–75, Ithaca. CLC. Büring, D. (2004). Crossover situations. Natural Language Semantics, 12(1):23–62. Büring, D. (2005). Binding Theory. Cambridge University Press, Cambridge. Butt, M. (1993). Object specificity and agreement in Hindi/Urdu. In Beals, K., Cooke, G., Kathman, D., Kita, S., McCullough, K.-E., and Testen, D., editors, Papers from the 29th Regional Meeting of the Chicago Linguistics Society, pages 89–103. Chicago Linguistics Society, Chicago. Butt, M. and King, T. (2004). The status of case. In Dayal, V. and Mahajan, A., editors, Clause Structure in South Asian Languages, pages 153–198. Kluwer, Boston. Carlson, G. (1977). Reference to Kinds in English. PhD thesis, University of Massachusetts, Amherst. Carlson, G. (2003). Weak indefinites. In Coene, M. and D’hulst, Y., editors, From NP to DP, pages 354–366. John Benjamins, Amsterdam/Philadelphia. Champollion, L. (2011). Quantification and negation in event semantics. In Partee, B., Glanzberg, M., and Skilters, J., editors, Baltic International Yearbook of Cognition, Logic and Communication, volume 6. New Prairie Press, Manhattan, KS. Champollion, L. (2015). The interaction of compositional semantics and event semantics. Linguistics and Philosophy, 38:31–66. Champollion, L. (2016a). Covert distributivity in algebraic event semantics. Semantics and Pragmatics, 9:15:1–65. Champollion, L. (2016b). Overt distributivity in algebraic event semantics. Semantics and Pragmatics, 9:16:1–65. Cheng, L. L.-S., Heycock, C., and Zamparelli, R. (2017). Two levels for definiteness. In Erlewine, M. Y., editor, Proceedings of GLOW in Asia XI, pages 79–93. MIT Working Papers in Linguistics, Cambridge, MA. Cheng, L. L.-S. and Sybesma, R. (1999). Bare and not-so-bare nouns and the structure of NP. Linguistic Inquiry, 30:509–542. Chierchia, G. (1984). Topics in the Syntax and Semantics of Infinitives and Gerunds. PhD thesis, University of Massachusetts Amherst. Chierchia, G. (1989). Structured meanings, thematic roles and control. In Chierchia, G., Partee, B., and Turner, R., editors, Properties, Types and Meanings II, pages 131–166. Kluwer Academic Publishers, Dordrecht. Chierchia, G. (1990). Anaphora and attitudes de se. In Bartsch, R., van Benthem, J., and van Emde Boas, P., editors, Semantics and Contextual Expression, pages 113–150. Foris, Dordrecht. Chierchia, G. (1995). Individual-level predicates as inherent generics. In Carlson, G. and Pelletier, F. J., editors, The Generic Book, pages 176–223. University of Chicago Press, Chicago. Chierchia, G. (1998). Reference to kinds across languages. Natural Language Semantics, 6:339–405. Cho, J. (2017). The acquisition of different types of definite noun phrases in L2-English. International Journal of Bilingualism, 21:367–382. Cho, J.-H. (1994). On scrambling. reconstruction, crossover, and anaphor binding. In KimRenaud, Y.-K., editor, Theoretical Issues in Korean Linguistics, pages 255–273. University of Chicago Press, Chicago.
328
REFERENCES
Choi-Jonin, I. (2008). Particles and postpositions in Korean. In Kurzon, D. and Adler, S., editors, Adpositions: Pragmatic, Semantic and Syntactic Perspectives, page 133–170. John Benjamins, Amsterdam/Philadelphia. Chomsky, N. (1973). Conditions on transformations. In Anderson, S. R. and Kiparsky, P., editors, A Festschrift for Morris Halle, pages 232–286. Holt, New York. Chomsky, N. (1980). On binding. Linguistic Inquiry, 11:1–46. Chomsky, N. (1981). Lectures on Government and Binding. de Gruyter, Berlin. Chomsky, N. (1986). Barriers. MIT Press, Cambridge. Chomsky, N. (1993). A minimalist program for linguistic theory. In Hale, K. and Keyser, S. J., editors, The View from Building 20: Essays in Linguistics in Honor of Sylvain Bromberger, pages 1–25. MIT Press, Cambridge. Chomsky, N. (1995). The Minimalist Program. The MIT Press, Cambridge, Massachusetts. Chomsky, N. (2000). Minimalist inquiries: The framework. In Martin, R., Michaels, D., and Uriagereka, J., editors, Step by Step: Essays on Minimalist Syntax in Honor of Howard Lasnik, pages 89–155. MIT Press, Cambridge, Massachusetts. Chomsky, N. (2001). Derivation by phase. In Kenstowicz, M., editor, Ken Hale: A Life in Language, pages 1–52. MIT Press, Cambridge, Massachusetts. Chung, I. (2009). Suppletive verbal morphology in Korean and the mechanism of vocabulary insertion. Journal of Linguistics, 45:533–567. Chung, S. (1978). Case Marking and Grammatical Relations in Polynesian. University of Texas Press, Austin, TX. Chung, S. and Ladusaw, W. A. (2004). Restriction and Saturation, volume 42 of Linguistic Inquiry Monograph. MIT Press, Cambridge, MA. Clemens, L. (2019). Prosodic noun incorporation: The relationship between prosody and argument structure in Niuean. Syntax, 22:337–377. Collins, J. N. (2017). Samoan predicate initial word order and object positions. Natural Language and Linguistic Theory, 35:1–59. Comrie, B. (1989). Language Universals and Linguistic Typology. Blackwell, Oxford. Coon, J. (2010). VOS as predicate fronting in Chol. Lingua, 120:354–378. Coon, J. and Preminger, O. (2013). Transivitiy in Chol: A new argument for the split VP hypothesis. In Fainleib, L., LaCara, N., and Park, Y., editors, Proceedings of the 41st Annual Meeting of the North East Linguistics Society, pages 127–142. GLSA Publications, Amherst. Coppock, E. and Champollion, L. (2017). Formal Semantics Bootcamp. Manuscript, 3 Sept 2018. Corbett, G. G. (2000). Number. Oxford University Press, Oxford. Cresti, D. (1995). Extraction and reconstruction. Natural Language Semantics, 3:79–122. Croft, W. (1988). Agreement vs. case marking and direct Objects. In Barlow, M. and Ferguson, C. A., editors, Agreement in Natural Language: Approaches, Theories, Descriptions, page 159–180. CSLI, Stanford. Croft, W. (1990). Typology and Universals. Cambridge University Press, Cambridge. Culicover, P. and Wexler, K. (1977). Some syntactic implications of a theory of language learnability. In Culicover, P., Wasow, T., and Akmajian, A., editors, Formal Syntax, pages 7–61. Academic Press, New York/San Francisco/London. Dalrymple, M. and Nikolaeva, I. (2011). Objects and Information Structure. Cambridge University Press, Cambridge. Danon, G. (2011). Agreement and DP-internal feature distribution. Syntax, 14:297–317. Davidson, D. (1967). The logical form of action sentences. In Resher, N., editor, The Logic of Decision and Action, pages 81–95. University of Pittsburgh Press, Pittsburgh.
REFERENCES
329
Dayal, V. (1999). Bare NP’s, reference to kinds, and incorporation. In Matthews, T. and Strolovitch, D., editors, Proceedings of SALT 19, pages 34–51. Cornell University, Ithaca, NY. Dayal, V. (2003). A semantics for pseudo-incorporation. Rutgers University. Dayal, V. (2004). Number marking and (in)definiteness in kind terms. Linguistics and Philosophy, 27:393–450. Dayal, V. (2011). Hindi pseudo-incorporation. Natural Language and Linguistic Theory, 29:123–167. Dayal, V. (2017). Determining (in)definiteness in the absence of articles. In Hohaus, V. and Rothe, W., editors, Proceedings of Triple A 3, pages 85–99. Universitätsbibliothek Tübingen. Deal, A. R. (2015). Interaction and satisfaction in ϕ-agreement. In Bui, T. and Ozyildiz, D., editors, Proceedings of NELS 45, volume 1, pages 179–192, Amherst. GLSA. Dede, M. (1986). Definiteness and referentiality in Turkish verbal sentences. In Karimi, S., editor, Studies in Turkish Linguistics, pages 147–165. John Benjamins, Amsterdam. Diesing, M. (1992). Indefinites. MIT Press, Cambridge. Dobrovie-Sorin, C., Bleam, T., and Espinal, M. T. (2006). Bare nouns, number and types of incorporation. In Vogeleer, S. and Tasmowski, L., editors, Non-definiteness and Plurality, pages 51–79. John Benjamins, Amsterdam/Philadelphia. Driemel, I. (2020). Pseudo-incorporation and its movement patterns. Glossa: A Journal of General Linguistics, 5:106. Driemel, I. and Lee, H. (2022). Pseudo-incorporation vs. differential argument marking in Korean. In Du, Y., Chen, Z., Li, X., Xu, Z., Pan, V.J., editors, Proceedings of GLOW in Asia XIII., pages 75–90, The Chinese University of Hongkong, Honkong. Driemel, I. and Tebay, S. E. (2022). Determiner removal in Balinese non-pivot agents. Syntax, 25:276–298. Elbourne, P. (2002). Situations and Individuals. PhD thesis, MIT. Elbourne, P. (2005). Situations and Individuals. Cambridge University Press, Cambridge. Elbourne, P. (2013). Definite Descriptions. Oxford University Press, Oxford. Embick, D. and Marantz, A. (2008). Architecture and blocking. Linguistic Inquiry, 39:1–53. Embick, D. and Noyer, R. (2001). Movement operations after syntax. Linguistic Inquiry, 32:555–595. Embick, D. and Noyer, R. (2007). Distributed morphology and the syntax-morphology interface. In Ramchand, G. and Reiss, C., editors, The Oxford Handbook of Linguistic Interfaces, pages 289–324. Oxford University Press, Oxford. Enc˛, M. (1991). The semantics of specificity. Linguistic Inquiry, 22:1–25. Epstein, S., Groat, E., Kawashima, R., and Kitahara, H. (1998). A Derivational Approach to Syntactic Relations. Oxford University Press, Oxford. Erguvanlı, E. (1984). The Function of Word Order in Turkish Grammar. University of California Press, Berkeley/Los Angeles/London. Ernst, T. (2002). The Syntax of Adjuncts. Cambridge University Press, Cambridge. Espinal, T. and McNally, L. (2011). Bare nominals and incorporating verbs in Spanish and Catalan. Journal of Linguistics, 47:87–128. Fábregas, A. (2013). Differential object marking in Spanish: State of the art. Borealis, 2:1–80. Fanselow, G. (1990). On the nature of proper government and syntactic barriers. In Abraham, W., Kosmeijer, W., and Reuland, E., editors, Issues in Germanic Syntax, pages 33–48. Mouton de Gruyter, Berlin. Fanselow, G. (1992). Deplazierte Argumente. Ms., Universität Stuttgart.
330
REFERENCES
Fanselow, G. (2002a). Münchausen-style head movement and the analysis of verb second. Manuscript, Universität Potsdam. Fanselow, G. (2002b). Quirky subjects and other specifiers. In Kaufmann, I. and Stiebels, B., editors, More than Words, page 227–250. Akademie Verlag, Berlin. Farkas, D. and de Swart, H. (2003). The Semantics of Incorporation: From Argument Structure to Discourse Transparency. CSLI Publications, Stanford. Féry, C. (1993). German Intonational Patterns. Max Niemeyer Verlag, Tübingen. Fintel, K. v. (1994). Restrictions on Quantifier Domains. PhD thesis, University of Massachusetts, Amherst. Fong, S. (2019). Proper movement through Spec-CP: An argument from hyperraising in Mongolian. Glossa, 4(30):1–42. Fox, D. (1999). Reconstruction, binding theory, and the interpretation of chains. Linguistic Inquiry, 30:157–196. Fox, D. (2002). Antecedent-contained deletion and the copy theory of movement. Linguistic Inquiry, 33:63–96. Frege, G. (1892). Über Sinn und Bedeutung. In Patzig, G., editor, Funktion, Begriff, Bedeutung, pages 40–65. Vandenhoeck and Ruprecht, Göttingen. Frey, W. (1993). Syntaktische Bedingungen für die semantische Repräsentation: Über Bindung, implizite Argumente und Skopus. Akademie Verlag, Berlin. Frey, W. (2006). Contrast and movement to the German prefield. In Molnar, V. and Winkler, S., editors, The Architecture of Focus, pages 235–264. Mouton de Gruyter, Berlin. Frey, W. (2015). NP-incorporation in German. In Borik, O. and Gehrke, B., editors, The Syntax and Semantics of Pseudo-Incorporation, pages 225–261. Brill, Leiden/Boston. Frey, W. and Tappe, T. (1991). Zur Interpretation der X-bar-Theorie und zur Syntax desMittelfeldes. Grundlagen eines GB-Fragmentes. Ms., Universität Stuttgart. Galdeano, R. B. (2000). Differential Object Marking in Catalan: Contexts of Appearance and Analysis. PhD thesis, Universitat Autònoma de Barcelona. Gambhir, V. (1981). Syntactic Restrictions and Discourse Functions of Word Order in Standard Hindi. PhD thesis, University of Pennsylvania. García García, M. (2018). Nominal and verbal parameters in the diachrony of differential object marking in Spanish. In Serzˇant, I. A. and Witzlack-Makarevich, A., editors, The Diachronic Typology of Differential Argument Marking, pages 209–242. Language Science Press, Berlin. Gebhardt, L. (2011). Classifiers are functional. Linguistic Inquiry, 42:125–130. van Geenhoven, V. (1998). Semantic Incorporation and Indefinite Descriptions. CSLI, Palo Alto. Gehrke, B. (2015). Adjectival participles, event kind modification and pseudoincorporation. Natural Language and Linguistic Theory, 33:897–938. Georgi, D. (2014). Opaque Interactions of Merge and Agree: On the Nature and Order of Elementary Operations. PhD thesis, Leipzig University, Leipzig. Georgi, D. and Müller, G. (2010). Noun phrase structure by reprojection. Syntax, 13:1–36. Georgi, D. and Salzmann, M. (2011). DP-internal double agreement is not double Agree: Consequences of Agree-based case assignment within DP. Lingua, 121:2069–2088. Gerstner-Link, C. (1995). Über Generizität: Generische Nominalphrasen in singulären Aussagen und generischen Aussagen. Wilhelm Fink Verlag, München. Göksel, A. and Kerslake, C. (2005). Turkish: A Comprehensive Grammar. Routledge, London. Goldberg, L. M. (2005). Verb-Stranding VP Ellipsis: A Cross-Linguistic Study. PhD thesis, McGill University.
REFERENCES
331
Gračanin-Yüksek, M. and İşsever, S. (2011). Movement of bare objects in Turkish. Dilbilim Araştırmaları, 22:33–49. Green, J. J. (2018). Adjunct Control: Syntax and Processing. PhD thesis, University of Maryland. Grewendorf, G. (1995). German. Syntactic sketches. In Jacobs, J., von Stechow, A., Sternefeld, W., and Vennemann, T., editors, Handbooks of Linguistics and Communication Science, volume 9, pages 1288–1319. De Gruyter Mouton, Berlin/Boston. Grewendorf, G. and Sternefeld, W. (1990). Scrambling theories. In Grewendorf, G. and Sternefeld, W., editors, Scrambling and Barriers, pages 3–37. John Benjamins, Amsterdam/Philadelphia. Gribanova, V. (2013). Verb-stranding verb phrase ellipsis and the structure of the Russian verbal complex. Natural Language and Linguistic Theory, 31:91–136. Guntsetseg, D. (2008). Differential object marking in Mongolian. In Schäfer, F., editor, Proceedings of the 5th Workshop on Altaic Formal Linguistics, pages 53–69. Working Papers of the SFB 732 Incremental Specification in Context. Guntsetseg, D. (2009). Differential object marking in (Khalkha-)Mongolian. In Shibagaki, R. and Vermeulen, R., editors, Proceedings of the 5th Workshop on Altaic Formal Linguistics, pages 115–129. MIT Working Papers in Linguistics. Guntsetseg, D. (2010). The function of accusative case in Mongolian. In Maezawa, H. and Yokogoshi, A., editors, Proceedings of the 6th Workshop on Altaic Formal Linguistics. MIT Working Papers in Linguistics. Guntsetseg, D. (2016). Differential Case Marking in Mongolian. Harrassowitz Verlag, Wiesbaden. Guntsetseg, D. and Klein, U. (2009). Two types of case alternation in Mongolian. In Current Issues in Unity and Diversity of Languages: Collection of the Papers Selected from the CIL 18, pages 3114–3128. Linguistic Society of Korea. Haider, H. (2010). The Syntax of German. Cambridge University Press, Cambridge. Halle, M. (1997). Distributed morphology: Impoverishment and fission. In Bruening, B., Kang, Y., and McGinnis, M., editors, Papers at the Interface, volume 30 of MIT Working Papers in Linguistics, pages 425–449. Academic Press, New York. Halle, M. and Marantz, A. (1993). Distributed morphology and the pieces of inflection. In Hale, K. and Keyser, S. J., editors, The View from Building 20, pages 111–176. MIT Press, Cambridge. Halle, M. and Marantz, A. (1994). Some key features of distributed morphology. In Carnie, A., Harley, H., and Bures, T., editors, Papers on Phonology and Morphology, volume 21 of MIT Working Papers in Linguistics, pages 275–288. Academic Press. Halpert, C. (2012). Argument Licensing and Agreement in Zulu. PhD thesis, MIT. Han, C.-h., Lidz, J., and Musolino, J. (2007). V-raising and grammar competition in Korean: Evidence from negation and quantifier scope. Linguistic Inquiry, 38:1–47. Hanink, E. (2018). Structural Sources of Anaphora and Sameness. PhD thesis, University of Chicago. Harley, H. (2004). Merge, conflation, and head movement: The First Sister Principle revisited. In Moulton, K. and Wolf, M., editors, Proceedings of NELS 34, pages 239–254, Amherst. GLSA. Harley, H. (2009). The morphology of nominalizations and the syntax of vp. In Giannakidou, A. and Rathert, M., editors, Quantification, Definiteness, and Nominalization, pages 321–343. Oxford University Press, Oxford.
332
REFERENCES
Harley, H. (2012). Semantics in distributed morphology. In Maienborn, C., Heusinger, K. v., and Portner, P., editors, Semantics, volume 3 of Handbooks of Linguistics and Communication Science, pages 2151–2172. Mouton de Gruyter, Berlin. Harley, H. (2013). External arguments and the Mirror Principle: On the distinctness of Voice and v. Lingua, 125:34–57. Harley, H. (2015). The syntax-morphology interface. In Kiss, T. and Alexiadou, A., editors, Syntax, Theory and Analysis: An International Handbook, volume II, pages 1128–1153. Mouton de Gruyter, Berlin. Harley, H. and Noyer, R. (1999). Distributed morphology. Glot International, 4:3–9. Harley, H. and Noyer, R. (2000). Formal versus encyclopedic properties of vocabulary: Evidence from nominalisations. In Peeters, B., editor, The Lexicon-Encyclopedia Interface, pages 349–374. Elsevier, Oxford. Hawkins, J. A. (1978). Definiteness and Indefiniteness. Croom Helm, London. Hawkins, J. A. (1999). Processing complexity and filler-gap dependencies across grammars. Language, 75:244–285. Heck, F. and Himmelreich, A. (2017). Opaque intervention. Linguistic Inquiry, 48:47–97. Heck, F. and Müller, G. (2003). Derivational optimization of Wh-movement. Linguistic Analysis, 33:97–148. Volume appeared 2007. Heck, F. and Müller, G. (2007). Extremely local optimization. In Brainbridge, E. and Agbayani, B., editors, Proceedings of WECOL 26, pages 170–183, Fresno. California State University. Heck, F. and Richards, M. (2010). A probe-goal approach to agreement and nonincorporation restrictions in Southern Tiwa. Natural Language Linguistic Theory, 28:681–721. Heim, I. (1982). The Semantics of Definite and Indefinite Noun Phrases. PhD thesis, University of Massachusetts. Heim, I. (1991). Artikel und Definitheit. In von Stechow, A. and Wunderlich, D., editors, Semantik: Ein internationales Handbuch der zeitgenössischen Forschung, pages 487–535, Berlin. de Gruyter. Heim, I. (2008). Features on bound pronouns. In Harbour, D., Adger, D., and Béjar, S., editors, Phi-Theory. Phi-features across Modules and Interfaces, pages 35–56. Oxford University Press, Oxford. Heim, I. and Kratzer, A. (1998). Semantics in Generative Grammar. Blackwell, Oxford. Herburger, E. (2002). The negative concord puzzle revisited. Natural Language Semantics, 9:289–333. von Heusinger, K. (2002). Specificity and definiteness in sentence and discourse structure. Journal of Semantics, 19:245–274. von Heusinger, K. (2008). Verbal semantics and the diachronic development of DOM in Spanish. Probus, 20:1–31. von Heusinger, K. (2018). The diachronic development of differential object marking in Spanish ditransitive constructions. In Serzˇant, I. A. and Witzlack-Makarevich, A., editors, Diachrony of Differential Argument Marking, pages 315–344. Language Science Press, Berlin. von Heusinger, K. and Bamyacı, E. (2016). Specificity effects of Turkish differential object marking. In Zidani-Eroğlu, L., Ciscel, M., and Koulidobrova, E., editors, Proceedings of the 12th Workshop on Altaic Formal Linguistics, pages 309–319. MIT Working Papers in Linguistics. von Heusinger, K. and Kaiser, G. A. (2007). Differential object marking and the lexical semantics of verbs in Spanish. In Kaiser, G. A. and Leonetti, M., editors, Proceedings of
REFERENCES
333
the Workshop ‘Definiteness, Specificity and Animacy in Ibero-Romance Languages’, pages 83–109. Fachbereich Sprachwissenschaft, Universität Konstanz. von Heusinger, K. and Kornfilt, J. (2005). The case of the direct object in Turkish: Semantics, syntax and morphology. Turkic Languages, 9:3–44. von Heusinger, K., Kornflit, J., and Kizilkaya, S. (2019). Differential object marking, partitivity and specificity in Turkish. In Bondarenko, T., Colley, J., Davis, C., and Privoznov, M., editors, Proceedings of the 14th Workshop on Altaic Formal Linguistics. MITWPL, Cambridge, MA. Higginbotham, J. (1985). On semantics. Linguistic Inquiry, 16:547–593. Hinterwimmer, S. (2015). A unified account of the properties of German demonstrative pronouns. In Grosz, P., Patel-Grosz, P., and Yanovich, I., editors, NELS 40: Semantics Workshop on Pronouns, pages 61–107. GLSA, Amherst. Hoekstra, T. (1988). Small clause results. Lingua, 74:101–139. Holmberg, A. (1991). Head scrambling. GLOW Handout, Leiden University. de Hoop, H. (1996). Case Configuration and Noun Phrase Interpretation. Garland Publishing, New York. de Hoop, H. and Narasimhan, B. (2008). Ergative case-marking in Hindi. In de Hoop, H. and de Swart, P., editors, Differential Subject Marking, pages 63–78. Springer, Dordrecht. Hornstein, N. and Nunes, J. (2002). On asymmetries between parasitic gap and across-theboard constructions. Syntax, 5:26–54. İşsever, S. (2003). Information structure in Turkish: the word order-prosody interface. Lingua, 113:1025–1053. İşsever, S. (2007). Towards a unified account to clause-initial scrambling in Turkish: a feature analysis. Turkic Languages, 11:93–123. İşsever, S. (2008). EPP-driven scrambling and Turkish. In Kurebito, T., editor, Ambiguity of Morphological and Syntactic Analyses, pages 390–398. Tokyo University of Foreign Studies (Research Institute for Languages of Asia and Africa [ILCAA]) Press, Tokyo. Im, Y.-J. (2007). On the use of the determiners and its application. Journal of Korean Language and Culture, 34. Ionin, T. and Matushansky, O. (2006). The composition of complex cardinals. Journal of Semantics, 23:315–360. Jackson, K. H. (2008). The Effect of Information Structure on Korean Scrambling. PhD thesis, University of Hawai’i. Jacobs, J. (1982). Syntax und Semantik der Negation im Deutschen. Wilhelm Fink Verlag, München. Janhunen, J. A. (2012). Mongolian. Benjamins, Amsterdam. Jayaseelan, K. A. (2001). IP-internal topic and focus phrases. Studia Linguistica, 55:39–75. Jayaseelan, K. A. (2004). Question movement in some SOV languages and the theory of feature checking. Language and Linguistics, 55:5–27. Jayaseelan, K. A. and Amritavalli, R. (2005). Scrambling in the cleft construction in Dravidian. In Sabel, J. and Saito, M., editors, The Free Word Order Phenomenon: Its Syntactic Sources and Diversity, pages 137–161. Mouton de Gruyter, Berlin/New York. Jayaseelan, K. A. and Amritavalli, R. (2017). Dravidian Syntax and Universal Grammar. Oxford University Press, Oxford. Jenks, P. (2015). Two kinds of definites in numeral classifier languages. In D’Antonio, S., Moroney, M., and Little, C. R., editors, Proceedings of SALT 25, pages 103–124. LSA and CLC Publications. Jenks, P. (2018). Articulated definiteness without articles. Linguistics and Philosophy, 26:287–350.
334
REFERENCES
Jiménez-Fernández, A. and İşsever, S. (2012). Deriving A/A’-effects in topic fronting: intervention of focus and binding. In Blaszczak, J., Rozwadowska, B., and Witkowski, W., editors, Current Issues in Generative Linguistics: Syntax, Semantics and Phonology, pages 8–25. CGCL, Wroclaw. Jo, J. and Palaz, B. (2018). Licensing pseudo-noun incorporation in Turkish. Poster at NELS 49. Jo, J. and Palaz, B. (2019). Non-canonical pseudo-incorporation in Turkish. Manuscript, Delaware University, July 2019. Johannessen, J. B. (2003). Negative polarity verbs in Norwegian. Working Papers in Scandinavian Syntax, 71:33–73. Kalin, L. (2014). Aspect and Argument Licensing in Neo-Aramaic. PhD thesis, University of California, Los Angeles. Kalin, L. (2018). Licensing and differential object marking: The view from neo-Aramaic. Syntax, 21:112–159. Kalin, L. and van Urk, C. (2001). Aspect-based agreement reversal in neo-Aramaic. In Arnett, N. and Bennett, R., editors, Proceedings of the 30th West Coast Conference on Formal Linguistics, pages 184–194. Cascadilla Proceedings Project, Somerville, MA. Kamali, B. (2015). Caseless direct objects in Turkish revisited. ZAS Papers in Linguistics, 58:107–123. Kamp, H. and Reyle, U. (1993). From Discourse to Logic. Springer, Dordrecht. Kan, S. (2010). Number marking, blocking effects, and Turkish noun phrases. Handout Ling 720: Plurality across languages, Fall 2010. Kang, A. (2015). (In)definiteness, Disjunction and Anti-specificity in Korean: A Study in the Semantics-Pragmatics Interface. PhD thesis, University of Chicago. Kang, A. (2018). Marking definiteness in articleless language: The role of the contextual domain restrictor KU in Korean. Ms. Kang, B.-M. (1994). Plurality and other semantic aspects of common nouns in Korean. Journal of East Asian Linguistics, 3:1–24. Kang, B.-M. (2002). Categories and meanings of Korean floating quantifiers – with some references to Japanese. Journal of East Asian Linguistics, 11:375–98. Katz, G. (2003). Event arguments, adverb selection, and the stative adverb gap. In Lang, E., Maienborn, C., and Fabricius-Hansen, C., editors, Modifying Adjuncts, pages 455–474. de Gruyter, Berlin. Kayne, R. (2017). Principles of particle constructions. In Guéron, J., Obenauer, H.-G., and Pollock, J.-Y., editors, Grammatical Representation, pages 101–140. Dordrecht, Foris. Keine, S. (2007). Reanalysing Hindi split-ergativity as a morphological phenomenon. In Trommer, J. and Opitz, A., editors, 1-2-many, pages 73–127. Linguistische Arbeitsberichte 85, Leipzig. Keine, S. (2010). Case and Agreement from Fringe to Core: A Minimalist Approach. De Gruyter, Berlin. Keine, S. and Bhatt, R. (2016). Interpreting verb clusters. Natural Language and Linguistic Theory, 34:1445–1492. Keine, S. and Müller, G. (2008). Differential argument encoding by impoverishment. In Richards, M. and Malchukov, A., editors, Scales, pages 83–136. Linguistische Arbeitsberichte 86, Leipzig. Keine, S. and Müller, G. (2011). Non-zero/non-zero alternations in differential object marking. In Lima, S., Mullin, K., and Smith, B., editors, Proceedings of NELS 39, pages 441–454. GLSA, Amherst.
REFERENCES
335
Keine, S. and Müller, G. (2015). Differential argument encoding by impoverishment. In Bornkessel-Schlesewsky, I., Malchukov, A., and Richards, M., editors, Scales and Hierarchies: A Cross-Disciplinary Perspective, page 75–130. de Gruyter, Berlin. Keine, S. and Poole, E. (2018a). Interpreting long scrambling in Hindi-Urdu. In Hucklebridge, S. and Nelson, M., editors, Proceedings of NELS 48, volume 2, pages 105–118, Amherst. GLSA. Keine, S. and Poole, E. (2018b). Not all reconstruction effects are syntactic. Manuscript, University of California, Los Angeles. Kelepir, M. (2001). Topics in Turkish Syntax: Clausal Structure and Scope. PhD thesis, MIT. Khym, H. (1998). Reanalysis of evidence for/against agrp in korean. In Kyle, J., editor, Kansas Working Papers in Linguistics, pages 47–57. University of Kansas, Berlin. Kidwai, S. (2020). A case for bare accusatives in Hindi-Urdu. Handout, Formal Approaches to South Asian Languages 10. Kim, K. and Melchin, P. B. (2018a). Modifying plurals, classifiers, and co-occurrence: The case of Korean. Glossa, 3(25):1–29. Kim, K. and Melchin, P. B. (2018b). On the complementary distribution of plurals and classifiers in East Asian classifier languages. Language and Linguistic Compass, 12(e12271):1–22. Kiparsky, P. (1973). Elsewhere. In Anderson, S. and Kiparsky, P., editors, A Festschrift for Morris Halle, pages 93–106. Academic Press, New York. Kiparsky, P. (2001). Structural case in Finnish. Lingua, 111:315–376. Kishimoto, H. (2000). Locational verbs, agreement, and object shift in Japanese. The Linguistic Review, 17:53–109. Kitahara, H. (1997). Elementary Operations and Optimal Derivations. MIT Press, Cambridge. Kluender, R. (1992). Deriving island constraints from principles of predication. In Goodluck, H. and Rochemont, M., editors, Island Constraints: Theory, Acquisition and Processing, pages 223–258. Kluwer, Dordrecht. Knecht, L. E. (1986). Subject and Object in Turkish. PhD thesis, MIT. Ko, H. (2007). Asymmetries in scrambling and cyclic linearization. Linguistic Inquiry, 38:49–83. Ko, H. (2014). Edges in Syntax. Oxford University Press, Oxford. Kobele, G. M. (2006). Generating Copies: An Investigation into Structural Identity in Language and Grammar. PhD thesis, University of California. Kobele, G. M. and Zimmermann, M. (2012). Quantification in German. In Paperno, D. and Keenan, E., editors, Handbook of Quantifiers in Natural Language: Volume I, pages 227–283. Springer, Cham. Koopman, H. (2005). Korean (and Japanese) Morphology from a syntactic perspective. Linguistic Inquiry, 36:609–665. Kornfilt, J. (1997). Turkish. Routledge, London. Kornfilt, J. (2003a). Scrambling, subscrambling, and case in Turkish. In Karimi, S., editor, Word Order and Scrambling, pages 125–155. Blackwell Publishing, Oxford. Kornfilt, J. (2003b). Subject case in Turkish nominalized clauses. In Junghanns, U. and Szucsich, L., editors, Syntactic Structures and Morphological Information, pages 129–215. Mouton de Gruyter, Berlin/New York. Kornfilt, J. (2005). Asymmetries between pre-verbal and post-verbal scrambling in turkish. In Sabel, J. and Saito, M., editors, The Free Word Order Phenomenon: Its Syntactic Sources and Diversity, pages 163–179. Mouton de Gruyter, Berlin/New York.
336
REFERENCES
Kornfilt, J. (2008). DOM and two types of DSM in Turkish. In de Hoop, H. and de Swart, P., editors, Differential Subject Marking, pages 79–112. Springer, Dordrecht. Koster, J. (1984). On binding and control. Linguistic Inquiry, 15:417–459. Kramer, R. (2015). A split analysis of plurality: Evidence from Amharic. In Arnett, N. and Bennett, R., editors, Proceedings of the 30th West Coast Conference on Formal Linguistics, pages 226–236. Cascadilla Proceedings Project, Somerville, MA. Kratzer, A. (1989). An investigation of the lumps of thought. Linguistics and Philosophy, 12:607–653. Kratzer, A. (1995). Stage-level and individual-level predicates. In Carlson, G. and Pelletier, F. J., editors, The Generic Book, pages 125–175. University of Chicago Press, Chicago. Kratzer, A. (1996). Severing the external argument from its verb. In Rooryck, J. and Zaring, L., editors, Phrase Structure and the Lexicon, pages 109–137. Kluwer, Dordrecht. Kratzer, A. (1998). Scope or pseudoscope? Are there wide scope indefinites? In Rothstein, S., editor, Events and Grammar, pages 163–196. Kluwer, Dordrecht. Kratzer, A. (2000). The event argument. Manuscript: University of Massachusetts, Amherst, MA. Kratzer, A. (2019). Situations in natural language semantics. In Zalta, E. N., editor, The Stanford Encyclopedia of Philosophy. Metaphysics Research Lab, Stanford University, Summer 2019 edition. Krifka, M. (1998). Scope inversion under the rise-fall contour in German. Linguistic Inquiry, 29:75–112. Krifka, M. (1999). At least some indefinites are not indefinites. In Turner, K., editor, The Semantics Pragmatics Interface from Different Points of View, pages 257–291. Elsevier Science, North Holland. Kripke, S. (1972). Naming and necessity. In Davidson, D. and Harman, G., editors, Semantics of Natural Languages, pages 253–355. Reidel, Dordrecht. Krishnamurti, B. (1998). Telugu. In Steever, S. B., editor, The Dravidian Languages, pages 202–240. Routledge, London/New York. Krishnamurti, B. (2003). The Dravidian Languages. Cambridge University Press, Cambridge. Kural, M. (1997). Postverbal Constituents in Turkish and the linear correspondence axiom. Linguistic Inquiry, 28:498–519. Kwon, S.-N. and Zribi-Hertz, A. (2006). Bare objects in Korean: (pseudo-)incorporation and (in)definiteness. In Vogeleer, S. and Tasmowski, L., editors, Non-definiteness and Plurality, pages 107–132. John Benjamins, Amsterdam/Philadelphia. Kwon, S.-N. and Zribi-Hertz, A. (2008). Differential function marking, case, and information structure: Evidence from Korean. Language, 84:258–299. Landau, I. (2013). Control in Generative Grammar. Cambridge University Press, Cambridge. Landau, I. (2015). A Two-Tiered Theory of Control. MIT Press, Cambridge. Landau, I. (2017). Direct variable binding and agreement in obligatory control. In PatelGrosz, P., Grosz, P. G., and Zobel, S., editors, Pronouns in Embedded Contexts, pages 1–41. Springer, Dordrecht. Landman, F. (2000). Events and Plurality: The Jerusalem Lectures. Blackwell Publishing, Oxford, UK. Larson, R. K. (1988). On the double object construction. Linguistic Inquiry, 19:335–391. Lasnik, H. (2002). Clause-mate conditions revisited. Glot International, 6:94–96.
REFERENCES
337
Lee, Chungmin (1989). (In)definites, case markers, classifiers and quantifiers in Korean. In Kuno, S., Whitman, J., Kang, Y.-S., Lee, I.-H., Bak, S.-Y., and Kim, Y.-J., editors, Harvard Studies in Korean Linguistics III, pages 469–487. Hanshin, Seoul. Lee, Chungmin (1992). Definite/specific and case marking in Korean. In Kim-Renaud, Y., editor, Theoretical Issues in Korean Linguistics, pages 325–341. Center for the Study of Language (CSLI) for Stanford Linguistics Society. Lee, EunHee (2019). Korean Syntax and Semantics. Cambridge University Press, Cambridge. Lee, Hanjung (2006). Iconicity and variation in the choice of object forms in Korean. Language Research, 42:323–355. Lee, Hanjung (2008). Quantitative variation in Korean case ellipsis: Implications for case theory. In de Hoop, H. and de Swart, P., editors, Differential Subject Marking, pages 41– 61. Springer, Dordrecht. Lee, Hanjung (2011). Gradients in Korean case ellipsis: An experimental investigation. Lingua, 121(1):20–34. Lee, Kum-Young (2009). Finite Control in Korean. PhD thesis, University of Iowa. Lees, R. B. (1965). Turkish nominalizations and a problem of ellipsis. Foundations of Language, 1:112–121. Lee, Youngjoo (1999). The Syntax and Semantics of Focus Particles. PhD thesis, Seoul National University. Legate, J. A. (2008). Morphological and abstract case. Linguistic Inquiry, 39:55–101. Legate, J. A. (2014). Voice and v. Lessons from Acehnese. MIT Press, Cambridge, MA. Lehmann, T. (1993). A Grammar of Modern Tamil. Pondicherry Institute of Linguistics and Culture, Pondicherry. Lehmann, T. (1998). Old Tamil. In Steever, S. B., editor, The Dravidian Languages, pages 75–99. Routledge, London/New York. Leonetti, M. (2004). Specificity and differential object marking in Spanish. Catalan Journal of Linguistics, 3:75–114. Levin, T. (2015). Licensing without Case. PhD thesis, MIT. Levin, T. (2016). Successive-cyclic case assignment: Korean nominative-nominative casestacking. Natural Language and Linguistic Theory, 35:447–498. Levin, T. (2019). On the nature of differential object marking: Insights from Palauan. Natural Language and Linguist Theory, 37:167–213. Li, Y. (1990). X0 -binding and verb incorporation. Linguistics, 21:399–426. Lidz, J. (2006). The grammar of accusative case in Kannada. Language, 82:10–32. Link, G. (1987). Generalized quantifiers and Plurals. In Gärdenfors, P., editor, Generalized Quantifiers, Linguistic and Logical Approaches, pages 151–180. Reidel, Dordrecht. Lobeck, A. (1995). Ellipsis: Functional Heads, Licensing, and Identification. Oxford University Press, Oxford. Longobardi, G. (1994). Reference and proper names: A theory of N-movement in syntax and logical form. Linguistic Inquiry, pages 609–665. Lo´pez, L. (2012). Indefinite Objects. Scrambling, Choice Functions, and Differential Marking. The MIT Press, Cambridge, Massachusetts. Lyutikova, E. and Pereltsvaig, A. (2013). Elucidating nominal structure in articleless languages: A case study of Tatar. In Faytak, M., Goss, M., Baier, N., Merrill, J., Neely, K., Donnelly, E., and Heath, J., editors, Proceedings of the 39th Annual Meeting of the Berkeley Linguistics Society, pages 123–136. Berkeley Linguistics Society, Berkeley, CA. Madhavan, P. (1987). Clefts and Pseudoclefts in English and Malayalam: A Study in Comparative Syntax. PhD thesis, CIEFL, Hyderabad.
338
REFERENCES
Mahajan, A. (2003). Word order and (remnant) VP movement. In Karimi, S., editor, Word Order and Scrambling, pages 217–237. Blackwell Publishing. Manzini, R. (1983). On control and control theory. Linguistic Inquiry, 14:421–446. Marantz, A. (1991). Case and licensing. In Westphal, G., Ao, B., and Chae, H.-R., editors, Proceedings of the Eigth Eastern States Conference on Linguistics, pages 234–253. University of Maryland. Marantz, A. (1993). Implications of asymmetries in double object constructions. In Mchombo, S. A., editor, Theoretical Aspects of Bantu Grammar, pages 113–150. CSLI Publications, Stanford, CA. Marantz, A. (1995). A late note on late insertion. In Kim, Y.-S., Lee, B.-C., Lee, K.-J., Yang, H.-K., and Yoon, J.-Y., editors, Explorations in Generative Grammar. A Festschrift for Dong-Whee Yang, pages 396–413. Hankuk, Seoul. Marantz, A. (2000). Roots: The universality of root and pattern morphology. Paper presented at the Conference on Afro-Asiatic Languages, University of Paris VII. Massam, D. (2001). Pseudo noun incorporation in Niuean. Natural Language and Linguistic Theory, 19:153–97. Massam, D. and Smallwood, C. (1997). Essential features of predication in English and Niuean. In Kusumoto, K., editor, Proceedings of NELS 27, pages 263–272, Amherst. GLSA. Matthewson, L. (1998). Determiner Systems and Quantificational Strategies: Evidence from Salish. Holland Academic Graphics, The Hague. Matthewson, L. (1999). On the interpretation of wide-scope indefinites. Natural Language Semantics, 7:79–134. Matushansky, O. (2006). Head movement in linguistic theory. Linguistic Inquiry, 37:69– 109. Matushansky, O. (2010). Same problem, different solution. Manuscript, Utrecht University/CNRS. May, R. (1977). The Grammar of Quantification. PhD thesis, MIT. May, R. (1985). Logical Form. MIT Press, Cambridge, MA. McCloskey, J. (2010). The shape of Irish clauses. In Carnie, A., editor, Proceedings of Formal Approaches to Celtic Linguistics. Cambridge Scholars Publishing, Newcastle upon Tyne. McNally, L. (2004). Bare plurals in Spanish are interpreted as properties. Catalan Journal of Linguistics, 3:115–133. Melis, C. (1995). El objeto directo personal en El Cantar de Mio Cid. Estudio sintácticopragmático. In Pensado, C., editor, El complemento directo preposicional, pages 133–163. Visor Libros, Madrid. Merchant, J. (2001). The Syntax of Silence: Sluicing, Islands, and the Theory of Ellipsis. Oxford University Press, New York. Merchant, J. (2002). Swiping in Germanic. In Zwart, C. J.-W. and Abraham, W., editors, Studies in Comparative Germanic Syntax: Proceedings from the 15th Workshop on Comparative Germanic Syntax, pages 289–315. John Benjamins, Amsterdam/ Philadelphia. Merchant, J. (2009). Polyvalent case, geometric hierarchies, and split ergativity. In Bunting, J., Desai, S., Peachey, R., Straughn, C., and Tomkova, Z., editors, Proceedings of the 42nd Annual Meeting of the Chicago Linguistics Society, pages 57–76. University of Chicago Press, Chicago. Milsark, G. (1977). Toward an explanation of certain peculiarities of the existential construction in English. Linguistic Analysis, 3:1–29. Mithun, M. (1984). The evolution of noun incorporation. Language, 60:847–894.
REFERENCES
339
Miyagawa, S. (1993). Case-checking and minimal link condition. In Phillips, C., editor, Papers on Case and Agreement II, volume 19, pages 213–254. MIT Working Papers in Linguistics. Miyagawa, S. (2011). Genitive subjects in Altaic and specification of phase. Journal of East Asian Linguistics, 121:1265–1282. Mohanan, T. (1995). Wordhood and lexicality: Noun incorporation in Hindi. Natural Language and Linguistic Theory, 13:75–134. Monahan, P. (2006). Control in Halh Mongolian. Manuscript, Variation in Control Structures Project. Montaut, A. (2018). The rise of differential object marking in Hindi and related languages. In Serzˇant, I. A. and Witzlack-Makarevich, A., editors, Diachrony of differential argument marking, pages 281–313. Language Science Press, Berlin. Moravcsik, E. A. (2012). The distribution of case. In Malchukov, A. L. and Spencer, A., editors, The Oxford Handbook of Case, pages 231–245. Oxford University Press, Oxford. Müller, G. (1995). A-bar Syntax. A Study in Movement Types. Mouton de Gruyter, Berlin. Müller, G. (1996). A constraint on remnant movement. Natural Language and Linguistic Theory, 14:355–407. Müller, G. (1998). Incomplete Category Fronting. Kluwer, Dordrecht. Müller, G. (2004a). A distributed morphology approach to syncretism in Russian noun inflection. In Arnaudova, O., Browne, W., Rivero, M. L., and Stojanovic, D., editors, Proceedings of FASL 12, pages 353–373. Michigan SLavic Publications. Müller, G. (2004b). Verb-second as vP-first. Journal of Comparative Germanic Linguistics, 7:179–234. Müller, G. (2005). Syncretism and iconicity in Icelandic noun declensions: A distributed morphology approach. In Booij, G. and van Marle, J., editors, Yearbook of Morphology 2004, pages 229–271. Springer Netherlands, Dordrecht. Müller, G. (2006). Subanalyse verbaler Flexionsmarker. In Breindl, E., Gunkel, L., and Strecker, B., editors, Grammatische Untersuchungen, pages 183–203. Narr, Tübingen. Müller, G. (2009). Ergativity, accusativity, and the order of Merge and Agree. In Grohmann, K. K., editor, Explorations of Phase Theory: Features and Arguments, pages 269–308. Mouton de Gruyter, Berlin/New York. Müller, G. (2010). On deriving CED effects from the PIC. Linguistic Inquiry, 41:35–82. Müller, G. (2011). Constraints on Displacement: A Phase-Based Approach. John Benjamins, Amsterdam. Müller, G. (2017). Structure removal: An argument for feature-driven Merge. Glossa, 2(28):1–35. Müller, G. (2018). Pseudo-incorporation by structure removal. Handout, Talk at CGSW 33, Göttingen. Müller, G. and Thomas, D. (2017). Three-way systems do not exist. In Coon, J., Massam, D., and Travis, L. D., editors, The Oxford Handbook of Ergativity, pages 279–307, Oxford. OUP. Murphy, A. and Puškar, Z. (2018). Closest conjunct agreement is an illusion. Natural Language and Linguistic Theory, 36(4):1207–1261. Murugesan, G. (2019). Predicting the Anaphor Agreement Effect and its Violations. PhD thesis, Universität Leipzig. Nevins, A. (2011). Marked targets versus marked triggers and impoverishment of the dual. Linguistic Inquiry, 42:413–444. Ngata, H. M. (1994). English-Maori dictionary. Learning Media, Wellington. Nichols, J. (1986). Head-marking and dependent-marking grammar. Language, 62:56–119. Nilsson, B. (1985). Case Marking Semantics in Turkish. PhD thesis, University of Stockholm.
340
REFERENCES
Noyer, R. (1998). Impoverishment theory and morphosyntactic markedness. In Lapointe, S. G., Brentari, D. K., and Farrell, P. M., editors, Morphology and Its Relation to Syntax, pages 264–285. CSLI Publications, Stanford, CA. Nuger, J. (2010). Architecture of the Palauan Verbal Complex. PhD thesis, University of California, Santa Cruz. Nunes, J. (2001). Sideward movement. Linguistic Inquiry, 32:303–344. Nunes, J. (2004). Linearization of Chains and Sideward Movement. The MIT Press, Cambridge, Massachusetts. Ochi, M. (2001). Move F and ga/no conversion in Japanese. Journal of East Asian Linguistics, 10:247–286. O’Grady, W. (1991). Categories and Case: The Sentence Structure of Korean. Benjamins, Amsterdam. Onea, E. and Geist, L. (2011). Indefinite determiners and the pragmatics of referential anchoring. International Review of Pragmatics, 3:194–227. Ott, D. and de Vries, M. (2016). Right-dislocation as deletion. Natural Language and Linguistic Theory, 34:641–690. Ouhalla, J. (1993). Subject-extraction, negation and the anti-agreement effect. Natural Language and Linguistic Theory, 11:477–518. Özge, U. (2011). Turkish indefinites and accusative marking. In Simpson, A., editor, Proceedings of the 7th Workshop on Altaic Formal Linguistics, pages 253–267. MIT Working Papers in Linguistics. Özsoy, S. (2001). On ‘small’ clauses, other ‘bare’ verbal complements and feature checking in Turkish. In Taylan, E. E., editor, The Verb in Turkish, pages 213–237. John Benjamins, Amsterdam. Öztürk, B. (2005). Case, Referentiality and Phrase Structure. John Benjamins, Amsterdam/Philadelphia. Öztürk, B. (2009). Incorporating agents. Lingua, 119:334–358. Öztürk, B. (2013). Rightward movement, EPP and specifiers: Evidence from Uyghur and Khalkha. In Webelhuth, G., Sailer, M., and Walker, H., editors, Rightward Movement in a Comparative Perspective, pages 175–210. John Benjamins, Amsterdam/Philadelphia. Özyıldız, D. (2017). Quantifiers in Turkish. In Paperno, D. and Keenan, E., editors, Handbook of Quantifiers in Natural Language: Volume II, pages 857–937. Springer, Cham. Panagiotidis, P. (2015). Categorial Features: A Generative Theory of Word Class Categories. Cambridge University Press, Cambridge. Park, Y. (2018). Overt subjects in obligatory control constructions in Korean. In Bennett, W. G., Hracs, L., and Storoshenko, D. R., editors, Proceedings of the 35th West Coast Conference on Formal Linguistics, pages 305–312. Cascadilla Proceedings Project, Somerville, MA. Parsons, T. (1990). Events in the Semantics of English. MIT Press, Cambridge, MA. Parsons, T. (1995). Thematic relations and arguments. Linguistic Inquiry, 26:635–662. Partee, B. (1975). Deletion and variable binding. In Keenan, E., editor, Formal Semantics of Natural Languages, pages 16–34. Cambridge University Press, Cambridge. Partee, B. (1986a). Ambiguous pseudoclefts with unambiguous be. In Berman, S., Choe, J.-W., and McDonough, J., editors, Proceedings of NELS 16, pages 354–366. University of Massachussetts, Amherst, Amherst,MA. Partee, B. (1986b). Noun phrase interpretation and type-shifting principles. In Groenendijk, J., de Jongh, D., and Stokhof, M., editors, Studies in Discourse Representation Theory and the Theory of Generalized Quantifiers, pages 115–143. Foris, Dordrecht.
REFERENCES
341
Partee, B. (1988). Many quantifiers. In Powers, J. and de Jong, K., editors, Proceedings of the Fifth ESCOL, pages 383–402. The Ohio State University, Columbus, OH. Patel-Grosz, P. and Grosz, P. G. (2017). Revisiting pronominal typology. Linguistic Inquiry, 48:259–297. Penka, D. (2010). Negative Indefinites. Oxford University Press, Oxford. Pensado, C. (1995). Estado de la cuestio´n y bibliografí a comentada. In Pensado, C., editor, El complemento directo preposicional, pages 179–233. Visor Libros, Madrid. Percus, O. (2006). Antipresuppositions. In Ueyama, A., editor, Theoretical and Empirical Studies of Reference and Anaphora: Toward the Establishment of Generative Grammar as an Empirical Science, page 52–73. Japan Society for the Promotion of Science. Percus, O. and Sauerland, U. (2003). Pronoun movement in dream reports. In Kadowaki, M. and Kawahara, S., editors, Proceedings of NELS 33, pages 265–284. GLSA Publications, Amherst. Pesetsky, D. (1989). Language-particular processes and the earliness principle. Ms., MIT. Pesetsky, D. (2010). Zero Syntax: Experiencers and Cascades. MIT Press, Cambrige. Pesetsky, D. and Torrego, E. (2001). T-to-C movement: Causes and consequences. In Kenstowicz, M., editor, Ken Hale: A Life in Language, pages 355–426. MIT Press, Cambridge, MA. Pfau, R. (2000). Features and Categories in Language Production. PhD thesis, GoetheUniversität Frankfurt. Pfau, R. (2009). Grammar as Processor: A Distributed Morphology Account of Spontaneous Speech Errors. John Benjamins, Amsterdam. Platzack, C. (2013). Head movement as a phonological operation. In Cheng, L. L.-S. and Corver, N., editors, Diagnosing Syntax, pages 21–43. Oxford University Press, Oxford. Polinsky, M. (1992). Maori ‘He’ revisited. Oceanic Linguistics, 31:229–250. Poole, E. (2017). Movement and the Semantic Type of Traces. PhD thesis, University of Massachusetts, Amherst. Poppe, N. (1951). Khalkha-mongolische Grammatik, mit Bibliographie, Sprachproben und Glossar. Franz Steiner, Wiesbaden. Prince, A. and Smolensky, P. (2004). Optimality Theory: Constraint Interaction in Generative Grammar. Blackwell, Malden, MA and Oxford. Revision of 1993 Technical Report TR-2, Rutgers University Center for Cognitive Science, Rutgers University, New Brunswick, NJ. Pylkkänen, L. (2008). Introducing Arguments. Linguistic Inquiry Monographs. MIT Press, Cambridge. Reinhart, T. (1997). Quantifier scope: How labor is divided between QR and choice functions. Linguistics and Philosophy, 20:335–397. Reinhart, T. (2006). Interface Strategies. The MIT Press, Cambridge, Massachusetts. Roberts, C. (2003). Uniqueness in definite noun phrases. Linguistics and Philosophy, 26:287–350. Rodrí guez-Mondoñedo, M. (2007). The Syntax of Objects: Agree and Differential Object Marking. PhD thesis, University of Connecticut. Romero, M. (1997). The correlation between scope reconstruction and connectivity effects. In Curtis, E., Lyle, J., and Webster, G., editors, Proceedings of the 16th West Coast Conference on Formal Linguistics, pages 354–366. CSLI Publications, Stanford. Romero, M. (1998). Focus and Reconstruction Effects in Wh-Phrases. PhD thesis, University of Massachusetts, Amherst. Ross, J. (1967). Constraints on Variables in Syntax. PhD thesis, Massachusetts Institute of Technology.
342
REFERENCES
Rullmann, H. (1995). Maximality in the Semantics of Wh-Constructions. PhD thesis, University of Massachusetts, Amherst. Russell, B. (1905). On denoting. Mind, 14:479–493. Sadock, J. M. (1980). Noun incorporation in Greenlandic: A case of syntactic word formation. Language, 56:300–319. Sag, I. and Pollard, C. (1991). An integrated theory of complement control. Language, 67:63–113. Saito, M., Lin, T.-H. J., and Murasugi, K. (2008). N’-ellipsis and the structure of noun phrases in Chinese and Japanese. Journal of East Asian Linguistics, 17:247–71. Sarma, V. (1999). Case, Agreement and Word Order: Issues in the Syntax and Acquisition of Tamil. PhD thesis, MIT. Sarma, V. (2003). Non-canonical word order: Topic and focus in adult and child Tamil. In Karimi, S., editor, Word Order and Scrambling, pages 238–272. Blackwell Publishing, Oxford. Sauerland, U. (1998). On the Making and Meaning of Chains. PhD thesis, MIT. Sauerland, U. (2003). A new semantics for number. In Young, R. and Zhou, Y., editors, Proceedings of SALT 13, pages 258–275, Ithaca, NY. CLC Publications. Sauerland, U. (2004). The interpretation of traces. Natural Language Semantics, 12:63–127. Sauerland, U. (2007). Flat binding: Binding without sequences. In Sauerland, U. and Gärtner, H.-M., editors, Interfaces + recursion = language? Chomsky’s Minimalism and the View from Syntax-Semantics, pages 197–253. Mouton de Gruyter, Berlin. Sauerland, U. (2008a). Implicated presuppositions. In Steube, A., editor, The Discourse Potential of Underspecified Structures, pages 581–600. de Gruyter, Berlin. Sauerland, U. (2008b). On the semantic markedness of Phi-features. In Harbour, D., Adger, D., and Béjar, S., editors, Phi-Theory. Phi-features across Modules and Interfaces, pages 35–56. Oxford University Press, Oxford. Sağ, Y. (2018). The semantics of Turkish numeral constructions. In Sauerland, U. and Solt, S., editors, Proceedings of Sinn und Bedeutung 22, vol. 2, ZASPiL 61, pages 307–324. ZAS, Berlin. Schein, B. (1993). Plurals and Events. MIT Press, Cambridge, MA. Schiffman, H. F. (1999). A Reference Grammar of Spoken Tamil. Cambridge University Press, Cambridge. Schneider-Zioga, P. (2007). Anti-agreement, anti-locality and minimality. Natural Language and Linguistic Theory, 25:403–446. Schoorlemmer, E. and Temmerman, T. (2012). Head movement as a PF-phenomenon: Evidence from identity under ellipsis. In Choi, J., Hogue, E. A., Punske, J., Tat, D., Schertz, J., and Trueman, A., editors, Proceedings of WCCFL 29, pages 232–240, Amherst. GLSA. Schwarz, F. (2009). Two Types of Definites in Natural Language. PhD thesis, University of Massachusetts Amherst. Schwarz, F. (2014). How weak and how definite are weak definites? In Aguilar-Guevara, A., Bruyn, B. L., and Zwarts, J., editors, Weak Referentiality, pages 213–235. John Benjamins, Amsterdam/Philadelphia. Schwarz, F. (2019). Weak vs. strong definite articles – Meaning and form across languages. In Aguilar-Guevara, A., Loyo, J. P., and Maldonado, V. V.-R., editors, Definiteness across Languages, pages 1–37. Language Science Press, Berlin. Schwarzschild, R. (2002). Singleton indefinites. Journal of Semantics, 19:289–314. Schweikert, W. (2005). The Order of Prepositional Phrases in the Structure of the Clause. John Benjamins, Amsterdam/Philadelphia.
REFERENCES
343
Seidel, E. (2019). Bare direct objects in Turkish: Pseudo-incorporated or weak arguments. In Bondarenko, T., Colley, J., Davis, C., and Privoznov, M., editors, Proceedings of the 14th Workshop on Altaic Formal Linguistics, pages 277–288, Cambridge, MA. MIT Working Papers in Linguistics. Selvanathan, N. (2017). Cleft constructions in Tamil and anti-agreement. Manuscript, lingbuzz/003681. Siddiqi, D. (2006). Minimize Exponence: Economy Effects on a Model of the Morphosyntactic Component of Grammar. PhD thesis, University of Arizona. Siddiqi, D. (2009). Syntax within the Word: Economy, Allomorphy, and Argument Selection in Distributed Morphology. John Benjamins, Amsterdam. Silverstein, M. (1976). Hierarchies of features and ergativity. In Dixon, R. M. W., editor, Grammatical Categories in Australian Languages, pages 112–171. Australian Institute of Aboriginal Studies, Canberra. Simpson, A. (2008). Classifiers and DP structure in Southeast Asia. In Cinque, G. and Kayne, R. S., editors, The Oxford Handbook of Comparative Syntax, pages 806–838. Oxford University Press, Oxford. Słodowicz, S. (2007). Complement control in Turkish. In Stiebels, B., editor, ZAS Papers in Linguistics 47: Studies in Complement Control, pages 125–157. Berlin. Sohn, H.-M. (1999). The Korean Language. Cambridge University Press, Cambridge. Sohn, K.-W. (1995). Negative Polarity Items, Scope, and Economy. PhD thesis, University of Connecticut, Storrs. Sportiche, D. (2016). Neglect. Manuscript, University of California, Los Angeles. Stabler, E. (1997). Derivational minimalism. In Retoré, C., editor, Logical Aspects of Computational Linguistics, pages 68–95, Heidelberg. Springer. Stabler, E. (2011). Computational perspectives on minimalism. In Boeckx, C., editor, Oxford Handbook of Linguistic Minimalism, pages 616–641, Oxford. Oxford University Press. Stalnaker, R. C. (1978). Assertion. Syntax and Semantics, 9:315–322. von Stechow, A. (1993). Die Aufgaben der Syntax. In Syntax: An International Handbook of Contemporary Research, volume 1. Halbband, pages 1–88. De Gruyter Mouton, Berlin/Boston. von Stechow, A. (2012). Syntax and semantics: An overview. In Maienborn, C., Heusinger, K. v., and Portner, P., editors, Semantics, Handbooks of Linguistics and Communication Science, pages 2173–2223. De Gruyter Mouton. von Stechow, A. and Sternefeld, W. (1988). Bausteine syntaktischen Wissens. Westdeutscher Verlag, Opladen. Steedman, M. (1985). Dependency and coordination in the grammar of Dutch and English. Language, 61:523–568. Sternefeld, W. (2006). Syntax: Eine morphologisch motivierte generative Beschreibung des Deutschen. Stauffenburg, Tübingen. Stiebels, B. (2006). Agent focus in Mayan languages. Natural Language and Linguistic Theory, 24:501–570. Stowell, T. (1981). Origins of Phrase Structure. PhD thesis, MIT. Stowell, T. (1983). Subjects across categories. The Linguistic Review, 2:285–312. Sundaresan, S. (2006). The argument structure of verbal alternations in Tamil. In Baumer, D., Montero, D., and Scanlon, M., editors, Proceedings of the 25th West Coast Conference on Formal Linguistics, pages 390–398. Cascadilla Proceedings Project, Somerville, MA. Sundaresan, S. (2012). Context and (Co)reference in the Syntax and Its Interfaces. PhD thesis, Universität Stuttgart/University of Tromso.
344
REFERENCES
Sundaresan, S. and McFadden, T. (2017). The articulated v layer: evidence from Tamil. In ´ ., editors, The Verbal Domain, pages 153–178. D’A lessandro, R., Franco, I., and Gallego, A Oxford University Press, Oxford. Swart, H. d. (1993). Adverbs of Quantification: A Generalized Quantifier Approach. Garland, New York. Swart, H. d. (2016). Negation. In Aloni, M. and Dekker, P., editors, The Cambridge Handbook of Formal Semantics, Cambridge Handbooks in Language and Linguistics, page 467–489. Cambridge University Press. Syed, S. and Simpson, A. (2017). On the DP/NP status of nominal projections in Bangla: Consequences for the theory of phases. Glossa, 2(68):1–24. Takano, Y. (1992). Unbound traces and indeterminacy of derivation. In Nakamura, M., editor, Current Topics in English and Japanese, pages 229–253. Hituzi Syobo, Tokyo. Tanaka, H. (2001). Right-dislocation as scrambling. Journal of Linguistics, 37:551–579. Termucu, C. (2005). The interaction of syntax and discourse in word order: data from Turkish. Dilbilim ve Uygulamaları, pages 123–159. Testelets, Y. G. and Arkadiev, P. M. (2014). Differential nominal marking: The pervasive case alternation in Circassian. Syntax of the World’s Languages VI. Pavia. Torrego Salcedo, E. (1999). El complemento directo preposicional. In Bosque, I. and Demonte, V., editors, Gramática descriptiva de la lengua española. Las construcciones sintácticas fundamentales. Relaciones temporales, aceptuales y modales, volume 2, pages 1779–1805. Espasa Calpe, Madrid. Trommer, J. (2001). Distributed Optimality. PhD thesis, Universität Potsdam. Tyler, M. (2019). Differential object marking by A’-status. In Baird, M. and Pesetsky, J., editors, Proceedings of the 49th Meeting of the North East Linguistics Society. GLSA Publications, Amherst. van Urk, C. (2019a). Constraining predicate fronting. GLOW 42, Handout. van Urk, C. (2019b). Object licensing in Fijian and the role of adjacency. Natural Language and Linguist Theory, 38:313–364. Uygun, D. (2006). Scrambling bare nominal objects in Turkish. Paper presented at ICTL 13, Uppsala University. Vallduví, E. and Engdahl, E. (1996). The linguistic realization of information packaging. Linguistics, 34:459–519. Weisser, P. (2017). A purely morphological approach to differential object marking in Tamil and Beyond. Handout, GGS 43. Weisser, P. (2018). NOM-NOM in Mari. Handout, Shrinking Trees Workshop. Wiese, B. (1999). Unterspezifizierte Paradigmen. Form und Funktion in der pronominalen Deklination. Linguistik Online, 4. Wiese, B. (2001). Zur lateinischen Nominalflexion: Die Form-Funktions-Beziehung. Manuscript. Wiklund, A.-L., Hrafnbjargarson, G. H., Bentzen, K., and Þorbjörg Hro´arsdo´ttir (2007). Rethinking Scandinavian verb movement. Journal of Comparative Germanic Linguistics, 10:203–233. Wiltschko, M. (2008). The syntax of non-inflectional plural marking. Natural Language and Linguistic Theory, 26:639–694. Winter, Y. (1997). Choice functions and the scopal semantics of indefinites. Linguistics and Philosophy, 20:399–467. Wojdak, R. (2008). The Linearization of Affixes: Evidence from Nuu-Chah-Nulth. Springer, Dordrecht.
REFERENCES
345
Wu, Y. and Bodomo, A. (2009). Classifiers ≠ determiners. Linguistic Inquiry, 40:487–503. Wunderlich, D. (1997). Cause and the structure of verbs. Linguistic Inquiry, 28:27–68. Wurmbrand, S. (2001). Infinitives: Restructuring and Clause Structure. Mouton de Gruyter, Berlin. Wurmbrand, S. (2017). Formal and semantic agreement in syntax: A dual feature approach. In Emonds, J. and Janebová, M., editors, Language Use and Linguistic Structure: Proceedings of the Olomouc Linguistics Colloquium 2016, pages 19–36. Palacky´ University, Olomouc. Yang, D.-W. (1985). On the integrity of control theory. In Stephen, B., Choe, J.-W., and McDonough, J., editors, Proceedings of NELS 15, pages 389–408. GLSA Publications, Amherst. Yoon, J. H. S. (1995a). Nominal, verbal, and cross-categorical affixation in Korean. Journal of East Asian Linguistics, 4:325–356. Yoon, J.-M. (1991). The Syntax of A-Chains: A Typological Study of ECM and Scrambling. PhD thesis, Cornell University. Yoon, J.-M. (1995b). Verb movement and structure of IP in Korean. Language Research, 26:343–371. Zeijlstra, H. (2004). Sentential Negation and Negative Concord. PhD thesis, University of Amsterdam. Zimmermann, E. (2017). Multiple reduplication as non-segmental affixation: A case study from Nuuchahnulth. In Lamont, A. and Tetzloff, K., editors, Proceedings of NELS 47. GLSA Publications, Amherst. Zimmermann, M. (2007). Overt existential closure in Bura (Central Chadic). In Friedman, T. and Gibson, M., editors, Proceedings of SALT 17, pages 333–350. Cornell University, Ithaca, NY. Zimmermann, M. (2008). Quantification in Hausa. In Matthewson, L., editor, Quantification: A Cross-Linguistic Perspective, pages 415–475. Emerald (North Holland Linguistic Series), East Sussex. Zwart, J.-W. (2017). An argument against the syntactic nature of verb movement. In Bailey, L. R. and Sheehan, M., editors, Order and Structure in Syntax I: Word Order and Syntactic Structure, pages 29–47. Language Science Press, Berlin.
Language Index Adyghe 5, 42–43, 303fn. 9.1 Cantonese 95 Catalan 45 Chamorro 299–301 English 18, 24, 36–37, 102, 133, 137, 138, 139, 180, 183fn. 5.22, 186, 201, 295 French 111 German 9, 25, 75fn. 4.23, 99fn. 4.34, 104fn. 4.38, 107–109, 110fn. 5.1, 111–113, 168, 179–188, 218–221, 224, 237fn. 7.9, 243, 271–277, 297, 303–304 Hebrew 114 Hindi 1, 3, 6, 8, 9, 10, 11, 12, 43fn. 4.4, 45–57, 110fn. 5.1, 115–116, 120–123, 129–130, 132, 133, 137, 138, 139, 140, 146, 168, 190–191, 222–223, 286, 298, 307–309 Hungarian 109 Japanese 93, 162, 243 Kannada 5, 310 Kiswahili 45
Korean 5, 8–9, 12fn. 2.3, 27, 29–30, 31, 34, 41, 43, 50, 58, 83–108, 125, 166–179, 195–196, 212–218, 224, 261–271, 286–291, 310 Mandarin 10, 60fn. 4.15, 93–95 Maori 35–36, 301–302, 310 Mongolian 8, 9, 44, 57, 59, 67–75, 144–154, 199–202, 224, 239–249 Niuean 44fn. 4.5, 305–307 Norwegian 110fn. 5.1, 114 Romanian 45 Sakha 3, 222, 292, 293, 298 Spanish 2, 5–7, 8, 9, 11, 45, 57, 59, 115, 118–119, 168, 185, 189–190, 192–195, 223, 224, 310–316 St’a´t’imcets 35–36, 310 Swedish 111 Tamil 3, 8, 9, 31–33, 41–42, 57, 59–67, 130–143, 146, 168fn. 5.13, 198–199, 222, 224, 227–239, 279–286, 292–298 Tatar 5–6 Thai 10, 60fn. 4.15 Turkish 2–3, 5–6, 9, 37, 40, 43–44, 50, 58, 59, 75–83, 124–125, 129, 154–166, 168, 202–212, 224, 249–261, 298–299, 310, 316–320
Subject Index Agree 15–16, 37 agreement default 231, 296–298 honorific 41 phi 16, 33, 39–42, 162, 175, 197, 206, 303fn. 9.1 animacy 1, 7, 10, 45–47, 57–58, 59, 61fn. 4.16, 67, 134, 303fn. 9.1, 322 binder prefix 22, 28, 191–192 binding 2, 6, 10, 14, 27–28, 189–196, 214–215, 232–234, 278, 314, 318 case assignment 14, 17, 33, 39, 42, 44–45, 303, 306, 309, 316 decomposition 34, 58 dependent case 14, 44–45 licensing 38fn. 3.12, 303fn. 9.1, 310fn. 9.5 category change 3, 36–38 choice function 23, 133, 168–169, 185, 302, 310–313 classifiers 83, 92–97, 175–177 cleft 230–233 control 2, 6, 10, 14, 28, 190–191, 196–197, 215, 217–218, 278, 285–286 conventionalized meaning 75fn. 4.23, 181–182, 190, 197, 210, 219, 220, 226, 303 c-selection 3, 13, 14, 19, 26, 38, 309, 321 definiteness anaphoric 4, 9, 10, 24–26, 60fn. 4.15, 68, 88, 102–103, 121, 323 scale 11–12, 34, 46, 56, 58 unique 4, 9, 10, 24–26, 61fn. 4.16, 68, 88, 89, 99–100, 323 weak 99fn. 4.34, 181fn. 5.20 demonstratives 9, 10, 54fn. 4.11, 60fn. 4.15, 88, 323 derived kind predication 136, 183fn. 5.22 detransitivization 33, 38, 42–44, 301, 302fn. 9.1, 307, 318 differential subject marking 1fn. 1.1, 12 discourse new 139–140, 148fn. 5.8, 157, 167 discourse referentiality 185, 318–319 Distributed Morphology 16, 19–20, 45–46
edge feature 225–226 event identification 21, 22, 29, 311 event semantics 20, 116–117 existential closure 20, 22, 29–30, 117, 150, 169, 299, 300, 301, 302, 313, 316 existential sentence 78, 139–140, 147–148, 168 Extension Condition 110 feature categorial 3, 13–14, 31, 33–34, 37, 321 stack 16, 17, 19, 26, 38 focus adverb 94, 293–294 freezing 235–236, 255 function composition 112 generic 133–134, 146, 156, 177, 183 harmonic alignment 46–47 head movement 110, 113, 292, 299, 314, 316fn. 9.8 homogeneity 137–138, 167 impoverishment 33–34, 45, 46, 49, 106fn. 4.39 indefinites 4, 9, 11, 23–24, 28–29, 35–36, 59, 81, 108, 130–131, 132, 301, 310 kind 136, 176, 181fn. 5.20, 182, 183fn. 5.22 layering accounts categorial 36–37 nominal 303fn. 9.1, 310fn. 9.5 left dislocation 6–7, 223 logophoric centre 217–218 manner adverb 108, 179–180, 184, 244, 293 markedness constraint 34, 46, 47, 49, 56, 58 Merge 15–16 modification 26, 175, 307 Move 15, 18–19, 22–23 movement restrictions 31–33, 222–224, 293, 303 nominalized clause 162, 166, 202, 231 noun incorporation 33, 292, 299–300, 307 NP-ellipsis 93, 95
348
SUBJECT INDE X
number marking 13, 26, 175, 305, 307 neutrality 175–176, 307fn. 9.2 numerals 4, 9, 11, 97fn. 4.33, 134, 146, 147–148 object shift 31, 224, 316 Optimality Theory 19, 56–57 partitivity 97fn. 4.33, 138 phase balance 225 phase theory 16, 225–227, 311, 313 pluractionality operator 307–308 presupposition 25, 99, 134fn. 5.4, 181–182, 186, 217 proper name 9, 10, 54fn. 4.11, 75fn. 4.23 quantifier adverbial 150 raising 20, 117 strong, weak 4, 9, 11, 156–157 reconstruction semantic 126, 128 syntactic 126, 127, 170, 172
Restrict 299–301, 302, 311, 313–314, 316 resultative 199, 201–202, 204, 294 rightward movement 230, 235, 236, 253, 255 scope inertness 1, 39, 115–117, 129 scope rigidity 84fn. 4.29, 168, 185 scrambling 31, 84fn. 4.29, 108, 124, 125, 167fn. 5.13, 168, 200, 224–227 size restriction 1, 7, 11, 56 small clause 294–295 specificity 10, 23–24, 42, 50, 55, 59, 65, 72–73, 80, 104–105, 107, 108, 185–187, 285–286, 296–298, 310fn. 9.5, 319 surface adjacency 222, 277, 293, 296, 298, 302fn. 9.1, 322 theta-roles 13, 14, 20, 26, 34, 117 topic 52, 96, 178, 251–252 topicalization 31, 108, 182fn. 5.21, 224, 235, 237fn. 7.9, 271, 272, 274 trace conversion 22fn. 3.4 type-shifter 99, 102 weak crossover 233
O X F O R D ST U D I E S I N T H E O R ET I C A L L I N GU IST I C S PUBLISHED 1 The Syntax of Silence Sluicing, Islands, and the Theory of Ellipsis by Jason Merchant 2 Questions and Answers in Embedded Contexts by Utpal Lahiri 3 Phonetics, Phonology, and Cognition edited by Jacques Durand and Bernard Laks 4 At the Syntax-Pragmatics Interface Concept Formation and Verbal Underspecification in Dynamic Syntax by Lutz Marten 5 The Unaccusativity Puzzle Explorations of the Syntax-Lexicon Interface edited by Artemis Alexiadou, Elena Anagnostopoulou, and Martin Everaert 6 Beyond Morphology Interface Conditions on Word Formation by Peter Ackema and Ad Neeleman 7 The Logic of Conventional Implicatures by Christopher Potts 8 Paradigms of Phonological Theory edited by Laura Downing, T. Alan Hall, and Renate Raffelsiefen 9 The Verbal Complex in Romance by Paola Monachesi 10 The Syntax of Aspect Deriving Thematic and Aspectual Interpretation Edited by Nomi Erteschik-Shir and Tova Rapoport 11 Aspects of the Theory of Clitics by Stephen Anderson 12 Canonical Forms in Prosodic Morphology by Laura J. Downing 13 Aspect and Reference Time by Olga Borik 14 Direct Compositionality edited by Chris Barker and Pauline Jacobson 15 A Natural History of Infixation by Alan C. L. Yu 16 Phi-Theory Phi-Features Across Interfaces and Modules edited by Daniel Harbour, David Adger, and Susana Béjar 17 French Dislocation Interpretation, Syntax, Acquisition by Cécile De Cat 18 Inflectional Identity edited by Asaf Bachrach and Andrew Nevins 19 Lexical Plurals by Paolo Acquaviva
20 Adjectives and Adverbs Syntax, Semantics, and Discourse Edited by Louise McNally and Christopher Kennedy 21 InterPhases Phase-Theoretic Investigations of Linguistic Interfaces edited by Kleanthes Grohmann 22 Negation in Gapping by Sophie Repp 23 A Derivational Syntax for Information Structure by Luis López 24 Quantification, Definiteness, and Nominalization edited by Anastasia Giannakidou and Monika Rathert 25 The Syntax of Sentential Stress by Arsalan Kahnemuyipour 26 Tense, Aspect, and Indexicality by James Higginbotham 27 Lexical Semantics, Syntax, and Event Structure edited by Malka Rappaport Hovav, Edit Doron, and Ivy Sichel 28 About the Speaker Towards a Syntax of Indexicality by Alessandra Giorgi 29 The Sound Patterns of Syntax edited by Nomi Erteschik-Shir and Lisa Rochman 30 The Complementizer Phase edited by Phoevos Panagiotidis 31 Interfaces in Linguistics New Research Perspectives edited by Raffaella Folli and Christiane Ulbrich 32 Negative Indefinites by Doris Penka 33 Events, Phrases, and Questions by Robert Truswell 34 Dissolving Binding Theory by Johan Rooryck and Guido Vanden Wyngaerd 35 The Logic of Pronominal Resumption by Ash Asudeh 36 Modals and Conditionals by Angelika Kratzer 37 The Theta System Argument Structure at the Interface edited by Martin Everaert, Marijana Marelj, and Tal Siloni
38 Sluicing Cross-Linguistic Perspectives edited by Jason Merchant and Andrew Simpson 39 Telicity, Change, and State A Cross-Categorial View of Event Structure edited by Violeta Demonte and Louise McNally 40 Ways of Structure Building edited by Myriam Uribe-Etxebarria and Vidal Valmala 41 The Morphology and Phonology of Exponence edited by Jochen Trommer 42 Count and Mass Across Languages edited by Diane Massam 43 Genericity edited by Alda Mari, Claire Beyssade, and Fabio Del Prete 44 Strategies of Quantification edited by Kook-Hee Gil, Steve Harlow, and George Tsoulas 45 Nonverbal Predication Copular Sentences at the Syntax-Semantics Interface by Isabelle Roy 46 Diagnosing Syntax edited by Lisa Lai-Shen Cheng and Norbert Corver 47 Pseudogapping and Ellipsis by Kirsten Gengel 48 Syntax and its Limits edited by Raffaella Folli, Christina Sevdali, and Robert Truswell 49 Phrase Structure and Argument Structure A Case Study of the Syntax-Semantics Interface by Terje Lohndal 50 Edges in Syntax Scrambling and Cyclic Linearization by Heejeong Ko 51 The Syntax of Roots and the Roots of Syntax edited by Artemis Alexiadou, Hagit Borer, and Florian Scha¨fer 52 Causation in Grammatical Structures edited by Bridget Copley and Fabienne Martin 53 Continuations and Natural Language by Chris Barker and Chung-chieh Shan 54 The Semantics of Evaluativity by Jessica Rett
55 External Arguments in Transitivity Alternations A Layering Approach by Artemis Alexiadou, Elena Anagnostopoulou, and Florian Scha¨fer 56 Control and Restructuring by Thomas Grano 57 The Interaction of Focus, Givenness, and Prosody A Study of Italian Clause Structure by Vieri Samek-Lodovici 58 The Morphosyntax of Gender by Ruth Kramer 59 The Morphosyntax of Imperatives by Daniela Isac 60 Sentence and Discourse edited by Jacqueline Guéron 61 Optimality-Theoretic Syntax, Semantics, and Pragmatics From Uni- to Bidirectional Optimization edited by Géraldine Legendre, Michael T. Putnam, Henriëtte de Swart, and Erin Zaroukian 62 The Morphosyntax of Transitions A Case Study in Latin and Other Languages by Vı´ctor Acedo-Matella´n 63 Modality Across Syntactic Categories edited by Ana Arregui, Marı´a Luisa Rivero, and Andrés Salanova 64 The Verbal Domain edited by Roberta D’A lessandro, Irene Franco, and Ángel J. Gallego 65 Concealed Questions by Ilaria Frana 66 Parts of a Whole Distributivity as a Bridge between Aspect and Measurement by Lucas Champollion 67 Semantics and Morphosyntactic Variation Qualities and the Grammar of Property Concepts by Itamar Francez and Andrew Koontz-Garboden 68 The Structure of Words at the Interfaces edited by Heather Newell, Ma´ire Noonan, Glyne Piggott, and Lisa deMenaTravis 69 Pragmatic Aspects of Scalar Modifiers The Semantics-Pragmatics Interface by Osamu Sawada
70 Encoding Events Functional Structure and Variation by Xuhui Hu
82 Pseudo-Noun Incorporation and Differential Object Marking by Imke Driemel
71 Gender and Noun Classification edited by Éric Mathieu, Myriam Dali, and Gita Zareikar
PUBLISHED IN ASSOCIATION WITH THE SERIES
72 The Grammar of Expressivity by Daniel Gutzmann 73 The Grammar of Copulas Across Language edited by Marı´a J. Arche, Antonio Fa´bregas, and Rafael Marı´n 74 The Roots of Verbal Meaning by John Beavers and Andrew Koontz-Garboden 75 Contrast and Representations in Syntax edited by Bronwyn M. Bjorkman and Daniel Currie Hall 76 Nominalization 50 Years on from Chomsky’s Remarks edited by Artemis Alexiadou and Hagit Borer 77 Majority Quantification and Quantity Superlatives A Crosslinguistic Analysis of Most by Carmen Dobrovie-Sorin and Ion Giurgea 78 The Grammar of the Utterance How to Do Things with Ibero-Romance by Alice Corr 79 The Derivational Timing of Ellipsis edited by Gu¨liz Gu¨nes¸ and Anikó Lipta´k 80 Negation and Negative Dependencies by Hedde Zeijlstra 81 Angles of Object Agreement edited by Andrew Nevins, Anita Peti-Stantic´, Mark de Vos, and Jana Willer-Gold
The Oxford Handbook of Linguistic Interfaces edited by Gillian Ramchand and Charles Reiss
IN PREPARATION The Place of Case in Grammar edited by Elena Anagnostopoulou, Dionysios Mertyris, and Christina Sevdali Phi Syntax A Theory of Agreement by Susana Béjar The Left Edge Ban and the Architecture of Grammar by Kenyon Branan Cartography and Explanatory Adequacy edited by Ángel Gallego and Dennis Ott Generality and Exception by Ivan Garcia-Alvarez Definiteness in Balkan Romance by Daniela Isac Non-interrogative Subordinate Wh-clauses edited by Łukasz Jędrzejowski and Carla Umbach Syntactic Reconstruction in Minimalism by Dominique Sportiche The Syntax of Perspectival Anaphora by Sandhya Sundaresan Nominalizations in Icelandic by Jim Wood