The Syntax and Semantics of Noun Modifiers and the Theory of Universal Grammar: A Korean Perspective [1st ed.] 978-3-030-05884-5, 978-3-030-05886-9

This book takes Korean as a basis to provide a detailed universal Determiner Phrase (DP) structure. Adnominal adjectival

1,156 133 28MB

English Pages XVI, 321 [331] Year 2019

Report DMCA / Copyright

DOWNLOAD FILE

Polecaj historie

The Syntax and Semantics of Noun Modifiers and the Theory of Universal Grammar: A Korean Perspective [1st ed.]
 978-3-030-05884-5, 978-3-030-05886-9

Table of contents :
Front Matter ....Pages i-xvi
Introduction (Min-Joo Kim)....Pages 1-15
Adnominal Adjectival Classes in Korean (Min-Joo Kim)....Pages 17-65
Adjective Ordering Restrictions: The View from Korean (Min-Joo Kim)....Pages 67-96
A New Formal Analysis of AOR (Min-Joo Kim)....Pages 97-157
Capturing the Korean Facts (Min-Joo Kim)....Pages 159-236
Extensions, Refinements, and Prospects (Min-Joo Kim)....Pages 237-299
Summary and Conclusion (Min-Joo Kim)....Pages 301-313
Back Matter ....Pages 315-321

Citation preview

Studies in Natural Language and Linguistic Theory 96

Min-Joo Kim

The Syntax and Semantics of Noun Modifiers and the Theory of Universal Grammar A Korean Perspective

The Syntax and Semantics of Noun Modifiers and the Theory of Universal Grammar

Studies in Natural Language and Linguistic Theory VOLUME 96

Managing Editors Marcel den Dikken, Research Institute for Linguistics, Hungarian Academy of Sciences and Department of English Linguistics, Eötvös Loránd University, Budapest, Hungary Liliane Haegeman, University of Gent, Gent, Belgium Joan Maling, Brandeis University, Waltham, MA, USA Maria Polinsky, University of Maryland, College Park, MD, USA Editorial Board Guglielmo Cinque, University of Venice, Italy Jane Grimshaw, Rutgers University, New Brunswick, USA Michael Kenstowicz, Massachusetts Institute of Technology, Cambridge, MA, USA Hilda Koopman, University of California, Los Angeles, USA Howard Lasnik, University of Maryland, College Park, MD, USA Alec Marantz, New York University, NY, USA John J. McCarthy, University of Massachusetts, Amherst, MA, USA Ian Roberts, University of Cambridge, UK

More information about this series at http://www.springer.com/series/6559

Min-Joo Kim

The Syntax and Semantics of Noun Modifiers and the Theory of Universal Grammar A Korean Perspective

123

Min-Joo Kim Department of English Texas Tech University Lubbock, TX, USA

ISSN 0924-4670 ISSN 2215-0358 (electronic) Studies in Natural Language and Linguistic Theory ISBN 978-3-030-05884-5 ISBN 978-3-030-05886-9 (eBook) https://doi.org/10.1007/978-3-030-05886-9 Library of Congress Control Number: 2018964255 © Springer Nature Switzerland AG 2019 This work is subject to copyright. All rights are reserved by the Publisher, whether the whole or part of the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations, recitation, broadcasting, reproduction on microfilms or in any other physical way, and transmission or information storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar methodology now known or hereafter developed. The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication does not imply, even in the absence of a specific statement, that such names are exempt from the relevant protective laws and regulations and therefore free for general use. The publisher, the authors, and the editors are safe to assume that the advice and information in this book are believed to be true and accurate at the date of publication. Neither the publisher nor the authors or the editors give a warranty, express or implied, with respect to the material contained herein or for any errors or omissions that may have been made. The publisher remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. This Springer imprint is published by the registered company Springer Nature Switzerland AG The registered company address is: Gewerbestrasse 11, 6330 Cham, Switzerland

To my beloved and loving daughter Serin

Foreword

There are some puzzles that, despite years of careful study and incremental progress, leave one longing for a deeper understanding. In generative grammar and formal semantics, the relative order of adnominal modifiers has a claim to being foremost among them. To be sure, a great deal of real progress has been achieved over the years, and the topic has garnered sustained examination in many places. Among many others, this includes Cinque’s landmark 2010 volume on adjectives, which one might regard as a kind of intellectual ancestor of this work. And of course, it has been a topic of typological research and research outside of the formal linguistic tradition. And yet, after all this, I can’t really claim to know why—at least in some truly deep sense of ‘why’—language after language seems to agree that an adjective of shape should occur higher in its nominal than an adjective of color. A priori, it’s a baffling fact. Why would languages be designed this way and not another? Why, in all their diversity, would they agree on this, something so seemingly arbitrary? In this work, Min-Joo Kim faces this problem and its cousins head-on, but in a particular instantiation that isn’t widely familiar. She approaches the issue from the vantage point of Korean. With a wealth of new data and important new observations, Kim sheds new light on old puzzles and offers fresh new ones that will no doubt continue to vex us for years. She proposes a generalized universal conception of the nominal extended projection, working broadly in the cartographic tradition, but characterized in a way that should be relevant to researchers approaching the topic from a variety of perspectives. For example, she splits the DP into high, middle, and low fields, each of them associated with particular families of syntactic and semantic roles. The terms are pretheoretical, and the work done in each syntactic neighborhood (the resolution of reference, quantification, and predication) could be fruitfully mapped onto non-cartographic conceptions of nominal structure. In one respect, the choice of Korean is surprising. Kim herself has in previous work questioned whether Korean has adjectives as an open-class lexical category at all. Nor does the language provide the richness of structural landmarks within the nominal extended projection that morphologically richer languages do, or indeed even ones with overt articles. But of course, this is precisely why there is a certain vii

viii

Foreword

kind of work that truly is best carried out by linguists working on their native languages. The required analytic subtlety, coupled with a corresponding subtlety of intuitions, may simply be unavailable to anyone less expert. The great problem in the distribution of modifiers in general—adjectives and adverbs both—is that the correspondence between position and interpretation is undeniable, clear, and yet vexingly imperfect. The facts defy clear, unambiguous characterization. There always seem to be exceptions, and often it’s not obvious if they are principled exceptions. It’s a risky domain in which to maneuver analytically. Kim faces this problem in part by taking an expansive view. She takes into account not just the relation between the syntactic properties of a modifier and its semantics, but also their phonology. Nor does she confine herself to the traditional tools of cartographic syntax: Optimality-theoretic principles also make an appearance. It’s this kind of ecumenical attitude that makes this work so potentially useful. The details of one analysis or another rarely survive perfectly intact over the decades, but the underlying analytical intuitions are lasting, and often influential long after the tides of intellectual history have turned and turned again. It’s largely for this reason that Jespersen and Bolinger are still widely cited. The analysis of the Korean extended DP—and of nominals across languages—offered here is of course important and interesting, but the insights that underlie it are where this volume ensures its lasting relevance. East Lansing, MI, USA

Marcin Morzycki

Preface

What does nominal structure look like in any given language? What fills it? And how and why? This book is an attempt to answer these questions by proposing a universal theory of noun modifiers from the perspective of Korean. Trying to construct a universal theory of noun modifiers from the perspective of a language which allegedly lacks an open class of adjective is seemingly an oxymoron. But if one firmly believes in Universal Grammar, then there cannot be a better testing ground. The genesis of this book goes all the way back to Fall 2000, when I took the proseminar on modifiers taught by Lisa Matthewson at UMass Amherst. If someone asked me at that time whether I believed Korean has the category of adjective or not, I would have said it does not—which is in fact the answer I gave in my 2002 MIT Working Papers in Linguistics paper. If someone asks the same question now, my answer would be it doesn’t really matter because Korean has other rich resources to express a full spectrum of adnominal meanings and the same is probably true of every other human language. In languages like Korean, relative clauses play an instrumental role in expressing various adnominal meanings, but identifying where exactly they are located inside what we call a determiner phrase (DP) is not a trivial task. Similarly, locating non-clausal, lexical noun modifiers inside a DP along with relative clauses and functional elements like demonstratives, articles, and classifiers is far from a simple task. Possibly for this reason, there isn’t a single, comprehensive work out there which provides an overview of the syntax and semantics of all three types of noun modifiers while also capturing their constituent order variation across languages within a formal framework. This is the gap the present work intends to fill, and I hope that I have succeeded in doing so, albeit partly. Writing this book has been a long and difficult process largely because noun modifiers are elusive species but more so because trying to look for adjective-like elements in a language that lacks a prototypical category of adjective can easily turn out to be futile efforts. That said, even though truth is still far away and the jury is still out, I now know so much better than before, so am thankful for having been given the opportunity to conduct this research. Whatever outcome this book has ix

x

Preface

yielded, it was only possible because it is standing on the shoulders of giants like Bolinger, Sproat and Shih, Cinque, Larson, Laenzlinger, and Svenonius. The intellectual debt it owes to their work is evident in every single page, and although I disagree with them in some places, I agree with them in all other places. Conducting the present research was also possible because of the financial support from the 2010 Academy of Korean Studies grant (#AKS-2010-R36), Texas Tech University’s 2013 CAHSS grant, a 2018 TTU Humanities Center Alumni College Fellowship, and a 2018 National Humanities Center Summer Residency Fellowship, all of which I gratefully acknowledge here. Receiving the Faculty Development Leave award in Fall 2012 was crucial in making headway with this book project. Therefore, I thank Texas Tech and the Texas Tech University System for yet another reason. I’m extremely grateful to the three anonymous reviewers of Studies in Natural Language and Linguistic Theory (SNLT) for their detailed and constructive feedback, and the SNLT series editors and the editorial staff for their constant support, encouragement, and assistance over the years. I learned a great deal from teaching the noun modifier seminar in Spring 2012 at Texas Tech University, so would like to take this opportunity to thank my graduate students in that seminar, in particular Rita Jhang and Olga Pahom. Thanks are also due to the participants and organizers of the meetings I presented parts of this work at: the Linguistics Colloquium at the Graduate Institute of Applied Linguistics (February 2011), the Korean Society of Language and Information Meeting (June 2012), the 31st WCCFL conference at Arizona State University (February 2013), the TTU Cognitive Brown Bag Series (May 2015), the 18th SICOGG conference at Sogang University, Seoul, Korea (August 2016), and the Semantics 2018: Looking Ahead workshop at UMass Amherst (March 2018). I’m especially indebted to Professors Steve Parker, Seungho Nam, and Angelika Kratzer for respectively inviting me to the first, the second, and the last events. Last but not least, I’m ineffably grateful to Marcin Morzycki for graciously writing the foreword of this book for me, and Lisa Matthewson for teaching the proseminar on modifiers in Fall 2000, which undoubtedly planted fruitful seeds in many of us. Every single book that is written by a woman who is also a mom mentions their children and I won’t be an except to that. Officially writing this book started in 2012 and since then, I have been mostly living in some atemporal world that my daughter had no access to. Yet she grew with this book and is still happy and healthy, and she just turned 11. I dedicate this book to her with love and thanks which no noun modifier in human language can begin describing. In the midst of all this, there was my husband Sung-Won, so many thanks are also due to him for his unwavering love and support. Lubbock, TX, USA October 2018

Min-Joo Kim

Contents

..... .....

1 1

..... ..... .....

4 6 7

.....

7

..... .....

9 13

...... ......

17 17

. . . . . . . . . .

22 24 29 33 39 46 54 59 63 63

1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.1 Setting the Stage . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.2 Aims, Motivations, and the Gist of the Analysis to Be Proposed . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.3 Organization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.4 Some Remarks on the Technical Terms to Be Used . . . . . . 1.4.1 What It Means for an ADJ to Be Intersective, Subsective, Modal, or Privative . . . . . . . . . . . . . . . 1.4.2 On the Predicative Versus Adnominal/Attributive, Intersective Versus Non-intersective, and Restrictive Versus Non-restrictive Distinctions . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2 Adnominal Adjectival Classes in Korean . . . . . . . . . . . . . . . 2.1 Some Background on Korean . . . . . . . . . . . . . . . . . . . . . 2.2 Major Korean Adjectival Classes and Their Characteristic Properties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.2.1 Adjectival Prefixes . . . . . . . . . . . . . . . . . . . . . . . . 2.2.2 Adjectival Ns . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.2.3 Attributive Determinatives . . . . . . . . . . . . . . . . . . 2.2.4 Expressions Ending in -Cek . . . . . . . . . . . . . . . . . 2.2.5 UN-XPs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.2.6 RCs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.3 Summary and Discussion . . . . . . . . . . . . . . . . . . . . . . . . 2.4 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

xi

xii

3 Adjective Ordering Restrictions: The View from Korean . 3.1 Correlation Between the Morpho-Syntactic Complexity and the Interpretive Possibilities of an N Modifier . . . . 3.2 Relative Ordering Among N Modifiers in Korean . . . . . 3.3 AOR on Simplex N Modifiers . . . . . . . . . . . . . . . . . . . 3.4 Relative Order Between DEM, NUM, and ADJ . . . . . . 3.5 What to Make of Our Findings . . . . . . . . . . . . . . . . . . 3.6 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Contents

........

67

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

70 72 80 87 88 94 95

4 A New Formal Analysis of AOR . . . . . . . . . . . . . . . . . . . . . . . 4.1 Lessons from Cinque (2010) and Larson (1998, 2000) . . . . 4.2 Lessons from Svenonius (2008) and Laenzlinger (2005) . . . 4.2.1 Svenonius’ (2008) Decompositional Analysis of DP 4.2.2 Laenzlinger’s (2005) Complex-DP Analysis . . . . . . 4.3 The New Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.3.1 Three-Way Classification of Nominals . . . . . . . . . . 4.3.2 Three Types of N Modifiers and Three Merge Sites Inside DP . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.3.3 The Full DP Internal Structure and a New Decompositional Split-DP Analysis . . . . . . . . . . . . 4.3.4 Interim Summary . . . . . . . . . . . . . . . . . . . . . . . . . . 4.4 Some Immediate Welcome Results of the New Analysis . . . 4.5 More on the “Low” Origin of DEMs and the Position of FocP . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.6 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

97 98 102 102 109 113 113

. . . . . . .

. . . . . 116 . . . . . 122 . . . . . 132 . . . . . 134 . . . . . 146 . . . . . 154 . . . . . 154

5 Capturing the Korean Facts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.1 Application of the Proposed Analysis to Korean: Some Positive Outcomes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.2 Challenges for a Derivational Analysis . . . . . . . . . . . . . . . . . . 5.2.1 Challenge 1: Surface Position of Certain UN-APs Relative to Other Lexical ADJs . . . . . . . . . . . . . . . . . . 5.2.2 Challenge 2: Ordering Between RCs . . . . . . . . . . . . . . . 5.3 Solving the Problems by Adding an Output Filter . . . . . . . . . . 5.3.1 Why Certain Lexical APs Surface in Such “Unexpected” Positions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.3.2 Ranking of the Constraints on AO in Korean . . . . . . . . 5.3.3 Solving the Problems Presented by Some UN-AP and ATT-DET/CEK-AP Co-occurrences . . . . . . . . . . . . 5.3.4 Ordering Restrictions on FRCs in Korean . . . . . . . . . . . 5.3.5 An Attempt to Formally Derive the Ordering Restrictions on Korean RCs . . . . . . . . . . . . . . . . . . . . .

. . 159 . . 161 . . 173 . . 173 . . 176 . . 179 . . 180 . . 183 . . 186 . . 194 . . 201

Contents

xiii

5.4 Why Co-occurring with Ku ‘the/that’ Engenders a Non-canonical ADJ Order . . . . . . . . . . . . . . . . . . . . . . . 5.4.1 The Cognitive Indexical Properties of Ku at Hand . . 5.4.2 Formally Capturing the “Special” Properties of KU . 5.5 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

207 208 221 232 234

. . . . . . . 237 . . . . . . . 237 . . . . . . . 241

6 Extensions, Refinements, and Prospects . . . . . . . . . . . . . . . 6.1 Ordering Restrictions on Mandarin FRCs . . . . . . . . . . . . 6.2 Integrated Non-restrictive RCs in Mandarin and Korean . 6.3 What Constitutes the Semantics of DEMs: The View from Korean and English . . . . . . . . . . . . . . . . . . . . . . . 6.4 More on the Internal Structure of a DemP: Is a Uniform Analysis Possible? . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6.5 Supplementary Versus Non-restrictive RC Distinction in Prenominal Modifier Languages . . . . . . . . . . . . . . . . 6.6 What Determines the Interpretive Possibilities of APs? . . 6.7 How Many FocPs Inside a DP? . . . . . . . . . . . . . . . . . . . 6.8 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

277 283 290 294 295

7 Summary and Conclusion . . . . . . . . . . . . . . . . . . . . . . . 7.1 Summary of the Main Findings and Claims . . . . . . . 7.2 What All This Leads Us To . . . . . . . . . . . . . . . . . . 7.3 Understanding (Korean) N Modifiers in a New Light 7.4 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

301 302 304 305 310 311

. . . . . .

. . . . . .

. . . . . .

. . . . . . . 250 . . . . . . . 267

Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 315

Abbreviations

3RD ACC ADN ANT ART ASSRT AUX CAUS CL CONN CONT.TOP COP DAT DECL DEF DEM EXCL EXP FMN FUT GEN GOAL HON HS IMP IMPRF IND INFL INFRML LOC

3rd person Accusative case Adnominal marker Anterior Article Assertive Auxiliary Causative marker Classifier Connective Contrastive topic Copula Dative case Declarative sentence ending Definite Demonstrative Exclamative Experiential Feminine Future Genitive case Goal Honorific Hearsay Imperative Imperfective Indicative mood Inflection Informal style Locative

xv

xvi

MSC NEG NML NOM N.PST PASS PL PRF PST QUOT REL RTRO SG TOP VS

Abbreviations

Masculine Negation Nominalizer Nominative case Non-past Passive Plural Perfective Past Quotative Relative clause marker Retrospective mood Singular Topic marker Verbal stem

Chapter 1

Introduction

1.1

Setting the Stage

Adnominal adjectival expressions are presumably not required by grammar, as evidenced by the apparent optionality of the adjective phrase (AP) really interesting and the relative clause (RC) who studies physics at MIT in (1). (1) Yesterday, John met a ([AP really interesting]) girl ([RC who studies physics at MIT]).

Yet their syntax and semantics have been shown to provide an important window on the internal structure of Determiner Phrase (DP) and the architecture of human language in general (see, a.o., Larson 1998, 2000; Bouchard 2002; Giusti 2005; Laenzlinger 2005; Svenonius 2008; Cinque 1994; 2005, 2010; Alexiadou et al. 2007, Part III; Ramaglia 2011; Cabredo Hofherr and Matushansky 2010; and the references there). There are by and large four well-established typological generalizations on the syntax and semantics of noun (N) modifiers that are relevant for our purposes. First of all, when a demonstrative (DEM), a numeral (NUM), and an adjective (ADJ) co-occur modifying the same N, they almost always occur in the order of ‘DEM > NUM > ADJ’, where > indicates syntactic hierarchy rather than linear precedence (see, e.g., Greenberg 1963; Hawkins 1983; Dryer 1992; Cinque 2005). This can be seen by the fact that the English example (2a) is grammatical but its variants with alternate word orders in (2b, c) are not. (2) a. those two interesting books b. *those interesting two books c. *interesting those two books

(DEM > NUM > ADJ) (DEM > ADJ > NUM) (ADJ > NUM > DEM)

© Springer Nature Switzerland AG 2019 M.-J. Kim, The Syntax and Semantics of Noun Modifiers and the Theory of Universal Grammar, Studies in Natural Language and Linguistic Theory 96, https://doi.org/10.1007/978-3-030-05886-9_1

1

2

1

Introduction

Secondly, when multiple ADJs co-occur modifying the same N, they typically occur in the order of ‘quality ADJ > size ADJ > shape ADJ > color ADJ > N’, as exemplified in (3) (e.g., Whorf 1945; Sproat and Shih 1988, 1990; Cinque 2010). And such adjective ordering restrictions (AORs) have been taken to suggest that ADJs describing ‘absolute’ properties (e.g., shape, color, or source terms) must occur closer to the N than those describing ‘relative’ properties (e.g., quality or size terms), or that N modifiers occurring higher in the DP structure require a more comparative or evaluative semantic computation on the part of the language user than those occurring lower in the structure (Sproat and Shih 1988, 1990; see also Cinque 2010). (3) a. the big red vase b. *the red big vase

(size ADJ > color ADJ > N) (color ADJ > size ADJ > N)

Thirdly, it is widely held that the AORs illustrated by (3) only hold for “true” ADJs which are morpho-syntactically simplex. That is, they do not hold for morpho-syntactically complex N modifiers (e.g., RCs) or those that have a clausal origin (see, a.o., Sproat and Shih 1988, 1990; Cinque 2010). To illustrate this, in Mandarin, mono-morphemic ADJs occur in ‘size > color’ orders, as shown in (4), but when converted into bi-morphemic expressions as a result of having the adnominalizing morpheme -de1 attached to them, they may occur in the reversed orders, as shown in (5). (4) a.

xiao lu hua-ping small green vase ‘small green vase’ b. * lu xiao hua-ping

(5) a.

xiao-de lu-de small green ‘small green vase’ b. lu-de xiao-de

(size > color)

(color > size) [Sproat and Shih 1988: ex. (3a, a’)]

hua-ping vase

(size > color)

hua-ping

(color > size) [Sproat and Shih 1988: ex. (2a, a’)]

Lastly, in numerous languages, the surface position of an adnominal ADJ tends to correlate with its semantics (see, e.g., Whorf 1945; Bolinger 1967; Vendler 1968; Sproat and Shih 1988, 1990; Larson 1998; Bouchard 2002; Larson and Marušič

1

This morpheme marks RCs in Mandarin, so the ADJs in (5) are considered as disguised or reduced RCs (Sproat and Shih 1988, 1990; Cinque 2010).

1.1 Setting the Stage

3

2004; Svenonius 2008; Cinque 2010; Ramaglia 2011). For example, in Romance and some Germanic languages, if an ADJ carries what Cinque (2010) calls ‘direct’ modificational (Mod) semantics, i.e., N-dependent, subsective, generic, characteristic, intensional, attributive, and/or reference-modifying in the sense of Bolinger (1967), then it almost always occurs prenominally; if it occurs postnominally, then it almost always carries what he calls ‘indirect’ Mod semantics, i.e., N-independent, intersective, non-generic, occasion-based, extensional, predicative, and/or referent-modifying in the sense of Bolinger.2 This is illustrated by the French data in (6): here, the ADJ grand receives a different construal, depending on whether it occurs prenominally or postnominally, i.e., ‘great’ versus ‘tall’ (Bouchard 2002). (Here and below, Art is short for article.) (6) a.

un grand homme a big man ‘a great man’ b. un homme grand a man big ‘a tall man’

(Art > N > ADJ)

(Art > ADJ > N)

Similarly, English ADJs may receive different interpretations depending on whether they occur before or after an N. This can be seen by the interpretive difference between (7a) and (7b), where the ADJ navigable occurs: (7a) is construed as referring to a river that is inherently navigable but (7b) is construed as referring to a river that is navigable on a particular occasion (Bolinger 1967). (7) a. b.

a navigable river a river navigable

(Art > ADJ > N) (Art > N > ADJ)

In recent years, such typological generalizations and the relevant cross-linguistic findings have led generative linguists to posit that different types of N modifiers are licensed by different types of functional phrases (FPs) which form a hierarchical structure inside a DP, contra the older beliefs held by authors like Chomsky (1965), Smith (1961, 1964), Jacobs and Rosenbaum (1968), Lakoff (1971), and Kayne (1994)3 (see, a.o., Cinque 1994, 2010; Scott 2002; Laenzlinger 2005; Svenonius 2008). Yet there is still a great deal of disagreement among authors as to precisely what kinds of FPs are present inside a DP, what their relative positions are, and how 2

In non-Romance or non-Germanic languages also, similar observations have been made about the correlation between the syntax and the semantics of adnominal ADJs. In Greek for example, only intersective or referent-modifying ADJs participate in what is known as the determiner spreading phenomenon: in this phenomenon, not only the head N but also its adjectival modifiers co-occur with the definite article i (see, a.o., Alexiadou and Wilder 1998; Plank 2003; Campos and Stavrou 2004; and Alexiadou 2014). 3 Such authors hold that (English) adnominal ADJs have a clausal origin; that is, they derive from RCs but appear prenominally via transformation (i.e., movement).

4

1

Introduction

exactly different surface orders involving DP-internal constituents come about, giving rise to different interpretive possibilities (for a good overview, see Alexiadou et al. 2007, Part III). Another outstanding question is whether it will be possible to capture AORs in various languages by assuming the same underlying DP structure and by applying the same transformational mechanisms (i.e., Move and Merge but nothing else).4 If the answer is in the negative, then the ensuing questions will include what the scope of the relevant cross-linguistic variation is and what is responsible for such a variation other than purely historical and/or lexical reasons.

1.2

Aims, Motivations, and the Gist of the Analysis to Be Proposed

This book aims to answer the above questions primarily through the lens of Korean but its ultimate goal is to present a viable universal DP structure and to help identify principles that govern the mapping between the surface position, the form, and the meaning of well-formed linguistic expressions. Korean provides a unique opportunity to address these questions because its AOR phenomenon has not been much studied in the literature.5 But there are other important reasons as well. First of all, Korean allegedly lacks an open class of ADJ (see, e.g., Martin 1992; Suh 1996; Yu 1998; Maling and Kim 1998; Kim 2002) and this raises the question of how such a language may express various types of adnominal meanings and how its N modifiers may behave with respect to various typological generalizations that have been made in the literature if they are applicable at all. Secondly, all adnominal categories in Korean occur prenominally, so it is not easy to differentiate between “true” APs and reduced RCs (RRCs), unlike the case with relatively well studied languages like English, Spanish, and Italian (more on this in Chap. 2); in the latter type of languages, “true” APs and RRCs have a distinct syntactic status, either linearly or hierarchically (see Cinque 2010 and the references there). Finally, Korean lacks both indefinite and definite articles and this, in conjunction with the prenominal-only positional possibilities for all its N modifiers, makes it extremely difficult to identify the exact syntactic positions of its N dependents, especially when compared to article-possessing languages like English and Italian. In what follows, I show that there is a correlation between the position of N modifiers and their semantics in Korean and the surface position of a Korean N modifier is determined in part by its morpho-syntactic complexity but in this language, being a morpho-syntactically complex N modifier does not necessarily

4

For recent such attempts, see, a.o., Svenonius 2008 and Cinque 2010. At the time of writing this book, Kang 2006 and Kim 2014a, b are the only published works that are devoted to analyzing the AOR phenomenon in Korean.

5

1.2 Aims, Motivations, and the Gist of the Analysis to Be Proposed

5

mean carrying indirect Mod semantics, unlike what has been claimed in the literature (e.g., Kang 2006; Cinque 2010). To capture such a partial correspondence between the form and the meaning of N modifiers in Korean as well as the partial ordering restrictions on them, I look at the phenomena at hand not only from morpho-syntactic and semantic standpoints but also from a phonological standpoint, and I offer a cartographic analysis which combines the insights of Laenzlinger’s (2005) and Svenonius’ (2008) analyses, among others, while also integrating lessons from other classical works on this topic (e.g., Bolinger 1967; Sproat and Shih 1988, 1990; Larson 1998, 2000; Cinque 1994, 2005, 2010). The main thesis to be proposed will be that the universal DP structure is comprised of three sub-DP layers, which I label as the High Field, the Middle Field, and the Low Field, and all DP-internal elements including lexical ADJs, articles, DEMs, and focused constituents originate from what I call the Low Field, as schematically represented in (8). Here, DPd/r, DPq, and DPp stand for DPdeictic/referential, DPquantificational, and DPpredicative, respectively, and FocP and LocP stand for Focus Phrase and Locative Phrase, respectively (more on this in Chap. 4). (8) Three fields inside a full-fledged DP under the new split-DP analysis: [DPq [DPp [FocP [LocP [UnitP/PlP [SortP [nP [√P N]]]]]]]]] [DPd/r [+referential;+/-deictic; +/-definite] [+quantificational]

[+predicative]

High Field

Low Field

Middle Field

I will also claim that any focused element may occur at [Spec, FocP], and what I call LocP introduces DEMs and articles as well p as possessive determiners. Furthermore, I borrow UnitP/PlP, SortP, nP, and P from Svenonius (2008) but assign slightly different functions p to them. For example, under my analysis, not only [Spec, nP] but also [Spec, P] may license thematic APs though the thematic roles they license are not identical to each other, as we will see in Chap. 4. As a consequence of positing such a structure, the present analysis will be shown to correctly predict that in any given language, no more than three ADJs typically occur inside the same minimal DP boundary and when they co-occur, they do so in the order of ‘focused ADJ > sortal ADJ > thematic ADJ’ at the underlying level. Another notable prediction of this analysis will be that depending on whether the DEM that co-occurs with the ADJs inside the same DP receives a focal stress or not, it may occur preceding or following the ADJs, and this will let us account for DP-internal constituent order variation in languages like Spanish and Italian in ways similar to but different from what has been suggested by other low-DEM origin analyses (e.g., Brugè 1996, 2002; Bernstein 1997, 2001). In this book, I also offer a new analysis of DEMs in Korean as well as in several other languages. In particular, a new affective/emotive meaning of DEMs will be identified in addition to a new set of binary features that constitute DEM meanings in human language. In addition, I will map out the position of various types of RCs relative to other N dependents, a topic that most (if not all) existing analyses of AOR have left open.

6

1

Introduction

In brief, I will argue that integrated RCs (which encompass both restrictive and nonrestrictive RCs) have three different merge sites, as depicted in (9a), and supplementary (Sppl) RCs select for a TP and the head of a SpplP hosts a comma in the sense of Potts (2003/2005), as sketched in (9b). (9) Position of various types of RCs under the new analysis: a. Position of integrated RCs: [DPq RC [DPp RC [FocP [LocP … [√P N]]]]]] [DPd/r RC High Field Middle Field Low Field b. Position of a Sppl RC: [SpplP RC [Sppl′ [Sppl , ] [TP …]]]

The proposed way to classify and analyze RCs will be shown to let us address some of the questions that have not received much attention in the literature such as why in languages like Mandarin and Korean, apparently restrictive RCs can modify proper names as well as DEM-marked nominals, and why Mandarin or Korean-type languages seem to lack English-type Sppl RCs.6 Furthermore, we will see that the decompositional split-DP analysis put forth here successfully implements various two-way distinctions that have been made on N modifier semantics such as Bolinger’s (1967) reference- versus referent-modifying distinction, Sproat and Shih’s (1988, 1990) and Cinque’s (2010) direct versus indirect N modifier distinction, and Larson’s (1998, 2000) NP-inner versus NP-outer spatial distinction.

1.3

Organization

This book is organized as follows: Chap. 2 introduces major adjectival classes in Korean, providing some relevant background information on Korean along the way. This chapter also takes up the question of how Korean expresses various types of ADJ meaning in the absence of a prototypical ADJ class, a question that has not been taken up in the extant generative linguistics literature.7 Chapter 3 takes a close look at how Korean behaves with regard to prevailing typological generalizations on adnominal adjectival categories, in particular the generalizations on AOR and the mapping between the form and the meaning of N modifiers in human language.

6

Lin and Tsai (2015) take up the question of why integrated RCs in Mandarin may modify proper names or DEM-marked nominals and they also cite numerous other relevant works on the topic but, to my knowledge, the question on the presence or absence of Sppl RCs in Mandarin has not been addressed in the literature. And in the case of Korean, such questions have not been addressed in the previous analyses at all. 7 For this reason, the content of this chapter will prove to be informative to anyone interested in the semantics of adjectival N modifiers but readers who are not so interested in the topic or those who do not wish to have such detailed information on Korean may skip this chapter without much harm being done in comprehending subsequent chapters.

1.3 Organization

7

Chapter 4 presents the new decompositional split-DP analysis briefly outlined above and demonstrates how it explains facts surrounding N modification in English as well as data drawn from Spanish, Italian, and French which involve focused DEMs and ADJs. Chapter 5 applies the proposed analysis to Korean and shows that when it comes to discourse-oriented, article-less languages like Korean, accounting for the relative ordering of DP-internal constituents may require combining a derivational mechanism with a representational one such as an optimality theoretic mechanism, as has been demonstrated by authors like Choi (1996) and Lee (2006), although such authors are concerned with different kinds of data than here, namely, data involving clausal structure and case marking, respectively. Chapter 6 extends the analysis developed in Chaps. 4 and 5 to other languages and other linguistic phenomena including RC ordering in Mandarin, the syntax and semantics of the article/DEM-like lexeme ta in Slovenian, and the apparent positional freedom of Romanian ADJs, and discusses its outcomes and prospects. Finally, Chap. 7 summarizes and concludes the book with questions and implications for future research.

1.4

Some Remarks on the Technical Terms to Be Used

Before proceeding to Chap. 2, I would like to make a few remarks on some of the technical terms to be used in this book, making some terminological clarifications and differentiations along the way.

1.4.1

What It Means for an ADJ to Be Intersective, Subsective, Modal, or Privative

In characterizing the semantics of an ADJ or adjectival N modifier, I will frequently resort to the four-way classification that is commonly used in the formal semantics literature, namely, the idea that ADJs fall into (i) intersective, (ii) subsective, (iii) modal, and (iv) privative sub-types or classes (see Kamp and Partee 1995; Partee 1995). As one can see from Table 1.1, this four-way taxonomy is based largely on set-theoretic semantic properties exhibited by an NP that is comprised of an ADJ and a common noun (CN). More concretely, an ADJ is considered intersective if its meaning holds irrespective of the meaning of the CN it modifies. For example, a carnivorous X is always carnivorous (i.e., ‘meat eating’) regardless of whether the X is pronounced as animal, fish, mammal, monkey, or plant. So intersective ADJs are considered N-independent and can be readily handled in an extensional semantics framework.

8

1

Introduction

Table 1.1 Four-way formal semantic classification of ADJ meaning Type

Examples

MP

Justification for the MP

Intersective

carnivorous, yellow, rectangular, German skillful, good, remarkable, typical

[[ADJ N]] = [[ADJ]] \ [[N]]a

A carnivorous animal is both carnivorous and is an animal.

[[ADJ N]]  [[N]] ([[ADJ N]] 6¼ [[ADJ]] \ [[N]])

A skillful doctor is skillful as a doctor but may not be as something else (e.g., as a teacher). An alleged murder may or may not be a murderer.

Subsective

Modal

alleged, None arguable, likely, potential, putative, questionable Privative former, [[ADJ N]] \ [[N]] = ∅ A fake gun is not a gun. fictitious, imaginary, past, fake, counterfeit a Unless otherwise specified, [[a]] stands for a function which delivers the extension of a linguistic expression or syntactic category a.

Unlike intersective ADJs, subsective ADJs (e.g., skillful) are N-dependent. This is exemplified by the fact that while a carnivorous doctor who is also a teacher is a carnivorous teacher, a skillful doctor who is also a teacher may not necessarily be a skillful teacher. Such a semantic property of subsective ADJs can be represented by the meaning postulate (MP) given in Table 1.1, namely that an NP comprised of a subsective ADJ plus a CN denotes a subset of the set denoted by that CN, rather than the intersection between the two sets denoted by the CN and the ADJ. What are classified as modal ADJs (e.g., alleged) are also non-intersective but unlike subsective ADJs, the properties they describe cannot be ascribed to an individual based solely on the state of affairs in the actual world. This is illustrated by the fact that while a skillful doctor is still a doctor, an alleged doctor may not be a doctor. And for this reason, modal ADJs are treated differently from subsective ones in the literature, and I will treat them differently as well. Finally, while the extension of an NP comprised of a modal ADJ and a CN may intersect with the extension of that CN depending on what the actual world looks like, the extension of an NP made up of a privative ADJ (e.g., former) and a CN never does. For example, while an alleged murderer may actually turn out to be a murderer, a former senator is no longer a senator in the actual world, so the intersection between the set denoted by the ADJ and the one denoted by the CN is always empty, and this justifies a separate treatment of privative ADJs from modal ones.8 8

I should note that this four-way classification of ADJs has recently been challenged. For example, Partee (2009a, b) claims that all ADJs are either subsective or modal. Furthermore, authors do not necessarily agree on the precise semantic types of the four ADJ meanings: some hypothesize that intersectives denote first-order predicates which denote sets as their extension and properties as

1.4 Some Remarks on the Technical Terms to Be Used

1.4.2

9

On the Predicative Versus Adnominal/Attributive, Intersective Versus Non-intersective, and Restrictive Versus Non-restrictive Distinctions

In describing the syntactic and semantic function of an ADJ, authors typically resort to dichotomous distinctions such as (i) predicative versus adnominal, (ii) predicative versus attributive, (iii) intersective versus non-intersective, and (iv) restrictive versus non-restrictive. But such terms are often used without clear definitions offered for them, and this may create a bit of confusion on the part of the reader (for relevant discussion, see Alexiadou et al. 2007, Part III, Ch. 1). Hence, there arises the need for clarifying what they will mean for us in the context of this book. First off, I will use the term ‘adnominal’ interchangeably with ‘attributive’ and use the term ‘predicative’ in contrast with both of them (compare Bolinger 1967). More specifically, I will consider an ADJ to be predicative if it occurs as part of a sentential predicate. To exemplify this by using English data, predicative ADJs occur as the complement to a linking verb (e.g., be, appear, look, and seem) as shown in (10); adnominal or attributive ADJs occur inside a nominal projection, modifying some nominal meaning, regardless of the syntactic size or semantic type of that nominal projection, as illustrated in (11). (10) Predicative use of an ADJ: a. The student is very kind. b. This problem appears challenging. c. John’s father was really proud of his son. d. John is polite. (11) Adnominal/attributive use of an ADJ: a. The (very) kind student even showed me where the building was. b. That challenging problem was eventually solved by Jill. c. The proud father was actually John’s father. d. I’m looking for a polite student.

Notably, in the literature, some DP-internally occurring ADJs have been treated as predicative. For example, Alexiadou et al. (2007) state that postnominally occurring N modifiers in Romance languages such as gentile and redondo in (12) are predicatively used because they are construed as intersective N modifiers.

their intension whereas subsectives, modals, and privatives denote higher-order operators which denote functions from intensions of CN phrases to intensions of CN phrases (for details see Partee 1995). On the other hand, authors like Larson (1998) have suggested that all four types of ADJ denote first-order predicates. Despite such controversy among authors, the four-way taxonomy of ADJ meaning outlined here is still widely used in describing the semantics of adnominal ADJs (see Cinque 2010; Morzycki 2016), so I adopt it here as well, but in Chap. 2, I briefly discuss why one may wish to streamline it in the way Partee (2009a, b) does.

10

1

(12) a. [DP une personne (très) gentille] a person (very) polite ‘a (very) polite person’ b. [DP el sombrero redondo] the hat round ‘the round hat’

Introduction

(French)

(Spanish)

[Alexiadou et al. 2007: 329, ex. (59b, c)]

But for our purposes, such remarks should be reinterpreted as follows: some postnominal ADJs in Romance languages are disguised or reduced RCs. And this means that they actually occur occupying the predicate position inside a clausal N modifier.9 That said, given the way ‘predicative’ is used in the extant literature, in order to use the term consistently while also capturing the semantics of postnominal modifiers such as those in (12), I adopt the ‘reference- versus referent-modifying’ semantic distinction suggested by Bolinger (1967), thereby refraining from using the term ‘predicative’ wherever possible (see Bouchard 2002 for a similar move). According to Bolinger, an ADJ is ‘reference-modifying’ if it qualifies the property denoted by the head N rather than the entire nominal projection’s denotation, so reference-modifying ADJs give rise to what are called ‘characteristic’ or ‘generic’ interpretations in the literature (e.g., Larson 1998; Cinque 2010). On the other hand, a ‘referent-modifying’ ADJ ascribes a property to the individual that is denoted by the entire DP that contains it, so referent-modifying ADJs engender what are referred to as ‘occasion-based’ or ‘non-generic’ interpretations. Turning now to the intersective versus non-intersective distinction, I will analyze an ADJ to be intersective if the MP for intersective ADJs in Table 1.1 holds for it; if not, it will be considered as non-intersective (i.e., subsective, modal, or privative). Given this, the postnominal ADJs in (12) do in fact instantiate intersective N modifiers [as correctly analyzed by Alexiadou et al. (2007)] because if X is a polite person, then X is both a person and is polite; similarly, if Y is a round hat, then it is both a hat and is round.

9

This characterization applies to only some instances of postnominal ADJs in Romance languages, however: as an anonymous reviewer points out, some postnominal ADJs may carry non-intersective meanings and therefore, they cannot be treated as predicates inside clausal modifiers. For example, the French sentence given in (i), which contains the postnominally occurring ADJ bleue, does not mean that ‘Jean has a blue fear’ (i.e., something that is both fear and is blue). Instead, the ADJ here helps engender an idiomatic (i.e., non-intersective) meaning, so one cannot postulate an RRC structure for it.

(i)

Jean a eu AUX have J. ‘Jean has a terrible fear.’

une one

peur fear

bleue. blue

1.4 Some Remarks on the Technical Terms to Be Used

11

As for non-intersective N modifiers, I should note that, in the literature, being a non-intersective ADJ is often equated with carrying what is referred to as ‘adverbial’ semantics (e.g., Alexiadou et al. 2007; Cinque 2010) and there are some good reasons for it. To see this, consider the sentences in (13): these sentences contain a subsective N modifier beautiful and here, the meaning of this ADJ is not ascribed to Olga; rather, it is ascribed to the dancing events involving her where she occurs as the agent. And since ascribing properties to events is what manner adverbs do (see Davidson 1967 and much subsequent work), we can treat such occurrences of beautiful as carrying adverbial semantics. (13) a. Olga is a beautiful dancer although she is not pretty as a woman. b. That girl over there dances really beautifully. And I want to know that beautiful dancer’s name.

Consider now (14) and (15). Here, the (a) examples are semantically equivalent to the (b) examples, and this shows that modal ADJs like alleged and privative ADJs like former can be rendered an adverbial construal as well. (14) a. John is an alleged murderer. b. John is allegedly a murderer. (15) a. John is a former senator. b. John was formerly a senator.

Taken together, then, this set of examples shows that non-intersective ADJs need to be differentiated from intersective ones and the former may well be treated on a par with adverbs. Turning now to the restrictive versus non-restrictive modifier distinction, for our purposes, being a restrictive N modifier will mean that the ADJ at hand reduces the size of the set denoted by the nominal it modifies, and being a non-restrictive N modifier means that it does not. Given this, one may think that the restrictive versus non-restrictive distinction is synonymous with the intersective versus non-intersective distinction because an intersective modifier is bound to reduce the size of the set denoted by the head N. But we will not treat them synonymously here (compare with Cinque 2010). In my analysis, whether an ADJ is intersective or non-intersective will be determined by looking at the semantic relation that holds between the ADJ and its head N and whether an ADJ is restrictive or not will be determined by looking at the discourse context. That is, the intersective versus non-intersective distinction is made within a sentence whereas the restrictive versus non-restrictive distinction is made cross-sententially or at the discourse level.

12

1

Introduction

To illustrate this, consider first (16). In this discourse, the English ADJ red occurs twice and, in both occurrences, it instantiates an intersective ADJ since red hat refers to the set of things that are both red and hats. Yet while the first occurrence of the ADJ is restrictive, the second occurrence of it is not: in the second sentence, the referent of the entire NP is already established, so there is no need for the ADJ to reduce the size of the set denoted by the head N, unlike the case with the first sentence. (16) John got me a really beautiful red hat for my birthday. And I’m going to wear that red hat only for some special occasions.

The independence of the intersective versus non-intersective distinction from the restrictive versus non-restrictive distinction is further exemplified by the data given in (11). According to the way we define the terms here, the ADJs in (11a, b, c) are inherently intersective but in this environment, they are construed as non-restrictive because the DPs containing them refer to discourse-old individuals. The ADJ polite in (11d), on the other hand, instantiates not only an intersective but also a restrictive N modifier because it is construed as reducing the size of the set denoted by student. In a similar vein but somewhat conversely, in (13a), the ADJ beautiful instantiates a non-intersective N modifier but its occurrence actually reduces the size of the set denoted by the head N: due to the presence of the ADJ, the nominal projection containing dancer denotes only a subset of individuals that have the property of being dancers, namely, those that are beautiful as dancers. In the case of (13b), the ADJ is used not only non-intersectively but also non-restrictively because here, the head N’s maximal projection denotes a discourse-old individual and the set reduction is already done by the presence of the manner adverb beautifully in the preceding sentence.

Table 1.2 Terms to be used in describing the syntactic and semantic functions of ADJs Dichotomy I

Predicative Attributive/adnominal

Dichotomy II

Reference-modifying Referent-modifying

Dichotomy III

Intersective Non-intersective

Dichotomy IV

Restrictive Non-restrictive

The ADJ syntactically occurs as part of a sentential predicate. The ADJ syntactically occurs inside a DP, modifying a nominal projection. The ADJ qualifies the property of the head N, i.e., its ‘intension’. The ADJ attributes a property to the individual denoted by the DP that hosts it, i.e., its ‘extension’. [[ADJ N]] = [[ADJ]] \ [[N]] The relation between the ADJ and the head N cannot be captured by the MP for intersective ADJs given in Table 1.1. The ADJ’s meaning reduces the size of the set denoted by the head N. The ADJ’s meaning does not reduce the size of the set denoted by the head N; the nominal’s referent is already established in the discourse.

1.4 Some Remarks on the Technical Terms to Be Used

13

To recapitulate, then, according to the terminological distinctions we make here, no DP-internally occurring ADJs will be analyzed as predicative unless there is good reason to believe that they actually occupy a clausal predicate position. That said, even among attributive/adnominal ADJs, distinctions should be made because some may modify the reference of their head N whereas some may modify the referent thereof, and while some may modify their head N’s meaning intersectively, some may modify it non-intersectively, and finally, while some may perform a restrictive Mod function, some may perform a non-restrictive Mod function. These ideas are summarized in Table 1.2 for ease of future reference.

References Alexiadou, Artemis. 2014. Multiple determiners and the structure of DPs. Amsterdam: John Benjamins. Alexiadou, Artemis, and Chris Wilder. 1998. Adjectival modification and multiple determiners. In Possessors, predicates and movement in the DP, ed. Artemis Alexiadou, and Chris Wilder, 303–332. Amsterdam: John Benjamins. Alexiadou, Artemis, Liliane Haegeman, and Melita Stavrou. 2007. Noun phrase in the generative perspective. Berlin: Walter de Gruyter. Bernstein, Judy B. 1997. Demonstratives and reinforcers in Romance and Germanic languages. Lingua 102: 87–113. Bernstein, Judy B. 2001. Focusing the ‘right’ way in Romance determiner phrases. Probus 13: 1–29. Bolinger, Dwight. 1967. Adjectives in English: Attribution and predication. Lingua 18: 1–34. Bouchard, Denis. 2002. Adjectives, number and interfaces: Why languages vary? Amsterdam: North-Holland. Brugè, Laura. 1996. Demonstrative movement in Spanish: A comparative approach. In University of Venice working papers in linguistics, vol. 6 (1), 1–53. Brugè, Laura. 2002. The positions of demonstratives in the extended nominal projection. In Functional structure in DP and IP: The cartography of syntactic structures, vol. 1, ed. Guglielmo Cinque, 15–53. New York: Oxford University Press. Cabredo Hofherr, Patricia, and Ora Matushansky (eds.). 2010. Adjectives: Formal analyses in syntax and semantics. Amsterdam: John Benjamins. Campos, Héctor, and Melita Stavrou. 2004. Polydefinite constructions in Modern Greek and in Aromanian. In Balkan syntax and semantics, ed. Olga Mišeska Tomić, 137–144. Amsterdam: John Benjamins. Choi, Hye-Won. 1996. Optimizing structure in context: Scrambling and information structure. Stanford, CA: Stanford University dissertation. Chomsky, Noam. 1965. Aspects of the theory of syntax. Cambridge, MA: MIT Press. Cinque, Guglielmo. 1994. On the evidence for partial N movement in the Romance DP. In Paths towards universal grammar, ed. Cinque, Guglielmo, Jan Koster, Jean-Yves Pollock, Luigi Rizzi, and Raffaella Zanuttini, 85–110. Georgetown, DC: Georgetown University Press. Cinque, Guglielmo. 2005. Deriving Greenberg’s Universal 20 and its exceptions. Linguistic Inquiry 36: 315–332. Cinque, Guglielmo. 2010. The syntax of adjectives. A comparative study. Cambridge, MA: MIT Press. Davidson, Donald. 1967. The logical form of action sentences. In The logic of decision and action, ed. Nicholas Rescher, 81–95. Pittsburgh: University of Pittsburgh Press. Dryer, Matthew. 1992. The Greenbergian word order correlations. Language 68: 865–895.

14

1

Introduction

Giusti, Giuliana. 2005. At the left periphery of the Romanian noun phrase. In On space and time in language, ed. Martine Coene, and Liliane Tasmowski, 23–49. Cluj-Napoca: Clusium. Greenberg, Joseph H. 1963. Some universals of grammar with particular reference to the order of meaningful elements. In Universals of language. Report of a conference held at Dobbs Ferry, NY, April 13–15, 1961, ed. Greenberg, Joseph H., 73–113. Cambridge, MA: MIT Press. Hawkins, John A. 1983. Word order universals. New York: Academic Press. Jacobs, Roderick, and Peter Rosenbaum. 1968. English transformational grammar. Waltham, MA: Blaisdell. Kamp, Hans, and Barbara H. Partee. 1995. Prototype theory and compositionality. Cognition 57: 29–191. Kang, Soon Haeng. 2006. The two forms of the adjective in Korean. In University of Venice working papers in linguistics, vol. 16, 137–163. Kayne, Richard S. 1994. The antisymmetry of syntax. Cambridge, MA: MIT Press. Kim, Min-Joo. 2002. Does Korean have adjectives. In MIT working papers in linguistics 43: Proceedings of HUMIT 2001, ed. Ionin, Tonia, Heejeong Ko, and Andrew Nevins, 71–89. Cambridge, MA: MITWPL. Kim, Min-Joo. 2014a. On the position of adnominal adjectival expressions in Korean. In Peaches and plums, ed. Huang, C.-T. James, and Feng-hsi Liu, 75–99. Taipei, Taiwan: Academia Sinica. Kim, Min-Joo. 2014b. Adjective ordering: A view from Korean. In Proceedings of the 31st west coast conference on formal linguistics, ed. Robert Santana-LaBarge, 285–294. Somerville, MA: Cascadilla Proceedings Project. Laenzlinger, Christopher. 2005. French adjective ordering: Perspectives on DP-internal movement types. Lingua 115: 645–689. Lakoff, George. 1971. On generative semantics. In Semantics: An interdisciplinary reader in philosophy, linguistics, and psychology, ed. Danny D. Steinberg, and Leon A. Jacobovits, 232– 296. Cambridge, MA: Cambridge University Press. Larson, Richard K., and Franc Marušič. 2004. Indefinite pronoun structures with APs. Linguistic Inquiry 35 (2): 268–287. Larson, Richard K. 1998. Events and modification in nominals. In Proceedings from semantics and linguistic theory (SALT) VIII, 145–168. Cornell University Press. Larson, Richard K. 2000. Temporal modification in nominals. Paper presented at the International Roundtable on the Syntax of Tense, University of Paris VII, France. Lee, Hanjung. 2006. Parallel optimization in case systems: Evidence from case ellipsis in Korean. Journal of East Asian Linguistics 15: 69–96. Lin, Jo-Wang, and Wei-Tian Dylan Tsai. 2015. Restricting non-restrictive relative clauses in Mandarin Chinese. In Chinese syntax in a cross-linguistic perspective. ed. Li, Yen-Hui Audrey, Andrew Simpson, and Wei-Tien Dylan Tsai, 100–127. Oxford: Oxford University Press. Maling, Joan, and Soowon Kim. 1998. Case assignment in the siphta-construction. In Description and explanation in Korean linguistics, ed. Ross King, 133–168. East Asia Program: Cornell University. Martin, Samuel E. 1992. A reference grammar of Korean. Tokyo, Japan: The Charles E. Tuttle Company. Morzycki, Marcin. 2016. Modification. Cambridge, UK: Cambridge University Press. Partee, Barbara H. 1995. Lexical semantics and compositionality. In An invitation to cognitive science, vol. I: Language, ed. Gleitman, Lila, and Mark Liberman, 311–360. Cambridge, MA: MIT Press. Partee, Barbara H. 2009a. The dynamics of adjective meaning. In Computational linguistics and intellectual technologies: Papers from the annual international conference “Dialogue 2009”, ed. A.E. Kibrik et al., 593–597. Moscow: Russian State Humanities University. Partee, Barbara H. 2009b. Formal semantics, lexical semantics, and compositionality: The puzzle of privative adjectives. Philologia 7: 11–23.

References

15

Plank, Frans. 2003. Double articulation. In Noun phrase structure in the languages of Europe, ed. Frans Plank, 337–395. Berlin: Mouton de Gruyter. Potts, Christopher. 2005. The logic of conventional implicatures. New York: Oxford University Press. (2003 University of California-Santa Cruz dissertation.) Ramaglia, Francesca. 2011. Adjectives at the syntax-semantics interface. Studies in generative grammar. Berlin: Mouton de Gruyter. Scott, Gary-John. 2002. Stacked adjectival modification and the structure of nominal phrases. In Functional structure in DP and IP: The cartography of syntactic structures, vol. 1, ed. Guglielmo Cinque, 91–120. New York: Oxford University Press. Smith, Carlota. 1961. A class of complex modifiers in English. Language 37: 342–365. Smith, Carlota. 1964. Determiners and relative clauses in a generative grammar of English. Language 40: 37–52. Sproat, Richard, and Chinlin Shih. 1988. Prenominal adjectival ordering in English and Mandarin. In Proceedings of NELS, vol. 18, 465–489. Amherst, MA: GLSA. Sproat, Richard, and Chinlin Shih. 1990. The cross-linguistics distribution of adjectival ordering restrictions. In Interdisciplinary approaches to language: Essays in honor of S-Y. Kuroda, ed. Georgopoulos, Carol, and Roberta Ishihara, 565–593. Dordrecht: Kluwer. Suh, Jeng-Soo. 1996. Kwuke mwunpep [Korean grammar]. Seoul: Hanyang University Press. Svenonius, Peter. 2008. The position of adjectives and other phrasal modifiers in the decomposition of DP. In Adjectives and adverbs: Syntax, semantics, and discourse, ed. Chris Kennedy, and Louise McNally, 16–42. New York: Oxford University Press. Vendler, Zeno. 1968. Adjectives and nominalizations. The Hague: Mouton. Whorf, Benjamin Lee. 1945. Grammatical categories. Language 21: 1–11. Yu, Hyen-Kyeng. 1998. Kwuke hyengyongsa yenkwu [A study on Korean adjectives]. Seoul: Hankwukmwunhwasa.

Chapter 2

Adnominal Adjectival Classes in Korean

This chapter introduces major adnominal adjectival classes in Korean, focusing on their basic morpho-syntactic and semantic properties. In so doing, we take a first step towards describing how Korean encodes various types of adnominal meaning in comparison to more well-studied languages like English.

2.1

Some Background on Korean1

Korean is an East Asian language which is mainly spoken in the Korean peninsula and some parts of China and Japan.2 It is considered a language isolate.3 But due to the active and long language contact between Korea and Japan, it displays numerous syntactic similarities to Japanese, which is also considered a language isolate. For historical and geographical reasons, there has been a significant influence from Classical Chinese, in particular in the lexicon, and for over a millennium, Korean was written with adapted Chinese characters called hanja until the phonetically-based writing system hangul was invented in the 15th century, becoming widely used in the 20th century. Morphologically, Korean is an agglutinative language and except for nouns and a few functional categories (e.g., DEMs, NUMs), all verbs, tense/aspect/mood (TAM) markers, complementizers, case particles, and adpositions are bound morphemes. In terms of syntactic properties, Korean is a head-final language, with SOV as the canonical constituent order. Scrambling is common, both DP internally and externally, but sentence structures exhibit rigid head-finality. 1

For a good overview of the Korean language, see, a.o., Chang 1996 and Sohn 1999. Korean is one of the two official languages in China’s Yanbian Korean Autonomous Prefecture, for example. 3 A few linguists claim that Korean belongs to the Altaic language family, which is typically represented by Manchu-Tungusic, Mongolian, and Turkic languages, but such a view is not widely adopted. 2

© Springer Nature Switzerland AG 2019 M.-J. Kim, The Syntax and Semantics of Noun Modifiers and the Theory of Universal Grammar, Studies in Natural Language and Linguistic Theory 96, https://doi.org/10.1007/978-3-030-05886-9_2

17

18

2 Adnominal Adjectival Classes in Korean

Because of the agglutinative nature, grammatical relations in Korean are marked by case particles or postpositions but depending on context, case markers can be dropped, as shown in (1).4 (1) a. b.

Na(-ka) Seyra(-rul) S.(-ACC) I(-NOM) ‘I met Seyra.’ Mina(-nun) hakyo(-ey) M.(-TOP) school(-GOAL) ‘Mina went to school.’

manna-ss-ta. meet-PST-DECL ka-ss-ta. go-PST-DECL

There are some non-trivial semantic consequences for phonologically realizing or unrealizing a case-marker. For example, an overtly-case-marked DP receives (contrastive) focus. And when no overt case particles appear in a sentence, the entire sentence can be interpreted as a non-partitioned informational unit which updates the hearer’s knowledge state. This can be illustrated by looking at (1a): when uttered without the case particles, the content of this sentence can be entirely new information to the hearer, and for this reason, it can be felicitously uttered in an out-of-the-blue context to start a conversation.5 As we will see below and in Chap. 5, case-particle omission possibility has an important consequence for our analysis of various types of DP structure in Korean. Among other syntactic properties of Korean, what will turn out to be the most relevant for our purposes is that it is an article-less and classifier (CL) language. There are determiner-like elements such as DEMs, possessives, and NUMs, but none of them behaves like a prototypical definite article which would correspond to English the. Furthermore, all of them occur prenominally, suggesting that they are phrasal constituents rather than heads, given that Korean is a head-final language. I should note, however, that Korean NUMs always form a cluster with a CL and together, they may occur either prenominally or postnominally (modulo some semantic consequences to be discussed in Chap. 5).6 This is shown in (2). (2) a.

b.

4

sey-kwon-uy three-CL-GEN ‘three books’ chayk book ‘three books’

chayk book

(NUM + CL > N)

sey-kwon three-CL

(N > NUM + CL)

In transcribing the Korean data presented in this book, following standard practice, I adopt Yale Romanization (Martin 1992). In addition, I use a hyphen to indicate a boundary between two lexemes and a dot to indicate a morpheme boundary within a single lexeme. 5 As far as I am aware, this property of Korean has gone unnoted in the literature. For other semantic, pragmatic, and/or syntactic properties of (structural) case omission, see Ahn and Cho 2007, and the references there. 6 When occurring postnominally, a ‘NUM + CL’ string behaves like a floated quantifier and in some cases, such postnominal occurrences of NUMs are strongly preferred over the prenominal occurrences thereof.

2.1 Some Background on Korean

19

It is also worth noting that Korean has a plural marker (PL) -tul and this morpheme is typically realized on a [+animate] nominal, as can be seen by comparing (3) with (4).7 (3) Sentences containing a plural inanimate NP: a. Chayksang-wuy-ey sey-kwon-uy chayk(?-tul)-i three-CL-GEN book(-PL)-NOM Desk-top-LOC ‘There are three books on the desk.’ sey-kwon-i b. Chayksang-wuy-ey chayk(*-tul) book(-PL) three-CL-NOM Desk-top-LOC ‘There are three books on the desk.’ (4) Sentences containing a plural animate NP: a. Pang-an-ey sey-myeng-uy Room-inside-LOC three-CL-GEN ‘There are three students in the room.’ b. Pang-an-ey hak.sayng(-tul) book(-PL) Room-inside-LOC ‘There are three students in the room.’

iss-∅-ta. exist-N.PST-DECL iss-∅-ta. exist-N.PST-DECL

hak.sayng(-tul)-i book(-PL)-NOM

iss-∅-ta. exist-N.PST-DECL

sey-myeng-i three-CL-NOM

iss-∅-ta. exist-N.PST-DECL

Another notable property of Korean is that all lexically contentful adnominal APs occur prenominally (just like the DEMs and possessive N modifiers in the language) but because APs and some RCs have the same ending, namely, -(u)n8 (which I gloss as adnominalizer (ADN) to stay neutral for the time being), it is not always easy to discern them from each other. To see this, compare (5a) and (5b). (Here and throughout, e indicates the gap inside an RC that is co-indexed with the head N of that RC.) (5) a.

b.

7

[ppalka.h-]-n cip house [red-]-ADN ‘a red house’ [ei cip.wung-i ppalka.h-∅-]-n [__ roof-NOM red-N.PST]-ADN ‘a house whose rood is red’

(putative AP)

cipi house

(full-fledged RC)

This is actually a rather unusual property that Korean has since CL languages usually do not have a PL marker, in addition to CL markers (see Svenonius 2008). 8 This suffix may be realized as -un or -n, depending on the ending of the predicative stem it is attached to: if the preceding stem ends in a consonant, it is realized as -un; if not, it is realized as -n.

20

2 Adnominal Adjectival Classes in Korean

Devising fool-proof diagnostics with which to differentiate true APs from RCs in Korean is not an easy task, as the existence of conflicting analyses in the extant literature suggests.9 Therefore, I do not attempt it here. For the sake of concreteness, however, I will assume that an adjectival expression ending in -(u)n can be given the categorial status of an AP if it meets the following criteria: (i) its root is an intransitive stative predicate; (ii) it has no overt TAM marking other than -(u)n10; and (iii) it has no overt case marking. According to these classificatory criteria, the N modifiers in (6) instantiate APs whereas those in (7) instantiate RCs. I assume that nominative case (NOM) is licensed by (finite) tense in Korean, so even though (6a) and (7a) are seemingly identical except for the presence versus absence of NOM inside the N modifier, I do not treat them on a par with each other. A differential treatment of (6a) and (7a) is also motivated by the fact that while (7a) is compatible with an overt TAM marking, as shown in (7b), (6a) is not, as exemplified in (6b). To reiterate, then, I take an overt NOM inside an N modifier in Korean as one of the indicators of a clausal structure and thus treat cases like (7a) as containing an RC even though they lack an overt TAM marking and are derived from an intransitive predicate in a way analogous to (6a). (6) APs ending in -(u)n in Korean: a. [khi-khu-]-n ai [height-big-]-ADN child ‘a/the tall child’ (Lit.: ‘a/the height-big child’) b. *[khi-khu-ess-te-]-n ai [height-big-PST-RTRO-]-ADN child Intended: ‘a/the formerly tall child’ (7) RCs ending in -(u)n in Korean containing an intransitive predicate: a. [ei khi-ka khu-]-n aii [__ height-NOM big-]-ADN kid ‘a/the child who is tall’ (Lit.: ‘a/the child whose height is big.’) b. [ei khi-ka khu-ess-te-]-n aii [__ height-NOM big-PST-RTRO-]-ADN kid ‘a/the kid who used to be tall’

Yet another characteristic of Korean that will prove to be relevant for our purposes is that, unlike languages like English, it has two copular predicates, namely, i- and ha-

9

For example, in the extant literature, authors like Choy (1971), Yu (1998), Sohn (1999), Mok (2002), and Kang (2005, 2006) maintain that Korean has an open class of ADJ and a proper subset of adjectival expressions ending in -(u)n are true APs although they never explicitly spell out how to differentiate them from RRCs. On the other hand, authors like Hwang (1990), Martin (1992), Suh (1996), Maling and Kim (1998), and Kim (2002) claim that Korean lacks an open class of ADJ and all adjectival expressions ending in -(u)n are better treated as RRCs. 10 As will be shown in Sect. 2.2.6, in Korean, the TAM marking inside an RC is sometimes realized by changing the form of the morpheme -(u)n and therefore not all RCs have the same ending.

2.1 Some Background on Korean

21

(Sohn 1999: 281–283), and while nominals may only occur with i-, as shown in (8), ADJs do not exhibit a uniform behavior: some predicative ADJ-like expressions occur only with i- and some may occur only with ha-, as shown in (9) and (10), respectively. Given this, a predicative complement’s compatibility with i- can help decide whether it belongs to a nominal category or not, and this will in turn help decide on the categorial status of some adjectival expressions, as we will see in Sect. 2.2. (8) Nominal predicative complements co-occurring with a copular verb: a. √Minho-nun ku tangsi haksayng*(-i)-ess-ta. (N + i-copula) that time student-COP-PST-DECL M.-TOP ‘Minho was a student at that time.’ a’. *Minho-nun ku tangsi haksayng-ha-yess-ta. (N + ha-copula) Intended: ‘Minho was a student at that time.’ b. √Minho-nun ney-ka kekceng*(-i)-∅-ta. (N + i-copula) you-Nom worry-COP-N.PST-DECL M.-TOP ‘As for Minho, you are a/the worry.’ b’. *Minho-nun ney-ka kekceng-ha-∅-ta. (N + ha-copula) Intended: ‘As for Minho, you are a/the worry.’ (9) Adjectival predicative complements co-occurring with a copular verb: (ADJ + i-copula) a. √Minho-nun kongkyekcek-i-∅-ta. M.-TOP aggressive-COP-N.PST-DECL ‘Minho is aggressive.’ (ADJ + ha-copula) b. *Minho-nun kongkyekcek-ha-∅-ta. Intended: ‘Minho is aggressive.’ (10) a. *Minho-nun ku tangsi yumyeng-i-ess-ta. M.-TOP that time famous-COP-PST-DECL Intended: ‘Minho was famous at that time.’ b. √Minho-nun ku tangsi yumyeng-ha-yess-ta. ‘Minho was famous at that time.’

(ADJ + i-copula)

(ADJ + ha-copula)

Finally, I should note that a large proportion of Korean ADJs are of Sino-Korean (SK) origin, i.e., borrowed from Classical Chinese (Sohn 1999),11 and although in many cases, an SK expression and its corresponding native Korean (NK) expression co-exist as illustrated in (11), native or nativized lexemes and their Sino-Korean counterparts exhibit different morpho-syntactic properties. For example, all NK or nativized verbal categories are bound morphemes but their SK counterparts are mostly free morphemes, so even if they are semantically verbal, SK lexemes behave more like nouns in terms of their morphology (recall that between nouns and verbs,

11

According to Sohn (1999: 86), contemporary Korean lexicon consists of roughly 60% of Sino-Korean, 35% of native Korean, and 5% of loan lexemes or borrowed elements.

22

2 Adnominal Adjectival Classes in Korean

only nouns can be free morphemes in Korean, as stated on p. 17). Below, we will see that this property of Korean lexemes partly explains the existence of two distinct adjectival classes in the language. (11) a.

b.

2.2

cang-sin (SK) long-body ‘a tall person’ khu-∅-]-n sarami (NK) [ei khi-ka [__ height-NOM big-N.PST-]-ADN person ‘a person who is tall’ (Lit.: ‘a person whose height is big’)

Major Korean Adjectival Classes and Their Characteristic Properties

The traditional linguistic literature on Korean suggests that there cannot be just one absolutely right way to identify its adjectival classes (e.g., Yu 1998 and the references there). Hence, one has to make some arbitrary choices in deciding which type of lexemes to include in the discussion of N modifiers in Korean and which not to. In this book, I adopt a form-based criterion over a more meaning or function-based one chiefly because it is more reliable. But another important reason is that doing so will make it easier for us to talk about the correspondence between the morpho-syntax and the semantics of adjectival expressions in subsequent pages. Applying a form-based criterion to Korean yields the following six adnominal adjectival classes:12

12

One may wish to add possessive or genitive-case (GEN) marked expressions to this list because some of them seem to carry adjectival semantics, as shown in (i).

(i) a.

b.

[mi-ci]-uy [non-knowledge]-GEN ‘unknown world’ [cang-sin]-uy [long-body]-GEN ‘tall man’

sey.key world namca man

I have not included GEN-marked adjectival expressions here because they are relatively small in number. Moreover, unlike other morpho-syntactic operations listed in (12) such as cek- or unsuffixation, GEN-marking is not exclusively used to create an adjectival category; for example, it is frequently used to create a quantificational category, as shown in (ii). (ii) tay-pwupwun-uy large-portion-GEN ‘most people’

saram-tul person-PL

2.2 Major Korean Adjectival Classes and Their Characteristic Properties

23

(12) Major adnominal adjectival classes in Korean: (i) adjectival prefixes: e.g., tay-cethayk ‘large house’ (ii) adjectival Ns: e.g., hankwuk kwukmin ‘Korean citizen’ (iii) so-called attributive determinatives (ATT-DETs): e.g., say cip ‘new house’ (iv) adjectival expressions ending in -cek (CEK-XPs): e.g., kasi-cek pyel ‘visible star’ (v) adjectival expressions ending in -un (UN-XPs): e.g., yumyengha-n saram ‘famous person’ (vi) RCs: e.g., yumyengha-ess-te-un saram ‘person who used to be famous’

To my knowledge, listing all these six classes under the rubric of major adnominal expressions in Korean is unprecedented; most reference grammars only consider what I call UN-XPs as true ADJs in Korean (see, e.g., Yu 1998 and the references there). But I have included adjectival prefixes and adjectival Ns here because much of N modification in Korean is actually done by them, as we will see shortly. Moreover, in numerous cases, their presence in the language obviates the need for using putative APs, and this sometimes makes it difficult to apply some of the well-established typological generalizations on AOR to Korean.13 As for what I have labeled as attributive determinatives (ATT-DETs) and expressions ending in -cek (CEK-XPs), I have included them in our discussion because Kang (2005, 2006) treats them as forming ADJ classes in Korean along with UN-XPs, and I agree with his treatment for the most part. That said, they are included here for two additional reasons: one is that they hold unique statuses in the Korean lexicon, and the other is that they make interesting comparison cases for UN-XPs, both of which we will see below.

13

To illustrate this a bit, anticipating our discussion in Chap. 3, in Korean, multiple putative adjectival phrases (i.e., UN-XPs) rarely occur modifying the same N, and this makes it difficult to test whether Korean ADJs occur in the order of ‘quality > quantity/size > color > nationality’, a cross-linguistic tendency observed by Sproat and Shih (1988, 1990). This can be seen by the fact that in Korean, one of those adjectival meanings is most likely expressed by a prefix, and in such cases, the resulting linear order can be ‘color > size’, rather than ‘size > color’, as shown in (i). We will revisit this issue in Chap. 3 and identify the “forces” behind such typologically unexpected AOR phenomena. (i) haya.h-n tay-ce.thayk large-mansion white-ADN ‘white, large mansion’

(color > size)

24

2 Adnominal Adjectival Classes in Korean

Finally, RCs are included here because as mentioned above, putative adnominal ADJs in Korean have been claimed to be verbal predicates inside RCs (e.g., Hwang 1990; Martin 1992; Suh 1996; Maling and Kim 1998; Kim 2002), so there is the need to examine whether it is possible to tease apart APs which look like RCs from true RCs or vice versa beyond the descriptive level of analysis I have offered above. In addition, RCs need to be looked at in the context of N modification in Korean in general because they play a major role in “compensating for” the absence of certain types of ADJs in this language. In what follows, I offer an overview of each of the six adjectival classes listed in (12), beginning with adjectival prefixes. At the outset of each subsection, I introduce the morpho-syntactic and semantic characteristics of the adjectival class at hand, thereby establishing their distinctiveness. After this, I examine their semantics from the viewpoint of the four-way formal semantic classification of ADJ meanings introduced in Chap. 1. I then turn to discussing the typological implications of the Korean facts carved out here in the larger context of theorizing the syntax and semantics of N modifiers in human language. What is undertaken in this chapter will show how each adjectival class in Korean behaves in isolation but it will also pave the road for investigating their behavior in more complex syntactic environments in Chaps. 3 through 5.

2.2.1

Adjectival Prefixes

Korean has a plethora of adjectival prefixes. They are typically mono-syllabic and mono-morphemic though some can be free morphemes. Related to this is the fact that most of them are of SK origin though some can be of NK (e.g., ches- ‘first’, hol- ‘alone’) or non-SK foreign origin (e.g., anthi- borrowed from anti- in English). Free-standing adjectival prefixes behave like nouns in that when occurring predicatively, they co-occur with the i-copula, not the ha-copula, as shown in (13). (13) a.

b.

cwung-hyeng (adjectival prefix + N) middle-shape ‘middle size/scale’ Ku os-uy saycu-nun cwung-i/*ha-∅-ta. (predicative use) That clothes-GEN size-TOP middle-COP-N.PST-DECL ‘Those clothes are of middle size.’

I assume that adding an adjectival prefix to an N occurs in the realm of morphology, rather than in syntax, since the resulting nominal expression exists in the lexicon. Therefore, for ease of reference, I will henceforth refer to the process of adding an adjectival prefix to a nominal root ‘a morphological N compounding’.

2.2 Major Korean Adjectival Classes and Their Characteristic Properties

25

Morphological N compounding can target any stratum in the lexicon and can cover a wide range of adnominal meaning including color, size, quantity, quality/ rank, temperature, temporal order, value, and gender, as shown in (14)–(20). It can also derive an idiomatic meaning, as shown in (21), or add a more grammatical meaning such as degree and proportion/quantification, as shown in (22). (14) Color-denoting prefixes: a. cheng/hong-kochwu blue/red-pepper ‘green/red pepper’ b. cheng-paci blue-pants ‘blue jeans’ (15) Size-denoting prefixes: a. tay/so-hyeng large/small-shape ‘large/small size/scale’ b. so-nye small-woman ‘girl’ (Lit.: ‘small woman’)

(SK)

(SK)

(SK)

(SK)

(16) Quality/rank-denoting prefixes: a. ko/ce/cwung-kup (SK) high/low/middle-class/level ‘high/low/middle class/level’ b. cwung-taycang (SK) middle-leader/general ‘captain’ (Lit.: ‘middle leader/general’) (17) Temperature-denoting prefix: a. nayng-khephi (SK) cold-coffee ‘cold coffee’ b. on-thang (SK) warm-tub ‘bath tub containing warm water found in a public bathing house’

26

2 Adnominal Adjectival Classes in Korean

(18) Temporal-order-denoting prefixes: a. cho-chang.ki (SK) early-stage ‘the early stage (of someone’s life or business)’ b. hwu-pan.pwu (SK) later-half.part ‘the latter part (of a show, game, concert, or novel)’ c. ches-ai (NK) initial-child ‘someone’s first-born’ d. mayn-twi (NK) very-back ‘the very last (minute/seat)’ (19) Value-denoting prefix: a. ho-sa good-event ‘good event’ b. ak-cay evil-occurrence ‘tragic incident’ (20) Gender-denoting prefix: a. am-thalk female-chicken ‘hen’ b. swus-thalk male-chicken ‘rooster’

(SK)

(SK)

(NK)

(NK)

(21) Prefixes engendering an idiomatic meaning: (NK) a. hol-api alone-father ‘widower’ (NK) b. hol-mom alone-body ‘(married) woman who is not pregnant’ c. wus-ton (NK) above-money ‘additional money offered in a business transaction (usually for a special treatment)’ (22) Prefixes with a grammatical meaning: a. choy-wusen (degree: SK) utmost-priority ‘the highest priority/the most pressing’ (proportion: SK) b. cen-kwuk.min entire-nation.citizen ‘the entire nation/people of some country’ (proportion: SK) c. chong-ciphap general-gathering ‘the gathering of the entire individuals relevant for the discourse at hand’ (proportion: NK) d. on-tongney entire-town ‘the entire town’

2.2 Major Korean Adjectival Classes and Their Characteristic Properties

27

Before proceeding further, I should note that some adjectival prefixes obligatorily engender definite and/or unique interpretations for the root Ns they combine with; for example, all the morphological N compounds in (18) and (22) receive definite interpretations regardless of the discourse context in which they occur, as indicated by their English translations. And I take this to suggest that adjectival prefixes in Korean are of two kinds, namely, lexical and grammatical, and the grammatical ones arguably intensify determiner meaning (a term borrowed from Bolinger 1967) in a way comparable to what English functional adjectives like very do in contexts like (23). (23) a. b.

in this very moment That is the very reason why we didn’t choose John for this job.

Turning now to re-examining Korean adjectival prefixes in light of the four-way classification of ADJ meanings introduced in Chap. 1: first, some of them may have intersective meanings and this is illustrated in (24). The prefixes cheng and nayng here can be analyzed as intersective modifiers because the MP that holds for the prototypical intersective ADJ carnivorous in English also holds for them, as one can see from comparing (25) and (26). (24) a.

cheng-kochwu blue-pepper ‘green pepper’ b. nayng-khephi cold-coffee ‘cold coffee’

(25) The MP for an intersective ADJ in English: a. [[ADJ N]] = [[ADJ]] ∩ [[N]] b. [[carnivorous animal]] = [[carnivorous]] ∩ [[animal]] (26) The MPs for Korean CNs modified by cheng- or nayng-: a. [[cheng N]] = [[cheng]] ∩ [[N]] b. [[nayng N]] = [[cheng]] ∩ [[N]]

Secondly, some adjectival prefixes can express subsective meaning, which is spelled out in (27). (27) The MP for a subsective ADJ in English: a. [[ADJ N]] ⊆ [[N]] b. [[skillful doctor]] ⊆ [[doctor]]

28

2 Adnominal Adjectival Classes in Korean

By way of illustration, the prefix tay- in (28a) instantiates a subsective N modifier because the denotation of tay-pepkwan constitutes a subset of the denotation of the root N, as shown in (29). That is, the morphological N compound refers to a set of individuals that are powerful as judges, rather than those that are powerful and are judges. Similarly, cham- in (28b) is best analyzed as a subsective modifier because the entire N compound refers to a set of individuals that are of a certain quality as combs, namely, very fine-toothed combs.14 (28) a.

b.

(29) a. b.

tay-pep.kwan large-judge ‘supreme court justice’ (Lit.: ‘large judge’) cham-pis true-comb ‘fine-toothed comb’ (Lit.: ‘true/genuine comb’) [[tay-pepkwan]] ⊆ [[pepkwan]] [[tay-pepkwan]] ≠ [[tay]] ∩ [[pepkwan]]

Finally, some adjectival prefixes can carry privative ADJ semantics, for which the MP is given in (30), along with an English example. (30) MP for privative ADJ meaning: a. [[ADJ N]] ∩ [[N]] = ∅ b. [[fake gun]] ∩ [[gun]] = ∅

To illustrate privative adjectival prefixes in Korean, consider the nominal data in (31). These examples show that adding pwul- ‘not’ or wi- ‘false’ to a noun amounts to negating the meaning of that noun because the derived compound N describes a property that is described by the complement of the set denoted by the root N. Consequently, the MPs in (32) hold in a manner identical to those given in (30). (31) a.

b.

(32) a. b.

pwul-kyuchik not-pattern/rule ‘irregularity’ wi-sen.ca false-good.person ‘hypocrite’ (Lit.: ‘falsely good person’) [[pwul-kyuchik]] ∩ [[kyuchik]] = ∅ [[wi-senca] ∩ [[senca]] = ∅

In Korean culture, these combs are made of bamboo and are much finer than what are called nit combs in English.

14

2.2 Major Korean Adjectival Classes and Their Characteristic Properties

29

Adjectival prefixes are highly robust and semantically versatile but I have not found cases where they carry modal adnominal semantics. This fact creates a curious gap in their distribution. We take up the question of why this may be the case in Sect. 2.3.

2.2.2

Adjectival Ns

What I call adjectival Ns are similar to adjectival prefixes in that they can be of SK, NK, or non-SK foreign origin, and they also form a semantic unit with the N they modify, some of which instantiate N compounds, as is the case with prefix plus N combinations. Yet there is reason to treat them separately from adjectival prefixes: while (almost) all adjectival prefixes are mono-syllabic and mono-morphemic, most adjectival Ns are bi-syllabic and/or bi-morphemic, and while an adjectival prefix plus head N combination always forms a single word which cannot be further divided at the word level, an adjectival N plus head N combination creates a word that is modified by another word; that is, they have a structure of a word embedded inside another word. To see this, compare (33a) and (33b), where [ ] indicates a word boundary. (33) a.

b.

[N cheng-kochwu] [ blue-pepper] ‘green pepper’ [N [N yuri] [N kurus]] [ [ glass] [ bowl]] ‘glass bowl’

(adjectival prefix + N)

(adjectival ADJ + N)

Given this, here and below, I will refer to cases involving adjectival Ns as ‘analytic N compounds’. Analytic N compounds are abundant in Korean and the compounding process involved may engender various types of N modificational meaning: as shown in (34a–d), adjectival Ns can describe color, quantity, quality, and gender, just like adjectival prefixes. They can also encode other attributive meanings or idiomatic meanings, as shown in (34e–f). In addition, they may describe temporal or spatial relations, as shown in (34g–h) although they do not express other grammatical relations as freely as adjectival prefixes do. Finally, unlike adjectival prefixes, they may even describe shape, nationality, and several thematic relations, as shown in (34i–n).

30

2 Adnominal Adjectival Classes in Korean

(34) a.

norang napi yellow butterfly ‘yellow butterfly’ b. tay.ryang sayngsan large.quantity production ‘mass production’ c. ko.kup yori high.class food ‘high-class food’ d. namca hwacangsil man bathroom ‘men’s room’ e. yerum kkoch summer flower ‘summer flower’ f. cirum kil diameter road ‘short cut’ g. cheum mannam first meeting ‘first encounter’ h. cwungkan cwasek middle seat ‘middle seat/row’ i. neymo sangca square box ‘square-shaped box’ j. hankwuk kwukmin South.Korea citizen ‘Korean citizen’ k. pwullanse-cey15 catongcha France-made car ‘car made in France’ phari chong-yengsa-kwan 15l. Paris general-consul-building ‘consulate general located in Paris’

(NK: color)

(SK: quantity)

(SK: quality)

(SK: gender)

(NK: attributive meaning)

(NK: idiomatic meaning)

(NK: temporal relation)

(SK: spatial relation)

(NK: shape)

(SK: nationality)

(SK: thematic, Source)

(Foreign: thematic, Location)

15

Categorially, the word pwullanse-cey is a nominal even though its English translation is not. This is evidenced by the fact that it can occur in argument positions, bearing a nominative or an accusative case marker, as shown in (ia). Furthermore, when occurring in predicate position, it combines with the i-copula, not the ha-copula, as shown in (ib).

(i) a.

b.

pwullanse-cey-ka coh- -ta. Catongcha-nun France-made-NOM be.good-N.PST-DECL Automobile-TOP ‘As for cars, France-made ones are good.’ I catongca-nun pwullanse-cey-i/*-ha- -ta. France-made-COP-N.PST-DECL This automobile-TOP ‘This car is France-made.’

2.2 Major Korean Adjectival Classes and Their Characteristic Properties

m. pam-namwu ipsakwi chestnut-tree leaf ‘leaves of chestnut trees’ n. kyengcey kayhyek economy renovation ‘renovation of the economy’

31

(NK: thematic, Possession/Part-whole)

(SK: thematic, Theme)

The fact that some adjectival Ns can encode nationality is worth noting here because this property is not shared by other adjectival classes in Korean. Also, the ability to encode various kinds of thematic relation makes adjectival Ns stand out among non-clausal adjectival expressions16 because even though some adjectival expressions ending in -cek (i.e., CEK-XPs) can bear a thematic relation to the head N they modify, Theme is the only relation they can express. Besides, the same CEK-XP may be construed as encoding a non-thematic meaning as well as a thematic one, as illustrated in (35) (more on this in Sect. 2.2.4). (35) kyengcey-cek kayhyek economy renovation Reading 1: ‘renovation of the economy’ Reading 2: ‘economical renovation’

(CEK-XP) (thematic relation: Theme) (non-thematic attributive meaning)

When viewed through the lens of the four-way formal semantic taxonomy of adjectival meaning given in Chap. 1, adjectival Ns turn out to encode three of the four types of ADJ meanings in a way similar to adjectival prefixes. First, they can express intersective meanings. This is illustrated in (36a, b). In these examples, the meaning of the entire N compound is computed by intersecting the sets that are denoted by the adjectival N and the head N, as given in (37). (36) a.

b.

(37) a. b.

16

cwungkwuk-san kochwu China-from hot.pepper ‘hot pepper from China’ cwungko catongcha used automobile ‘used car’ [[cwungkwuksan N]] = [[cwungkwuksan]] ∩ [[N]] [[cwungko N]] = [[cwungko]] ∩ [[N]]

RCs should be excluded from its comparison set because, given their clausal nature, they can express various thematic relations holding between their head Ns and the sentential predicates inside them. In fact, in Korean, the gap inside an RC which bears the same index as the head N can bear any grammatical relation that is listed on Keenan and Comrie’s (1977) NP Accessibility Hierarchy except for the object of comparison relation, and this provides further evidence for their syntactic and semantic versatility. For relevant discussion and data, see Kwon 2008 (Ch. 2).

32

2 Adnominal Adjectival Classes in Korean

Secondly, some adjectival Ns exhibit subsectivity and this is exemplified in (38): the analytic N compound tayri kyosa ‘substitute teacher’ refers to a subset of teachers, namely, individuals who are substitutes as teachers but not necessarily as something else, as stated in (39). And this shows that the adjectival N tayri is a non-intersective, N-dependent modifier, unlike the adjectival Ns like cwungkwuksan and cwungko. (38) tayri kyosa substitute teacher ‘substitute teacher’ (39) [[tayri kyosa]] ⊆ [[kyosa]]

The non-intersectivity of tayri ‘substitute’ is further evidenced by the fact that what is given in (40) is invalid in a manner analogous to inference patterns involving prototypical subsective ADJs like skillful in English.17 (For simplicity, I present the inference patterns by using English as the meta-language since what matters for our purposes is the meaning, not the linguistic expressions themselves.) (40) Premise: John is a substitute teacher. Premise: John is a pianist. _________________________________________ Conclusion: John is a substitute pianist. (INVALID)

Thirdly, adjectival Ns can be used to express privative modificational meanings, as exemplified by (41). The privative property of the adjectival Ns given here is evidenced by the coherence of the sentences in (42) and the MPs given in (43). (41) a.

b.

(42) a.

b.

17

wico ci.phye counterfeit paper.money ‘counterfeit paper money’ kacca chong fake gun ‘fake gun’ Wico ci.phye-nun ci.phye-ka an-i-∅-ta. paper.money-NOM not-COP-N.PST-DECL Counterfeit paper.money-TOP ‘Counterfeit paper money is not paper money.’ Kacca chong-un chong-i an-i-∅-ta. gun-NOM not-COP-N.PST-DECL Fake gun-TOP ‘A fake gun is not a gun.’

Using this inference pattern as a way to test the subsectivity of an ADJ is due largely to works by Clark (1970), Montague (1970), Parsons (1970), and Kamp (1975).

2.2 Major Korean Adjectival Classes and Their Characteristic Properties

(43) a. b.

33

[[wico ciphye]] ∩ [[ciphye]] = ∅ [[kacca chong]] ∩ [[chong]] = ∅

As is the case with adjectival prefixes, adjectival Ns are not seen to carry modal ADJ semantics, despite their robustness and semantic versatility, and again, we will address why this might be the case in Sect. 2.3.

2.2.3

Attributive Determinatives

What are referred to as attributive determinatives (ATT-DETs) in traditional grammars18 are mono-morphemic lexemes of either NK or SK origin. What is most noteworthy about ATT-DETs is that unlike the other adjectival classes, they form a closed class (see, e.g., Suh 1996; Sohn 1999; Mok 2002). The extant literature lists the following lexemes under this category: say ‘new’ (NK), hyen ‘present/incumbent’ (SK), cen ‘former/previous time’ (SK), ye/yec ‘old/ former/long ago’ (NK), and swun ‘pure/complete’ (SK) (see, a.o., Suh 1996; Sohn 1999; Mok 2002),19 and some illustrative cases are given in (44).

Some authors call ATT-DETs ‘adnominal modifiers’ (kwan-hyeng-e in Korean) (e.g., Kang 2005). But I do not adopt that nomenclature here because it will not let us differentiate ATT-DETs from other types of N modifiers, as any lexeme that modifies an N or a nominal may be considered an adnominal modifier. 19 In addition, many reference grammars treat hen ‘worn out/used’ as an ATT-DET. But I analyze it differently and therefore do not list it here. I take hen to be resulting from combining the verbal stem hel- ‘to wear out’ with the adnominalizing suffix -un. That is, it originates from a morphologically complex expression that is best analyzed as instantiating what we call an UN-XP here. Evidently, this erstwhile UN-XP has developed into two distinct lexemes: (i) the contracted form hen, which means ‘being used (but not necessarily being worn out)’, and (ii) the non-contracted form hel-un, which means ‘worn out’ and which retains its original meaning. Support for this complex origin analysis of hen comes from the fact that it can co-occur with ATT-DETs like say, as shown in (i), whereas lexemes that are unquestionably ATT-DETs such as hyen ‘present/incumbent’ cannot, as shown in (ii). Note that the grammatical data in (i)–(ii) are obtained via Google searches conducted in September, 2013. Therefore, they reflect actual usage although one of the reviewers disagrees with the judgments on (ib) and (ic). 18

(i) a. say/hen os new/used clothes ‘new/used clothes’ b. say hen os ‘used clothes that are newly purchased/obtained’ (Lit.: ‘new used clothes’) c.?hen say os ‘pre-owned but barely worn, so essentially new clothes’ (Lit.: ‘used new clothes) or ‘used clothes that are newly purchased/obtained’

34

2 Adnominal Adjectival Classes in Korean

(44) a.

b.

c.

d.

say moca new hat ‘new hat’ yec chinkwu old/former/long.ago friend ‘old friend or a friend one has known for a long time/known in the past’ swun papo complete idiot ‘complete idiot’ hyen cengpwu present government ‘the present government’

Given their mono-morphemic nature, ATT-DETs have occasionally been treated as a sub-type of adjectival prefixes (e.g., Yu 1998; Lee 2004). But there are reasons to treat them differently. One reason is that while adjectival prefixes do not form separate words from their head N, as shown in (33a), ATT-DETs do, as shown in (45). (45) a.

b.

c.

d.

[say] [moca] [new] [hat] ‘a new hat’ [yec] [chinkwu] [old/former/long.ago] [friend] ‘an old friend/a friend known for a long time/known in the past’ [swun] [papo] [complete] [idiot] ‘a complete idiot’ [hyen] [cengpwu] [present] [government] ‘the present government’

Another more important reason is that while adjectival prefixes cannot occur preceding an analytic N compound, ATT-DETs can. To see this, consider first (46). These data instantiate cases where an adjectival N occurs modifying a morphological N compound which is comprised of an adjectival prefix and an N, and both examples are judged fine.

(ii) a. say/hyen taythongryeng new/incumbent president ‘the new/incumbent president’ b.*say hyen taythongryeng Intended: ‘the new incumbent president’ c.*hyen say taythongryeng Intended: ‘the new incumbent president’ or ‘the incumbent new president’

2.2 Major Korean Adjectival Classes and Their Characteristic Properties

35

(46) Analytic N compounds embedding morphological N compounds: a. kyengki hwu-panpwu (Adjectival N > Adjectival prefix > Root N) game later-part ‘the latter part of a/the game’ b. taynamwu cham-pis (Adjectival N > Adjectival prefix > Root N) bamboo true-comb ‘a fine-toothed comb made with bamboo’

Consider now (47a, b). The ungrammaticality of these examples shows that an adjectival prefix cannot occur structurally higher than an adjectival N. That is, ‘Adjectival prefix > Adjectival N > Root N’ order does not obtain. (47) Adjectival prefixes modifying analytic N compounds: a. *hwu-[kyengki panpwu] (Adjectival prefix > Adjectival N > Root N) later-[game part] Intended: ‘the latter part of a/the game’ b. *cham-[taynamwu pis] (Adjectival prefix > Adjectival N > Root N) true-[bamboo comb] Intended: ‘a fine-toothed comb made with bamboo’

Finally, compare (46)–(47) with (48). This leads us to see that unlike adjectival prefixes, ATT-DETs can occur structurally higher than analytic N compounds, instantiating ‘ATT-DET > Adjectival N > Root N’ order, but ‘Adjectival N > ATT-DET > Root N’ does not obtain. (48) ATT-DETs modifying analytic N compounds: a. hyen [kyengki hwu-panpwu] (ATT-DET > Adjectival N > Root N) present [game later-part] ‘the present latter part of the game’ b. say [taynamwu pis] (ATT-DET > Adjectival N > Root N) new [bamboo comb] ‘a new bamboo comb’ a’. *kyengki hyen hwu-panpwu (Adjectival N > ATT-DET > Root N) Intended: ‘the present latter part of the game’ b’. *taynamwu say pis (Adjectival N > ATT-DET > Root N) Intended: ‘a new bamboo comb’

In the literature, attempts have also been made to subsume ATT-DETs under the rubric of adjectival Ns. Some authors (e.g., Sohn 1999; Mok 2002) analyze say ‘new’ and cwu ‘primary/chief’ as adjectival Ns, for example, based on the fact that they can co-occur with -(u)ro, a morpheme20 that converts a nominal category into an adverbial one, as shown in (49)–(50).

20

Depending on how one looks at it, this morpheme can be analyzed as a postposition or a derivational suffix. More on this in Sect. 2.2.4.

36

2 Adnominal Adjectival Classes in Korean

(49) a. b.

(50) a. b.

mal vs. ‘word’ kwi vs. ‘ear’

mal-ro ‘verbally’ kwi-ro ‘through the ear’

say-ro ‘newly’ cwu-ro ‘chiefly/mainly/primarily’

However, treating ATT-DETs as a subclass of adjectival Ns is not tenable for at least two reasons. Firstly, unlike adjectival Ns, ATT-DETs can be modified by the degree adverb (ADV) acwu ‘very’,21 which typically occurs modifying lexical ADVs (e.g., acwu cal ‘very well’) or psych-predicates (e.g., acwu choh-ta ‘be very happy/good’) (see Kang 2005 for a similar observation). To see this, compare (51) and (52). (51) Degree ADV acwu + Adjectival N: a. [(*acwu) wico] ci.phye [( very) counterfeit] paper.money b. [(*acwu) kacca] chong [( very) fake] gun c. [(*acwu) ceng.ki] moim [( very) regularly.held] meeting (52) Degree ADV acwu + ATT-DET: a. [(√acwu) say] moca [( very) new] hat ‘a ( very) new hat’ b. [(√acwu) yec] kiek [( very) long.ago] memory ‘a ( very) old memory (Lit.: ‘a very long-ago memory’) c. [(√acwu) cen] iyaki [( very) previous] story ‘a ( very) old story’ (‘a story from a very long time ago’)

Secondly, while adjectival Ns can occur predicatively when co-occurring with the copular verb i-, ATT-DETs cannot, as shown in (53) and (54), respectively.

21

Some ATT-DETs are not compatible with acwu but this is for semantic reasons. For example, hyen ‘present/incumbent’ cannot be graded, so it resists degree modification, as exemplified in (i).

(i) [(*acwu) hyen] cengpwu [( very) present] government Intended: ‘the (highly) present government’

2.2 Major Korean Adjectival Classes and Their Characteristic Properties

(53) Adjectival Ns: ci.phye a. wico counterfeit paper.money ‘counterfeit paper money’ b. ku ci.phye-nun wi.co-i/*ha-∅-ta. counterfeit-COP-N.PST-DECL That paper.money-TOP ‘That paper money is counterfeit.’ (54) ATT-DETs: a. say ci.phye new paper.money ‘new paper money’ ci.phye-nun b. *Ku say-i/ha-∅-ta. new-COP-N.PST-DECL That paper.money-TOP Intended: ‘That paper money is new.’

37

(attributive use)

(predicative use)

(attributive use)

(predicative use)

Taken together, this set of facts shows that ATT-DETs form a distinctive class from both adjectival prefixes and adjectival Ns. What is particularly noteworthy is that they can be modified by the degree ADV acwu,22 but they can occur only attributively. A closer look at their semantic properties further shows that they indeed form a unique class of lexemes: they cover a semantic space that no other Korean adjectival classes can. Moreover, they behave more like functional categories than lexical ones. To be more concrete, some ATT-DETs have the semantics of temporal or degree modifiers; e.g., yec ‘old/former/a long time ago’, cen ‘former/previous’, and hyen ‘present/incumbent’, swun ‘pure/complete’. Correlatively, they have what is referred to as ‘adverbial’ semantics of ADJs in the generative linguistics literature (e.g., Larson 1998; Alexiadou et al. 2007, Part III; Cinque 2010) in the sense that they are construed as modifying the eventuality associated with the head N’s meaning. To illustrate this, an ATT-DET plus N string say moca ‘a new hat’ can be construed as meaning a hat that is newly purchased. Similarly, cen taythongryeng ‘a former president’ is synonymous with ‘a person who was formerly a president’, and yec chinkwu ‘an old friend’ is synonymous with ‘a person who was one’s friend a long time ago’. Finally, swun papo ‘a complete idiot’ can be paraphrased as ‘someone whose characteristic properties completely match those of an idiot’. Possibly for these reasons, ATT-DETs instantiate only two of the four types of adjectival meaning introduced in Chap. 1, namely, subsective and privative, and

22

On the basis of the fact that ATT-DETs can be modified by a degree ADV, Kang (2005) claims that they are “true syntactic ADJs in Korean”. But such a treatment cannot be readily justified, as we will see below.

38

2 Adnominal Adjectival Classes in Korean

some of them even resist any of the classifications that would be available under the four-way taxonomy. For example, say ‘new’ can be analyzed as instantiating a subsective modifier when modifying an N like kyosa ‘teacher’ because syllogism (55) fails.23 (55) Premise: John is a new teacher. Premise: John is a Ph.D. ________________________________________ Conclusion: John is a new Ph.D. (INVALID)

The ATT-DET cen ‘former’ in (56) can, on the other hand, be classified as a privative modifier because the MP in (57) holds. (56) cen sangwonuywon former senator ‘a former senator’ (57) [[cen sangwonuywon]] ∩ [[sangwonuywon]] = ∅

Notice now that in contrast to both say and cen, yec partly defies the four-way classification because depending on context, it can mean either ‘former’ or ‘a long time ago’: when construed as meaning ‘former’ as in yec ayin ‘an ex-boy/ girlfriend’, it can be treated as a privative modifier, but when construed as meaning ‘a long time ago’ as in yec chinkwu in (45b), it cannot; yec chinkwu would mean something like ‘a person whose friendship goes back to a long time ago’.24 Similarly, hyen ‘present’ and swun ‘complete/pure’ do not fit into any of the four ADJ classes given in Table 1.1 since the semantic relation they bear to their head N is neither intersective nor subsective, privative, or modal.25

23

An anonymous reviewer suggests that despite the syllogism failure given in (55), say in (45a) is better analyzed as an intersective modifier because if X is a new hat, then it belongs to the intersection of the sets of things that are hats and those that are new. But I should point out that say moca actually denotes a set of things that are new as hats. This is evidenced by the fact that one can call something a new hat even if it is made with old cloth or fabric as long as it is newly made (see also footnote 19). Therefore, I claim that even in cases like (45a), say is best analyzed as a subsective modifier. 24 The ATT-DET yec exhibits yet another type of ambiguity when co-occurring with possessives like na-uy ‘my’. For details, see Larson and Cho 2003. 25 An anonymous reviewer suggests that hyen ‘present’ and swun ‘complete’/‘pure’ are better analyzed as modal N modifiers in a way similar to ADJs like potential and alleged in English. But I disagree with him/her on this point because while ‘potential N’ may or may not instantiate the meaning of that N in the actual world, ‘hyen N’ or ‘swun N’ does, so there is no need to resort to a possible world semantics to deal with their semantics.

2.2 Major Korean Adjectival Classes and Their Characteristic Properties

2.2.4

39

Expressions Ending in -Cek

What I call CEK-XPs are typically created by adding the SK suffix -cek (CEK)26,27 to a bi-morphemic root of the same origin, which may or may not be a free morpheme, as shown in (58)–(59). Consequently, CEK-XPs are predominantly of SK origin whereas the three adjectival classes surveyed thus far can be of either SK or NK origin. In some cases, however, the root of a CEK-XP may be a non-SK loan word, as shown in (60), though it cannot be a purely NK lexeme, as exemplified in (61). (58) a.

non-li argue-reason ‘logic’ b. non.li-cek selmyeng logic-CEK explanation ‘a logical explanation’

(59) a. *cam-ceng temporarily-decided Intended: ‘being temporarily decided’ b. cam.ceng-cek kyellon being.temporarily.decided-CEK conclusion ‘a tentative conclusion’ (60) a.

b.

opama-cek sako thinking Obama-CEK ‘an Obama-style thinking’ lathin ameyrikha-cek mwunhwa culture Latin America-CEK ‘a Latin American-style culture’

(61) a. *erini-cek sako child-CEK thinking Intended: ‘a child-style/like thinking’ b. *mengchengi-cek hayngtong idiot-CEK behavior Intended: ‘an idiot-like behavior’

26

(SK-N; a word by itself)

(SK-N + CEK)

(SK-root; not a word by itself)

(SK-N + CEK)

(Non-SK-loan-N + CEK)

(Non-SK-loan-N + CEK)

(NK-N + CEK)

(NK-N + CEK)

I treat -cek as a de-adjectival suffix but alterative analyses have also been suggested (e.g., Sohn 1999: 217). 27 The morpheme -cek has the same origin as the adnominalizing morpheme -de in Chinese but it has a far more limited distribution and function in Korean. For example, unlike -de, it does not occur on RCs. Consequently, we cannot gloss it as ADN since we reserve this label for suffix -un and its morphophonemic variants (i.e., -n and -(u)l). And for this reason, here and throughout, I gloss -cek as CEK.

40

2 Adnominal Adjectival Classes in Korean

CEK-XPs have several additional properties that set them apart from ATT-DETs. First of all, being comprised of a bi-morphemic root plus the suffix cek, they are morphologically more complex than ATT-DETs. Secondly, unlike ATT-DETs, they form an open class (modulo the restriction on their roots mentioned above). Third, none of them carries ‘adverbial’ semantics. Fourth, while ATT-DETs can only occur attributively, CEK-XPs can occur both attributively and predicatively: they occur predicatively by co-occurring with the icopula, as shown in (62). (62) CEK-XPs in predicate positions: a. I kyellon-un cam.ceng-cek-i/*ha-∅-ta. this conclusion-TOP temporally.decided-CEK-COP-N.PST-DECL ‘This conclusion is tentative.’ b. Ku-uy sarang-un mayng.mok-cek-i/*ha-ess-ta. he-GEN love-TOP blind-CEK-COP-PST-DECL ‘His love was blinded.’

Finally, unlike ATT-DETs, CEK-XPs cannot be directly modified by the degree ADV acwu ‘very’. Notably, they can be modified by acwu if they are augmented into a verbal category by co-occurring with the i-copula, but in that case, they are better analyzed as ADJs ending in -(u)n, which form another adjectival class listed in (12), namely, UN-XPs. To see this, consider (63) and (64): (63) shows that when occurring adnominally, the CEK-XPs nonli-cek and sengkong-cek can be modified by acwu but only if they have the copular verb plus the adnominalizing suffix -(u)n attached to them. On the other hand, (64) shows that CEK-XPs can occur in predicate position modified by a degree ADV but that is because they occur as part of a verbal predicate that is headed by the copular verb i-. (63) Compatibility with a degree ADV in adnominal position: a. Mina-uy acwu non.li-cek*(-i-n) selmyeng M.-GEN very logic-CEK(-COP-ADN) explanation Intended: ‘Mina’s very logical explanation’ b. Ku acwu seng.kong-cek*(-i-n) kayhyek renovation That very success-CEK(-COP-ADN) Intended: ‘that very successful renovation’ (64) Compatibility with a degree ADV in predicate position: a. Mina-uy selmyeng-un acwu nonli-cek-i-ess-ta. M.-GEN explanation-TOP very logic-CEK-COP-PST-DECL ‘Mina’s explanation was very logical.’ b. Ku kayhyek-un acwu sengkong-cek-i-ess-ta. That renovation-TOP very success-CEK-COP-PST-DECL ‘That renovation was very successful.’

2.2 Major Korean Adjectival Classes and Their Characteristic Properties

41

When we compare CEX-XPs with adjectival Ns, they exhibit more affinities than differences. First, just like adjectival Ns, CEK-XPs can occur predicatively if they occur with the i-copula and this can be seen from comparing (53) and (62). Also notable is the fact that when occurring in predicative position, neither adjectival Ns nor CEK-XPs are compatible with the ha-copula, as shown in (65). (65) Incompatibility of adjectival Ns and CEK-XPs with the ha-copula: a. *Ku mokkeri-nun kacca-ha-∅-ta. (Adjectival N + ha-copula) fake-COP-N.PST-DECL That necklace-TOP Intended: ‘That necklace is fake.’ b. *I kyellon-un cam.ceng-cek-ha-∅-ta. (CEK-XP + ha-copula) This conclusion-TOP temporally.decided-CEK-COP-N.PST-DECL Intended: ‘This conclusion is tentative.’

Next, both adjectival Ns and CEK-APs resist direct modification by the degree ADV acwu, as exemplified in (51) and (63). Finally, as mentioned in Sect. 2.2.2, both adjectival Ns and CEK-XPs may express thematic relations, although CEK-XPs’ ability to do so is delimited to cases that encode a Theme relation only (more on this property of CEK-XPs in Chap. 3). That said, there are at least two reasons to treat CEK-XPs distinctly from adjectival Ns. One is that when a CEK-XP co-occurs with an adjectival N within the same nominal boundary, the former must occur preceding the latter, as shown in (66). (66) Relative ordering between a CEK-XP and an adjectival N: a. hyek.sin-cek kyeng.cey kayhyek (CEK-XP > Adj N) innovation-CEK economy renovation ‘an innovative renovation of the economy’ b. *kyeng.cey hyek.sin-cek kayhyek (Adj N > CEK-XP) renovation economy innovation-CEK Intended: ‘an innovative renovation of the economy’

The other reason is that even though both adjectival Ns and CEK-XPs behave like nominals, they target different syntactic positions. To illustrate this, both the adjectival N kyengcey and the CEK-XP hyeksin-cek in (66a) can occur as the complement to the postposition -(u)ro ‘with’, which assigns an Instrument role to the nominal it selects for, as exemplified in (67)–(69),28 but while the adjectival N may also occur in object or subject position, as shown in (70), the CEK-XP cannot, as shown in (71).

28

In this regard, both adjectival Ns and CEK-XPs are similar to ATT-DETs since the latter may also occur as complement to the postposition -(u)ro, as shown in (50).

42

(67) Postposition -(u)ro selecting for a DP: Minho-nun swu.ken-uro meri-rul towel-with head-ACC M-TOP ‘Minho covered his head with a towel.’

2 Adnominal Adjectival Classes in Korean

tep-ess-ta. cover-PST-DECL

Ability to occur as complement to postposition -(u)ro: (68) Adjectival N: Cengpwu-nun kyeng.cey-ro nara-rul sal-li-ess-ta. economy-with nation-ACC live-CAU-PST-DECL Government-TOP ‘The government revived the nation with the economy.’ (69) CEK-XP: Cengpwu-nun hyek.sin-cek-uro kayhyek-ul tanhayngha-yess-ta. Government-TOP innovation-CEK-with renovation-ACC perform-PST-DECL ‘The government performed the renovation innovatively.’ Ability to occur in object or subject position: (70) Adjectival N: a. Kwu cengpwu-nun kyeng.cey-rul Previous government-TOP economy-ACC ‘The previous government revived the economy.’ b. Kyeng.cey-ka choy-wusen-i-∅-ta. utmost-priority-COP-N.PST-DECL Economy-NOM ‘The economy is the highest priority.’

sal-li-ess-ta. live-CAU-PST-DECL

(71) CEK-XP: a. Kwu cengpwu-nun hyek.sin(*-cek)-ul tanhayngha-yess-ta. Previous government-TOP innovation-CEK-ACC perform-PST-DECL Intended: ‘The previous government performed an innovation.’ b. Hyek.sin(*-cek)-i choy-wusen-i-∅-ta. innovation-CEK-NOM utmost-priority-COP-N.PST-DECL Intended: ‘An innovation is the highest priority.’

This set of facts shows that CEK-XPs have conflicting properties: in some respects, they behave like nominals but in other respects, they behave like ADJs. In addition, the fact that they resist direct modification by degree ADVs makes it difficult to treat them as prototypical ADJs because gradability is considered a hallmark of the ADJ category across languages (see, e.g., Kennedy and McNally 2005; cf. Morzycki 2016). Given this, we conclude that they are best analyzed as nominal ADJs or nominal adjectival expressions, and that is the label we will be using below. Taking a closer look at the semantic properties of CEK-XPs further reveals that they have a unique status among the adjectival classes in Korean: they are usually employed to modify abstract nominals or technical terms used in academia or government affairs and they tend to describe highly conceptual properties (e.g., non. li-cek ‘logical’ or ‘pertaining to logic’). This partly explains why -cek rarely occurs attached to Ns describing color, size, quantity, or quality even if they are of SK

2.2 Major Korean Adjectival Classes and Their Characteristic Properties

43

origin; presumably, such lexemes are not conceptually abstract enough. To see this, consider (72).29 (72) a.*/??cek.sayk-cek (cf. cek.sayk ‘red’; SK) Intended: ‘pertaining to red’ b. ??chwung.hyeng-cek (cf. chwung.hyeng ‘medium size’; SK) Intended: ‘pertaining to medium size’ c. ??so.ryang-cek (cf. so.ryang ‘small quantity’; SK) Intended: ‘pertaining to small quantity’ d. ?ko.kup-cek (cf. ko.kup ‘high quality’; SK) Intended: ‘pertaining to high quality’

What is also notable at this point is that CEK-XPs cannot modify proper names whereas putative ADJs in Korean can. To exemplify this, the CEK-XP non.li-cek ‘logical’ can modify a common N in.kan ‘person’ but it cannot modify a proper name Minswu, so while the nominal string non.li-cek inkan ‘a logical person’ is a well-formed expression, non.li-cek Minswu is not. Interestingly, however, the verbalized form non.li-cek-i-n, with the copular verb i- followed by the adnominalizer -un attached to the CEK-XP, can modify a proper name, as the goodness of non.li-cek-i-n Minswu shows (more on this in Sect. 2.2.5). Even though CEK-XPs undoubtedly have a unique status in the Korean lexicon, when considered under the microscope of the four-way formal semantical classification of ADJs, they turn out to be similar to adjectival Ns in that they can also express intersective, subsective, and privative adnominal relations, but not modal ones. Their usage as intersective modifiers is exemplified by (73): nonlicek here instantiates an intersective modifier because if X is a logical explanation, then it has to be both logical and an explanation, as stated in (74). (73) non.li-cek sel.myeng logic-CEK explanation ‘a logical explanation’ (74) [[nonli-cek N]] = [[nonli-cek]] ∩ [[N]]

Now, instances where CEK-XPs carry subsective semantics are given in (75). The CEK-XP cenceycek in (75a) clearly displays subsectivity because cenceycek kwuncwu refers to someone who is tyrannical as a monarch but not necessarily as something else, so the two premises in (76) yield an invalid conclusion.

29

Even though this description generally holds, recent Google search results suggest that ceksuffixation is becoming more productive in contemporary Korean. This is evidenced by the fact that what used to be judged plain ungrammatical is becoming more acceptable these days [e.g., (72d)] or acceptable to some speakers though still not to others [e.g., (72a)]. I thank an anonymous reviewer for drawing this interspeaker variation to my attention.

44

2 Adnominal Adjectival Classes in Korean

(75) a. cen.cey-cek kwuncwu monarch tyranny-CEK ‘a tyrannical monarch’ b. ka.si-cek hayngseng visibility-CEK planet ‘a visible planet’ (76) Premise: Edward was a tyrannical monarch. Premise: Edward was Catherine’s husband. ________________________________________ Conclusion: Edward was Catherine’s tyrannical husband. (INVALID)

The subsective nature of the CEK-XP kasicek in (75b) may be less clear at first glance because visible in English can receive an intersective construal when occurring postnominally (e.g., There were many stars visible tonight.), as observed by Bolinger (1967). Unlike visible, however, when occurring adnominally and without the copular verb attached to it, kasicek can only describe properties that are inherent, generic, reference-modifying, and/or enduring. Therefore, kasicek pyel refers to a subset of stars that are inherently visible. Similarly, the nominal comprised of the CEK-XP pikasicek (the antonym of kasicek) and the N pyel refers to a subset of stars that are inherently invisible. The subsective nature of pikasicek as well as kasicek is further evidenced by the fact that in Korean, one cannot utter (77) without contradicting him/herself. By contrast, its English counterpart given in (78) is coherent, and this is because the stressed invisible is construed as an intersective modifier which describes non-inherent, referent-modifying, and/or temporary properties (for details, see Larson 1998, 2000; Larson and Marušič 2004) (more on this in Chap. 3). (Note: here and below, capitalization indicates focal stress.) (77) (Context: It is cloudy tonight and there are no stars visible in the sky.) #Khaphella-nun pi-ka.si-cek ka.si-cek pyel-i-ta.30 K.-TOP not-visibility-CEK visibility-CEK star-COP-DECL Intended: ‘Capella is an INVISIBLE visible star.’ (Note: stress on pi-kasi-cek) 30

30

In order to convey the intended meaning of (77), pikasicek has to be augmented into an UN-XP, as will be shown in Sect. 2.2.5

2.2 Major Korean Adjectival Classes and Their Characteristic Properties

45

(78) (Context: It is cloudy tonight and there are no stars visible in the sky.) a. Capella is an INVISIBLE visible star. b. VISIBLE visible stars include Capella and INVISIBLE visible stars include Sirius. (Adapted from Larson 1998)

In light of this, I conclude that not only cenceycek ‘tyrannical’ but also (pi)kasicek ‘(in)visible’ are subsective modifiers, as more formally stated in (79). (79) a. b.

[[cencye-cek N]] ⊆ [[N]] [[(pi)kasi-cek N]] ⊆ [[N]]

Lastly, (80) instantiates a case where a CEK-XP carries a privative adnominal meaning. Here, the CEK-XP plus N string entails the negation of the N’s denotation, as stated in (81),31 so we are led to conclude that it is an example of a privative ADJ.32

31

This intuition has been confirmed by both the dictionary definitions for the CEK-XP hekwu-cek and five native speakers of Korean. 32 As far as I can see, CEX-XPs do not carry modal semantics, and this brings up the need to clarify Kang’s (2006) remarks on them: Kang states that the nominal string chwuceng-cek salinca is a well-formed expression. Furthermore, by translating this nominal as ‘an alleged murderer’ in English, he claims that CEK-XPs can express a modal meaning. However, from what I have found by consulting native Korean speakers and Korean dictionaries as well as conducting Google searches, chwuceng-cek salinca is not a well-formed expression: my consultants call it a non-word and furthermore, Google searches yield zero results for it. I should also note that chwuceng-cek does not mean ‘alleged’; rather, it means something like ‘constructed by inferential reasoning’. So even in cases where chwuceng-cek occurs modifying abstract Ns like cungke ‘evidence’, as shown in (i) (which appear in actual discourses but extremely rarely), the semantic relation that holds between the ADJ and the head N is that of intersection. That is, if X is chwuceng-cek cungke, then it instantiates a type of evidence in the actual world, so the MP given in (ii) holds, and this makes the ADJ an intersective one according to what is stated in Table 1.1 in Chap. 1. (i) chwuceng-cek cungke inference-CEK evidence ‘evidence that is constructed via inference’ (Lit.: ‘inferential evidence’) (ii) [[chwuceng-cek N]] = [[chwuceng-cek]]

[[N]]

Importantly, this intersective property of chwuceng-cek lets us further explain why it does not occur modifying Ns like salinca ‘murderer’ or totwuk ‘thief’ (e.g., *chwuceng-cek salinca, *chwuceng-cek totwuk): individuals instantiating murderer or thief cannot have the property of being constructed via inferential reasoning and hence the semantic incompatibility between the CEK-XP and such Ns.

46

2 Adnominal Adjectival Classes in Korean

(80) he.kwu-cek sasil fact false.construction-CEK ‘a falsely constructed fact’ (81) [[hekwu-cek sasil]] ∩ [[sasil]] = ∅

2.2.5

UN-XPs

As briefly mentioned above, what I call UN-XPs are created by adding the suffix (u)n to intransitive, stative verbal predicates. The adnominalizing morpheme -(u)n is purely of NK origin, and all Korean verbal predicates are either native or nativized lexemes. Consequently, all UN-XPs can be considered NK expressions, unlike members of the other adjectival classes surveyed thus far. Nominals comprised of an UN-XP and an N are exemplified in (82)–(84). These paradigms show that some UN-XPs are made up of ha- copular phrases and some are comprised of a CEK-XP plus the i-copula. Importantly, the fact that some UN-XPs are derived from CEK-XP plus i-copula clusters shows that they are structurally bigger than any other adjectival classes we have looked at thus far. It also suggests that their syntactic size has to be at least as big as the projection headed by the copula i- (whatever the exact node label is33). (82) a.

b.

(83) a.

b.

twungkul-n thakca table round-ADN ‘a round table’ ari.ttap-un yeca attractive-ADN woman ‘an attractive woman’ yu.myeng-ha-n famous-COP-ADN ‘a famous actor’ ppalka-h34-n red-COP-ADN ‘a red roof’

paywu actor cipwung roof

34

33

Some authors treat i- as a verb (e.g., Martin 1992) but others treat it as an ADJ (e.g., Yu 1998; Sohn 1999). Given this, I will remain agnostic as to the exact node label of UN-XPs. 34 According to Sohn (1999: 283), -h is the contracted form of the ha-copula.

2.2 Major Korean Adjectival Classes and Their Characteristic Properties

(84) a.

b.

non.li-cek-i-n logic-CEK-COP-ADN ‘a logical explanation’ wi.sen-cek-i-n hypocrisy-CEK-COP-ADN ‘a hypocritical person’

47

selmyeng explanation inkan person

The verbal origin of UN-XPs correlates with two other properties they exhibit. First, they can freely occur in predicate position but only if the suffix -(u)n is removed from them, and in this regard, they display the exact opposite behavior of adjectival Ns and CEK-XPs; recall that the other non-clausal adjectival classes require an additional morpho-syntactic device (i.e., the copular verb i-) to be able to occur predicatively if they can at all. Secondly, due to their verbal nature, when occurring in predicate position, UN-XPs can directly bear a TAM marking regardless of whether they derive from a stative or a non-stative verbal stem.35 To see this, consider (85)–(87).

35

In terms of TAM marking, stative/adjectival verb stems and non-stative verbal stems in Korean differ in that only the latter are compatible with imperfective and progressive aspect marking, which are done by the morpheme -n- and the periphrastic marker -ko iss-, respectively. In addition, unlike stative/adjectival verbal stems, non-stative verbal stems are not compatible with the null tense/aspect marker; they require an overt TAM marking in order to occur as part of an actual utterance. To see this, compare (i) and (ii).

(i) TAM marking on a stative/adjectival verbal stem: (simple non-past) a. Ku paywu-nun ci.kum paykophu- -ta. right.now hungry-N.PST-DECL That actor-TOP ‘That actor is/feels hungry right now.’ may-il paykophu-n- -ta. (non-past imperfective) b. *Ku paywu-nun every-day hungry-IMPRF-N.PST-DECL That actor-TOP Intended: ‘That actor feels hungry every day.’ ci.kum paykophu-ko iss- -ta. (non-past progressive) c. *Ku paywu-nun right.now hungry-CONN exist-N.PST-DECL That actor-TOP Intended: ‘That actor is being/feeling hungry right now.’ (ii) TAM marking on a non-stative verbal stem: may-il a. *Ku paywu-nun every-day That actor-TOP Intended: ‘That actor runs every day.’ may-il b. Ku paywu-nun every-day That actor-TOP ‘That actor runs every day.’ ci.kum c. Ku paywu-nun now That actor-TOP ‘That actor is running right now.’

talli- -ta. run-N.PST-DECL

(simple non-past)

talli-n- -ta. run-IMPRF-N.PST-DECL

(non-past imperfective)

talli-ko iss- -ta. (non-past progressive) run- CONN exist-N.PST-DECL

48

2 Adnominal Adjectival Classes in Korean

(85) a. twungkul-n thakca round-ADN table ‘a round table’ b. Ku thakca-nun That table-TOP ‘That table is round.’ c. Ku thakca-nun That table-TOP ‘That table was round.’ (86) a.

b.

c.

(87) a.

b.

c.

(attributive use) twungkul-∅-ta. round-N.PST-DECL

(predicative use)

twungkul-ess-ta. round-PST-DECL

(tense marking)

yumyeng-ha-n paywu actor famous-COP-ADN ‘a famous actor’ Ku paywu-nun yumyeng-ha-∅-ta. That actor-TOP famous-COP-N.PST-DECL ‘That actor is famous.’ Ku paywu-nun yumyeng-ha-yess-ta. famous-COP-PST-DECL That actor-TOP ‘That actor was famous.’ nonli-cek-i-n selmyeng explanation logic-CEK-COP-ADN ‘a logical explanation’ Ku selmyeng-un nonli.cek-i-∅-ta. logical-COP-N.PST-DECL That explanation-TOP ‘That explanation is logical.’ Ku selmyeng-un nonli.cek-i-ess-ta. That explanation-TOP logical-COP-PST-DECL ‘That explanation was logical.’

(attributive use)

(predicative use)

(tense marking)

(attributive use)

(predicative use)

(tense marking)

UN-XPs also differ from members of other adjectival classes in Korean in that they can be modified by any type of degree ADVs; this is shown in (88). Recall that CEK-XPs cannot be directly modified by any degree ADVs including acwu ‘very’ as illustrated in (63). And while ATT-DETs can be modified by acwu, they cannot be modified by other degree ADVs such as maywu and cham, as exemplified in (89). If we are to use gradability as a diagnostic with which to determine a lexical item’s adjective-hood, then UN-XPs’ compatibility with all types of degree ADV partly explains why they are considered as forming the ADJ class in Korean in both traditional (e.g., Choy 1971; Yu 1998; Sohn 1999) and more modern linguistics literatures (Kang 2005, 2006).

2.2 Major Korean Adjectival Classes and Their Characteristic Properties

(88) Degree ADV + UN-XP: a. [√acwu/√maywu/√cham [ very/very/truly ‘a very suspected person’ b. [√acwu/√maywu/√cham [ very/very/truly ‘a very red roof’ c. [√acwu/√maywu/√cham [ very/very/truly ‘a very big earthquake’

uysimsurep-]-un suspected-]-ADN

49

saram person

ppalka.h-]-n red-]-ADN

cipwung roof

khu-]-n big-]-ADN

cicin earthquake

(89) Degree ADV + ATT-DET: a. [√acwu/*maywu/*cham say] moca [ very/very/truly new] hat Intended: ‘a very new hat’ b. [√acwu/*maywu/*cham yec] kiek [ very/very/truly old] memory Intended: ‘a very old memory’ c. [√acwu/*maywu/*cham swun] papo [ very/very/truly complete] idiot Intended: ‘a very complete idiot’

Turning now to taking a closer look at their semantic properties: as a class, UN-XPs can modify any type of Ns whereas members of the other non-clausal adjectival classes are rather “picky” about what they modify. Related to this, UN-XPs can express any type of N modificational meaning except for the ‘adverbial’ semantics that is expressed by ATT-DETs and the grammatical meanings expressed by some adjectival prefixes. To illustrate this, first of all, the data in (82)–(84) show that UN-XPs can describe shape, physical appearance, color, and abstract conceptual properties as well as human propensities. Secondly, what is given in (90) shows that they can describe size, length, physical properties of an object, as well as age, value, and speed.

50

2 Adnominal Adjectival Classes in Korean

san (90) a. khu-n mountain big-ADN ‘a big mountain’ haru b. ki-n day long-ADN ‘a long day’ c. ttakttak.ha-n uyca be.hard-ADN chair ‘a hard chair’ saram d. celm-un person young-ADN ‘a young person’ kyelkwa e. co.h-un result be.good-ADN ‘a good/positive result’ hoypok f. pparu-n fast-ADN recovery ‘a fast recovery’

(size)

(length)

(physical property)

(age)

(value)

(speed)

Thirdly, (91) shows that some UN-XPs can engender idiomatic meanings. (91) a. khu-n ai child big-ADN ‘the eldest child/daughter-in-law’ (Lit.: ‘a big child’) b. eri-n i young-ADN person ‘a child’36 c. pparu-n kil road fast-ADN ‘a shortcut’ (Lit.: ‘a fast road’) 36

Next, as illustrated in (92), UN-XPs can describe an individual’s intelligence as well as their physical, sensory, cognitive, and/or emotive states. And notably, this is an ability that no other Korean adjectival classes introduced above are seen to have.

The noun erin-i literally means ‘a young person’ but in Korean culture, it is used as an honorific for children, so there is a bit of discrepancy between its literal meaning and its actual meaning. 36

2.2 Major Korean Adjectival Classes and Their Characteristic Properties

(92) a.

b.

c.

d.

yengli.ha-n ai child be.smart-ADN ‘a smart child’ yenyak.ha-n ai child be.sick-ADN ‘a healthy child’ tep-un saram warm-ADN person ‘a person who feels warm’ picham.ha-n saram be.miserable-ADN person ‘a person who feels miserable’

51

(intelligence)

(physical state)

(sensory state)

(cognitive/emotive state)

Finally, unlike members of the other non-clausal adjectival classes in Korean, UN-XPs can readily modify proper names, as briefly mentioned in Sect. 2.2.4 and as can be further exemplified by comparing (93) and (94). (93) Adj prefix/Adj N/ATT-DET/CEK-XP + proper name: a. *cang-Minswu (Adj prefix + proper name) long-M. Intended: ‘a tall Minswu’ b.??cang.sin Minswu37 (Adj N + proper name) tall M. Intended: ‘tall Minswu’ c. *say Minswu (ATT-DET + proper name) new M. Intended: ‘a new Minswu’ d. *cek.kuk-cek Minswu (CEK-XP + proper name) proactive-CEK M. Intended: ‘proactive Minswu’ 37 (94) UN-XP + proper name: a. cang.sin-i-n Minswu tall-COP-ADN M. ‘tall Minswu’ b. say-rop-un Minswu M. new-VS-ADN ‘a new Minswu’ c. cek.kuk-cek-i-n Minswu proactive-CEK-COP-ADN M. ‘proactive Minswu’ d. pwulssang.ha-n Minswu be.poor-ADN M. ‘poor Minswu’

37

An expression like this may be found in newspaper headlines but it is not used in ordinary discourses.

52

2 Adnominal Adjectival Classes in Korean

While we are on this topic, let me also point out that an UN-XP which directly embeds a CEK-XP plus copula cluster may have a different semantics than the CEK-XP that is embedded inside of it. To see this, compare (95a) and (95b): (95a), which contains the CEK-XP keyngcey-cek, denotes a renovation of the economy itself or a renovation of something that will impact the economy. On the other hand, (95b), which contains the UN-XP keyngcey-cek-i-n, denotes a renovation that is economical, rather than a renovation of the economy. And this shows that while a CEK-XP may bear a thematic relation to the head N or describe properties pertaining to it, the UN-XP that is derived from the CEK-XP does not. (95) a.

b.

CEK-XP + N: [kyeng.cey]-cek kay.hyek [economy]-CEK renovation Reading 1: ‘a renovation of the economy’ Reading 2: ‘a renovation pertaining/related to economy’ Corresponding UN-XP + N: [kyeng.cey-cek-i-]-n kay.hyek [economy-CEK-COP-]-ADN renovation ‘an economical renovation’

In a similar vein, when an UN-XP embeds a CEK-XP plus copula cluster, it may give rise to additional meanings. To see this, compare (96a) with (96b). As Kang (2006) observes, while kasi-cek can only describe an inherent and enduring property, its UN-XP counterpart kasi-cek-in can describe both a temporary and an enduring property. That is, the latter can receive both reference-modifying and referent-modifying interpretations in the sense of Bolinger (1967) (more on this in Chap. 3). (96) a.

b.

CEK-XP + N: Antromeyta-nun kasi-cek pyel-i-ta. Andromeda-TOP visible star-COP-DECL ‘The Andromeda is an inherently visible star.’ (reference Mod) Corresponding UN-XP + N: Antromeyta-nun kasi-cek-i-n pyel-i-ta. Andromeda-TOP visible-COP-ADN star-COP-DECL ‘The Andromeda is an inherently visible star.’ (reference-Mod) ‘The Andromeda is among the stars visible (tonight).’ (referent-Mod) [Kang 2006: 159, ex. (37)]

The foregoing shows that UN-XPs form a distinctive adjectival class among the non-clausal adjectival classes in Korean. However, in view of the four-way formal semantic classification of ADJ meanings, they turn out to be not so different from CEK-XPs. First of all, UN-XPs can be employed to express intersective semantics, as illustrated in (97). One may suspect that UN-XPs like nop-un ‘high’ are subsective

2.2 Major Korean Adjectival Classes and Their Characteristic Properties

53

but their meanings are simply vague in a way analogous to ADJs like big and small in English.38 That is, the nominal nop-un kenmwul refers to an individual that is tall for a building. (97) a.

b.

haya-h-n napi butterfly be.white-ADN ‘a white butterfly’ nop-un kenmwul high-ADN building ‘a tall building’

Secondly, UN-XPs carry subsective meanings and this is exemplified by (98): in each of these examples, the UN-XP plus an N cluster yields ‘ADJ as an N’ reading, as more formally stated in (99). (98) a.

b.

c.

(99) a. b. c.

chongmyeng.ha-n haksayng be.brilliant-ADN student ‘a brilliant student’ ttwiena-n hakca scholar remarkable-ADN ‘a remarkable scholar’ nungswuk.ha-n uysa be.skillful-ADN doctor ‘a skillful doctor’ [[chomyengha-n haksayng]] ⊆ [[haksayng]] [[ttwiena-n hakca]] ⊆ [[hakca]] [[nungswukha-n uysa]] ⊆ [[uysa]]

Finally, instances where UN-XPs express privative modificational relations are given in (100). (100) a.

kasang-cek-i-n hyensil false.imagination-CEK-COP-ADN reality ‘a falsely imagined reality’ b. hekwu-cek-i-n seykye false.construction-CEK-COP-ADN world ‘an imaginary world’ c. wico-toy-n ciphye paper.money counterfeit-become-ADN ‘counterfeit paper money’ d. nalco-toy-n sasil fact fabricated-become-ADN ‘a fabricated fact’

38

For the vagueness of ADJs like big and small, see, a.o., Siegel 1976, Partee 1995, and Heim and Kratzer 1998.

54

2 Adnominal Adjectival Classes in Korean

Before closing this subsection, I should mention that although UN-XPs like uysimsurep-un ‘suspected’ may seem like a modal N modifier, there is reason to think that they are not. To see this, consider (101). Here, the DP uysimsurep-un saram refers to a set of people who are suspected to have some property P and this shows that the UN-XP uysimsurep-un stands in an intersective relation to the N saram, rather than a modal one. (101) uysim.surep-un saram person suspected-ADN ‘a suspected person’

What is also notable is that unlike its apparent English counterpart suspected, uysimsurep-un cannot modify Ns like salinca ‘murderer’, towuk ‘thief’, and kangto ‘robber’, paralleling the behavior of chwuceng-cek noted in Footnote 32 in Sect. 2.2.4. Such properties of adnominal expressions like uysimsurep-un and chwuceng-cek raise the question of how Korean may express modal modificational relations illustrated by English examples like suspected/alleged thief and suspected/alleged murderer. The answer to this question is found in the next subsection, to which we now turn.

2.2.6

RCs

RCs form the syntactically largest and the most complex adnominal class of all in Korean (as is the case with any other human language that has RC constructions). This is evidenced by the fact that they can embed transitive verbal structures as well as non-stative, intransitive verbal structures and seemingly adjectival or stative/ intransitive ones. Moreover, regardless of its type, the embedded predicate of an RC can bear an overt TAM marking. To see this, consider (102)–(104).39

39

The way Korean marks the TAM of an embedded clause is rather complex. First, the null morpheme ∅ indicates non-past tense in the embedding clause but past tense or perfective aspect in the embedded clause, as shown in (i).

(i) a.

b.

(non-past, non-embedded clause) Na-nun Mina-ka coh- -ta. M.-NOM like-N.PST-DECL I-TOP ‘I like Mina.’ [Chelswu-ka ei cak-nyen-ey sakwi- ]-n yecai (perfective/past RC) [C.-NOM __ last-year-LOC date-PRF]-REL woman ‘The woman that Chelswu dated last year’

Furthermore, depending on whether the embedded clause’s predicate is semantically more verbal (i.e., active, dynamic, imperfective) or adjectival (i.e., stative, non-dynamic, perfective), an apparently identical morpheme may encode different TAM information. For example, while a null morpheme plus -(u)n cluster indicates past tense in verbal RCs, as shown in (ib) above, when occurring in adjectival RCs, it indicates non-past tense, as shown in (102a) in the text. For further details and discussion on the TAM marking of RCs in Korean, see, a.o., Lee 1993 and Ko 2007.

2.2 Major Korean Adjectival Classes and Their Characteristic Properties

(102) Adjectival RCs: a. [ei (cikum) yeppu-∅]-n aii [__ (now) pretty-N.PST]-REL child ‘a child who is pretty (now)’ b. [ei (han ttay) yeppu-ess-te-]-n aii [__ (once time) pretty-PST-RTRO-]-REL child ‘a child/person who used to be pretty (at some point)’ c. [ei (aphuro) yeppu-]-l aii [__ (forward) pretty-]-REL.IRR kid ‘a child/person who will become pretty (in the future)’

55

(non-past)

(past)

(irrealis)

(103) Non-stative, intransitive verbal RCs: a. [ei (cikum) ca-n-]-un sarami (non-past imperfective) person [__ (now) sleep-IMPRF-]-REL ‘a person who is sleeping (now)’ b. [ei (ku ttay) ca-te-]-n sarami (past imperfective) [__ (that time) sleep-RTRO-]-REL person ‘a person who was sleeping (at that time)’ c. [ei (aphuro) cal-]-l sarami (irrealis) [__ (forward) sleep-]-REL.IRR person ‘a person who will go to bed (shortly/in the future)’ (104) Transitive verbal RCs: a. [Chelswu-ka ei (yocum) sakwi-n-]-un yecai (imperfective) __ (these.days) date-IMPRF-]-REL woman [C.-NOM ‘the woman that Chelswu is dating (these days)’ b. [Chelswu-ka ei (cen-ey) sakwi-ess-te-]-n yecai (perfective) [C.-NOM __ (former-Loc) date-PST-RTRO-]-REL woman ‘the woman that Chelswu had dated before’ sakwi-]-l yecai (irrealis) c. [Chelswu-ka ei mi.ray-ey __ future-LOC date-]-REL.IRR woman [C.-NOM ‘the woman that Chelswu will date in the future’

On the face of it, UN-XPs constitute a subset of what I call adjectival RCs such as those given in (102). But the fact that RCs contain an overt TAM marking, whereas UN-XPs do not, gives us the justification to treat them separately, despite their apparently similar ending in some cases, namely, -(u)n. To further illustrate this, let us first consider the sentences in (105). These two sentences contain a string-identical adjectival expression which I analyze as an UN-XP here. And they show that when the embedding clause has imperfective/ habitual aspect, the dative DP containing the UN-XP only receives a generic/ non-specific interpretation, yet when the embedding clause has perfective or episodic aspect, the same DP may receive either a generic/non-specific or a non-generic/specific interpretation, but importantly, on both readings, the UN-XP in (105b) is temporally bound by the embedding clause’s tense.

56

2 Adnominal Adjectival Classes in Korean

(105) a. Chelswu-nun [DP [XP hwana-]-n saram]-eykey chincelha-∅-ta. be.kind-N.PST-DECL C.-Top [ [ angry-]-ADN person]-DAT ‘Chelswu is kind to (any) angry people.’ b. Chelswu-nun [DP [XP hwana-]-n saram]-eykey chincelha-yess-ta. be.kind-PST-DECL C.-Top [ [ angry-]-ADN person]-DAT Reading 1: ‘Chelswu was kind to (any) person/people who had the property of being habitually angry at that time.’ Reading 2: ‘Chelswu was kind to some (specific) person/people who was/were angry at that time.’

To compare the behavior of UN-XPs with that of indisputably full-fledged RCs (FRCs), consider now (106)–(107). Just like the UN-XP in (105b), the FRCs occurring in these sentences are also embedded inside a past-tense marked clause, but the events they describe can be evaluated relative to a time that is different from the embedding clause’s reference time, and this shows that unlike UN-XPs, FRCs need not be temporally bound by their embedding clauses. (106) Chelswu-nun [DP [RC ei hwa-ka na-ess-te-]-un [ [ ___ anger-NOM come.out-PST-RTRO-]-REL C.-TOP saram]-eykey chincelha-yess-ta. be.kind-PST-DECL person]-DAT Reading 1: ‘Chelswu was kind to some (specific) person who was angry at that time.’ Reading 2: ‘At some past time t1, Chelswu was kind to some (specific) person who was angry at some other past time t2.’ (107) Chelswu-nun [DP [RC Minswu-ka ei sakwi-n-]-un yecai-rul [ [ M.-NOM __ date-IMPRF-]-REL woman-ACC C.-TOP sarangha-yess-ta. love-PST-DECL Reading 1: ‘Chelswu loved the woman that Minswu was dating at that time.’ Reading 2: ‘Chelswu loved the woman that Minswu is dating now.’

FRCs display several other notable semantic properties as well. First and foremost, they can express any type of adnominal semantics except for the determiner-intensifying grammatical semantics that some ATT-DETs and adjectival prefixes are seen to carry. To illustrate this, compare (108) and (109). These paradigms show that RCs can even contribute what would be comparable to ‘adverbial’ ADJ meanings although they do so by bearing an intersective relation to their head Ns. (Here and below, pro indicates a null category that is distinct from the gap inside an RC that is co-indexed with the head N.)

2.2 Major Korean Adjectival Classes and Their Characteristic Properties

57

(108) ATT-DET + N: a. say kyengcey cengchayk new economy policy ‘a/the new economic policy’ b. yec iyaki old story ‘an old story or a story heard a long-time-ago’ (109) Corresponding RC + N: a. [DP [hyen cengpwu-ka ei say-ro kyeypalha-n-]-un [ [present government-NOM __ new-with develop-IMPRF-]-REL kyengcey cengchayki] economy policy] ‘a/the economic policy that the present government is newly developing’ b. [DP [pro ei yec nal-ey tul-ess-te-]-n iyakii [ [__ __ old day-LOC hear-PST-RTRO-]-REL story ‘a story that I heard a long time ago’

Secondly, RCs can freely modify proper names, as exemplified in (110), thereby taking on the semantics of non-restrictive N modifiers (although unlike English-type languages, Korean does not formally differentiate restrictive RCs from non-restrictive ones or vice versa; more on this in Chap. 6). (110) RCs modifying proper names: a. [ei nay.il hankwuk-uro ttena-n-]-un Minswui [__ tomorrow Korea-GOAL leave-IMPRF-]-REL M. ‘Minswu, who is leaving for Korea tomorrow’ sakwi-ess-te-]-un Minswui b. [Seyra-ka cen-ey ei [S.-NOM former-LOC __ date-PST-RTRO-]-REL M. ‘Minswu, whom Sara used to date’

Thirdly, FRCs can help express what would be comparable to subsective and privative semantics, not to mention intersective meanings. By way of illustration, the RCs in (111) are interpreted by combining with their head N’s denotations intersectively. The RCs in (112) also combine with their head Ns intersectively, but here, the ADJs embedded in them can only bear a subset relation to their head N’s denotations, so they can be said to contribute ostensibly subsective semantics. Similarly, the RCs in (113) end up contributing privative adnominal meanings since the ‘RC plus N’ combinations entail the negation of the head N’s denotation, befitting the definition of privative modifiers assumed in the formal semantics literature (e.g., Partee 1995; Morzycki 2016).

58

2 Adnominal Adjectival Classes in Korean

(111) RCs contributing an intersective adnominal meaning: a. [ei cipwung]-i ppalkah-∅]-n cipi [__ roof]-NOM be.red-N.PST-]-REL house ‘a house whose roof is red’ b. [ei huymang-ey cha-∅]-n sarami [__ hope-with be.full-N.PST-]-REL person ‘a person who is full of hope’ (112) RCs indirectly contributing a subsective adnominal meaning: epcek]-i hwullwungha-∅]-n hakcai a. [ei [DP proi [__ [ __ accomplishment]-NOM remarkable-N.PST-]-REL scholar √‘a scholar whose accomplishments as a scholar are remarkable’ *‘a scholar whose accomplishments as something else (e.g., as an educator) are remarkable’ ttay ttwiena-ess-te-]-n uysai b. [ei han [__ one time stand.out-ANT-RTRO-]-REL doctor ‘a doctor who used to be outstanding as a doctor, not as something else’ (113) RCs indirectly contributing a privative adnominal meaning: a. [ei cin.cca-ka an-i-∅]-n cincwui [__ real.thing-NOM not-COP-N.PST-]-REL pearl ‘a fake pearl’ (Lit.: ‘pearl that is not real’) b. [ei wico-ro mantul-∅]-n ciphyei [__ counterfeit-as made-PRF-]-REL paper.money ‘counterfeit paper money’ (Lit.: ‘paper money that was made as a counterfeit’)

Finally, even though FRCs do not carry modal adnominal semantics themselves, they provide a way to express it. To illustrate, consider (114). (114) RCs indirectly contributing a modal adnominal meaning: a [ei sarami salinca-ro cwuceng-toy-n- ]-un [__ murderer-as inference-PASS-IMPRF-N.PST-]-REL person ‘a person who is (being) inferred to be a murderer’ b. [ei salinca-ro uysim pat-ko iss-n- ]-un exist-IMPRF-N.PST-]-REL [__ murderer-as suspicion receive-CONN sarami person ‘a person who is (being) suspected to be a murderer’ c. [ei salin hyemuy-rul pat-ko iss-n- ]-un [__ murder suspicion-ACC receive-CONN exist-IMPRF-N.PST-]-REL sarami person ‘a person who is (being) suspected of murder’

2.2 Major Korean Adjectival Classes and Their Characteristic Properties

59

In all three examples above, the RCs and their head Ns stand in an intersective relation. Nevertheless, the RC plus head N strings approximate the semantics of nominals containing modal ADJs like alleged and suspected in English because the head N saram denotes a set of humans and the RCs denote a set of individuals who are suspected to be murderers or suspected of murder, so when they combine, they are construed as referring to the set of humans who are suspected to be murderers or are suspected of murder, as spelled out in (115) for (114b). (115)

A

B

A = {x x is a human} (head N’s denotation) B = {x x is suspected to be a murderer} (RC’s denotation) A B = {x x is a human and is suspected to be a murderer} (RC + N’s denotation)

Collectively, this set of facts shows that FRCs can express a wider range of adnominal meanings than UN-XPs. And this is most likely to be due to the clausal structure they have which allows them to spell out the content of the intended modificational meaning in ways that the smaller syntactic size of UN-XPs (or other adjectival classes) cannot. This also shows that the presence of FRCs in Korean, in a way, “makes up” for the absence of prototypical ADJs in the language, so even in this language, the full gamut of adjectival meanings can be expressed.

2.3

Summary and Discussion

In this chapter, I have introduced six major adnominal classes in Korean, focusing on their basic morpho-syntactic and semantic properties. As summarized in Table 2.1, each of these adjectival classes has a distinct status. Looking at the six adjectival classes as a whole by way of Table 2.1 leads us to see that they differ in their morpho-syntactic complexity: those introduced earlier in this chapter tend to be morpho-syntactically more simplex than those introduced later. For example, between CEK-XPs and UN-XPs, the former are more simplex than the latter, or the latter are more complex than the former. Another notable pattern that emerges from this comparison is that in Korean, morpho-syntactically simplex adjectival expressions tend to behave like nominal categories and morpho-syntactically complex ones tend to behave like verbal categories. Given this, we obtain the paradigm given in (116).

60

2 Adnominal Adjectival Classes in Korean

Table 2.1 Summary of characteristic properties of major adjectival classes in Korean Adj prefixes

Adj Ns

ATT-DETs

CEK-XPs

UN-XPs

FRCs

Mono-morphemic Bi-morphemic

Mono-morphemic Tri-morphemic Poly-morphemic Poly-morphemic

SK, NK

SK, NK

SK, NK

SK

NK

Open class

Open class

Closed class

Open class

Open class

N/A

Semi-nominal

Nominal

Nominal-like

Nominal-like

Verbal

Clausal

Non-gradable

Non-gradable

Partly gradable

Non-gradable

Fully gradable

N/Aa

Some are lexical and some are functional categories.

Largely lexical but some are functional categories.

Mostly functional Almost though some are exclusively lexical. lexical category

Lexical category

Non-lexical category but lexical in meaning.

Can express various lexical (color, size, shape) and grammatical (quantity) adnominal meanings. Can have a DET-intensifying meaning.

Can express color, size, quantity, shape, quality, nationality, and thematic relations. Can have a DET-intensifying meaning.

Temporal, degree, and ‘adverbial’ ADJ semantics. Some have a DET-intensifying meaning.

Can express any type of adnominal meaning except for modal, grammatical, and adverbial ADJ meaning. Can be non-restrictive.

Can express any type of lexical adnominal meaning. Can be non-restrictive.

Highly specified semantics: abstract, conceptual, technical meaning only.

NK

a Although the notion ‘gradability’ cannot be applied to RCs, the predicates inside them are fully gradable in that they can be modified by any type of degree ADV.

(116) Morpho-syntactic complexity vs. categorial status of Korean adjectival classes: FRCs > UN-XPs > CEK-XPs > Adj Ns > ATT-DETs/Adj prefixes Clausal verbal nominal nominal nominal/nominal (Here, ‘α > β’ indicates that α has a greater degree of morpho-syntactic complexity than β.)

As for their semantic properties, we have seen that all six adjectival classes in Korean may encode subsective and privative adnominal meanings, and most of them may encode intersective meanings, but none of them may carry modal adnominal semantics; a modal adnominal meaning is expressed indirectly via a relativization strategy. These findings are summarized in Table 2.2. Taken together, what we have uncovered here has important implications for the mapping between the morpho-syntax and the semantics of ADJs across languages. First of all, our findings show that if we apply purely morpho-syntactically based criteria, all adjectival classes in Korean are either nominal or verbal, rather than adjectival, but between themselves, they can sufficiently cover the syntactic and semantic space of N modification. Notably, the fact that Korean adjectival classes are either nominal or verbal also lends support to the widely-held belief that the category of ADJ has both [+N] and [+V] features (Chomsky 1970; Baker 2003). In addition, it lends indirect support to the claim that among syntactic categories, only N and V are truly universal (see, e.g., Van Valin 2008).

2.3 Summary and Discussion

61

Table 2.2 Korean adjectival classes versus the four-way adnominal semantic taxonomy Adjectival classes Prefixes Adj Ns ATT-DETs CEK-XPs UN-XPs FRCs

Modificational relations Intersective Subsective p p p p p * p p p p p p

Modal * * * * * *

Privative p p p p p p

Furthermore, this finding validates Dixon and Aikhenvald’s (2004) conclusion that from a holistic standpoint, any language can be said to have an ADJ category, and importantly, this validation trivializes the question of whether Korean has a proto-typical open class of ADJ or not—a question that figured prominently in the Korean linguistics literature for quite some time (see Footnote 9). Next, the way in which Korean expresses ADJ meaning sheds new light on the ADJ hierarchy given in (17), which has been traditionally assumed in the formal semantics literature. (117) Adjective hierarchy: Intersective < Subsective < Modal < Privative (Here ‘α < β’ indicates that β has a greater degree of semantic complexity than α.) (taken from Partee 2009a)

According to this hierarchy, among the four adnominal semantic relations introduced in Chap. 1, intersective relations are the least complex and privative ones are the most complex. But this makes our findings summarized in Table 2.2 look rather curious and here is why: First of all, our finding that in Korean, subsective N modifiers have a wider distribution than intersective ones is puzzling because according to (117), subsective N modifiers are supposed to have more complex semantic types than intersective ones, so one may expect the former to be less commonly attested if there is an inverse correlation between the semantic complexity of an ADJ and its frequency. Secondly, in view of the above ADJ hierarchy, our finding that all Korean adjectival classes including FRCs can encode privative ADJ semantics is somewhat surprising because it is presumably the most complex ADJ meaning. Thirdly, our observation that no adjectival expression in Korean has the semantics of a modal ADJ raises the question of why such a lacuna may exist. This question arises because in the literature, privative adnominal meanings and modal ones are often treated on a par with each other under the rubric of ‘intensionality’ (e.g., Clark 1970; Kamp 1975; Montague 1970) and therefore privative and modal modifiers are predicted to pattern together, at least in some respects, yet while Korean

62

2 Adnominal Adjectival Classes in Korean

adjectival expressions can freely express privative semantics, none of them may directly express modal meanings. While we cannot do full justice to these questions, we can find possible answers to them in Partee’s recent work on ADJs: Partee (2009a, b) points out that the traditional four-way classification of ADJ meaning cannot fully capture the behavior of “privative” ADJs like fake in English. By way of illustration, if fake is truly a privative ADJ, then (118) should be incoherent but that is contrary to fact. To be more concrete: on the traditional view, a gun cannot be a fake gun since the set of guns and the set of fake things do not intersect, so it will be redundant or even absurd to ask someone whether some gun is real or fake. However, (118) can be felicitously uttered in (almost) any discourse context. (118) Is that gun real or fake?

As a way to solve this conundrum, Partee claims that all “normal” ADJs (i.e., non-modal ones) are actually subsective; that is, what have been referred to as intersective or privative ADJs in the traditional semantics literature are just subclasses of subsectives. Under this view, then, (118) is semantically well-formed because depending on context, the denotation of gun can include any type of gun, both real and fake, and fake gun can refer to a specific subtype of guns. Partee further submits that unlike what are traditionally known as intersective, subsective, and privative ADJs, modal ADJs are not “normal”, and to substantiate this claim, she draws our attention to NP-split phenomena in Slavic languages: In a Slavic NP-split phenomenon, certain adnominal ADJs occur sentenceinitially split from the N they modify, but certain others do not, and interestingly enough, those that can undergo an NP-split are intersective, subsective, or privative ADJs, and those that cannot are modal ones.40 On the basis of the Slavic facts, Partee concludes that modal ADJs are truly distinct from other types of ADJ and therefore they merit a separate treatment even if we subsume intersectives and privatives under the rubric of subsectives. In light of Partee’s revised classification of ADJs, our finding that subsectives are the most commonly attested N modifiers in Korean is no longer surprising because all “normal” adjectival expressions are supposed to belong to the subsective class. Under this more streamlined taxonomy of adnominal semantics, the distributional parallel between subsective modifiers and privative ones in Korean is no longer curious either because privatives are in fact special sub-cases of subsective modifiers. As for the absence of modal ADJs in Korean, it can now be understood as a manifestation of a fundamental difference between modal N modifiers and the other types of N modifiers.

40

This phenomenon has been studied by Siewierska (1984), Sekerina (1997), Nowak (2000), and Pereltsvaig (2008), among others. See Partee 2009b for additional references.

2.4 Conclusion

2.4

63

Conclusion

To conclude this chapter, Korean has abundant resources for N modification and even though no adjectival class in Korean behaves like proto-typical ADJs (which exhibit the ability to directly modify an N while also being gradable), collectively, they provide rich enough resources with which to express the full spectrum of adjectival meanings. Despite their morpho-syntactic and semantic differences, all Korean adjectival classes may encode subsective adnominal semantics but none of them may directly encode modal semantics, and these findings provide additional empirical support for Partee’s (2009a, b) claim that all “normal” ADJs are subsectives under which rubric both intersective and privative ADJs are subsumed, and there is something truly special about modal adnominal modifiers both syntactically and semantically. Our probing in this chapter has shown that RCs play an important role in filling certain semantic gaps in Korean and there is an interesting correlation between the morpho-syntactic complexity of an N modifier and its syntactic type and the range of possible adnominal meanings it can carry. In the next chapter, we take a closer look at how such correlations may be best understood in the context of addressing the question of how Korean behaves with respect to several typological generalizations that have been made on the morpho-syntax and semantics of N modifiers.

References Ahn, Hee-Don, and Sungeun Cho. 2007. Subject-object asymmetries of morphological case realization. Language and Information 11: 53–76. Alexiadou, Artemis, Liliane Haegeman, and Melita Stavrou. 2007. Noun phrase in the generative perspective. Berlin: Walter de Gruyter. Baker, Mark. 2003. Verbs, nouns, and adjectives: Their universal grammar. Cambridge, MA: Cambridge University Press. Bolinger, Dwight. 1967. Adjectives in English: Attribution and predication. Lingua 18: 1–34. Chang, Suk-Jin. 1996. Korean. London Oriental and African Library 4. Amsterdam: John Benjamins. Chomsky, Noam. 1970. Remarks on nominalization. In Readings in English transformational grammar, ed. Roderick Jacobs, and Peter Rosenbaum, 184–221. Waltham: Ginn and Company. Choy, Hyen-Pay. 1971. Wuri malpon [Our grammar]. Seoul: Jengumsa. Cinque, Guglielmo. 2010. The syntax of adjectives. A comparative study. Cambridge, MA: MIT Press. Clark, Romane L. 1970. Concerning the logic of predicate modifiers. Noûs 4: 311–335. Dixon, R.M.W., and Alexandra Y. Aikhenvald (eds.). 2004. Adjective classes: A cross-linguistic typology. Explorations in linguistic typology. Oxford: Oxford University Press. Heim, Irene, and Angelika Kratzer. 1998. Semantics in generative grammar. Malden, MA: Blackwell. Hwang, Shin-Ja Joo. 1990. The relative clause in narrative discourse. Language Research 26 (2): 373–400.

64

2 Adnominal Adjectival Classes in Korean

Kamp, Hans. 1975. Two theories about adjectives. In Formal semantics of natural languages, ed. Edward Keenan, 123–155. Cambridge, MA: Cambridge University Press. Kang, Soon Haeng. 2005. On the adjectives in Korean. In University of Venice working papers in linguistics, vol. 15, 153–169. Kang, Soon Haeng. 2006. The two forms of the adjective in Korean. In University of Venice working papers in linguistics, vol. 16, 137–163. Keenan, Edward L., and Bernard Comrie. 1977. Noun phrase accessibility and universal grammar. Linguistic Inquiry 8 (1): 63–99. Kennedy, Chris, and Louise McNally. 2005. Scale structure and the semantic typology of gradable predicates. Language 81: 354–381. Kim, Min-Joo. 2002. Does Korean have adjectives. In MIT working papers in linguistics 43: Proceedings of HUMIT 2001, ed. Ionin, Tonia, Heejeong Ko, and Andrew Nevins, 71–89. Cambridge, MA: MITWPL. Ko, Youngkeun. 2007. Hankwukeuy sicey sepep tongcaksang [The aspect of the tense usage in Korean], 2nd ed. Seoul: Thayhaksa. Kwon, Nayoung. 2008. Processing of syntactic and anaphoric gap-filler dependencies in Korean: Evidence from self-paced reading time, ERP and eye-tracking experiments. San Diego, CA: University of California-San Diego dissertation. Larson, Richard K. 1998. Events and modification in nominals. In Proceedings from semantics and linguistic theory (SALT) VIII, 145–168. Cornell University Press. Larson, Richard K. 2000. Temporal modification in nominals. Paper presented at the International Roundtable on the Syntax of Tense, University of Paris VII, France. Larson, Richard K., and Sungeun Cho. 2003. Temporal adjectives and the structure of possessive DPs. Natural Language Semantics 11: 217–247. Larson, Richard K., and Franc Marušič. 2004. Indefinite pronoun structures with APs. Linguistic Inquiry 35 (2): 268–287. Lee, Hyo Sang. 1993. The temporal system of noun-modifying (attributive) clauses in Korean from a typological perspective. Studies in Language 17: 75–110. Lee, Yeong-Hyun. 2004. k-categorial grammar for Korean restrictive adjectives. Ms., Chosun University (Kwangju, South Korea). Maling, Joan, and Soowon Kim. 1998. Case assignment in the siphta-construction. In Description and explanation in Korean linguistics, ed. Ross King, 133–168. East Asia Program: Cornell University. Martin, Samuel E. 1992. A reference grammar of Korean. Tokyo, Japan: The Charles E. Tuttle Company. Mok, Jung-Soo. 2002. A study of the category ‘kwanhyeongsa’ [adnominals] and ‘hyeongyongsa’ [adjectives] in Korean grammar. Ennehak [Linguistics] 31: 71–99. Montague, Richard. 1970. English as a formal language. In Linguaggi nella società e nella tecnica, ed. Visentini, Bruno et al., 189–224. Milan: Edizioni di Comunità. (Reprinted in Montague 1974.) Morzycki, Marcin. 2016. Modification. Cambridge, UK: Cambridge University Press. Nowak, Anita. 2000. On split PPs in Polish. Ms., Univeresity of Massachusetts-Amherst. Parsons, Terence. 1970. Some problems concerning the logic of grammatical modifiers. Synthese 21: 320–334. Partee, Barbara H. 1995. Lexical semantics and compositionality. In An invitation to cognitive science, vol. I: Language, ed. Gleitman, Lila, and Mark Liberman, 311–360. Cambridge, MA: MIT Press. Partee, Barbara H. 2009a. The dynamics of adjective meaning. In Computational linguistics and intellectual technologies: Papers from the annual international conference “Dialogue 2009”, ed. A.E. Kibrik et al., 593–597. Moscow: Russian State Humanities University. Partee, Barbara H. 2009b. Formal semantics, lexical semantics, and compositionality: The puzzle of privative adjectives. Philologia 7: 11–23. Pereltsvaig, Asya. 2008. Split phrases in colloquial Russian. Studia Linguistica 62: 5–38.

References

65

Sekerina, Irina A. 1997. The syntax and processing of scrambling constructions in Russian. New York City, NY: City University of New York dissertation. Siewierska, Anna Maria. 1984. Phrasal discontinuity in Polish. Australian Journal of Linguistics 4: 57–71. Siegel, Muffy Emily Ann. 1976. Capturing the adjective. Amherst, MA: The University of Massachusetts-Amherst dissertation. Sohn, Ho-Min. 1999. The Korean language. Cambridge, UK: Cambridge University Press. Sproat, Richard, and Chinlin Shih. 1988. Prenominal adjectival ordering in English and Mandarin. In Proceedings of NELS, vol. 18, 465–489. Amherst, MA: GLSA. Sproat, Richard, and Chinlin Shih. 1990. The cross-linguistics distribution of adjectival ordering restrictions. In Interdisciplinary approaches to language: Essays in honor of S-Y. Kuroda, ed. Georgopoulos, Carol, and Roberta Ishihara, 565–593. Dordrecht: Kluwer. Suh, Jeng-Soo. 1996. Kwuke mwunpep [Korean grammar]. Seoul: Hanyang University Press. Svenonius, Peter. 2008. The position of adjectives and other phrasal modifiers in the decomposition of DP. In Adjectives and adverbs: Syntax, semantics, and discourse, ed. Chris Kennedy, and Louise McNally, 16–42. New York: Oxford University Press. Van Valin, Robert D. Jr. 2008. RPs and the nature of lexical and syntactic categories in role and reference grammar. In Investigations of the syntax-semantics-pragmatics interface, ed. Van Valin, Robert D. Jr., 161–78. Amsterdam/Philadelphia: John Benjamins. Yu, Hyen-Kyeng. 1998. Kwuke hyengyongsa yenkwu [A study on Korean adjectives]. Seoul: Hankwukmwunhwasa.

Chapter 3

Adjective Ordering Restrictions: The View from Korean

As briefly introduced in Chap. 1, the existing literature provides several typological generalizations (TGs) on the morpho-syntax and semantics of N modifiers. First of all, across languages, ADJs tend to occur following both DEMs and NUMs but preceding the N they modify, as given in (1) (see, a.o., Greenberg 1963; Hawkins 1983; Dryer 1992; Cinque 2005). (Again, unless otherwise noted, > indicates a hierarchical relation, not (just) a linear order.) (1) TG #1 Relative ordering between DEM, NUM, ADJ, and N: DEM > NUM > ADJ > N

Secondly, when a nominal has multiple N modifiers, morpho-syntactically simplex modifiers occur closer to the head N than morpho-syntactically complex ones do, as given in (2) (e.g., Bolinger 1967; Sproat and Shih 1988, 1990; Larson 1998, 2000; Bouchard 2002; Cinque 2005, 2010). (2) TG #2 Morpho-syntactic complexity of modifiers and their relative ordering: Complex Modifier > Simplex Modifier > N

Thirdly, modifiers occurring closer to the head N carry what Cinque (2010) calls ‘direct’ adnominal semantics and those occurring farther away from it carry what he calls ‘indirect’ adnominal semantics, a generalization that he offers based on numerous

This chapter is based on Kim (2014a, b), but most of the content has been revised and clarified. © Springer Nature Switzerland AG 2019 M.-J. Kim, The Syntax and Semantics of Noun Modifiers and the Theory of Universal Grammar, Studies in Natural Language and Linguistic Theory 96, https://doi.org/10.1007/978-3-030-05886-9_3

67

68

3 Adjective Ordering Restrictions: The View from Korean

existing works (e.g., Whorf 1945; Bolinger 1967; Sproat and Shih 1988, 1990; Larson 1998; Bouchard 2002; Larson and Takahashi 2007). This is given in (3).1 (3) TG #3 Correlation between the position of N modifiers and their semantics: Indirect Modifier > Direct Modifier > N Stage (S)-level Individual (I)-level Restrictive Non-restrictive Intersective Non-intersective (e.g., subsective, modal) Relative/Comparative Absolute Deictic Generic Referent-modifying Reference-modifying Extensional Intensional Predicative Attributive

Next, the prevailing generalization is that the morpho-syntactic complexity of an N modifier correlates with its interpretive possibilities (e.g., Sproat and Shih 1988, 1990; Cinque 2010; Kang 2006). According to Cinque (2010, Chap. 3) in particular, morpho-syntactically simplex N modifiers (i.e., “true” APs in his terms) carry what he calls direct adnominal semantics and morpho-syntactically complex ones (i.e., RRCs or FRCs) carry what he calls indirect adnominal semantics, as given in (4). (4) TG #4 Correlation between the morpho-syntactic complexity, the semantic possibilities, and the surface position of an N modifier relative to the head N: a. Simplex modifier (i.e., true APs) Direct adnominal semantics Indirect adnominal semantics b. Complex modifier (i.e., RRCs or FRCs) c. Indirect (reduced RC) modification Direct modification [Det. [S-level (or I-level) [I-level] NP]]] (non-adverbial) (‘adverbial’)2 [Det. [restrictive [non-restrictive3 NP]]] 2 3

The ‘direct’ versus ‘indirect’ N modifier distinction actually originates from Sproat and Shih (1988, 1990) but while they use these terms synonymously with morphologically simplex and morphologically complex N modifiers, respectively, Cinque (2010) uses them to refer to different types of adnominal semantics. 2 The ‘adverbial’ versus ‘non-adverbial’ adnominal meaning distinction is not included in the schema Cinque (2010: 33) offers, but he mentions it in the text (p. 27), treating adverbial adnominal meaning on a par with individual-level adnominal meaning. Besides, as we observed in Chap. 2, some Korean N modifiers carry adverbial semantics. Hence, I have included it here for ease of reference in the pages that follow. 3 I have included a restrictive (RS) versus non-restrictive (NRS) distinction here because Cinque (2010) makes this distinction for N modifiers and assumes that RS modifiers encode intersective/ predicative meaning whereas NRS ones encode non-intersective/attributive meaning. Notably, however, I use these terms differently: as stated in Chap. 1, I will be using the term ‘non-restrictive’ to refer to the function of an N modifier that modifies a nominal whose referent is already established in the discourse at hand and therefore, under our analysis, an NRS modifier cannot occur below an RS modifier, unlike what Cinque assumes, as one can see from (4c). 1

3 Adjective Ordering Restrictions: The View from Korean

[Det. [implicit RC [Det. [intersective [Det. [relative (to a comparison class) [Det. [comparative (with superlatives) [Det. [(non)specificity-inducing [Det. [epistemic ‘unknown’ [Det. [discourse anaphoric ‘different’ [Det. [deictic [Det. [literal interpretation possible in predicate position

69

[modal NP]]] [non-intersective NP]]] [absolute NP]]] [absolute (with superlatives) NP]]] [specificity-inducing NP]]] [evaluative ‘unknown’ NP]]] [NP dependent ‘different’ NP]]] [generic NP]]] [possible idiomatic interpretation NP]]] not possible in predicate position (adapted from Cinque 2010: (23), p. 33)

Lastly, across languages, it has been found that morpho-syntactically simplex N modifiers are subject to the AOR in (5), but morpho-syntactically complex ones (e.g., de-ADJs in Mandarin) are not; the latter type of modifiers may occur in any order relative to each other (e.g., Whorf 1945; Sproat and Shih 1988, 1990; Cinque 2010). (5) TG #5 AOR on simplex modifiers: Quality > Size > Shape > Color > Provenance/Source > N

Taken together, these five typological generalizations suggest that the morpho-syntactic complexity of N modifiers correlates with their semantics as well as their positional possibilities, and only simplex N modifiers are subject to certain AORs.4 In what follows, I test these generalizations against Korean, with a view to answering the following questions. First, does a Korean N modifier’s morpho-syntactic complexity correlate with its interpretive possibilities? If so, does being complex mean carrying indirect adnominal semantics whereas being simplex means carrying direct adnominal semantics, as assumed in Cinque 2010, which is considered the standard analysis of the syntax of N modifiers in the current literature? Second, are there any AORs in Korean? If so, does the surface position of an N modifier correlate with its morpho-syntactic complexity, as has been claimed for languages like English and Italian?

4

For this reason, Cinque (2010, Chap. 2) assumes that only simplex Ns are licensed by certain functional projections which form a hierarchy among themselves. We will come back to this issue in Chap. 4 and review his analysis more in detail there.

70

3.1

3 Adjective Ordering Restrictions: The View from Korean

Correlation Between the Morpho-Syntactic Complexity and the Interpretive Possibilities of an N Modifier

If morpho-syntactic complexity of an N modifier goes hand in hand with its interpretive possibilities, as assumed by authors like Cinque (2010) and Kang (2006), and if the TGs given in (3)–(4) are universally valid, then, given the conclusion we arrived at in Chap. 2, we should expect (6) to hold for Korean. That is, morpho-syntactically simplex modifiers like ATT-DETs and CEK-XPs (henceforth CEK-APs) carry what Cinque calls direct modificational semantics and morpho-syntactically complex ones like UN-XPs (henceforth UN-APs) and FRCs carry what he calls indirect modificational semantics. (For simplicity, I do not include adjectival prefixes and adjectival Ns in this schema.) (6) Prediction on Korean N modifiers under a Cinquerian analysis: FRC UN-AP CEK-AP ATT-DET Indirect Mod Direct Mod Referent-modifying Reference-modifying Extensional Intensional Intersective Non-intersective S-level I-level N-independent (non-subsective) N-dependent (subsective) Relative/Comparative Absolute Deictic/Non-generic Generic Predicative Attributive

The above prediction is borne out, but only partly so. In the case of ATT-DETs, they are seen to behave as expected because they mostly carry N-dependent, temporal, ‘adverbial’ semantics, and they only occur adnominally (or ‘attributively’ in Cinque’s terminology). (7) ATT-DET + N: a. say ci.phye new paper.money ‘new paper money’ (‘newly made money’) b. *Ku ci.phye-nun say-(i)- -ta. new-(COP)-N.PST-DECL That paper.money-TOP Intended: ‘That paper money is new.’

(attributive use)

(predicative use)

The behavior of CEK-APs also accords with (6) because they only carry attributive, I-level, N-dependent, and/or reference-modifying semantics, as illustrated by the interpretive behavior of kasi-cek ‘visible’ in (8).

3.1 Correlation Between the Morpho-Syntax and Semantics

(8) CEK-AP + N: kasi-cek pyel visible-CEK star ‘an inherently visible star’ *‘a star that is visible on a particular occasion’

71

(enduring property; I-level) (temporary property; S-level)

The behavior of UN-APs does not necessarily match what is depicted in (6), however: they are undoubtedly morpho-syntactically more complex than CEK-APs, yet they can carry direct as well as indirect adnominal semantics. To see this, consider first (9). Here, the UN-AP kasi-cek-i-n ‘visible’, which stems from kasi-cek, can be construed as describing an enduring property as well as a temporary one. (9) The semantic versatility of an UN-AP that is derived from a CEK-AP: kasi-cek-i-n pyel visible-CEK-COP-ADN star ‘an inherently visible star.’ (enduring property; I-level) ‘a star visible on a particular occasion.’ (temporary property; S-level)

Consider now (10) and (11). These data show that, contrary to (6), UN-APs can receive both absolute (i.e., direct) and relative (i.e., indirect) adnominal interpretations. (10) Mina-ka [kacang noph-un] san-ul olla-ss-ta. climb-PST-DECL M.-Nom [most high-ADN] mountain-ACC Reading 1: ‘Mina climbed the highest mountain in the world.’ (absolute) Reading 2: ‘Mina climbed the highest mountain under discussion.’ (relative) (11) Chelswu-nun [hwana-n] saram-eykey chincelha-yess-ta. be.kind-PST-DECL C.-Top [angry-ADN] person-DAT Reading 1: ‘Chelswu was kind to any angry individual(s).’ (generic) Reading 2: ‘Chelswu was kind to some specific person who was angry at some past time.’ (non-generic/deictic)

Consider now (12). This datum shows that some UN-APs may only receive direct Mod interpretations: when uttered out of the blue, the UN-AP ttwiena-n ‘remarkable’ can only be construed as describing a subsective, N-dependent interpretation, rather than a non-subsective, N-independent one. (12) Context: Neutral or out of the blue. ttwiena-n uysa remarkable-ADN doctor ‘a doctor who is remarkable as a doctor’ (subsective) *‘a doctor who is remarkable as something else (e.g., as a pianist)’ (non-subsective)

Such interpretive behavior of UN-APs, in particular what is exemplified by (12), casts serious doubt on analyses like Kang 2006 which adopts a Cinquerian theory of

72

3 Adjective Ordering Restrictions: The View from Korean

N modifiers for Korean without making any adjustments—that is, by treating UN-APs as inherently indirect N modifiers, based solely on the fact that they are comprised of a greater number of morphemes than CEK-APs are. Furthermore, it shows that while there is a strong correlation between the morpho-syntactic complexity of an N modifier and its interpretive possibilities, the correspondence between them is not a perfect one, unlike what a Cinquerian analysis would predict. The interpretive behavior of FRCs in Korean raises similar issues in that they may receive what Cinque (2010) calls direct adnominal interpretations in addition to what he calls indirect interpretations. To see this, consider (13). This example can be felicitously uttered in answer to ‘What kind of man are you looking for?’. And this shows that depending on context, FRCs may receive either reference-modifying or referent-modifying interpretations. (13) [ei khi-ka maywu khu-]-n namcai [__ height-NOM very big-]-REL man Reading 1: ‘a very tall man (i.e., tall for a man)’ Reading 2: ‘the very tall man (i.e., the man, who is very tall)’

(reference-Mod) (referent-Mod)

Similarly, (14) can be uttered in a context where the speaker is looking for a non-specific student who has a good scholastic aptitude, and this shows that the FRC here can receive a reference-modifying/generic interpretation as well as a referent-modifying/deictic/non-generic interpretation. (14) [ei kongpwu-rul cal ha-n-]-un haksayngi [__ study-ACC well do-IMPRF-]-REL student Reading 1: ‘a/any student who excels in studying’ (reference-Mod) Reading 2: ‘the student, who excels in studying’ (referent-Mod)

3.2

Relative Ordering Among N Modifiers in Korean

In view of (2), if Korean is a typologically well-behaving language in all respects, then (15) should hold. But in reality, it does not. (15) Prediction on the AOR in Korean based on the morpho-syntactic complexity of N modifiers involved: FRC > UN-AP > CEK-AP > ATT-DET > [(Adjectival N) Head/Root N]

When we first look at the relative ordering between an ATT-DET or a CEK-AP and an adjectival N (which forms a compound N with the head/root N), the picture looks neat and tidy since, under all circumstances, only ‘ATT-DET > Adjectival N’ or ‘CEK-AP > Adjectival N’ orders are attested, as shown in (16)–(19).

3.2 Relative Ordering Among N Modifiers in Korean

73

Relative ordering between ATT-DETs and adjectival Ns: (16) a. say(,) [yenge sensayng-nim] (ATT-DET > Adjectival N) new [English teacher-HON] ‘a new English teacher’ b. *yenge(,) say sensayng-nim (Adjectival N > ATT-DET) (17) a.

cen(,) [sangwon uywon] (ATT-DET > Adjectival N) former [higher.seat congressman] ‘a former senator’ (Lit.: ‘a former higher-seat congressman’) b. *sangwon(,) cen uywon (Adjectival N > ATT-DET)

Relative ordering between CEK-APs and adjectival Ns: (18) a. sengkong-cek [kyengcey cengchayk] (CEK-AP > Adjectival N) successful [economy policy] ‘a successful economy/economic policy’ b. *kyengcey(,) sengkong-cek cengchayk (Adjectival N > CEK-AP) (19) a.

phyenghwa-cek [nam-pwuk thongil] (CEK-AP > Adjectival N) peaceful [south-north unification] ‘a peaceful reunification of South and North Koreas’ (Lit.: ‘a peaceful South-North reunification) b. *nam-pwuk(,) phyenghwa-cek thongil (Adjectival N > CEK-AP)

The relative ordering between an UN-AP/FRC and an adjectival N also accords well with what is depicted in (15): ‘UN-AP > Adjectival N’ or ‘FRC > Adjectival N’ orders are attested but the reversed orders are not, as shown in (20)–(23). Relative ordering between UN-APs and adjectival Ns: (20) a. ywunungha-n [yenge sensayng-nim] [English teacher-HON] capable-ADN ‘a capable English teacher’ b. *yenge(,) ywunungha-n sensayng-nim

(UN-AP > Adjectival N)

(Adjectival N > UN-AP)

(21) a. conkyenghalmanha-n [sangwon uywon] (UN-AP > Adjectival N) respectable-ADN [higher.seat congressman] ‘a respectable senator’ (Lit.: ‘a respectable higher-seat congressman’) b. *sangwon(,) conkyenghalmanha-n uywon (Adjectival N > UN-AP) Relative ordering between FRCs and adjectival Ns: (22) a. [ei icey mak tayhak-ul colepha- -]-n [yenge sensayng-nim]i [__ now just college-ACC graduate-PRF]-REL [English teacher-HON] ‘an English teacher who has just graduated from college’ (FRC > Adjectival N) b. *yenge(,) [e icey mak tayhak-ul colepha- -]-n sensayng-nim (Adjectival N > FRC) (23) a. [ei wuri-ka ciciha-n-]-un [__ we-NOM support-IMPRF-]-REL ‘the economic policy that we support’ b. *kyengcey(,) [e wuri-ka ciciha-n-]-un

[kyengcey [economy cengchayk

cengchayk]i policy] (FRC > Adjectival N) (Adjectival N > FRC)

74

3 Adjective Ordering Restrictions: The View from Korean

When we look at the relative ordering between CEK-APs and ATT-DETs, however, some complications arise: according to (15), only ‘CEK-AP > ATT-DET’ should obtain but in some cases, only the reverse orders are permitted, as exemplified by (24). Relative ordering between CEK-APs and ATT-DETs: kayhyek (24) a. say(,) kyengcey-cek renovation new, economy-CEK Intended: ‘a new renovation of the economy’ b. ??kyengcey-cek, say kayhyek Intended: ‘a new renovation of the economy’

(ATT-DET > CEK-AP) (CEK-AP > ATT-DET)

Although cases like (24) are unexpected in view of (15), they are nonetheless informative for our purposes because they show that, in any human language, semantic properties of ADJs may play an important role in linearizing them. More specifically, all else being equal, an ADJ that receives a theta role from the head N occurs closer to it than an ADJ that does not (Svenonius 2008). To illustrate this with (24), here, the CEK-AP is construed as bearing a thematic relation to the head N whereas the ATT-DET describes a sortal property of individuals, and the two ADJs may occur only in the ‘ATT-DET > CEK-AP’ order. The flipside of this generalization is that if a sortal CEK-AP co-occurs with a sortal ATT-DET, then ‘CEK-AP > ATT-DET’ should obtain, conforming to (15). Such cases are exemplified by (25). Here, the CEK-AP kyengcey-cek is construed as meaning ‘economical’ rather than functioning as the semantic object of the de-verbal N kayhyek ‘renovation’ and since it is comprised of three morphemes, so is longer than the ATT-DET, it occurs farther away from the head N. (25) When a sortal CEK-AP co-occurs with a sortal ATT-DET: kyengcey-cek say kayhyek (CEK-AP > ATT-DET) economy-CEK new renovation ‘an economical new renovation (of something)’

Likewise, (26) shows that if a CEK-AP which co-occurs with an ATT-DET is construed as describing a sortal property of individuals, then it occurs preceding the ATT-DET, as expected by (15). (26) a.

hapli-cek(,) say cengchayk (CEK-AP > ATT-DET) reason-CEK new policy ‘a rational new policy’ or ‘a new rational policy’ b. cengchi-cek(,) say ciphyeng (CEK-AP > ATT-DET) politics-CEK new horizon ‘a politically new horizon’ or ‘a new political horizon’ c. tokcay-cek(,) cen kwuncwu (CEK-AP > ATT-DET) tyranny-CEK former monarch ‘the tyrannical former monarch’ (i.e., ‘the former monarch who was tyrannical’)

3.2 Relative Ordering Among N Modifiers in Korean

75

This set of facts leads us to conclude that when a CEK-AP and an ATT-DET co-occur modifying the same N and if both describe sortal properties of individuals, then they occur in the ‘CEK-AP > ATT-DET’ order but if the CEK-AP receives a thematic role from the head N, then they occur in the reversed order. I should note that even when a sortal CEK-AP occurs with an ATT-DET, ‘ATT-DET > CEK-AP’ orders may obtain but such cases are best analyzed as instantiating marked orders. To see this, consider (27) in comparison to (24a). Note that capitalization indicates phonological prominence or focal stress (FOC). (27) a.

SAY*(,) hapli-cek cengchayk (ATT-DET > CEK-AP) new rational policy ‘a NEW rational policy (not an old one)’ b. SAY*(,) cengchi-cek ciphyeng (ATT-DET > CEK-AP) new political horizon Intended: ‘a NEW political horizon (not an old one)’ c. ?CEN*(,) tokcay-cek kwuncwu (ATT-DET > CEK-AP) former tyrannical monarch ‘the FORMER tyrannical monarch (not the current one)’

Comparing (27) and (24a) leads us to see that although they all instantiate superficially identical surface orders, they actually exhibit different phonological properties: while the ATT-DET in (24a) does not require a FOC nor does it have to be followed by an intonational break, those in (27) require both of them. In addition, unlike (24a), the data in (27) require focus contexts: for example, (27a) can be felicitously uttered only when a new rational policy is being compared with an old one. Similarly, (27c) can be judged fine only if the property of being a former monarch is being emphasized by the speaker. When we turn to the relative ordering between UN-APs and ATT-DETs, again, the actual picture turns out to be more complex than what meets the eye. Here too, the relative morpho-syntactic complexity of the UN-AP and the ATT-DET at hand does matter, but there are two additional factors at work: one is whether the hosting DP is uniquely referring or not, and the other is whether the UN-AP is phonologically heavier than the ATT-DET or not. To see this, consider first (28). Here, the DP is necessarily uniquely referring and both ‘UN-AP > ATT-DET’ and ‘ATT-DET > UN-AP’ orders are permitted although the example instantiating the ‘UN-AP > ATT-DET’ order is judged to be the neutral one; when occurring in the other order as in (28b), in order for the datum to be judged grammatical, either a pause is obligatory after the ATT-DET or a FOC marking on the ATT-DET is necessary, and the UN-AP is construed as bearing narrow scope. Relative ordering between an UN-AP and an ATT-DET: (28) When the hosting DP refers to a unique entity: a. pwu.phay.ha-n(,) hyen cengkwon corrupt-ADN present government ‘the corrupt present government’ b. hyen*(,) pwu.phay.ha-n cengkwon ‘the present corrupt government’

(UN-AP > ATT-DET)

(ATT-DET > UN-AP)

76

3 Adjective Ordering Restrictions: The View from Korean

Consider now (29). This paradigm shows that when the hosting DP is not uniquely referring, only ‘UN-AP > ATT-DET’ is allowed for (modulo the fact that a different ATT-DET occurs in this data set) (more on this in Chap. 5). (29) When the hosting DP refers to a non-unique entity: a. khe.ta.rah-n kapang say new bag large-ADN ‘a new large bag’ khe.ta.rah-n kapang b. *say bag new large-ADN Intended: ‘a new large bag’ khe.ta.rah-n kapang b . *SAY bag new large-ADN Intended: ‘a NEW large bag (not a small one)’

(UN-AP > ATT-DET)

(ATT-DET > UN-AP)

There is additional complexity to cases where an UN-AP co-occurs with an ATT-DET inside a non-uniquely referring DP, however: if the UN-AP is multi-syllabic and thus is decidedly longer than the ATT-DET, then only ‘UN-AP > ATT-DET’ is permitted, as shown in (29).5 But if the UN-AP is mono-syllabic, just like the co-occurring ATT-DET, then data instantiating ‘UN-AP > ATT-DET’ are judged acceptable but only marginally so, despite the fact that they would bear out (15). To see this, consider (30). This paradigm shows that, as is, the datum instantiating the ‘UN-AP > ATT-DET’ order is not judged good, but if the UN-AP bears a FOC, then the grammaticality judgment improves, as one can see from comparing (30a) and (30a′). This paradigm further shows that when the ATT-DET bears a FOC, then, even the ‘ATT-DET > UN-AP’ order may be judged to be marginally acceptable, unlike the case with (29b′). (30) When the UN-AP is mono-syllabic and when the DP is non-uniquely referring: a. ??khu-n say kapang (UN-AP > ATT-DET) large-ADN new bag Intended: ‘a large new bag’ a . KHU-N say kapang (UN-AP with FOC > ATT-DET) large-ADN new bag Intended: ‘a LARGE new bag (not a small one)’ b. *say khu-n kapang (ATT-DET > UN-AP) new large-ADN bag Intended: ‘a new large bag’ b .??SAY khu-n kapang (ATT-DET with FOC > UN-AP) new large-ADN bag Intended: ‘a NEW large bag (not an old one)’

The grammaticality judgements on the data in (29) and (30) are obtained by consulting five (three female and two male) native speakers of Korean.

5

3.2 Relative Ordering Among N Modifiers in Korean

77

Collectively, the data in (28)–(30) lead us to conjecture that the relative phonological weight of an UN-AP and an ATT-DET may matter but only when the hosting DP is not necessarily uniquely denoting. And this conjecture is affirmed by what is given in (31): this paradigm shows that when a mono-syllabic UN-AP occurs inside a non-uniquely denoting DP accompanied by its own degree ADV modifier, it occurs preceding an ATT-DET, unlike the case with (30). If the hosting DP is uniquely denoting, however, then the ATT-DET may still occur before the UN-AP, as exemplified in (32). When an UN-AP occurs with its own degree Adv modifier: (31) When the hosting DP refers to a non-unique entity: a. [acwu khu-]-n say kapang (Adv + UN-AP > ATT-DET) [very big-]-ADN new bag ‘a very large new bag’ b. *say [acwu khu-]-n kapang (ATT-DET > Adv + UN-AP) Intended: ‘a new, very large bag’ b . *SAY [acwu khu-]-n kapang Intended: ‘a new, very large bag (not an old one)’ (32) When the hosting DP refers to a unique entity: a. [maywu pwuphayha-]-n, hyen cengkwon [very corrupt-]-ADN current government ‘the very corrupt current government’ b.(?)hyen, [maywu pwuphayha-]-n cengkwon Intended: ‘the current, very corrupt government’

(Adv + UN-AP > ATT-DET)

(ATT-DET > Adv + UN-AP)

Turning now to the relative ordering between UN-APs and CEK-APs, things also turn out to be more complex than what is depicted in (15): if the CEK-AP that co-occurs with an UN-AP is construed as bearing a thematic relation to the head N, then they always occur in the order of ‘UN-AP > CEK-AP’ irrespective of their relative phonological weights, as shown in (33)–(34)6. If the CEK-AP describes a sortal property of individuals, so is attributive, then, typically, ‘UN-AP > CEK-AP’ obtains, but in such cases, putting a FOC on the CEK-AP may allow for a ‘CEK-AP > UN-AP’ order. Besides, even without a FOC, data instantiating ‘CEK-AP > UN-AP’ order is not absolutely out. To see this, consider (35) in comparison with (33)–(34).

6 (33b′) and (34b′) would be judged better if the CEK-APs are construed as attributive, i.e., meaning ‘scientific’ and ‘economical’, respectively, but that is not what we are concerned with here.

78

3 Adjective Ordering Restrictions: The View from Korean

Relative ordering between an UN-AP and a CEK-AP: When the CEK-AP is construed as thematic: kwahak-cek palkyen (UN-AP > CEK-AP) (33) a. nollawu-n discovery surprising-ADN science-CEK ‘a surprising discovery of science’ b. *kwahak-cek nollawu-n palkyen (CEK-AP > UN-AP) surprising-ADN discovery science-CEK Intended: ‘a surprising discovery of science’ b . *KWAHAK-CEK nollawu-n palkyen Intended: ‘a surprising discovery of SCIENCE (not something else)’ (34) a. kwalmokhalmanha-n kyengcey-cek sengcang (UN-AP > CEK-AP) economy-CEK growth remarkable-ADN ‘a remarkable growth of economy’ b. *kyengcey-cek kwalmokhalmanha-n sengcang (CEK-AP > UN-AP) remarkable-ADN growth economy-CEK Intended: ‘a remarkable growth of economy’ b . *KYENGCEY-CEK kwalmokhalmanha-n sengcang Intended: ‘a remarkable growth of ECONOMY (not something else)’ When the CEK-AP at hand is construed as attributive: sayrop-un hapli-cek cengchayk (UN-AP > CEK-AP) new-ADN reason-CEK policy Intended: ‘a new rational policy’ b. ??hapli-cek sayrop-un cengchayk (CEK-AP > UN-AP) reason-CEK new-ADN policy Intended: ‘a rational, new policy’ c. ??HAPLI-CEK sayrop-un cengchayk reason-CEK new-ADN policy Intended: ‘a RATIONAL, new policy (not an irrational one)’

(35) a.

These facts reiterate that data containing an UN-AP and a CEK-AP require a more nuanced treatment than what has been done in the existing analyses such as Kang 2006: Kang assumes that all data instantiating the ‘UN-AP > CEK-AP’ order are of the same kind. But our probing shows that some of them are attested because the CEK-AP bears a thematic relation to the head N while others are attested because the UN-AP at hand is morpho-syntactically more complex than the co-occurring CEK-AP and additionally the discourse context is a neutral one. When we look at the relative ordering between FRCs and other adjectival classes in Korean, focusing on cases where the RCs occur only with lexical N dependents rather than functional ones such as DEMs, we observe that the prediction laid out in (15) is largely borne out7: when an FRC co-occurs with other N modifiers, it tends

7

In Sect. 3.5, we will see that when an FRC co-occurs with other N modifiers and the hosting DP also contains a DEM, the FRC may not necessarily occur in the left-most periphery of the DP, and in Chap. 5, I will claim that this phenomenon has something to do with the affective function that some Korean DEMs may perform.

3.2 Relative Ordering Among N Modifiers in Korean

79

to occur in the left-most position inside the nominal projection containing it regardless of whether the other N modifiers bear a FOC or not, although, in some cases, it needs to be followed by an intonational break, as shown in (36)–(43). Ordering between FRCs and ATT-DETs: hakkyo-rul palcen-sikhi-]-l say cengchayki (36) a. [ei wuri [__ we school-ACC develop-CAUS-]-REL new policy ‘a new policy that will develop our school’ (RC > ATT-DET) hakkyo-rul palcen-sikhi-]-l cengchayki b. *SAY, [ei wuri Intended: ‘a NEW policy that will develop our school (not an/the old one)’ (ATT-DET > RC) (37) a.

[ei mincwucwuuy-rul thanapha-n-]-un hyen cengkwoni [__ democracy-ACC oppress-N.PST-]-REL current government ‘the current government which oppresses democracy’ (RC > ATT-DET) b. *HYEN, [ei mincwucwuuy-rul thanapha-n-]-un cengkwoni Intended: ‘the CURRENT government which oppresses democracy (not the previous one)’ (ATT-DET > RC)

Ordering between an FRC and a CEK-AP: (38) a. [ei hakkyo-rul palcen-sikhi-]-l hapli-cek cengchayki [__ school-ACC develop-CAUS-]-REL rational-CEK policy ‘a rational policy that will develop our school’ (RC > CEK-AP) b. *HAPLI-CEK, [ei hakkyo-rul palcen-sikhi-]-l cengchayki Intended: ‘a RATIONAL policy that will develop our school (as opposed to a non-rational one)’ (CEK-AP > RC) (39) a.

[wuri motwu-ka ei para-n-]-un isang-cek sahoyi [we all-NOM __ desire- IMPRF-]-REL ideal-CEK society ‘an/the ideal society that all of us desire’ (RC > CEK-AP) sahoyi b. *ISANG-CEK, [wuri motwu-ka ei para-n-]-un Intended: ‘an/the IDEAL society that all of us desire (as opposed to a non-ideal one)’ (CEK-AP > RC)

Ordering between an FRC and an UN-AP: hakkyo-rul palcen-sikhi-]-l sayrop-un cengchayki (40) a. [ei wuri policy [__ we school-ACC develop-CAUS-]-REL new-ADN ‘a new policy that will develop our school’ (RC > UN-AP) hakkyo-rul palcen-sikhi-]-l cengchayki b.??SAYROP-UN, [ei wuri Intended: ‘a NEW policy that will develop our school (not an old one)’ (UN-AP > RC)

80

3 Adjective Ordering Restrictions: The View from Korean

(41) a.

[ei kongpwu-rul cal ha-n-]-un, chakha-n haksayngi [__ study-ACC well do-IMPRF-]-REL good.hearted-ADN student ‘a good-hearted student who is academically excellent’ (RC > UN-AP) b.??CHAKHA-N, [ei kongpwu-rul cal ha-n-]-un haksayngi Intended: ‘a GOOD-HEARTED student who is academically excellent (as opposed to an ill-hearted one)’ (UN-AP > RC)

(42) Ordering among an FRC, a CEK-AP, and an ATT-DET: a. [ei hakkyo-rul palcen-sikhi-]-l, hapli-cek, say cengchayki [__ school-ACC develop-CAUS-]-REL rational-CEK new policy ‘a new rational policy that will develop our school’ (RC > ATT-DET > CEK-AP) cengchayki b.??HAPLI-CEK, [ei hakkyo-rul palcen-sikhi-]-l, say Intended: ‘a new RATIONAL policy that will develop our school’ (CEK-AP > RC > ATT-DET) c. *SAY, [ei hakkyo-rul palcen-sikhi-]-l, hapli-cek cengchayki Intended: ‘a NEW rational policy that will develop our school’ (ATT-DET > RC > CEK-AP) (43) Ordering among an FRC, an UN-AP, and a CEK-AP: palcen-sikhi-]-l, sayrop-un, hapli-cek a. [ei kyengcey-rul [__ economy-ACC develop-CAUS-]-REL new-ADN rational-CEK cengchayki policy ‘a new rational policy that will develop economy’ (RC > UN-AP > CEK-AP) b.??HAPLI-CEK, [ei kyengcey-rul palcen-sikhi-]-l, sayrop-un cengchayki Intended: ‘a new RATIONAL policy that will develop our school’ (CEK-AP > RC > UN-AP) c. ?SAYROP-UN, [ei kyengcey-rul palcen-sikhi-]-l, hapli-cek cengchayki Intended: ‘a NEW rational policy that will develop our school’ (UN-AP > RC > CEK-AP)

To summarize this subsection, then, the morpho-syntactic complexity of Korean N modifiers serves as a determinant for their surface relative ordering, as expected in view of (15), but other factors such as the semantic properties of the N modifiers, the denotation of the hosting DP, and discourse contexts also have a role to play. Importantly, in some cases, such semantic and pragmatic factors may even override weight-based constraints. Therefore, we conclude that TG #2 does not always hold for Korean.

3.3

AOR on Simplex N Modifiers

If TG #5 holds true of Korean, then, simplex Korean N modifiers whose meanings have to do with quality, size, shape, color, and provenance/source are expected to occur in the order given in (44) but complex ones will not since they are allegedly immune to any type of AOR (e.g., Sproat and Shih 1988, 1990; Cinque 2010).

3.3 AOR on Simplex N Modifiers

81

(44) Prediction on the relative ordering of simplex N modifiers in Korean: Quality > Size > Shape > Color > Provenance/Source

We cannot test this AOR against CEK-APs and ATT-DETs, however, because their semantics has nothing to do with quality, size, shape, color, and/or source; in Korean, such attributive meanings are expressed by UN-APs, adjectival prefixes, or adjectival Ns, as shown in Chap. 2. Given this, if we delimit our attention to the behavior of UN-APs for now, then the prediction is that they will not comply with the AOR in (44). But this is not what we find, and the way multiple UN-APs co-occur inside the same nominal projection challenges the prevailing views on morpho-syntactically complex N modifiers in several respects: First of all, multiple UN-APs rarely co-occur modifying the same nominal without being conjoined by a coordinator such as -ko ‘and’ and -ciman ‘but’ but if they ever do, they are invariably accompanied by intonational breaks between them. Such a phonological property is characteristic of what Sproat and Shih (1988, 1990) call ‘parallel modifiers’, yet UN-APs cannot be treated as such because they turn out to be subject to the AOR in (44) whereas parallel N modifiers are supposed to be free from any type of AOR (Sproat and Shih 1988, 1990). To illustrate this, consider (45)–(47). These paradigms show that when uttered out of the blue, UN-APs describing color, size, and/or shape may occur only in orders that conform to what is depicted in (44), modulo the fact that they have to occur with intonational breaks between them. Ordering restrictions on two UN-APs: (45) a. khu-n*(,) palkah-n kapang big-ADN, red-ADN bag ‘a big, red bag’ b.??ppalkah-n, khu-n kapang Intended: ‘a big, red bag’ (46) a.

khu-n, twunkul-un thakca circle.shaped-ADN table big-ADN ‘a big, round table’ b. ??twunkul-un, khu-n thakca

(47) a.

kappissa-n, palkah-n kapang expensive-ADN, red-ADN bag ‘an expensive, red bag’ b. *ppalkah-n, kappissa-n kapang Intended: ‘an expensive, red bag’

(size > color)

(color > size)

(size > shape)

(shape > size) (quality > color)

(color > quality)

Consider now (48)–(50). In these examples, the UN-APs occur in ‘color/shape > size’ and ‘color > quality’ orders, contrary to what is expected given (44). But notably, the hosting nominals require focus contexts, as indicated by the English translations

82

3 Adjective Ordering Restrictions: The View from Korean

as well as the FOC marking on the relevant UN-APs.8 And this leads us to conclude that ‘size > color/shape’ and ‘quality > color’ are the unmarked orders for multiple occurrences of UN-APs in Korean. (48) Context: The speaker is trying to clarify to the hearer that she wants a red big bag, not a blue big one. Na-nun PPALKA-N, khu-n kapang-i (color > size) I-TOP red-ADN, big-ADN bag-NOM philyoha- -ta, PHARA-N khu-n kapang-i an-i-ra. need-N.PST-DECL blue-ADN big-ADN bag-NOM not-COP-CONN ‘I need a RED big bag, not a BLUE big one.’ (49) Context: The speaker is trying to clarify that she wants a round big table, not a square big one. Na-nun TWUNGUL-UN, khu-n thakca-ka (shape > size) I-TOP round.shaped-ADN, big-ADN table-NOM philyoha- -ta, NEMONA-N khu-n thakca-ka an-i-ra. need-N.PST-DECL square-shaped-ADN big-ADN table-NOM not-COP-CONN ‘I need a ROUND big table, not a SQUARE big one.’ (50) Context: The speaker is trying to clarify that she wants a red expensive bag, not a blue expensive one. Na-nun PPALKA-N, kappissa-n kapang-i (color > quality) expensive-ADN bag-NOM I-TOP red-ADN, philyoha- -ta, PHARA-N kappissa-n kapang-i an-i-ra. need-N.PST-DECL blue-ADN expensive-ADN bag-NOM not-COP-CONN ‘I need a RED expensive bag, not a BLUE expensive one.’

I should also point out that UN-APs may occur in apparent violation of (44) if one of them occurs with an overt degree modifier but not the others. To see this, consider (51)–(53). In these paradigms, nominal data instantiating ‘color/ shape > size’ are judged fine and those instantiating the reversed order are not, exemplifying the opposite pattern of (45)–(47). But here, the UN-APs occurring more to the left of the head N are accompanied by a degree ADV, unlike the case

8

As will be further discussed in Chap. 4, information structural factors can incur violation of the AOR in (44) in other languages as well. For example, as Svenonius (2008) points out, FOC-marked ADJs in English may freely occur in non-canonical orders. By way of illustration, (i) below shows that when bearing a FOC, a shape-denoting ADJ may occur preceding a size-denoting one and a material- or a source-denoting ADJ may occur preceding a quality-denoting one. (i) a. big square table; *square big table; SQUARE big table b. expensive wooden table; *wooden expensive table; WOODEN expensive table c. tasty French cheese; *French tasty cheese; FRENCH tasty cheese. [Svenonius 2008: 35, ex. (35)]

3.3 AOR on Simplex N Modifiers

83

with (45)–(47). And given this, we are led to conclude that a weight-based constraint is at work in linearizing UN-APs as well as what is stated in (44). Ordering between two UN-APs, where one of them contains a degree ADV: (51) a. [acwu ppalkah-n], khu-n kapang (color > size) [very red-ADN] big-ADN bag ‘a very red, big bag’ b. ??khu-n, [acwu ppalkah-n] kapang (size > color) bag big-ADN [very red-ADN] Intended: ‘a big, very red bag’ (52) a.

[acwu twunkul-un], khu-n [very circle.shaped-ADN] big-ADN ‘a very round, big table’ b. ??khu-n, [acwu twunkul-un] big-ADN [very circle.shaped-ADN] Intended: ‘a big, very round table’

(53) a.

[acwu ppalkah-n], kappissa-n [very red-ADN] expensive-ADN ‘a very red, expensive bag’ b. ??kappissa-n, [acwu ppalkah-n] expensive-ADN [very red-ADN] Intended: ‘an expensive, very red bag’

thakca table

(shape > size)

thakca table

(size > shape)

kapang bag

(color > quality)

kapang bag

(quality > color)

To summarize the foregoing, then, in principle, multiple occurrences of UN-APs abide by the AOR in (44), yet when one of them becomes phonologically heavier as a result of bearing a FOC or being modified by a degree ADV, an apparent violation of the AOR may be incurred. These facts show that UN-APs exhibit a behavior that is not typical of morphologically complex N modifiers, unlike what has been claimed by authors like Kang (2006). They further show that, in Korean, a phonological weight-based constraint may override other typologically well-established AORs such as (44). Turing now to examining how FRCs in Korean behave with regard to the predictions made about morpho-syntactically complex N modifiers: they turn out to be free from the AOR given in (44), which is as expected. And illustrative data are given in (54): this paradigm shows that when two FRCs co-occur, if their meanings have to do with size and color, then they may occur in either order relative to each other, unlike the case with the corresponding UN-APs given in (45). (54) a. [ei momcip-i acwu khu-n], [ei thel-i onthong large-REL] [__ fur-NOM all [__ body-NOM very soi cow ‘a cow whose body is really big whose fur is all black’ b. [ei thel-i onthong kkamah-n], [ei momcip-i acwu soi ‘a cow whose fur is all black whose body is really big’

kkamah-n] black-REL] (size > color)

khu-n] (color > size)

84

3 Adjective Ordering Restrictions: The View from Korean

Notably, however, it is not the case that FRCs may occur in any order relative to each other. To see this, consider (55), which is taken from Larson and Takahashi 2007. Here, the (b) example is decidedly worse than the (a) example, and this can be taken to suggest that Korean FRCs are in fact subject to some sort of AOR. (55) Data suggesting some AOR on FRCs: a. [RC1 nay-ka ei ecey manna- -n], [ I-NOM __ yesterday meet-PRF-REL] ei phyengso tampay-rul manhi phiwu-n-un] [RC2 [ __ usually cigarette-ACC a.lot smoke-IMPRF-REL] Chelswu-i-ta. sarami-un person-TOP C.-COP-DECL ‘The person that I met yesterday who smokes a lot is Chelswu.’ (RC1 > RC2) b.?? [RC2 ei phyengso tampay-rul manhi phiwu-n-un], ei ecey manna- -n] sarami-un Chelswu-i- -ta. [RC1 nay-ka (RC2 > RC1)

Larson and Takahashi take the badness of data like (55b) to suggest that Korean FRCs are linearized according to whether their embedded predicate is S-level or I-level, more specifically, whether they are ordered in such a way that an RC with an S-level predicate (RCS) invariably occurs farther away from the head N than an RC with an I-level predicate (RCI) does in a manner analogous to what Del Gobbo (2005) has claimed about Mandarin FRCs. In other words, in languages like Korean and Mandarin, if RCs containing a similar type of predicates co-occur, they can occur in any order relative to each other, but if their predicates belong to different “levels” (i.e., one is an S-level and the other is an I-level predicate), then they must occur instantiating the ‘RCS > RCI’ order. At first glance, the descriptive generalization Larson and Takahashi offer seems to be empirically adequate since, in both (56) and (57) below, the ungrammatical cases instantiate ‘RCI > RCS’ whereas the grammatical cases instantiate ‘RCS > RCI’ orders. Cases where an RCS and an RCI co-occur: (56) a. [DP [RC1 nay-ka ei ecey manna- ]-n, __ yesterday meet-PRF]-REL [ [ I-NOM [RC2 ei phyengso tampay-rul manhi phiwu-n-]-un a.lot smoke-IMPRF-]-REL [ __ usually cigarette-ACC sarami]-un Chelswu-i- -ta. (RCS > RCI) C.-COP-N.PST-DECL person]-TOP ‘The person that I met yesterday who smokes a lot is Chelswu.’ b. *[DP [RC2 ei phyengso tampay-rul manhi phiwu-n-]-un, sarami]-un Chelswu-i- -ta. manna- ]-n [RC1 nay-ka ei ecey (RCI > RCS) Intended: ‘The person that smokes a lot who I met yesterday is Chelswu.’

3.3 AOR on Simplex N Modifiers

85

(57) a.

[DP [RC1 Mina-ka yocum ei sakwi-n-]-un, these.days __ date-IMPRF-]-REL [ [ M.-NOM [ RC2 ei elkwul-i nemwuna calsayngki- ]-n namcai]-nun [ __ face-NOM truly be.handsome-N.PST]-REL man]-TOP Cinho-i- -ta. (RCS > RCI) C.-COP-N.PST-DECL ‘The man that Mina is dating these days who is so handsome is Cinho.’ b. *[DP [RC2 ei elkwul-i nemwuna calsayngki- ]-n, [RC1 Mina-ka yocum ei sakwi-n-]-un namcai]-nun (RCI > RCS) Cinho-i- -ta. Intended: ‘The man who is so handsome that Mina is dating these days is Cinho.’

A closer examination reveals, however, that Larson and Takahashi’s generalization cannot encompass the full spectrum of facts. To see this, consider first (58): this pair of examples shows that in some cases, ‘RCI > RCS’ is not only permitted but also strongly preferred over ‘RCS > RCI’. [DP [RC1 Mina-ka phyengso ei culkye-pwuru-n-]-un, usually __ enjoy-singing-IMPRF-]-REL [ [ M.-NOM [RC2 ei chilsipnyentay-ey ywuhaynghay-ss-te-]-n be.popular-ANT-RTRO-]-REL [ __ 70’s-LOC norayi]-nun leylitpi-i- -ta. (RCI > RCS) let.it.be-COP-N.PST-DECL song-TOP ‘The song that Mina likes to sing which used to be popular in the 1970s is Let It Be.’ b.?/ [DP [RC2 ei chilsipnyentay-ey ywuhaynghay-ss-te-]-n, [RC1 Mina-ka phyengso ei culkye-pwuru-n-]-un norayi]-nun leylitpi-i- -ta. (RCS > RCI) Intended: ‘The song that used to be popular in the 1970s which Mina likes to sing is Let It Be.’

(58) a.

Consider now (59) and (60). These paradigms show that, unlike what Larson and Takahashi report, RCs containing similar types of predicates do not always occur in any random order either. (59) Cases where two RCI’s co-occur: a. [DP [RC1 Mina-ka ei i seysang-eyse kacang __ this world-LOC most [ [ M.-TOP cohaha-n-]-un, [RC2 ei maumssi-ka nemwuna chakha- ]-n be.good-N.PST]-REL like-IMPRF-]-REL [ __ heart-NOM really chinkwui]-nun Cinhi-i- -ta. (RC1 > RC2) friend]-TOP C.-COP-N.PST-DECL ‘The friend that Mina likes the most in this world who has a truly beautiful heart is Cinhi.’ ei maumssi-ka nemwuna chakha- ]-n, b. ??[DP [RC2 Mina-ka ei i seysang-eyse kacang cohaha-n-]-un [RC1 chinkwui]-nun Cinhi-i- -ta. (RC2 > RC1) Intended: ‘The friend who has a truly beautiful heart who Mina likes the most in this world is Cinhi.’

86

3 Adjective Ordering Restrictions: The View from Korean

(60) Cases where two RCS’s co-occur: a. [DP [RC1 Minaj-ka ei kacang cal kiekha-n-]-un, __ most well remember-IMPRF-]-REL [ [ M.-NOM [RC2 ei kwusipnyentay-chopan-ey iss-ess-te-]-n exist-ANT-RTRO-]-REL [ __ 1990’s-early-LOC ili]-un [proj emma]-hako-uy yehayng-i- -ta. (RC1 > RC2) Mom]-with-GEN trip-COP-N.PST-DECL event]-TOP [__ ‘The event that Mina remembers the best which happened in the early 1990s is the trip with her mom.’ ei kwusipnyentay-chopan-ey iss-ess-te-]-n, b. *[DP [RC2 Minaj -ka ei kacang cal kiekha-n-]-un ili]-un [RC1 [proj emma]-hako-uy yehayng-i- -ta. (RC2 > RC1) Intended: ‘The event that happened in the early 1990s which Mina remembers the best is the trip with her mom.’

Taken together, this set of facts shows that while the S- versus I-level predicate distinction that Larson and Takahashi (2007) discuss seems to play some role in linearizing FRCs in Korean, there are other factors at work as well.9 In conjunction with the data containing UN-APs, they also show that the prevailing views about the AOR on morpho-syntactically complex N modifiers merit rethinking.

9

This is not just a Korean-specific phenomenon: as Juchin Jhang (2012) observes, Mandarin Chinese presents similar challenges to Larson and Takahashi’s (2007) analysis, as shown in (i) and (ii), taken from her work. I revisit the Chinese cases in Chap. 6, after presenting how to best deal with the Korean data in Chap. 5, and suggest a way to account for them within a formal syntactic framework.

(i) Cases where an RC with an I-level predicate co-occurs with an RC with an S-level predicate: a. [Ta xihuan de] [Jay xie de] na shou ge DE] [J. wrote DE] DEM CL song [She likes hen you-ming (RCI > RCS) very have-fame ‘The song that she likes which Jay wrote is famous’ b. *[Jay xie de] [ta xihuan de] na shou ge DE] [she likes DE] DEM CL song [J. wrote hen you-ming (RCS > RCI) very have-fame Intended: ‘The song that she likes which Jay wrote is famous’ (ii) Cases where two RCs with S-level predicates co-occur: de] [ni shang-ci shuoguo de] ren (RC1 > RC2) a. [Wo zuotian yudao DE] [you last-time talk.EXP DE] person [I yesterday met ‘The person that I met yesterday whom you mentioned last time’ b. *[Ni shang-ci shuoguo de] [wo zuotian yudao de] ren (RC2 > RC1) DE] person [You last-time talk.EXP DE] [I yesterday met Intended: ‘The person that I met yesterday whom you mentioned last time’

3.4 Relative Order Between DEM, NUM, and ADJ

3.4

87

Relative Order Between DEM, NUM, and ADJ

Since Korean is a head-final, CL language whose CLs and PL marker are bound morphemes, in view of TG #1, we predict that (61) will hold true of the language: (61) Prediction on the linear ordering of DP-internal elements in Korean: DEM > NUM + CL > ADJ > N + PL

This prediction is borne out but interestingly, when all three types of N dependents (i.e., a DEM, a NUM, and an ADJ) co-occur, they need to be accompanied by intonational breaks. To see this, consider (62) and (63), which contain an animate and an inanimate N, respectively. Here, the pauses between the NUM and the ADJ are obligatory whereas those between the DEM and the NUM are not. (62) Ordering among DEM, NUM, ADJ, and an animate N: DEM > NUM + CL > ADJ > N + PL ku(,) sey-myeng-uy*(,) khikhu-n haksayng-tul tall-ADN student-PL that three-CL-GEN ‘Those three tall students’ (63) Ordering among DEM, NUM, ADJ, and an inanimate N: DEM > NUM + CL > ADJ > N + PL ku(,) sey-kwon-uy*(,) caymiissnu-n chayk-tul that three-CL-GEN interesting-ADN book-PL ‘Those three interesting books’

What is also worth pointing out is that the N dependents in (62)–(63) may occur in alternate orders, as shown in (64)–(65), although, again, intonational breaks are necessary in some locations but not in others.10,11

10

I address why some of these are necessary in Chap. 5. The idea I put forward there will be that pauses indicate phrasal movement or the presence of an adjunction structure inside a DP structure. 11 Based on two psycholinguistic experiments, Joo (2013) claims that Korean speakers actually most strongly prefer the surface order given in (64d) and (65d) (i.e., ‘DEM > A > N + PL > NUM + CL’), preferring the order given in the (a) examples (i.e., ‘DEM > A > NUM + CL > N + PL’) next, and disliking the orders given in the (b)–(c) examples. To account for this, she resorts to Hawkins’ (2004) domain minimization theory of processing, suggesting that, for processing purposes, all else being equal including compositionality related factors, the surface order that minimizes the distance between the head N and its dependents is the most preferred one. But her experiment data were given without any pauses between the nominal constituents, unlike the way the data in (64)–(65) are constructed, even though doing so would greatly impact grammaticality judgments for Korean, as we see here and will continue to see in subsequent pages. Besides, they were presented without any contextualization. Therefore, even though I am sympathetic to the conclusions she draws, her experiment results do not undermine the grammaticality of the data provided here. For relevant discussion, see also Lee 2000.

88

3 Adjective Ordering Restrictions: The View from Korean

(64) Alternate orders for (62): a. DEM > ADJ > NUM + CL > N + PL: ku(,) khikhu-n*(,) sey-myeng-uy that tall-ADN three-CL-GEN b. ADJ > DEM > NUM + CL > N + PL: (?)khikhu-n*(,) ku*(,) sey-myeng-uy tall-ADN that three-CL-GEN c. NUM + CL > DEM > ADJ > N + PL: sey-myeng-uy*(,) ku*(,) khikhu-n that tall-ADN three-CL-GEN d. DEM > ADJ > N + PL > NUM + CL: ku(,) khikhu-n haksayng-tul(,) student-PL that tall-ADN ‘Those three tall students’

haksayng-tul student-PL haksayng-tul student-PL haksayng-tul student-PL sey-myeng three-CL

(65) Alternate orders for (63): a. DEM > ADJ > NUM + CL > N + PL: ku(,) caymiissnu-n*(,) sey-kwon-uy chayk-tul three-CL-GEN that interesting-ADN book-PL b. ADJ > DEM > NUM + CL > N + PL: (?)caymiissnu-n*(,) ku*(,) sey-kwon-uy chayk-tul interesting-ADN student-PL that three-CL-GEN c. NUM + CL > DEM > ADJ > N + PL: sey-kwon-uy*(,) ku*(,) caymiissnu-n chayk-tul that interesting-ADN book-PL three-CL-GEN d. DEM > ADJ > N + PL > NUM + CL: ku(,) caymiissnu-n chayk-tul(,) sey-kwon that interesting-ADN book-PL three-CL ‘Those three interesting books’

As far as I can tell, there are no truth-conditional differences associated with these variant orders. This means that their difference is most likely information-structural, with the examples given in (62)–(63) representing the canonical constituent order which would be uttered in neutral contexts, and those in (64)–(65) representing non-canonical constituent orders. That said, exactly what would yield such variant orders is a question that merits an answer. I take up this question in Chap 5 (Sect. 5.1) after presenting an analysis that articulates the base-positions of DEMs, NUMs/CLs, and PL in Korean. For now, I turn to discussing what to make of the findings we have made so far.

3.5

What to Make of Our Findings

Applying the five well-established TGs introduced at the beginning of this chapter to Korean has yielded several interesting findings, which can be summarized as follows:

3.5 What to Make of Our Findings

89

(66) Summary of the findings about Korean N modifiers made in this chapter: a) Korean DEMs, NUMs, and ADJs occur observing TG #1, but they may occur in various other orders as well (see Sect. 3.4). b) Morpho-syntactically more complex N modifiers tend to occur farther away from the head N than morpho-syntactically simplex ones do, in agreement with TG #2, but this constraint may be overridden by other factors such as focus, relative phonological weight, and the [+/_referential] properties of the hosting DP (see Sect. 3.2). c) An N modifier that carries ‘direct’ adnominal semantics does not always occur farther away from the head N than an N modifier that carries ‘indirect’ adnominal semantics does, unlike what is predicted by TG #3 (see the behavior of FRCs in Sect. 3.1). d) Morpho-syntactically complex N modifiers may carry direct adnominal semantics, contrary to TG #4 (see the behavior of UN-APs and FRCs in Sect. 3.1). e) Even complex N modifiers are subject to some sort of AOR, unlike what is predicted by TG #5 (see the behavior of UN-APs and FRCs in Sect. 3.3). f) AORs obtain even for parallel modification cases (see the behavior of UN-APs).

These findings show that Korean defies all the solidly established typological generalizations on N modification. In particular, the Korean facts teach us that one cannot assume that being morpho-syntactically simplex means carrying direct adnominal semantics and being morpho-syntactically complex means carrying indirect adnominal semantics, contra what has been claimed in the literature (e.g., Cinque 2010; Kang 2006). Nevertheless, what is also clear is that, even in Korean, all else being equal, simplex modifiers tend to occur closer to the head N than complex modifiers do. That is, both phonological weight and morpho-syntactic complexity matter in linearizing Korean N modifiers, just like how such factors may impact the linear ordering of N modifiers in relatively well-studied languages like English and Italian. Additional evidence that a phonological weight-based AOR figures prominently in Korean comes from the fact that when different types of N modifiers co-occur modifying the same nominal, what determines their surface order is the number of morphemes comprising them, and for this reason, the cross-linguistically well-attested AOR in (5) may be violated by simplex N modifiers in Korean. To see this, consider (67). Here, a source-term and a size-term co-occur modifying the same N, so in view of (5), the ‘size > source’ order should obtain but only the reversed order is permitted. What is notable here, however, is that one N modifier is comprised of three morphemes (i.e., cwung-kwuk-cey) but the other is comprised of just two (i.e., so-hyeng), and what is judged to be good instantiates the case where the heavier one occurs more to the left. Therefore, we conclude that at the end of the day, what determines the relative ordering between the two N modifiers is actually their relative weight, that is, the heavier one first and the lighter one next.

90

3 Adjective Ordering Restrictions: The View from Korean

(67) a. cwung-kwuk-cey12 so-hyeng ratio center-country-made small-shape radio Intended: ‘a small-sized radio manufactured in China’ b.??so-hyeng cwung-kwuk-cey ratio Intended: ‘a small-sized radio manufactured in China’ 12

(source > size)

(size > source)

Essentially the same argument can be made with (68)–(69). Given TG #5, we expect ‘shape > color’ but not ‘color > shape’, so the data in (69) comes as a surprise for us at first glance. Crucially, however, in both (68) and (69), the good cases instantiate the ‘heavier modifier > lighter modifier’ order and therefore they abide by the weight-based AOR we have just identified, which I tentatively label as ‘CONSweight’ for convenience. sam-kak-hyeng-uy huy-n-sayk13 white-ADN-color three-point-shape-GEN ‘a white, triangle-shaped chair’ b. *huy-n-sayk(,) sam-kak-hyeng-uy white-ADN-color three-point-shape-GEN Intended: ‘a white, triangle-shaped chair’

(68) a.

uyca chair

(shape > color)

uyca chair

(color > shape)

uyca chair

(shape > color)

uyca chair

(color > shape)

13

(69) a. *sam-kak(,) huy-n-sayk-uy three-point white-ADN-color-GEN Intended: ‘a white, triangle-shape chair’ b. huy-n-sayk-uy sam-kak three-point white-ADN-color-GEN ‘a white, triangle-shape chair’

By the same reasoning, the contrast between (70a) and (70b) makes sense if we consider the fact that even though both huynsayk ‘white’ and samkak ‘triangle’ are comprised of two syllables, one is morpho-syntactically more complex than the other. More concretely, while samkak is composed by having an adjectival prefix sam ‘three’ attached to the bound root N -kak ‘angle’, huynsayk is composed by having an intransitive NK verbal root huy- ‘be.white’14 combine with the adnominalizer -un, so together, they form an UN-AP, which modifies an N of Sino-Korean origin, namely, sayk.

As I noted in footnote 15, Chap. 2, by using pwullanse-cey ‘France made’, cwungkwuk-cey is a nominal even though its English translation is not. Therefore, (67a) instantiates a case where two adjectival Ns modify a noun (which happens to be a loan word). 13 Supporting evidence for this analysis of huy- comes from the fact that it can occur predicatively without being accompanied by a copular verb, as exemplified in (i). 12

(i) Ku uyca-nun huy- -ta. be.white-N.PST-DECL That chair-TOP ‘That chair is white.’ 14 The fact that huy-n is an UN-AP also lets us make sense of why, when the shape-denoting N modifier in (68) occurs without the genitive marker -uy, the color-denoting term huy-n-sayk may occur either before or after it, as exemplified in (ii). The reason for this “flexible” word order is that

3.5 What to Make of Our Findings

91

(70) a.??sam-kak huy-n-sayk uyca three-point white-ADN-color chair Intended: ‘a white, triangle-shaped chair’ b. huy-n-sayk sam-kak uyca white-ADN-color three-point chair ‘a white, triangle-shaped chair’

(shape > color)

(color > shape)

The idea that Consweight is at work in Korean receives further support from paradigms like (71): (71a) is judged far worse than (71b), despite the fact that the former accords with TG #5 and the latter does not. But notably, while (71a) has two UN-APs modifying an N, (71b) contains just one UN-AP and the other N modifier occurs in the form of an adjectival prefix. And this suggests that Korean has a strong tendency to employ N modifiers of different weights for different types of adnominal meaning and then linearize them accordingly by putting heavier constituents farther away from the head N. (71) a.??khu-n haya-n cethayk large-ADN white-ADN mansion Intended: ‘a large, white mansion’ b. haya-n tay-cethayk white-ADN large-mansion ‘a large, white mansion’

(size > color: ‘UN-AP > UN-AP > N’)

(color > size: ‘UN-AP > Prefix > N’)

Imposing Consweight on Korean alone cannot capture the full range of facts, however, since we have already seen in Sect. 3.2 that the surface orders between an ATT-DET, a CEK-AP, and an UN-AP are actually more fluid than what is predicted by (15). In Sect. 3.2, I suggested that such unexpected linear orders are due in part to pragmatic factors, namely, whether some adjectival meaning is being focused (or foregrounded) in the given discourse context. In this section, I would like to show that similar factors may also trigger relatively light N modifiers to occur in non-canonical surface positions. To illustrate this, consider (72)–(75). In these paradigms, the DEM ku occurs in the left-most position of the DP and the presence of this DEM somehow lets a CEK-AP or an UN-AP occur preceding an FRC, unlike what we saw in (38)–(42). What is also notable is that here, the CEK-APs or the UN-APs must be followed by a

the two N modifiers are both comprised of three morphemes, unlike in (68), where the shape-term consists of four morphemes. (ii)

a. sam-kak-hyeng, huy-n-sayk white-ADN-color three-point-shape-GEN ‘a white, triangle-shaped chair’ b. huy-n-sayk, sam-kak-hyeng three-point-shape white-ADN-color ‘a white, triangle-shaped chair’

uyca chair

(shape > color)

uyca chair

(color > shape)

92

3 Adjective Ordering Restrictions: The View from Korean

pause and when they occur instantiating the ‘DEM > CEK-AP/UN-AP > RC’ order, there has to be no pause after ku (see the (b) examples). Moreover, there are some semantic differences correlated with the different surface orders. For instance, (72a) can be felicitously uttered in a context where the content of the RC is construed as contributing backgrounded and/or supplementary information. By contrast, (72b) can be uttered only if the RC’s content is construed as an integrated part of the message being conveyed and the content of the AP is being emphasized. Co-occurrence of an FRC and a CEK-AP preceded by a DEM: (72) a. ku*(,) [RC ei hakkyo-rul palcen-sikhi-]-l*(,) [AP hapli-cek] that [ __ school-ACC develop-CAUS-]-REL [ rational-CEK] cengchayki (DEM > RC > CEK-AP) policy ‘the/that rational policy, which will develop our school’ b. ?ku [AP halpli-cek]*(,) [RC ei hakkyo-rul palcen-sikhi-]-l (DEM > CEK-AP > RC) cengchayki Intended: ‘that rational policy that will develop our school’ (73) a.

b.

ku*(,) [RC ei mincwucwuuy-rul thanapha- ]-n*(,) that [ __ democracy-ACC oppress-PRF]-REL [AP tokcay-cek] kwuncwui (DEM > RC > CEK-AP) [ tyrant-CEK] monarch ‘the/that tyrannical monarch, who oppressed democracy’ ku [AP tokcay-cek]*(,) [RC ei mincwucwuuy-rul thanapha- ]-n kwuncwui (DEM > CEK-AP > RC) Intended: ‘that tyrannical monarch who oppressed democracy’

Co-occurrence of an FRC and an UN-AP preceded by a DEM: (74) a. ku*(,) [RC Mina-ka ei sa-o- ]-n*(,) [AP kappissa-n] __ buy-come-PRF]-REL [ expensive-ADN] that [ M.-NOM (DEM > FRC > UN-AP) mwulken-tuli item-PL ‘the/those expensive things, which Mina bought and brought’ b. ku [AP kappissa-n]*(,) [RC Mina-ka ei sa-o- ]-n mwulken-tuli (DEM > UN-AP > FRC) Intended: ‘those expensive things that Mina bought and brought’ ku*(,) [RC Mina-ka ei sakwi-ess-te-]-n*(,) [AP calsayngki-n] that [ M.-NOM __ date-PRF-RTRO]-REL [ handsome-ADN] (DEM > FRC > UN-AP) namca-tuli man-PL ‘the/those handsome men, who Mina used to date’ b. ku [AP calsayngki-n]*(,) [RC Mina-ka ei sakwi-ess-te-]-n namca-tuli (DEM > UN-AP > FRC) Intended: ‘those handsome men that Mina used to date’

(75) a.

3.5 What to Make of Our Findings

93

Taken together, these facts show that the presence of a DEM in the same DP structure and occurring adjacent to it enables a morpho-syntactically light N modifier to occur preceding morphologically heavy ones, and since typical DEM-marked DPs in Korean are all definite and/or deictic, we are led to hypothesize that the phenomenon at hand may be triggered by some discourse prominent meaning that DEM-marking on a DP is supposed to encode (whatever it is). If this idea is on the right track, then it will show yet again that discourse-pragmatic factors play a non-trivial role in linearizing N modifiers in Korean. Therefore, I revisit this issue in Chap. 5, taking a close look at the phenomenon at hand. Before closing this chapter, I would like to identify still another property of Korean N modifiers by drawing the reader’s attention to the minimal pairs in (76) and (77): both of these data sets instantiate a case where a noun is modified by two nominal APs, one quality-denoting and the other size-denoting, but while (76) meets our expectations, (77) does not. Given (5) and Consweight, (77a) is expected to be grammatical and (77b) is expected to be ungrammatical, contrary to fact. (76) a.

[N choy-sin-sik], [N tay-hyeng] thipi [ ultimate-new-style] [ big-shape] television ‘a most-recent-style, big-sized television’ b. ??[N tay-hyeng], [N choy-sin-sik] thipi Intended: ‘a big, most-recent style television’

(77) a. *[N choy-sin-sik], [UN-AP khu-n] [ ultimate-new-style] [ big-ADN] Intended: ‘a most-recent-style, big television’ b. ?[UN-AP khu-n], [N choy-sin-sik] Intended: ‘a big, most-recent style television’

(quality > size > N)

(size > quality > N)

thipi (quality > size > N) television thipi

(size > quality > N)

At first glance, the mirror image between (76) and (77) is puzzling, but when we pay attention to the categorial status of each of the APs, things begin to make sense because we quickly realize that khu-n ‘big’ is a verbal AP whereas the other APs in (76)–(77) are all nominal categories (see Chap. 2 for the details), and this realization leads to the hypothesis that the AOR in Korean may be governed by the categorial status of the modifiers at hand as well, and all else being equal, it is more desirable for a verbal N modifier to occur preceding a nominal N modifier. When we incorporate this idea as well as our findings in Sect. 3.2 into a version of (15), we obtain (78) as the unmarked ordering among the four major adjectival classes in Korean. (Here, CEK-APATT is short for ‘an attributive CEK-AP’ and CEK-APh is short for ‘a thematic CEK-AP’.)

94

3 Adjective Ordering Restrictions: The View from Korean

(78) Correlation between the morpho-syntactic complexity, categorial status, and the relative position of major adjectival classes in Korean: FRC > UN-AP > CEK-APATT > ATT-DET > CEK-AP > (Compound) N Clausal verbal nominal nominal nominal farther away from/str (Here, > indicates a greater degree of morpho-syntactic complexity.)

What is depicted in (78) reflects the correlation between the morpho-syntactic complexity of an N modifier and its relative surface position. In addition, it informs us about the correlation between the categorial status of an N modifier and its relative surface position since it can be read as: ‘In Korean, all else being equal, a more nouny modifier occurs closer to the head N than a more adjectival or verby one does; an adjectival modifier occurs closer to the head N than a verby one does; and a verby modifier occurs closer to the head N than a clausal one does’.

3.6

Conclusion

In this chapter, I have shown that in Korean, the TGs given in (1), (3), (4), and (5) do not necessarily hold, but the TG given in (2) does to the extent that is possible. This means that one of the deciding factors in linearizing Korean N modifiers is not which modifier is a color- versus a shape-term or a size- versus a quality-term, but rather which of them is phonologically heavier or morpho-syntactically more complex. We have also observed, however, that meaning-related factors (e.g., whether one of the co-occurring APs is thematic or not, whether the hosting DP is uniquely referring or not, and whether some particular property of individuals is being focused or not) may result in apparent violation of (2), although in such cases, the hosting DP may have to contain an obligatory intonational break after the “offending” AP, or a focused item, and/or the DEM ku. Another important revelation has been that in Korean, the categorial status of N modifiers serves as another determinant in linearizing them if they co-occur inside the same nominal projection. In view of these findings, we can conclude that the AOR phenomenon in Korean is regulated (in part) by the following set of constraints: (79) Constraints regulating the adjective ordering in Korean: a. Heavier Modifier > Lighter Modifier > N b. Discourse-Old Modifier > Discourse-New Modifier > N c. Verbal Modifier > Nominal Modifier > N d. Compositional semantics (e.g., scope, relative vs. absolute semantics)

Obviously, we still need to find out exactly how these constraints interact with each other. Related to this, what is particularly pressing is the question of which constraint may override which others and under what circumstances. Furthermore,

3.6 Conclusion

95

we need to address what governs the ordering of FRCs when multiple of them co-occur modifying the same nominal, why occurring adjacent to the DEM ku ‘the/ that’ followed by a pause may let APs violate Consweight, and why an intonational break is necessary when an N modifier occurs in a non-canonical position, possibly carrying a focused meaning. Providing answers to these questions is the goal of Chap. 5. But some of them will already be being answered in Chap. 4, as I develop the theoretical framework within which to capture the AOR phenomenon across languages by articulating the positions of clausal and focused N modifiers inside a presumably universal DP structure.

References Bolinger, Dwight. 1967. Adjectives in English: Attribution and predication. Lingua 18: 1–34. Bouchard, Denis. 2002. Adjectives, number and interfaces: Why languages vary? Amsterdam: North-Holland. Cinque, Guglielmo. 2005. Deriving Greenberg’s Universal 20 and its exceptions. Linguistic Inquiry 36: 315–332. Cinque, Guglielmo. 2010. The syntax of adjectives. A comparative study. Cambridge, MA: MIT Press. Del Gobbo, Francesca. 2005. Chinese relative clauses: Restrictive, descriptive or appositive? In Contributions to the XXX Incontro di Grammatica Generativa, ed. Brugè, Laura, Giuliana Giusti, Nicola Munaro, Walter Schweikert, and Giuseppina Turano, 287–305. Venezia: Cafoscarina. Dryer, Matthew. 1992. The Greenbergian word order correlations. Language 68: 865–895. Greenberg, Joseph H. 1963. Some universals of grammar with particular reference to the order of meaningful elements. In Universals of language. Report of a conference held at Dobbs Ferry, NY, April 13–15, 1961, ed. Greenberg, Joseph H., 73–113. Cambridge, MA: MIT Press. Hawkins, John A. 1983. Word order universals. New York: Academic Press. Hawkins, John A. 2004. Efficiency and complexity in grammar. Oxford, UK: Oxford University Press. Jhang, Juchin. 2012. Multiple prenominal adjectival modifiers’ ordering restrictions in Mandarin Chinese. Ms., Texas Tech University. Joo, Kum-Jeong. 2013. Universal 20 and word order variation in Korean. In Working papers in linguistics: University of Hawai‘i at Mānoa, vol. 44 (2). http://www.ling.hawaii.edu/research/ WorkingPapers/wp-KumJeongJoo.pdf. Accessed 28 Mar 2018. Kang, Soon Haeng. 2006. The two forms of the adjective in Korean. In University of Venice working papers in linguistics, vol. 16, 137–163. Kim, Min-Joo. 2014a. On the position of adnominal adjectival expressions in Korean. In Peaches and plums, ed. Huang, C.-T. James, and Feng-hsi Liu, 75–99. Taipei, Taiwan: Academia Sinica. Kim, Min-Joo. 2014b. Adjective ordering: A view from Korean. In Proceedings of the 31st west coast conference on formal linguistics, ed. Robert Santana-LaBarge, 285–294. Somerville, MA: Cascadilla Proceedings Project. Larson, Richard K. 1998. Events and modification in nominals. In Proceedings from semantics and linguistic theory (SALT) VIII, 145–168. Cornell University Press. Larson, Richard K. 2000. Temporal modification in nominals. Paper presented at the International Roundtable on the Syntax of Tense, University of Paris VII, France.

96

3 Adjective Ordering Restrictions: The View from Korean

Larson, Richard K., and Naoko Takahashi. 2007. Order and interpretation in prenominal relative clauses. In Proceedings of the workshop on Altaic formal linguistics II. MIT working papers in linguistics, vol. 54, 101–120. Cambridge, MA: MITWPL. Lee, Chungmin. 2000. Numeral classifiers, (in-)definites and incremental theme in Korean. In Korean syntax and semantics: LSA institute workshop, Santa Cruz, ’91, ed. Chungmin Lee, and John Whitman. Seoul: Thaehaksa. Sproat, Richard, and Chinlin Shih. 1988. Prenominal adjectival ordering in English and Mandarin. In Proceedings of NELS, vol. 18, 465–489. Amherst, MA: GLSA. Sproat, Richard, and Chinlin Shih. 1990. The cross-linguistics distribution of adjectival ordering restrictions. In Interdisciplinary approaches to language: Essays in honor of S-Y. Kuroda, ed. Georgopoulos, Carol, and Roberta Ishihara, 565–593. Dordrecht: Kluwer. Svenonius, Peter. 2008. The position of adjectives and other phrasal modifiers in the decomposition of DP. In Adjectives and adverbs: Syntax, semantics, and discourse, ed. Kennedy, Chris, and Louise McNally, 16–42. New York: Oxford University Press. Whorf, Benjamin Lee. 1945. Grammatical categories. Language 21: 1–11.

Chapter 4

A New Formal Analysis of AOR

In this chapter, I develop a theoretical framework within which to account for the behavior of N modifiers in Korean as well as other languages by taking Cinque’s (2010) and Larson’s (1998, 2000) works as the points of departure. Cinque’s work merits our initial consideration because it is by far the most extensive analysis of the syntax of ADJs that is currently available in the generative linguistics literature (at least at the time of writing this book). It also integrates the findings of numerous other important works predating it (e.g., Greenberg 1963; Bolinger 1967; Vendler 1968; Hetzron 1978; Givón 1980; Dixon 1982, 2004; Cinque 1990, 1994, 2005; Sproat and Shih 1988, 1990; Larson 1998, 2000; Beck 1999; Giusti 2002; Bouchard 2002; Scott 2002; Laenzlinger 2005; Alexiadou et al. 2007). Furthermore, the pied-piping analysis it offers is shown to successfully capture the relevant cross-linguistic data drawn from Romance and Germanic languages, so it is likely to yield similarly positive results for other languages or language families as well. Larson’s work, on the other hand, deserves our consideration here because it is the only analysis I know of which provides a way to capture the direct versus indirect adnominal semantic distinction within a formal semantic framework. Besides, it has the potential to capture the syntax and semantics of RCs as they pertain to AOR, a relatively poorly understood area or topic, not only in English but also in other languages (see Larson and Takahashi 2007 for details). This chapter, therefore, begins with a brief review of these two works in light of the Korean facts carved out in the previous chapter. After this, we take a close look at Svenonius’ (2008) decompositional DP analysis and Laenzlinger’s (2005) complex DP analysis, as they will prove to provide additional insights and tools with which to account for the Korean data as well as data drawn from other languages. We then turn to presenting the new syntactic analysis of N modifiers, demonstrating how it works for English and for some Romance language data involving DEMs and ADJs.

© Springer Nature Switzerland AG 2019 M.-J. Kim, The Syntax and Semantics of Noun Modifiers and the Theory of Universal Grammar, Studies in Natural Language and Linguistic Theory 96, https://doi.org/10.1007/978-3-030-05886-9_4

97

98

4.1

4 A New Formal Analysis of AOR

Lessons from Cinque (2010) and Larson (1998, 2000)

Taking essentially the same position as Sproat and Shih (1988, 1990), Cinque (2010) assumes that there is a direct correspondence between the morpho-syntactic complexity of an adjectival expression and its semantics: in brief, morpho-syntactically simplex attributive APs carry what he calls ‘direct’ adnominal semantics and morpho-syntactically complex ones carry what he calls ‘indirect’ adnominal semantics. He also states that direct adnominal semantics encompasses non-intersective, attributive, non-restrictive, I-level, modal, absolute, evaluative, and NP-dependent interpretations whereas indirect adnominal semantics encompasses intersective, restrictive, S-level, comparative, epistemic, discourse-anaphoric, and speaker-oriented interpretations, as we briefly talked about in Chap. 3. Another tenet of Cinque’s analysis is that DP structure contains various FPs that are dedicated to different types of adnominal semantics—a view that is now considered standard due to his own seminal work (Cinque 1994) and much subsequent work after it (see, a.o., Brugè 1996, 2002; Bernstein 2001; Isac 2003; Giusti 2002; Scott 2002; Laenzlinger 2005). More specifically, he claims that direct N modifiers (i.e., “true APs” in his terms) are licensed by FPs that are immediately above NP, whereas indirect N modifiers (i.e., RRCs and FRCs) are licensed by FPs that are located much higher in the DP structure but still below NumP. In addition, he hypothesizes that more than one FP of the same type can be projected within the same DP and hence multiple adnominal elements may co-occur modifying the same nominal, as exemplified by a happy young blind Belgian sheep dog (Hetzron 1978: 179) and my big white old round wooden table in English. Furthermore, drawing on his own previous work on RCs (Cinque 2003, 2008a, b), he suggests that a DP contains a small indefinite FP, which he labels as ‘dP’, and d “carries some sort of referential import though not the uniquely individuating referential import of the higher D” (p. 34), and while direct N modifiers merge below dP, indirect ones merge above it.1 The key elements of Cinque’s theory of the syntax and semantics of N modifiers are schematically represented in (1), which contains two APs base-generated under dP and two RCs base-generated above it.

1

Cinque (2010) also states that D marks the (maximal) intersection between the set contributed by dP and the set contributed by the RC(s) above it (p. 34), so no modifiers that occur below dP will have intersective or restrictive adnominal semantics, which is why he treats “true” APs as non-intersective modifiers. But in Sect. 4.3, I show that things are a bit more complicated than the picture he presents to us by way of (1).

4.1 Lessons from Cinque (2010) and Larson (1998, 2000)

99

(1) Cinque’s universal DP structure with two types of N modifiers (Mods): DP NumP FP2 FRC

FP2

Indirect Mod

dP RRC

d

FP1

AP1

FP1

Direct Mod AP2

NP

Cinque’s analysis provides an elegant account of why seemingly identical adjectival expressions may receive different interpretations depending on their syntactic position. To illustrate, consider (2), which is adapted from Larson 1998. In this sentence, the unstressed visible is construed as describing an inherent, reference-modifying, and attributive property of individuals whereas the stressed visible and invisible are construed as describing temporary, referent-modifying, and intersective properties, and according to Cinque, this is so because while ADJs occurring closer to an N are true APs, those occurring far-away from it are disguised RCs. (2) (Context: The speaker is looking at the night sky on a cloudy day.) VISIBLE visible stars include Capella and INVISIBLE visible stars include Sirius.

When applied to Korean, Cinque’s analysis yields several positive results. For example, it lets us capture the semantic difference between the two UN-APs that appear in (3): in this sentence, the outer UN-AP can only be construed as describing a temporary, occasion-based, and referent-modifying property of individuals in a way analogous to the stressed occurrence of invisible in (2), whereas the inner one can only be construed as describing an inherent, semi-permanent, and reference-modifying property of individuals, resembling the unstressed occurrences of visible in (2). And under his analysis, the two UN-APs receive such construals because while one is a true AP that is base-generated under dP, the other is a disguised or reduced RC which is base-generated above dP.

100

4 A New Formal Analysis of AOR

(3) Onulpam, pi-kasi-cek-i-n kasi-cek-i-n tonight, not-visible-CEK-COP-ADN visible-CEK-COP-ADN pyel-un Antromeyta-i- -ta. star-TOP Andromeda-COP-N.PST-DECL ‘Tonight, a star that is inherently visible but is temporally invisible is Andromeda.’

In addition, his analysis provides a straightforward answer to why morpho-syntactically heavy N modifiers in Korean tend to occur in the left periphery of a DP whereas morpho-syntactically light ones tend to occur closer to the head N: in his analysis, the two types of N modifiers behave the way they do because they simply have different merge sites, namely, above versus below dP, as depicted in (1). Despite such merits, Cinque’s analysis cannot be adopted for our purposes as is, because it incorrectly predicts that morpho-syntactically complex N modifiers will only carry indirect adnominal semantics, contrary to fact; FRCs can describe characteristic properties of individuals, as we observed in Chap. 3 and as further exemplified by (4)–(5).2 Hence, according to the way he defines ‘direct adnominal semantics’, they should be analyzed as being able to carry reference-modifying semantics, but that would contract his analysis of indirect N modifiers. (4) a. John is [NP a student [who excels in school]]. b. Mary is [NP a person [who teaches music]]. (5) Korean: a. Cinho-nun [NP [ei hakep sengcek-i C.-TOP [ [__ scholastic grade-NOM wuswuha- ]-n haksayngi]-i- -ta. be.remarkable-N.PST]-REL student]-COP-N.PST-DECL ‘Cinho is a student who excels in school.’ b. Mina-nun [NP [ei umak-ul karuchi-n-]-un teach-IMPRF-]-REL M.-TOP [ [__ music-ACC sarami]-i- -ta. person]-COP-N.PST-DECL ‘Mina is a person who teaches music.’

2

The non-referential status of the nominals that are modified by the RCs in (4)–(5) is evidenced by the fact that the nominal projection containing the RC in (4a) cannot be referred back to by using a pronoun. To see this, consider (i). (More on this in Sect. 4.3.1.)

(i) a. John is a student who excels in school. He, that is, John, aced this semester again. b. John is a student who excels in school. *He, that is, the student, aced this semester again.

4.1 Lessons from Cinque (2010) and Larson (1998, 2000)

101

Relatedly, the fact that kasicekin in (3) is construed as describing an inherent property presents a problem to a Cinquerian analysis. The reason is that UN-APs are morpho-syntactically more complex than CEK-APs, so under his line of analysis (e.g., Kang 2006), they would be predicted to carry only indirect adnominal semantics, but that is not what we see in their actual interpretive behavior (more on this in Chap. 5). Unlike Cinque, Larson does not subscribe to the idea that only a functional category can introduce an N modifier to a nominal structure. Moreover, he hypothesizes that there is no positional restriction on N modifiers but they may receive only a certain type of interpretation depending on where they occur inside an NP (or a DP): if an N modifier occurs in what he calls ‘NP-inner space’, it receives a direct adnominal interpretation in the sense of Sproat and Shih (1988, 1990) regardless of whether it occurs before or after the head N. If it occurs in what he calls ‘NP-outer space’, it receives an indirect adnominal interpretation irrespective of its position relative to the head N. Notably, Larson’s syntax-semantics mapping hinges on the assumption that the generic operator (GEN) in the sense of Chierchia (1995) is present at the NP level and this operator will bind any adjectival expression that occurs inside the NP but not those that occur beyond it. Consequently, under his analysis, if an N modifier is located in the inner space of an NP, then it is bound to receive an intensional, N-dependent, reference-modifying interpretation but if it is located in the outer space, then it receives an extensional, N-independent, referent-modifying construal, as schematically given in (6). (Here, XP stands for any type of N modifier including FRCs.) (6) Larson’s analysis of direct vs. indirect N modifiers: [DP XP5 [ [XP3 [NP GEN XP1 XP2] ] XP4] ] Indirect Mod Direct Mod Indirect Mod

Larson’s analysis provides a way to explain why even morpho-syntactically complex expressions may describe inherent or generic properties of individuals, as shown in (4)–(5). It can also readily handle cases like (7) in which two FRCs that occur modifying the same N are both construed as carrying direct adnominal semantics, describing characteristic properties of individuals.

102

4 A New Formal Analysis of AOR

(7) a. I am looking for a student [who receives excellent grades at school][who provides a good example for other students]. b. Korean: Na-nun [ei hakep sengcek-i wuswuha- ]-n, [__ scholastic grade-NOM be.remarkable-N.PST]-REL I-TOP [ei tha-uy mopem-i toy-n-]-un haksayngi-ul good.example-NOM become-IMPRF-]-REL student-ACC [__ others-GEN chac-ko iss- -ta. exist-N.PST-DECL look.for-CONN ‘I am looking for a student who receives excellent grades who provides a good example for other students.’

However, a purely Larsonian analysis will fall short of capturing the partly phonology-driven aspect of AOR in Korean and in other languages, namely, the fact that cross-linguistically, heavier N modifiers tend to occur farther away from the head N than lighter ones do. Therefore, we need to complement his analysis by marrying it with a version of Cinquerian analysis. In sum, then, the foregoing shows that, for our purposes, what is in need is a mechanism that captures the semantic versatility of complex N modifiers (as Larson’s analysis does) which also ensures that, all else being equal, they occur in the left periphery of a DP (as Cinque’s analysis does). In addition, the new analysis needs to provide an answer to how and why discourse related factors such as focus and definiteness may impact the surface ordering of N modifiers, as we saw in Chap. 3. And since where exactly discourse salient elements occur inside a DP can be best answered by spelling out the positions of all relevant DP-internal functional elements including DEMs and NUMs, any adequate new analysis will have to take a more decompositional approach to AOR than what Cinque’s and Larson’s analyses have done. And this is where more recent works like Svenonius 2008 and Laenzlinger 2005 come into play.

4.2 4.2.1

Lessons from Svenonius (2008) and Laenzlinger (2005) Svenonius’ (2008) Decompositional Analysis of DP

Svenonius holds the mainstream view that DP contains a layered functional structure and N modifiers enter the DP structure by merging at the specifier (Spec) position of the relevant FPs. But departing from authors like Cinque (1994) and Scott (2002), he claims that FPs which introduce ADJs and which are also part of UG are just a few in number. To illustrate this by briefly comparing his analysis with Scott’s (2002): while Scott proposes (8) as the hierarchy that is formed by

4.2 Lessons from Svenonius (2008) and Laenzlinger (2005)

103

ADJ-introducing FPs in all languages,3 Svenonius argues that the categories posited in such fine-grained analyses “are not well-motivated outside of the adjectival ordering phenomenon that they are introduced to describe” (p. 35). (8) Scott’s (2002) proposal for a universal hierarchy of AP-related FPs inside DP: Determiner > Ordinal > Cardinal > Subjective Comment > Evidential > Size > Length > Height > Speed > Depth > Width > Temperature > Wetness > Age > Shape > Color > Nationality/Origin > Material > Compound Element > NP

Svenonius further claims that actually attested ADJ orderings are not as rigid as what would be predicted by hierarchies like (8), and this is seen by the fact that while it is true that ‘length > width’ is commonly attested across languages as shown in (9), the reversed order is also permitted as exemplified in (10), unlike what would be predicted by Scott’s line of analysis. (9) Data instantiating the ‘length > width’ order: a. a long thin knife. b. *a thin long knife. (Svenonius 2008: (46); emphasis mine) (10) Data instantiating the ‘width > length’ order: a. a long thick rope b. a thick long rope c. thin long strands of pasta d. a thin long charm necklace (adapted from Svenonius 2008, (47), and parts of his (49); emphasis mine)

To propose a universal DP structure that is motivated outside of ADJ ordering phenomena, Svenonius examines the relative ordering among various functional N dependents, and by integrating the findings of the relevant cross-linguistic research, he proposes several smaller hierarchies for various DP-internal functional categories, which collectively elaborate on Greenberg’s (1963) Universal 20. First, he treats an article (Art), a plural marker (Pl), and an N as heads and offers (11) as the relative hierarchy between them. (11) Art > Pl > N

3

Scott (2002) develops this idea by extending Cinque’s (1999) treatment of ADVs. In Cinque’s theory, ADVs are introduced by FPs which form a fixed universal hierarchy at the clausal level.

104

4 A New Formal Analysis of AOR

Second, he identifies three levels of classifier, namely, (i) UNIT, which refers to a numeral classifier (which can be equated with Borer’s (2005) #), (ii) SORT, which refers to a sortal classifier (which can be equated with Borer’s (2005) Cl), and (iii) n, which refers to an N classifier (given Marantz 2001). And he suggests that these three classifiers are functional heads which form the following hierarchy among themselves: (12) UNIT > SORT > n

Thirdly, based on the observation that a plural marker and a sortal classifier rarely co-occur, he puts Pl and SORT in the same slot, thereby suggesting the hierarchy given in (13). (13) Art > UNIT > Pl/SORT > n > N

Fourth, he treats DEMs, NUMs, and ADJs as phrasal elements, which respectively modify a DP, a PlP, and an NP, although he also notes that in some languages, DEMs can be lexicalized as D heads, just as how the English numeral one sometimes seems to occur as a lexicalized head. Next, regarding the position of DEM, he concludes that ‘Dem > Art’ is the basic order although he acknowledges that in some languages (e.g., Berbice Dutch Creole, Galela, Samoan), ‘Art > Dem’ may also obtain, as shown in (14).4 (14) o

tahu mane house this ‘this house’

(Art > N > DEM)

ART

(Rijkhoff 2002:184)

In addition, he concludes that NUM is higher than both Pl and UNIT and therefore ‘Num > UNIT > Pl’ has to be the underlying hierarchy among them. Finally, he posits that ADJs can occur not only above and below Pl/SORT but also above and below n (and possibly in other places as well). When we combine these hierarchies, we obtain (15) as the universal hierarchy among various N dependents that would be predicted by Svenonius (2008). (15) Svenonius’ (2008) proposal for a universal hierarchy among various N dependents: Dem > Art > Num > UNIT > Adj > Pl/SORT > Adj > n > N

4

In this context, Svenonius (2008) also mentions Brugè’s (2002) claim that, in all languages, DEMs originate low but they move to [Spec, DP] either in overt syntax or at LF, but he states that her claim is “hard to reconcile with Greenberg’s (1963) Universal 20” (p. 26, footnote 6). I make a note of his remarks here because, below, I argue for a lower origin analysis of DEMs as well (see Sect. 4.4), but my analysis will be shown to be not subject to the problems that afflict Brugè’s analysis (see Sect. 4.5).

4.2 Lessons from Svenonius (2008) and Laenzlinger (2005)

105

Comparing this hierarchy with what is given in (8) leads us to see that, in Svenonius’ analysis, only two or three FPs are responsible for introducing adnominal ADJs, unlike the case with more mainstream analyses of N modifiers such as Scott 2002 and Cinque 2010. What is also notable is that, under his analysis, a DP is comprised of just a handful of FPs, and each of them introduces a different type of N-dependent. More specifically, the functional head Art introduces DEMs, which are phrasal categories; UNIT introduces NumP; Pl introduces plural morphology; SORT introduces a gradable and subsective AP; n introduces a non-gradable and intersective APpwhose meaning has to do with ‘nationality’, ‘source’, and ‘material’; and finally, , i.e., ‘root’ in the sense of Marantz (2001), introduces an ADJ which forms an idiom with the head N. In order to yield various surface constituent orders that differ from what is given in (15), Svenonius adopts Aboh’s (2004) ‘roll-up’ movement approach, under which phrasal movements occur in sequence, largely for cluster-formation purposes—that is, to ensure that certain heads occur adjacent to each other. He explicitly remarks, however, that all languages share a certain basic functional skeleton for DP and yet the exact internal DP structure may vary from language to language because some languages may innovate a language-particular functional head if needed (p. 40).5 Another important aspect of Svenonius’ analysis is the idea that all types of N modifiers may occur anywhere inside the DP as long as the resulting structure is compatible with compositional semantics, an idea that echoes Larson’s (1998, 2000) analysis but differs from Cinque’s (2010). To demonstrate how Svenonius’ analysis may actually work, let us consider (16). This datum shows that English has a different surface order than what is given in (15), namely, ‘DEM > Num > AP > ADJ > N + Pl’, and under his analysis, this surface order is derived as in (17). That is, once the basic structure is formed in accordance with (15), the SortP undergoes a roll-up movement for cluster formation purposes; more specifically, it raises to PlP+ (i.e., an expansion of PlP; a notational convention he borrows from Koopman and Szabolcsi 2000). Consequently, the plural suffix gets realized on the head N even though it originates higher than SortP. (16) English datum instantiating ‘DEM > Num > AP > ADJ > N + Pl’: these three very lovely Chinese vases

In essentially the same spirit, Svenonius (2008) also states that “UG provides the basic ingredients for the category Adjective, and all languages appear to avail themselves of it in one form or another” but it is also possible that a language has its own innovative ways to modify each layer of the DP, either with specialty inventories of modificational elements for each layer or by virtue of having a large class of ADJs that can occur in multiple functional projections, as is the case in English (p. 41-42). 5

106

4 A New Formal Analysis of AOR

(17) Derivation of (16) under Svenonius’ analysis: ArtP Dem these

UnitP NumP

PlP+

three

PlP Pl -s

SortP AP

very lovely Sort

Mov’t to form a cluster with the Pl marker Sort nP Chinese vase

Consider now the Norwegian datum in (18), which instantiates the ‘DEM > Num > ADJ > N + Pl + Art’ order. According to Svenonius, this surface order obtains because in Norwegian, DEM is projected above ArtP (most likely as a head) and the plural marker and the definite article form a cluster with the head N, as depicted in (19). (18) Norwegian datum instantiating ‘DEM > Num > ADJ > N + Pl + Art’: disse tre berømte bøk-e-ne these three famous book- PL-DEF ‘these three famous books’ [Svenonius 2008: ex. (23)]

4.2 Lessons from Svenonius (2008) and Laenzlinger (2005)

107

(19) Derivation of (18) under Svenonius’ analysis: DemP

Dem

ArtP

disse ‘these’ UnitP

Art

tUnitP

-ne NumP

tre ‘three’

PlP

nP

DEF

Pl

tnP

-e AdjP

berømte ‘famous’

NP

PL

bøk‘book’ [Svenonius 2008: (24)]

These two sample derivations show (sufficiently in my view) that applying roll-up movements to a well-motivated decompositional DP structure can yield positive results in capturing DP-internal constituent order variation across languages. I will therefore take a similar approach in developing my own analysis of N modifiers below. I will also borrow the following ideas from Svenonius: First, even though he does not take a stand on the exact syntactic position of RCs, he implicitly suggests that there is no a priori reason to treat RCs differently than “true” APs in the sense of Cinque (2010). In fact, some of the tree diagrams he offers for languages like Burmese, Vietnamese, and Thai contain RCs, and in these diagrams, RCs merge at [Spec, SortP], just like typical adnominal APs; see his (28)–(30), for example. Secondly, he states that RCs may be attached at different levels under different circumstances (see his Footnote 10), and by making such remarks, he suggests that clausal N modifiers have a more versatile syntax than what is assumed in the literature (e.g., Cinque 2010). Thirdly, Svenonius points out that focus may salvage an otherwise illicit ADJ order, as shown (20) (see also Sproat and Shih 1988 for a similar observation), and to account for such data, he assumes that a DP contains a FocusP which is projected

108

4 A New Formal Analysis of AOR

above Kind Phrase (KiP) in the sense of Zamparelli (2000)6 and focused ADJs move to [Spec, FocP], giving rise to a non-canonical surface order, as depicted in (21). (20) Focus-induced non-canonical ADJ ordering in English: a. big square table; *square big table; SQUARE big table b. expensive wooden table; *wooden expensive table; WOODEN expensive table c. tasty French cheese; *French tasty cheese; FRENCH tasty cheese. [Svenonius 2008: ex. (35)] (21) English DP structure with a focused ADJ under Svenonius’ analysis: 7 FocP AP2 SQUARE

Foc Foc

KiP SortP AP1 big

Sort Sort

nP tAP2

n n

P table

7

To summarize this subsection, then, Svenonius’ decompositional DP analysis provides an important guiding principle for developing a formal analysis of N modifiers since, evidently, any successful attempt at it must yield outcomes that are compatible with what is given in (15). Furthermore, the formal framework he assumes is flexible enough to capture the relevant cross-linguistic variation. That said, his analysis leaves open the question of where exactly discourse-linked but unfocused N modifiers (e.g., non-restrictive and referent-modifying ones) may occur inside the DP structure. This question finds an answer in Laenzlinger 2005, to which we now turn.

6

For similar assumptions, see also Guttièrez-Rexach and Mallen 2002, Truswell 2004, and Demonte 2005. 7 Note that this derivation is adapted from Svenonius’ (2008) diagram (36). In his original diagram, FocP is absent and the focused ADJ moves to [Spec, KiP], unlike what he states in the text (see his p. 35). Hence, in order to make his diagram match his explanation, I took the liberty to make the appropriate amendments, that is, adding a FocP to (21) and representing the focused AP as having moved to the Spec position of FocP.

4.2 Lessons from Svenonius (2008) and Laenzlinger (2005)

4.2.2

109

Laenzlinger’s (2005) Complex-DP Analysis

Laenzlinger’s approach to AOR is, in spirit, not so different from Scott’s (2002) or Cinque’s (2010) since it also assumes that adnominal ADJs merge at the Spec position of relevant FPs inside DP which form a strictly hierarchical structure among them. Furthermore, he claims that differences between Romance and Germanic languages with regard to ADJ ordering boil down to the parametric variation of the obligatory NP-raising in Romance versus no NP-raising in Germanic.8 To illustrate this, under his analysis, the surface order exemplified by the French datum in (22a) is derived as in (23). This diagram shows that at the initial stage of the derivation, an NP and two FPs that introduce an AP (which he labels as FPadj) are projected inside the DP, and after this, the NP moves to the Spec of FPAgr(NP), an FP devoted to checking the agreement between an N and an ADJ that is located right above the lowest FPadj. This is followed by the FPAgr(NP) undergoing a ‘snowballing movement’9 to the Spec of the next highest FPAgr(NP), and this yields the ‘N > AP2 > AP1’ order, which is the mirror image of a Germanic language; in English or German, ‘AP1 > AP2 > NP’ obtains instead. (22) French data instantiating the ‘N > ADJ > ADJ’ surface order: italienne magnifique a. une voiture a car Italian beautiful ‘a beautiful Italian car’ b. une table ronde magnifique a table round beautiful ‘a beautiful round table’ (23) [DP [FPAgr(NP) [FPAgr(NP)

[FP magnifique [FPAgr(NP) [FP italienne [NP voiture]]]]]]]

Even though Laenzlinger’s analysis is among the descendants of Cinque 1994, it stands out for several reasons. First of all, it simplifies Scott’s (2002) ADJ hierarchy given in (8) by dividing ADJs into five ‘semantic meta-classes’, as shown in (24).

8

This idea was actually adopted by Cinque (2010). In fact, Laenzlinger’s (2005) analysis is among the many reasons why Cinque (2010) revises the N-raising analysis proposed in Cinque 1994 to a phrasal movement analysis. 9 What is called ‘snowballing movement’ is comparable to Aboh’s (2004) ‘roll-up’ movement.

110

4 A New Formal Analysis of AOR

(24) Laenzlinger’s re-organization of Scott’s (2002) ADJ hierarchy: [ QUANTIF Ordinal > Cardinal] > [ SPEAKER-ORIENT Subjective Comment > Evidential] > [ SCALAR PHYSICAL PROPERTY Size > Length > Height > Speed > Depth > Width] > [ MEASURE Temperature > Wetness > Age] > [ NON-SCALAR PHYSICAL PROPERTY Shape > Color > Nationality/Origin > Material] [Laenzlinger 2005: (14)]

Secondly, Laenzlinger claims that the syntactic hierarchy of ADJs is the reflex of their semantic hierarchy (p. 655) and in doing so, he (implicitly) emphasizes the role that semantics plays in deriving various surface ADJ orderings; authors working in the Cinquerian tradition typically invoke movement for largely morpho-syntactic or morpho-phonological reasons than for semantic reasons (e.g., Brugè 1996, 2002; Giusti 2002, 2005; Cinque 2010). Next, to capture the AOR phenomena in Romance languages, he adopts what is known as the ‘Split-DP’ hypothesis,10 namely, the idea that the topmost FP that makes up a nominal projection is not a simple DP but a Rizzian (Rizzi 1997) CP-like projection. To be more specific, following authors like Ihsane (2000), Ihsane and Puskas (2001), Kariaeva (2004), and Aboh (2004), he assumes that what is traditionally referred to as a DP is comprised of DPdeixis (= external) and DPdetermination (= internal), as given in (25),11 and while the internal layer expresses [±definiteness] and/or partitivity, the external layer expresses relevant pragmatic meanings such as referentiality and deixis. (25) Split-DP structure proposed by Laenzlinger (2005): [DPdeixis (= external) [Ddeixis [DPdetermination (= internal) [D determination

[ [NP]]]]]]

Finally, Laenzlinger proposes that, in Romance languages,12 an FP that is devoted to checking gender and number agreement between an N and a D (i.e., FPAgr(NP)) is located above DPinternal, and this FP attracts the NP for agreement feature checking. But notably, such an additional NP movement would result in a ‘Dexternal > N > Dinternal’ configuration which is not attested in French, so when the agreement features are checked off between the NP and the D in the middle field of his split-DP structure, the head of DPinternal raises to the Dexternal position, thereby yielding a ‘D > N’ order, as schematically represented in (26).

10 For works that endorse what is known as the Split-DP Hypothesis, see, a.o., Kayne 1994, Starke 1995, Bosque and Picallo 1996, Guiterrez-Rexach and Mallen 2001, Vangsnes 2004, and Giusti 2005. 11 According to Laenzlinger (2005), DPdeixis (= external) is analogous to Rizzi’s (1997) ForceP (which determines a clause’s type) and DPdetermination (= internal) is analogous to Rizzi’s FinP (which presumably determines a clause’s [±finiteness]). 12 Laenzlinger does not specify to what extent his claim may extend beyond French, but it is contextually implied that it will carry over to other Romance languages such as Spanish and Italian, and I have taken this implication at face value here.

4.2 Lessons from Svenonius (2008) and Laenzlinger (2005)

111

(26) Split-DP structure and D raising following NP raising to DPexternal layer: [DPdeixis (= external) [Dexternal [FPAgr(NP) [DPdetermination (= internal) [Dinternal [[NP ] ]]]]]]

As for the syntax of prenominal modifiers (whose prenominal position is non-canonical in Romance), Laenzlinger first points out that French ADJs may occur prenominally only if they are in weak forms, or they carry an emphatic/strong subjective interpretation or function as quantifiers, as exemplified in (27). He then posits that DPexternal contains three FPs, namely, QuantP, SubjP, and WeakP, and these FPs respectively attract a weak ADJ, a subjective ADJ, and a quantificational ADJ from below. Under this idea, then, nombreuses in (27) occurs at [Spec, QuantP], superbes at [Spec, SubjP], and petites at [Spec, WeakP]. (27) ces nombreuses superbes petites these numerous wonderful small ‘these numerous wonderful small red cars’

voitures cars

rouges red

Laenzlinger further observes that French permits maximally three postnominal ADJs but only if the rightmost one is focused, as schematized in (28) and as exemplified in (29). To capture this constrained behavior of French ADJs, he suggests that in the (French) DP structure, focused ADJs are base-generated at the Spec of FPFoc, which is the highest FPadj below Dinternal, but any unfocused ADJs raise to an FPAgr(NP) that is higher than FPFoc, and consequently, a focused ADJ ends up occurring in the rightmost position inside the DP.13 (28) Constraint on postnominal ADJ ordering in French: [N < Adj < Adj < *Adj/ ADJ] [Laenzlinger 2005: (35)] (29) a. *une voiture rouge italienne magnifique a car red Italian beautiful Intended: ‘a beautiful red Italian car’ b. ?une voiture rouge italienne MAGNIFIQUE a car red Italian BEAUTIFUL Intended: ‘a BEAUTIFUL red Italian car’ [Laenzlinger 2005: (50)-(51)]

13

To ensure that the number of unfocused postnominal ADJs in French does not exceed two, Laenzlinger adopts Pearson’s (2000: 339) Feature Inheritance Hypothesis, which essentially dictates that, in French, an FPadj must inherit a lexical feature from a lexical N via Spec-Head agreement but once this is done, the NP need not raise any further, so unless a snowballing movement of an extended NP occurs, an N may contain maximally two unfocused postnominal ADJs within its maximal projection.

112

4 A New Formal Analysis of AOR

Finally, based on the observation that adnominal PPs in French tend to occur following an unstressed postnominal ADJ but preceding a focused ADJ, as shown in (30), Laenzlinger suggests that (French) adnominal PPs are base-generated in the Spec of an FP that is located right above DPinternal and below the highest FPAgr(NP), yet due to the movements depicted in (31), ‘D > Adj > Adj > Adj > N > Adj > PP > FOC-Adj’ obtains. (30) les the [PP [

nombreuses superbes petites [ NP voitures] rouges numerous wonderful small [ cars] red de 1958] ITALIENNES from 1958] Italian numerous wonderful small red ITALIAN cars from 1958

(31) Derivation of (30) under Laenzlinger’s analysis: DPexternal QuantP SubjP WeakP FPAgr(NP) FPpp PP de 1958 D

DPinternal FPAgr(NP) FPFoc ITALIENNES

FPadj

petites splendides nombreuses

FPAgr(NP) FPadj rouges

NP voitures

4.2 Lessons from Svenonius (2008) and Laenzlinger (2005)

113

Comparing (31) with (21) leads us to see that Svenonius’ and Laenzlinger’s analyses are similar with regards to their treatment of focused ADJs, but by splitting the DP structure into two layers, namely, DPinternal and DPexternal, Laenzlinger offers a more straightforward way to incorporate Larson’s (1998, 2000) insights on the correlation between the position of N modifiers and their interpretive possibilities within a Cinquerian framework. Therefore, given the conclusion we arrived at in Sect. 4.1, it seems worth applying his analysis to the Korean data on AOR. That said, Laenzlinger is agnostic about where exactly FPs introducing NUM-marking, DEMs, and RCs are located relative to each other inside a DP, not to mention where they are located relative to FPs introducing adnominal APs and PPs. He also (tacitly) treats DEMs on a par with articles—more specifically, as heads that are base-generated under what he calls Dinternal [see his trees (43), (55), and (74)], but such treatments will not work for all languages, though they may for French, since DEMs are typically phrasal species whereas articles are (almost) always heads. In light of these considerations, it is concluded that the best way to make use of Laenzlinger’s analysis for our purposes is to amend it, incorporating some of the lessons we have learned from reviewing Svenonius’ analysis. And that is exactly what we will do in subsequent pages.

4.3

The New Analysis

Building on the insights of authors like Larson (1998, 2000) and Laenzlinger (2005), I posit that the space inside a DP may fall into two sub-spaces and while lexical meanings are composed in the lower space, functional meanings are composed in the higher space. However, I classify N modifiers into three kinds and posit that they have different merge sites and for this reason, DP-internal space may be broken into three parts when approached from the standpoint of N modifiers. I also suggest that there are at least three types of nominal meaning and they are constructed in different regions inside the DP. Below, I elaborate on these ideas one by one though not necessarily in this order.

4.3.1

Three-Way Classification of Nominals

First of all, I classify nominals into three sub-types, namely, (i) predicative, (ii) quantificational, and (iii) referential. By ‘predicative’, I mean syntactic constituents that are analyzed as of type or linguistic expressions which denote

114

4 A New Formal Analysis of AOR

characteristic functions of sets of individuals. By ‘quantificational’, I mean constituents of type or expressions which denote functions from sets of individuals to truth-values. And by ‘referential’, I mean constituents of type or expressions which denote individuals. These three types of nominals are exemplified by the bracketed items in (32)–(34).14 (32) Predicative nominals: of type a. John is [a doctor]. b. John and Mary are [doctors]. (33) Quantificational nominals: of type a. [Every doctor (from that hospital)] came to the party. b. [Most doctors (who work at the hospital)] came to the party. (34) Referential nominals: of type a. [The (tall) doctor (who works at the ER)] came to the party. b. [A (well-known) doctor (who works at the ER)] came to the party.

Of these three types of nominals, I suggest that predicative ones have the least layered syntactic structure. One reason to think so is that, unlike quantificational or referential nominals, predicative ones cannot contain a numeral modifier, a quantificational DET, or a referential DEM, not only in English but also in Korean. To see this, consider (35)–(37).

14

In addition to these three sub-types, another nominal category can be identified which is comprised of nominals that occur as pivots in there-BE-existential sentences, as illustrated in (i). I do not include such nominals in our discussion here because they have been claimed to be neither predicative nor quantificational (see Hartmann 2008) and possibly for this reason, their exact categorial status and grammatical function have been a topic of intense debate in the literature and therefore trying to do justice to them will take us too far afield (see, a.o., Williams 1984; McNally 1992/1997, 1998; Moro 1997; Felser and Rupp 2001; see also Francez 2007, 2009; Hartmann 2008; and the references there). That said, I think of such nominals as falling somewhere between predicative and quantificational categories though a fuller exploration of this idea has to be left for future research. (i) a. There was [every kind of doctor you can imagine] at the party. b. There were [all sorts of doctors] at the party. c. There were [several doctors] at the party.

4.3 The New Analysis

115

(35) Predicative nominals: a. John and Mary are (*two/*a few/*those) doctors. b. Korean: [Con-kwa Meyri]-nun *twu-myeng-uy/*myech-myeng-uy/*ku [John-and Mary]-TOP two-CL-GEN/a.few-CL-GEN/that uysa(-tul)-i- -ta. doctor(-PL)-COP-N.PST-DECL Intended: ‘John and Mary are two/a few/those doctors.’ (36) Quantificational nominals: a. Every teacher met two students. b. Korean: Motun sensayng-nim(-tul)-i twu-myeng-uy Every teacher- HON(-PL)-NOM two-CL-GEN manna-ss-ta. meet- PST-DECL ‘Every teacher met two students.’

haksayng(-tul)-ul student(-PL)-ACC

(37) Referential nominals: a. They need to rely on those {two/few} examples that John came up with. b. These are my {two/few} contributions to the project. Korean: c. [Con-i ei sayngkakhaynay- ]-n ku myech-kaci-uy think.of-ANT]-REL that few-kind-GEN [John-NOM __ yemwun-tul example-PL ‘those few examples that John came up with’ d. na-uy {twu-/myech}-myeng-uy celchin-tul {two-/a.few}-CL-GEN close.friend-PL I- GEN ‘two/a few close friends of mine’ (Lit.: ‘my two/few close friends’)

Another reason to postulate a smaller structure for predicative nominals than for referential ones is that they cannot be referred to by using a definite description. To illustrate this, in each discourse in (38)–(39), the first sentence cannot be followed by the second sentence if it begins with a definite description that is co-indexed with the subject complement of the preceding sentence, and this shows that the internal structure of a predicative nominal lacks space for a DET or a DEM, unlike the case with referential nominals.15

15

One may think that the badness of some of the continuations in (38)–(39) is due to the fact that predicative nominals are inherently non-referential, as established by authors like Williams (1983) and Partee (1986). But I take the ungrammaticality of these data to be indicative of a structural difference between predicative nominals and referential ones—a claim which I believe has not been made in the existing literature.

116

4 A New Formal Analysis of AOR

(38) a. Johni is a doctor. *The doctori/ He i is from Dallas. b. [John and Mary] i are doctors. *The doctorsi/ Theyi are married to each other. (39) Korean: uysa-i- -ta. *Ku uysai-nun/ Kui-nun a. Coni -un That doctor-TOP/He-TOP John-TOP doctor-COP-N.PST-DECL tallasu-eyse o-ass-ta. Dallas-from come-PRF-DECL Intended: ‘Johni is a doctor. *The doctori/ Hei is from Dallas.’ uysa-tul-i- -ta. b. [Con-kwa Meyri]i-nun doctor-PL-COP-N.PST-DECL [John-and Mary]-TOP *Ku uysa-tul i-un/ Ku-tuli -un pwupwu-i- -ta. That doctor-PL-TOP/That- PL-TOP couple-COP-N.PST-DECL Intended: ‘[John and Mary]i are doctors. *The doctorsi/theyi are married to each other.’

Taking the above contrasts into consideration, I submit that predicative meanings are composed in a DP-internal region that is lower than where quantificational or referential meanings are composed, as roughly given in (40). (40) Partition inside a full-fledged nominal projection (first pass): [DP Referential/Quantificational meaning [ Predicative meaning]]

On the basis of (40), one may think that, under the present analysis, referential nominals and quantificational ones are treated on a par with each other. But in the next subsection, I argue that they have different internal structures and crucially for our purposes, the basis for this idea will be coming from the way N modifiers are distributed inside a DP.

4.3.2

Three Types of N Modifiers and Three Merge Sites Inside DP

In classifying N modifiers, I draw on Huddleston and Pullum’s (2005) analysis of RCs but I also make a few departures from them. First of all, Huddleston and Pullum classify RCs into two kinds, namely, integrated and supplementary, and they state that integrated RCs express “information that is presented as an integral part of the larger message” whereas supplementary

4.3 The New Analysis

117

RCs convey information that is construed as “supplementary to that is expressed in the rest of the sentence: it is additional, often parenthetical, material” (p. 187).16 I adopt their classification of RCs but taking one step further, I assume that any type of N modifier can be classified either as integrated or as supplementary, depending on the type of information they are construed as contributing. That is, if the content of an N modifier is construed as forming an integral part of the larger message conveyed, then it exemplifies an integrated N modifier; if its content is construed as supplying something supplementary, additional, or parenthetical, then it exemplifies a supplementary N modifier. In the case of English, the semantic differences between the two types of N modifiers are formally indicated, so it is relatively easy to differentiate one type from the other: integrated N modifiers are treated as part of the intonation unit of the nominal they modify, so they occur without a pause between them and their head N. By contrast, supplementary modifiers are treated as a separate intonation unit, so they are marked off by pauses, which are notated as commas, dashes, or parentheses in writing. To see this, compare (41) and (42). The (a) examples here are taken from Huddleston and Pullum 187: (10). (41)

Integrated N modifiers: a. Politicians [who make extravagant promises] aren’t trusted. b. The doctor [who likes Let It Be] works at the University Medical Center. c. The doctor [with no name tag on] treated Mary. d. John fell in love with a [really tall] woman.

(42)

Non-integrated N modifiers: a. Politicians, [who make extravagant promises], aren’t trusted. b. John, [who likes Let It Be], doesn’t really like the Beatles. c. The doctor, [with no name tag on], got into serious trouble. d. John fell in love with Mary, [a very tall blonde].

Incorporating such formal differences between the two types of N modifiers, I propose that integrated modifiers are base-generated DP-internally by merging at some [Spec, FP], as assumed in much recent work (e.g., Cinque 1994, 2010; Laenzlinger 2005; Svenonius 2008), whereas supplementary modifiers are base-generated DP-externally by merging at the Spec position of what I call ‘Supplementary Phrase’ (SpplP for short). I further posit that in all languages that have supplementary N modifiers, the head of SpplP hosts a formal feature that is pronounced as a pause, which is indicated as a comma in writing (cf. Potts 2003/2005), but depending on the

Alternatively, it can be said that integrated modifiers contribute to the ‘at issue’ meaning of the utterance in the sense of Potts (2003/2005), whereas supplementary ones contribute to the ‘non-at-issue’ meaning or they provide background information on what is being expressed by the sentence at hand. 16

118

4 A New Formal Analysis of AOR

language, Sppl may select for a TP or a DP,17 and this has something to do with the surface position of the N modifiers in that language. For example, if the language has only postnominal N modifiers, then Sppl selects for a TP; if the language has only prenominal N modifiers, then it selects for a DP.18 These ideas are schematically represented in (43), where XP represents a supplementary N modifier and YP represents an integrated one.

17

By positing such syntax for sentences containing supplementary RCs, I am in a way lending support to the widely-held view that non-restrictive RCs which occur accompanied by commas are asymmetrically conjoined to sentences (e.g., Demirdache 1991). But unlike the standard approach, I claim that what Sppl selects for may vary from language to language. Furthermore, I assume that in English-type languages, Sppl selects for a TP, rather than a CP (compare Demirdache 1991). I make this assumption because, as far as I can see, there is no empirical evidence that the sentential structure that embeds a supplementary RC in English is a CP as opposed to a TP. 18 I offer a more streamlined analysis of supplementary RCs in Chap. 6, especially for Korean-type languages. But even for now, the current analysis lets us capture the relevant data in ways that the existing analyses of supplementary RCs cannot. For example, as will be demonstrated below as well as in Chap. 5, the idea that Sppl selects for a TP in postnominal modifier languages but a DP in prenominal modifier languages will let us explain why supplementary RCs in English may have non-DP antecedents (e.g., a predicative nominal, an AP, and a CP), as shown in (i), but their Korean counterparts may only have full-fledged DP antecedents, as shown in (ii). In (ii), all examples except for (a) instantiate a coordinate construction, and this is because, in Korean, the meanings intended by (ib, c, d) cannot be conveyed by using an RC construction (which in turn has something to do with the surface position of RCs in that language). (i) a. b. c. d.

[John]i, [who]i is a doctor, is Bill’s friend. John is [a doctor]i, whichi Bill is not. John is [handsome]i, whichi Bill is not. John said [that he was from Dallas]i, whichi his own brother denied later.

(ii) Korean: RC construction: a. Uysa-i-n Con-un Pil-uy chinkwu-i-ta. John-TOP Bill-GEN friend-COP-DECL Doctor-COP-REL ‘John, who is a doctor, is Bill’s friend.’ Coordinate structures: b. Con-un uysa-i-ntey Pil-un uysa-ka ani-ta. doctor-COP-but Bill-TOP doctor-TOP not.be-DECL John-TOP ‘John is a doctor but Bill is not.’ calsayngkiess-nuntey Pil-un kurehci anh-ta. c. Con-un handsome-but Bill-TOP so not.be-DECL John-TOP ‘John is handsome but Bill is not.’ cakii-ka Tallas-eyse o-ass-ta-ko d. Coni-un self-NOM Dallas-from come-PRF-IND-COMP John-TOP kukes-ul pwuinhay-ss-ta. malhay-ss-nuntey kui-uy chin-hyeng-i he-GEN real-brother-NOM that-ACC deny-PST-DECL say-PST-but ‘Johni said that hei is from Dallas, but hisi own brother denied it.’

4.3 The New Analysis

119

(43) Position of supplementary N modifiers and that of integrated N modifiers: [SpplP [XP] [Sppl [Sppl , ] [TP/DP … [DP … [FP YP] … N] …]]] Supplementary Mod Integrated Mod

While Huddleston and Pullum do not further classify RCs (or other types of N modifiers), I divide integrated N modifiers into two sub-types, namely, restrictive (RS) and non-restrictive (NRS).19 Therefore, my analysis differs from theirs in yet another respect. As stated in Chap. 1, I define a restrictive N modifier as any adnominal category that reduces the size of the set denoted by the head N of a DP including those that place a restriction on the domain of quantification such as the RCs in (44); conversely, a non-restrictive N modifier is defined as any category that modifies the referent of a DP without reducing the size of the set that is denoted by its head N. Hence, under the present analysis, integrated but appositive modifiers such as those in (45) fall under the rubric of ‘non-restrictive’. Notice that each example in (45) can be felicitously uttered only when some individual which fits the description expressed by the RC has already been picked out by the speaker, and the speaker asserts that the property denoted by the VP holds true of that individual, so the RC receives a non-restrictive and appositive interpretation even though its meaning is processed not as supplementary to what is being expressed in the rest of the sentence. (44) Restrictive N modifiers: Illustration with English a. John is a [tall] man. b. John is a student [who studies physics]. c. Every student [who likes Let it Be] was there. d. Every [sincere] person [who attended the party] can tell you that. (45) (Integrated) non-restrictive N modifiers: Illustration with English a. John is the man [who likes Let It Be]. b. The students [from Dr. Hutchison’s class] took all the prizes. c. The student [who likes Let It Be] is John. d. That [old] book [that has been sitting there forever] is from my great-grandfather.

I should note that Huddleston and Pullum do not make a restrictive versus non-restrictive distinction for N modifiers because according to them, it is not always possible to tease them apart from each other on the basis of the form alone. By way of illustration, the RCs in (46)–(47) (taken from Huddleston and Pullum, p. 188) are formally alike (in the sense that there are no pauses around them) even though the RCs in (46) reduce the size of the head N and the one in (47) does not. (46) Martha has two sons [who are still at school] and two [who are at university].

19

Supplementary modifiers are always non-restrictive, so this distinction does not apply to them.

120

4 A New Formal Analysis of AOR

(47) Martha has two sons [she can rely on] and hence is not unduly worried.

But just because differentiating between restrictive RCs and non-restrictive RCs is sometimes difficult does not mean that the distinction between them does not exist. I should also point out that such difficulties seem to arise only when the RC at hand occurs inside a nominal projection that is large enough to occupy an argument position. That is, whenever an RC occurs inside a predicative nominal, it only receives a restrictive interpretation (as is the case with (44b), for example). Therefore, an N modifier’s ability to carry non-restrictive semantics may in fact serve as a useful diagnostic with which to determine the syntactic size of the nominal at hand, that is, how full-fledged it is. Furthermore, the fact that restrictive N modifiers can occur inside any type of nominal projection including predicative ones suggests that they are located lower than non-restrictive N modifiers, contrary to what has been claimed in the literature (e.g., Cinque 201020). If what I am claiming here is correct, then, making the restrictive versus non-restrictive distinction for RCs or any type of N modifiers is both empirically and theoretically called for, rather than “misleading”, unlike what Huddleston and Pullum claim. Given this, we can state that all N modifiers can be classified as either supplementary or integrated at the macro-level, but since integrated ones are either restrictive or non-restrictive, RCs may fall into three sub-types, namely, (i) supplementary, (ii) (integrated) restrictive, and (iii) (integrated) non-restrictive. And given the assumptions made here, these three types of N modifiers merge in different locations: supplementary modifiers merge DP-externally whereas integrated N modifiers merge DP-internally, and among integrated N modifiers, restrictive ones merge lower than non-restrictive ones, as depicted in (48). (48) Three-way classification of N modifiers and their positions inside DP: Supplementary Mod Integrated Mod NRS Mod DP-EXTERNAL

RS Mod

DP-INTERNAL

The way I classify and analyze N modifiers here gives us an important motivation to postulate a middle field inside DP in which quantificational meanings are constructed. And on this basis, I suggest that DP-internal space falls into three sub-spaces, namely, what I call ‘High Field’, ‘Middle Field’, and ‘Low Field’, and each of them is concerned with constructing referential, quantificational, and predicative nominal meanings, respectively. Such partitioning of DP space is

20 Cinque (2010: 24) claims that non-restrictive N modifiers occur closer to the head N than restrictive ones do, which is actually the exact opposite of what is being claimed here. See Sect. 4.1.

4.3 The New Analysis

121

deemed necessary because although restrictive modifiers may modify any type of non-referential nominals, be they quantificational or predicative, quantificational nominals cannot be treated on a par with predicative nominals because the latter type of nominals do not tolerate numeral ADJs, as we saw in (35). In light of the proposed way to sub-divide DP-internal space, then, we arrive at something like (49) as an elaboration of what is given in (40). (49) DP-internal structure and three fields inside of it: [DP Referential meaning [ Quantificational meaning [ Predicative meaning]]] High Field Middle Field Low Field NRS Mod RS Mod RS Mod

Needless to say, the above picture begs several questions including what kinds of FPs may comprise each field inside a DP. But before proceeding to address such questions, I would like to briefly demonstrate how the analysis we have developed thus far may let us capture the relevant data. And to that end, I would like to outline a few additional assumptions we will be making in the rest of this book. First of all, I adopt a Kaynean (Kayne 1994) anti-symmetry approach to phrase structure and therefore will assume that all movements are leftward. I also assume that there is just one Spec position for each FP. But in order to accommodate the fact that multiple RCs may co-occur stacked on top of each other and even ADJs may occur in an unordered manner as originally observed by Dixon (1977), and as illustrated by the goodness of both (50a) and (50b) below, I hypothesize that FPs of the same kind can be created in the left periphery of a DP structure as needed, forming an adjunction structure, provided that the outcome is compatible with compositional semantics and it does not violate certain universal or languagespecific AORs (cf. Cinque 201021). (50) a. brave clever man b. clever brave man

In addition, I assume that an intonational break or a pause inside a nominal projection has a syntactic (as well as semantic) import. More concretely, I take a pause to indicate the presence of an adjunction structure in the surrounding syntactic environment.

21 Cinque (2010) also allows for an adjunction structure inside DP though he does not explicitly say so; for example, in diagram (24) on p. 34 of his book, the FPs that introduce what he calls ‘direct modifiers’ are not formally differentiable, so they can be analyzed as forming an adjunction structure. But he does not discuss under what circumstances an adjunction structure can be created or what consequences it may have for a mainstream syntactic analysis that assumes a Kaynean framework. I discuss these matters below, as well as in Chap. 5, by resorting to some concrete data that are best analyzed as containing an adjoined configuration.

122

4 A New Formal Analysis of AOR

In light of these assumptions we are making, then, under the present analysis, the ‘head N + RC’ strings in (41a), (42a), and (42d) will have the syntactic derivations sketched in (51), (52a), and (52b), respectively. (51) Derivation of the integrated RC in (41a) under the present analysis: [FP [FP [who make extravagant promises] [NP politicians]]]

(52) a. Derivation of the supplementary subject RC in (42a) under the present analysis: [SpplP+ [SpplP [who make extravagant promises] [Sppl ,] [TP [DP politicians] [T …]]]]

b. Derivation of the supplementary object RC in (42d) under the present analysis: [SpplP+ [SpplP [who was a very tall blonde] [Sppl ,] [TP John fell in love with Mary]]]

To briefly explain how to read these diagrams: (51) shows that in English, the head N of an integrated RC moves across the RC, landing in a Spec position that is created right above it, thereby giving rise to an ‘N > RC’ surface order. On the other hand, (52a) shows that if the head N of an English supplementary RC is a sentential subject, then once a Sppl selects for a TP, the maximal projection of the head N, which contains the supplementary RC, moves to an adjunct position that is created right above the SpplP, which I call the Spec of SpplP+ (i.e., an extension of SpplP), borrowing the notational device from Svenonius (2008) and Koopman and Szabolcsi (2000). Consequently, the RC surfaces postnominally, both preceded and followed by a pause: the first pause is due to the adjunction structure created by the head N’s movement and the second pause is due to the comma that is hosted by Sppl. Finally, (52b) shows that if the head N of a supplementary RC is a sentential object, then the entire TP that Sppl selects for moves to [Spec, SpplP+], presumably triggered by some language-specific need to create a postnominal RC structure. This movement creates an adjunction structure which is indicated by a pause (i.e., a comma) that follows the TP; the head of the SpplP also hosts a pause but since the RC occurs sentence-finally, it is conflated with a period, as one can see in (42d).

4.3.3

The Full DP Internal Structure and a New Decompositional Split-DP Analysis

In fleshing out the entire DP internal structure, I take a Rizzian (Rizzi 1997) cartographic approach in a manner similar to Laenzlinger (2005). But I combine it with Svenonius’ (2008) line of analysis for the reasons given in Sect. 4.1 and 4.2. And this means that I will also assume that:

4.3 The New Analysis

123

(i)

what derives different surface ADJ orders across languages is some type of leftward phrasal movement inside DP, which Laenzlinger calls ‘snow-balling’ movement and which Svenonius calls ‘roll-up’ movement; (ii) a DP contains a FocP; and (iii) any focused N modifier may occur at [Spec, FocP]. The actual split-DP structure I argue for will look rather different from what has been proposed in the literature, however. And here is how and/or why: First and foremost, I split what Laenzlinger calls DPdetermination/internal into two sub-layers, namely, what I call ‘DPquant’ (DPq) and ‘DPpredicate’ (DPp). Additionally, I relabel what Laenzlinger calls DPdeixis/external as ‘DPdeixis/referential’ (DPd/r for short). I posit that Dq licenses [+quantificational] feature and Dd/r licenses not only [+deictic] and [+referential] but also [±definite] features. Unlike Laenzlinger, I do not differentiate between DPdeixis and DPdetermination because, as far as I can see, there is no empirical evidence with which to tell that they are actually distinct heads. Furthermore, any non-predicative or non-quantificational DP is bound to be [+referential] regardless of whether it is definite or indefinite, so under the tri-partite DP structure I am proposing here, there is no need to postulate a D head which licenses [±definite] features but does not license [+referential] feature. To recap, then, under the present analysis, what is standardly referred to as a nominal may be syntactically realized as a DPd/r, a DPq, or a DPp, depending on whether it is referential, quantificational, or predicative, and each of these phrases are constructed in different fields inside DP, as sketched in (53).

(53) New split-DP analysis and the three fields inside a full-fledged DP: [DPd/r [DPq [DPp ]]] [+referential;+/-deictic; +/-definite]

[+quantificational]

[+predicative]

High Field

Middle Field

Low Field

Given the above picture, one question that arises is: where would D elements such as articles, DEMs, and quantificational DETs merge if what is standardly known as a DP is indeed comprised of three sub-fields, as suggested here? In answer to this question, I submit that all N dependents originate from what I call the Low Field but some of them may later raise to what I call the Middle or the High Fields, thereby licensing the relevant D features (cf. Chomsky 1995).22 To illustrate, I assume that the lexeme every merges in the Low Field of a DP, and if it

22

The underlying assumption here is that D features are licensed when both the Goal and the Probe in the sense of Chomsky (1995) have the same sets of features and the Goal occurs in the local domain of the Probe. In the present analysis, Goals are assumed to be located below the Low Field of a DP and their corresponding Probes are located in the Middle or the High Field of a DP. Hence, they raise to a higher field at a later point in the derivation if they carry [+quantificational] or [+referential] feature.

124

4 A New Formal Analysis of AOR

stays there, it functions like an ADJ, as shown in (54a), but if it raises to [Spec, DPq], then it behaves like a quantificational DET, as shown in (54b). (54) a. You are my everything. b. Every student came to the party.

Under the present analysis, then, functional elements become what they are in overt syntax by virtue of occurring at the Spec position of the relevant FPs and thereby entering into an Agree relation with their heads.23 And to execute this idea, I propose that the low region of a DP has a similar but different internal structure than what is suggested by Svenonius. More specifically, all the FPs that he postulates are projected inside what I call the Low Field, but in the structure proposed here, a FocP is located higher than a UnitP, rather than right above a KiP (KindP), which is presumably located lower than where NUM-related functional categories are located (compare with Svenonius’ tree (21) given in Sect. 4.2.1). I also posit that a DP contains a Location Phrase (LocP), which is projected between a FocP and a UnitP, and this FP is where possessives, DEMs, and articles merge. Summarizing thus far, then, under the present analysis, the Low Field of a DP has the following internal structure: (55) Overview of the DP internal structure in the present analysis: [DPd/r [DPq [DPp [FocP [LocP [UNITP [PlP [SORTP [nP [ P N]]]]]]]]]]

I suggest that a FocP is located higher than both UNITP and PlP, unlike what Svenonius assumes, because focused ADJs can modify plural nominals, as shown in (56); given the plural morphology on the head N that is modified by a focused ADJ, it has to be the case that Pl-marking is first picked up by the head N before the focus movement of the ADJ occurs and, since Pl-marking is licensed by PlP in the formal framework adopted here, it will be logical to think that a FocP is located higher than a PlP inside the DP structure.

23

In this connection, I do not rule out the possibility that all functional elements including articles and DEMs are actually represented as bundles of features in narrow syntax with possibly complex internal structures, as has been suggested for DETs and DEMs in the recent literature. For example, numerous authors have claimed that DEMs are comprised of an adjectival component and a definite element (e.g., Dryer 1992; Delsing 1993; Chomsky 1995; Bernstein 1997; Julien 2005; Elbourne 2005; Leu 2008, 2015) and in the case of Leu (2008, 2015), he even hypothesizes that not just DEMs but also DETs are morpho-syntactically complex. For the sake of concreteness, however, in this book, I will treat functional elements as lexemes and represent them as such in the syntactic trees/diagrams. That said, in Chap. 6, I revisit the status of DEMs and compare my analysis with Leu’s (2008, 2015), which treats DEMs as extended adjectival projections which are minimally comprised of a definite marker, an agreement head, and a null locative element.

4.3 The New Analysis

125

(56) a. John bought SQUARE wooden tables. b. I need WOODEN luxurious tables.

As for the rationale for postulating a LocP inside a DP: Rijkhoff (2002) observes that across languages, the DP layer manifests a ‘Location > Quantity > Quality’ surface order, where DEMs or possessive elements take up the Location slot. Given this, if we assume that what I call DPp contains a LocP,24 which c-commands a UNITP/PlP as well as a SORTP, and that DemPs are base-generated at [Spec, LocP], then we can capture the typological finding that ‘Location > Quantity > Quality’ is the most commonly attested order although in some languages, DEMs may surface following quantity or quality-denoting terms, rather than preceding them (compare Brugè 1996, 2002) (more on this in Sect. 4.5). Regarding the position of LocP, I posit that it is located above UNITP/PlP but below FocP. The basis for this idea comes from the fact that even in morphologically impoverished languages like English, all occurrences of DEMs are NUM-marked but not all of them are focused, as shown in (57)–(60). Put another way, comparing (59) with (57), (58), and (60) leads us to see that FOC-marking on a DEM has to occur after NUM-marking on it is completed, which presumably happens in the vicinity of UNITP/PlP (in whatever way it is executed). Therefore, if a DP contains both a FocP and a LocP and if plural marking is indeed syntactically licensed, then a LocP has to be projected somewhere between a FocP and a UNITP/PlP.

24

If the DP at hand is non-referential (i.e., if it is just predicative or quantificational), then, I assume that the head of a LocP is either empty or it hosts [-referential] feature since indexical, DET-like DEMs never occur inside predicative nominals (e.g., *John and Mary are those doctors.) nor can they co-occur with a quantificational DET (e.g., *Every that student came.; *That every student came.). Similarly, given that numeral ADJs are banned inside predicative nominals, as shown in (35), I assume that what I call a DPp lacks a UNITP even though it contains a PlP and a FocP; recall that predicative nominals may have a Pl-marking and they may also contain a focused ADJ (e.g., John and Mary are doctors.; John is a COMPETENENT doctor, not an incompetent one.).

126

4 A New Formal Analysis of AOR

(57) Illustration of a definite and indexical DEM: a. Who is this stranger here? b. That is John. c. These/Those are socks. (58) Illustration of an indefinite/specific DEM: a. I went to the mall the other day, and there was this strange man talking really loudly in the shoes section. b. (Context: During a fund-raising season, a radio program host is talking to the listeners.) Please pick up that phone and call us now! (59) Illustration of a definite and contrastive-focused DEM: Which book do you want? THIS one or THAT one? Or THESE or THOSE? (60) Illustration of a definite, anaphoric, and unfocused DEM: a. In my neighborhood, there is [a really tall building]i. And thisi/*thesei building has all sorts of interesting and cool stores inside. b. Once upon a time, there lived [a lovely little princess]i. Thati/*thosei little princess was mostly happy but needed a friend.

To briefly demonstrate how the proposed analysis may let us capture some of the relevant data involving DEMs, consider what is given (61): this diagram is meant to represent, albeit roughly, the derivation of the DEM-containing nominal in (57a), namely, this stranger. Here, the DemP containing this merges at [Spec, LocP], and this is followed by the entire LocP raising to [Spec, DPd/r], which licenses the [+deictic; +definite; +referential; +specific] features for the DP (more on this in Sect. 4.4). (61) Derivation of a deictic DP in the present analysis: [DPd/r [DPq [DPp [FocP [LocP [DemP this ] [UNITP NUM [PlP PL [SORTP [nP [ P [ ]]]]]]]]]]]

Notably, positing a LocP inside a DP in the way we do here allows us to give a more nuanced analysis of DEMs than Svenonius does. And in so doing, it also lets us improve on Laenzlinger’s analysis: As mentioned in Sect. 4.2.1, in Svenonius’ structure, DEMs are base-generated above articles by either occurring at the Spec of a (referential) DP or projecting their own phrase above a (referential) DP (see (19) above, for example). But such a treatment of DEMs is afflicted by the fact that DEMs can be focused as shown in (59). More specifically, the problem is that if DEMs merge above articles, as Svenonius suggests, then there has to be a FocP located above where English articles are base-generated, but in his structure given in (21) above, a FocP is located below where articles are base-generated. Hence, given the presence of data like (59) as well as data like (62) (where what is focused is an ADJ and it is pronounced below where English indefinite article a is pronounced), his line of analysis has to posit another FocP inside DP, namely, above

4.3 The New Analysis

127

where DemPs are base-generated, in addition to keeping one below where articles are base-generated. By contrast, if we assume that DEMs merge low as I suggest here, then there is no need to postulate two separate FocPs inside a DP since in our structure, any type of focused N dependent may stop by at [Spec, FocP] on its way up to a higher position, as sketched in (63) for the derivation of the focused that in (59) (more on the number of focus positions inside DP in Chap. 6). (62) a. John is a HAPPY person (not an unhappy one). b. They are MEDICAL doctors, not doctors of PHILOSOPHY. (63) Derivation of a DP containing a focused DEM in the present analysis: [ LocP that [UNITP NUM [PlP [ SORTP [nP [ P [ ]]]]]]]]]]] [DPd/r [DPq [DPp [FocP

Svenonius’ or Laenzlinger’s lines of analyses also have trouble accounting for the fact that some DEMs are not only indefinite but also non-referential, as observed by Doran and Ward (2015) (D&W) and as illustrated by (64). (64) a. Kagan joked. “I’m this Jewish girl from New York City, and this is really not what we did on the weekends.” [taken from D&W (1c)] b. Griselda is the wife of one of Kevin’s law school classmates and she is awesome. She is this amazing little spitfire, full of so much life and energy. [adapted from D&W (2a)]

According to Doran and Ward, nominals containing proximal DEM this such as those in (64a, b) are “not used by the speaker to refer to any particular entities; rather, they are being used to predicate a property of the referent” (p. 61).25 From what I have found, this intuition is shared by other native English speakers. Given this, we cannot assume that all nominals containing a DEM modifier have an internal structure where the DEM surfaces either under a [+referential] D or in a position that is an extension of a [+referential] D, unlike what authors like Laenzlinger and Svenonius suggest. And this provides additional support for a low-origin analysis of DEMs. Notice now that the syntactic mechanism proposed here can in fact let us derive data like (64a, b): in cases like (64a, b), the proximal DEM occurs inside what I call a DPp, which is the name for the FP that represents a predicative nominal. Hence, the nominal projection in which it occurs does not denote anything referential but there is still a place for it to occur in because, under our analysis, DEMs originate below DPp, though they may raise to [Spec, DPp] to pronounce that FP, as sketched 25 Doran and Ward (2015) further argue that such uses of a DEM classify the referent with respect to a discourse-new property and they also indicate that the speaker is conveying additional information about the referent beyond the category membership. For details, see their paper.

128

4 A New Formal Analysis of AOR

in (65). And this accords well with Doran and Ward’s intuition that such uses of this have something to do with predicating a property of an individual that is denoted by the sentential subject. (65) Derivation of a property-denoting DEM in (64b) under the present analysis: [DPp [FocP [LocP [DemP this ] [ PlP [SORTP [nP [ P [ ]]]]]]]]

While we are on this topic, let me also mention that in our framework, depending on the language, definite articles carrying [+referential] may be base-generated either at Loc or at [Spec, LocP], but they later raise to Dd/r or [Spec, DPd/r], and this is to pronounce the DPd/r in compliance with the Principle of Economy of Lexical Insertion in the sense of Dimitrova-Vulchanova and Giusti (1998) given in (66). (66) Principle of Economy of Lexical Insertion: A functional projection must be licensed at all levels of representation by (a) making the specifier visible or by (b) making the head visible. (Dimitrova-Vulchanova and Giusti 1998)

According to Dimitrova-Vulchanova and Giusti, a language may satisfy this principle either by satisfying (a) or (b) or by satisfying both (a) and (b), that is, by interpreting the ‘or’ in (66) exclusively or inclusively. In keeping with Dimitrova-Vulchanova and Giusti’s spirit and in light of the findings of much subsequent work (e.g., Brugè 2002; Giusti 2002; Bernstein 2001; Alexiadou et al. 2007), I assume that (66) is a universal principle. But I also posit that it does not apply to FPs whose heads host no syntactic features. Consequently, under the present analysis, such heads need not be visible (i.e., pronounced) at any level of representation, as stated in (67), which I offer as the revised version of (66) that we will adopt in this book. In addition, I suggest that, all else being equal, what I call DPd/r is pronounced by a locative element (e.g., an article or a DEM) and what I call DPq is pronounced by a numeral element (e.g., a quantificational DET), but depending on the resources a language has, what actually pronounces these FPs may vary. (67) Revised Principle of Economy of Lexical Insertion (henceforth PELI): A functional projection (FP) that has any syntactic features must be licensed at all levels of representation by: (a) making the specifier visible (i.e., pronounced) or (b) making the head visible. But an FP whose head hosts no syntactic features need not be visible at any level of representation.

4.3 The New Analysis

129

Assuming that PELI is the driving force behind N dependents occurring in what I call the High Field of a DP lets us capture the relevant cross-linguistic variation on articles and/or DEMs. To quickly illustrate this, in so-called ‘article-less’ languages like Japanese, Korean, Mandarin, and Russian, in the absence of a definite article, non-restrictive RCs or APs may pronounce the DPd/r if a DEM is not suitable to occur inside the same DPd/r for some semantic reason (more on this in Chap. 5). In languages like Spanish and Italian, typically, the definite article pronounces the DPd/r but the FP can be pronounced by a DEM as well (e.g., Brugè 1996, 2002; Giusti 2002; Bernstein 1997, 2001). And under the present analysis, such article-possessing languages behave the way they do because their DPd/r can be pronounced either by having a definite article occupy its head position or by having a definite DEM occupy its Spec position (more on this in Sect. 4.4). In languages like Jordanian Arabic, Syrian Arabic, Hungarian, Greek, and Romanian, a definite article and a definite DEM may co-occur modifying the same N, as exemplified in (68) (see Cowell 1964; Spencer 1992; Szabolcsi 1994; Horrocks and Stavrou 1987; Giusti 1993). And under the present analysis, this is because in such languages, PELI has to be satisfied conjunctively rather than disjunctively, so not only the head but also the Spec position of the DPd/r needs to be pronounced, as depicted in (69) for the Greek datum in (68c). (68) a. hada el-bate this the-house ‘this house’ b. ez a könyv this the book ‘this book’ c. afto to vivlio this the book ‘this book’ d. om-ul acesta man-the this ‘this man’

(Jordanian Arabic26)

(Hungarian; Szabolcsi 1994)

(Greek; Horrocks and Stavrou 1987)

(Romanian; Giusti 1993)

26

)

(69) Co-occurrence of a DEM and a definite article and their low origin: [DPd/r [Dd/r’ [Dd/r [DPq [DPp [FocP [LocP [afto ] [Loc’ [Loc to [UNITP [PlP vivlio ]]]]]]]]]]]

Before closing this subsection, two additional remarks are in order, which further show how the present analysis departs from Svenonius’.

26

I thank Khaleel Abusal for supplying the Jordanian Arabic datum presented here.

130

4 A New Formal Analysis of AOR

First, recall that under Svenonius’ analysis, thematic APs merge at [Spec, nP]27 whereas those receiving a sortal property or a compositional interpretation (e.g., French whole-wheat toast, i.e., ‘French-style whole-wheat toast’) merge at [Spec, SortP] (see his p. 36–37), but other than this, he does not mention any other Spec position in which an AP maypreceive a thematic role from the head N. Unlike him, I explicitly argue that [Spec, P] is where internal theta roles of de-verbal Ns are assigned. This idea is motivated by the fact that in Korean, an adjectival N may receive an internal theta role from the head N and in such cases, nothing can intervene between the adjectival N and the head N, as shown in (70)–(71). Korean data instantiating adjectival Ns bearing an internal theta role: (70) a. [DP [N kyengcey] [N kayhyek]] [[ economy] [ renovation]] ‘a renovation of the economy’ b. [DP sin-[N kyengcey] [N kayhyek]] [[ new-[economy] [ renovation]] ‘a new renovation of the economy’ c. *[DP [N kyengcey] [N sin-kayhyek]] [[ economy] [ new-renovation]] Intended: ‘a new renovation of the economy’ (71) a. [DP [N swuchwul] [N sengcang]] [[ export] [ growth]] ‘a growth in export’ b. [DP tay-[N swuchwul] [N sengcang]] [ big-[ export] [ growth]] ‘a big growth in export’ c. *[DP [N swuchwul] [N tay-sengcang]] [[ export] [ big-growth]] Intended: ‘a big growth in export’

I take the badness of (70c) and (71c) to suggest that an adjectival N forms an N compound with the head N (as I already noted in Chap. 2). In addition, I posit that APs which form an N compound with a root N stay in situ. And this is based on the fact that whenever an adjectival lexeme that forms a compound with an N occurs non-adjacent to it, the compound meaning disappears, as exemplified by (72b), in contrast with (72a).

27

We can infer this from the fact that in Svenonius’ (2008) analysis of English data like Indian wild rice and very long French shoes, in which the APs Indian and French receive the ‘origin’ or ‘source’ construal, the APs merge at [Spec, nP]. See his tree diagrams (42) and (45) in the original work.

4.3 The New Analysis

131

(72) a. (artificial) nervous system = system of nerves (which is artificial) b. nervous artificial system = system that is artificial and is nervous (Svenonius 2008, ex. (38); emphasis mine)

p On this revised view on things, then, [Spec, P] licenses not only APs forming idioms but also those that form N compounds with a head N or receive an internal theta role from it. Additionally, [Spec, nP] is reserved not only for non-gradable and intersective APs but also for thematic APs that are structurally larger than adjectival Ns. As for [Spec, SortP], it is reserved for gradable and subsective APs, as originally assumed by Svenonius. The other departure I make from Svenonius is that while he is agnostic about the exact syntactic position of RCs (though he has a relatively flexible view on them), I claim that all RCs enter a nominal structure by merging at the Spec position of the relevant DP layer. To be more specific, I posit that RCs describing sortal properties of individuals merge at [Spec, DPp];28 RCs restricting the domain of quantification for a quantifier merge at [Spec, DPq]; and non-restrictive but integrated RCs merge at [Spec, DPd/r]. I further suggest that depending on the language, RCs may even merge at an adjoined Spec position of a DPd/r, a DPq, or a DPp, which is created atop the relevant DP layer. That is, there is a parametric variation across languages with regard to whether their RCs may merge in an adjoined Spec position or not. In the tri-partite split-DP structure I am advancing here, then, integrated RCs may merge in three different kinds of Spec positions, as depicted in (73). (73) Merge sites of integrated RCs inside the proposed split-DP structure: [DPd/r RC [DPq RC [DPp RC [… [SortP [nP [ P]]]]]]] High Field Middle Field Low Field

As we will see below, one of the most notable consequences of assuming (73) will be that RCs always merge after all the non-clausal N dependents inside the same DP field have merged, and this will let us explain why RCs tend to occur in the peripheries of DP in any human language.

28

Under Svenonius’ analysis, such RCs merge at [Spec, SortP], as exemplified by his diagram (28) for Burmese data given on p. 31 of his paper, and this implies that RCs may occupy the same syntactic positions as APs, unlike what I suggest here. Whether RCs and APs can ever target the same syntactic position is an empirical question that I have to leave for future research. But having RCs base-generated in the periphery of each DP layer in a tri-partite DP structure like the way we do here will be shown to yield desirable results for several languages, as we will see below. Moreover, treating restrictive RCs on a par with restrictive APs as Svenonius does will run afoul of the fact that languages like English treat RCs differently from APs; in English, “true” APs occur prenominally but RCs do not.

132

4.3.4

4 A New Formal Analysis of AOR

Interim Summary

When we combine all the ideas presented in the preceding three subsections, we obtain (74) as the universal DP structure containing both lexical and functional N dependents that I would like to propose and defend in this book and (75) as the schema for the syntax of supplementary N modifiers including RCs. (Here and below, [+Deg] represents the semantic feature of gradability and [+h] represents the thematic property of the AP at hand.) Since thematic N modifiers typically occur in the form of Ns, the picture depicted below lets us capture the relative ordering between nominal versus verbal N modifiers in Korean (an observation we made in Chap. 3), as we will see in Chap. 5. (74) My proposal for the universal DP structure: the entire picture DPd/r

High Field Middle Field

DPq Locus of NRS Mod

DPp Low Field FocP

Loci of RS Mod

LocP DemP

UnitP NumP

PlP Pl

SortP AP([+Deg]) AP([+

nP P

])

AP([+

])

(75) Position of supplementary N modifiers in the present analysis: [SpplP [XP] [Sppl [Sppl , ] [TP/DP … [DPd/r [DPq [DPp [FocP [LocP [ …[ P N] …]]]]]]]]]

The structure given in (74) adequately reflects the ‘Location > Quantity > Quality’ surface order exhibited by the world’s languages (Rijkhoff 2002) at both the macro and

4.3 The New Analysis

133

the micro levels: inside the Low Field, DEMs (or possessive expressions) fill the Location slot,29 numeral expressions fill the Quantity slot, and descriptive or sortal N modifiers fill the Quality slot. A similar structure is also found beyond the Low Field: Location related elements such as DEMs occur in the highest region, Quantity related elements fill the middle region, and lexical elements which give a descriptive content to the nominal at hand occur in the lowest region. Furthermore, in this structure, there is a clear division between restrictive modifiers and non-restrictive ones: if a modifier restricts the domain of quantification, then it occurs in the Middle Field, but if not, it occurs in the Low Field. If it carries a non-restrictive meaning, then it occurs in the High Field either by occupying a Spec position of DPd/r or by merging at [Spec, SpplP], depending on whether its meaning is integrated into what is being expressed by the rest of the sentence or not. In morphologically rich languages (e.g., French), the actual DP-internal structure may be more elaborate than what is depicted here since such languages’ DP may contain additional FPs which license person/gender/number agreement at each relevant step in a manner analogous to what Laenzlinger’s FPAgr(NP)s do. Also, in some languages, the DPd/r region may contain FPs that host discourse prominent ADJs [i.e., those that are previously mentioned and thus are referent-modifying in the sense of Bolinger (1967)] and such FPs may exhibit a behavior that is similar to what Laenzlinger calls QuantP, SubjP, and WeakP; that is, they attract relevant D-linked N modifiers to their Spec positions from below or let them base-generated there. For now, however, what is given above should give us a sufficient idea about the new decompositional split-DP analysis. Therefore, we turn now to applying it to additional linguistic data on AOR, identifying some of its immediate welcome results along the way.

29

To keep things manageable, in this book, I remain agnostic as to the precise relative ordering between a DEM and a possessive expression (Poss) since for our purposes, it suffices to assume that both categories occupy the Location slot inside the DP structure although the prevailing view is that ‘DEM > Poss’ is the underlying order (e.g., Giusti 2002). Another reason why I do not take a stance on this matter is that at least in Korean, they rarely co-occur and if they (ever) do, the possessive element almost always occurs preceding the DEM (i.e., ‘Poss > DEM’). Moreover, there has to be a pause between them and they are always construed as standing in an appositive relation. To see this, consider (i): here, the (a) example can be felicitously uttered in a context where a subset of individuals instantiating clothes is already carved out and it has also been established that they are many in number and they belong to Mina. I should also note that in ordinary contexts, (ib) is definitely judged marginal or unacceptable but FOC-marking on the possessive element may improve on its grammaticality; for example, if it is uttered in a context where the speaker is comparing Mina’s clothes with someone else’s. (i) a.

Mina-uy*(,) ku(,) manh-un that a.lot-ADN M.-GEN b. ??ku manh-un*(,) Mina-uy Intended: ‘Those many clothes of Mina’s’

os-tul clothes-PL os-tul

(Poss > DEM) (DEM > Poss)

134

4.4

4 A New Formal Analysis of AOR

Some Immediate Welcome Results of the New Analysis

The new analysis presented in Sect. 4.3 incorporates various two-way distinctions that have been made on the semantics of N modifiers in the literature, such as Bolinger’s (1967) reference- versus referent-modification distinction, Larson’s (1998) NP-inner versus NP-outer spatial distinction, and Cinque’s (2010) or Sproat and Shih’s (1988, 1990) direct versus indirect modifier distinction. And it does so within a single formal framework which adopts largely standard assumptions about phrase structure, an achievement that is not shared by alternative analyses of N modifiers. Below, I demonstrate this by using English data before we look at its applicability to other languages in Sect. 4.5 in the context of discussing the “low” origin of DEMs and the position of FocP inside DP. To start, according to Bolinger, referent-modifying APs describe temporary properties of individuals whereas reference-modifying ones describe inherent properties. Yet he does not provide a detailed account of exactly how such interpretative differences are structurally correlated other than stating that, in English, reference-modifying APs are always prenominal and postnominal ones are always referent-modifying. The present analysis offers a much more detailed view on how the syntactic position of an N modifier may map onto its meaning and vice versa. To see this, compare the English sentences in (76)–(77): the highlighted ADJs in (76) are construed as reference-modifying whereas their counterparts in (77) are construed as referent-modifying. (76) Data instantiating reference-modifying ADJs in English: a. John is [a skillful doctor]. b. Why can’t astronomers agree on the number of [visible stars]? Because the number can change depending on the latitude or the season of the year. (77) Data instantiating referent-modifying ADJs in English: a. There was a very skillful doctor at the clinic. [That skillful doctor] was from China. b. Only a few stars were visible tonight. And among [those few visible stars] were Capella and Sirius.

In the proposed framework, all the ADJs in (76)–(77) enter into the DP structure by merging at [Spec, SortP] since all of them have subsective and potentially gradable semantics. They end up receiving apparently different types of

4.4 Some Immediate Welcome Results of the New Analysis

135

interpretation, however, because the nominal projections containing them denote different types of things and this is reflected in their syntax. More specifically, the nominal a skillful doctor in (76a) is predicative, and therefore in our analysis, it constitutes a DPp. Furthermore, this nominal contains no DET or DEM, so its LocP is assumed to be empty (or it is not even projected). Given this, nothing inside of the Low Field raises to the High Field, as sketched in (78). Here, the indefinite article a is base-generated under the head position of PlP which hosts [-Pl] feature and then it raises to Dp to pronounce that FP in compliance with PELI; the idea that when occurring as part of a predicative nominal, a only encodes [-Pl], rather than carrying any referential import, is supported by the fact that it can be replaced by the numeral one (e.g., John is one skillful doctor.) modulo some pragmatic meaning difference that comes about when one replaces a in sentences like (76a).30 In addition to the raising of a to the head position of DPp, the SortP raises to [Spec, PlP+] for number agreement purposes. But other than that, no visible changes are made to the underlying structure. Consequently, the AP skillful receives a reference-modifying interpretation. (78) Derivation of a skillful doctor in (76a): DPp Dp Dp

LocP PlP+ PlP Pl [-PL(URAL)] a AP skillful

SortP nP doctor

In the case of (76b), the nominal visible stars refers to a non-specific plural individual, so it constitutes a DPd/r whose LocP is phonologically empty but which hosts [+ref(erential); −deic(tic); −def(inite); −spec(ific)] features, as depicted in (79). As a result, the null morpheme under Loc raises to the Dd/r position for feature valuation purposes, and the AP visible receives what is known as a reference-modifying interpretation since it occurs at [Spec, SortP], just like skillful in (76a). And I further assume that the entire DPp subsequently raises to [Spec, DPd/r] to pronounce the FP. (See (83) for a similar structure.) 30 Notably, Heycock and Zamparelli (2005) offer a similar treatment of a but while they assume that the indefinite article in English always surfaces under PlP, on the present analysis, some occurrences of it would surface at [Spec, DPd/r], namely cases where the nominal headed by it denotes a specific, referential individual that takes wide scope in the sentence as in A friend of mine bought a really nice car today.

136

4 A New Formal Analysis of AOR

(79) Derivation of visible stars in (76b): DPd/r Dd/r Dd/r [+REF] [-DEIC] [-DEF] [-SPEC]

DPp LocP Loc Loc [+REF] [-DEIC] [-DEF] [-SPEC]

PlP+ PlP Pl [+PL] -s AP visible

SortP nP star

As for (77a), the bracketed nominal that skillful doctor contains an overt DemP, which is assumed to be base-generated at [Spec, LocP], as given in (80), so this nominal carries not only referential but also definite, specific, and deictic semantics, and as a result, the AP skillful is felt to be receiving a referent-modifying construal even though it stays inside the Low Field. (Essentially the same explanation applies to the interpretation of visible in (77b), so I do not provide a separate tree for it.) Notably, comparing (79) and (80) shows that English fulfills PELI disjunctively by having either [Spec, DPd/r] or Dd/r pronounced but not both.

4.4 Some Immediate Welcome Results of the New Analysis

137

(80) Derivation of that skillful doctor in (77a): DPd/r Dd/r Dd/r [+REF] [+DEIC] [+DEF] [+SPEC]

DPp LocP DemP that

Loc [+REF] [+DEIC] [+DEF] [+SPEC]

Loc PlP+ PlP Pl [-PL]

SortP AP

nP

skillful

doctor

Turning now to comparing the present analysis with Larson’s: Larson incorporates Bolinger’s reference- versus referent-modifier distinction by suggesting that modifiers receiving generic interpretations occur in the NP-inner space and those receiving non-generic interpretations occur in the NP-outer space, but he leaves open the question of exactly where in the nominal structure the two types of modifiers occur. In the present analysis, this question is answered as follows: modifiers receiving generic construals occur below SortP if they are APs, but between DPp and SortP if they are RCs. On the other hand, modifiers receiving non-generic construals occur above DPp if they are RCs; if they are APs, they may still stay inside DPp but the nominal projection in which they occur has to be bigger than a DPp. Taken together, these show that in the current framework, the DPp level serves as a demarcation between what Larson calls the NP-internal space and the NP-external space. The present analysis also gives us a handle on data like (81), where two potentially subsective ADJs occur inside the same DP and while the AP that occurs closer to the head N receives a generic interpretation, the one that occurs farther from it carrying a focal stress receives a non-generic interpretation. (81) English data instantiating focused ADJs with non-generic construals only: a. [Most UNSKILLFUL skillful doctors] were found at that clinic; unskillful in the sense that they just didn’t know how to relate to patients. b. Tonight, [every VISIBLE visible star] was found in the eastern sky.

Under our analysis, the lower ADJs in (81) merge at [Spec, SortP] and the higher ones merge directly at [Spec, FocP], as depicted in (82). Notably, I treat most in

138

4 A New Formal Analysis of AOR

(82a) as a numeral AP which is base-generated at [Spec, UnitP] but later moves to [Spec, DPq] to license the relevant D features; I do not treat it as heading UnitP or DPq because, unlike quantificational DETs like every and each, it cannot be folp lowed by a numeral ADJ (e.g., Similar problems arose every three months/every single month versus *Similar problems arose most three months), so it is most likely a phrasal element that is base-generated at [Spec, UnitP], which is reserved for numeral ADJs in the current framework. Apart from the raising of most, no other movement occurs. Consequently, ‘Quantifier > focused ADJ > unfocused ADJ > N’ obtains as the surface order for (81a). Note also that for simplicity, I have omitted a DPp in (82), but the fact that a non-generic AP is sitting inside a DPp here should not run afoul of my statement above that in the present analysis, DPp serves as a dividing line between the NP-internal and NP-external spaces because the DP structure given in (82) is larger than a DPp and therefore it in fact shows that for any nominal to contain an N modifier that receives a non-generic interpretation, it has to be minimally comprised of a DPq. Finally, I do not provide a separate tree for (81b) here, but that is because we will be looking at a case where every occurs as a quantificational DET shortly. (82) Derivation of most UNSKILLFUL skillful doctors in (81a): DPq Dq Dq

FocP AP

LocP UnitP

UNSKILLFUL AP

PlP+

most

PlP Pl -s

SortP AP

nP

skillful

doctor

Turning now to comparing the present analysis with Cinque’s and Sproat and Shih’s analyses: in Cinque’s analysis, what he calls ‘dP’ serves as the dividing line for direct N modifiers and indirect N modifiers since the former type of N modifiers merge below dP and stay there whereas the latter type of N modifiers merge above it. But adopting his line of analysis cannot square with the fact that even RCs, which are indirect N modifiers in the way Cinque and Sproat and Shih use the term, may receive reference-modifying interpretations, as we saw in Chap. 3 as well as in Sect. 4.1.

4.4 Some Immediate Welcome Results of the New Analysis

139

Under our analysis, RCs receiving so-called direct adnominal interpretations merge at the DPp level, so we now have a way to explain why being a “direct” modifier does not necessarily mean that it is morpho-syntactically simplex. To illustrate this, let us first reconsider (41a), which is repeated below for convenience: in this sentence, the subject DP contains an RC and it receives a direct adnominal interpretation since it is construed as characterizing people who are known as politicians. (41) a. Politicians [who make extravagant promises] aren’t trusted.

Under the present analysis, the “direct” semantic construal of the RC in (41a) follows from the fact that the DP that embeds it is derived as in (83): once the structure inside the DPp region is constructed, the RC is base-generated at the [Spec, DPp] which is created right above the pre-existing DPp. After this, the null morpheme under Loc raises to the head position of DPd/r to have the [+referential; −deictic; −definite; −specific] features checked off. This is followed by the lower DPp raising to the Spec of DPd/r to pronounce the FP. Consequently, the RC surfaces postnominally even though its base position is prenominal. (83) Derivation of the subject DP in (41a) under the present analysis: DPd/r D d/r Dd/r [+REF] [-DEIC] [-DEF] [-SPEC]

DPp+ DPp RC

DPp LocP

who make extravagant promises Loc [+REF] [-DEIC] [-DEF] [-SPEC]

PlP+ PlP Pl -s

SortP nP politician

Reconsider now (44b), which is again repeated below. The predicative nominal in this sentence has the derivation given in (84): here, the head N is singular and the

140

4 A New Formal Analysis of AOR

size of the nominal is smaller than that in (41a), and being a non-referential nominal, the LocP is empty, so it does not raise to the head position of DPd/r. (44) b. John is a student [who studies physics]. . (84)

Structure of the predicative nominal in (44b) under the present analysis: DPp+ DPp RC

DPp

who studies physics

Dp Dp

LocP PlP+ PlP [-PL] a

SortP nP student

Turning to cases where an RC occurs inside a quantificational DP such as (44c, d): (44) c. Every student [who likes Let it Be] was there. d. Every [sincere] person [who attended the party] can tell you that.

In our analysis, if an RC at hand is construed as describing an inherent or generic property of individuals as in (44c), then it is base-generated at the Spec position of an adjoined DPp, as depicted in (85). If the N modifier is construed as describing an occasion-based or temporary property of individuals as in (44d), then it is base-generated at the Spec position of an adjoined DPq, as shown in (86). The RCs in these data surface postnominally, following other N dependents, however, because after every moves from Unit to Dq, the lower DPp raises to [Spec, DPp+] in (44c), and the lower DPq raises to [Spec, DPq+] in (44d), as depicted below, and such additional movements are presumably triggered by language-specific needs to prevent phonologically heavy constituents from occurring to the left of functional heads.

4.4 Some Immediate Welcome Results of the New Analysis

(85)

141

Derivation of the Subject DP of (44c): DPq Dq Dq

DPp+ DPp RC

DPp

who likes “Let It Be”

LocP UnitP Unit every

PlP+ PlP Pl [-PL]

SortP

student (86)

Derivation of the Subject DP of (44d): DPq+ DPq RC who attended the party

DPq Dq

DPp UnitP Unit every

PlP+ PlP Pl [-PL]

SortP

sincere person

Notably, the proposed way of handling things receives further support from the fact that when the two RCs in (44c, d) co-occur inside the same DP, they occur only in a certain order. To see this, consider (87).

142

4 A New Formal Analysis of AOR

(87) a. Every man [RC1 who likes Let It BE] [RC2 who attended the party] was from Dallas. (RC 1 > RC2) b.??Every man [RC2 who attended the party] [RC1who likes Let It BE] was from Dallas. (RC2 > RC1)

Under the present analysis, such ordering restrictions on English RCs stem from the following reasons: RC1 in (87) is reference-modifying, so it is base-generated at an adjoined [Spec, DPp]. RC2 is referent-modifying but it still restricts the domain of quantification for the quantificational DET every, so it is base-generated at an adjoined [Spec, DPq]. Despite their base-positions, they surface in the ‘RC1 > RC2’ order, however, because after they merge, the lower DPp raises to [Spec, DPp+], yielding an ‘N > RC1’ order, and that movement is followed by the lower DPq raising to [Spec, DPq+], which yields ‘every > N > RC1 > RC2’ order, as depicted in (88). (88)

Derivation of the Subject DP of (87a) under the present analysis: DPq+ DPq RC2

DPq

who attended the party

Dq Dq

DPp+ DPp RC1

DPp

who likes “Let It Be”

UnitP Unit every

PlP+ PlP Pl [-PL]

SortP man

The analysis presented and defended here can also readily handle data containing integrated but non-restrictive modifiers as well as a DEM and an AP. To see this, reconsider (45d).

4.4 Some Immediate Welcome Results of the New Analysis

143

(45) d. That [old] book [that has been sitting there forever] is from my greatgrandfather.

Under the present analysis, the ADJ old inside the subject DP in (45d) has the same syntax as the ADJ skillful in (80) since it also occurs following the DEM that and for it to be used felicitously, it must have been the case that it was mentioned in the previous utterance as a characteristic property of the individual at hand. As for the RC, given our assumptions about the syntax of integrated but non-restrictive RCs, it is base-generated at an adjoined [Spec, DPd/r] that is created atop the pre-existing DPd/r. But the entire DP instantiates ‘DEM > ADJ > N > RC’ in the end because, after the SortP moves to [Spec, PlP+] for number agreement purposes and the LocP moves to [Spec, DPd/r] for [+referential; +definite; +deictic; +specific] feature licensing purposes, the lower DPd/r raises to DPd/r+ in a manner similar to the roll-up movements in (88), as pictured in (89). (89)

Derivation of the Subject DP of (45d) under the present analysis: DPd/r+ DPd/r RC that has been sitting there forever

DPd/r DPp LocP DemP

PlP+

that

PlP [-PL]

SortP AP

nP

old

book

Related to (87a) and (45d), it should be noted that the proposed analysis can deliver desirable results for data like (90) as well. In (90), two RCs occur inside the same definite referential DP modified by the and only but in a fixed order. (90) a. This is the only river [RC1 that is currently navigable] [RC2 that will take you to the city]. (RC1 > RC2) b.??This is the only river [RC2 that will take you to the city] [RC1 that is currently navigable]. (RC2 > RC1)

144

4 A New Formal Analysis of AOR

Under our analysis, the fixed order between the two RCs in (90) is due to the way the intended meaning of the DP containing them is composed: in computing the DP’s meaning, one needs to first combine RC1’s meaning with the head N’s since it describes a more inherent property of individuals than RC2 does. Given this, the postcopular DP in (90a) is derived as in (91). That is, at the initial stage of the derivation, being a numeral ADJ, only is base-generated at [Spec, UnitP] and the is base-generated under Loc. After this, being a reference-modifier in the sense of Bolinger, RC1 merges at an adjoined [Spec, DPp], and being a referent-modifier, RC2 merges at an adjoined [Spec, DPd/r]. This is followed by the SortP moving to [Spec, PlP+] for number agreement and the moving to Dd/r for [+referential; +definite; −deictic; +specific] feature licensing. This is in turn followed by the lower DPp raising to DPp+, which is followed by the lower DPd/r raising to DPd/r+. Consequently, the two RCs occur not only postnominally but also in the order of ‘RC1 > RC2’. (91)

Derivation of the postcopular DP in (90a) under the present analysis: DPd/r+ DPd/r RC2

DPd/r

Dd/r that is currently [+REF] navigable [-DEIC] [+DEF] [+SPEC]

DPp+ DPp RC1

DPp LocP

that will take you to the city

Loc UnitP [+REF] PlP+ [-DEIC] AP [+DEF] PlP [+SPEC] only the SortP [-PL] river

For the sake of completeness, let me turn now to demonstrating how the proposed analysis may capture the syntax of English supplementary RCs. To see this, let us first consider the two sentences in (92).

4.4 Some Immediate Welcome Results of the New Analysis

145

(92) a. This river, which is navigable right now, will definitely take you to Cairo. b. You should take this river , which is navigable right now.

To begin with the derivation of (92a), in this sentence, the supplementary RC occurs between the sentential subject and the modal verb will, and it is marked off by pauses. In our analysis, this surface form obtains because after Sppl selects for the TP-level structure represented by the matrix clause, the subject DP raises to an adjoined [Spec, SpplP], giving rise to an ‘N > RC > will (Tense)’ order, as shown in (93), and since we assume that Sppl hosts a pause and the adjunction configuration given in (93) creates another pause after it, the RC surfaces both preceded and followed by a pause. (93)

Derivation of (92a) under the present analysis: SpplP SpplP RC

Sppl

Sppl , which is navigable DPd/r right now this river

TP T T will

VP

definitely take you to Cairo

In the case of (92b), after Sppl selects for a TP, the object of the sentence raises to an adjoined [Spec, SpplP], but there occurs yet another movement, namely, the entire TP raising to the Spec position of SpplP+, as given in (94). Consequently, the supplementary RC occurs following the TP, and there is an intonational break between them, which is due to the adjunction structure that raising the object DP has created. Here too, the Sppl hosts a pause but it is conflated with a period, given its sentence-final location. But other than that, the surface order of the RC relative to its head N is identical to what is exhibited by (92a); that is, in both cases, the RC occurs immediately after the head N, with an intonational break between them.

146

(94)

4 A New Formal Analysis of AOR

Derivation of (92b) under the present analysis: SpplP+ SpplP SpplP RC

Sppl

Sppl , which is navigable DPd/r right now you

TP T T should

VP V take

DPd/r

this river

To summarize this subsection, then, the proposed decompositional split-DP analysis is capable of capturing the syntax and semantics of various N modifiers in English without losing the key insights of the relevant existing analyses in the process of implementing them into the syntactic framework adopted here. What is particularly noteworthy is that the present analysis incorporates Bolinger’s insightful distinction between reference- versus referent-modification without incorrectly assuming that all reference-modifiers are direct N modifiers and all referent-modifiers are indirect N modifiers, thereby improving on Sproat and Shih’s and Cinque’s analyses. Furthermore, it makes correct predictions about the position of various types of RCs inside a DP, and in so doing, it improves on Larson’s and Svenonius’ analyses as well as all the other works on the syntax of N modifiers that are currently available in the literature including Laenzlinger’s and Cinque’s.

4.5

More on the “Low” Origin of DEMs and the Position of FocP

In this section, I provide additional support for the relatively low origin of DEMs and the need to posit a FocP above what I call LocP by comparing my analysis with two existing analyses of DEMs in Romance languages, namely, Brugè 1996, 2002 and Bernstein 2001. Brugè (1996, 2002) is among the first and strongest advocates of the low origin analysis of DEMs. The upshot of her analysis is that DEMs are base-generated below all adnominal ADJs but above possessive modifiers and the head N. This idea is motivated by the fact that DEMs occur DP-finally in Spanish, as exemplified in (95).

4.5 More on the “Low” Origin of DEMs and the Position of FocP

147

(95) Spanish DEM occurring DP-finally: El libro viejo este (Art > N > ADJ > DEM) the book old this

As Brugè herself acknowledges, however, in numerous languages, DEMs obligatorily occur in DP-initial position, and even in Spanish, they may occur DP-initially, as shown in (96). (96) Spanish DEM occurring DP-initially: Este libro viejo (DEM > N > ADJ) this book old’ ‘this old book’

To account for cases like (96), Brugè hypothesizes that in every language, DEMs must check their [+referential; +deictic] features by moving to [Spec, DP] at LF, but depending on the language, the feature checking can be done before Spell-Out (as well as at LF). In other words, data like (96) obtain because in Spanish, DEMs may raise to [Spec, DP] in narrow syntax to check their [+referential; +deictic] features, an option that is not available for languages like English. Viewed in this way, then, pre- versus postnominal occurrences of DEMs result from different parameter settings across languages. Brugè’s analysis has several problems, however. First of all, under her analysis, DEMs are underlyingly c-commanded by lexical ADJs and in the case of Spanish, they are in the scope of lexical ADJs even at the surface level, but DEMs are inherently D-linked whereas lexical ADJs are not, so from the semantic point of view, lexical ADJs should in fact be in the scope of a DEM instead; that is, the c-command relation should be the other way around. Secondly, Brugè treats prenominal and postnominal DEMs merely as syntactic variants of each other with no semantic differences between them, but from what I have found by consulting several Spanish speakers, while postnominal DEMs carry a contrastive focus meaning, prenominal ones only carry a purely anaphoric or unfocused deictic meaning. For example, according to my informants, (95) can be felicitously uttered in a context where the speaker is comparing two old books that are visible in the discourse context, but (96) cannot be; conversely, while (96) can be uttered in a context where some old book has been introduced to the discourse and the speaker is referring to it, (95) cannot be.31 Thirdly, Brugè’s analysis is troubled by the fact that not all occurrences of DEMs in Spanish are DP-initial or DP-final; sometimes, a DEM may occur in an

31

I wish to thank Diego Pascual, Olga Pahom, Idoia Elola, Cole Morrison, and Luis Alonso-Ovalle for the judgments on the Spanish data presented here, which are provided by Diego Pascual unless otherwise noted.

148

4 A New Formal Analysis of AOR

intermediate position, preceding an ADJ, as observed by Roca (1996) and as illustrated by (97). (97) Spanish DEM occurring DP-medially: ?el libro ese viejo (Art > N > DEM > ADJ) the book that old ‘that old book’

The existence of data like (97) casts further doubt on the idea that DEMs are generated below any type of ADJs; if the intermediate position for the DEM ese in (97) results from moving across the ADJ viejo, then it is not clear what might trigger this movement since, given the way things work in Brugè’s analysis, when occurring DP-medially, a DEM cannot have its [+referential; +deictic] features checked off, nor can it receive a focus interpretation. Finally, to support her low origin analysis of DEMs, Brugè points out that even in languages like French and Italian, DEM reinforcers (RFs) occur DP-finally despite the fact that such languages do not permit postnominal DEMs. But what she misses out on is the fact that DPs containing DEM RFs in French and Italian always receive a focus interpretation, as noted by Bernstein (2001) and as illustrated by the data given in (98)–(99). (98) French DEM co-occurring with a RF: cette femme intelligente ci this woman intelligent RF ‘THIS (not that) intelligent woman’ (99) Italian DEM co-occurring with a RF: questo libro vecchio qui this book old RF ‘THIS (not that) old book’

(DEM > N > ADJ > RF)

(DEM > N > ADJ > RF)

These problems do not arise in the present analysis even though it postulates a “low” origin for DEMs as well. First of all, in the proposed analysis, DEMs (and definite articles) merge in a position that c-commands restrictive ADJs [see (74)], and therefore, unlike the case with Brugè’s analysis, configurations where D-linked elements are c-commanded by non-D-linked ones do not come about. Secondly, our analysis correctly predicts that DP-final versus DP-initial position of a DEM correlates with focused versus non-focused semantics. To demonstrate this with (95) (ignoring, for simplicity, precisely how gender and number agreements are done in Spanish), the focused interpretation associated with the postnominal DEM is derived as in (100). That is, after the definite article el, the DEM este, the ADJ viejo, and the N libro merge under Loc, at [Spec, LocP], at [Spec, SortP], and inside the nP, respectively, the nP moves to [Spec, SortP+], giving rise to an ‘N > ADJ’ order. This is followed by the raising of este to [Spec, FocP] to have its [+focus] feature checked off, which is in turn followed by the remnant

4.5 More on the “Low” Origin of DEMs and the Position of FocP

149

movement of the LocP to [Spec, DPd/r] to check off the [+referential; +definite; +deictic; +specific] features. Consequently, the DP surfaces with ‘Art > N > ADJ > focused DEM’ linear order. (100) Derivation of the DP-final DEM order in (95):32 DPd/r FocP LocP

DemP2 este

t2

Loc Loc [+REF] [+DEIC] [+DEF] [+SPEC] el

UnitP PlP SortP+ SortP

nP1 libro

AP

t1

viejo 32

:

As for the derivation of (96) where a DEM surfaces DP-initially, here too, the nP raises to [Spec, SortP+], giving rise to an ‘N > ADJ’ order, but this DEM does not have the [+focus] feature. Hence, the DemP raises to [Spec, DPd/r], thereby having the [+referential; +definite; +deictic; +specific] features licensed, as depicted in (101). As a result, the entire DP receives an unfocused but still referential interpretation. Notably, in this particular case, the definite article el is not pronounced and we can attribute this to the Doubly Filled DP Filter in the sense of Giusti (1993), which dictates that either the Spec or the head position of some

32

Given the derivation in (100), the reader may wonder why English never allows for the definite article the to co-occur with a focused DEM, as shown in (ia), even though DEMs can be focused in English as well, as shown in (ib).

(i) a. *the old book THIS/THAT b. THIS/THAT old book

Under the present analysis, cases like (ia) are not attested because while English DEMs always carry [+deictic], its definite article the always carries [−deictic] feature [compare (80) and (91)] and as a result, they would never be base-generated inside the same LocP from the get-go. Under this way of viewing things, then, the cross-linguistic variation on the co-occurrence between a definite article and a DEM may boil down to whether the definite article in the language at hand may carry [+deictic] or not. I revisit this issue in Chap. 6.

150

4 A New Formal Analysis of AOR

FP be pronounced but not both; unlike in (96), the definite article can be pronounced in (95) because the DemP occupies [Spec, FocP] at Spell-Out, not [Spec, DPd/r]. (101) Derivation of the DP-initial DEM order in (96): [DPd/r [DPp [FocP [LocP [este] [UnitP/PlP [SortP+ [SortP viejo [nP libro]]]]]]]]

When it comes to cases like (97), further investigation will be necessary to determine precisely what kind of semantic consequences such non-canonical constituent order may have for the entire DP. But about such data, Bernstein (2001) states that while it is hard to precisely characterize their meanings, they can be labeled as some sort of ‘non-neutral unfocused’ interpretation. Furthermore, I should point out that such cases contain a definite article which occurs DP-initially. Given this, we can hypothesize that in cases like (97), the nP surfaces at [Spec, FocP] as a result of moving across the DEM and the ADJ, and the definite article raises from Loc to Dd/r to license the [+referential; +definite; +deictic; +specific] features, thereby complying with PELI, as depicted in (102). (102) Derivation of the intermediate postnominal DEM order in (97): [DPd/r el [FocP [LocP [DemP ese ]

[UnitP/PlP [SortP viejo [nP libro]]]]]]]]

Finally, under the present analysis, data (98)–(99), which exemplify the DP-final position and the obligatory focused interpretation of DEM RFs in French and Italian, can be derived as follows: at the initial point in the derivation, the N and the ADJ are base-generated inside the nP and at [Spec, SortP], respectively, and the N raises to [Spec, SortP+], yielding an ‘N > ADJ’ order, in a manner similar to Spanish. But when it comes to DEMs, things look a bit different than they do in Spanish: French and Italian DEMs always occur in complementary distribution with the definite article, as one can see from (103)–(104), so they are best treated as heads than as phrases (see also Svenonius 2008 for a similar view of French DEMs). Given this, in our framework, they are base-generated under Loc and their RFs are base-generated at [Spec, LocP]. Yet ‘DEM > N > ADJ > RF’ obtains as their surface order in the end, and this is because, at the next point in the derivation, the RF moves to [Spec, FocP] to check off the [+focus] feature, and after this, the LocP undergoes a remnant movement to [Spec, DPd/r], consequently licensing the [+deictic; +referential; +definite; +specific] features of the DP as well as fulfilling PELI, as sketched in (105).

4.5 More on the “Low” Origin of DEMs and the Position of FocP

151

(103) Complementary distribution between DEMs and definite articles in French: a. ce livre (DEM > Msc.N) ‘this/that book’ b. le livre (Def.Art > Msc.N) ‘the book’ c. *ce le livre (DEM > Def.Art > Msc.N) c’.*le ce livre (Def.Art > DEM > Msc.N) Intended: ‘the/this/that book’ d. cette fille (DEM > Fmn.N) ‘this/that girl’ e. la fille (Def.Art > Fmn.N) ‘the girl’ la f. *cette fille (DEM > Def.Art > Fmn.N) f’. *la cette fille (Def.Art > DEM > Fmn.N) Intended: ‘the/this/that girl’ (104) Complementary distribution between DEMs and definite articles in Italian: a. questo libro (DEM > Msc.N) ‘this book’ b. il libro (Def.Art > Msc.N) ‘the book’ libro (DEM > Def.Art > Msc.N) c. *questo il c’.*il questo libro (Def.Art > DEM > Msc.N) Intended: ‘the/this book’ d. questa casa (DEM > Fmn.N) ‘this house’ e. la casa (Def.Art > Fmn.N) ‘the house’ f. *questa casa (DEM > Def.Art > Fmn.N) la f’. *la questa casa (Def.Art > DEM > Fmn.N) Intended: ‘the/this house’ (105) Derivation of the Italian DP in (99) with a DP-final RF: DPd/r FocP LocP

RF2 qui

Loc

t2 Loc questo

UnitP PlP SortP+ SortP

nP1 libro

AP vecchio

t1

152

4 A New Formal Analysis of AOR

As things turn out, the present analysis partly recalls Bernstein’s (2001) analysis but it also fares better than hers in several respects: Bernstein points out the semantic difference between DP-finally occurring DEMs and DP-initially occurring ones in Romance languages by labeling the former ‘focused’ and the latter ‘neutral’. And to capture this, she postulates, in essentially the same spirit as Roca (1996), that DEMs are generated in the Spec of an FP that is below DP, c-commanding all lexical ADJs and the head N, but the phrase that contains the head N and its adjectival modifiers scrambles to the left of the DEM, and consequently, the DEM becomes stranded in a position where it receives nuclear stress, namely, DP-final position,33 as given in (106). When we compare (106) with (100), we see that the end product of her analysis is not so different from ours. (106) Derivation of a DP-final focused DEM under Bernstein’s analysis: el [XP libroi interestante ti]k este [XP t]k the [ book interesting] ‘this interesting book’ [Bernstein 2001: (28)]

What is also notable is that, under her analysis too, focused DEMs are generated higher than all restrictive N modifiers from the get-go, so the scope relation between them falls out naturally, unlike the case with Brugè’s analysis. Furthermore, by assuming that unfocused DEMs merge at [Spec, DP] whereas focused DEMs merge in a lower position, Bernstein provides a way to capture the correlation between the syntactic position of a DEM inside a DP and its focused or unfocused interpretation in a way similar to what we have done here. That said, Bernstein’s analysis has difficulty accounting for cases involving focused ADJs such as those in (107)–(109): according to her, an N dependent receives a focus interpretation only when some syntactic phrase containing the head N and all its adjectival modifier [i.e., what she labels as ‘XP’ in (106)] has moved to a higher position across that N dependent, thereby making it occur DP-finally. But in sentences like (107)–(109), what is left behind is an ADJ, not a DEM, and this raises the question of how such derivations may come about, given the syntactic configuration she assumes. To take (107) for example, what could be that XP which c-commands the head N mesa ‘table’, which can move across the ADJ redonda ‘round’ but still containing the N mesa and its PP modifier madera, whereby making the ADJ surface DP-finally? As far as I can see, such a phrase does not exist in syntactic structures like (106).

33

In Spanish, DP-final constituents receive nuclear stress, due to the Nuclear Stress Rule in the sense of Zubizaretta (1998).

4.5 More on the “Low” Origin of DEMs and the Position of FocP

153

(107) Juan necesita una mesa de madera Juan need-N.PST.3RD.SG a table of wood ‘Juan needs a ROUND wooden table (not a square one).’ (108) Juan compró una mesa a table Juan buy-PST.3RD.SG no una CUADRADA. not a square ‘Juan bought a ROUND table, not a SQUARE one.’

REDONDA. round

REDONDA, round,

(109) Diego compró la mesa REDONDA, Diego buy-PST.3RD.SG the table round, no la CUADRADA. not the square ‘Diego bought the ROUND table, not the SQUARE one.’

Another challenge for Bernstein’s analysis is presented by data like (110) which involve predicative nominals: as we saw in Sect. 4.3.1, there is reason to think that predicative nominals are comprised of a smaller structure than referential nominals (see also Dayal 2009). And if that is indeed the case, then we expect them to be devoid of the FP that Bernstein assumes to be present to the left of where DEMs merge, to which the maximal projection containing an N and all its non-focused dependents move, because such an FP would be found only in a referential DP. And if this is a correct reasoning and if the structure she assumes is all there is to it, then data like (110) should not obtain because the material that appears to the left of the focused ADJ inside the subject predicate would have no place to move to. (110) Esa fue una mesa de madera REDONDA, table of wood round, That be-PST.3RD.SG a no una mesa de madera CUADRADA. not a table of wood square. ‘That was a ROUND wooden table, not a SQUARE wooden table.’

Under the present analysis, any DP internal element with [+focus] feature can move to [Spec, FocP] as long as it is base-generated below FocP, and after this, a remnant movement can happen across the focused item. Hence, the grammaticality of data like (107)–(109) does not come as a surprise and they can be readily derived. Moreover, under our analysis, a FocP is projected inside what I call a DPp, so predicative nominals are correctly predicted to host a focused-marked ADJ, as attested by (110).

154

4.6

4 A New Formal Analysis of AOR

Summary

In this chapter, drawing on several influential works on the syntax and semantics of N modifiers, I have presented what I call a new decompositional split-DP analysis under which the universal DP structure is comprised of three DP layers (DPd/r, DPq, and DPp) and all DP-internal functional elements originate from what I call the Low Field (i.e., inside DPp). The proposed analysis offers a more nuanced treatment of DEMs, definite articles, quantificational DETs, and focused N modifiers than have been previously possible. It also articulates the surface syntactic positions of various types of RCs relative to other DP-internal elements in ways that have not been attempted before. Furthermore, the current framework incorporates various two-way semantic/morpho-syntactic distinctions of N modifiers that have been made in the literature without losing their core insights. The proposed formal mechanism has been shown to yield desirable results for English data containing different kinds of RCs and APs as well as different types of functional elements. In addition, it has been shown to deliver promising results for data drawn from Spanish, French, and Italian in which a focused or an unfocused DEM co-occurs with an AP inside the same referential DP.

References Aboh, Enoch Oladé. 2004. The morphosyntax of complement-head sequences: Clause structure and word order patterns in Kwa. Oxford: Oxford University Press. Alexiadou, Artemis, Liliane Haegeman, and Melita Stavrou. 2007. Noun phrase in the generative perspective. Berlin: Walter de Gruyter. Beck, David. 1999. The typology of parts of speech systems: The markedness of adjectives. Toronto: University of Toronto dissertation. Bernstein, Judy B. 1997. Demonstratives and reinforcers in Romance and Germanic languages. Lingua 102: 87–113. Bernstein, Judy B. 2001. Focusing the ‘right’ way in Romance determiner phrases. Probus 13: 1–29. Bolinger, Dwight. 1967. Adjectives in English: Attribution and predication. Lingua 18: 1–34. Borer, Hagit. 2005. Structuring sense, vol. I: In name only. Oxford: Oxford University Press. Bosque, Ignacio, and Carme Picallo. 1996. Postnominal adjectives in Spanish DPs. Journal of Linguistics 32: 349–385. Bouchard, Denis. 2002. Adjectives, number and interfaces: Why languages vary? Amsterdam: North-Holland. Brugè, Laura. 1996. Demonstrative movement in Spanish: A comparative approach. In University of Venice working papers in linguistics, vol. 6 (1), 1–53. Brugè, Laura. 2002. The positions of demonstratives in the extended nominal projection. In Functional structure in DP and IP: The cartography of syntactic structures, vol. 1, ed. Guglielmo Cinque, 15–53. New York: Oxford University Press. Chierchia, Gennaro. 1995. Individual-level predicates as inherent generics. In The generic book, ed. Gregory Carlson, and Jeffry Pelletier, 176–223. Chicago: Chicago University Press. Chomsky, Noam. 1995. The minimalist program. Cambridge, MA: MIT Press.

References

155

Cinque, Guglielmo. 1990. Agreement and head-to-head movement in the Romance NP. Paper presented at the 20th Linguistic Symposium in Romance Languages, University of Ottawa. Cinque, Guglielmo. 1994. On the evidence for partial N movement in the Romance DP. In Paths towards universal grammar, ed. Guglielmo Cinque, Jan Koster, Jean-Yves Pollock, Luigi Rizzi, and Raffaella Zanuttini, 85–110. Georgetown, DC: Georgetown University Press. Cinque, Guglielmo. 1999. Adverbs and functional heads: A cross-linguistic perspective. Oxford, UK: Oxford University Press. Cinque, Guglielmo. 2003. The prenominal origin of relative clauses. Paper presented at the Workshop on Antisymmetry and Remnant Movement. New York University, Oct 31–Nov 1, 2003. Cinque, Guglielmo. 2005. Deriving Greenberg’s Universal 20 and its exceptions. Linguistic Inquiry 36: 315–332. Cinque, Guglielmo. 2008a. More on the indefinite character of the head of restrictive relatives. In Selected proceedings of the 34th Incnotro di Grammatica Generativa, ed. Benincà, Paola, Federico Damonte, and Nicoletta Penello, 3–24. Padua: Unipress. Cinque, Guglielmo. 2008b. Two types of nonrestrictive relatives. In Empirical issues in syntax and semantics, vol. 7, ed. Bonami, Oliver, and Patricia Cabredo Hofherr, 99–137. Paris: CNRS. Cinque, Guglielmo. 2010. The syntax of adjectives. A comparative study. Cambridge, MA: MIT Press. Cowell, Mark W. 1964. A reference grammar of Syrian Arabic. Washington: Georgetown University Press. Dayal, Veneeta. 2009. Bare noun phrases. Ms., Rutgers University. Delsing, Lars-Olof. 1993. The internal structure of noun phrases in the Scandinavian languages. Lund, Sweden: University of Lund dissertation. Demirdache, Hamida. 1991. Resumptive chains in restrictive relatives, appositives, and dislocation structure. Cambridge, MA: MIT dissertation. Demonte, Violeta. 2005. Meaning-form correlations and the order of adjectives in Spanish. Ms., Universidad Autónoma de Madrid. Dimitrova-Vulchanova, Mila, and Giuliana Giusti. 1998. Fragments of Balkan nominal structure. In Possessors, predicates and movement in the determiner phrase, ed. Artemis Alexiadou, and Chris Wilde, 333–360. Amsterdam: John Benjamins. Dixon, R.M.W. 1977. Where have all the adjectives gone? Studies in Language 1: 19–80. Dixon, R.M.W. 1982. Where have all the adjectives gone? The Hague: Mouton. Dixon, R.M.W. 2004. Adjective classes in typological perspective. In Adjective classes: A cross-linguistic typology. Explorations in linguistic typology, ed. Dixon, R.M.W., and Alexandra Y. Aikhenvald, 1–49. Oxford: Oxford University Press. Doran, Ryan B., and Gregory Ward. 2015. Proximal demonstratives in predicate NPs. In Proceedings of the annual meeting of the Berkeley linguistics society, vol. 41. ed. by Jurgensen, Anna E., Hannah Sande, Spencer Lamoureux, Kenny Baclawski, and Alison Zerbe, 61–70. Berkeley, CA: Berkeley Linguistic Society. Dryer, Matthew. 1992. The Greenbergian word order correlations. Language 68: 865–895. Elbourne, Paul. 2005. Presuppositional demonstratives. Ms., Universität Potsdam. Felser, Claudia, and Laura Rupp. 2001. Expletives as arguments: Germanic existential sentences revisited. Linguistische Berichte 187: 289–324. Francez, Itamar. 2007. Existential propositions. Stanford, CA: Stanford University dissertation. Francez, Itamar. 2009. Existentials, predication, and modification. Linguistics and Philosophy 32: 1–50. Giusti, Giuliana. 1993. La sintassi dei determinant [The syntax of the determiner]. Padova: Unipress. Giusti, Giuliana. 2002. The functional structure of noun phrases: A bare phrase structure approach. In Functional structure in DP and IP: The cartography of syntactic structures, vol. 1, ed. Guglielmo Cinque, 54–90. New York: Oxford University Press. Giusti, Giuliana. 2005. At the left periphery of the Romanian noun phrase. In On space and time in language, ed. Martine Coene, and Liliane Tasmowski, 23–49. Cluj-Napoca: Clusium.

156

4 A New Formal Analysis of AOR

Givón, Talmy. 1980. Notes on the semantic structure of English adjectives. Language 46: 816–837. Greenberg, Joseph H. 1963. Some universals of grammar with particular reference to the order of meaningful elements. In Universals of language. Report of a conference held at Dobbs Ferry, NY, Apr 13–15, 1961, ed. Greenberg, Joseph H., 73–113. Cambridge, MA: MIT Press. Guttièrez-Rexach, Javier, and Enrique Mallen. 2001. NP movement and adjective positions in the DP phases. In Features and interfaces, ed. Julia Herschensohn, Enrique Mallen, and Karen Zagona, 107–132. Amsterdam: John Benjamins. Guttièrez-Rexach, Javier, and Enrique Mallen. 2002. Toward a unified minimalist analysis of prenominal adjectives. In Structure, meaning, and acquisition in Spanish: Papers from the IV hispanic linguistics symposium, ed. Lee, James F., Kimberly L. Geeslin, and J. Clancy Clements, 178–192. Somerville, MA: Cascadilla Press. Hartmann, Jutta M. 2008. Expletives in existentials: English there and German da. Tilburg: University of Tilburg dissertation. Hetzron, Robert. 1978. On the relative order of adjectives. In Language universals, ed. Hansjakob Seiler, 165–184. Tübingen: Gunter Narr Verlag. Heycock, Caroline, and Roberto Zamparelli. 2005. Friends and colleagues: Coordination, plurality, and the structure of DP. Natural Language Semantics 13: 201–270. Horrocks, Geoffrey, and Melita Stavrou. 1987. Bounding theory and Greek syntax: Evidence from wh-movement in NP. Journal of Linguistics 23: 79–108. Huddleston, Rodney, and Geoffrey Pullum. 2005. A student’s introduction to English grammar. Cambridge, UK: Cambridge University Press. Isac, Daniela. 2003. Restrictive relative clauses vs. restrictive adjectives: An asymmetry within the class of modifiers. In Asymmetry in grammar, vol. 1: Syntax and semantics, ed. Di Sciullo, Anna Maria, 27–49. Amsterdam: John Benjamins. Ihsane, Tabea. 2000. Three types of possessive modifiers. In Generative Grammar in Geneva 1: 21–54. Geneva: University of Geneva. Ihsane, Tabea, and Genoveva Puskas. 2001. Specific is not definite. Generative Grammar in Geneva 2: 39–54. Julien, Marit. 2005. Nominal phrases from a Scandinavian perspective. Amsterdam: John Benjamins. Kang, Soon Haeng. 2006. The two forms of the adjective in Korean. In University of Venice working papers in linguistics, vol. 16, 137–163. Kariaeva, Natalia. 2004. Determiner spreading in Modern Greek: Split DP hypothesis. Ms., Rutgers University. Kayne, Richard S. 1994. The antisymmetry of syntax. Cambridge, MA: MIT Press. Koopman, Hilda, and Anna Szabolcsi. 2000. Verbal complexes. Current studies in linguistics, vol. 34. Cambridge, MA: MIT Press. Laenzlinger, Christopher. 2005. French adjective ordering: Perspectives on DP-internal movement types. Lingua 115: 645–689. Larson, Richard K. 1998. Events and modification in nominals. In Proceedings from semantics and linguistic theory (SALT) VIII, 145–168. Cornell University Press. Larson, Richard K. 2000. Temporal modification in nominals. Paper presented at the International Roundtable on the Syntax of Tense, University of Paris VII, France. Larson, Richard K., and Naoko Takahashi. 2007. Order and interpretation in prenominal relative clauses. In Proceedings of the workshop on Altaic formal linguistics II. MIT working papers in linguistics, vol. 54, 101–120. Cambridge, MA: MITWPL. Leu, Thomas. 2008. The internal syntax of determiners. New York City, NY: New York University dissertation. Leu, Thomas. 2015. The architecture of determiners. New York: Oxford University Press. Marantz, Alec. 2001. Words. Ms., MIT. McNally, Louise. 1997. An interpretation for the English existential construction. New York: Garland. (1992 University of California-Santa Cruz dissertation.)

References

157

McNally, Louise. 1998. Existential sentences without existential quantification. Linguistics and Philosophy 21: 353–392. Moro, Andrea. 1997. The raising of predicates: Predicative noun phrases and the theory of clause structure. Cambridge, UK: Cambridge University Press. Partee, Barbara H. 1986. Noun phrase interpretation and type-shifting principles. In Studies in discourse representation theory and the theory of generalized quantifiers, ed. Groendijk, Jeroen, Dick de Jongh, and Martin Stokhof, 115–143. Dordrecht: Foris. Pearson, Matt. 2000. Two types of VO languages. In The derivation of VO and OV., ed. Svenonius, Peter, 327–363. Amsterdam/Philadelphia: John Benjamins. Potts, Christopher. 2005. The logic of conventional implicatures. New York: Oxford University Press. (2003 University of California-Santa Cruz dissertation.) Rijkhoff, Jan. 2002. The noun phrase. Oxford, UK: Oxford University Press. Rizzi, Luigi. 1997. The fine structure of the left periphery. In Elements of grammar, ed. Liliane Haegeman, 281–337. Dordrecht: Kluwer Academic Publishers. Roca, Francesc. 1996. La determinación y la modificación nominal en español. Barcelona, Spain: Universidad Autónoma de Barcelona dissertation. Scott, Gary-John. 2002. Stacked adjectival modification and the structure of nominal phrases. In Functional structure in DP and IP: The cartography of syntactic structures, vol. 1, ed. Guglielmo Cinque, 91–120. New York: Oxford University Press. Spencer, Andrew. 1992. Morphological theory: An introduction to word structure in generative grammar. Oxford: Basil Blackwell. Sproat, Richard, and Chinlin Shih. 1988. Prenominal adjectival ordering in English and Mandarin. In Proceedings of NELS, vol. 18, 465–489. Amherst, MA: GLSA. Sproat, Richard, and Chinlin Shih. 1990. The cross-linguistics distribution of adjectival ordering restrictions. In Interdisciplinary approaches to language: Essays in honor of S-Y. Kuroda, ed. Georgopoulos, Carol, and Roberta Ishihara, 565–593. Dordrecht: Kluwer. Starke, Michal. 1995. On the format for small clauses. In Syntax and semantics, vol. 28: Small clauses, ed. Cardinaletti, Anna, and Maria Teresa Guasti, 237–269. New York: Academic Press. Svenonius, Peter. 2008. The position of adjectives and other phrasal modifiers in the decomposition of DP. In Adjectives and adverbs: Syntax, semantics, and discourse, ed. Chris Kennedy, and Louise McNally, 16–42. New York: Oxford University Press. Szabolcsi, Anna. 1983. The possessor that ran away from home. The Linguistic Review 3 (1): 89–102. Szabolcsi, Anna. 1994. The noun phrase. In Syntax and semantics, vol. 27: The syntactic structure of Hungarian, ed. Kiefer, Ferenc, and Katalin É. Kiss, 179–274. San Diego: Academic Press. Truswell, Robert. 2004. Attributive adjectives and the nominals they modify. Oxford University master’s thesis. Vangsnes, Øystein A. 2004. Rolling up the Scandinavian noun phrase. Paper presented at the 27th GLOW Colloquium. Athens, April 19–21, 2004. Vendler, Zeno. 1968. Adjectives and nominalizations. The Hague: Mouton. Williams, Edwin. 1983. Semantic vs. syntactic categories. Linguistics and Philosophy 6: 423–446. Williams, Edwin. 1984. There-insertion. Linguistic Inquiry 15: 131–153. Zamparelli, Roberto. 2000. Layers in the determiner phrase. New York: Garland. Zubizaretta, María Luisa. 1998. Prosody, focus, and word order. Cambridge, MA: MIT Press.

Chapter 5

Capturing the Korean Facts

In the preceding two chapters, I first showed that the linear order of N modifiers in Korean is regulated in part by the workings of the constraints given in (1) (Chap. 3). After this, I proposed what is given in (2) as the tri-partite DP structure shared by all languages and demonstrated how assuming this structure and certain head/phrasal movements may let us capture constituent order variations involving RCs and other focused or unfocused N dependents in languages like English, Spanish, French, and Italian (Chap. 4). (1) Constraints regulating the adjective ordering in Korean: a. Relatively Heavy Modifier > Relatively Light Modifier > N b. Discourse-Prominent Modifier > Non-Discourse-Prominent Modifier > N c. Verbal Modifier > Nominal Modifier > N d. Compositional semantics (e.g., scope, relative vs. absolute semantics)

© Springer Nature Switzerland AG 2019 M.-J. Kim, The Syntax and Semantics of Noun Modifiers and the Theory of Universal Grammar, Studies in Natural Language and Linguistic Theory 96, https://doi.org/10.1007/978-3-030-05886-9_5

159

160

5 Capturing the Korean Facts

(2) My proposal for the universal DP structure: SpplP Locus of Sppl Mod … DPd/r

High Field DPq

Locus of NRS Mod

Middle Field DPp

Low Field FocP

Loci of RS Mod

LocP DemP

UnitP/PlP NumP/Pl

SortP

AP([+Deg])

nP

AP([+θ])

P AP([+θ])

In this chapter, I apply this DP structure and the derivational mechanism that comes with it to Korean and examine how it fares with a language whose N modifiers have not been much studied within a formal framework, especially when compared to Indo-European languages like English, Spanish, French, and Italian. Our examination will show that the proposed analysis captures several important aspects of N modification phenomena in Korean, yet it falls short of capturing certain AORs attested by the language. To resolve these issues without abandoning the decompositional split-DP analysis we have developed (which has been shown to yield numerous positive results already), I propose an output-based filtering device within an optimality theoretic framework and show how doing so yields results that adopting a derivational approach alone cannot. In this context, I answer two outstanding questions raised in Chap. 3, namely, (i) what governs the ordering of FRCs inside the same nominal in Korean and (ii) why in Korean, occurring immediately after the DEM ku ‘the/that’ may allow for an N modifier to occur preceding a morpho-syntactically more complex N modifier in apparent violation of the typologically robust constraint given in (1a). We will see that answering the first question involves identifying constraints that are sensitive to discourse structure and temporal structure of FRCs, and answering the second question involves establishing a new affective function for DEM ku and articulating how such a pragmatic meaning may be encoded by grammar.

5.1 Application to Korean: Some Positive Outcomes

5.1

161

Application of the Proposed Analysis to Korean: Some Positive Outcomes

As we observed in Chap. 3, in Korean, when an adjectival N that forms a compound N with another N co-occurs with an ATT-DET or a CEK-AP, it must occur following the ATT-DET or the CEK-AP, as illustrated in (3)–(4). (3) Ordering between an ATT-DET and an adjectival N: a. say mohyeng catongcha (ATT-DET > Adjectival N) new miniature car ‘a new miniature car’ b. *mohyeng(,) say catongcha (Adjectival N > ATT-DET) (4) Ordering between an attributive CEK-AP and an adjectival N: a. cen-kuntay-cek mokcay kenmwul (CEK-AP > Adjectival N) pre-modern.time-CEK wooden architecture ‘a pre-modern wooden architecture’ b. *mokcay(,) cen-kuntay-cek kenmwul (adjectival N > CEK-AP)

In our current system, material forming a compound N with a root N merges at [Spec, √P], and non-gradable nominal APs denoting properties of individuals merge at [Spec, nP]. Therefore, the adjectival Ns and the ATT-DET/CEK-AP in (3) and (4) are base-generated at [Spec, √P] and [Spec, nP], respectively, as schematically given in (5). (Here and below, for expository simplicity, FPs not directly relevant to the discussion at hand will be omitted in the tree diagrams.) In light of (5), (3b) and (4b) are predicted to be ungrammatical in contrast to (3a) and (4a) because an adjectival N merged inside a √P cannot move out of it, so it is bound to surface following an ATT-DET or a CEK-AP if there are any. (5) Position of the lexical APs in (3) and (4): [DPp [FocP [LocP [SortP [nP ATT-DET/CEK-AP [

P

Adjectival N]]]]]]

Assuming (5) as the underlying structure for data like (3) and (4) also lets us account for why a thematic AP may occur between an ATT-DET/CEK-AP and an adjectival N but only if there is an intonational break after the ATT-DET/CEK-AP. To see this, consider (6)–(7). (6) DP instantiating ‘ATT-DET > Thematic AP > Adjectival N > Head N’: ?say*(,) tokil-cey [mohyeng catongcha] new German-made [miniature car] ‘a new German-made miniature car’

162

5 Capturing the Korean Facts

(7) DP instantiating ‘CEK-AP > Thematic AP > Adjectival N > Head N’: cen-kuntay-cek*(,) namhan-socay [mokcay kenmwul] South.Korea-located [wooden architecture] pre-modern.time-CEK ‘a pre-modern South-Korea located wooden architecture’

Under the present analysis, nP hosts N modifiers which receive a thematic role from the head N but which do not form a compound N with it. Therefore, thematic modifiers like tokilcey ‘German-made’ and namhansocay ‘South-Korea located’ merge at the Spec of nP that is already present in the DP structure, and as a result, the co-occurring ATT-DET or CEK-AP merges at an adjoined [Spec, nP] which is created atop the pre-existing [Spec, nP]. Given our assumptions laid out in Sect. 4.3.2, the adjunction structure created by the merge of the ATT-DET or CEK-AP will be phonologically realized as a pause. Consequently, an intonational break is obligatory after the ATT-DET or CEK-AP in contexts like (6)–(7). (8) Position of the APs in (6) and (7): [DPp [FocP [SortP [nP ATT-DET/CEK-AP [nP Thematic AP [ P Adjectival N]]]]]]

The present analysis further captures the fact that, when bearing a focal stress, a thematic AP may occur preceding an ATT-DET or an attributive CEK-AP, as exemplified in (9)–(10), although ordinarily it cannot, as shown in (6)–(7). Such data are predicted to be attested because any AP may occur at [Spec, FocP] if that position is available, as sketched in (11). Notably, in cases like (9)–(10), the pause that would have been present after the ATT-DET or CEK-AP in cases like (6)–(7) is not readily palpable because the focus movement of the thematic AP presumably takes it with it or absorbs it. (9) Variant of (6) with a FOC on the thematic AP: TOKIL-CEY say [mohyeng catongcha] German-made new [miniature car] ‘a GERMAN-MADE new miniature car’ (10) Variant of (7) with a FOC on the thematic AP: NAMHAN-SOCAY cen-kuntay-cek [mokcay kenmwul] [wooden architecture] South.Korea-located pre-modern.time-CEK ‘a SOUTH-KOREA-LOCATED pre-modern wooden architecture’ (11) Derivation of (9) under the present analysis: [DPd/r [DPp [FocP TOKIL-CEY [SortP [nP say [nP __ [

P

mohyeng catongcha]]]]]]]

On the present analysis, not every thematic AP has the same syntactic status, however: for example, while APs bearing Location or Source roles merge at [Spec, nP] as shown in (8), APs bearing Theme roles are assumed to merge at [Spec, √P] (see Sect. 4.3.3). Making such a positional differentiation between different types of

5.1 Application to Korean: Some Positive Outcomes

163

thematic APs lets us capture the ordering restrictions on certain CEK-APs. By way of illustration, in (12), the thematic CEK-AP kyengceycek ‘of economy’ cannot occur before the attributive CEK-AP cemcincek ‘gradual’ even if it bears a focal stress, and in our framework, this fixed order arises because kyengceycek receives an internal theta role from the de-verbal N sengcang, so it is base-generated at [Spec, √P] and this blocks it from moving to a higher position, just like the adjectival Ns in (3)–(4), as schematized in (13). (12) Ordering between an attributive CEK-AP and a thematic CEK-AP: a. Datum instantiating ‘CEK-APATT > CEK-AP[+ ]’: cemcin-cek kyengcey-cek sengcang slow.progress-CEK economy-CEK growth ‘a gradual growth of economy’ b. Ungrammaticality of a ‘CEK-AP[+ ] with FOC > CEK-APATT’ order: *KYENGCEY-CEK(,) cemcin-cek sengcang Intended: ‘a gradual growth of ECONOMY’

(13) Reason for the ungrammaticality of (12b): [DPp [FocP KYENGGEY-CEK [LocP [SortP [nP cemcin-cek [

P

__ sengcang]]]]]]

The decompositional derivational analysis advanced here can also yield desirable results for more complex data containing RCs, DEMs, and/or adnominal APs. To see this, consider first the definite and referential plural DP given in (14), which contains a non-restrictive RC, the DEM ku ‘the/that’, a sortal property denoting UN-AP, and an attributive CEK-AP. (14) A plural DP containing an NRS-FRC, a DEM, an UN-AP, and a CEK-APATT: [[RC ei kyengcey-rul sal-li-]-l*(,) [Dem ku] [[ __ economy-ACC live-CAUS-]-REL.FUT [ that] [CEK-AP hapli-cek] [cengchayk-tuli]] [UN-AP sayrop-un] [ new-ADN] [ rational-CEK] [policy-PL]] ‘the/those new rational policies which will boost our/the economy’

Under our analysis, this DP is derived as follows: at the initial stage of the derivation, the head N and the CEK-AP are base-generated inside √P and nP, respectively, and being a gradable AP, the UN-AP is base-generated at [Spec, SortP]. Additionally, the plural marker -tul is base-generated under the head position of PlP, and the DEM ku merges at the Spec of LocP whose head hosts the same set of features as Dd/r, namely, [+REF; +DEF; +DEIC; +SPEC]. At the next stage, the SortP raises to [Spec, PlP+], resulting in the plural marker suffixed to the head N. In addition, the DemP raises to [Spec, DPd/r], thereby licensing the relevant D features as well as complying with PELI (see Chap. 4, Sect. 4.3.5). Finally, the non-restrictive FRC merges at the [Spec, DPd/r] that is created right above the

164

5 Capturing the Korean Facts

pre-existing DPd/r. Apart from this, no changes are made to the underlying structure. Consequently, except for the plural marker, all N modifiers including the RC surface preceding the head N. (15) Derivation of (14): ‘NRS-RC > DEM > UN-AP > CEK-AP > N + PL’ DPd/r RC

DPd/r Dd/r

kyengceyrul sallil Dd/r [+REF] [+DEF] [+DEIC] [+SPEC]

LocP DemP

ku

Loc Loc [+REF] [+DEF] [+DEIC] [+SPEC]

PlP+ PlP Pl [+PL] -tul

SortP AP

nP AP

P

haplicek

cengchayk

sayropun

The derivation given in (15) enables us to explain why the RC in (14) is followed by an obligatory pause; in this configuration, the RC occupies an adjoined position, reminiscent of the ATT-DET or the CEK-AP in (6)–(7). It also captures the fact that despite its morpho-syntactic complexity relative to a CEK-AP, the UN-AP sayropun is construed as carrying a direct N modificational meaning, contrary to what would be predicted by a Cinquerian (Cinque 2010) analysis (e.g., Kang 2005, 2006); under our analysis, this apparent mismatch between the morpho-syntactic complexity and the possible semantic construal for an N modifier is expected because any kind of sortal AP may occur occupying [Spec, SortP]. Our derivational system can readily handle Korean nominal data involving NUMs and CLs as well. In particular, it provides a systematic account for how and/ or why the surface constituent order variation may correlate with truth-conditional meaning differences. To see this, consider the two sentences in (16), focusing on their object DPs. These two DPs have different constituent orders and they also have slightly different meanings: the object DP in (16a) is non-specific whereas the one in (16b) is specific. This is evidenced by the fact that (16a) can make a good answer to the question ‘What did you do yesterday?’ but (16b) cannot, as exemplified in (17); on the other hand, while (16b) can make a good answer to ‘How many linguistics students among the ones we know did you meet?’, (16a) cannot, as

5.1 Application to Korean: Some Positive Outcomes

165

illustrated in (18). These semantic differences suggest that while the RC in (16a) is construed as restrictive, the one in (16b) can be construed as non-restrictive; that is, (16b) can be uttered in a context where the property described by the RC is already established in the discourse, so it does not reduce the size of the set denoted by the head N. What is also notable is that the object DPs in (16a) and (16b) have different phonological contours: the RC in (16a) has to be followed by a pause but the one in (16b) cannot be. Conversely, while there is no pause after the NumP in (16a), there is one after the NumP in (16b).

1

(16) Word order variation and truth-conditional meaning difference: a. ‘RC > NUM + CL > N + PL’ order and a non-specific interpretation: Na-nun [DP [RC ei enehak-ul cenkongha-n-]-un*(,) I- TOP [ [ __ linguistics-ACC major.in-IMPRF-]-REL [ NumP sey-myeng-uy1] haksayngi-tul]-ul manna-ss-ta. [ three- CL-GEN] student-PL]-ACC meet-PST-DECL ‘I met with three non-specific students who major in linguistics.’ b. ‘NUM + CL > RC > N + PL’ order and a specific interpretation: Na-nun [DP [NumP sey-myeng-uy]*(,) [RC ei enehak-ul cenkongha-n-]-un haksayngi-tul]-ul manna-ss-ta. ‘I met with three specific/particular students who major in linguistics.’ (17) Appropriate discourse context for (16a): ecey mwues-ul hay-ss-ni? A: Ne-nun You-TOP yesterday what-ACC do-PST-Q.INFRML ‘What did you do yesterday? B: Na-nun ecey [DP [RC ei enehak-ul cenkongha-n-]-un*(,) I-TOP yesterday [ [ __ linguistics-ACC major.in-IMPRF-]-REL [NumP sey-myeng-uy] haksayngi-tul]-ul manna-ss-e. student-PL]-ACC meet-PST-DECL.INFRML [ three-CL-GEN] ‘Yesterday I met with three non-specific students who major in linguistics.’ B :#Na-nun ecey [DP [NumP sey-myeng-uy]*(,) enehak-ul cenkongha-n-]-un haksayngi-tul]-ul [RC ei manna-ss-e. ‘Yesterday I met with three specific/particular students who major in linguistics.’

Following standard practice, I am glossing morpheme -uy which occurs on a ‘NUM + CL’ sequence as a genitive marker here. But I assume that it is actually a dummy element. Therefore, I do not represent it in the syntactic trees here and below. One reason to think that such occurrences of -uy are dummy elements is that their meaning has nothing to do with possession. Furthermore, the apparent GEN marker does not show up when the same ‘NUM + CL’ sequence occurs DP-finally, as shown in (20b). That said, to keep things manageable, I do not further discuss the status of -uy here. Instead, I refer the interested reader to An 2014, wherein it is claimed that some occurrences of -uy is a dummy agreement marker.

1

166

5 Capturing the Korean Facts

(18) Appropriate discourse context for (16b): enehak-ul cenkongha-n-]-un [DP [RC ei A: Ne-nun You- TOP [ [ __ linguistics-ACC major.in-IMPRF-]-REL haksayngi-tul]-ul myech-myeng manna-ss-ni? how.many-CL meet-PST-Q.INFRML student-PL]-ACC ‘How many linguistics students did you meet?’ enehak-ul cenkongha-n-]-un*(,) [DP [RC ei B: #Na-nun [ [ __ linguistics-ACC major.in-IMPRF-]-REL I- TOP [ NumP sey-myeng-uy] haksayngi-tul]-ul manna-ss-e. [ three- CL-GEN] student-PL]-ACC meet-PST-DECL.INFRML ‘I met with three non-specific students who major in linguistics.’ B : Na-nun [DP [NumP sey-myeng-uy]*(,) [RC ei enehak-ul cenkongha-n-]-un haksayng i-tul]-ul manna-ss-e. ‘I met with three specific/particular students who major in linguistics.’

Under the tri-partite DP analysis promoted here, the semantic and phonological differences between (16a) and (16b) fall out naturally from the differing surface positions of the NumPs as well as the differing merge sites of the RCs, which in fact stems from the different nominal statuses of the two object DPs: The object DP in (16a) instantiates what I call a DPq and the one in (16b) instantiates what I call a DPd/r and they are derived as in (19a) and (19b), respectively. These derivations show that in both cases, the NumP is base-generated at [Spec, UnitP], but while it moves to [Spec, DPq] in (16a), thereby licensing the [+QUANT] under Dq (as well as fulfilling PELI), it occurs at an adjoined [Spec, DPd/r] in (16b) because the pre-existing [Spec, DPd/r] is already occupied by the non-restrictive RC.2 Consequently, while the NumP in (16a) surfaces DP-medially, the one in (16b) surfaces DP-initially. Furthermore, only the latter NumP occurs followed by an intonational break which is created by an adjunction structure. Finally, in (16a), the RC occupies an adjoined position, so it is followed by a pause; on the other hand, the RC in (16b) does not occupy an adjoined position, so it is not accompanied by a pause. (Note: Here, a strike-through on a feature in a tree diagram indicates feature licensing or valuation.)

2

I posit that the RC here is base-generated at the pre-existing [Spec, DPd/r] rather than at an adjoined position because, as stated in Chap. 4 (Sect. 4.3.3), I assume that in article-less languages, PELI is satisfied for DPd/r if a locative element (e.g., a DEM) can pronounce it within its projection, but if not, then a non-restrictive RC may pronounce it by occupying its Spec position. In the case of (14), the DP at hand contains a DEM. Therefore, the DPd/r is pronounced by the DEM, and this makes the RC adjoined to DPd/r. However, that is not the case with (16b) and hence a slightly different merge site for its RC.

5.1 Application to Korean: Some Positive Outcomes

(19)

167

a. Derivation of the object DP in (16a): ‘RS-RC > NUM + CL > N + PL’ DPq RC

DPq NumPi

Dq

enehakul cenkonghanun [+QUANT] Dq seymyeng [+QUANT]

UnitP ti

PlP+ SortP2

haksayng

PlP Pl -tul

t2

b. Derivation of the object DP in (16b): ‘NUM + CL > NRS-RC > N + PL’ DPd/r DPd/r RC Dd/r enehakul cenkonghanun [+REF] [-DEF] [-DEIC] [+SPEC]

Dd/r UnitP NumP

PlP+

seymyeng haksayng-tul

Consider now the plural definite DPs in (20a, b). These DPs exhibit rather different constituent orders but they are truth-conditionally identical in that both may be felicitously uttered only when the discourse participants are already aware

168

5 Capturing the Korean Facts

that Mina was helped by three specific individuals.3 Moreover, in both cases, the NumP can carry a focal stress.4 (20) Word order variation for a plural, definite DP with a NUM and a CL: a. ‘RC > DEM > NumP > UN-AP > N + PL’ order: [DP [RC ei Mina-rul towacwu- ]-n*(,) [DP ku [ that [ [ __ M.-ACC help-PRF]-REL haksayng-tul]i] [NumP sey-myeng-uy]*(,) [AP khikhu-un] [ tall-ADN] student-PL]] [ three-CL-GEN] ‘those three tall students who helped Mina’ b. ‘RC > DEM > UN-AP > N + PL > NumP’ order: [DP ku*(,) [DP [RC ei Mina-rul towacwu- ]-n*(,) haksayng-tuli [NumP sey-myeng]] [AP khikhun] ‘those three tall students who helped Mina’

Under the present analysis, the two DPs in (20) have no truth-conditional differences but they have different surface constituent orders for the following reasons: in both cases, the RC merges at an adjoined [Spec, DPd/r] and the NumP raises from [Spec, UnitP] to [Spec, FocP]. But while only the DemP raises to [Spec, DPd/r] to license the relevant D features in (20a), in (20b), the entire LocP containing the trace of the moved NumP raises to [Spec, DPd/r], instantiating a type of remnant movement that is reminiscent of the Spanish cases discussed in Chap. 4. To see this, compare (21a) and (21b). (Note that again, I treat the apparent genitive marker -uy on seymyeng in (20a) as a dummy element; see Footnote 1.)

For example, they can both occur in contexts where the speaker wants to find out the names of the three students who helped Mina, as shown in (i), and since wh-questions presuppose the existence of the individuals in the denotation of the wh-constituent at hand, these DPs can be said to denote discourse-old individuals.

3

(i) a.

b.

4

Possible discourse context for (20a): towacwu- ]-n, [DP ku [NumP sey-myeng-uy], [DP [RC ei Mina-rul help-PRF]-REL [ that [ three-CL-GEN] [ [ __ M.-ACC haksayng-tul]i]-un nwukwu-i-ni? [AP khikhu-un] student-PL]]-TOP who-COP-Q.INFRML [ tall-ADN] ‘Who are those three students who helped Mina? Possible discourse context for (20b): [DP [RC ei Mina-rul towacwu- ]-n, [DP ku, [AP khikhun] [NumP sey-myeng]]-un nwukwu-i-ni? haksayng-tuli ‘Who are those three students who helped Mina?

In fact, sentences (20a, b) can be most naturally uttered with a focal stress on seymyeng even though they do not have to carry a contrastive-focus meaning. And according to my intuition, the presence of seymyeng in these sentences implicates that the fact that the number of students who helped Mary is three is somehow noteworthy.

5.1 Application to Korean: Some Positive Outcomes

(21)

169

a. Derivation of (20a): ‘NRS-FRC > DEM > NUM + CL > UN-AP > N + PL’ DPd/r RC

DPd/r Dd/r

Minarul towacun Dd/r FocP [+REF] LocP [+DEF] NumPi [+DEIC] [+SPEC] DemP UnitP seymyeng ti ku

PlP+

khikhun haksayng-tul b. Derivation of (20b): ‘NRS-FRC > DEM > UN-AP > N + PL > NUM + CL’ DPd/r RC

DPd/r Dd/r

Minarul towacun Dd/r [+REF] [+DEF] [+DEIC] [+SPEC]

FocP NumPi

LocP DemP

UnitP

seymyeng ti

PlP+

ku khikhun haksayng-tul

The derivations given in (21a, b) not only yield the correct surface orders for the DPs in (20) but also provide us with a way to account for their phonological contours: in both cases, the RCs occur followed by an obligatory pause because they merge in an adjoined position, namely, the [Spec, DPd/r] that is created atop the pre-existing [Spec, DPd/r].5 Apart from this, however, the two DPs are pronounced rather differently. First, while there is no pause after the DEM ku in (20a), there is one after the DEM in (20b), and this is because the trace of the raised NumP is present inside the ‘DEM + AP’ string and its presence is phonologically “audible” in the form of a short pause. Secondly, the NumP in (20a) is followed by a pause

5

And given what is stated in Footnote 2, the RCs merge at an adjoined position here because the DP already contains a DEM, unlike the case with (16b), but resembling cases like (14).

170

5 Capturing the Korean Facts

because there are two traces of moved items after it, namely, the trace of the moved DemP and the trace of the moved NumP; in the case of (20b), we do not detect a pause after the NumP even though it is immediately followed by the trace of the moved LocP because it surfaces in a DP-final position. Applying the present analysis to data like (22)–(25) also yields positive results in that it lets us explain some of the puzzling phenomena pertaining to the relative orders between certain ATT-DETs and UN-APs which we identified in Chap. 3. More specifically, we can now account for why ATT-DETs like hyen ‘present’ may occur preceding an UN-AP whereas ATT-DETs like say ‘new’ or yec ‘old’ cannot, and how their contrastive behavior may correlate with their hosting DPs having uniquely-referring semantics or not. When the hosting DP is non-uniquely referring: kapang (22) a. khetarah-n say large-ADN new bag ‘a new large bag’ b. *say, khetarah-n kapang Intended: ‘a new large bag’ (23) a. kuriwun yec chinkwu old friend much.missed- ADN ‘a much-missed old friend’ b. *yec, kuriwun chinkwu Intended: ‘a much-missed old friend’ When the hosting DP is uniquely referring: (24) a. pwuphayha-n, hyen cengkwon corrupt-ADN present government ‘the corrupt present government’ b. hyen, pwuphayha-n cengkwon ‘the present corrupt government’

(UN-AP > ATT-DET) (ATT-DET > UN-AP)

(UN-AP > ATT-DET) (ATT-DET > UN-AP)

(UN-AP > ATT-DET) (ATT-DET > UN-AP)

(25) a. sengkongceki-n, hyen kyengcey cengchayk (UN-AP > ATT-DET) economy policy successful-ADN present ‘the successful current economic policy’ b. hyen, sengkongceki-n kyengcey cengchayk (ATT-DET > UN-AP) ‘the current successful economic policy’

Under our decompositional split-DP analysis, the contrast between (22)/(23) and (24)/(25) obtains because the ATT-DETs occurring in them have different merge sites: say and yec are attributive nominal APs, so they merge at [Spec, nP], as we assumed for cases like (6)–(7). On the other hand, hyen is inherently non-attributive and furthermore, it carries a built-in [+DEF] semantics. Therefore, it merges at [Spec, LocP], which is where DEMs are base-generated.

5.1 Application to Korean: Some Positive Outcomes

171

Support for this analysis of the ATT-DETs at hand comes from two sources. One is that when occurring by itself in an out-of-the-blue context, the string of words comprised of hyen and an N may only be construed as ‘the present/current N’ (i.e., a definite nominal), as indicated by the English translations for (24)–(25); in contrast, when occurring in isolation and in an out-of-the-blue context, the string of words comprised of say/yec and an N may only be construed as ‘a new/old N’ (i.e., an indefinite nominal), as indicated by the English translations for (22)–(23). The other supporting evidence comes from the fact that while say or yec can never take scope over a DEM, hyen can, and for this reason, (26d) is ungrammatical whereas (27d) is not: (27d) can be uttered in a context where the speaker is contrasting two different policies of a similar kind, one of which was instituted sometime in the past and the other of which is in effect now. Notably, when hyen occurs following ku as in (27c), it is under the scope of the DEM, and such a nominal can be uttered in a context where there are several policies being used at the present time which both the speaker and the hearer are familiar with, and the speaker is talking about some particular one among them. (26) a. say cengchayk new policy ‘a new policy’ b. ku cengchayk that policy ‘the/that policy’ c. ku say cengchayk that new policy ‘the/that new policy’ d. *say, ku cengchayk Intended: ‘the/that policy, which is new’ (27) a. hyen cengchayk present policy ‘the present policy’ b. ku cengchayk that policy ‘the/that policy’ c. ku, hyen cengchayk that present policy ‘that present policy’ d. hyen, ku cengchayk ‘the present policy of that kind’

(DEM > ATT-DET) (ATT-DET > DEM)

(DEM > ATT-DET) (ATT-DET > DEM)

On the present analysis, then, hyen and ku have the same syntax even though one is categorially classified as an ATT-DET and the other is classified as a DEM. And this means that (27a) and (27b) have the same underlying structure except that what occupies [Spec, LocP] in (27a) is hyen but it is ku in (27b), as depicted in (28).

172

(28)

5 Capturing the Korean Facts

Derivation of (27a, b): ‘hyen/DEM > N’ [DPd/r __ [FocP [LocP hyen/ku [Loc [Loc [+REF;+DEIC;+DEF;+SPEC] [UnitP [PlP+ cengchayk]]]]]]]

Similarly, (27c) and (27d) have essentially the same underlying structures as (27a, b) except that in such cases, either hyen or ku is base-generated at [Spec, LocP] and then raises to [Spec, DPd/r] and after this, either ku or hyen merges at an adjoined [Spec, DPd/r] position that is created atop the pre-existing [Spec, DPd/r] position, as given in (29) and (30), respectively. The resulting configurations capture the fact that ku in (27c) and hyen in (27d) are both pronounced with a pause after them; such a phonological contour obtains because they occur in adjoined positions. (29)

Derivation of (27c): ‘DEM > hyen > N’ [DPd/r ku [DPd/r hyeni [FocP [LocP ti [Loc [+REF;+DEIC;+DEF;+SPEC] [UnitP [PlP+ cengchayk]]]]]]]

(30)

Derivation of (27d): ‘hyen > DEM > N’ [DPd/r hyen [DPd/r kui [FocP [LocP ti [Loc [+REF;+DEIC;+DEF;+SPEC] [UnitP [PlP+ cengchayk]]]]]]]

In sum, then, in all cases, hyen can occur in a position that a DemP can occur whereas the other ATT-DETs merge at [Spec, nP], which is lower than where a DemP can be base-generated (see tree (11), for example). As a corollary, while ATT-DETs like say are always construed as being under the scope of a DEM if they co-occur, hyen may take scope over a DEM, as exemplified in (27d). Notably, the special syntactic and semantic status of hyen we argue for here lets us explain why it may occur preceding an UN-AP, as shown in (24b) and (25b), even though that would violate constraints given in (1a) and (1c), namely, ‘Relatively Heavy Modifier > Relatively Light Modifier > N’ and ‘Verbal Modifier > Nominal Modifier > N’. Furthermore, it lets us capture how and why the (a) and the (b) data in (24)–(25) differ from each other. In cases like (24a) and (25a), the ATT-DET is base-generated at [Spec, LocP] and then moves to the pre-existing [Spec, DPd/r] to license the relevant D features as well as to fulfill PELI. After this, an UN-AP merges at an adjoined [Spec, DPd/r] that is created above the surface position of hyen, as depicted in (31). Consequently, the two ADJs occur with a pause between them, and they are also construed as standing in an appositive relation to each other. As for data like (24b) and (25b), we can in fact derive them from what is given in (31) by scrambling hyen across the UN-AP, as shown in (32). And this agrees with native Korean speakers’ intuition that, unlike (24a) and (25a), (24b) and (25b) require a non-neutral discourse context to be judged felicitous, namely, one in which the meaning of hyen is being foregrounded.

5.1 Application to Korean: Some Positive Outcomes

(31)

173

Derivation of (24a): ‘UN-AP > ATT-DET > N’ surface order DPd/r AP

DPd/r Dd/r

pwuphayhan Dd/r [+REF] [+DEF] [+DEIC] [+SPEC]

FocP LocP AP

UnitP PlP+

hyen cengkwon (32)

Derivation of (24b): ‘ATT-DET > UN-AP > N’ surface order [DPd/r __ [DPd/r pwuphayhan [DPd/r hyen [FocP [LocP [UnitP [PlP+ cengkwon]]]]]]]

5.2

Challenges for a Derivational Analysis

The previous section has shown that the proposed derivational analysis has a positive prospect for capturing the syntax of N modifiers and their ordering restrictions in a relatively understudied, head-final, article-less, and scrambling language. Under scrutiny, however, it turns out that there are several empirical challenges confronting it and they are coming from the behavior of UN-APs and FRCs in Korean.

5.2.1

Challenge 1: Surface Position of Certain UN-APs Relative to Other Lexical ADJs

The first set of challenges that confront us has to do with the surface position of certain UN-APs relative to other lexical ADJs. To see this, consider (33)–(34), which are repeated from Chap. 3. These two data sets contain khetaran and khun, which are semantically synonymous but differ in length, i.e., tri-syllabic versus mono-syllabic.

174

5 Capturing the Korean Facts

(33) Ordering between a tri-syllabic UN-AP and an ATT-DET: a. khe.ta.rah-n say kapang (UN-AP > ATT-DET) large-ADN new bag Intended: ‘a large new bag’ b. KHE.TA.RAH-N say kapang (UN-AP with FOC > ATT-DET) large-ADN new bag Intended: ‘a LARGE new bag (not a small one)’ c. *say khe.ta.rah-n kapang (ATT-DET > UN-AP) new large-ADN bag Intended: ‘a new large bag’ d. *SAY khe.ta.rah-n kapang (ATT-DET with FOC > UN-AP) new large-ADN bag Intended: ‘a NEW large bag (not a small one)’ (34) Ordering between a mono-syllabic UN-AP and an ATT-DET: a.??khu-n say kapang (UN-AP > ATT-DET) large-ADN new bag Intended: ‘a large new bag’ b. KHU-N say kapang (UN-AP with FOC > ATT-DET) large-ADN new bag Intended: ‘a LARGE new bag (not a small one)’ c. *say khu-n kapang (ATT-DET > UN-AP) new large-ADN bag Intended: ‘a new large bag’ d.??SAY khu-n kapang (ATT-DET with FOC > UN-AP) new large-ADN bag Intended: ‘a NEW large bag (not an old one)’

The comparison between (33a, b) and (34a, b) shows that while a tri-syllabic UN-AP may freely occur before an ATT-DET, a mono-syllabic one cannot unless it bears a focal stress. And at the moment, we have no way of forbidding a mono-syllabic UN-AP to occur before an ATT-DET (while permitting a tri-syllabic one to do so) since, being a gradable ADJ, khun is expected to merge at [Spec, SortP], so it would not be competing for the same position as say, which would occur at [Spec, nP]. In addition to the contrast between (33a, b) and (34a, b), the grammaticality judgments on (33d) and (34d) are problematic for us since, under the present analysis, any focused AP should be able to occur preceding an UN-AP by moving to [Spec, FocP], so the two examples are predicted to be equally grammatical, contrary to fact. Furthermore, if an ATT-DET cannot occur preceding an UN-AP for some reason, then both (33d) and (34d) should be judged unacceptable but (34d) is actually judged better than (33d) (though it is only marginally acceptable) and currently, we do not have an account of such gradient grammaticality judgments. In the face of data like (33c, d) and (34c, d), one cannot simply resort to an ad hoc rule which dictates that when two APs co-occur, the lighter one must occur closer to the head N, since doing so will run afoul of data like (35).

5.2 Challenges for a Derivational Analysis

(35) Relative ordering between two UN-APs: ppal.kah-n kapang a. khu-n, red-ADN bag big-ADN, ‘a big, red bag’ b.??ppal.kah-n, khu-n kapang Intended: ‘a red, big bag’

175

(size > color)

(color > size)

Moreover, as we already observed in Chap. 3, the relative ordering between two UN-APs is governed by more than one factor: when two UN-APs co-occur inside the same DP, their ordering is subject to the typologically well-established semantic constraint given in (36), and this is in fact what the contrast between (35a) and (35b) illustrates. (36)

Typological generalization on the ordering between simplex APs: Quality > Size > Shape > Color > Provenance/Source

However, this AOR is not always respected: when two UN-APs co-occur and if only one of them has a degree modifier, then that UN-AP may occur preceding the other one even if the outcome does not fit the above generalization. To see this, consider (37). Here, the (a) example is judged better than the (b) example even though it instantiates ‘color > size’ in violation of (36).6 (37) Relative ordering between an UN-AP with a degree modifier and one without: a. [acwu ppalkah-n], khu-n kapang (color > size) big-ADN bag [very red-ADN] ‘a very red, big bag.’ b.??khu-n, [acwu ppalkah-n] kapang (size > color) Intended: ‘a big, very red bag.’

Taken together, this set of facts shows that different “forces” are at work in linearizing lexical ADJs in Korean, and some of them are of semantic nature [e.g., (36)] while others are of phonological nature [e.g., (1a)]. Obviously, just writing a few ad hoc rules and applying them blindly will not work for our purposes. Furthermore, doing so will not let us capture the gradient grammaticality that data on adjective ordering in Korean are shown to exhibit. Given this, the question that arises is how to account for the AOR phenomena in Korean while maintaining the derivational analysis proposed in Chap. 4 and assuming what is given in (2) as the universal DP structure.

6 In answer to ‘What did you buy yesterday?’, a Korean speaker can utter (37a) but not (37b), and this shows that between the two examples, the former instantiates the more canonical constituent order.

176

5.2.2

5 Capturing the Korean Facts

Challenge 2: Ordering Between RCs

In Chap. 3, we observed that when two RCs co-occur modifying the same DP in Korean, they do not occur in any random order. In this context, we also got to see that one cannot account for the relevant facts by simply dictating that in Korean (and in some other languages), RCs with an S-level predicate occur in NP-outer space in the sense of Larson (1998, 2000) and RCs with an I-level predicate occur in NP-inner space, unlike what authors like Larson and Takahashi (2007) have claimed. In this section, I would like to point out that data instantiating ordering restrictions on Korean RCs actually exhibit a more gradient grammaticality than what has been acknowledged in the literature, and in this regard, they are analogous to what we saw with data involving lexical Korean APs in Sect. 5.2.1. To illustrate this, let us first take a look at paradigms (38)–(42), which are repeated from Chap. 3 though not in the same order. In these paradigms, all the (a) examples are judged good but the (b) examples have varying degrees of ungrammaticality. What is particularly noteworthy is that, unlike the (b) examples in (38) through (41), (42b) is judged not so bad, and depending on context, it may be judged perfectly grammatical and/or even better than (42a) (more on this in Sect. 5.3.4). (38) a. [RC1 nay-ka ei ecey manna- -n], __ yesterday meet-PRF-REL] [ I-NOM phyengso tampay-rul manhi phiwu-n-un] [RC2 ei a.lot smoke-IMPRF-REL] [ __ usually cigarette-ACC Chelswu-i-ta. sarami-un person-TOP C.-COP-DECL ‘The person that I met yesterday who smokes a lot is Chelswu.’ (RC1 > RC2) b.??[RC2 ei phyengso tampay-rul manhi phiwu-n-un], ei ecey manna- -n] sarami-un Chelswu-i- -ta. [RC1 nay-ka Intended: ‘The person that smokes a lot who I met yesterday is Chelswu.’ (RC2 > RC1) Mina-ka yocum ei sakwi-n-]-un, (39) a. [DP [RC1 these.days __ date-IMPRF-]-REL [ [ M.-NOM [RC2 ei elkwul-i nemwuna calsayngki- ]-n namcai]-nun truly be.handsome-N.PST]-REL man]-TOP [ __ face-NOM (RC1 > RC2) Cinho-i- -ta. C.-COP-N.PST-DECL ‘The man that Mina is dating these days who is so handsome is Cinho.’ elkwul-i nemwuna calsayngki- ]-n, b. *[DP [RC2 ei ei sakwi-n-]-un, namcai]-nun [RC1 Mina-ka yocum (RC2 > RC1) Cinho-i- -ta. Intended: ‘The man who is so handsome that Mina is dating these days is Cinho.’

5.2 Challenges for a Derivational Analysis

177

(40) a. [DP [RC1 Minaj-ka ei kacang cal kiekha-n-]-un, most well remember-IMPRF-]-REL, [ [ M.-NOM __ [RC2 ei kwusipnyentay-chopan-ey iss-ess-te-]-n exist-ANT-RTRO-]-REL [ __ 1990’s-early-LOC [proj emma]-hako-uy yehayng-i- -ta. (RC1 > RC2) ili]-un [__ Mom]-with-GEN trip-COP-N.PST-DECL event]-TOP ‘The event that Mina remembers the best which happened in the early 1990s is the trip with her mom.’ b. *[DP [RC2 ei kwusipnyentay-chopan-ey iss-ess-te-]-n, kacang cal kiekha-n-]-un ili]-un [RC1 Minaj-ka ei emma]-hako-uy yehayng-i- -ta. (RC2 > RC1) [proj Intended: ‘The event that happened in the early 1990s which Mina remembers the best is the trip with her mom.’ ei i seysang-eyse kacang (41) a. [DP [RC1 Mina-ka __ this world-LOC most [ [ M.-TOP maumssi-ka nemwuna chakha- ]-n cohaha-n-]-un, [RC2 ei heart-NOM really be.good-N.PST]-REL like-IMPRF-]-REL [ __ chinkwui]-nun Cinhi-i- -ta. (RC1 > RC2) C.-COP-N.PST-DECL friend]-TOP ‘The friend that Mina likes the most in this world who has a truly beautiful heart is Cinhi.’ maumssi-ka nemwuna chakha- ]-n, b.??[DP [RC2 ei i seysang-eyse kacang cohaha-n-]-un [RC1 Mina-ka ei Cinhi-i- -ta. (RC2 > RC1) chinkwui]-nun Intended: ‘The friend who has a truly beautiful heart who Mina likes the most in this world is Cinhi.’ (42) a. [DP [RC1 Mina-ka phyengso ei culkye-pwuru-n-]-un, [ [ M.-NOM usually __ enjoy-singing-IMPRF-]-REL, chilsipnyentay-ey ywuhaynghay-ss-te-]-n [RC2 ei be.popular-ANT-RTRO-]-REL [ __ 70’s-LOC (RC1 > RC2) norayi]-nun leylitpi-i- -ta. let.it.be-COP-N.PST-DECL song-TOP ‘The song that Mina likes to sing which used to be popular in the 1970s is Let It Be.’ b.?/ [DP [RC2 ei chilsipnyentay-ey ywuhaynghay-ss-te-]-n, Mina-ka phyengso ei culkye-pwuru-n-]-un [RC1 leylitpi-i- -ta. (RC2 > RC1) norayi]-nun Intended: ‘The song that used to be popular in the 1970s which Mina likes to sing is Let It Be.’

Notably, in the above data sets, the grammatical examples have the template of an RC with an object gap followed by an RC with a subject gap.7 But even if we look at cases where the co-occurring RCs contain different types of gap, similar

7

This was noted by an anonymous reviewer, who raised the possibility that the grammaticality or ungrammaticality of data containing RCs may be impacted by what type of gap the RCs contain.

178

5 Capturing the Korean Facts

issues arise. To see this, consider (43)–(46). In these examples, the RCs contain the same type of gap; that is, they either contain only subject gaps or object gaps. Yet they do not occur in any random order. What is also worth noting is that even though (43) seems to fit the descriptive generalization that Larson and Takahashi (2007) offer in that its grammatical datum instantiates ‘RCI-level > RCS-level’ order, the data given in (44) through (46) do not; these three data sets contain RCs with S-level predicates, so according to Larson and Takahashi, they should be able to occur in any order relative to each other, but that is not what we see here. (43) Data containing an RC with an I-level predicate and an RC with an S-level predicate: a. [DP [RC1 ei yocumtule tampay-rul manhi phiwu-n-]-un, [ [ __ these.days cigarette-ACC a.lot smoke-IMPRF-]-REL, ei khi-ka maywu khu- ]-n [RC2 [ __ height-NOM very big-N.PST]-REL Cinho-i- -ta. sarami]-un person]-TOP C.-COP-N.PST-DECL ‘The person who smokes a lot these days who is very tall is Cinho.’ (RC1 > RC2) b. *[DP [RC2 ei khi-ka maywu khu- ]-n, [RC1 ei yocumtule tampay-rul manhi phiwu-n-un] sarami]-un Cinho-i- -ta. Intended: ‘The person who is very tall who smokes a lot these days is Cinho.’ (RC2 > RC1) Data containing RCs with S-level predicates and their ordering restrictions: (44) a. [DP [RC1 ei wuri hoysa-ey cinan cwu-ey ipsaha- ]-n, week-LOC enter-ANT]-REL, [ [ __ our company-LOC last [RC2 ei onul kwacang-nim-ul towa- -turi- ]-n [ __ today director-HON-ACC help-CONN-give.HON-ANT]-REL cikwoni]-un Pakkinam-ssi-i- -ta. employee]-TOP P.-Mr.-COP-N.PST-DECL ‘The employee who started working in our company last week who helped my department’s director today is Mr. Kinam Park.’ (RC1 > RC2) b.??[DP [RC2 ei onul kwacang-nim-ul towa- -turi- ]-n, [RC1 ei wuri hoysa-ey cinan cwu-ey ipsaha- ]-n cikwoni]-un Pakkinam-ssi-i- -ta. Intended: ‘The employee who helped my department’s director today who started working in our company last week is Mr. Kinam Park.’ (RC2 > RC1)

5.2 Challenges for a Derivational Analysis

179

(45) a. [DP [RC1 Mina-ka ku nal ei malha- ]-n, [ [ M.-NOM that day __ mention-ANT]-REL, [RC2 nay-ka ei ecey manna- ]-n sarami]-un __ yesterday meet-ANT]-REL person]-TOP [ I-NOM Cinho-i- -ta. C.-COP-N.PST-DECL ‘The person that Mina mentioned that day who I met yesterday is Cinho.’ (RC1 > RC2) b. *[DP [RC2 nay-ka ei ecey manna- ]-n, [RC1 Mina-ka ku nal ei malha- ]-n sarami]-un Cinho-i- -ta. Intended: ‘The person that I met yesterday who Mina mentioned that day is Cinho.’ (RC2 > RC1) (46) a. [DP [RC1 Chelswu-ka ecey ei Mina-eykey sa-ta [ [ C.-NOM yesterday __ M.-to buy-CONN cwu- ]-n, [RC2 nay-ka cikum ei kac-ko give-ANT]-REL, [ I-NOM now __ have-CONN iss-n-]-un cangnankami]-un ropothu-i- -ta. exist-IMPRF-]-REL toy]-TOP robot-COP-N.PST-DECL ‘The toy that Chelswu bought and brought home for Mina yesterday which I have with me now is a robot.’ (RC1 > RC2) [RC2 nay-ka cikum ei kac-ko iss-n-]-un, b.??[DP ecey ei Mina-eykey sa-ta cwu- ]-n [RC1 Chelswu-ka cangnankami]-un ropothu-i- -ta. Intended: ‘The toy that I have with me now which Chelswu bought and brought home for Mina yesterday is a robot.’ (RC2 > RC1)

As far as I can see, the data given in (38)–(46) would present a problem to any derivational analysis of N modifiers at least for three reasons: First of all, no derivational analysis can explain how and why such data exhibit varying degrees of grammaticality. Secondly, no extant analysis offers a way to order RCs differently according to whether they contain an I-level or an S-level predicate. Thirdly, even if one devises such an analysis, they would have trouble accounting for data like (45)– (46) since they involve relative ordering of RCs of similar kinds—similar in the sense that they are all comprised of S-level predicates.

5.3

Solving the Problems by Adding an Output Filter

So, how do we go about dealing with the challenges presented by the Korean data discussed in the previous two subsections? In answer to this question, what we first need to recognize is that our derivational system slightly over-generates for Korean and therefore we need a way to filter out linguistic expressions it generates with respect to certain AOR constraints that Korean seems to be subject to. Since our goal is to offer a formal system that works for all languages, such constraints should

180

5 Capturing the Korean Facts

not be language-specific, however. Given this, what is called for is an analysis that is couched within Optimality Theory (OT) (Prince and Smolensky 1993). An OT analysis would be optimal for our purposes because OT assumes that Universal Grammar (UG) is comprised of a set of violable constraints which are ranked differently from language to language, so if we apply such constraints and their rankings to language X, then we can ensure that among all the possible linguistic forms that our derivational system generates which we can call the inputs, only the ones that satisfy the highest-ranked constraints in that language will be filtered through and surface as outputs. And since every language or grammar has different constraint rankings, we can explain why language X may behave differently than the next language Y with respect to adjective ordering or any other related phenomena. In brief, what I would like to suggest is that when it comes to AOR, what is called GEN (generator) in OT has the shape of the decompositional split-DP-based derivational system we developed in Chap. 4 and this component in UG works together with CON (the universal constraint set) and EVAL (evaluator) to produce optimal candidates for any human language.

5.3.1

Why Certain Lexical APs Surface in Such “Unexpected” Positions

I propose that adjective ordering in human language is partly governed by the workings of the four constraints given in (47). I further hypothesize that these constraints target the DPp region inside a DP and they are applied to each pair of APs occurring in that region. The rationale for thinking that they target the Low Field of a DP is that the ordering restrictions we observed in Sect. 5.2.1 only apply to pairs of lexical ADJs that occur inside a predicative nominal or those occurring inside an indefinite, non-specific DP, modifying their head N’s reference rather than the referent thereof. (47) A partial list of constraints regulating adjective ordering in human language: a. *NA > NNA: An ADJ sequence where a nominal AP (NA) occurs before a non-nominal one (NNA) inside a DPp is banned. b. DEGLFT: Every degree adverb is at the left edge of a DPp. Every AP with relative semantics is at the left edge of a DPp. c. RELLFT: d. HVYLFT: For any pair of APs occurring inside a DPp, the heavier one is at the left edge of the DPp.

I should note that applying HVYLFT to linguistic data will involve comparing the number of syllables of the two APs under evaluation and given the way it is defined here, one violation of this constraint would mean that the two APs have the same number of syllables, so the constraint is not satisfied; two violations of it would mean that the AP on the left is one syllable shorter than the one on the right; three

5.3 Solving the Problems by Adding an Output Filter

181

violations would mean that the AP on the left is two syllables shorter than the one on the right, and so on and so forth.8 I will further assume that a pause that is created by an adjunction structure that is present immediately after an AP will have the effect of adding one syllable to the AP’s length and similarly, putting a focal stress on an AP would amount to adding one syllable to the AP. Turning now to the motivations for proposing the constrains in (47), of the four constraints, invoking the alignment constraints, namely, DEGLFT, RELLFT, and HVYLFT, is inspired by Grimshaw (2001). Grimshaw captures constituent order variation across languages by postulating a set of alignment constraints such as HEADLFT (‘Every X-zero is at the left edge of an X-max’), SPECLFT (‘Every specifier is at the left edge of an X-max’), and COMPLFT (‘Every complement is at the left edge of an X-max’) and then by ranking them differently for different languages in accordance with their constituent order types. To illustrate this, under her analysis, languages like English have the ‘SPECLFT [[ HEADLFT [[ COMPLFT’ ranking in their grammar, so they instantiate SVO as their surface order, as shown in tableau (48); languages like Japanese, on the other hand, have the ‘SPECLFT [[ COMPLFT [[ HEADLFT’ ranking and therefore they instantiate the reversed ordering between V and O at the surface level, as shown in tableau (49). (48)

Reason for the SVO order in English under Grimshaw’s (2001) analysis: Ranking: SPECLFT >> HEADLFT >> COMPLFT input: Subj V Obj

(49)

SPECLFT

HEADLFT

COMPLFT

a.

[XP Spec [X Head Complement]]

*

**

b.

[XP Spec [X Complement Head]]

**!

*

Reason for the SOV order in Japanese under Grimshaw’s analysis: Ranking: SPECLFT >> COMPLFT >> HEADLFT input: Subj V Obj

SPECLFT

COMPLFT

HEADLFT

a.

[XP Spec [X Head Complement]]

**!

*

b.

[XP Spec [X Complement Head]]

*

**

That said, invoking each of the constraints in (47) is empirically motivated as well. First of all, the idea that the markedness constraint *NA > NNA regulates adjective ordering in human language is based on our finding in Chap. 3 that verbal APs tend to occur preceding non-verbal ones in Korean but it is also motivated by

8

This suggests that violations of HVYLFT are calculated in a cumulative manner but this will not be the same kind of cumulatively as authors like Keller (2000) and Sorace and Keller (2005) assume in analyzing linguistic data within an OT framework.

182

5 Capturing the Korean Facts

the fact that even in other languages, nominal APs tend to occur closer to the root N than non-nominal APs do when both types of AP co-occur within the same DPp boundary. To see this, consider the English data given in (50). Here, the (d) example, which instantiates ‘NA > NNA’ order, cannot occur in the postcopular position indicated by a blank.9 (50) Ordering between NNAs and NAs in English: That’s ______________. a. a silicon shoe b. an expensive shoe c. an expensive silicon shoe d. *a silicon expensive shoe

(NNA > NA) (NA > NNA)

Secondly, postulating DEGLFT as a member of the universal constraint set CON is prompted by our observation in Chap. 3 that ADJs with degree modifiers in Korean tend to occur in the left-edge of DPp but this idea also receives support from the fact that even in languages like English, such ADJs tend to occur preceding unfocused ADJs or ADJs with no degree modifiers. To see this, consider (51): as the candidate for filling in the blank in the example sentence given below, (51e) is decidedly worse than (51c, d). (51) Ordering between a modified ADJ and an unmodified one in English: My daughter is dating ___________________. a. a (very) cute boy b. a (very) tall boy c. a very tall, cute boy (modified AP > unmodified AP) d. a very cute, tall boy (modified AP > unmodified AP) e.??a cute, very tall boy (unmodified AP > modified AP)

9

An anonymous reviewer claims that this constraint is contradicted by the existence of English data like (i), in which a more nominal AP occurs farther away from the head N than a less nominal one does. (i) a gold musical box.

What is important to note, however, is that every OT constraint is violable and every language is assumed to have potentially different constraint rankings. To apply this to the data at hand: in (i), musical forms a compound N with box and gold modifies the entire compound N, resembling the structure of the Korean example given in (3a), i.e., say mohyeng catongcha ‘a new miniature car’. Given this, on the present analysis, musical would merge at [Spec, √P] and gold would merge at [Spec, nP] and between musical and gold, the latter has a more relative semantics (i.e., a gold musical box need not be (purely) gold). Therefore, the surface order in which they occur would satisfy RELLFT even though it may violate *NA > NNA. If correct, then, the present analysis suggests that English has a different constraint ranking than Korean but that is expected under an OT-based analysis. I should also point out that, in lieu of musical, music can occur in (i) (i.e., a music box), as numerous Google searches I have conducted validate, and in such cases, neither RELLFT nor *NA > NNA would be violated. I therefore conclude that the existence of data like (i) does not undermine the analysis put forward here.

5.3 Solving the Problems by Adding an Output Filter

183

Third, the motivation for invoking RELLFT comes from the fact that across languages, APs tend to occur observing the AOR given in (36), repeated below for convenience, and even in Korean, all else being equal, UN-APs tend to occur complying with it as well. (36) Typological generalization on the ordering between simplex APs: Quality > Size > Shape > Color > Provenance/Source

Fourth, HVYLFT is postulated for two reasons: one is that in Korean, phonologically heavier APs tend to occur father away from the head N than phonologically lighter ones do. The other reason is that in numerous other languages too, morpho-syntactically complex APs tend to occur farther away from the head N than morpho-syntactically simplex ones (see, a.o., Bolinger 1967; Sproat and Shih 1988, 1990; Larson 1998, 2000; Bouchard 2002; Cinque 2005, 2010).

5.3.2

Ranking of the Constraints on AO in Korean

I argue that in Korean, the four constraints listed in (47) are ranked as given in (52). This means that *NA > NNA is the highest-ranked and HVYLFT is the lowest-ranked among them, and DEGLFT and RELLFT are not ranked relative to each other although they both outrank HVYLFT. (52) Ranking among the constraints in (47) for Korean: *NA > NNA >> {DEGLFT, RELLFT} >> HVYLFT

I rank RELLFT higher than HVYLFT because grammatical data in which two unmodified UN-APs occur side by side may comply with RELLFT but they may violate HVYLFT. To see this, consider (53) and assume that it is uttered in an out-of-the-blue context. This data set exemplifies that in neutral contexts, ‘size > shape’ is strongly preferred in Korean even if the shape-term is phonologically heavier, and this suggests that in Korean, satisfying RELLFT is more “important” than satisfying HVYLFT and hence to assume ‘RELLFT [[ HVYLFT’ for the grammar of Korean. (53) a. ?khu-n, tongkura-n big-ADN circle.shaped-ADN ‘a big, round tray’ b. *tongkura-n, khu-n Intended: ‘a round, big tray’

cayngpan tray

(size > shape)

cayngpan

(shape > size)

Essentially the same argument can be made with the data given in (35), which are repeated below as (54). These examples contain two UN-APs, a size-term and a

184

5 Capturing the Korean Facts

color-term, and here again, in the good case, the shorter AP occurs to the left of the longer one, suggesting that ‘RELLFT [[ HVYLFT’ holds for Korean. (54) a. khu-n, ppalkah-n red-ADN big-ADN, ‘a big, red bag’ b.??ppalkah-n, khu-n Intended: ‘a red, big bag’

kapang bag

(size > color)

kapang

(color > size)

Next, I posit that in Korean, *NA > NNA outranks RELLFT on the basis of the fact that when a verbal AP co-occurs with a nominal one, the verbal AP always occurs preceding the nominal one, regardless of which between the two has a more relative semantics. By way of illustration, consider (55), where a size-term co-occurs with a quality-term and one of them is a nominal AP and the other is a verbal AP. In this paradigm, the grammatical datum satisfies *NA > NNA but violates RELLFT. In contrast, the ungrammatical datum violates *NA > NNA but satisfies RELLFT; We reach this verdict because quality-terms have more relative semantics than size-terms. Given this, it is concluded that the correct partial ranking for Korean has to be ‘*NA > NNA [[ RELLFT’ and since we have already decided above that ‘RELLFT [[ HVYLFT’ is the correct ranking between the two constraints, we obtain ‘*NA > NNA [[ RELLFT [[ HVYLFT’ for Korean. (55) a. [UN-AP khu-n] [N ko-kup] [ high-quality] [ big-ADN] ‘a large, high-quality car’ b. *[N ko-kup [UN-AP khu-n] Intended: ‘a high-quality, big car’

cha car

(size > quality)

cha

(quality > size)

For the sake of completeness, let us now consider (56), which is identical to (55) except that here, a nominal AP, rather than a verbal one, occurs as the size-term. In this paradigm, both the optimal and the non-optimal candidates are at a tie with regard to HVYLFT and they also vacuously satisfy *NA > NNA. As a result, what gets to be the deciding constraint here is RELLFT, as shown in tableau (57). And this provides additional support for positing ‘*NA > NNA [[ RELLFT [[ HVYLFT’ for Korean. (56) a. [N ko-kup], [N tay-hyeng] [ high-class] [ big-shape] ‘a high-quality, big-sized car’ b. *[N tay-hyeng], [N ko-kup] [ big-shape] [ high-class] Intended: ‘a big-sized, high-quality car’

cha car

(quality > size)

cha car

(size > quality)

5.3 Solving the Problems by Adding an Output Filter

185

(57) Consequence of assuming ‘*NA > NNA >> RELLFT >> HVYLFT’ for Korean: input: [DPp kokup tayhyeng N]

*NA > NNA

a.

[DPp [nP ko.kup, [nP tay.hyeng [ P N]]]]

b.

[DPp [nP tay.hyeng, [nP ko.kup [ P N]]]]

RELLFT

HVYLFT

*!

Turning now to the relative ranking between DEGLFT and RELLFT, at first glance, it may seem that DEGLFT outranks RELLFT in Korean because the highlighted part in (58a) violates RELLFT but complies with DEGLFT whereas (58b) exemplifies the opposite case. (58) a. [acwu ppalkah-n], khu-n [very red-ADN] big-ADN ‘a very red, big bag.’ b.??khu-n, [acwu ppalkah-n] Intended: ‘a big, very red bag.’

kapang bag

(color > size)

kapang

(size > color)

But considering a wider range of data leads us to see that ranking DEGLFT higher than RELLFT would in some cases give rise to undesirable results. To illustrate this, Korean speakers judge (59c) to be pretty good but (59b) to be not so good. In fact, they judge that (59c) is just as good as (59a), and crucially, here, the first ADJ is lengthened as a result of pronouncing the first syllable longer than it is in (59b). (59) Relative ordering between an UN-AP with a degree word and one without: a. [acwu ppalkah-n], khetara-n kapang (color > size) [very red-ADN] big-ADN bag Intended: ‘a very red, big bag.’ b.??khetara-n, [acwu ppalkah-n] kapang (size > color) Intended: ‘a big, very red bag.’ c. khe.e.tara-n, [acwu ppalkah-n] kapang (size w/t vowel lengthening) Intended: ‘a big, very red bag.’

If we assume ‘DEGLFT [[ RELLFT’ for Korean, then (59c) will be filtered out from the get-go, as shown in (60); if we assume, on the other hand, that the two constraints are at a tie for Korean, then we can correctly predict that both (59a) and (59c) are judged grammatical, as shown in (61). Even under the latter ranking scenario, (59b) will still come out ungrammatical because it violates HVYLFT, which the other two candidates do not. Given this, we conclude that in Korean, DEGLFT and RELLFT are not ranked relative to each other. Notably, treating the two constraints as such will make correct predictions about the data in (58) as well: as shown in (62), (58a) is predicted to be the more optimal candidate than (58b), and even when compared against a candidate in which khun is lengthened by one syllable, it comes out as the winner.

186

5 Capturing the Korean Facts

(60) Consequence of assuming ‘D EGLFT >> RELLFT’ for (59): Incorrect prediction input: [DPp khetaran acwu ppalkahn N] a. b. c.

DEGLFT

[DPp [SortP a.cwu ppal.kan, [SortP khe.ta.ran [nP [ P N]]]]] [DPp [SortP khe.ta.ran, [SortP a.cwu ppal.kan [nP [ P N]]]]] [DPp [SortP khe.e.ta.ran, [SortP a.cwu ppal.kan [nP [ P N]]]]]

RELLFT

HVYLFT

* *!

*

*!

(61) Consequence of assuming {DEGLFT, RELLFT} for (59): Correct prediction input: [DPp khetaran acwu ppalkahn N] a. b. c.

DEGLFT

[DPp [SortP a.cwu ppal.kan, [SortP khe.ta.ran [nP [ P N]]]]] [DPp [SortP khe.ta.ran, [SortP a.cwu ppal.kan [nP [ P N]]]]] [DPp [SortP khe.e.ta.ran, [SortP a.cwu ppal.kan [nP [ P N]]]]]

RELLFT

HVYLFT

* *

*!

*

(62) Consequence of assuming {DEGLFT, RELLFT} for (58): Correct prediction input: [DPp khun acwu ppalkahn N]

DEGLFT

a.

[DPp [SortP a.cwu ppal.kan, [SortP khun [nP [ N]]]]]

b.

[DPp [SortP khun, [SortP a.cwu ppal.kan [nP [ P N]]]]] [DPp [SortP khu.un, [SortP a.cwu ppal.kan [nP [ P N]]]]]

c.

5.3.3

RELLFT

HVYLFT

*

P

*

***!

*

**!

Solving the Problems Presented by Some UN-AP and ATT-DET/CEK-AP Co-occurrences

The OT analysis I have just put forward lets us account for the problematic data in (33)–(34), which are repeated below as (63)–(64). It further lets us explain why between two semantically similar lexemes (e.g., khetaran ‘large/big’ vs. khun ‘large/big’), the longer one may sometimes be chosen over the shorter one.

5.3 Solving the Problems by Adding an Output Filter

187

(63) Ordering between a tri-syllabic UN-AP and an ATT-DET: say kapang (UN-AP > ATT-DET) a. khe.ta.rah-n large-ADN new bag Intended: ‘a large new bag’ b. KHE.TA.RAH-N say kapang (UN-AP with FOC > ATT-DET) new bag large-ADN Intended: ‘a LARGE new bag (not a small one)’ c. * say khe.ta.rah-n kapang (ATT-DET > UN-AP) new large-ADN bag Intended: ‘a new large bag’ d. *SAY khe.ta.rah-n kapang (ATT-DET with FOC > UN-AP) bag new large-ADN Intended: ‘a NEW large bag (not a small one)’ (64) Ordering between a mono-syllabic UN-AP and an ATT-DET: a.??khu-n say kapang (UN-AP > ATT-DET) large-ADN new bag Intended: ‘a large new bag’ b. KHU-N say kapang (UN-AP with FOC > ATT-DET) large-ADN new bag Intended: ‘a LARGE new bag (not a small one)’ c. *say khu-n kapang (ATT-DET > UN-AP) new large-ADN bag Intended: ‘a new large bag’ d.??SAY khu-n kapang (ATT-DET with FOC > UN-AP) new large-ADN bag Intended: ‘a NEW large bag (not an old one)’

Under the proposed OT analysis, (63a) is grammatical but (64a) is not because while the former satisfies HVYLFT, the latter incurs one violation of it, as shown in (65).10 (65) Reason for the contrast between (63a) and (64a): input: [DPp khun/khetaran say N] a.

[DPp [SortP khe.ta.ran [nP say [ P N]]]]

b.

[DPp [SortP khun [nP say [ P N]]]]

*NA > NNA

RELLFT

HVYLFT

*!

Turning now to explaining the grammaticality judgments on (63a, b) and (64a, b): since (63a) is already judged grammatical, the goodness of (63b) does not surprise us. That is, putting a FOC on khetaran would not make any difference in this

10

Here and below, I only provide partial tableaux to stay focused on the issues at hand. For instance, the data in (63)–(64) all vacuously satisfy DEGLFT, so this constraint is omitted in tableaux (65)–(67).

188

5 Capturing the Korean Facts

particular case because it is already heavier than the ADJ that follows it. In the case of (64), however, FOC-marking on the first ADJ does make a difference. More concretely, given what we assume about the effect of placing a FOC on an ADJ, stated under (47), (64b) incurs just one violation of HVYLFT, which is one less than what (64a) incurs. Hence, when the two data are being compared by EVAL, the former surfaces as the winner, as indicated in tableau (66). Nevertheless, (64b) is judged to be a less optimal candidate than (63a, b), and this explains why Korean speakers would use either (63a) or (63b) to convey the intended meaning.11 (66) Reason for the contrast between (64a) and (64b): input: [DPp khun say N]

*NA > NNA

a.

[DPp [FocP [SortP khun [nP say [ P N]]]]]

b.

[DPp [FocP KHUN [SortP __ [nP say [ N]]]]]

RELLFT

HVYLFT *!

P

Finally, on the present analysis, the four examples given in (63c, d) and (64c, d) are all expected to be judged bad because all of them violate *NA > NNA, the highest-ranking constraint. Example in (64d) is predicted to be judged better than the other three, however, because it does not violate HVYLFT whereas the others do, as indicated in (67). Therefore, if (63c, d) and (64c, d) were the only candidates being evaluated, then (64d) would surface as the winner though such cases would not be attested since there will always be more optimal candidates competing with them, i.e., (63a, b) and (64b), and this provides a way to explain its marginal acceptability as noted above.

Essentially the same reasoning explains why sayropun ‘new’ would be preferred over say ‘new’ in certain contexts, as shown in (i): sayropun is a tri-syllabic AP whereas say is just mono-syllabic. For this reason, when they occur preceding a CEK-AP, sayropun (almost) always occurs instead of say. And under the present analysis, this is expected because while (ia) and (ib) do not comply with HVYLFT (the first example incurs two violations and the second one incurs one violation of this constraint), the (c) and the (d) examples do. Note that with regard to the other constraints, they are at a tie since they (vacuously) satisfy them. 11

(i) a .??say, hapli-cek kyengcey cengchayk policy new reason-CEK economy Intended: ‘a new rational economic policy’ b. ?SAY, hapli-cek kyengcey cengchayk Intended: ‘a NEW rational economic policy’ c. sayrop-un hapli-cek cengchayk reason-CEK policy new-ADN Intended: ‘a new rational policy’ d. SAYROP-UN hapli-cek cengchayk Intended: ‘a NEW rational policy’

(ATT-DET > CEK-AP) (ATT-DET with FOC > CEK-AP) (UN-AP > CEK-AP) (UN-AP with FOC > CEK-AP)

5.3 Solving the Problems by Adding an Output Filter

189

(67) Reason for the grammaticality judgement difference between (63c, d)/(64c) and (64d): input: [DPp khetaran/khun say N]

*NA > NNA

RELLFT

HVYLFT

a.

[DPp [FocP [SortP say [SortP khe.ta.ran [nP__ [ P N]]]]]]

*

***!

b.

[DPp [FocP SAY [SortP khe.ta.ran [nP __ [ P N]]]]] [DPp [FocP [SortP say [SortP khun [nP __ [ P N]]]]]]

*

**!

*

*!

[DPp [FocP SAY [SortP khun [nP __ [ N]]]]]

*

c. d.

P

The analysis proposed here makes correct predictions as to when FOC-marking may allow for an ADJ to occur at the left periphery of a DP and when it may not. By way of illustration, the nominal APs in (68)–(69) cannot occur preceding the UN-APs, despite the fact that they bear a FOC, so are phonologically heavier than the other APs. Under our analysis, this is so because their occurrences at the left edge of the DP will incur a violation of *NA > NNA, as shown in (70). (Recall that these APs are nominal whereas UN-APs are verbal, as we established in Chap. 2, based on their compatibility versus incompatibility with the i- copula verb when occurring in predicate position.) (68) Context: A asks B what he bought yesterday and he says the following in answer to the question. a.??TOKIL-CEY sayrop-un [catongcha] (NA with FOC > NNA) [car] German-made new-ADN Intended: ‘a GERMAN-MADE new car (not a French-made one)’ b. sayrop-un tokil-cey [catongcha] (NNA > NA) German-made [car] new-ADN ‘a new German-made car’ (69) Context: A asks B where he went yesterday and he says the following in answer to that question. a.??ILPON-SIK sayrop-un [siktang] (NA with FOC > NNA) Japan-style new-ADN [restaurant] Intended: ‘a JAPANESE new restaurant (not a Chinese style)’ b. sayrop-un ilpon-sik [siktang] (NNA > NA) Japan-style [restaurant] new-ADN ‘a new Japanese restaurant’

190

5 Capturing the Korean Facts

(70) Reason for the contrast between (68a) and (68b): input: [DPp sayropun tokilcey N] a. b.

*NA > NNA

[DPp [SortP say.ro.pun [nP tok.il.cey [ P N]]]] [DPp [FocP TOK.IL.CEY [SortP say.ro.pun [nP __ [ P N]]]]]

RELLFT

HVYLFT *

*!

*

Reconsider now (9), which is repeated below as (71). This datum shows that the focused nominal AP in (6) can occur preceding the ATT-DET, and this is as expected under our analysis because such cases involve ordering two nominal APs. More concretely, as shown in tableau (72), the candidates satisfy both *NA > NNA and RELLFT (though the first constraint is satisfied vacuously), so the constraint that ends up making the call is HVYLFT, and while the example given in (71) satisfies it, its competitor given in (72b) does not.12 (71) TOKIL-CEY say [mohyeng catongcha] German-made new [miniature car] ‘a GERMAN-MADE new miniature car’ (72) Reason for the goodness of (71): input: [DPp say tokilcey N] a. b.

[DPp [FocP TOK.IL.CEY [nP say [nP __ [ P N]]]]] [DPp [FocP [nP say, [nP TOK.IL.CEY [ P N]]]]]

*NA > NNA

RELLFT

HVYLFT

***!

In the present framework, the grammaticality differences between data like (73) and (74) also receive a straightforward account: these cases are concerned with relative ordering between two UN-APs, and because these APs’ semantics has to do with size, shape, or color, their relative ordering is subject to RELLFT, which outranks HVYLFT. As a result, even when focused, the color- or shape-term may not readily occur preceding the size-term, as shown in (75); that is, all else being equal,

12

Essentially the same logic explains the goodness of (10), which exemplifies that a FOC on the thematic AP may allow it to occur preceding a CEK-AP.

5.3 Solving the Problems by Adding an Output Filter

191

the examples instantiating ‘size > color/shape’ are judged better.13 And this in turn provides us with a way to explain why the FOC-marked ADJs in (73a) and (74a) may stay in situ rather than moving to [Spec, FocP] even though that position is available. (73) Context: The speaker is trying to clarify to the hearer that the bag she needs is different from what he thinks. a. Na-nun [DP khu-n, PPALKA-N kapang]-i (size > color with FOC) I-TOP [ big-ADN, red-ADN bag]-NOM philyoha- -ta, khu-n, PHARA-N kapang-i an-i-ra. bag-NOM not-COP-CONN need-N.PST-DECL big-ADN blue-ADN ‘I need a big RED bag, not a big BLUE one.’ b.??Na-nun [DP PPALKA-N, khu-n kapang]-i (color with FOC > size) philyoha- -ta, PHARA-N , khu-n kapang-i an-i-ra. Intended: ‘I need a RED big bag, not a BLUE big one.’ (74) Context: The speaker is trying to clarify to the hearer that the tray she needs is different from what he thinks. TONGKURA-N cayngpan-i (size > shape with FOC) a. Na-nun [DP khu-n, I-TOP [ big-ADN, round-ADN tray]-NOM philyoha- -ta, khu-n, NEYMONA-N cayngpan-i need-N.PST-DECL big-ADN square.shaped-ADN bag-NOM an-i-ra. not-COP-CONN ‘I need a big ROUND tray, not a big SQUARE one.’ b.??Na-nun [DP TONGKURA-N, khu-n cayngpan]-i (shape with FOC > size) philyoha- -ta, NEYMONA-N, khu-n cayngpan-i an-i-ra. Intended: ‘I need a ROUND big tray, not a SQUARE big one.’

13

I thank an anonymous reviewer for drawing this to my attention, that is, the fact that not all focused ADJs in Korean need to occur in the left-periphery of a DP. Notably, this is also true of other languages. For example, in English too, placing a focal stress on an ADJ alone can indicate focus-marking and therefore a FOC-marked ADJ need not appear at the left-periphery of the DP although it can, as shown in (i); in fact, (ib) is judged less good than (ia), paralleling the behavior of the Korean data given in (73)–(74).

(i) a. John lives in a large, luxurious, WHITE house (not a yellow one). b. ?John lives in a WHITE, large, luxurious house (not a yellow one).

192

5 Capturing the Korean Facts

(75) Reason for the contrast between (73a) and (73b): input: [DPp khun ppalkan N] a. b.

*NA > NNA

[DPp [FocP [SortP khun, [SortP PPAL.KAN [ P N]]]]] [DPp [FocP PPAL.KAN [SortP khun, [SortP ___ [ P N]]]]]

RELLFT

HVYLFT **

*!

Adding an OT component to our derivational system also lets us explain why (6) is slightly marginal but it becomes perfect if the UN-AP sayropun ‘new’ occurs instead of the ATT-DET say ‘new’, followed by a pause, as shown in (76). The grammaticality changes as a result of this lexical substitution simply because, unlike (6), (76) does not violate HVYLFT, as indicated in (77b). Notice further that even if we lengthen the vowel of say, mimicking what we did in (59c), the judgment does not become as impeccable as (76), as shown in (77c), and this also follows from our proposed OT analysis. (6) DP containing a compound N, an ATT-DET, and a thematic AP in-between: ?say, [mohyeng catongcha] tokil-cey new German-made [miniature car] ‘a new German-made miniature car’ (76) Variant of (6) with UN-AP sayropun ‘new’ occurring in lieu of ATT-DET say: sayrop-un, tokil-cey [mohyeng catongcha] ‘a new German-made miniature car’ (77) Reason for the contrast between (6) and (76): input: [DPp say/sayropun tokilcey N]

*NA > NNA RELLFT

a.

[DPp [FocP [nP say, [nP tok.il.cey [ P N]]]]]

b.

[DPp [FocP [nP say.rop.un, [nP tok.il.cey [

HVYLFT **!

P

N]]]]] c.

[DPp [FocP [nP say.ay, [nP tok.il.cey [ P N]]]]]

*!

Taken together, the grammaticality differences between minimal pairs like (6) and (76) as well as (73a, b), (74a, b), and (59b, c) show that the need to comply with a constraint like HVYLFT can impact how N modifiers are pronounced in actual discourses when they co-occur inside the same DPp boundaries, and I take this to be an additional welcome result of the present analysis. Lastly, refining our machinery by combining a derivational analysis with an OT analysis can readily handle data like (24)–(25), in which the ATT-DET hyen ‘present’ occurs both preceding and following an UN-AP. At first glance, the

5.3 Solving the Problems by Adding an Output Filter

193

surface order of ‘ATT-DET > UN-AP > N’ instantiated by (24b) and (25b) may appear to violate HVYLFT. But being a functional category, hyen is not subject to the same filtering system as lexical ADJs are. Similarly, due to their non-restrictive semantics, the UN-APs in (24)–(25) are base-generated outside of DPp and therefore they are not subject to the four constraints listed in (47); these constraints would only target the DPp region, as stated in Sect. 5.3.1 and as indicated by the dotted circle in (78). (78) Derivation of (24b) and the effects of the OT constraints listed in (47): DPd/r+ DPd/r AP

DPd/r

puwphayhan

D d/r Dd/r [+REF] [+SPEC] [+DEF] [+DEIC]

DPp LocP AP hyen

UnitP/PlP+ Unit/PlP SortP P cengkwon

*NA > NNA >> {DEGLFT, RELLFT} >> HVYLFT

Relatedly, the present analysis also lets us explain why the data in (79) are judged fine even though they both instantiate cases where a mono-syllabic N modifier co-occurs with an ADJ which has its own degree modifier and therefore is relatively long. The reason for their grammaticality is simply this: these DPs have the same syntax as those in (24)–(25) and therefore the constraints listed in (47) do not apply to them and this lets the APs occur in alternate orders relative to each other. (79) When hyen co-occurs with an UN-AP which has its own degree modifier: a. Degree Adv + UN-AP > ATT-DET: [maywu pwuphayha-]-n, hyen cengkwon [very corrupt-]-ADN current government ‘the very corrupt current government’ b. ATT-DET > Degree Adv + UN-AP: hyen, [maywu pwuphayha-]-n cengkwon ‘the current, very corrupt government’

194

5.3.4

5 Capturing the Korean Facts

Ordering Restrictions on FRCs in Korean

In resolving the empirical challenges presented by RC ordering in Korean, I should first point out that all the “problematic” data we have looked at have one common property: the DPs that directly embed the RCs at issue function as a discourse topic. This was already evident in data (38)–(46) since all the RC-containing DPs in them bear the topic marker -(n)un. But there are other reasons to think that ordering restrictions on Korean RCs are due (in part) to the discourse topic status of the DP they occur within. To see this, let us first consider the exchanges between the two speakers in (80). In this discourse, A informs B that she met someone yesterday and that person is a heavy smoker, and A’s second utterance exemplifies that (38a) can be felicitously uttered in a context where the DP containing the two RCs serves as the discourse topic and the RCs describe properties of individuals that are discourse-old and these properties hold true of the hosting DP’s referent. Notice also that replacing A’s second utterance with (38b), which exhibits the reversed order for the RCs, results in ungrammaticality, not to mention pragmatic anomaly. (80) Felicitous discourse context for (38a): A: Na-nun ecey etten saram-ul manna-ss-e. I-TOP yesterday some person-ACC met-PST-DECL.INFRML ‘I met someone yesterday.’ Kurentey ku saram-un phyengso tampay-rul manhi And that person-TOP usually cigarette-ACC a.lot phiwu- -e. smoke-N.PST-DECL.INFRML ‘And that person smokes a lot.’ machwu-e-poa- . Ku saram-i nwukwu-i- -nci That person-NOM who-COP-N.PST-whether guess-CONN-try-IMP.INFRML ‘Try to guess who that person is.’ B: Hoksi, pro Minswu-an-i- -ya? By.any.chance ___ M.-not-COP-N.PST-Q.INFRML ‘Isn’t it Minswu by any chance? A: Ani. [DP [RC1 Nay-ka ei ecey manna- ]-n, __ yesterday meet-PRF]-REL, No. [ [ I-NOM [RC2 ei phyengso tampay-rul manhi phiwu-n-]-un a.lot smoke-IMPRF-]-REL [ __ usually cigarette-ACC sarami]-un paro Chelswu-i- -ya. C.-COP-DECL.INFRML person-TOP exactly ‘No. The person that I met yesterday who smokes a lot is Chelswu.’ A .*Ani, [DP [RC2 ei phyengso tampay-rul manhi phiwu-n-]-un, No. [ [ __ usually cigarette-ACC a.lot smoke-IMPRF-]-REL [RC1 nay-ka ei ecey manna- ]-n [ I-NOM __ yesterday meet-PRF]-REL sarami]-un paro Chelswu-i- -ya. person-TOP exactly C.-COP-DECL.INFRML Intended: ‘No. The person who smokes a lot that I met yesterday is Chelswu.’

5.3 Solving the Problems by Adding an Output Filter

195

Regarding the badness of (38b), one may suspect that it will be judged grammatical if it is uttered in a context where A introduces a new discourse topic by first mentioning that the person she met yesterday is a heavy smoker and then commenting that she met him yesterday. But even in such a context, the sentence does not sound good, as exemplified by (81): this discourse shows that responding to B’s utterance with (38b) yields ungrammaticality as well as infelicity whereas doing so with (38a) does not. And this suggests that there is something not quite right about the way the two RCs in (38b) are ordered on the grammatical level, not just pragmatically. (81) Logically conceivable discourse context for (38b): A: Etten saram-i phyengso tampay-rul manhi cigarette-ACC a.lot some person-NOM usually phiwu- -e. smoke-N.PST-DECL.INFRML ‘There is someone who smokes a lot.’ Kurentey nay-ka ku saram-ul ecey manna-ss-e. And I-NOM that person- ACC yesterday met-PST-DECL.INFRML ‘And I met that person yesterday.’ machwu-e-poa- . Ku saram-i nwukwu-i- -nci That person-NOM who-COP-N.PST-whether guess-CONN-TRY-IMP.INFRML ‘Try to guess who that person is.’ B: Hoksi, pro Minswu-an-i- -ya? By.any.chance ___ M.-not-COP-N.PST-Q.INFRML ‘Isn’t it Minswu by any chance? A. *Ani, [DP [RC2 ei phyengso tampay-rul manhi phiwu-n-]-un, No. [ [ __ usually cigarette-ACC a.lot smoke-IMPRF-]-REL [RC1 nay-ka ei ecey manna- ]-n [ I-NOM __ yesterday meet-PRF]-REL sarami]-un paro Chelswu-i- -ya. C.-COP-DECL.INFRML person-TOP exactly Intended: ‘No. The person who smokes a lot that I met yesterday is Chelswu.’ A : Ani. [DP [RC1 Nay-ka ei ecey manna- ]-n, No. [ [ I-NOM __ yesterday meet-PRF]-REL, phyengso tampay-rul manhi phiwu-n-]-un [RC2 ei a.lot smoke-IMPRF-]-REL [ __ usually cigarette-ACC sarami]-un paro Chelswu-i- -ya. person-TOP exactly C.-COP-DECL.INFRML ‘No. The person that I met yesterday who smokes a lot is Chelswu.’

Consider now (82). In this paradigm, the same RCs as those in (39) occur but this time, their hosting DP is not a discourse topic; the topic of these sentences is manifested as a pro, which can be glossed as ‘as for me’. Interestingly, in these cases, the RCs can occur in either order relative to each other, and this provides additional evidence that the discourse status of the hosting DP plays a role in linearizing Korean RCs.

196

5 Capturing the Korean Facts

(82) Context: The speaker cannot remember things very well these days. So out of frustration, she says something out of the blue. ei sakwi-n-]-un, a. proj [DP2 [DP1 [RC1 Mina-ka yocum __ [ [ [ M.-NOM these.days __ date-IMPRF-]-REL, elkwul-i nemwuna calsayngki- ]-n [RC2 ei truly be.handsome-]-REL [ __ face-NOM ku namcai]-uy irum]-i kapcaki sayngkak-i name]-NOM suddenly thought-NOM that man]-GEN an-na-n- -ta. not-come.out-IMPRF-N.PST-DECL ‘I don’t remember the name of the guy Mina is dating these days who is so handsome.’ b. proj [DP2 [DP1 [RC2 ei elkwul-i nemwuna calsayngki- ]-n, truly be.handsome-N.PST]-REL, __ [ [ [ __ face-NOM [RC1 Mina-ka yocum ei sakwi-n-]-un [ M.-NOM these.days __ date-IMPRF-]-REL ku namcai]-uy irum]-i kapcaki sayngkak-i that man]-GEN name]-NOM suddenly thought-NOM an-na-n- -ta. not-come.out-IMPRF-N.PST-DECL ‘I don’t remember the name of the guy Mina is dating these days who is so handsome.’

Now, I would like to point out that three additional factors play a role in RC ordering in Korean. First of all, there is reason to think that whether the RCs at hand are derived from episodic event descriptions or not affects their surface order. Descriptively speaking, an RC with an episodic content tends to occur preceding an RC with a non-episodic content, as illustrated by (83) as well as (38): in both (38) and (83), the (a) examples instantiate cases where the first RC has an episodic (E) content and the second RC has a non-episodic (NE) content (i.e., ‘RCE > RCNE’) and the (b) examples instantiate cases where an RC with an NE content occurs preceding an RC with an E content (i.e., ‘RCNE > RCE’). (83) Context: Both the speaker and the hearer know which child they are talking about. a. [DP [RC1 Nay-ka ecey ei satacwu- ]-n, yesterday __ buy-CONN give-PRF]-REL [ [ I-NOM [RC2 ku ay-ka ei phyengso chohaha-n-]-un [ that child-NOM __ ordinary.time like-IMPRF-]-REL umsik]-un phicca-i- -ta. food]-TOP pizza-COP-N.PST-DECL ‘The food that I bought and brought for the child yesterday which he likes a lot is pizza.’ (RC1 > RC2) ay-ka ei phyengso chohaha-n-]-un, b. *[DP [RC2 Ku ei ecey satacwu- ]-n [RC1 nay-ka umsiki]-un phicca-i- -ta. Intended: ‘The food that the child likes a lot which I bought and brought (to him) yesterday is pizza.’ (RC2 > RC1)

5.3 Solving the Problems by Adding an Output Filter

197

The contrast between (84) and (85) further illustrates the workings of the constraint I have just identified, which we will refer to as ‘the RCE > RCNE condition’ for now. In (84), the two RCs are both derived from an NE event description and they can occur in either order. In (85), on the other hand, the (b) example is judged bad, and this can be attributed to the fact that the presence of the temporal adverb yocumtule ‘these days’ in the first RC makes its content more episodic than otherwise, so given that the second RC’s content is still non-episodic, the more desirable surface order for them would be what is exemplified by (85a), namely, ‘RC1 > RC2 > N’ or ‘RCE > RCNE > N’. (84) Context: The discourse participants are trying to identify the person who has the property of being tall and who is also a heavy smoker. a. [DP [RC1 ei phyengso tampay-rul manhi phiwu-n-]-un, a.lot smoke-IMPRF-]-REL [ [ __ usually cigarette-ACC [ RC2 ei khi-ka maywu khu- ]-n, big-N.PST]-REL, [ __ height-NOM very saram i]-un Cinho-i- -ta. person]-TOP C.-COP-N.PST-DECL ‘The person who smokes a lot who is very tall is Cinho.’ (RC1 > RC2) b. [DP [RC2 ei khi-ka maywu khu- ]-n, [RC1 ei phyengso tampay-rul manhi phiwu-n-]-un Cinho-i- -ta. sarami]-un ‘The person who is very tall who smokes a lot is Cinho.’ (RC2 > RC1) (85) Context: The discourse participants are trying to identify the person who has the property of being tall and who has recently become a heavy smoker. a. [DP [RC1 ei yocum-tule tampay-rul manhi phiwu-n-]-un, [ [ __ these.days cigarette-ACC a.lot smoke-IMPRF-]-REL, ei khi-ka maywu khu- -]-n [RC2 [ __ height-NOM very big-N.PST]-REL, sarami]-un Cinho-i- -ta. person]-TOP C.-COP-N.PST-DECL ‘The person who smokes a lot these days who is very tall is Cinho.’ (RC1 > RC2) b. *[ DP [RC2 ei khi-ka maywu khu- ]-n, ei yocum-tule tampay-rul manhi phiwu-n-un] [RC1 sarami]-un Cinho-i- -ta. Intended: ‘The person who is very tall who smokes a lot these days is Cinho.’ (RC2 > RC1)

Another descriptive generalization I would like to offer regarding the way RCs are linearized in Korean is that if two RCs co-occur and they both have episodic contents, then, all else being equal, the RC whose event time precedes the other RC’s event time occurs more to the left in the DP structure. To see this, consider first (86)–(88). In these paradigms, the (a) examples are all grammatical, and what cuts across them is that the time of their first RC (t1) is before the time of their second RC (t2)—that is, ‘t1 < t2’ holds, where < indicates a precedence relation. Conversely, in all the less desirable cases, t1 follows t2; that is, ‘t1 > t2’ holds.

198

5 Capturing the Korean Facts

(86) a. [DP [RC1 Mina-ka ku nal ei malha- ]-n, day __ mention-ANT]-REL, [ [ M.-NOM that [RC2 nay-ka ei ecey manna- ]-n sarami]-un __ yesterday meet-ANT]-REL person]-TOP [ I-NOM Cinho-i- -ta. C.-COP-N.PST-DECL ‘The person that Mina mentioned that day who I met yesterday is Cinho.’ (t1 < t2) b. *[DP [RC2 Nay-ka ei ecey manna- ]-n, Mina-ka ku nal ei malha- ]-n [RC1 Cinho-i- -ta. sarami]-un Intended: ‘The person that I met yesterday who Mina mentioned that day is Cinho.’ (t1 > t2) (87) a. [DP [RC1 Chelswu-ka ecey ei Mina-eykey sata[ [ C.-NOM yesterday __ M.-to buy-CONN cwu- ]-n, [RC2 nay-ka cikum ei kac-ko give-ANT]-REL, [ I-NOM now __ have-CONN iss-n-]-un cangnankami]-un ropothu-i- -ta. exist-IMPRF-]-REL toy]-TOP robot-COP-N.PST-DECL ‘The toy that Chelswu bought and brought to Mina yesterday which I have with me now is a robot.’ (t1 < t2) b.?[DP [RC2 Nay-ka cikum ei kac-ko iss-n-]-un, [RC1 Chelswu-ka ecey ei Mina-eykey satacwu- ]-n ropothu-i- -ta. cangnankami]-un Intended: ‘The toy that I have with me now which Chelswu bought and brought to Mina yesterday is a robot.’ (t1 > t2) (88) a. [DP [RC1 ei wuri hoysa-ey cinan cwu-ey ipsaha- ]-n, [ [ __ our company-LOC last week-LOC enter-ANT]-REL, onul kwacang-nim-ul towa- -turi- ]-n [RC2 ei [ __ today director-HON-ACC help-conn-give.HON-ANT]-REL cikwoni]-un Pakkinam-ssi-i- -ta. employee]-TOP P.-Mr.-COP-N.PST-DECL ‘The employee who started working in our company last week who helped the department’s director today is Mr. Kinam Park.’ (t1 < t2) b.??[ DP [RC2 ei onul kwacang-nim-ul towa- -turi- ]-n, [RC1 ei wuri hoysa-ey cinan cwu-ey ipsaha- ]-n cikwoni]-un Pakkinam-ssi-i- -ta. Intended: ‘The employee who helped the department’s director today who started working in our company last week is Mr. Kinam Park.’ (t1 > t2)

Thirdly, when two Korean RCs co-occur modifying a DP that denotes a discourse topic, the RC whose content is personally more familiar to the relevant discourse participant at the time of discourse occurs more to the left inside the DP. Notably, the need to meet this condition may even override the need to meet the other two conditions we have identified, namely, ‘the RCE > RCNE condition’ and what we can tentatively label as ‘the t1 < t2 condition’. This is illustrated by the contrast between the two sentences in (89): in this paradigm, RC1’s content is

5.3 Solving the Problems by Adding an Output Filter

199

personally familiar to the speaker—in fact, she probably has first-hand experience to assert that the content of this RC is true in the actual world. On the other hand, RC2 contains a hearsay marker, which indisputably indicates that its content is less familiar to the speaker than RC1’s content is. And this correlates with the fact that while (89a) is judged fine, (89b) is not. I should also note that (89a) is judged good even though RC1 has a non-episodic content and RC2 has an episodic one in violation of what we call the RCE > RCNE condition. (89) Context: The speaker and the hearer are both friends with Mina. The speaker is especially close to her, so she spends a lot of time with her every single day. a. [DP [RC1 Mina-ka ei cohaha-n-]-un, [ [ M.- NOM __ like-IMPRF-]-REL, han-ttay ywuhaynghay-ss-ta-ko-ha-]-nun [ RC2 yec-nal-ey [ old-day- LOC one-time be.popular-PST-IND-COMP-HEARSAY-]-REL norayi]-nun paro i noray-i- -ta. exactly this song-COP-N.PST-DECL song]-TOP ‘The song that Mina likes which is said to have been popular sometime in the past is this song.’ (RC1 > RC2) b.??[DP [RC2 yec-nal-ey han-ttay ywuhaynghay-ss-ta-ko-ha-]-nun, Mina-ka ei cohaha-n-]-un [RC1 norayi]-nun paro i noray-i- -ta. Intended: ‘The song that is said to have been popular sometime in the past which Mina likes is this song.’ (RC2 > RC1)

Essentially the same points can be made with the data given in (90): in this paradigm, the (a) example has RC1 in (89a) occur in the second position, and the (b) example has that RC occur in the first position. Yet here, the (b) example is judged less good. We can attribute the marginality of the (b) example to the fact that its first RC’s content is deemed less familiar to the speaker than its second RC’s content is. That is, the property of ‘being liked by Mina’ makes the hosting DP less familiar to the speaker than the property of ‘being made by me’ does, and this is what makes the two RCs occur in the order they do, namely, as in the (a) example. (90) Context: The speaker and the hearer are both friends with Mina. The speaker is especially close to her, so she spends a lot of time with her every single day. a. [DP [RC1 Nay-ka ei mantul- ]-n, __ make-ANT-]-REL, [ [ I-NOM Mina-ka ei cohaha-n-]-un [RC2 [ [ M.-NOM __ like-IMPRF-]-REL norayi]-nun napi-i- -ta. butterfly-COP-N.PST-DECL song]-TOP ‘The song that I wrote which Mina likes is Napi.’ (RC1 > RC2) b.??[DP [RC2 Mina-ka ei cohaha-n-]-un, [RC1 nay-ka ei mantul- ]-n norayi]-nun napi-i- -ta. Intended: ‘The song that Mina likes which I wrote is Napi.’ (RC2 > RC1)

200

5 Capturing the Korean Facts

Consider now (91). In this case, the two RCs respectively denote the properties of ‘being liked by my younger brother’ and ‘being recalled to have been made by my mom and offered to us sometime in the past’, and between these two properties, the first one makes the hosting DP’s referent more familiar to the speaker since it is derived from an eventuality that is experienced by her at speech time. The second property is also derived from an eventuality that is familiar to the speaker but this eventuality is something that occurred in a relatively distant past, as indicated by the retrospective marker -te on the verbal stem. For this reason, while ‘RC2 > RC1’ is not entirely out, ‘RC1 > RC2’ is judged decidedly better. Notably, the surface order instantiated by (91a) violates what we call the RCE > RCNE condition, and this, in conjunction with the goodness of (90a), suggests that the third condition we have just identified may override the other two and therefore a weighted treatment of the three constraints will be necessary. (91) a. [DP1 [RC1 nayj nam-tongsayng-i ei cohaha-n-]-un, [ [ my male-younger.sibling-NOM __ like-IMPRF-]-REL [RC2 yec-nal-ey [DP2 proj emma]-ka wuri-eykey ei __ mom]-NOM we-to __ [ old-day-LOC [ mantul-e cwu-si-te-]-n umsiki]-un mantwu-i- -ta. make-CONN give-HON-RTRO-]-REL food]-TOP dumpling-COP-N.PST-DECL ‘The food that my younger brother likes which my mom used to make for us is dumplings.’ (RC1 > RC2) [DP2 proj emma]-ka wuri-eykey b.??[DP1 [RC2 yec-nal-ey mantul-e cwu-si-te-]-n, [RC1 nayj nam-tongsayng-i ei cohaha-n-]-un umsiki]-un mantwu-i- -ta. ei Intended: ‘The food that my mom used to make for us which my younger brother likes is dumplings.’ (RC2 > RC1)

Before leaving this topic, I should point out that the third condition is at work even if the relevant discourse participant is construed as the hearer rather than the speaker. To see this, consider (92). Here, the speaker (S) is asking a question of the hearer (H) about something that pertains to him (i.e., H). In this discourse, the content of RC1 is deemed personally more familiar to H at speech time than the content of RC2 is. Therefore, in the given discourse context, the third condition would dictate that the two RCs occur in the order of ‘RC1 > RC2’, and this is exactly what is attested by the grammaticality and the ungrammaticality of (92a) and (92b).

5.3 Solving the Problems by Adding an Output Filter

201

(92) Context: S knows that H has written several songs and S is familiar with all of them. She also knows that their mutual friend Mina likes one of the songs written by H but she does not know which song it is. In this context, she says one of the following sentences to H. a. [DP [RC1 Ney-ka ei mantul- ]-n, __ make-ANT-]-REL, [ [ you-NOM Mina-ka ei cohaha-n-]-un [RC2 [ [ M.-NOM __ like-IMPRF-]-REL mwues-i- -ni? norayi]-nun what- COP-N.PST-Q.INFRML song]-TOP ‘What is the song that you wrote which Mina likes?’ (RC1 > RC2) b. *[DP [RC2 Mina-ka ei cohaha-n-]-un, ney-ka ei mantul- ]-n [RC1 norayi]-nun mwues-i- -ni? Intended: ‘What is the song that Mina likes which you wrote?’ (RC2> RC1)

In sum, then, what I have presented above leads us to conclude that RC ordering in Korean is governed by factors that are largely DP-external: first of all, much of the phenomenon has to do with the discourse-familiar topic status of the DP that embeds the RCs at hand. Secondly, when two RCs co-occur inside the same DP which functions as a discourse-familiar topic, the first RC’s content has to be personally more familiar to the relevant discourse participant at the time of discourse than the second RC’s content is and this obviously requires a semantic computation at the utterance-level. Moreover, we have also observed that some of the relevant data may exhibit gradient grammaticality and the need to fulfill one constraint may override the need to fulfill other constraints. Collectively, these findings suggest that RC ordering in Korean can be best handled in an optimality theoretic analysis under which DP-internal positions of RCs are determined by the interactions between different constraints, similarly to the way adjective ordering in the DPp region is determined. That said, offering a fully satisfactory analysis of RC ordering in Korean within an OT framework will be beyond the scope of this book, as it inevitably requires a more in-depth discussion of Topic (which is a massively elusive topic in and of itself) and a direct engagement with other OT-based work (e.g., Choi 1996). Therefore, aiming for a relatively modest goal, in the next subsection, I will only make an initial attempt to offer such an analysis.

5.3.5

An Attempt to Formally Derive the Ordering Restrictions on Korean RCs

Incorporating some of the findings we have just made, I would like to start this subsection by suggesting that when two RCs co-occur inside the same DP which functions as a discourse topic, they are both base-generated at a [Spec, DPd/r] that is created atop the pre-existing one, forming an adjunction structure, as pictured in (93).

202

(93)

5 Capturing the Korean Facts

Position of RCs inside a topic DP: DPd/r RC

DPd/r RC

DPd/r DPq DPp

I would like to suggest next that despite their adjunct status, such RCs are subject to certain ordering restrictions because, given their hosting DP’s discourse status, whatever semantic contributions they make has to be deemed desirable at the discourse level. In other words, they occur in certain orders not because of their intrinsic syntactic properties but because of the discourse function of the DP they are meant to modify. Thirdly, as formal renditions of the three conditions we identified in the previous subsection, I propose (94) and suggest that the constraints listed here form a subset of CON (and therefore are part of UG) and they only target the left edge of a topic DP; this means that they do not apply to RCs that occur inside a non-discourse topic DP or those that occur below the DPd/r level. (94) Constraints on RCs which modify the same discourse-familiar sentential topic: a. EPCRCLFT: An RC derived from an episodic event description occurs at the left edge of a topic DP d/r. b. PRERCLFT: When two RCs modify the same topic DPd/r, the RC whose event time is earlier occurs at the left edge of the DPd/r. c. FMLRCLFT: When two RCs modify the same topic DPd/r, the RC whose content is more familiar to the relevant discourse participant (i.e., S or H) occurs at the left edge of the DPd/r.

Fourth, I propose (95) as the ranking for these three constraints. I posit that FMLRCLFT outranks both PRERCLFT and FMLRCLFT on the basis of the fact that it overrides the other two conditions, as we observed in the previous subsection. As for EPCRCLFT versus PRERCLFT, I suggest that they are unranked relative to each other because what I called ‘the t1 < t2 condition’ in the previous subsection is contingent upon what I called ‘the RCE > RCNE condition’ there. (95) Ranking between the proposed three constraints on topic-modifying RCs: FMLLFT >> {EPCLFT, PRELFT}

Finally, I assume that when looking at sentential data, the relevant component of UG (e.g., EVAL) knows which DP is construed as the discourse-familiar topic at the

5.3 Solving the Problems by Adding an Output Filter

203

given discourse time and evaluates the relevant output candidates that GEN generates with respect to the constraint ranking in (95). I further assume that a DP which functions as a discourse topic occurs at the Spec of a Topic Phrase (TopP) in the sense of Rizzi (1997), which is, according to Rizzi, located somewhere above Inflection Phrase (IP) and below Force Phrase (ForceP), as schematically given in (96). (96) Position of a discourse-familiar topic inside a clausal structure: ForceP TopP DP

Top Top

XP IP

To recap, then, the three constraints I have proposed target the left edge of a topic DPd/r, as indicated in (97), and consequently, RCs occurring below the DPd/r level will be unaffected by them. (97) The operation domain of the constraints listed in (94): TopP DPd/r RC

Top

DPd/r Top RC

DPd/r

XP X

IP

DPp

FMLRCLFT >> {PRERCLFT, EPCRCLFT}

Turning now to demonstrating how the proposed analysis may capture the relevant facts: first, in this analysis, the two RCs in (38) occur in the order they do because data instantiating ‘RC2 > RC1’ would incur a violation of the highest-ranked relevant constraint, namely, FMLRCLFT, whereas data instantiating the other order (i.e., ‘RC1 > RC2’) would not, as shown in tableau (98). (Here and below, for ease of exposition, in the OT tableaux, I provide the English translations for the Korean data at issue. In addition, I do not represent all the DPd/r boundaries since in all cases, the less optimal candidate will always incur at least one more violation of the highest-ranked constraint than the optimal candidate does, so in the end, the presence of all the DPd/r boundaries in the diagrams will not impact the final outcome.)

204

5 Capturing the Korean Facts

(98) Reason for the surface order of the two RCs in (38): input: [DPd/r [RC1 nayka ecey mannan] [RC2 phyengso tampayrul manhi phiwunun] [DPp [ P saram]]] a.

FMLRC LFT

EPCRC LFT

*!

*

PRERC LFT

[DPd/r [RC1 ‘I met yesterday’] [RC2 ‘usually smokes a lot’] [DPp [ P ‘person’]]]

b.

[DPd/r [RC2 ‘usually smokes a lot’] [RC1 ‘I met yesterday’] [DPp [ P ‘person’]]]

Similarly, we can ascribe the contrast between the two sentences in (40) to the fact that (40b) would incur a violation of the highest-ranked constraint whereas (40a) would only violate the next highest-ranked constraint, namely, PRERCLFT, as shown in (99). (40a) does not violate FMLRCLFT because here, the two RCs occur in the order of ‘RC1 > RC2’, and the content of RC1 should be more familiar to the speaker than the content of RC2 is; the RC1 is about someone S personally knows whereas RC2 is not. (99) Reason for the surface order of the two RCs in (40): input: [DPd/r [RC1 Minaka kacang cal kiekhanun] [RC2 kwusipnyentaychoey issessten] [DPp [ P il]]] a.

[DPd/r [RC1 ‘Mina remembers’] [RC2 ‘occurred in the early 1990s’] [DPp [ P ‘event’]]]

b.

[DPd/r [RC2 ‘occurred in the early 1990s’] [RC1 ‘Mina remembers’] [DPp [ P ‘event’]]]

FMLRC LFT

EPCRC LFT

PRERC LFT *

*!

As for (39), only the (a) example is judged grammatical because RC1 here has an episodic content but RC2 has a non-episodic one, so having them occur in the ‘RC2 > RC1’ order as in the (b) example would incur one violation of EPCRCLFT whereas having them occur in the ‘RC1 > RC2’ order would not, as indicated in (100). (100) Reason for the surface order of the two RCs in (39): input: [DPd/r [RC1 Minaka yocum sakwinun] [RC2 elkwuli nemwuna calsayngkin] [DPp [ P namca]]] a. b.

[DPd/r [RC1 ‘Mina is dating’] [RC2 ‘is goodlooking’] [DPp [ P ‘man’]]] [DPd/r [RC2 ‘is good-looking’] [RC1 ‘Mina is dating’] [DPp [ P ‘man’]]]

FMLRC LFT

EPCRC LFT

PRERC LFT

*!

The proposed analysis also provides a way to account for variable grammaticality judgments exhibited by data like (42): in some contexts, (42b) will be judged

5.3 Solving the Problems by Adding an Output Filter

205

even better than (42a) because, due to the presence of the retrospective/evidential marker -te, RC2 can be construed as denoting a property that is derived from an eventuality that the speaker recalls from the past,14 and on such a construal, it can actually make the hosting DP’s referent more familiar to S than what the property denoted by RC1 would. Besides, unlike (42a), the surface order exhibited by (42b) does not violate EPCRCLFT or PRERCLFT. Therefore, when uttered in a discourse context like (101b), (42b) may be judged more grammatical than (42a) although when uttered in contexts like (101a), the judgement will be reversed. (101) Reason for the variable grammaticality judgements on (42a, b): a. Context: The speaker was born in the mid 1990s and is not very familiar with songs that used to be popular in the early 1970s but she is good friends with Mina and so knows that Mina likes certain songs from the early 1970s. input: [DPd/r [RC1 Minaka phyengso culkyepwurun] [RC2 chilsipnyentayey ywuhaynghayssten] [DPp [ P noray]]] a.

[DPd/r [RC1 ‘Mina likes to sing’] [RC2 ‘was popular in the 1970s’] [DPp [ P ‘song’]]]

b.

[DPd/r [RC2 ‘was popular in the 1970s’] [RC1 ‘Mina likes to sing’] [DPp [ P ‘song’]]]

FMLRC LFT

EPCRC LFT

PRERC LFT

*

*

*!

b. Context: The speaker was born in the mid 1950s and really liked all the songs that were popular in the early 1970s, as she listened to them, growing up. As a result, she can recite almost all of them even now. FMLRC EPCRC PRERC input: [DPd/r [RC2 chilsipnyentayey ywuhaynghayssten] [RC1 Minaka phyengso LFT LFT LFT culkyepwurun] [DPp [ P noray]]] a.

[DPd/r [RC1 ‘Mina likes to sing’] [RC2 ‘was popular in the 1970s’] [DPp [ P ‘song’]]]

b.

[DPd/r [RC2 ‘was popular in the 1970s’] [RC1 ‘Mina likes to sing’] [DPp [ P ‘song’]]]

*!

*

Notably, the paradigm in (102) may at first glance seem similar to (42) but the present analysis correctly predicts that the RCs in this paradigm occur in the other surface order: Here, RC2 contains the hearsay-marker -hanun, rather than the retrospective/evidential marker -te. As a result, the ‘RC2 > RC1’ order would incur a violation of FMLRCLFT whereas the other order would not. Therefore, (102a) is judged to be markedly better than (102b).

14

See Kim 2016a and the references there regarding the semantics of -te that occurs as part of an RC marker and its contribution to bringing about such a speaker-oriented recollective meaning.

206

5 Capturing the Korean Facts

(102) a. [DP [RC1 Mina-ka ei chohaha-n-]-un, [ M.-NOM __ like-IMPRF-]-REL, [RC2 ei yec-nal-ey hanttay ywuhaynghayss-ta-ko-ha-]-nun [ __ old-day-LOC once was.popular-IND-COMP-HEARSAY]-REL norayi]-nun leylitpi-i- -ta. song-TOP let.it.be-COP-N.PST-DECL ‘The song that Mina likes which is said to have been popular sometime in the past is Let It Be.’ (RC1 > RC2) b.??[DP [RC2 ei yec-nal-ey hanttay ywuhaynghayss-ta-ko-ha-]-nun, [RC1 Mina-ka ei chohaha-n-]-un norayi]-nun leylitpi-i- -ta. Intended: ‘The song that is said to have been popular in the past which Mina likes is Let It Be.’ (RC2 > RC1) (103) Reason for the fixed surface order of the two RCs in (102): input: [DPd/r [RC1 Minaka cohahanun] [RC2 yecnaley hanttay ywuhaynghaysstakohanun] [DPp [ P noray]]] a.

b.

[DPd/r [RC1 ‘Mina likes’] [RC2 ‘is said to be popular at some time in the past’] [DPp [ P ‘song’]]] [DPd/r [RC2 ‘is said to be popular at some time in the past’] [RC1 ‘Mina likes’] [DPp [ P‘song’]]]

FMLRC LFT

EPCRC LFT

PRERC LFT

*

*

*!

Finally, the proposed way to account for the linear ordering restrictions on RCs lets us capture paradigms like (45)–(46), where all the RCs contain object gaps and they all contain S-level predicates. To take (45) for example, under the present analysis, the (a) example is judged to be good but the (b) example is not because the latter incurs one violation of PRERCLFT, which the former does not:15 (104) Reason for the surface order of the RCs in (45): input: [DPd/r [RC1 Minaka kunal malhan] [RC2 nayka ecey mannan] [DPp [ P saram]]] a.

[DPd/r [RC1 ‘Mina mentioned that day’] [RC2 ‘I met yesterday’] [DPp [ P ‘person’]]]

b.

[DPd/r [RC2 ‘I met yesterday’] [RC1 ‘Mina mentioned that day’] [DPp [ P ‘person’]]]

FMLRC LFT

EPCRC LFT

PRERC LFT

*!

On the basis of the foregoing, I conclude that adding an optimality theoretic apparatus to the derivational mechanism we have developed yields desirable results

15

One may wish to challenge this explanation by claiming that the speaker meeting the person under discussion describes a more familiar experience to her than her merely hearing Mina mention him; that is, (45a), which instantiates the ‘RC1 > RC2’ order, violates FMLRCLFT. But one cannot deny that the event of Mina mentioning that person to S is what S directly experienced. Therefore, I conclude that both examples in (45) satisfy FMLRCLFT, which makes PRERCLFT the deciding constraint in this case.

5.3 Solving the Problems by Adding an Output Filter

207

in accounting for ordering restrictions on non-restrictive RCs in Korean. I also take the results obtained here to suggest that output candidates generated by a derivational system always undergo an additional filtering process performed by what is referred to as EVAL in OT and therefore, a successful account of any constituent ordering phenomenon in human language may require both derivational and representational treatments.

5.4

Why Co-occurring with Ku ‘the/that’ Engenders a Non-canonical ADJ Order16

At the end of Chap. 3, we observed that in ordinary circumstances, when non-clausal lexical N modifiers co-occur with an RC inside the same DP, they always occur following the RC, but when occurring immediately after the distal DEM ku ‘the/that’ followed by a pause, they may occur preceding the RC, as shown in (105)–(108). And on the basis of these observations, I concluded that the phenomenon at hand is most likely due to some discourse prominent meaning that a DEM-marking inside a nominal projection is meant to encode (whatever it might be). Co-occurrence of an FRC and a CEK-AP with a DEM: [RC ei hakkyo-rul palcen-sikhi-]-l*(,) [AP hapli-cek] (105) a. ku*(,) that [ __ school-ACC develop-CAUS-]-REL [ rational-CEK] cengchayki policy ‘the/that rational policy, which will develop our school’ (DEM > RC > CEK-AP) [RC ei hakkyo-rul palcen-sikhi-]-l b. ?ku [AP hapli-cek]*(,) cengchayki Intended: ‘that rational policy that will develop our school’ (DEM > CEK-AP > RC) (106) a. ku*(,) [RC ei mincwucwuuy-rul thanapha- ]-n*(,) that [ __ democracy-ACC oppress-PRF]-REL [AP tokcay-cek] kwuncwui [ tyrant-CEK] monarch ‘the/that tyrannical monarch, who oppressed democracy’ (DEM > RC > CEK-AP) ei mincwucwuuy-rul thanapha- ]-n b. ku [AP tokcay-cek]*(,) [RC kwuncwui Intended: ‘that tyrannical monarch who oppressed democracy’ (DEM > CEK-AP > RC)

16

Part of the material presented in this section has been published in Kim 2016b but I offer a revised analysis here. I also provide additional data and walk the reader through each example more slowly. In Chap. 6, I present a more comprehensive analysis of DEMs in Korean, comparing them with the DEMs in English and in other languages.

208

5 Capturing the Korean Facts

Co-occurrence of an FRC and an UN-AP with a DEM: [AP kappissa-n] [RC Mina-ka ei sa-o- ]-n*(,) (107) a. ku*(,) buy-come-PRF]-REL [ expensive-ADN] that [ M.-NOM __ mwulken-tuli item-PL ‘the/those expensive things, which Mina bought and brought’ (DEM > FRC > UN-AP) mwulken-tuli kappissa-n]*(,) [RC Mina-ka ei sa-o- ]-n b. ku [AP Intended: ‘those expensive things that Mina bought and brought’ (DEM > UN-AP > FRC) (108) a. ku*(,) [RC Mina-ka ei sakwi-ess-te-]-n*(,) [AP calsayngki-n] date-PRF-RTRO]-REL [ handsome-ADN] that [ M.-NOM __ namca-tuli man-PL ‘the/those handsome men, who Mina used to date’ (DEM > FRC > UN-AP) b. ku [AP calsayngki-n]*(,) [RC Mina-ka ei sakwi-ess-te-]-n namca-tuli Intended: ‘those handsome men that Mina used to date’ (DEM > UN-AP > FRC)

In this section, I propose a derivational analysis which accounts for the goodness of data like (105b)–(108b) by spelling out the role that the DEM ku plays in such data. The upshot of the analysis will be that when occurring in such environments, ku selects for an AP and carries what I call a ‘cognitive indexical’ meaning. More concretely, I claim that computing the meaning of such a DEM involves locating the property of individuals denoted by its complement in the mental domain of the speaker and then pointing to it as if it were visible even though it is not in the actual discourse context. We will see that such occurrences of ku encode a type of speaker affect and the split-DP analysis advanced here provides a way to formally derive it even though it is a largely pragmatic phenomenon.

5.4.1

The Cognitive Indexical Properties of Ku at Hand

In Chap. 4, I argued for a more nuanced analysis of DEMs than the existing analyses by suggesting that there can be more than one merge site for them. And in this context, I demonstrated that English DEMs can carry at least four different types of meaning, as exemplified in (109)–(112). (109) Illustration of a definite and indexical DEM in English: a. Who is this stranger here? b. That is John. c. These/Those are socks.

5.4 Why Co-occurring with Ku ‘the/that’ Engenders a Non-canonical ADJ Order

209

(110) Illustration of an indefinite/specific DEM in English: a. I went to the mall the other day, and there was this strange man talking really loudly in the shoes section. b. (Context: During a fund-raising season, a radio program host is talking to the listeners.) Please pick up that phone and call us now! (111) Illustration of a definite and contrastive-focused DEM in English: Which book do you want? THIS one or THAT one? Or THESE or THOSE? (112) Illustration of a definite, anaphoric, and unfocused DEM in English: a. In my neighborhood, there is [a really tall building]i. And thisi building has all sorts of interesting and cool stores inside. b. Once upon a time, there lived [a lovely little princess]i. Thati little princess was mostly happy but needed a friend.

Korean DEMs can also encode similar meanings, as illustrated in (113)–(115).17 (113) Definite and indexical usage of DEMs in Korean: Context: A and B are looking at someone at a party. A: Ce saram-un nwukwu-i- -ni? That person-TOP who-COP-N.PST-Q.INFRML ‘Who is that person over there?’ B: *(Ku) saram-un Con-i- -ya. That person-TOP John-COP-N.PST-DECL.INFRML ‘That person is John.’ (114) Definite and contrastive-focused usage of DEMs in Korean: Context: A is a sales associate at a bookstore and B is a customer. A: I chayk-kwa ce chayk-cwung pro etten chayk-ul This book-and that book-between __ which book-ACC wonhasipnikka? want-CONN Q.HON ‘Between this book and that one over there, which one would you like?’ B: *(Ku) chayk-i- -yo. That book-COP-N.PST-ASSRT.HON ‘I’d like that book (i.e., the one that is close to you but far away from me).’

17

Interestingly, Korean lacks specific but indefinite DEMs which would correspond to the occurrences of this and that in data like (110). In the literature, English DEMs which occur in contexts like (110) have been characterized as performing an ‘affective’ function [a term due to Liberman (2008)] (see, e.g., Potts and Schwarz 2010; Acton and Potts 2014; and the references there). According to Potts and Schwarz (2010: 2), English speakers use DEMs affectively to “foster a sense of closeness and shared sentiment” with other discourse participants and such a pragmatic function is most likely a universal property of DEMs. Given this, the fact that Korean DEMs do not perform the same function as the English DEMs in (110) raises an interesting question of why that might be the case. I take up this question in Chap. 6.

210

5 Capturing the Korean Facts

(115) Definite, anaphoric, and unfocused usage of DEMs in Korean: Context: The speaker is telling a story to a child. Yes-nal-yes-nal han yes-nal-ey etten maul-ey Old-day-old-day big old-day-LOC some village-LOC [acwu chakha-n sonye-ka han-myeng]i sal-ko one-CL] live-CONN [very good.hearted-ADN little.girl-NOM iss-ess-eyo. exist-PST-DECL.HON ‘Once upon a time, there lived a very good-hearted young girli in some village.’ Enu-nal *(kui) (chakhan) sonye-nun chinkwu cip-uro house-to Some-day that (good.hearted) little.girl-TOP friend nol-re ka-ss-eyo. play-CONN go-PST-DECL.HON ‘One day thati (good-hearted) girl went to a friend’s house to play.’

But the DEM ku that occurs in (105b)–(108b) (henceforth KU for ease of reference) cannot be adequately characterized by using labels like ‘definite and indexical’, ‘definite and contrastive-focused’, and ‘definite, anaphoric, and unfocused’. One reason for this difficulty comes from the fact that while other occurrences of ku modify a predicate-level expression as shown in (113)–(115), KU modifies an already referring expression. To see this, consider (116). Comparing the (a) and the (b) examples in this paradigm shows that while the more “ordinary” DEM ku cannot modify a proper name, what I call KU can; Korean speakers’ intuition here is that KU and the AP that follows it together act like a non-restrictive N modifier that stands in an appositive relation to the DP they modify. Compare now the (c) and the (d) examples in (116). This time the comparison shows that, unlike ku, KU can co-occur with what appears to be a supplementary RC. Evidence that KU helps modify an already referring expression: (116) Context: The speaker and the hearer have a mutual friend named Mina, and she used to date Chelwsu, who is well-known to be arrogant and rude. The speaker tells the hearer that she accidentally had dinner with Chelswu last night. Upon hearing it, the hearer says out of disbelief, “Who did you say you had dinner with?” In response, the speaker says one of the following. a. ku > Proper Name: [DP (*ku) Chelswu] [ (that) C.] Intended: ‘Chelswu’ b. KU + UN-AP > Proper Name: [DP Chelswu] ku [AP omanpwulsonha-n] KU [ arrogant.and.rude-ADN] [ C.] ‘that arrogant and rude Chelswu’ c. Sppl-RC > ku > Proper Name: [DP ku Chelswui ] *[RC Mina-ka ei sakwi-te-]-n, [ M.-NOM __ date-RTRO-]-REL, [ that C.] Intended: ‘that Chelswu, who Mina used to date’ d. Sppl-RC > KU + UN-AP > Proper Name: ku [AP omanpwulsonha-n] [RC Mina-ka ei sakwi-te-]-n, [ M.-NOM __ date-RTRO-]-REL, KU [ arrogant.and.rude-ADN] [DP Chelswui ] [ C.] ‘that arrogant and rude Chelswu, who Mina used to date’

5.4 Why Co-occurring with Ku ‘the/that’ Engenders a Non-canonical ADJ Order

211

I should note that in data like (105b)–(108b), KU does not occur preceding a proper name but even in such cases, the characterization I have just offered fits its behavior. This is illustrated by (117). In this paradigm, what KU modifies is already a definite and referring DP (which is due to the presence of an episodic RC inside of the DP18). And here too, the ‘KU + AP’ string contributes an appositive modificational meaning, as indicated by the English translations for (117b, c). (117) Context: A is looking for something. B asks whether she (i.e., A) has lost something. And A says ‘yes’, and adds that what she has lost is: a. Definite NP: [DP [RC Mina-ka pro mokkerii] ei sa-cwu- ]-n [ [ M.-NOM __ __ buy-give-PRF]-REL necklace] ‘the necklace that Mina has bought for me’ b. KU + UN-AP > Definite NP: ei sa-cwu- ]-n ku [AP kappissa-n]*(,) [DP [RC Mina-ka pro KU [ expensive-ADN], [ [ M.-NOM __ __ buy-give-PRF]-REL mokkerii] necklace] ‘that expensive necklace that Mina has bought for me’ c. Sppl-RC > KU + UN-AP > Definite NP: ei puracil-eyse mantul- ]-n*(,) [ku [AP kappissa-n]]*(,) [RC pro make-PRF]-REL, [KU [ expensive-ADN]], [ __ __ Brazil-LOC Mina-ka pro ei sa-cwu- ]-n mokkerii] [DP [RC [ [ M.-NOM __ __ buy-give-PRF]-REL necklace] ‘that expensive necklace that Mina has bought for me, which was made in Brazil’

Let me now point out that unlike more “ordinary” Korean DEMs, KU can occur recursively. To see this, compare (118) with (119)–(120). Non-recursivity of “ordinary” occurrences of DEMs in Korean: (118) a. (*i,) (*i,) i moca This, this, this hat Intended: ‘this hat’ b. (*ku,) (*ku,) ku moca That, that, that hat Intended: ‘that hat’

18

Without the RC, the N mokkeri in (117a) will be construed as referring to some or any non-specific necklace (i.e., ‘a necklace’) though in some contexts (e.g., if some necklace has already been introduced to the discourse), it may be construed as referring to some specific/definite necklace (i.e., ‘the necklace’).

212

5 Capturing the Korean Facts

Recursivity of ‘KU + AP’ strings: (119) a. KU + AP > Proper Name: [ku [AP omanpwulsonha-n]] [DP Chelswu] [ arrogant.and.rude-ADN]] [ C.] [KU ‘that arrogant and rude Chelswu’ b. KU + AP > KU + AP > Proper Name: [AP yoksimmahn-un]] [ku [AP omanpwulsonha-n]]*(,) [ku [KU [ arrogant.and.rude-ADN]], [KU [ greedy-ADN]] [DP Chelswu] [ C.] ‘that arrogant and rude, and that greedy Chelswu’ c. KU + AP > KU + AP > KU + AP > Proper Name: [AP yoksimmahn-un]]*(,) [ku [AP omanpwulsonha-n]]*(,) [ku [KU [ arrogant.and.rude-ADN]], [KU [ greedy-ADN]], [DP Chelswu] [ku [AP sengcilkoyakha-n]] [KU [ ill-tempered-ADN]] [ C.] ‘that arrogant and rude, that greedy, and that ill-tempered Chelswu’ (120) a. KU + AP > Definite NP: [ku [AP kappissa-n]]*(,) [DP [RC Mina-ka ei sa-o- ]-n [ expensive-ADN]], [ [ M.-NOM __ buy-come-PRF]-REL [KU mokkerii] necklace] ‘that expensive necklace that Mina has bought and brought (for me)’ b. KU + AP > KU + AP > Definite NP: [ku [AP kwiha-n]]*(,) [ku [AP kappissa-n]]*(,) [ expensive-ADN]], [KU [ precious-ADN]] [KU [DP [RC Mina-ka ei sa-o- ]-n mokkerii] buy-come-PRF]-REL necklace] [ [ M.-NOM __ ‘that expensive, and that precious necklace that Mina has bought and brought (for me)’ c. KU + AP > KU + AP > KU + AP > Definite NP: [ku [AP kwiha-n]]*(,) [ku [AP kappissa-n]]*(,) [KU [ expensive-ADN]], [KU [ precious-ADN]], [DP [RC Mina-ka ei sa-o- ]-n [ku [AP yeppu-n]]*(,) [ pretty-ADN]], [ [ M.-NOM __ buy-come-PRF]-REL [KU mokkerii] necklace] ‘that expensive, that precious, and that pretty necklace that Mina has bought and brought (for me)’

Yet another reason to think that KU is a different kind of DEM than other occurrences of ku comes from the fact that it requires a different kind of discourse context than the others. To see this, first reconsider (113) and compare it with (121): (113B) contains ku and this sentence is uttered in a context where the speaker is looking at and/or pointing at someone. (121), on the other hand, contains KU and it

5.4 Why Co-occurring with Ku ‘the/that’ Engenders a Non-canonical ADJ Order

213

is occurring in a context where the referent of the KU-marked DP is not perceptible to S. Notably, (113) cannot be attested in a context where the discourse participants cannot see or hear the individual by the name of John; conversely, (121) cannot occur in a context where the speaker can see the items that she is referring to by using the noun mwulkentul. (113) Definite and indexical usage of DEMs in Korean: Context: A and B are looking at someone at a party. A: Ce saram-un nwukwu-i- -ni? That person-TOP who-COP-N.PST-Q.INFRML ‘Who is that person over there?’ B: Ku saram-un Con-i- -ya. That person-TOP John-COP-N.PST-DECL.INFRML ‘That person is John.’ (121) Context: Today Mina bought and brought some expensive items for the speaker and she left them in the kitchen. The speaker is currently in her bedroom, which is not next to the kitchen, so she cannot see the items that Mina brought for her. pro [DP [ku [AP kappissa-n], [RC Mina-ka ei sa-o- ]-n [ expensive-ADN] [ M.-NOM __ buy-come-PRF]-REL __ [ [KU mwulken-tuli]-ul eti-ey twue-yahal-kka? where-LOC keep-must-Q? item-PL-]-ACC ‘Where should I keep those expensive items that Mina bought and brought for me?’

Compare now (121) with (114), which is repeated below for convenience. Comparing these two discourses already shows that KU does not require a contrastive-focus context. But let me further point out that uttering KU in a contrastivefocus context actually yields pragmatic anomaly. This is shown in (122): here, the second utterance cannot be followed by the third sentence, which contains KU, even though the object DP bears the contrastive focus/contrastive topic marker -un. (114) Definite and contrastive-focused usage of DEMs in Korean: Context: A is a sales associate at a bookstore and B is a customer. A: I chayk-kwa ce chayk-cwung pro etten chayk-ul This book-and that book-between __ which book-ACC wonhasipnikka? Q.HON want-CONN ‘Between this book and that one over there, which one would you like?’ B: Ku chayk-i- -yo. That book-COP-N.PST-ASST.HON ‘I’d like that book (i.e., the one that is close to you but far away from me).’

214

5 Capturing the Korean Facts

(122) Context: The speaker has two siblings, Mina and Cinho. Today, both Mina and Cinho brought her some precious gifts. And she is now wondering where to store them without mixing them up and says the following. Onul Mina-ka [DP [AP acwu kappissa-n] mwulken-tul]-ul Today M.-NOM [ [ very expensive-ADN thing-PL]-ACC sao-ass-ta. buy-CONN come-PST-DECL ‘Today Mina bought and brought to me very expensive items.’ Kuriko Cinho-to [DP [AP acwu kappissa-n] mwulken-tul]-ul And C.-also [ [ very expensive-ADN thing-PL]-ACC sao-ass-ta. buy-CONN come-PST-DECL ‘And Cinho also bought and brought to me very expensive items.’ #Kurentey, pro [ku [AP kappissa-n], [DP [RC Mina-ka [ expensive-AND] [ [ M.-NOM But __ [KU ei sa-o- ]-n mwulken-tuli]-un eti-ey __ buy-come-PRF]-REL item-PL]-CONT.TOP where-LOC twue-yahal-ci cal morukeyss- -ta. well not.know-N.PST-DECL keep-must-CONN Intended: ‘But as for those (expensive) items that Mina bought and brought for me, I don’t know where to keep them.’

Based on the contrast between (113) and (121), one might think that KU has a built-in anaphoric meaning, and while it is true that what I call KU is typically used in anaphoric contexts, comparing it with a purely anaphoric ku reveals some important differences between them: while the anaphoric ku need not co-occur with an AP modifier, as indicated by the parentheses around chakhan ‘good-hearted’ in (115), KU requires a property-denoting AP to co-occur with it, as shown in (123).19 (115) Definite, anaphoric, and unfocused usage of DEMs in Korean: Context: The speaker is telling a story to a child. Yes-nal-yes-nal han yes-nal-ey etten maul-ey Old-day-old-day big old-day-LOC some village-LOC [acwu chakha-n sonye-ka han-myeng]i sal-ko one-CL] live-CONN [very good.hearted-ADN little.girl-NOM iss-ess-eyo. exist-PST-DECL.HON ‘Once upon a time, there lived a very good-hearted young girli in some village.’ Enu-nal *(kui) (chakhan) sonye-nun chinkwu cip-uro house-to Some-day that (good.hearted) little.girl-TOP friend nol-le ka-ss-eyo. go-PST-DECL.HON play-CONN ‘One day thati (good-hearted) girl went to a friend’s house to play.’

19

Note that sentence (123b) can be felicitously uttered in a context where there are multiple individuals who bear the name Chelswu but that is not the intended context here, as indicated in the context for (123).

5.4 Why Co-occurring with Ku ‘the/that’ Engenders a Non-canonical ADJ Order

215

(123) KU’s need to co-occur with an AP: Context: The speaker and the hearer have a mutual friend named Mina, and she used to date Chelwsu, who is well-known to be arrogant and rude. Yesterday, the speaker got involved in something that included Chelswu. Now she is informing the hearer about what happened to her yesterday by saying one of the following. Chelswu]-wa a. Na-nun ecey [ku [AP omanpwulsonha-n] I-TOP yesterday [KU [ arrogant.and.rude-ADN] C.]-with hamkkey siksa-rul ha-key-toy-ess-ta. together meal-ACC do-CONN-become-PST-DECL ‘Yesterday, I happened to dine with that awful arrogant and rude Chelswu.’ b. *Na-nun ecey [ku Chelswu]-wa hamkkey yesterday [KU C.]-with together I-TOP siksa-rul ha-key-toy-ess-ta. meal-ACC do-CONN-become-PST-DECL Intended: ‘Yesterday, I happened to dine with that Chelswu.’

What is also notable is that, unlike a purely anaphorically used ku, KU triggers a strong emotive implicature. For example, the ‘KU + AP’ string in (117b) signals that the speaker feels strongly about the fact that the necklace that Mina bought and brought for her is very expensive; by contrast, the anaphoric ku in (115) does not engender such a pragmatic meaning even if the AP is present in the sentence. This set of facts provides sufficient reasons to treat KU differently from other uses of ku but there are still other reasons to treat them differently. First of all, as already suggested by the badness of (123b), KU and the AP that occurs adjacent to it form a syntactic unit and this is further evidenced by the fact that deletion targets both of them, as illustrated in (124)–(128). Context for (124-125): The discourse participants are talking about Chelswu, whom their mutual friend Mina used to date but not any longer. (124) a. Mina-nun ecey [ku [AP omanpwulsonha-n] Chelswu]-wa M.-TOP yesterday [KU [ arrogant.and.rude-ADN] C.]-with hamkkey kuk.cang-ey ka-ss-ta. together movie.theater-to go-PST-DECL ‘Yesterday, Mina went to the movies with that arrogant and rude Chelswu.’ b. *Mina-nun ecey [ku Chelswu]-wa hamkkey M.-TOP yesterday [KU C.]-with together kuk.cang-ey ka-ess-ta. movie.theater-to go-PST-DECL Intended: ‘Yesterday, Mina went to the movies with that Chelswu.’ c. *Mina-nun ecey [[AP omanpwulsonha-n] Chelswu]-wa yesterday [[ arrogant.and.rude-ADN] C.]-with M.-TOP hamkkey kuk.cang-ey ka-ess-ta. together movie.theater-to go-PST-DECL Intended: ‘Yesterday, Mina went to the movies with arrogant and rude Chelswu.’

216

5 Capturing the Korean Facts

(125) [ku *([AP omanpwulsonha-n])], [ku *([AP yoksimmahn-un])], [ arrogant.and.rude-ADN]], [KU [ greedy-ADN]], [KU [ku *([AP sengcilkoyakha-n]]) [DP Chelswu] [ ill-tempered-ADN]] [ C.] [KU Intended: ‘that (arrogant and rude), that (greedy), and that (ill-tempered) Chelswu’ Context for (126-127): Today Mina bought and brought an expensive necklace for the speaker. The speaker is recalling that event now, expressing how pleased she is that the necklace that Mina brought for her is an expensive one. [DP [RC Mina-ka ei sa-o- ]-n (126) a. ku [AP kappissa-n], KU [ expensive-ADN] [ [ M.-NOM __ buy-come-PRF]-REL mokkerii] necklace] ‘that expensive necklace that Mina bought and brought (for me)’ b. *[AP kappissa-n], [DP [RC Mina-ka ei sa-o- ]-n buy-come-PRF]-REL [ expensive-ADN], [ [ M.-NOM __ mokkerii] necklace] 20 Intended: ‘that expensive necklace that Mina bought and brought’20 (127) [ku *([ AP kappissa-n])], [ku *([AP kwiha-n])], [KU [ expensive-ADN]], [KU [ precious-ADN]], yeppu-n])], [DP [RC Mina-ka ei sa-o- ]-n [ku *([ AP [KU [ pretty-ADN]], [ [ M.-NOM __ buy-come-PRF]-REL mokkerii] necklace] Intended: ‘that (expensive), that (precious), and that (pretty) necklace that Mina bought and brought (for me)’ (128) Context: Speaker is telling a story about some particular tyrant who oppressed democracy for a long time. [RC ei mincwucwuuy-rul thanapha- ]-n a. [ku [AP tokcay-cek]], [ tyrant-CEK]] [ __ democracy-ACC oppress-PRF]-REL [ KU kwuncwui ‘that tyrannical monarch who oppressed democracy’ b.*[ [AP tokcay-cek]], [RC ei mincwucwuuy-rul thanapha- ]-n kwuncwui Intended: ‘that tyrannical monarch who oppressed democracy’

(126b) can be judged acceptable if it is construed as denoting ‘the expensive necklace that Mina bought and brought, that is, if it is interpreted as simply referring to—without the emotive connotation—some discourse prominent necklace about which the fact that it is expensive and it was bought and brought by Mina has been established in the discourse context. 20

5.4 Why Co-occurring with Ku ‘the/that’ Engenders a Non-canonical ADJ Order

217

Secondly, paradigms (129) and (130) show that KU and its co-occurring AP have to move together, and this provides additional evidence for their constituency.21 (129) a. [ku [AP omanpwulsonha-n]]*(,) [ku [AP yoksimmahn-un]]*(,) [KU [ arrogant.and.rude-ADN]], [KU [ greedy-ADN]], [DP Chelswu] [ku [AP sengcilkoyakha-n]] [KU [ ill-tempered-ADN]] [ C.] ‘that arrogant and rude, that greedy, and that ill-tempered Chelswu’ b. [ku [AP yoksimmahn-un]], [ku [AP omanpwulsonha-n]], [ greedy-ADN]], [KU [ arrogant.and.rude-ADN]], [KU [DP Chelswu] [ku [AP sengcilkoyakha-n]] [ ill-tempered-ADN]] [ C.] [KU ‘that greedy, that arrogant and rude, and that ill-tempered Chelswu’ c. *kui, [ku [AP omanpwulsonha-n]], [ti [AP yoksimmahn-un]], KU, [KU [ arrogant.and.rude-ADN]], [ [ greedy-ADN]], [DP Chelswu] [ku [AP sengcilkoyakha-n]] [ ill-tempered-ADN]] [ C.] [KU Intended: ‘that arrogant and rude, that greedy, and that ill-tempered Chelswu’ d. *[AP yoksimmahn-un]i], [ku [AP omanpwulsonha-n]], [ku ti], [ greedy-ADN]], [KU [ arrogant.and.rude-ADN]], [KU ], [DP Chelswu] [ku [AP sengcilkoyakha-n]] [KU [ ill-tempered-ADN]] [ C.] Intended: ‘that arrogant and rude, that greedy, and that ill-tempered Chelswu’ (130) a. [ku [AP1 kappissa-n]], [ku [AP2 kwiha-n]], [ expensive-ADN]], [KU [ precious-ADN]], [KU [ku [AP3 yeppu-n]], [DP [RC Mina-ka ei sa-o- ]-n [ pretty-ADN]], [ [ M.-NOM __ buy-come-PRF]-REL [KU mokkerii] necklace] ‘that expensive, that precious, and that pretty necklace that Mina bought and brought (for me)’ b. [ku [AP2 kwiha-n]], [ku [AP1 kappissa-n]], [ precious-ADN]], [KU [ expensive-ADN]], [KU [DP [RC Mina-ka ei sa-o- ]-n [ku [AP3 yeppu-n]], [ pretty-ADN]], [ [ M.-NOM __ buy-come-PRF]-REL [KU mokkerii] necklace] ‘that precious, that expensive, and that pretty necklace that Mina bought and brought’

21

The data in (129)–(130) may all sound grammatical if they are just read out loud because some of the free-standing occurrences of ku can be construed as the ordinary DEM ku, but as indicated in the English translations, each occurrence of ku here is meant to modify the property that is denoted by the AP that it co-occurs with.

218

5 Capturing the Korean Facts

c. *[ku j, [ku [AP1 kappissa-n]], [tj [AP2 kwiha-n]], [KU, [KU [ expensive-ADN]], [ [ precious-ADN]], [DP [RC Mina-ka ei sa-o- ]-n [ku [AP3 yeppu-n]], [KU [ pretty-ADN]], [ [ M.-NOM __ buy-come-PRF]-REL mokkerii] necklace] Intended: ‘that expensive, that precious, and that pretty necklace that Mina bought and brought (for me)’ [ku [AP1 kappissa-n]], [ku tj], d. *[[AP2 kwiha-n] j, [[ precious-ADN]], [KU [ expensive-ADN]], [KU ], [DP [RC Mina-ka ei sa-o- ]-n [ku [AP3 yeppu-n]], [ pretty-ADN]], [ [ M.-NOM __ buy-come-PRF]-REL [KU mokkerii] necklace] Intended: ‘that expensive, that precious, and that pretty necklace that Mina bought and brought (for me)’

Finally, as exemplified by all the grammatical data containing a ‘KU + AP’ string above, KU and its accompanying AP occur without a pause between them. Furthermore, if there is another N modifier co-occurring with them, it needs to be phonologically set apart from them, as indicated by the commas given above. And together, these phonological properties provide yet additional evidence that KU and the AP that follows it form a syntactic constituent. To summarize thus far, then, what I call KU differs from other occurrences of ku in the following ways: (i) it modifies an already referential DP; (ii) it does not require a physical pointing or a contrastive-focus context to be licensed; (iii) it selects for a property-denoting AP; and (iv) they form a prosodic unit as well as a syntactic unit. In light of these findings, the data in (105b)–(108b) can be interpreted slightly differently than the way we did earlier: in Chap. 3, we took them to suggest that co-occurring with KU may let certain light APs occur preceding heavier N modifiers. While such a characterization is not exactly incorrect, given what we have uncovered here, a more accurate way to state the facts would be that: KU takes a property-denoting expression as its complement and together, they form a single FP, and for this reason, any property-denoting category may occur immediately after it without any pause in-between, and this gives rise to a surface order that is seemingly in violation of the weight-based AOR given in (1a). Under this revised way of looking at things, KU is best analyzed as some sort of ‘affective’ DEM in the sense of Liberman (2008) since the reason why it behaves in the way it does is that its occurrence in a discourse lets the speaker comment on some discourse familiar entity that has a noteworthy property. More specifically, it indicates that the degree to which the property denoted by its complement holds

5.4 Why Co-occurring with Ku ‘the/that’ Engenders a Non-canonical ADJ Order

219

true of the referent denoted by its hosting DP is very high and the speaker can visualize it even though she cannot actually see the discourse referent at the time/ location of the discourse (compare Kang 201822). To further establish this by way of exemplification, notice that in (121), the RC non-restrictively modifies the referent of the DP, i.e., the items that Mina bought and brought for the speaker, and the presence of KU engenders three implicatures: (i) the items Mary bought and brought for the speaker are highly expensive; (ii) the

22 At the final stage of writing this book, I came to learn that Kang (2018) makes similar observations about certain occurrences of the distal DEM ku in Korean and claims that such occurrences of ku are best analyzed as a scalar, additive focus modifier which is comparable to English even. To illustrate this, she discusses how the presence of ku in data like (i) encodes the speaker’s unexpected surprise, and offers (ii) as the meaning of ku in (i), where C represents a contextually constructed subset of alternatives.

(i) Context: Yesterday, Mary was invited to Kim’s place. Kim made dinner with tofu. Today, Ann asks Mary how the dish tasted. Mary tells Ann that the cooking was bad. Mary was surprised at the fact that the dish made with tofu could be unsavory because she likes tofu and tofu is delicious. Mary says: masiss-nun twupwu-ro (kim-i) KUH% delicious-REL tofu-with Kim-NOM KUFOCUS masep-nun yori-rul mantul-ess-e. dish-ACC cook-PST-DECL unsavory-REL ‘(It is unbelievable that) Kim cooked an unsavory dish with that much delicious tofu.’ [Kang 2018: 604, (11a)] (ii)

y[y tofu C(y) Kim cooked an unsavory dish with delicious y] y[y tofu likelihood(Kim cooked an unsavory dish with delicious y) > likelihood(Kim cooked an unsavory dish with delicious tofu)] [Kang 2018: 604, (11b)]

However, Kang mainly discusses the occurrence of ku in exclamative contexts and how it encodes mirativity in the sense of DeLancey (1997). Furthermore, as far as I can see, the occurrences of KU we are concerned with here have nothing to do with additivity nor do they evoke alternatives and then rank them on a scale, which is what a scalar focus modifier would do (see, a.o., Traugott 2006 and the references therein). To take (126a) in the text for example, by uttering it, the speaker does not implicate that there are other (expensive) items that Mina bought and brought for her and the necklace at issue is the least likely one that Mina was expected to buy and bring for her. In fact, such an implicature is impossible to draw for this datum. Given this, while I do see that the data that Kang discusses are bound to receive mirative interpretations, and I also agree with her that the meaning of non-physically deictic, non-anaphoric, and seemingly emphatically used ku has something to do with a scale, I believe that her analysis and mine are rather different in nature and scope although future research may show that a uniform analysis is possible for the occurrences of ku that she talks about and those that I do here. Moreover, I suspect that the pragmatic effects that Kang talks about may arise not (just) because of the presence of the DEM in the sentences but because of the contrastive meaning that is literally conveyed by the sentences themselves, i.e., the mismatch between the speaker’s expectations about some state of affairs and their actual realizations. For instance, in (i) above, the speaker expects any dish with tofu to be delicious but the dish that Kim cooked was not, and hence the surprise on her part.

220

5 Capturing the Korean Facts

speaker thinks it is worth mentioning it; and (iii) she can visualize how the items that Mina brought for her are highly expensive even though she cannot see them at the time of speech. To look at (124a) for comparison, in this case, there is no RC but the presence of KU brings about similar effects: it implicates that (i) Chelswu is highly arrogant and rude; (ii) the proposition that he is highly arrogant and rude is deemed noteworthy by the speaker; and (iii) she can picture the way Chelswu is so highly arrogant and rude even though she does not see him when uttering the sentence. Notably, the goodness of (131) below shows that even when the subject of the sentence is someone other than the speaker, a ‘KU + AP’ string contributes something that can be best analyzed as a speaker comment. That is, in the first sentence of this discourse, the person who thinks that Chelswu is extremely arrogant and rude is the speaker, not Mina, and she can visualize it when she utters the sentence; otherwise, the discourse cannot be judged coherent since the second sentence explicitly states that Mina does not hold the same belief as the speaker. (131) Context: The discourse participants have a mutual friend named Chelswu, who is well-known to be full of himself and rude. They also know someone by the name of Mina. And the speaker is telling the hearer about what Mina did yesterday. Mina-nun ecey [ku [AP omanpwulsonha-n] Chelswu]-wa yesterday [KU [ arrogant.and.rude-ADN] C.]-with M.-TOP hamkkey kukcang-ey ka-ss-ta. together movie.theater-to go-PST-DECL ‘Yesterday, Mina went to the movies with that arrogant and rude Chelswu.’ Na-nun oay Mina-nun Chelswu-ka omanpwulsonha-ta-ko why M.-CONT.TOP C.-NOM proud.and-rude-IND-COMP I-TOP sayngkakha-ci anh-nun-ci cal morukeyess- -ta. do.not-IMPRF-CONN well not.know-N.PST-DECL think-CONN ‘I don’t understand why Mina doesn’t think that Chelswu is arrogant and rude (when everybody else does).’

Similarly, the pragmatic anomaly of (132) shows that the speaker cannot deny what is encoded by the ‘KU + AP’ string that is present in the sentence she has just uttered. (132) Context: Same as (131). Mina-nun ecey [ku [AP omanpwulsonha-n] Chelswu]-wa M.-TOP yesterday [KU [ arrogant.and.rude-ADN] C.]-with hamkkey kukcang-ey ka-ss-ta. together movie.theater-to go-PST-DECL ‘Yesterday, Mina went to the movies with that arrogant and rude Chelswu.’ #Kurentey, na-nun Chelswu-ka omanpwulsonha-ta-ko But, I-TOP C.-NOM proud.and-rude-IND-COMP sayngkakha-ci anh-nun-ta. do.not-IMPRF-DECL think-CONN ‘But I don’t think that Chelswu is arrogant and rude.’

5.4 Why Co-occurring with Ku ‘the/that’ Engenders a Non-canonical ADJ Order

221

Collectively, this set of facts leads us to conclude that a ‘KU + AP’ string encodes speaker affect even though there is no doubt that it occurs within a DP boundary, given the ungrammaticality of the data in (133): (133a, b) are ungrammatical in contrast to sentences like (124a) because here, KU is occurring to the left of the temporal adverbial ecey ‘yesterday’ or onul ‘today’, which has to be adjoined at the sentential level. (Note that moving both KU and the AP across the temporal ADVs will yield equally ungrammatical sentences.) (133) a. *Mina-nun kui, ecey [DP ti [AP omanpwulsonha-n] M.-TOP KU, yesterday [ [ arrogant.and.rude-ADN] Chelswu]-wa hamkkey kukcang-ey ka-ss-ta. C.]-with together movie.theater-to go-PST-DECL Intended: ‘Yesterday Mina went to the movies with that arrogant and rude Chelswu.’ [DP tj [AP kappissa-n], b. *Mina-nun kuj, onul M.-TOP KU, today [ [ expensive-ADN] [RC Cinho-ka na-hanthey ei kacie-o- ]-n mwulken-tuli]-ul [ C.-NOM I-to __ bring-come-PRF]-REL item-PL-]-ACC pwuek-ey noh-a-twu-ess-ta. kitchen-LOC leave-CONN-AUX-PST-DECL Intended: ‘Today Mina left those expensive items that Cinho brought for me in the kitchen.’

In sum, then, placing a ‘KU + AP’ string in front of a DP is a strategy that Korean speakers may utilize to encode a certain attitude toward an entity that is known to him/her. Importantly, however, KU is licensed only when the individual it helps to comment on is not perceptible to the speaker and this suggests that it operates in a cognitive domain rather than in a physical domain (more on this in Chap. 6, Sect. 6.3). Given this, I posit that KU locates some property of individuals in the cognitive domain of the speaker and points to how highly that property holds true of some discourse-salient entity. I also posit that from such cognitive and scalar characteristics of KU, the noteworthiness implicature associated with the property denoted by its complement falls out, as well as the intuition that the speaker can visualize the property holding true of the entity at issue even though she cannot see the entity at the time of discourse. Viewed this way, then, there is little doubt that KU performs some sort of deictic/indexical function even though the domain in which it operates is not a physical one. That said, this lexeme is clearly not a run-of-the-mill indexical element, so from now on, I call it a ‘cognitive indexical marker’ (for lack of a better term).

5.4.2

Formally Capturing the “Special” Properties of KU

In the previous subsection, we have seen that KU (i) selects for an AP, (ii) forms a prosodic unit with it, (iii) modifies an already referential DP, (iii) operates in the

222

5 Capturing the Korean Facts

cognitive domain, and (iv) is invariably linked to the speaker, yet (v) it is syntactically located inside a DP boundary. To derive such properties of the DEM within a formal syntactic framework, I first propose that the FP that is headed by KU, which I henceforth call ‘Cognitive DemP’ (CogDemP), has the internal structure given in (134). In this structure, the head position of the CogDemP under which KU is base-generated always hosts a feature that I notate as ‘[+STS distant]’. Consequently, whatever node the FP c-commands denotes an entity that is spatio-temporally distant from the speaker, which is what [+STS distant] in fact stands for.23 In addition, all three levels of the CogDemP here has the subscript S, which is short for ‘speaker’, and this is meant to indicate that whatever KU and its extended projection modify is known to the speaker even if it may not be to the hearer. (134) Internal structure of a ‘KU + AP’ string: CogDemPS CogDem S CogDemS [+STS DISTANT] KU

AP

As for the external syntax, I hypothesize that a CogDemP hosting a ‘KU + AP’ string merges at an adjoined Spec position of a [+referential; +deictic; +definite; +specific] DPd/r, as schematized in (135a), or at the pre-existing Spec position of a [+referential; +deictic; +definite; +specific] LocP, as pictured in (135b). The first possibility is available because a ‘KU + AP’ string acts like an appositive N modifier, and in the syntactic framework assumed here, any integrated non-restrictive N modifier may occur in an adjoined [Spec, DPd/r] position. The second possibility is also available, however, because any type of DemP should be able to merge at [Spec, LocP]. Additionally, I posit that any DPd/r containing a ‘KU + AP’ string is construed as referring to an individual that is known to S but is absent in the discourse context because the CogDemP headed by KU always c-commands the head of that DP in the surface structure, as we will see shortly.

23

In Chap. 6, I revisit the semantics of DEMs and dissect their features. Under the more detailed analysis offered there, what I refer to as [+STS distant] will be comprised of two different features. But for now, what we have here will suffice.

5.4 Why Co-occurring with Ku ‘the/that’ Engenders a Non-canonical ADJ Order

223

(135) External syntax of a ‘KU + AP’ string: a. When a ‘KU + AP’ string is adjoined to a DPd/r: DP d/r … DPd/r CogDemPS

DPd/r

[+STS DISTANT] KU + AP

Dd/r Dd/r [+REF] [+DEIC] [+DEF] [+SPEC]

LocP Loc Loc [+REF] [+DEIC] [+DEF] [+SPEC]

UnitP/PlP

b. When a ‘KU + AP’ string is base-generated at [Spec, LocP]: DPd/r Dd/r LocP Dd/r [+REF] CogDemPS [+DEIC] [+DEF] Loc [+SPEC] [+STS DISTANT] KU + AP [+REF] [+DEIC] [+DEF] [+SPEC]

Loc UnitP/PlP

Positing such an external syntax for a CogDemP lets us immediately account for the distribution of a ‘KU + AP’ string relative to other N modifiers. First of all, we can now explain why it invariably occurs following a seemingly Sppl-RC if there is any.24 To see this, consider (136). In this discourse, the first sentence can

In Chap. 6 (Sect. 6.5), I take up the question of whether Korean has “truly” supplementary RCs or not, and I show that what we call Sppl-RCs here are actually not prototypical Sppl-RCs when compared to English Sppl-RCs such as those presented in Chap. 4. That said, treating the RCs in data like (136) as Sppl-RCs for the time being will not change things for us. Hence, for referential convenience, I will keep using this label, deferring a more refined analysis of Korean RCs until later.

24

224

5 Capturing the Korean Facts

be continued by (a) but not by (b), which respectively instantiate ‘Sppl-RC > KU + AP’ and ‘KU + AP > Sppl-RC’ orders. The RCs here can be analyzed as supplementary because they add discourse-new information that is not at issue, which is how Potts (2003/2005) characterizes supplementary or parenthetical N modifiers. Besides, their contents are not conveyed as an integral part of the larger message conveyed, which is how Huddleston and Pullum (2005: 187) define supplementary or parenthetical N modifiers. (136) Context: Speaker is telling her friend what happened to her today. Onul Mina-ka [DP [AP acwu kappissa-n] mokkeri]-rul Today M.-NOM [ [ very expensive-ADN] necklace]-ACC pro semwul-ro kaci-ko o-ass-e. come-PST-DECL.INFRML __ present-as have-CONN ‘Today Mina brought me a very expensive necklace as a present.’ a. Kurentey na-nun [DP [Sppl-RC Minswu-ka puracil-eyse ei But I-TOP [ [ M.-NOM Brazil-LOC __ sa-o-ass-ta-ko-ha-]-nun, [ku [AP kappissa-n]] buy-come-PST-IND-COMP-QUOT-]-REL, [KU [ expensive-ADN]] mokkerii]-rul pangkum il-e peri-ess-e. just.now lose-CONN AUX-PST-DECL.INFRML necklace]-ACC ‘But I just lost that expensive necklace, which I hear that Minswu bought and brought from Brazil.’ (Sppl-RC > KU + AP) b. #Kurentey na-nun [DP [ku [AP kappissa-n]], [Sppl-RC Minswu-ka puracil-eyse ei sa-o-ass-ta-ko-ha-]-nun mokkerii]-rul pangkum il-e peri-ess-e. Intended: ‘But I just lost that expensive necklace, which I hear that Minswu bought and brought from Brazil.’ (KU + AP > Sppl-RC)

Under the present analysis, the fixed order between a Sppl-RC and a ‘KU + AP’ string obtains because while a Sppl-RC merges DP-externally, as depicted in (2), being an NRS but integrated N modifier which contributes something that is presented as an integral part of the larger message, a ‘KU + AP’ string merges inside a DP boundary. Hence, we can posit (137) as the syntactic derivation of the object DP in (136a). Notably, this derivation correctly captures the way the DP is pronounced: the RC is pronounced with a pause after it because it merges at [Spec, SpplP] and the head of a SpplP hosts a pause, given our assumptions laid out in Chap. 4. The CogDemP is not pronounced with a pause after it, on the other hand, and this is because it does not occupy an adjoined position. Notice also that in this configuration, the CogDemP surfaces at [Spec, DPd/r], so all its features get inherited by the DP, making it construed as denoting an individual that is known to S but is absent in the discourse context.

5.4 Why Co-occurring with Ku ‘the/that’ Engenders a Non-canonical ADJ Order

225

(137) The internal structure of the object DP in (136a): SpplP RC

Sppl

Sppl Minswuka … , saoasstakohanun CogDemPS1 [+STS DISTANT] ku kappissan

DPd/r Dd/r Dd/r [+REF] [+DEIC] [+DEF] [+SPEC]

LocP t1

Loc Loc [+REF] [+DEIC] [+DEF] [+SPEC]

UnitP/PlP

mokkeri

The proposed analysis also lets us readily handle cases where multiple Sppl-RCs co-occur with a ‘KU + AP’ string. To see this, consider (138). (138) Context: Speaker is telling her friend what happened to her today by uttering the following sentences in the order they occur. a. Ecey Mina-ka [DP [AP acwu kappissa-n] mokkeri]-rul Yesterday M.-NOM [ [ very expensive-ADN] necklace]-ACC pro senmwul-ro kaci-e o-ess-e. come-PST-DECL.INFRML __ present-as have-CONN ‘Yesterday Mina brought me a very expensive necklace as a present.’ d. Kurayse na-nun nemwuna kippu-ess-e. Therefore I-TOP very.much be.happy-PST-DECL.INFRML ‘So I was so thrilled.’ [DP [Sppl-RC cinan-cwu-ey Minswu-ka ei c. Kurentey na-nun But I-TOP [ [ last-week-LOC M.-NOM __ [Sppl-RC ecey puracil-eyse sa-o-ass-ta-ko-ha-]-nun, yesterday Brazil-from buy-come-PST-IND-COMP-QUOT-]-REL, [ Mina-ka ei newmuwna yeppukey phocanghay-se pro really beautifully gift.wrapped-CONN __ M.-NOM __ kacie-o- ]-n, [Sppl-RC nay-ka phyengso-ey ei bring.CONN-come-PRF]-REL, [ I-NOM ordinary.time-LOC __ newmuwna kac-ko siph-ess-te-]-n [ku [AP kappissa-n]] really possess-CONN want-ANT-RTRO-]-REL [KU [ expensive-ADN]] mokkerii]-rul onul il-e peri-ess-e. today lose-CONN AUX-PST-DECL.INFRML necklace]-ACC ‘But today, I lost that expensive necklace, which I hear that Minswu bought and brought from Brazil last week, which Mina brought to me yesterday after giftwrapping it so beautifully, which I had been so wanting to have for so long.’ (Sppl-RC > Sppl-RC > Sppl-RC > KU + AP > N)

226

5 Capturing the Korean Facts

Under the present analysis, data like (138c) are expected to be attested because Sppl-RCs can be recursively added to the DP modified by a ‘KU + AP’ string by creating an adjunction structure, as schematically given in (139). (139) Structure of the object DP in sentence (138c): SpplP RC ei

SpplP RC ei

SpplP RC ei CogDemPS

ku kappissan

DPd/r Dd/r

mokkerii

The analysis proposed here provides us with a straightforward account of the relative order between a ‘KU + AP’ string and an NRS-RC as well. To see this, consider first (140). Here, the first sentence can be followed by either the (a) or the (b) sentence, unlike the case with (136), and this shows that a ‘KU + AP’ string may occur either before or after an NRS-RC.25 (140) Context: Speaker is telling her friend what happened to her today. Onul Mina-ka [DP [AP acwu kappissa-n] mokkeri]-rul Today M.-NOM [ [ very expensive-ADN ] necklace]-ACC pro senmwul-ro kaci-ko o-ass-e. come-PST-DECL.INFRML __ present-as have-CONN ‘Today Mina brought me a very expensive necklace as a present.’ a. Kurentey na-nun [DP [ku [AP kappissa-n]], [NRS-RC Mina-ka ei But I-TOP [ [KU [ expensive-ADN]], [ M.-NOM __ mokkerii]-rul pangkum il-e kaci-e o- -]-n have-CONN come-PRF]-REL necklace]-ACC just.now lose-CONN peri-ess-e. AUX-PST-DECL.INFRML ‘But I just lost that expensive necklace that Mina brought for me.’ (KU + AP > NRS-RC) [NRS-RC Mina-ka ei kaci-e o- -]-n, b. Kurentey na-nun [DP [ku [AP kappissa-n]] mokkerii]-rul pangkum il-e peri-ess-e. ‘But I just lost that expensive necklace that Mina brought for me.’ (NRS-RC > KU + AP)

25

The RC here is a non-restrictive RC rather than a supplementary one because it essentially repeats what is conveyed by the preceding sentence rather than adding anything new to the discourse.

5.4 Why Co-occurring with Ku ‘the/that’ Engenders a Non-canonical ADJ Order

227

Under our analysis, both ‘KU + AP > NRS-RC’ and ‘NRS-RC > KU + AP’ orders are permitted for the following reasons: an NRS-RC appositively modifies an already established discourse referent, conveying information that is presented as an integral part of the larger message, so it targets a [Spec, DPd/r] position. Being an appositive integrated N modifier, a ‘KU + AP’ string may occur at a [Spec, DPd/r] as well, but being a DemP, it may also merge at [Spec, LocP]. Hence, in the case of (140), the (a) sentence obtains if the RC merges at the pre-existing [Spec, DPd/r] and then the CogDemP headed by KU merges at the [Spec, DPd/r] that is created atop the pre-existing DPd/r, as given in (141a). If the CogDemP is base-generated at [Spec, LocP] but raises to [Spec, DPd/r] to pronounce the D features and then the RC merges at an adjoined [Spec, DPd/r] after it, as given in (141b), then the (b) sentence obtains instead. Notably, on either scenario, there will be a pause between the two N modifiers because one of them will occupy an adjoined position and this accords with how sentences (140a, b) are pronounced. (141) a. Structure of the object DP in (140a): ‘KU + AP > NRS-RC > N’ surface order DPd/r CogDemPS

DPd/r

CogDem S CogDemS [+STS DISTANT] ku

NRS-RC

AP

Dd/r Dd/r [+REF] [+DEIC] [+DEF] [+SPEC]

ei kappissan

LocP

mokkerii

b. Derivation of the object DP in (140b): ‘NRS-RC > KU + AP > N’ surface order DPd/r NRS-RC

DPd/r Dd/r

ei Dd/r [+REF] [+DEIC] [+DEF] [+SPEC]

LocP CogDemPS

ku kappissan

Loc Loc [+REF] [+DEIC] [+DEF] [+SPEC]

UnitP/PlP

mokkerii

In addition to capturing the distribution of ‘KU + AP’ strings relative to NRS- or Sppl-RCs, the proposed analysis enables us to account for the syntactic and

228

5 Capturing the Korean Facts

phonological properties of data containing multiple occurrences of ‘KU + AP’ such as those given in (119)–(120). For example, the data in (119c) and (120c) can be derived as schematically represented in (142) and (143), respectively. (119) c. [ku [AP omanpwulsonha-n]]*(,) [ku [AP yoksimmahn-un]]*(,) [ arrogant.and.rude-ADN]], [KU [ greedy-ADN]], [KU [DP Chelswu] [ku [AP sengcilkoyakha-n]] [ ill-tempered-ADN]] [ C.] [KU ‘that arrogant and rude, that greedy, and that ill-tempered Chelswu’ (120) c. [ku [AP kappissa-n]]*(,) [ku [AP kwiha-n]]*(,) [ expensive-ADN]], [KU [ precious-ADN]], [KU [ku [AP yeppu-n]]*(,) [DP [RC Mina-ka ei sa-o- ]-n [ pretty-ADN]], [ [ M.-NOM __ buy-come-PRF]-REL [KU mokkerii] necklace] ‘that expensive, that precious, and that pretty necklace that Mina has bought and brought (for me)’ (142) Structure of (119c): ‘KU + AP > KU + AP > KU + AP > Proper Name’ DPd/r CogDemP

DPd/r

CogDemP

DPd/r

CogDemP

DPd/r

Chelswu (143) Structure of (120c): ‘KU + AP > KU + AP > KU + AP > NRS-RC > N’ DPd/r CogDemP

DPd/r

CogDemP

DPd/r

CogDemP

DPd/r

NRS-RC ei

Dd/r Dd/r [+REF] [+DEIC] [+DEF] [+SPEC]

LocP Loc Loc [+REF] [+DEIC] [+DEF] [+SPEC]

Unit/PlP mokkerii

5.4 Why Co-occurring with Ku ‘the/that’ Engenders a Non-canonical ADJ Order

229

Relatedly, the present analysis correctly predicts that an NRS-RC may occur intervening between multiple occurrences of ‘KU + AP’. To see this, consider (144), a variant of (130a). (144) [ku [AP kappissa-n]], [DP [RC Mina-ka ei sa-o- -]-n, [KU [ expensive-ADN]], [ [ M.-NOM __ buy-come-PRF]-REL], [ku [AP yeppu-n]] mokkerii] [ku [AP kwiha-n]], [KU [ precious-ADN]], [KU [ pretty-ADN]] necklace] ‘that expensive, that precious, and that pretty necklace that Mina bought and brought (for me)’

In our analysis, data like (144) obtain because an NRS-RC and a CogDemP both target [Spec, DPd/r] positions, so their surface positions may be altered if the NRS-RC that has merged at an adjoined [Spec, DPd/r] scrambles to a higher [Spec, DPd/r], skipping over the CogDemP that has merged at a higher adjoined [Spec, DPd/r] than the RC, and there is yet another CogDemP occupying the pre-existing [Spec, DPd/r], as shown in (145); alternatively, such a surface order as (144) can obtain by merging the lowest CogDemP at the pre-existing [Spec, DPd/r], then merging another CogDemP at the next [Spec, DPd/r] that is created atop the pre-existing [Spec, DPd/r], followed by merging the NRS-RC at the next [Spec, DPd/r] and another CogDemP at the next [Spec, DPd/r]. Data like (130a) obtain, on the other hand, if the same NRS-RC merges at the pre-existing [Spec, DPd/r] and all the three CogDemPs merge at adjoined [Spec, DPd/r] positions created above it, followed by no further movement. (145) Derivation of (144): ‘KU + AP > NRS-RC > KU + AP > KU + AP > N’ DPd/r CogDemP

DPd/r DPd/r CogDemP

DPd/r

NRS-RC ei

DPd/r

CogDemP1 Dd/r [+REF] [+DEIC] [+DEF] [+SPEC]

Dd/r LocP t1

Loc Loc [+REF] [+DEIC] [+DEF] [+SPEC]

Unit/PlP

mokkerii

Yet another welcome result of the proposed analysis is that it lets us account for a minimal pair like (146a, b), which contains an anaphoric ku and KU.

230

5 Capturing the Korean Facts

(146) a. Discourse containing a DP comprised of an anaphoric ku + N: Pihayngki-an-eyse nay yeph-cari-ey [etten namca-ai]i -ka my next.seat-LOC [some male-kid]-NOM Plane-inside-LOC anc-key-toy-ess-ta. sit-CAU-happened.to.be-PST-DECL ‘In the plane, a young boyi happened to be sitting next to me.’ Kurentey, nacwung-ey al-ko po-ni, But, later.time-LOC know-CONN see-COMP, [DP kui namca-ai]-nun Chelswu-uy tongsayng-i-ess-ta. [ that male-kid]-TOP C.-GEN younger.sibling-COP-PST-DECL ‘But later, I learned that thati young boy was Chelswu’s younger brother’ b. Discourse containing a DP comprised of ‘KU + AP > anaphoric ku + N’: Pihayngki-an-eyse nay yeph-cari-ey [[maywu omanpwulsonha-n] Plane-inside-LOC my next.seat-LOC [[very arrogant.rude-ADN] anc-key-toy-ess-ta. etten namca-ai]i -ka some male-kid]-NOM sit-CAU-happened.to.be-PST-DECL ‘In the plane, a very arrogant and rude young boyi happened to be sitting next to me.’ Kurentey, nacwung-ey al-ko po-ni, know-CONN see-COMP, But, later.time-LOC [DP [kui [AP omanpwulsonha-n]], [DP kui namca-ai]]-nun [ arrogant.and.rude-ADN]], [ that male-kid]]-TOP [ [KU paro Chelswu-uy tongsayng-i-ess-ta. younger.sibling-COP-PST-DECL exactly C.-GEN ‘But later, I learned that thati arrogant and rude boy is actually Chelswu’s younger brother.’

Under the analysis put forth here, the topic DPs in the second sentences of (146a, b) are derived as in (147a, b). These two derivations are identical except that in (147b), the DPd/r has two Spec positions and the outer position is occupied by a CogDemP. (147b) is also essentially identical to (141b) except that here, what merges at [Spec, LocP] and then raises to [Spec, DPd/r] is a more “ordinary” DemP, not a CogDemP. (147) a. Derivation of the topic DP in the 2nd sentence of (146a): ‘Anaphoric ku > N’ DPd/r Dd/r Dd/r [+REF] [+DEIC] [+DEF] [+SPEC]

LocP DemP ku

Loc [+REF] [+DEF] [+DEF] [+SPEC]

Loc UnitP/PlP

namcaai

5.4 Why Co-occurring with Ku ‘the/that’ Engenders a Non-canonical ADJ Order

231

b. Derivation of the topic DP in the 2nd sentence of (146b): ‘KU + AP > ku > N’ DPd/r CogDemPS

DPd/r

CogDem S CogDemS AP [+STS DISTANT] KU omanpwulsonhan

Dd/r Dd/r [+REF] [+DEIC] [+DEF] [+SPEC]

LocP DemP ku

Loc [+REF] [+DEIC] [+DEF] [+SPEC]

Loc UnitP/PlP

namcaai

Correlatively, the present analysis offers a principled way to differentiate between an anaphoric ku and what I call KU. To see this, let us reconsider (107a, b), which are repeated below. [AP kappissa-n] (107) a. ku*(,) [RC Mina-ka ei sa-o- ]-n*(,) that [ M.-NOM __ buy-come-PRF]-REL], [ expensive-ADN] mwulken-tuli item-PL ‘those expensive things, which Mina bought and brought’ (DEM > FRC > UN-AP) b. ku [AP kappissa-n]*(,) [RC Mina-ka ei sa-o- ]-n mwulken-tuli Intended: ‘those expensive things that Mina bought and brought’ (DEM > UN-AP > FRC)

As mentioned above, although they look alike, these two examples do not mean the same, which is indicated by their English translations. In particular, unlike (107a), (107b) implicates that S feels strongly about the fact that the items under description or at issue are very expensive,26 and this already gives us the justification to treat the two occurrences of ku in (107) rather differently. On top of that, one occurrence of ku occurs away from an AP but the other occurrence does not.

26

My intuition about (107a) is that it can be interpreted in such a way that ku indicates that the DP it modifies has some salient discourse-old property and the content of that property is spelled out by the N modifier that immediately follows it, i.e., the RC in this case. But further investigation will be necessary to fully articulate the semantic intuition I have about such data. Hence, I defer doing so to another occasion. See Kim 2018 for a tentative account, under which the role of the RC in such contexts is comparable to the role of an actual demonstration that is required to license DEMs in English, as pointed out by Kaplan (1977/1989).

232

5 Capturing the Korean Facts

Given this, and given the fact that both of these examples are uttered in contexts that do not involve pointing or contrastive focus, we can conclude that while (107a) contains ku, (107b) contains KU. In light of these considerations, then, we can derive (107b) as sketched in (141a), which shows that the CogDemP here merges at [Spec, LocP] and then raises to the Spec of the pre-existing DPd/r. As for (107a), it is derived as in (148): here, the DemP headed by ku merges at [Spec, LocP] first and then stops by the pre-existing [Spec, DPd/r] next. This is followed by the DemP scrambling to the [Spec, DPd/r] that is created atop the [Spec, DPd/r] that the NRS-RC adjoins at. Consequently, ‘anaphoric ku > NRS-RC > N’ results as the surface order. Notably, what is given in (148) lets us capture the fact that in (107a), both ku and the RC are followed by obligatory pauses: these pauses are obligatory because, in the derived structure, both the DEM and the RC occupy adjoined positions. On the other hand, (107b) is predicted to be pronounced with no pause after the DEM but with a pause after the AP, and this is because, in (141a), while the CogDemP occupies an adjoined position, the RC does not. (148) Derivation of (107a): ‘Anaphoric ku > NRS-RC > N’ surface order DPd/r DemP1

DPd/r

NRS-RC

DPd/r

ku t1

Dd/r

ei Dd/r [+REF] [+DEIC] [+DEF] [+SPEC]

5.5

LocP t1

Loc Loc PlP+ [+REF] [+DEIC] [+DEF] [+SPEC] kappissan mwulkentuli

Summary

In this chapter, I applied the new decompositional split-DP analysis presented in the previous chapter to Korean and showed that while the proposed derivational system can adequately handle a wide range of relevant data, it falls short of accounting for certain cases, in particular those that exhibit gradient grammaticality. Confronted with such empirical challenges, I proposed a partial set of OT constraints as a way

5.5 Summary

233

to filter out non-optimal candidates, modeling after the alignment constraints Grimshaw (2001) postulates to capture constituent order variation at the clausal level. After demonstrating that adding such an output filter to our analytical toolkit enables us to account for some of the problematic data pertaining to AOR in Korean, I turned to capturing the relative ordering between FRCs, and by positing alignment constraints that are sensitive to RCs’ contents as well as the discourse status of their hosting DPs, I demonstrated that an OT-based system has wider applicability in dealing with constituent order variations than has been acknowledged in the extant generative linguistics literature. This discussion was followed by taking up a new AOR problem that had been identified in Chap. 3, namely, the fact that when occurring adjacent to the DEM ku, a relatively light AP may occur preceding an FRC, constituting an apparent exception to weight-based typological generalizations reported in the literature. In resolving this puzzle, I first pointed out that the phenomenon at hand actually stems from the pragmatic function of ku that is encoded in grammar. More specifically, I argued that in contexts like (105b)–(108b), the DEM functions as a cognitive indexical marker which is employed when the speaker wishes to comment on some specific/definite entity that is absent in the discourse context but has some noteworthy property. In order to derive such properties of KU within a formal syntactic framework, I have proposed that DEMs may be base-generated in the head position of what I call a CogDemP with a property-denoting AP occurring as their complement. Additionally, I have suggested that a CogDemP may enter a DP structure either by adjoining to a DPd/r or by merging at [Spec, LocP]. The proposed analysis of a ‘KU + AP’ string has been shown to make correct predictions as to where in the DP structure a CogDem may occur relative to non-restrictive and supplementary RCs. It also provides a principled way to differentiate between a more “ordinary” DEM and what I call a CogDem. Moreover, it has led us to see that Korean DEMs may take on affective meanings and carrying such a pragmatic meaning may impact the surface order of a DEM relative to other N modifiers. Recent linguistic research has shown that natural language encodes pragmatic meanings in various ways. For example, the binding literature makes frequent references to pragmatic notions like SELF, consciousness, and point of view (e.g., Sells 1987; Iida and Sells 1988; Zribi-Hertz 1989; Minkoff 1994, 2004; Stirling 2005). And due largely to the advances made in such works, it is now commonplace to assume that point of view is syntactically encoded (e.g., Speas and Tenny 2003; Tenny 2006; Sato and Kishida 2009; Chou 2012; Kim 2013). Independently of this, the cartographic syntactic literature suggests that the left-periphery of sentential structure is comprised of projections that are discourse-sensitive, namely, FPs like Evidential Phrase, Topic Phrase, and Focus Phrase (e.g., Rizzi 1997; Cinque 1999; Belleti 2004; Cinque and Rizzi 2008). In addition, in the formal pragmatics literature too, it is now firmly established that intonational breaks and contours encode meaning and they should be given formal syntactic/semantic treatments (e.g., Gunlogson 2001; Potts 2003/2005).

234

5 Capturing the Korean Facts

Given this, the findings we have made in this chapter will have implications that go beyond the syntax and semantics of Korean N modifiers and therefore they need to be discussed in a broader light. In the next chapter, I say more about the implications and the outcome of the present analysis by making extensions to other languages while taking a closer look at some of the phenomena we have introduced in this chapter as well as in the preceding chapters.

References Acton, Eric K., and Christopher Potts. 2014. That straight talk: Sarah Palin and the sociolinguistics of demonstratives. Journal of Sociolinguistics 18 (1): 3–31. An, Duk-Ho. 2014. Genitive case in Korean and its implications for noun phrase structure. Journal of East Asian Linguistics 23 (4): 361–392. Belletti, Andriana. 2004. Aspects of the low IP. In The structure of CP and IP: The cartography of syntactic structures, vol. 2, ed. Luigi Rizzi, 16–51. New York: Oxford University Press. Bolinger, Dwight. 1967. Adjectives in English: Attribution and predication. Lingua 18: 1–34. Bouchard, Denis. 2002. Adjectives, number and interfaces: Why languages vary? Amsterdam: North-Holland. Choi, Hye-Won. 1996. Optimizing structure in context: Scrambling and information structure. Stanford, CA: Stanford University dissertation. Chou, Chao-Ting Tim. 2012. Syntax-pragmatics interface: Mandarin Chinese why-the-hell and point-of-view operator. Syntax 15 (1): 1–24. Cinque, Guglielmo. 1999. Adverbs and functional heads: A cross-linguistic perspective. Oxford, UK: Oxford University Press. Cinque, Guglielmo. 2005. Deriving Greenberg’s Universal 20 and its exceptions. Linguistic Inquiry 36: 315–332. Cinque, Guglielmo. 2010. The syntax of adjectives. A comparative study. Cambridge, MA: MIT Press. Cinque, Guglielmo, and Luigi Rizzi. 2008. The cartography of syntactic structures. In CISCL working papers on language and cognition, vol. 2, ed. Vincenzo Moscati, 43–59. University of Siena. DeLancey, Scott. 1997. The mirative and evidentiality. Journal of Pragmatics 3: 371–384. Grimshaw, Jane. 2001. Economy of structure in OT. Rutgers Optimality Archive 434. Gunlogson, Christine. 2001. True to form: Rising and falling declaratives as questions in English. Santa Cruz, CA: University of California-Santa Cruz dissertation. Huddleston, Rodney, and Geoffrey Pullum. 2005. A student’s introduction to English grammar. Cambridge, UK: Cambridge University Press. Iida, Masayo, and Peter Sells. 1988. Discourse factors in the binding of zibun. In Papers from the second international workshop on Japanese syntax, ed. William J. Poser, 23–46. Stanford, CA: CSLI. Kang, Arum. 2018. Unexpected effect: The emphatic determiner with gradable NPs in Korean. The Journal of Studies in Language 33 (4): 595–615. Kang, Soon Haeng. 2005. On the adjectives in Korean. In University of Venice working papers in linguistics, vol. 15, 153–169. Kang, Soon Haeng. 2006. The two forms of the adjective in Korean. In University of Venice working papers in linguistics, vol. 16, 137–163.

References

235

Kaplan, David. 1977. Demonstratives: An essay on the semantics, logic, metaphysics, and epistemology of demonstratives and other indexicals. Ms., published in Themes from Kaplan. 1989. ed. Almog, Joseph, John Perry, and Howard Wettstein, 481–564. New York: Oxford University Press. Keller, Frank. 2000. Gradience in grammar: Experimental and computational aspects of degrees of grammaticality. Edinburgh, UK: University of Edinburgh dissertation. Kim, Jaieun. 2013. Subject and point-of-view in Korean: The syntax-discourse interface. Seoul, Korea: Sogang University master’s thesis. Kim, Min-Joo. 2016a. The ‘imperfective’ in attributive clauses in Korean as a window into the evidential past and the metaphysical future. Studies in Language 40 (2): 340–379. Kim, Min-Joo. 2016b. Cognitive indexical usage of demonstrative ku in Korean and a split DP analysis. In Proceedings of the 18th Seoul international conference on generative grammar, ed. Kim, Tae Sik, and Seungwan Ha, 175–194. Seoul: The Korean Generative Grammar Circle. Kim, Min-Joo. 2018. What does a demonstrative do when it co-occurs with a relative clause and an adjective phrase in Korean? Paper presented at Semantics 2018: Looking Ahead. University of Massachusetts-Amherst, March 10, 2018. Larson, Richard K. 1998. Events and modification in nominals. In Proceedings from semantics and linguistic theory (SALT) VIII, 145–168. Cornell University Press. Larson, Richard K. 2000. Temporal modification in nominals. Paper presented at the International Roundtable on the Syntax of Tense. University of Paris VII, France. Larson, Richard K., and Naoko Takahashi. 2007. Order and interpretation in prenominal relative Clauses. In Proceedings of the workshop on altaic formal linguistics II. MIT working papers in linguistics, vol. 54, 101–120. Cambridge, MA: MITWPL. Liberman, Mark. 2008. Affective demonstratives. Language Log, http://languagelog.ldc.upenn. edu/nll/?p=674. Accessed 28 Mar 2018. Minkoff, Seth. 1994. How some so-called “thematic roles’’ that select animate arguments are generated, and how these roles inform binding and control. Cambridge, MA: MIT dissertation. Minkoff, Seth. 2004. Consciousness, backward coreference, logophoricity. Linguistic Inquiry 34 (3): 485–494. Potts, Christopher. 2005. The logic of conventional implicatures. New York: Oxford University Press. (2003 University of California-Santa Cruz dissertation.) Potts, Christopher, and Florian Schwarz. 2010. Affective ‘this’. Linguistic Issues in Language Technology 5: 1–29. Prince, Alan, and Paul Smolensky. 1993. Optimality theory: Constraint interaction in generative grammar. Rutgers University Center for Cognitive Science Technical Report 2. Rizzi, Luigi. 1997. The fine structure of the left periphery. In Elements of grammar, ed. Liliane Haegeman, 281–337. Dordrecht: Kluwer Academic Publishers. Sato, Yosuke, and Maki Kishida. 2009. Psychological predicates and the point-of-view hyperprojection. Gengo Kenkyu [Language Research] 135: 123–150. Sells, Peter. 1987. Aspects of logophoricity. Linguistic Inquiry 18 (3): 445–479. Sorace, Antonella, and Frank Keller. 2005. Gradience in linguistic data. Lingua 115 (11): 1497–1524. Speas, Peggy, and Carol Tenny. 2003. Configurational properties of point of view roles. In Asymmetry in grammar, ed. Anna Maria Di Sciullo, 315–344. Amsterdam: John Benjamins. Sproat, Richard, and Chinlin Shih. 1988. Prenominal adjectival ordering in English and Mandarin. In Proceedings of NELS, vol. 18, 465–489. Amherst, MA: GLSA. Sproat, Richard, and Chinlin Shih. 1990. The cross-linguistics distribution of adjectival ordering restrictions. In Interdisciplinary approaches to language: Essays in honor of S-Y. Kuroda, ed. Georgopoulos, Carol, and Roberta Ishihara, 565–593. Dordrecht: Kluwer. Stirling, Lesley. 2005. Switch-reference and discourse representation. Cambridge, UK: Cambridge University Press. Tenny, Carol L. 2006. Evidentiality, experiencers, and the syntax of sentience in Japanese. Journal of East Asian Linguistics 15: 245–288.

236

5 Capturing the Korean Facts

Traugott, Elizabeth Closs. 2006. The semantic development of scalar focus modifiers. In The handbook of the history of English, ed. van Kemenade, Ans, and Bettelou Los, 335–359. Malden, MA: Basil Blackwell. Zribi-Hertz, Anne. 1989. Anaphor binding and narrative point of view: English reflexive pronouns in sentence and discourse. Language 65 (4): 695–727.

Chapter 6

Extensions, Refinements, and Prospects

This chapter discusses consequences of the analysis presented in the previous two chapters by extending it to several other languages or linguistic phenomena than what we have already looked at. Our discussions will show how the ordering restrictions on Mandarin RCs and their other seemingly “peculiar” behaviors may be explained under the proposed analysis; how Korean and English DEMs differ from each other in their semantics and pragmatics and how relevant cross-linguistic variation on DEM meanings may be captured within a binary feature-based framework; whether all languages have supplementary RCs or not and if not, what type of languages tend to have them; what might be the internal syntax of DemPs in all human languages and whether a uniform analysis may be possible for them; how Korean CogDem KU and a determiner-like lexeme ta in Slovenian compare with each other, and whether the cognitive indexical analysis of KU presented in Chap. 5 can be carried over to ta. On the way to discussing these topics, we compare the present analysis with Lin and Tsai’s (2015) analysis of Mandarin RCs as well as Leu’s (2008, 2015) analysis of the syntax of DemPs and Marušič and Žaucer’s (2007, 2013, 2014) analyses of ta in Slovenian. After discussing such topics in the light of the proposed analysis, the chapter ends by taking up two additional issues relevant for our purposes, namely, (i) factors determining the interpretative possibilities of adnominal APs and (ii) the number of focus positions inside a DP. In these contexts, we also discuss some Romanian, Korean, and English data that will bear on these issues.

6.1

Ordering Restrictions on Mandarin FRCs

In Chap. 3, we observed that Korean FRCs occur only in certain orders, and in Chap. 5, I proposed a way to account for their surface orders by restoring to an OT-based analysis. In this section, using Mandarin data, I aim to show that similar © Springer Nature Switzerland AG 2019 M.-J. Kim, The Syntax and Semantics of Noun Modifiers and the Theory of Universal Grammar, Studies in Natural Language and Linguistic Theory 96, https://doi.org/10.1007/978-3-030-05886-9_6

237

238

6 Extensions, Refinements, and Prospects

phenomena are found in other languages as well, and they can be captured by applying essentially the same mechanism as what did to Korean in Chap. 5. Jhang (2012) observes that Mandarin FRCs may occur only in certain orders and their ordering restrictions have little to do with whether the RCs at hand contain an I-level predicate or not, unlike what authors like Del Gobbo (2005) and Larson and Takahashi (2007) have claimed. To see this, consider (1)–(4), where RCI stands for an RC with an I-level predicate and RCS for an RC with an S-level predicate. Except for the sentences in (1), all the examples in (2)–(4) are judged acceptable or unacceptable, irrespective of whether they instantiate ‘RCS > RCI’ order or not.1 (Here and below, following standard practice in Chinese linguistics, I do not gloss the RC marker de. In addition, for simplicity, I do not indicate the tonal properties of the Mandarin words presented here.) (1) a.

[ei bei shiren yiwang de] [ei cengjing weida de] PASS world.people forget DE] [__ once great DE] [__ zuopini art.work ‘The art piece that has been forgotten by the world which was once great’ (RCS > RCS) b. *[ei cengjing weida de] [ei bei shiren yiwang de] Intended: ‘The art piece that has been forgotten by the world which was once great’ (RCS > RCS)

(2) a. [Ta xihuan ei de] [Jay xie ei de] [She like __ DE] [J. write __ DE] ‘The song that she likes which Jay wrote’ b. *[Jay xie ei de] [ta xihuan ei de] Intended: ‘The song that she likes which Jay wrote’

na

shou

DEM

CL

na

gei song (RCI > RCS) shou gei (RCS > RCI)

(3) a. [Wo zuotian yudao ei de] [ni shang-ci shuoguo ei de] reni [I yesterday meet __ DE] [you last-time mention __ DE] person ‘The person that I met yesterday who you mentioned the other day’ (RCS > RCS) b. *[Ni shang-ci shuoguo ei de] [wo zuotian yudao ei de] Intended: ‘The person that I met yesterday who you mentioned the other day’ (RCS > RCS) (4) a. [Ta sheji ei de] [ei de shoujiang de] [He design __ DE] [__ win first.prize DE] ‘The dress that he designed which won the first prize’ b. *[ei de shoujiang de] [ta sheji ei de] Intended: ‘The dress that he designed which won the first prize’

1

yangzhuangi dress (RCS > RCS) yangzhuangi (RCS > RCS)

I took this data set from Jhang 2012 but am presenting it with slightly modified glosses to make things more consistent with the way Mandarin data are presented in the literature. Note also that the grammaticality judgments on these Mandarin data are Jhang’s but she had them verified by 12 Mandarin speakers, and I also got them confirmed by two additional Mandarin speakers, namely, C.-T. James Huang and Feng-hsi Liu, whose help I gratefully acknowledge here.

6.1 Ordering Restrictions on Mandarin FRCs

239

The data set given above is admittedly too small to draw any decisive conclusions from. But all the examples here involve non-restrictive RCs, as was the case with the Korean examples we looked at in Chap. 5, and their contrasts can be explained by adopting the same line of OT-based analysis we employed for the Korean cases. More concretely, if we apply the constraints given in (5) and their relative rankings given in (6) and also assume that the RCs in (1)–(4) are embedded inside referential DPs and therefore they occupy what we call [Spec, DPd/r] positions, then all the good examples turn out to comply with FMLRCLFT, i.e., the highest-ranked constraint, but all the bad ones violate it. (5) Constraints on RCs which modify the same discourse-familiar sentential topic: a. EPCRCLFT: An RC derived from an episodic event description occurs at the left edge of a topic DPd/r. b. PRERCLFT: When two RCs modify the same topic DPd/r, the RC whose event time is earlier occurs at the left edge of the DPd/r. c. FMLRCLFT: When two RCs modify the same topic DPd/r, the RC whose content is more familiar to the relevant discourse participant (i.e., S or H) occurs at the left edge of the DPd/r. (6) Ranking between the constraints in (5): FMLLFT >> {EPCLFT, PRELFT}

To take a look at each case one by one, tableau (7) shows that (1a) is the more harmonic candidate than (1b) because the relevant discourse participant (i.e., S) does not remember this great work of art that is being talked about, just like the rest of the world, and therefore at the time of discourse, RC1’s content is deemed more familiar to her than RC2’s content is. (7)

Reason for the surface order of the two RCs in (1): input: [DPd/r [RC1 bei shiren yiwang de] [RC2 cengjing weida de] [DPp [ P zuopin]]] a.

[DPd/r [RC1 ‘has been forgotten by the world’] [RC2 ‘was once great’] [DPp [ P ‘art work’]]]

b.

[DPd/r [RC2 ‘was once great’] [RC1 ‘has been forgotten by the world’] [DPp [ P ‘art work’]]]

FMLRC LFT

EPCRC LFT

PRERC LFT *

*!

In (2) also, the (a) example surfaces as the optimal candidate, as shown in tableau (8), although it violates EPCRCLFT, rather than PRERCLFT, unlike the case with (1a): Here, RC1 contains the pronoun ta ‘she’, and given the felicity conditions on pronoun usage, it must have been the case that the referent of ta was already introduced to the discourse before (2) was uttered. This makes RC1’s content more familiar to the relevant discourse participant than RC2’s content is at the time of discourse, and this in turn makes the datum instantiating the ‘RC1 > RC2’ order more desirable than the one that instantiates the ‘RC2 > RC1’ order.

240

(8)

6 Extensions, Refinements, and Prospects

Reason for the surface order of the two RCs in (2): input: [DPd/r [RC1 ta xihuan de] [RC2 Jay xie de] [DPd/r [ P na shou ge]]] a.

[DPd/r [RC1 ‘she likes’] [RC2 ‘Jay wrote’] [DPd/r [ P ‘the song]]]

b.

[DPd/r [RC2 ‘Jay wrote’] [RC1 ‘she likes’] [DPd/r [ P ‘the song’]]]

FMLRC LFT

EPCRC LFT

PRERC LFT

* *!

In (3) too, (a) surfaces as the more harmonic candidate which better satisfies FMLLFT because RC1 describes an event that the speaker directly experienced, so at the time of discourse, its content is deemed more familiar to the relevant discourse participant (namely, S) than the content of RC2 is. (9)

Reason for the surface order of the two RCs in (3): input: [DPd/r [RC1 wo zuotian yudao de] [RC2 ni shangci shuoguo de] [DPp [ P ren]]]] a.

[DPd/r [RC1 ‘I met yesterday’] [RC2 ‘you mentioned last time’] [DPp [ P ‘person’]]]

b.

[DPd/r [RC2 ‘you mentioned last time’] [RC1 ‘I met yesterday’] [DPp [ P ‘person’]]]

FMLRC LFT

EPCRC LFT

PRERC LFT *

*!

Finally, in the case of (4), RC1 contains a pronoun whereas RC2 does not, so for the same reason as why (2a) is the more harmonic candidate than (2b) is, here too, (a) surfaces as the winner. Besides, here, the (b) example violates two constraints whereas the (a) example violates none, as shown in (10). (10)

Reason for the surface order of the two RCs in (4): input: [DPd/r [RC1 ta sheji de] [RC2 de shoujian de] [DPp [ P yangzhuang]]] a. b.

[DPd/r [RC1 ‘he designed’ [RC2 ‘won first prize’ [DPp [ P ‘dress’]]] [DPd/r [RC2 ‘won first prize’ [RC1 ‘he designed’ [DPp [ P ‘dress’]]]

FMLRC LFT

*!

EPCRC LFT

PRERC LFT

*

In sum, then, even though further research will be necessary to draw more definitive conclusions about what exactly governs the ordering of FRCs in Mandarin as well as in other languages, the foregoing shows that marrying an OT analysis with a derivational one has a great potential to yield positive results beyond Korean.

6.2 Integrated Non-restrictive RCs in Mandarin and Korean

6.2

241

Integrated Non-restrictive RCs in Mandarin and Korean

Under the present analysis, integrated RCs have three different merge sites, namely, [Spec, DPp], [Spec, DPq], and [Spec, DPd/r], and while those that merge at [Spec, DPp] and [Spec, DPq] receive restrictive modificational interpretations, those that merge at [Spec, DPd/r] receive integrated non-restrictive modificational interpretations, as given in (11), repeated from Chap. 4. (11) Position of RCs inside the proposed split-DP structure: [DPq RC [… [SortP [ DPd/r RC [DPp RC High Field Middle Field Low Field NRS Mod RS Mod

[nP [ P]]]]]]]

Furthermore, in the present framework, all languages are subject to the principle we refer to as PELI (Principle of Economy of Lexical Insertion) and in article-less languages, this principle dictates that the Spec of a DPd/r be filled by an N modifier which receives a non-restrictive interpretation. The version of PELI we have adopted is given in (12), repeated from Chap. 4. (12) Principle of Economy of Lexical Insertion (PELI): A functional projection (FP) that has any syntactic features must be licensed at all levels of representation by (a) making the specifier visible (i.e., pronounced) or (b) making the head visible. But an FP whose head hosts no syntactic features need not be visible at any level of representation. (Revised from Dimitrova-Vulchanova and Giusti 1998)

Positing such syntax for RCs and other types of N modifiers has non-trivial implications for the status of RCs which occur modifying proper names and pronouns in languages like Mandarin and Korean. The reason is that the ability to modify a proper name or a pronoun is considered a hallmark of non-restrictive or supplementary modifiers in the literature (e.g., Potts 2003/2005; Cinque 2006, 2008, 2010), yet upon closer inspection, RCs in Mandarin and Korean turn out to behave differently than the more familiar English-type non-restrictive or supplementary N modifiers, as Lin and Tsai (L&T) (2015) have shown, and this raises the question of what counts as a non-restrictive or a supplementary RC. In addition, even though the fact that Mandarin RCs can occur modifying proper names and

242

6 Extensions, Refinements, and Prospects

pronouns is well-documented in the Chinese linguistics literature, there has been a lack of consensus among authors as to whether such RCs are restrictive or non-restrictive (e.g., Chao 1968; Huang 1982; Tsai 1994; Zhang 2001; Lin 2003; Del Gobbo 2003, 2004, 2005, 2010; Constant 2011) (see Lin and Tsai 2015 for an overview and discussion). Therefore, taking a closer look at the relevant data and identifying the source of such “mysterious” properties of certain RCs in Mandarin is considered not only important but also necessary if one cares about such issues. In the pages that follow, I will treat Mandarin RCs which occur modifying proper names and/or pronouns as integrated RCs, following Del Gobbo (2010) and Lin and Tsai (2015). I follow such authors, as they offer convincing arguments to treat them as integrated RCs rather than non-integrated ones. But another more important reason is that adopting their line of analysis as a point of departure will turn out to eliminate the need to decide whether such RCs are restrictive RCs or non-restrictive ones in the end, as we will see shortly. To get us started on this discussion, let us first consider (13)–(16).2 These data show that Mandarin RCs can indeed occur modifying proper names or pronouns and when they do, they are not accompanied by pauses, unlike the case with supplementary RCs in English.3 (13) [Zheng zai jingzuo yundong de] Zhou Botong PROG sit-in stage.a.demonstration DE] Zhou Botong [right turan da jiao yi sheng, tiao-le qilai. suddenly big scream one voice jump-ASP up.come ‘Zhou Botung, who was sitting in to stage a demonstration, suddenly screamed loudly and jumped up.’ (Shi 2010: 327; cited as (27) in L&T 2015) (14) Nimen hui kandao [pao de you will see [run DE ‘You will see a Liuxiang who runs faster.’

geng more

kuai fast

de] DE]

Liuxiang. Liuxiang [L&T 2015: (28)]

2

For the Mandarin data presented here which are taken from the literature, I have retained the original morpheme-by-morpheme glosses and also have not added the gaps inside the RCs that would be co-indexed with the head Ns, since such gaps were not found in the original glosses. But for expository convenience, I have added square brackets to indicate the RC boundaries and also highlighted the RCs and their head Ns. 3 In fact, in cases like (13)–(16), there cannot be a pause between the RCs and their head Ns because that would yield ungrammaticality, as I have verified with Feng-hsi Liu via personal communication.

6.2 Integrated Non-restrictive RCs in Mandarin and Korean

243

(15) [Juyou hei ren xietong de] Obama dangxuan meiguo [possess black people blood DE] Obama elected America di sishisiren zongtong. the forty.fourth president ‘Obama, who has black blood, was elected the 44th president.’ [L&T 2015: (32)] (16) [Shii nüren de] women juede [shi nanren de] nimen DE] we think [be man DE] you [be women dei zou. should leave ‘We women think that you men should leave.’ (Hu 2008: 50; cited as (33) in L&T 2015)

Consider now (17)–(18). These data show that Korean RCs pattern with Mandarin RCs (modulo the different word orders at the clausal level and the case-marking that appears on argument DPs in Korean). What is particularly noteworthy is that just like the Mandarin RCs in (13)–(16) (see Footnote 3), the Korean RCs here cannot be followed by a pause, which clearly indicates that they are integrated rather than supplementary N modifiers, although the English translation for (17) may inadvertently suggest that the RC in it is a supplementary N modifier. (Note that to make the Korean data form minimal papers with the corresponding Mandarin data, I have not indicated gaps inside the RCs which would be co-indexed with their head Ns.) (17) [huk-in-i- -n](*,) Obama-ka mikwuk-uy [black-person-N.PST-COP-REL] Obama-NOM United.States-GEN taythongryeng-uro tangsen-toy-ess-ta. president-as elect-become-PST-DECL ‘Obama, who is a black person, got elected as United States’ President.’ wuri-tul-un [namca-i- -n](*,) nehuy-tul-i (18) [yeca-i- -n](*,) [woman-COP-N.PST-REL] we-PL-TOP [man-COP-N.PST-REL] you.all-PL-NOM ttena-ya.han.ta-ko sayngkakha-n-ta. leave-should-COMP think-N.PST-DECL ‘We women think that you men should leave.’

About Mandarin RCs like those in (13)–(16), Lin and Tsai (2015) claim that they have the same syntax as run-of-the-mill restrictive RCs but they can modify proper

244

6 Extensions, Refinements, and Prospects

names or pronouns because, in article-less languages, proper names and pronouns are predicate-denoting rather than referring (cf. Matushansky 20064). More concretely, the idea they promote is that, in Mandarin, proper names denote singleton sets (i.e., something of type but with just one member in the set) and because RCs also denote predicates of type , they can combine with each other. Despite their predicative semantics, however, a string comprised of an RC plus a proper name ends up denoting an individual (i.e., something of type ) because a null D head in Mandarin denotes the iota-operator and this operator closes set-denoting expressions, as suggested by Matushansky (2006). (19) a. Lin and Tsai’s treatment of proper-names in Mandarin: DP

D

b. c. d.

NP

Zhangsan Zhangsan = x.x = Zhangsan = P. the x such that P(x) D ] [DP] = [ P.the x such that P(x)]( x.x = Zhangsan) = the x such that x = Zhangsan [L&T 2015: (65)]

If proper names in Mandarin indeed denote predicates, as Lin and Tsai suggest, then they can readily combine with RCs since the latter are also predicate-denoting. Hence, in computing the meaning of a DP which has an internal structure where an RC modifies a proper name, the semantic operation Predicate Modification in the sense of Heim and Kratzer (1998) applies, yielding another predicate-denoting element, and to this, the null iota operator applies, giving rise to an individual-denoting expression, as exemplified in (21).

4

For the treatment of proper names, Lin and Tsai (2015) build on Matushansky’s (2006) analysis but while she claims that all proper names in all languages are inherently of type , they take a more eclectic position and thereby hypothesize that in some languages, they may be of type , as Longobardi (1994, 1999) suggests for English, but in other languages, they may be of type (e.g., Mandarin), and in still other languages, they may be ambiguous between of type and of type (e.g., Amharic).

6.2 Integrated Non-restrictive RCs in Mandarin and Korean

245

(20) Predicate Modification (PM) and are both If is a branching node, { , } is the set of ’s daughters, and in D, then = x De • (x) = (x) = 1. (Heim and Kratzer 1998: 65) (21) Lin and Tsai’s treatment of Mandarin datum (15): a. DP

D

NP2

CP

b. c. d. e.

NP1

juyou heren xietong de Obama Obama = x.x = Obama [CP juyou heren xietong de] = x.x has the blood of black people [NP2] = x.[ [CP] (x) = 1 and [NP1] (x) = 1] (via PM) = x.x has the blood of black people and x = Obama [DP] = the x such that x has the blood of black people and x = Obama [L&T 2015: (67)]

Under the split-DP analysis we proposed in Chap. 4, the internal syntactic structure of the DPs like those in (13)–(18) will be a bit more complex than what is depicted in (21), but we can achieve the same results and here is how: if we assume with Matushansky (2006) and Lin and Tsai (2015) that proper names and pronouns in Mandarin/Korean-type languages are predicate-denoting, then in our framework, the head N in examples like (15) will be located under DPp and, being a restrictive N modifier, the RC will merge at [Spec, DPp]. Consequently, we obtain a configuration in which both the head N and the RC are c-commanded by the null iota-operator which is sitting under Dd/r, as depicted in (22). Therefore, apart from the different node labels and more layered DP-internal structure that (22) has compared to (21a), the meaning of the DP is computed in essentially the same manner as in Lin and Tsai’s analysis.

246

6 Extensions, Refinements, and Prospects

(22) Underlying structure of Mandarin datum (15) under the present analysis: DPd/r, e Dd/r , e Dd/r,

DPp2,

RC,

DPp1,

juyou heren xietong de Obama (23) Computation of the meaning of (15) based on (22): a. Obama = x.x = Obama b. [RC juyou heren xietong de] = x.x has the blood of black people c. [DPp2] = x.[ [RC] (x) = 1 and Obama (x) = 1] (via PM) = x.x has the blood of black people and x = Obama = P. the x such that P(x) d. [Dd/r] = e. [DPd/r] = the x such that x has the blood of black people and x = Obama

One thing I need to point out before proceeding further is that, in view of the presumably universal constraint PELI given in (12), in an article-less language like Mandarin or Korean, the Spec of DPd/r in (22) has to be filled by an RC which acts like a non-restrictive modifier, so under the present analysis, the surface structure of (15) will actually look like what is given in (24). That is, the RC has to raise to the Spec of DPd/r, thereby pronouncing the DPd/r in compliance with PELI; this FP has to be pronounced because its head hosts uninterpretable features, namely, [+REFERENTIAL; +/_DEICTIC; +/_DEFINITE; +/_SPECIFIC], and without having a phonologically overt D in the language, something else has to pronounce it by moving to its Spec position (see Chap. 4, Sect. 4.3.3). (24) Surface structure of (15) under the present analysis: DPd/r, e Dd/r , e Dd/r, RC,

DPp, DPp,

juyou heren xietong de Obama

However, the raising of the RC to [Spec, DPd/r] in cases like (15) does not create a problem for us because, according to Lin and Tsai (2015), what we call restrictive modifiers and what we call non-restrictive ones have a nearly identical syntax and semantics; both types of RCs combine with set-denoting expressions but while

6.2 Integrated Non-restrictive RCs in Mandarin and Korean

247

restrictive RCs combine with non-singleton set denoting expressions, non-restrictive RCs combine with singleton set denoting ones (L&T 2015: 122). Furthermore, deriving the meaning of (15) based on (24) actually yields the same results as deriving it based on (22) does, and here is why: in standard formal semantics, movement creates an index right below the moved item, and the trace of the moved constituent is interpreted as denoting a free variable of the relevant kind, dictated by type-driven semantics (Klein and Sag 1985). Additionally, this free variable gets bound by a lambda operator contributed by the Predicate Abstraction (PA) rule, which applies when computing the syntactic node that immediately c-commands the index node (Heim and Kratzer 1998). Consequently, the sister node of the raised RC in (24) denotes a functor which takes the RC’s denotation as its argument. That is, the meanings of the raised RC and its sister node get computed via Functional Application (FA) rather than via Predicate Modification, as shown in (26f). But notably, the last line of (26f) is identical to (23e) and this shows that the raising of the RC does not impact the overall semantics of the DP in which it occurs. (25) Predicate Abstraction Rule (PA) Let be a branching node with daughter and , where dominates only a a ax/i numerical index i. Then, for any variable assigner a, = x De • . (Heim and Kratzer 1998: 186) (26) Computation of the surface structure of (15) under the present analysis: a. DPd/r, e RC,

?, 1

juyou heren xietong de

Predicate Abstraction over [t1]

Dd/r , e Dd/r, t1,

b. c. d. e. f.

DPp2, DPp1,

Obama [DPp2] = via PM = x.[ [t1] (x) = 1 and Obama (x) = 1] = x.Q1(x) and x = Obama [Dd/r] = = P. the x such that P(x) [Dd/r ] = via FA = the x such that Q1(x) and x = Obama [?] = via PA = Q1. the x such that Q1(x) and x = Obama [DPd/r] = via FA = [ Q1. the x such that Q1(x) and x = Obama]( [RC] ) = via lexical entry of RC and via -reduction = the x such that x has the blood of black people and x = Obama

Before closing this section, I should note that the present analysis can also readily deal with data like (16), (18), (27), and (28), which contain pronouns or definite descriptions marked by a DEM, and this can be a potential improvement on Lin and Tsai’s (2015) analysis since while they mention such Mandarin data, they

248

6 Extensions, Refinements, and Prospects

do not discuss their syntax and semantics in (any) detail, and it is not clear how their analysis may handle such cases. (27) Mandarin RC modifying a definite description containing a DEM: [dai yanjing de] nei-wei xiansheng [wear glasses DE] that-CL mister ‘the/that gentleman who is wearing glasses’ [adapted from L&T 2015: (2a)] (28) Korean RC modifying a definite description containing a DEM: [han-ttay Mina-rang sakiwi-ess-te-n] ce namca-pwun that male-HON [one-period M.-with go.out-PST-RTRO-REL] ‘that gentleman (over there) who used to go out with Mina’

To begin with (16) and (18), which have a pronoun modified by an integrated RC, under the present analysis, such DPs have the same syntax and semantics as cases like (15) and (17) except that the pronouns in them have a more deictic interpretation than proper names like Zhangsan or Obama do, as shown in (29) and (30). Note that for simplicity, I compute the RCs in their base-positions here because raising them to [Spec, DPd/r] will not engender any semantic difference, as we just observed by way of (23) and (26). (29) Structure of the first DP in Mandarin datum (16) under the present analysis: DPd/r, e Dd/r,

DPp2,

RC, shii nüren de ‘be women DE’

DPp1, women ‘we’

(30) Computation of the meaning of (16) based on (29): a. [DPp1] = women ‘we’ = x.x = Speaker and x  15 b. [RC shii nüren de] = x.x is a woman c. [DPp2] = via PM = x.[ [shii nüren de] (x) = 1 and women (x) = 1] = x.x is a woman and x = Speaker and x  1 d. [Dd/r] = = P. the x such that P(x) e. [DPd/r] = [Dd/r ( [DPp] ) = via FA and via lexical entries = the x such that x is a woman and x = Speaker and x  1 5

This is a rather cursory analysis of the pronoun women meaning ‘we’ but for our purposes, assuming this lexical entry for it will still let us make the case that the present analysis provides us with a way to handle data like (16) and (18).

5

6.2 Integrated Non-restrictive RCs in Mandarin and Korean

249

As for cases like (27) and (28), we need to posit a slightly more elaborate syntax and semantics for the head Ns because they contain a DEM (and a classifier). But if we assume that DEMs originate from [Spec, LocP] and further assume that given the way DEMs are treated in recent formal semantics literature (e.g., Patel-Grosz and Grosz 2017 and the references there), a distal DEM like ce in Korean hosts an index and it picks out a contextually salient set of individuals that are far away from both S and H in accordance with Korean speakers’ intuitions about the lexeme, then, we obtain something like (32) as the meaning of the DP in (28) based on the syntactic structure given in (31). Note that in (32), (superscripted) g indicates the assignment function. (31) Structure of Korean datum (28) under the present analysis: DPd/r, e Dd/r,

DPp2,

RC,

DPp1, LocP,

PlP,

han-ttay Mina-rang sakiwi-ess-te-n ce 1

namca

(32) Computation of the meaning of (28) based on (31): a. [PlP] = namca = x.x is a man b. [LocP] = ce g = x.x = g(1) and x is far away from both S and H c. [DPp1] g = ce namca = via PM = x.[ ce g(x) = 1 and namca (x) = 1] = via lexical entries = x.x = g(1) and x is far away from both S and H and x is a man d. [RC han-ttay Mina-rang sakiwi-ess-te-n = x.x used to go out with Mina e. [DPp2] g = via PM = x.[ [RC] (x) = 1 and ce namca (x) = 1] = x.x used to go out with Mina and x = g(1) & x is far away from both S and H and x is a man f. [Dd/r] = = P. the x such that P(x) g. [DPd/r]]g = [Dd/r ( [DPp] ) = via FA and via lexical entries = the x such that x used to go out with Mina and x = g(1) & x is far away from both S and H and x is a man

250

6 Extensions, Refinements, and Prospects

In sum, then, the present analysis can capture the syntax and semantics of integrated RCs modifying proper names or definite descriptions in a uniform way, incorporating the insights of Lin and Tsai (2015) while also complying with independently motivated universal constraints like PELI. In so doing, it also provides a way to explain why there is no formal difference between restrictive and non-restrictive RCs in languages like Mandarin and Korean and furthermore, why RCs like those in (13)–(16) and (27) may have been treated as restrictive modifiers by some authors (e.g., Lin 2003; Del Gobbo 2003; Shi 2010) but as non-restrictive modifiers by others (e.g., Tsai 1994; Lin 2003; Del Gobbo 2010) (see Lin and Tsai 2015 for discussion and additional references). Under our analysis, such RCs are actually both restrictive and non-restrictive at the same time because they start out in a position that is designated for restrictive modifiers, i.e., at [Spec, DPp], but they become pronounced in a position that is designated for non-restrictive modifiers, i.e., at [Spec, DPd/r], which is distinctively different from the way truly supplementary N modifiers behave (more on this in Sect. 6.5). To my knowledge, clarifying the status of Mandarin and Korean RCs in this way is something novel and it can therefore be considered another welcome result of the proposed split-DP analysis.

6.3

What Constitutes the Semantics of DEMs: The View from Korean and English

In Chap. 5, I identified a new affective function of the DEM ku in Korean and attributed its behavior to what I referred to as ‘the [+STS distant] feature’ (short for ‘spatio-temporally distant from the speaker’). In this section, by looking at the DEM system of Korean in comparison to that of English, I explore whether this feature can be reduced to or replaced by a more familiar set of semantic features or not. In this context, I identify the similarities and differences between the two DEM systems and also discuss how cross-linguistic variation pertaining to DEM meanings can be accounted for in an analysis that invokes a limited set of binary semantic features. To begin by providing some relevant background on the languages to be discussed: Korean DEMs instantiate a person-oriented system in the sense of Anderson and Keenan (1985: 280–288) in a manner similar to English, but while English has just proximal and distal DEMs (i.e., this/these versus that/those), Korean makes a three-way distinction for deictic DEMs, as outlined in (33). (33) Three-way DEM system in Korean: a. i ‘this’: the proximal form; refers to an entity that is perceived to be proximal to S. b. ku ‘that’: the neutral form; refers to an entity that is perceived to be distal from S but proximal to H. c. ce ‘that over there’: the distal form; refers to an entity that is perceived to be distal from both S and H. (adapted from Sohn 1999: 210)

6.3 What Constitutes the Semantics of DEMs

251

About the three-way DEM system in Korean, the first thing to note is that ce never occurs anaphorically whereas both i can ku can. To see this, consider (34). In this discourse, the bold-faced DP refers to an entity that has been introduced by the mention of the preceding utterance. And depending on how S perceives the problems at hand (e.g., whether she feels cognitively close to them or not), either i or ku may be used but ce cannot be. (34) Context: Speaker is talking to a friend about what’s been going on in her life lately. Minho-ka na-eykey [myechkaci mwunceycem-tuli]--ul cicekhaycwu-ess-ta. [several problem-PL]-ACC point.out-PST-DECL M.-NOM I-DAT ‘Minho has pointed out several problems to me.’ mwunceycem-tul-ul ettehkey Kurentey na-nun ii/kui/*cei But I-TOP this/that/that problem-PL-ACC how haykeyl-hayyahal-ci cal morukess-ta. solve-must-COMP well not.know-DECL But I don’t know how to solve these/those problems.’

Given the impossibility of using ce in contexts like (34), we can state that, for anaphoric purposes, Korean also makes a two-way distinction for its DEMs similarly to English. Therefore, I posit (35) as a first pass at the characteristic properties of purely anaphoric DEMs in Korean. (35) Purely anaphoric DEMs in Korean (first pass): a. i ‘this’ proximal b. ku ‘that’: distal/neutral

Making a close comparison between Korean DEMs and their English counterparts leads us to see, however, that what is given in (35) merits revision. To see this, consider first (36). This data set shows that while English this may occur in presentational contexts, referring to a specific but indefinite entity,6 Korean i cannot, even though its occurrence in such contexts will meet the characterization given in (35a) since the referent of the DP containing it is known to S, so is cognitively close to her; in such contexts, Korean employs either the numeral han ‘one’ or the indeterminate N modifier etten ‘some/which’, as shown below.

6

This strange man in (36) denotes a specific but indefinite entity because it is known to S but unknown to H (Lakoff 1974; Prince 1981; Maclaran 1982). Furthermore, its existence is not presupposed by both S and H. According to Ionin (2006) and Ionin et al. (2012), if a nominal is definite, then both S and H presuppose the existence of a unique individual associated with it.

252

6 Extensions, Refinements, and Prospects

(36) Indefinite and specific use: English this vs. Korean i Context: Speaker is talking to a friend about what happened to her recently. a. I went to the mall the other day and there was this strange man talking really loudly in the shoes section.7 b. Ecey __ mol-ey ka-ss-nuntey han/etten/*i one/some/this yesterday pro mall-to go-PST-but isangha-n namca-ka sinpal kakey-eyse strange-ADN man-NOM shoe store-LOC khun-sori-ro malulha-ko iss-ess-e. exist-PST-DECL.INFRML big-noise-with talk-CONN Intended: ‘I went to the mall yesterday and there was this strange man talking really loudly in the shoes store.’ 7

Consider now (37). This data set shows that, unlike English that, Korean ku cannot occur in out-of-the-blue contexts even though the intended referent of the DP containing it is distal from S in accordance with (35b) and the entire utterance is made with an exclamation, i.e., a heightened emotion, which is known to facilitate the occurrence of affective DEMs in English (see Potts and Schwarz 2010 and the references there, e.g., Lakoff 1974; Prince 1981; Liberman 2008); in such contexts, Korean requires a bare nominal, as indicated by the parentheses in (37b).8 (37) Occurrence in out-of-the-blue contexts: that vs. ku Context: A radio program host is talking to the listeners during a fund-raising season. a. Please pick up that phone and call us now! b. Cikum [(*ku) swuhwaki]-rul tul-ko cehuy pansongkwuk-uro Now [(that) phone-ACC] lift-CONN our.HON studio-to paro cenhwahay-cwu-sip-siyo! right.now call-give-HON-IMP.HON Intended: ‘Please pick up that phone and call us now!’

Even if we delimit our attention to anaphoric contexts, some additional differences between English and Korean still emerge, in particular with regard to the way

7 This property of this is shared by its German counterpart, as shown in (i), so it is not an English-specific fact.

(i) Gestern kam ich in eine Bar und da war dieser Fremde, der mich die ganze Zeit anstarrte. ‘Yesterday I walked into a bar and there was this stranger who stared at me all the time.’ (von Heusinger 2011: ex. (29); emphasis mine) 8

Some authors may postulate a zero determiner inside such nominal structures (e.g., Jo 2000; Simpson 1998; Suh 2005) but whether they contain a null determiner or not has no direct bearing on the issues at hand, so I use the term ‘bare nominal’ here.

6.3 What Constitutes the Semantics of DEMs

253

proximal DEMs behave. To see this, compare the (a) and the (b) examples in (38). Here, DEMs are being used anaphorically, carrying definite and specific meanings, and while English speakers only allow for this, Korean speakers only permit ku, despite the fact that the intended referent of the DP containing the DEM can be perceived as proximal to S. (38) Definite and specific use: English this vs. Korean i a. I’ve got a new roommate. I’ll ask this/*that guy if he’d be interested in buying your iPad. [adapted from Oshima and McCready 2017: (41)] b. Na sayrowun rwummeythu sayngkiess-e. I new roommate got.to.have- DECL.INFRML ‘I’ve got a new roommate.’ __ * i/ ku saram-hanthey __ ney aiphaytu pro this/that person-to pro your iPad kwansimissnun-ci mwulepwacwu-l-key. be.interested- COMP ask-will-ASSRT.INFRML Intended: ‘I’ll ask this/that guy if he’d be interested in your iPad.’

Consider now (39) and (40). The goodness of (39a) and (40a) shows that in English, both this and that may be licensed in certain anaphoric contexts although in contexts like (39a), the use of this would engender a more vivid narration of the story. Interestingly, in the same set of contexts, Korean speakers would strongly prefer ku over i, as shown in (39b), or they would only use ku, as shown in (40b). (39) Context: Speaker is at work, talking to a colleague. a. My neighbor has a dog. {This/that} dog kept me awake last night. (Gundel et al. 1993: 279) b. Nay iwus-hanthey kay-ka han-mari iss-e. dog-NOM one-CL exist-DECL.INFRML My neighbor-DAT ‘My neighbor has a dog.’ Kurentey ??i/ ku kay-ka ecey-pam-ey keysokhayse na-rul And this/that dog-NOM last-night-LOC continuously I-ACC kkaywu-ess-e.9 awake- PST-DECL.INFRML Intended: ‘And this/that dog kept me awake last night.’ 9

9

(39b) may be judged okay if it is uttered at the speaker’s home but in this discourse context, she is at work and this makes the use of i judged to be infelicitous.

254

6 Extensions, Refinements, and Prospects

(40) a. A: Do you remember that you, I, and [a student of mine]i had some discussion on null anaphora at the last conference? B: Sure. A: Well, {this/that}i student is going to finish his thesis, and he is hoping to have you as an external committee member. [adapted from Oshima and McCready 2017: (49)] b. A: Ne cinan hakhoy-eyse ne-rang, na-rang, kuriko You last conference-LOC you-with, I-with, and [nay haksayng han-myeng]i-irang kong taymyengsa-ey tayhay null pronoun-LOC about [my student one-CL]-with iyakiha-n kes sayngkakna-ni? NML remember-Q.INFRML converse-ADN ‘Do you remember that [a student of mine]i, you, and I talked about null anaphora at the last conference?’ B: Kurem. Sure ‘Sure.’ i pen-ey nonmwun-ul A: {*i/ku}i haksayng-i This/that student-NOM this time-LOC thesis-ACC simsa-pat-nuntey ne-rul oypwu wuywon-uro examination-receive-CONN you-ACC external member-as mosi-ko sip-tay. invite.HON-CONN wish-HS.INFRML Intended: ‘{This/that}i student is defending his thesis soon and hei says he wishes to have you as an external member (on hisi thesis committee).’

Taken together, this set of facts raises at least three (inter-related) questions: First, why Korean DEMs cannot carry indefinite and specific meanings in so-called presentational contexts whereas English DEMs can? Secondly, what prevents Korean proximal DEM i from (readily) occurring in anaphoric contexts like (38), (39), and (40), given that it can occur in contexts like (34)? Finally, what is the semantic property that is shared by all DEMs in human language? And what is/are the reason(s) for their cross-linguistic variation? In answer to these questions, I would like to first suggest that DEMs in any human language exhibit indexicality (i.e., pointing), yet they exhibit different properties because pointing can be done in different domains, aiming at different types of entities. More specifically, pointing can be done in either a physical or a non-physical (i.e., cognitive) domain, and its target can be specific or non-specific, discourse-old (DO) or discourse-new, and known or unknown to some relevant attitude holder (AH) (which may not be the same as the speaker) as well as known or unknown to the hearer. Furthermore, the target of the pointing may be perceived as proximal to both AH and the hearer, or just to one of them, or to neither of them. To put these ideas in more formal terms, I propose (41) as a (partial) list of semantic features that comprise DEMs.

6.3 What Constitutes the Semantics of DEMs

255

(41) List of semantic features DEMs can have (first pass): a. [+/_Phys]: the entity is presupposed to be present in the physical location of the discourse or not. b. [+/_DO]: the entity is discourse old or not. c. [+/_KtAH]: the entity is presupposed to be known to AH or not. d. [+/_KtH]: the entity is presupposed to be known to H or not. e. [+/_PrxAH]: the entity is perceived to be proximal to AH or not. f. [+/_PrxH]: the entity is perceived to be proximal to H or not.

I posit [+/_Phys] as a primary feature comprising DEM meanings in human language because not only in Korean but also in other languages, DEMs may behave differently according to whether they refer to an entity that is in a physically deictic space or not. To illustrate: recall that Korean ce cannot be used when pointing at something that exists in a non-physical domain, as shown in (35). Similarly, in its purely deictic use, the distal DEM ano (or a-series) in Japanese may be construed as pointing at an entity that is perceived as far away from both S and H (Oshima and McCready 2017: 804), but when used anaphorically, its intended referent has to be known to and thus cognitively proximal to both S and H (Kuno 1973: 29010; Ono 1994; Oshima and McCready 2017: 804). In other words, what appears to be the same lexeme in a language may function as a distal DEM when referring to an entity that exists in a physical domain but as a medial DEM when referring to an entity that exists in a non-physical domain (or is absent in the physical location of the discourse at hand). I posit [+/_DO] next because doing so will enable us to capture some of the differences between English-type languages and Korean-type ones with regard to DEM meanings: if we assume that English DEMs may be either [+DO] or [_DO] but Korean DEMs may only be [+DO], then we can readily explain why English DEMs may introduce new discourse referents but their Korean counterparts cannot. Furthermore, [+/_DO] subsumes [+/_definite], so positing this binary set of features as part of the core meanings of DEMs eliminates the need to address the controversial question of whether English DEMs inherently carry [+unique] semantics or just [+definite], a question that has figured prominently in the semantics literature but has not been settled as yet (see, a.o., Hawkins 1991; Roberts 2002; Wolter 2006); if some of them are [+definite] because they are [+DO] and if neither [+definite] nor [+unique] is part of the core meanings of DEM meanings in human language, then it follows that [+unique] meanings associated with English DEMs arise from a different source than what induces [+definite] meanings for them. As for positing the other features listed in (41), similar ideas have been suggested by other authors also (e.g., Oshima and McCready 2017 and the references there) but what is novel about the present analysis is that I am using AH rather than S as the point of view (POV) holder, by which I mean the individual who serves as

10 To be exacting, Kuno (1973) does not use the term ‘cognitively proximal’; instead, he characterizes the anaphoric uses of the a-series as referring to something that both S and H know personally (i.e., are acquainted with).

256

6 Extensions, Refinements, and Prospects

the perspectival center in deciding whether some entity is perceived as proximal or not. I differentiate between AH and S because when DEMs occur inside clausal complements of propositional attitude predicates, depending on whether the AH is construed as S or as the root clause’s subject, the felicity of the data containing them may vary as observed by Elbourne (2008) for English. To see this, consider (42) and (43). These paradigms show that both in English and Korean, if the POV holder is construed as S, then the distal DEM is preferred over the proximal one (i.e., the (a) cases), but if the POV holder is construed as the root clause’s subject, then the proximal DEM is (strongly) preferred (i.e., the (b) cases).11 12

Context for (42)-(43): Mary met with several senators recently and S is talking about it to a friend. (42) a. Maryi talked to no senator without declaring afterwards that ?this/ that senator was the one who would cosponsor heri bill. b. Maryi talked to no senator without thinking at the time that this/?that senator was the one who would cosponsor heri bill. [adapted from Elbourne 2008: (86)] etten sangwonuywon-to manna-ko nase ?i/ ku (43) a. Maryi-nun every senator-also meet-CONN after this/that M.-TOP sangwonuywon-i casini-uy pepan-ul cicihaycwu-l-ke-rako senator-NOM self-GEN bill-ACC support-will-NML-COMP palphyoha-ci-ahn-un saram-i ep-ta.12 declare-COMP-AUX.NEG-ADN person-NOM not.exist-DECL Intended: ‘There is no senator Maryi talked to without declaring afterwards that this/that senator was the one who would cosponsor heri bill.’ etten sangwonuywon-to mithing-cwung-ey b. Maryi-nun M.-TOP every senator-also meeting-in.the.middle-LOC i/*ku sangwonuywon-i casini-uy pepan-ul self-GEN bill-ACC this/that senator-NOM cicihaycwul-ke-rako mit-ci-ahn-un saram-i person-NOM support-will-NML-COMP believe-COMP-AUX.NEG-ADN ep-ta. not.exist-DECL Intended: ‘There is no senator Maryi talked to without believing at the time that this/that senator was the one who would cosponsor heri bill.’

11

These two paradigms also show that both in English and Korean, DEMs can have bound variable interpretations. 12 Here, the choice of DEM i would be judged acceptable only if Mary was standing by the senators that she had a visit with when she was making the public announcement.

6.3 What Constitutes the Semantics of DEMs

257

Differentiating between AH and S is further motivated by the fact that not only in English but also in Korean, proximal DEMs may be licensed even if the POV holder is not construed as S. By way of illustration, in (44) and (45), the AH is construed as Cinderella who is the protagonist of the story being told, and the choice of this over that and i over ku encodes how the intended referent of the DP is perceived as cognitively proximal to her, not to S. (44) Cinderellai’s godmother told heri that someday heri Prince Charming will come and recue heri and shei wonders when {this/that} Prince Charming will come. (45) Sinteyreylra-uy taymo-nim-un __ kot mesci-n soon handsome-ADN Cinderella-GEN godmother-HON-TOP pro wangca-nim-i Sinterelra-rul kwuha-re o-l-kerako prince-HON-TOP Cinderella-ACC rescue-to come-will-COMP malhay-ss-ta. Kurentey Sinteyreylra-nun {i/ku} say-PST-DECL But Cinderella-TOP this/that wangca-nim-i encey o-l-ci kwungkumha-ta. come-will-COMP wonder-DECL prince-HON-NOM when Intended: ‘Cinderellai’s godmother told heri that a handsome prince will come and recue heri soon and shei wonders when this/that prince will come.’

On the basis of these empirical facts, I suggest that English and Korean DEMs have the feature specifications given in (46) and (47). Comparing (46A) and (47A) leads us to see that when it comes to physically deictic DEMs, the only difference between the two languages is that what would be expressed by that in English is expressed by two different DEMs in Korean, namely, ku and ce. When it comes to non-physically-referring or anaphoric DEMs, things look similar except that in English, DEMs may be [+/_DO] but in Korean, they may only be [+DO]. (46) English DEMs and their feature specifications (first pass): A. Physically deictic Phys

DO

KtAH

KtH

This

+

N/A

+

+

PrxAH +

PrxH _

That

+

N/A

+

+

_

+/_

B. Non-physically deictic: Anaphoric or presentational Phys

DO

KtAH

KtH

PrxAH

PrxH

This

_

+/_

+/_

+/_

+

_

That

_

+/_

+/_

+/_

_

+/_

258

6 Extensions, Refinements, and Prospects

(47) Korean DEMs and their feature specifications (first pass): A. Physically deictic Phys DO KtAH KtH PrxAH

PrxH

i

+

N/A

+

+

+

_

ku

+

N/A

+

+

_

+

ce

+

N/A

+

+

_

_

B. Non-physically deictic: Purely anaphoric Phys DO KtAH

KtH

PrxAH

PrxH

i

_

+

+/_

+/_

+

_

ku

_

+

+/_

+/_

_

+/_

Under the present analysis, then, the meanings of the indefinite and specific DEMs this and that in (36a) and (37a) can be represented as in (48), where AH is construed as S. (48) Analysis of indefinite and specific this and that in (36a) and (37a): Non-physically deictic: Presentational or out-of-the-blue usage AH = S Phys DO KtAH KtH PrxAH

PrxH

This

_

_

+

_

+

_

That

_

_

_

+

_

+

If we are to represent the meanings of the i and ku that occur in (36b) and (37b), they would have the feature specifications given in (49), just like what we see for this and that in (36a) and (37a). And this explains why sentences (36b) and (37b) are judged bad when they are read with the DEMs occurring in them: Korean DEMs may only carry [+DO], so they are not licensed in contexts where their intended referents have not been introduced to the discourse—that is, when they do not have linguistic antecedents. Consequently, Korean expresses the intended determiner meanings by employing the NUM han ‘one’ or the indeterminate ADJ etten ‘some/ which’ for contexts like (36b) (i.e., cases where the intended referent of the DP is not presupposed to be known to H) and a bare nominal for contexts like (37b) (i.e. cases where the intended referent of the DP is presupposed to be known to H).

6.3 What Constitutes the Semantics of DEMs

259

(49) Intended meanings of i and ku in (36b) and (37b): Non-physically deictic: Presentational or out-of-the-blue usage AH = S Phys DO KtAH KtH PrxAH

PrxH

*i

_

_

+

_

+

_

*ku

_

_

_

+

_

+

What I have just put forward also correctly predicts that contexts like (34) will license both i and ku because such environments are [+DO], as given in (50). Furthermore, it lets us capture the subtle meaning differences associated with the choice between i and ku in (34): the choice of i indicates that S perceives the problems introduced in the preceding sentence as cognitively close to her. For instance, i can be used in a context where S has been struggling to resolve them, so they are felt to be “right with her” (so to speak). The choice of ku, on the other hand, indicates that she has not done much to actually resolve the problems, and this agrees with Korean native speakers’ intuitions.13 (50) Analysis of definite and specific i and ku in (34): Non-physically deictic: Purely anaphoric AH = S Phys DO KtAH KtH

PrxAH

PrxH

i

_

+

+

_

+

_

ku

_

+

+

_

_

_

The proposed analysis provides a natural account of the distribution of the DEMs in contexts like (42)–(43) as well. In (42a) and (43a), the non-proximal DEM is better suited because here, the AH is construed as S, and from her POV, the senators Mary talked to are distal entities, as shown in (51). If the AH is construed as Mary instead, then, this or i may be used but the presence of the bold-faced temporal adverbial clause in the sentence makes such an interpretation difficult to obtain. In the case of (42b) and (43b), the AH is readily construed as Mary and from her POV, each senator she is having a meeting with is perceived as close to her, so in such discourses, a proximal DEM is judged to be more felicitous than a distal one, as shown in (52).

13

I thank Duk-Ho An, Hee-Rahk Chae, Heejeong Ko, and Sung-Won Lee for the grammaticality judgements on the Korean data presented here.

260

6 Extensions, Refinements, and Prospects

(51) Analysis of bound DEMs that and ku in (42a) and (43a): Anaphoric AH = S Phys DO KtAH KtH PrxAH that/ku

_

+

+

_

(52) Analysis of bound DEMs this and i in (42b) and (43b): Anaphoric AH = Mary Phys DO KtAH KtH this/i

_

+

+

_

_

PrxAH +

PrxH _

PrxH _

In a similar vein, under our analysis, the distribution of the DEMs in narrative contexts like (44)–(45) can be explained as follows: in (44), this is preferred over that because S is telling a story from Cinderella’s POV and since her godmother told her that her Prince Charming will appear sometime soon, she can perceive him as a cognitively proximal entity. If, on the other hand, she believes that there are slim chances for him to come and rescue her, then that can be used felicitously, as I have verified with several English speakers.14 The foregoing suggests that the analysis I have put forth has positive prospects, but as things currently stand, this analysis wrongly predicts that both English this and Korean i will occur in contexts like (38), (39), and (40). This is because, in all these contexts, the intended referent of the DP containing the DEM can be analyzed as cognitively proximal to S, who serves as the AH, as indicated in the following tables:

14

I wish to thank Angela Carpenter, Kyle B. Johnson, Keir Moulton, and Anne-Michelle Tessier for the judgments on the English data.

6.3 What Constitutes the Semantics of DEMs

261

(53) Analysis of definite and specific DEMs in (38): Anaphoric AH = S Phys DO KtAH KtH this/i that/ku

_ _

+ +

+ +

_ _

(54) Analysis of definite and specific DEMs in (40): Anaphoric AH = S Phys DO KtAH KtH this/i that/ku

_ _

+ +

+ +

+ +

PrxAH

PrxH

+ _

_ _

PrxAH

PrxH

+ _

_ _

So, why does Korean only allow for distal DEMs in certain anaphoric contexts where English would permit both proximal and distal DEMs? Put differently, why does English this may occur in anaphoric contexts like (38)–(40) but Korean i cannot? I suggest that things are the way they are because purely anaphorically used Korean proximal DEM is subject to a licensing condition that its English counterpart is not. More concretely, a DP containing an anaphoric i requires that the sentence containing its immediate antecedent be temporally overlapping with the sentence in which it occurs. Support for this idea comes from the fact that (55), a variant of (38b), is judged fine even though it contains i. In this discourse, the highlighted part overtly indicates that the sentence containing the immediate antecedent of i temporally overlaps with the sentence containing the DEM. By contrast, in (38b), the situation denoted by the first sentence obtains sometime before now whereas the situation denoted by the second sentence will obtain sometime after now. Therefore, the DEM i is not licensed in that discourse. (55) Na sayrowun rwummeythu sayngkiess-e. I new roommate got.to.have-DECL.INFRML ‘I’ve got a new roommate.’ Cikum __ thonghwaha-ko iss-nuntey Now pro talk.on.the.phone-CONN exist-CONN __ i saram-hanthey __ ney aiphaytu pro this person-DAT pro your iPad kwanshimissnun-ci mwulepwacwu-l-key. be.interested-COMP ask-will-ASSRT.INFRML ‘I’m talking to him on the phone right now. Let me ask this guy if he’d be interested in your iPad.’

262

6 Extensions, Refinements, and Prospects

Comparing (56) with (40b) points to essentially the same pattern. In (56) too, the highlighted parts overtly establish a temporal link between the sentences containing i and its immediate antecedent in a manner analogous to (55), unlike the case with (40b) or (38b). (56) A: Ne cinan hakhoy-eyse ne-rang, na-rang, kuriko You last conference-LOC you-with, I-with, and [nay haksayng han-myeng]i-irang kong taymyengsa-ey tayhay [my student one-CL]-with null pronoun-LOC about iyakiha-n kes sayngkakna-ni? converse-ADN NML remember-Q.INFRML ‘Do you remember that [a student of mine]i, you, and I talked about null anaphora at the last conference?’ B: Kurem. Sure ‘Sure.’ A: Ku ihwu-ro kui haksayng-hako il-cwuil-ey That after-since that student-with one-week-LOC han-pen-ssik myentam-ul hay-o-ko iss-nuntey kapcaki one-CL-each meeting-ACC do-come-CONN exist-and suddenly haksayng-i ne-rul cakii nonmwun oypwu ii this student-NOM you-ACC self thesis external simsa-wuywon-uro mosi-ko sip-tay. examination-member-as invite.HON-CONN wish-HS.INFRML ‘Since then, thati student and I have been having weekly meetings and suddenly, thisi student says that hei wants to have you as an external member on hisi thesis committee.’

Turning now to (39b), the situation is slightly different here because the first sentence in this discourse is in the present tense whereas the second sentence is in the past tense, so in order to license i, we need to have a way to overtly indicate that the second sentence’s meaning is temporally accessible from the first sentence’s time, namely now, and by inserting a past imperfective narrative marker in the second sentence, we can achieve such a result. To see this, consider (57). In this discourse, the second sentence contains the past imperfective narrative marker -nunkey-ya, and this gives rise to an impression that S is telling a story about what happened to her last night as if she is re-experiencing it by placing her POV in the middle of that eventuality. Such a vivid story-telling strategy allows the first and second sentences’ times overlap in some non-physical space. Furthermore, it lets the dog under description perceived as cognitively proximal to S, who serves as the AH. Consequently, i is licensed here, unlike in (39b). Notably, such usage of i can be considered affective, and this goes hand-in-hand with the fact that (57) contains the epithet nom-uy, which is literally glossed as ‘guy-genitive’ but which actually carries a pejorative meaning as well as the conventional implicature that S has a negative attitude toward its referent, as its English translation ‘awful’ suggests. Besides, the second utterance of this discourse is made in an extremely heightened

6.3 What Constitutes the Semantics of DEMs

263

tone of voice accompanied by an exclamation, a property that is shared by the affective use of this in English (Potts and Schwarz 201015). (57) Context: Speaker is at work, talking to a colleague. Nay iwus-hanthey kay-ka han-mari iss-e. My neighbor-DAT dog-NOM one-CL exist-DECL.INFRML ‘My neighbor has a dog.’ nom-uy kay-ka ecey-pam-ey keysokhayse Kurentey i dog-NOM last-night-LOC continuously And this guy-GEN mak cise-tay-nun-ke-ya! relentlessly bark-attack-ADN.IMPRF-NML-COP.ASSRT.INFRML ‘And I’m telling you this awful dog was constantly barking all night last night!’

In contrast to Korean i, anaphoric this may be licensed in any contexts as long as its referent is perceived as proximal to AH. To see this, consider (58). In this discourse, the intended referent of this student is absent in the discourse context. It is also contextually obvious that S is not well acquainted with him. Furthermore, its intended referent is not cognitively proximal to the relevant POV holder (namely, H). Even so, the proximal DEM this can be licensed here. (58) A: Do you remember that you, I, and a student of yours had some discussion on null anaphora at the last conference? B: Sure. A: Is this student still around? Our project team needs some help from a native speaker of Japanese. [adapted from Oshima and McCready 2017: (48)]

In light of this, I conclude that anaphorically used Korean proximal DEM i differs from English this in that, in addition to being invariably [+DO], it carries what I abbreviate as a [+Tm] feature, which is short for ‘temporal’ and which dictates that its antecedent refer to an entity that is temporally accessible to AH. And given this conclusion, I offer (59), (60), and (61) as amendments of (41), (46) and (47), respectively. I posit that in any language, physically deictic DEMs carry a [+Tm] feature because their referents have to be present in the discourse context. As for that, I assume that just like this, it carries [+/_Tm] because it may be licensed regardless of whether its referent is temporally accessible to AH or not, as one can see from the goodness of both (37a) and (62): in (37a), the referent of that

15

Potts and Schwarz (2010: 6) note that many of the utterances containing affective this that have been documented in the literature are exclamatives or contain exclamative markers (e.g., really), as shown in (i).

(i) a. This Henry Kissinger is really something! (Lakoff 1974; emphasis mine) b. [In front of a computer] These IBM ThinkPads are amazing! (Bowdle and Ward 1995; emphasis mine)

264

6 Extensions, Refinements, and Prospects

is temporally accessible to AH but in (62), it is temporally inaccessible to her since the sentence containing its immediate antecedent is temporally not linked to the sentence in which it occurs. As for anaphorically used ku in Korean, I suggest that it is [+/_Tm] because it can occur in contexts like (63) as well as in contexts like (34) and (45): in (63), due to the presence of the temporal adverb yocum ‘these days’ and the present progressive tense/aspect marking in the first sentence, the two sentences are temporally linked to each other and for this reason, while i is preferred over ku, the distal DEM is not absolutely out in the second sentence. (59) Semantic features DEMs can have (revised): a. [+/_Phys]: the entity is presupposed to be present in the physical location of the discourse or not. b. [+/_DO]: the entity is discourse old or not. c. [+/_KtAH]: the entity is presupposed to be known to AH or not. d. [+/_KtH]: the entity is presupposed to be known to H or not. e. [+/_PrxAH]: the entity is perceived to be proximal to AH or not. f. [+/_PrxH]: the entity is perceived to be proximal to H or not. g. [+/_Tm]: the entity is temporally accessible to AH or not. (60) English DEMs and their feature specifications (revised): A. Physically deictic Phys DO KtAH KtH PrxAH PrxH

Tm

This

+

N/A

+

+

+

_

+

That

+

N/A

+

+

_

+/_

+

B. Non-physically deictic: Anaphoric or presentative Phys DO KtAH KtH PrxAH

PrxH

Tm

This

_

+/_

+/_

+/_

+

_

+/_

That

_

+/_

+/_

+/_

_

+/_

+/_

(61) Korean DEMs and their feature specifications (revised): A. Physically deictic: Purely deictic Phys DO KtAH KtH PrxAH PrxH

Tm

i

+

N/A

+

+

+

_

+

ku

+

N/A

+

+

_

+

+

ce

+

N/A

+

+

_

_

+

PrxH

Tm

B. Non-physically deictic: Purely anaphoric Phys DO KtAH KtH

PrxAH

i

_

+

+/_

+/_

+

_

+

ku

_

+

+/_

+/_

_

+/_

+/_

6.3 What Constitutes the Semantics of DEMs

265

(62) John brought up three issues in our meeting last week and I don’t know how to address those issues. (63) Context: Speaker is visiting with a friend at a café. Na yocum etten saram-kwa manna-ko I these.days some person-with see-CONN iss- -e. exist-N.PST-DECL.INFRML ‘I’m seeing someone these days.’ Kun-tey na i/?ku saram manhi coh- -a. And I this /that person a.lot like-N.PST-DECL.INFRML ‘And I really like this or the/that guy.’

Turning now to taking up the question of whether the analysis I have offered here can capture the semantic properties of what I call CogDems in Korean: our discussion in Chap. 5 shows that the CogDem ku, which we notate as KU, is licensed when the intended referent of the DP containing it is absent in the spatio-temporal location of the discourse. Therefore, we can posit that it carries a [_Phys] feature. We can also hypothesize that it is inherently [+DO; +KtAH; +/_KtH; +PrxAH; +/_PrxH] on the basis of the fact that its referent has to be discourse-old (or has been introduced to the cognitive domain in which it is being talked about),16 known to S although not necessarily to H, and cognitively proximal to S (who serves the AH in discourses containing KU) though not necessarily to H. In addition, we can posit that KU carries [_Tm] because the sentence containing its immediate antecedent has to be temporally non-intersecting with the sentence in which it occurs. Notably, the feature specification given above for KU is similar to what I have proposed for a purely anaphoric ku in (61B) except that while ku may be positively or negatively valued with regard to [KtAH], KU has to be only positively valued for this feature, and while ku is always negatively valued with respect to [PrxAH], KU has to be always positively valued for it, and finally, while ku can be either positively or negatively valued for [Tm], KU has to be invariably negatively valued for it (modulo the fact that it selects for an AP complement, which ku does not). When viewed in this way, what I referred to as the [+STS distant] feature in Chap. 5 can be simply re-labeled as [_Tm], which is one of the features that every instance of ku may potentially carry anyway. That said, KU is still different from the more “ordinary” purely anaphoric ku because it carries [+PrxAH] and it is

16

This means that even in contexts like (66) below, where there is no discourse participant other than S, KU carries [+DO] feature.

266

6 Extensions, Refinements, and Prospects

always [+KtAH]. Therefore, I offer (64) as an amendment of (61), with the feature specifications for cognitive indexically used i and ku added to it.17 (64) Korean DEMs and their feature specifications (re-revised): A. Physically deictic: Purely deictic Phys DO KtAH KtH PrxAH PrxH

Tm

i

+

N/A

+

+

+

_

ku

+

N/A

+

+

_

+

+

ce

+

N/A

+

+

_

_

+

B. Non-physically deictic: Purely anaphoric Phys DO KtAH KtH

PrxAH

+

PrxH

Tm

i

_

+

+/_

+/_

+

_

+

ku

_

+

+/_

+/_

_

+/_

+/_

C. Cognitive indexical uses (i.e., as CogDems) Phys DO KtAH KtH PrxAH

PrxH

Tm

i

_

+

+

+/_

+

+/_

+

ku

_

+

+

+/_

+

+/_

_

Notably, positing [+PrxAH] for what I call KU lets us explain why Korean speakers intuit that the referent of a DP it modifies is felt to be “visible” in the mental domain of the speaker even though it is neither spatially nor temporally close to the discourse location, as exemplified by (66). And this provides further justification for positing [+/_PrxAH] as a separate set of features from both [+/_Tm] and [+/_KtAH].

17

As far as I am aware, what I have presented here is a novel analysis of DEMs in Korean as well as English DEMs, and it differs from any other existing analysis in non-trivial ways. For instance, about the distal DEM ku, Kang (2012, 2015) claims that regardless of whether it is used deictically or anaphorically, a felicitous use of it requires that its referent be familiar within the discourse and/ or maximally identifiable to the hearer but her analysis makes rather different predictions on its behavior. In addition, unlike here, she does not differentiate between physically deictic, anaphoric, and cognitively deictic occurrences of ku, although the data we have looked at thus far seems to call for such a fine-grained analysis of the DEM.

6.3 What Constitutes the Semantics of DEMs

267

(66) Context: Yesterday Mina bought and brought some expensive items for the speaker and she left them in the kitchen. The speaker is currently in her bedroom, which is far away from the kitchen, so she cannot see the items Mina brought for her. And she is talking to herself. pro [DP [ku [AP kappissa-n], [RC Mina-ka ei sa-o- ]-n [ expensive-ADN] [ M.-NOM __ buy-come-PRF]-REL __ [ [KU mwulken-tuli]-ul eti-ey twue-yahal-kka? where-LOC keep-must-Q? item-PL-]-ACC ‘Where should I keep those expensive items that Mina bought and brought for me?’

In sum, then, even though presenting a fuller picture of the semantics and pragmatics of DEMs has to await another occasion, what I have proposed here can serve as a good starting point for that investigation. If correct, the present analysis shows that what cuts across all DEMs in human language is pointing but DEMs can be divided into subcategories, depending on which of the following feature combinations fits their semantic characteristics: [+/_Phys; +/_DO; +/_KtAH; +/_KtH; +/_PrxAH; +/_PrxH; +/_Tm].

6.4

More on the Internal Structure of a DemP: Is a Uniform Analysis Possible?

In Chap. 4, I proposed that if a referential DP contains a DEM, the DemP headed by the DEM merges at [Spec, LocP], and this FP later raises to [Spec, DPd/r] to license the relevant D features of the DPd/r. Then in Chap. 5, I argued that Korean DEMs may function as what I call CogDems and when they do, they select for an AP as given in (67). In addition, I suggested that CogDemPs also have a slightly different external syntax than more “ordinary” DemPs in the sense that they may be base-generated at [Spec, LocP], later moving to [Spec, DPd/r], or they may directly merge at a [Spec, DPd/r] that is created atop the pre-existing [Spec, DPd/r]. (67) Internal structure of a CogDemP headed by KU in Korean: CogDemPS CogDem S CogDemS KU

AP

In the previous section, however, I have shown that there is a way to treat more “ordinary” DemPs and CogDemPs on a par with each other. Given this, in this section, I would like to explore the possibility that all types of DemP in human language have a uniform structure. More specifically, I wish to examine whether they all have the structure sketched in (68), in which they select for an AP

268

6 Extensions, Refinements, and Prospects

complement, although the precise feature specifications constituting their semantics may vary and this may be reflected in their morpho-phonological and external syntactic properties. (68) Hypothesis on the internal structure of DemP in human language: DemP Dem Dem

AP

If we assume (68), then the DemP headed by ku in the Korean data (69) will be rendered to contain a silent AP, and since the DEM performs an anaphoric function here, it will be assumed to carry the same set of features as what I posited for ku in (64B), as schematically represented in (70). (69) Thom-i ecey nay-key chayk-ul han-kwon sa-cwu-ess-ta. book-ACC one-CL buy-give-PST-DECL Tom-NOM yesterday I-NOM ‘Tom bought me a book. chayk]-i acwu caymiiss-ess-ta. Kurentey [[DemP ku] And [[ that] book]-NOM very be.interesting-PST-DECL ‘And the book was very interesting.’ (70) Internal structure of the DemP in (69): A hypothesis DemP Dem Dem [_Phys] [+DO] [+KtAH] [_KtH] [_PrxAH] [_PrxH] [_Tm] ku

AP

At first glance, what is depicted in (70) seems implausible because Korean DEMs can occur inside a DP without ever having to be accompanied by an AP, as illustrated in (71). (71) Korean DPs comprised only of a DEM and an N: ku/i/ce saram/namca/sonye/chayk that/this/that person/man/girl/book ‘that/this/that person/man/girl/book’

6.4 More on the Internal Structure of a DemP

269

When approached from a more cross-linguistic perspective, however, this idea does not seem so unlikely. The reason is that in several languages or language varieties, referential nominals receiving indexical interpretations contain what Bernstein (1997) calls ‘a reinforcer’ (which amounts to here or there in English) and this suggests that, underlyingly, deictic DPs contain a slot for a locative element, as have been argued by numerous authors (e.g., Delsing 1993; Donaldson 1993; Bernstein 1997; Vangsnes 2004; Julien 2005; Roehrs 2010; Leu 2008, 2015): (72) Non-standard American English (Bernstein 1997): this here book (73) Afrikaans [Leu 2015: ex. (2.19)]: a. Ek het hier-die huis gebou. I have here-the house built ‘I built this house.’ b. Ek het daar-die man gesien. I have there-the man seen ‘I saw that man.’ (74) Colloquial Norwegian [Leu 2015: ex. (2.48)]: a. den her(r)-e klokka the here-INFL watch ‘this watch’ b. den her(r)-e huset the there-INFL house ‘that house’ (75) Swiss German [Leu 2015: ex. (2.53)]: das hus da/dettä this/that house here/there ‘this/that house’

On the basis of data like (72)–(75) as well as additional empirical evidence, Leu (2008, 2015) proposes that DemPs are extended projections of ADJs (xAPs for short), which are comprised of (i) a definite marker, (ii) an ADJ agreement head (AGRca), and (iii) a null locative element which he notates as ‘HERE’ or ‘THERE’, as given in (76). (76) Leu’s (2008, 2015) proposal for the internal structure of a DemP: [xAP the-AGRca HERE/THERE] N

In addition, he assumes that not only DemPs but also ‘definite article + ADJ’ strings occupy [Spec, DP] and that the head positions of such DPs are occupied by a null determiner which he notates as ‘THE’. To illustrate, under his analysis, the two English DPs in (77) have essentially the same internal structure except that (77a) contains an open class ADJ large whereas (77b) contains a closed class ADJ HERE, as given in (78a, b), respectively.

270

6 Extensions, Refinements, and Prospects

(77) a. the large house b. this house (78) a. Structure of the large house [Leu 2015, (3.6b)]: DP

xAP D0

NP

THE

house

the (AgrA) large

b. Structure of this house [Leu 2015, (3.5b)]: DP

xAP D0

NP

THE

house

the (AgrA) HERE

In view of such advances made in recent syntactic literature, then, there is nothing so unusual about the idea that a DemP contains a null AP; obviously, such an idea has been floating around for quite some time. Notably, however, the present analysis differs from the existing analyses in several important respects. To show this by comparing it with Leu’s (2008, 2015): first, in his analysis, a DemP is headed by an adjectival element HERE but in my analysis, it is headed by a DEM. Secondly, under his analysis, the underlying structure of any definite description contains a null definite article THE, be it comprised of a DEM plus an N or a definite article plus an ADJ and an N. Under my analysis, definite descriptions may have rather different internal structures depending on what constitutes them. By way of illustration, on the present analysis, the large house in (77a) is rendered the structure given in (79a), where the definite article the heads the LocP and the lexical ADJ large is base-generated inside the Low Field of the DP along with the head N. On the other hand, the deictic DP this house in (77b) is rendered the structure sketched in (79b), in which the DemP is base-generated at [Spec, LocP], with the head N base-generated inside the Low Field of the DP. (For simplicity, I abstract away from the exact feature specifications of the and this here. Also, in (79b), I do not represent the null AP inside the DemP headed by this.)

6.4 More on the Internal Structure of a DemP

271

(79) Structure of English definite descriptions under the present analysis: a. Structure of the large house: DPd/r Dd/r Dd/r [+REF] [+DEF] [-DEIC] [+SPEC]

LocP Loc Loc the

PlP+

large house b. Structure of this house: DPd/r Dd/r Dd/r [+REF] [+DEF] DemP [+DEIC] [+SPEC] this

LocP Loc Loc

PlP+

(large) house

Given the limited space we have in this book, we cannot fully address the question of which between these two lines of analysis may better capture the syntax and semantics of DemPs and their cross-linguistic variation. But I would like to point out that the behavior of the adjectival definite article ta in Slovenian presents at least four problems to Leu’s analysis. First of all, Leu treats ta as a DEM, yet as Marušič and Žaucer (M&Z) (2007, 2013, 2014) note, this lexeme is never stressed whereas a “true” DEM in Slovenian can be. Therefore, treating it as a DEM merits rethinking. Secondly, ta always co-occurs with an open class, property-denoting ADJ, as shown in (80), yet this property does not fall out naturally from Leu’s analysis of DEMs because, for him, DEMs are definite adjectival phrases headed by a functional ADJ HERE/THERE. (80) Ta’s syntactic property #1: obligatory co-occurrence with a property-denoting AP ta *(velika) knjga the (big) book ‘the big book’ (M&Z 2007: ex. (1); emphasis mine)

272

6 Extensions, Refinements, and Prospects

Thirdly, ta may occur inside an indefinite DP, immediately following an indefinite article, as shown in (81), and such a behavior is problematic for Leu’s analysis because under his analysis, DemPs contain a definite article, so they cannot be compatible with DPs that are headed by an indefinite article (cf. Marušič and Žaucer 201418). (81) Ta’s syntactic property #2: ability to occur inside an indefinite DP Lihkar ja mim prdirkal en ta hiter avto. Just-now AUX by speeded a TA fast car ‘Some fast car has just sped by.’ (M&Z 2007: ex. (2); emphasis mine)

Fourth, ta can co-occur with a “true” DEM such as tá (the stressed ta meaning ‘this’), tisti ‘that’, and tistele ‘those’, as shown in (82), and under Leu’s analysis, their co-occurrences cannot receive a satisfactory account: if we take (78b) as the structure for (77b), then with the xAP position occupied by tá, the ‘ta + AP’ strings in (82a, b) will have no suitable position to occur in because his xAPs are projections for “true” DEMs and the rest of the DP structure has space for D0 and NP only. (82) Ta’s syntactic property #3: ability to co-occur with “true” DEMs [ta zelen] svin nik a. tá green] pencil this [TA ‘this green pencil’ tistele [ta zelene] hlace. b. Prinesi mi green] pants bring me those [TA ‘Bring me those green pants.’ (M&Z 2007: ex. (12a) and (25a); emphasis mine)

At this juncture, I should note that ta resembles what I call CogDem KU in Korean in several ways (although an astute reader might have noticed it already). Firstly, the fact that ta requires an AP to occur inside a DP structure as shown above is reminiscent of what we observed with KU in Chap. 5 (see Sect. 5.4). Secondly, ta may occur iteratively and when it does, the APs it accompanies may occur in non-canonical orders (Marušič and Žaucer 2007) and this is analogous to what we saw with data containing KU. To illustrate, comparing (83) and (84) shows that, ordinarily, Slovenian ADJs may occur only in certain orders

Marušič and Žaucer (2014) offer a similar critique of Leu’s analysis although they end up analyzing ta as some sort of definite article that occurs inside an extended projection of an ADJ as well.

18

6.4 More on the Internal Structure of a DemP

273

(e.g., ‘size > color’) but when co-occurring with ta, they can enjoy more syntactic freedom.19 (83) Strict ordering of Slovenian ADJs: a. debeu zelen svin nik thick green pencil ‘a thick green pencil’ b. *zelen debeu svin nik

(size > color)

(color > size) (M&Z 2007: ex. (13); emphasis mine)

(84) Non-strict ordering of Slovenian ADJs when co-occurring with ta: [ta debeu] [ta zelen] svin nik a. tá (size > color) this [TA thick] [TA green] pencil ‘this thick green pencil’ b. tá [ta zelen] [ta debeu] svin nik (color > size) [adapted from M&Z 2007: ex. (12)]

Next, as shown in (82), ta can co-occur with a “true” DEM and this parallels the behavior of KU exemplified in (85) (repeated from (146b), Chap. 5): in (85), the first occurrence of ku is what I call KU and the second one is the “true” distal DEM ku, and while the first one can be omitted along with the AP it selects for, the second one cannot, respectively resembling the behavior of ta and the “true” DEM it co-occurs with. (85) Discourse containing a DP comprised of ‘KU + AP > anaphoric ku + N’: Pihayngki-an-eyse nay yeph-cari-ey [[maywu omanpwulsonha-n] Plane-inside-LOC my next.seat-LOC [[very arrogant.rude-ADN ] etten namca-ai]i -ka anc-key-toy-ess-ta. some male-kid]-NOM sit-CAU-happened.to.be-PST-DECL ‘In the plane, a very arrogant and rude young boyi happened to be sitting next to me.’ Kurentey, nacwung-ey al-ko po-ni, But, later.time-LOC know-CONN see-COMP, *(kui) namca-ai]-nun [DP ([kui [AP omanpwulsonha-n]],) (that) male-kid]-TOP [ ([KU [ arrogant.and.rude-ADN]],) paro Chelswu-uy tongsayng-i-ess-ta. younger.sibling-COP-PST-DECL exactly C.-GEN ‘But later, I learned that thati (arrogant and rude) boy is actually Chelswu’s younger brother.’

19

According to Lanko Marušič (personal communication), co-occurring with ta does not let ADJs occur in any free order relative to each other; rather, it allows them to occur at the beginning of a DP. Yet, given the existence of data like (84), it cannot be denied that co-occurrence with ta impacts their positional possibilities.

274

6 Extensions, Refinements, and Prospects

Finally, Marušič and Žaucer (2007, 2014) remark that adding ta to a nominal lets the speaker indicate to what degree the property that is denoted by the co-occurring ADJ holds true of the entity denoted by the nominal, and this is also reminiscent of what I claimed about data containing KU in Chap. 5, namely that “its occurrence in a discourse lets the speaker comment on some discourse familiar entity that has a noteworthy property. More specifically, it indicates that the degree to which the property denoted by its complement holds true of the referent denoted by its hosting DP is very high and the speaker can visualize it even though she cannot actually see the discourse referent at the time/location of the discourse” (p. 218-9).20 In light of such parallels, even though further investigation will be necessary to have a more accurate and comprehensive understanding of ta,21 treating it on a par with KU seems like a highly promising approach, and in fact, when we treat it as a subcategory of what I call a CogDem, many of its seemingly mysterious properties are no longer mysterious. First off, treating ta as some sort of CogDem immediately explains why it must co-occur with a property-denoting lexical ADJ; this is because a CogDem selects for an AP. Secondly, under the present analysis, a CogDemP can enter the DP structure by merging at a [Spec, DPd/r] that is created atop the existing DPd/r and therefore, if a ‘ta + AP’ string merges by occupying an adjoined [Spec, DPd/r] position, then it should be able to occur iteratively, since multiple [Spec, DPd/r] positions can be created on top of each other, as we saw in Chap 5.22 Moreover, being adjuncts, ‘ta + AP’ strings are expected to occur in any order relative to each other (provided that they meet certain language-specific constraints on AOR if there are any). Thirdly, nothing in our analysis dictates that every CogDemP be [+definite], so if ta encodes [+specific] meaning, as Marušič and Žaucer (2007) suggest, then we can posit that it carries [_DO; +KtAH] features, unlike KU, and this will explain its occurrence inside an indefinite DP, as exemplified in (81). Finally, the fact that a ‘ta + AP’ string may occur following a “true” DEM can receive an account if we assume that the true DEM tá merges at [Spec, LocP] and the ‘ta + AP’ merges at an adjoined [Spec, DPd/r] but the former later raises to a

Notably, I reached this conclusion about KU entirely independently of Marušič and Žaucer’s work; I was not aware of the behavior of Slovenian ta when I first looked at the relevant Korean data and began developing the formal framework presented here. Therefore, the fact that there is actually some unexpected convergence between what Marušič and Žaucer concluded about ta and what I have about KU is rather significant. And it also lends indirect support to the proposed analysis of KU or what I call a CogDem. 21 For example, the fact that a ‘ta + ADJ’ string carries some sort of contrastive focus meaning is among the outstanding issues in the literature (Lanko Marušič, personal communication) and giving a satisfactory account of this property of ta will require extensive research in and of itself. 22 Modulo the fact that multiple occurrences of a ‘ta + AP’ string need not be accompanied by pauses between them (Lanko Marušič, personal communication), unlike the case with multiple occurrences of a ‘KU + ADJ’ string. At the moment, I do not have an account of this fact. 20

6.4 More on the Internal Structure of a DemP

275

higher [Spec, DPd/r] position (for some compositional semantic or language-specific linear order related reasons), as sketched in (86). (86) Derivation of (82a): ‘tá > ta ADJ > N’ surface order23 DPd/r DPd/r CogDemP

DPd/r DemP1

Dd/r

ta zelen LocP

Dd/r tá t1

Loc Loc

UnitP/PlP

svin nik 23

Before leaving this topic, let me briefly compare the present analysis of ta with Marušič and Žaucer’s (2007) in order to narrow down possible analyses for ‘ta + ADJ’ strings: Marušič and Žaucer suggest that ta and its co-occurring ADJ form a small clause (XP) or a reduced RC structure in which, being a pronominal element, ta occupies the subject position (i.e., [Spec, XP]) and the ADJ occupies the predicate position (i.e., sister to X), as depicted in (87) for the DP in (80).24 But under their analysis, ta is supposed to refer to the same entity as the DP that embeds it, and this incurs an i-within-i violation in the sense of Chomsky (1981); that is, their analysis creates a configuration in which a pronoun is contained within its own antecedent.

23

This derivation is identical to the way we delivered the surface order for Korean data like (146b) in Chap. 5 (see diagram (147b) in that chapter) except that the lower DEM undergoes one more raising here. 24 Marušič and Žaucer (2007) explicitly acknowledge that they are adopting Campos and Stavrou’s (2004) analysis of Aromanian polydefiniteness phenomenon, under which essentially the same idea was posited for the ‘DEM + ADJ’ string that occurs inside definite Aromanian DPs.

276

6 Extensions, Refinements, and Prospects

(87) M&Z’s (2007) analysis of the DP in (80) [adapted from their (42)]: DPi FP XP tai

FP X

X

NP AP

N knjga

velika

Under our analysis, such a problem does not arise (cf. Marušič and Žaucer 201425). The reason is that we treat ta as heading a CogDemP which selects for an AP in a manner analogous to how a degree adverb may select for an AP, as depicted in (88). Consequently, there is no co-indexation relation established between the DEM and the DP within which it occurs and thus no i-within-i violation is incurred.

25 This problem does not arise under Marušič and Žaucer’s subsequent analysis, Marušič and Žaucer 2014, since, in that work, they claim that ta is a definite article which occurs at the Spec position of an extended projection of an AP which they call ADP, as shown in (i), and this functional lexeme ranges over degrees and it also encodes what they call ‘type-definiteness’.

(i) [ADP ta [AD [AQP zelo ‘very’ [AQ [DegP [Deg [AP [A dobra ‘good’]]]]]]]] [M&Z 2014: (29)]

I find Marušič and Žaucer’s (2014) analysis of ta outlined above to be promising, but I cannot fully evaluate it here because I have learned about it only recently and fully evaluating it will require extensive research that will go well beyond the scope of this book. That said, one problem I can think of for their analysis is that ta never occurs in a predicative position (Lanko Marušič, personal communication) although if it is indeed embedded inside an AP, then it should be able to, because any ADJ can be used predicatively as long as they are not inherently delimited for attributive uses only (e.g., mere, sheer, only as in I’m an only child in English). Interestingly, Korean KU cannot occur predicatively either, as shown in (ii). Given this, I conjecture that treating ta as some sort of ADJ dependent, as Marušič and Žaucer (2014) do, may merit rethinking. But I will have to leave this issue to future research. And I also need to acknowledge that, upon close examination, no two lexemes in any language or in any two languages would turn out to behave in an identical manner, so what works for Korean KU may not necessarily work for Slovenian ta or vice versa. (ii) Mokkeri-ka (*ku) pissa-ni? Necklace-NOM (that) expensive-Q.INFRML Intended: ‘Is a/the necklaced that expensive?’

6.4 More on the Internal Structure of a DemP

277

(88) a. Possible internal structure for the degree adverb very plus ADJ good sequences: [DegP [Deg [Deg very] [AP good]]] b. Possible internal structure of the DP in (80) under the present analysis: DPd/r CogDemP1

Dd/r

CogDem CogDem ta

Dd/r AP

LocP t1

Loc Loc

PlP+

velika knjga

In light of the foregoing, then, treating ta as instantiating some sort of CogDem seems like a plausible analysis to pursue. What I have shown above also suggests that all types of DemP in human language may have the internal syntax schematized in (68), where the AP it selects for is either lexical or functional, or overt or covert, and depending on the language, it may also contain an AGR FP, as Leu (2008, 2015) suggests. That said, if the present analyses of KU and ta are correct, then we should also expect that what we call CogDemPs exhibit a slightly different syntax than more “ordinary” DemPs do. Moreover, even among CogDemPs, some variation is to be expected because some may only occur inside a definite DP, as we have seen with Korean KU, but others may not have such a restriction, as we have seen with Slovenian ta, although what cuts across them is that they all perform some sort of dual function in the sense that, by selecting for an AP, they modify an adjectival meaning but by occurring at a Spec position of a full-fledged DP, they also modify a nominal meaning.

6.5

Supplementary Versus Non-restrictive RC Distinction in Prenominal Modifier Languages

In Chap. 4, I suggested that supplementary modifiers merge DP-externally whereas non-restrictive modifiers merge DP-internally. More concretely, the idea I put forth there was that while non-restrictive modifiers may merge at [Spec, DPd/r], as

278

6 Extensions, Refinements, and Prospects

sketched in (89), supplementary modifiers are introduced by what I call a SpplP, whose head hosts a pause or a comma and which selects for a TP or a DP, depending on whether the language at hand permits only postnominal N modifiers or prenominal ones, as given in (90). (In (89) and (90), XP stands for a non-restrictive or a supplementary RC.) (89) NRS RCs: [DPd/r [XP] … [DPp … N]] (90) Supplementary RCs: [SpplP [XP] [Sppl [Sppl , ]

[TP/DP … [… [FP YP] … N] …]]] Integrated RC

After submitting these ideas, in both Chaps. 4 and 5, I demonstrated how assuming such a syntax for the two types of N modifiers may let us account for the relevant data drawn from English and Korean. One thing I did not emphasize in that context, though, was that prenominal-modifier languages like Korean, Japanese, and Mandarin do not make a formal distinction between supplementary RCs and non-restrictive ones, especially when one considers cases where multiple RCs co-occur before the same DP. To see this, consider the Korean sentence in (91). This sentence contains three RCs and the first two are followed by a pause, and it can be given two construals: on one reading (Reading 1), two of the RCs may be construed as supplementary modifiers and on the other reading (Reading 2), all three RCs may be analyzed as non-restrictive modifiers, as indicated by the English translations. Since we differentiate supplementary RCs from non-restrictive ones by looking at whether their contents are understood as supplementary to what is being expressed by the rest of the sentence or not, in languages like Korean, the only way to tell which is which is to look at the discourse context and the speaker’s intention to use them. According to Korean speakers’ intuitions, (91) can be uttered in a context where the RCs’ contents are perceived as ‘by the way’ kind of remarks made by the speaker or in a context where their contents are perceived as directly relevant to the purpose of the discourse at hand. On either scenario, however, this sentence can be felicitous only when the speaker is talking about some particular necklace that has been singled out from a set of necklaces. Hence, the semantic contribution of the RCs here is definitely not to reduce the set of individuals denoted by the head N.

6.5 Supplementary Versus Non-restrictive RC Distinction

279

(91) Na-nun [RC1 cinan-cwu-ey Minswu-ka puracil-eyse ei I-TOP [ last-week-LOC M.-NOM Brazil-from __ sa-o-ass-ta-ko-ha-]-nun*(,) [RC2 ecey Mina-ka yesterday M.-NOM buy-come-PST-IND-COMP-QUOT-]-REL, [ nay-key ei kacieta-cwu- ]-n*(,) [RC3 nay-ka phyengso-ey [ I-NOM ordinary.time-LOC I-to __ bring-give-PRF]-REL, ei nemwunato kac-ko siph-ess-te-]-n mokkerii-rul want-ANT-RTRO-]-REL necklace-ACC __ too.much possess-CONN onul il-e peri-ess-ta. AUX-PST-DECL today lose-CONN Reading 1: ‘Today, I lost the necklace which I hear that Minswu bought and brought from Brazil last week, which Mina brought to me yesterday, which I had been really wanting to have for such a long time.’ Reading 2: ‘Today, I lost the necklace which I hear that Minswu bought and brought from Brazil last week which Mina brought to me yesterday which I had been really wanting to have for such a long time.’

Under the present analysis, the lack of a formal distinction between supplementary RCs and non-restrictive ones in Korean-type languages can be explained as follows: if we assume that the head of a SpplP may select for a DP and supplementary RCs merge at a [Spec, SpplP], as schematically represented in (90), then the object position of (91) can be rendered the structure sketched in (92), where the first two RCs have the status of supplementary N modifiers and the last one has the status of a non-restrictive N modifier. (92) Structure of the object DP in (91): Hypothesis I SpplP SpplP

RCSppl ei

RCSppl ei

DPd/r

RCNRS ei

Dd/r mokkerii

But nothing stops us from treating all three RCs in (91) as non-restrictive N modifiers since in our framework, non-restrictive RCs merge either at a pre-existing [Spec, DPd/r] or at an adjoined [Spec, DPd/r] position and they may occur stacked on top of each other, accompanied by pauses, as sketched in (93).

280

6 Extensions, Refinements, and Prospects

(93) Structure of the object DP in (91): Hypothesis II DPd/r RCNRS ei

DPd/r RCNRS ei

DPd/r

RCNRS ei

Dd/r mokkerii

Whichever is the correct analysis for (91), the surface structure obtained will have the first two RCs occur followed by a comma because, on our analysis, merging at an adjoined [Spec, DPd/r] or at a [Spec, SpplP] position is bound to yield such a phonological contour. That said, since assuming (93) will yield equally desirable results as assuming (92) does in deriving data like (91), we can take one step further and hypothesize that Korean-type languages lack supplementary RCs and “true” supplementary RCs are attested only in English-type languages in which all clausal N modifiers occur postnominally. If we make such a hypothesis, then as a corollary, we also need to posit that the head of a SpplP selects for a TP but not a DP, as sketched in (94), revising (90). (94) Syntax of supplementary RCs (revised): [SpplP [XP] [Sppl [Sppl , ] [TP …]]] Supplementary Mod

Assuming (94) lets us account for why English supplementary RCs may have non-DP antecedents but putative supplementary RCs in Korean may not. To see this, consider (95) and (96) (which are repeated from Footnote 17, Chap. 4). (95) shows that English supplementary relative pronouns can be co-indexed with a referential DP, a predicative nominal, an AP, or a CP. On the other hand, (96) shows that, in Korean, what can be expressed by supplementary RCs in English may only be expressed by coordinate structures unless the intended head N for the RC is a referential DP, as is the case with (96a). (95) a. b. c. d.

[John]i, [who]i is a doctor, is Bill’s friend. John is [a doctor]i, whichi Bill is not. John is [handsome]i, whichi Bill is not. John said [that he was from Dallas]i, whichi his own brother denied later.

6.5 Supplementary Versus Non-restrictive RC Distinction

281

(96) RC construction: a. Uysa-i-n Con-un Pil-uy chinkwu-i-ta. Bill-GEN friend-COP-DECL Doctor-COP-REL John-TOP ‘John, who is a doctor, is Bill’s friend.’ Coordinate structures: b. Con-un uysa-i-ntey Pil-un uysa-ka ani-ta. John-TOP doctor-COP-but Bill-TOP doctor-TOP not.be-DECL ‘John is a doctor but Bill is not.’ calsayngkiess-nuntey Pil-un kurehci anh-ta. c. Con-un Bill-TOP so not.be-DECL John-TOP handsome-but ‘John is handsome but Bill is not.’ Tallas-eyse o-ass-ta-ko d. Coni-un cakii-ka Dallas-from come-PRF-IND-COMP John-TOP self-NOM malhay-ss-nuntey kui-uy chin-hyeng-i kukes-ul he-GEN real-brother-NOM that-ACC say-PST-but pwuinhay-ss-ta. deny-PST-DECL ‘Johni said that hei is from Dallas, but hisi own brother denied it.’

In view of (94), the versatility of English supplementary RCs exemplified by (95a–d) is expected because any constituent inside a TP may raise to an adjoined [Spec, SpplP] which immediately c-commands the TP (thereby c-commanding the relative pronoun as well as its own trace). In the case of sentences like (95b–d), the raising of the intended head N of the supplementary RC is followed by a remnant movement of the TP to [Spec, SpplP+], as sketched in (97b–d), and this yields a surface structure where a supplementary RC occurs following a full-fledged sentence. (97) The syntax of English supplementary RCs a. Derivation of (95a): [SpplP ___ , [SpplP [CP who1 …] [Sppl [Sppl , [TP [DP John]1…]

b. Derivation of (95b): [SpplP+ ___ [SpplP [DPp a doctor]1, [SpplP [CP which1 …] [Sppl [Sppl [TP … t1]

c. Derivation of (95c): [SpplP+ ___ [SpplP [AP handsome]1, [SpplP [CP which1 …] [Sppl [Sppl [TP … t1]

d. Derivation of (95d): [SpplP+ ___ [SpplP [CP that he …]1, [SpplP [CP which1 …] [Sppl [Sppl [TP … t1]

In Korean, RCs do not exhibit similar versatility to English supplementary RCs (and consequently, the intended meanings of (96b–d) have to be expressed in a different construction) because the head of a SpplP only selects for a TP, so there is no way a language only permits prenominal N modifiers but it also possesses supplementary RCs at the same time: If we assume that Korean also has SpplPs, so

282

6 Extensions, Refinements, and Prospects

data like (96a) are derived from a Sppl selecting for a TP, then we cannot account for the fact that there cannot be a pause between the RC and the subject DP Con, as shown in (96a′),26 even though the head of a SpplP would host a comma. (96) a. *Uysa-i-n, Con-un Pil-uy chinkwu-i-ta. Doctor-COP-REL John-TOP Bill-GEN friend-COP-DECL Intended: ‘John, who is a doctor, is Bill’s friend.’

Note that we cannot assume that (96a) results from a Sppl selecting for a TP followed by the raising of the subject DP to [Spec, SpplP+] either, because doing so would wrongly predict that Korean permits postnominal RCs, apart from the fact that such a derivation would yield pauses both before and after the RCs. Given the structural differences between (96a) and (96b–d), one may think that Korean-type languages actually do have supplementary RCs but they can only have a nominal antecedent because their Sppl selects for a DP, as sketched in (90). However, if that is indeed the case, then a datum like (96a′) should be judged grammatical, contrary to fact; as comparison between (96a) and (96a′) shows, this sentence would be judged grammatical only if the comma is absent before the RC, yet assuming that Sppl selects for a DP in Korean would wrongly predict its obligatory presence in that position. In addition, raising the subject DP to [Spec, SpplP+] would run into the same problem as stated above by yielding postnominal RCs for a language that completely lacks them. If Sppl only selects for a TP in all languages, then how do we account for the grammaticality of data like (96a)? My answer is that such RCs are best analyzed as superficially non-restrictive but underlyingly restrictive N modifiers which are base-generated at a [Spec, DPp] but which later move to [Spec, DPd/r] to fulfill PELI. That is, they are no different than the proper-name-modifying RCs in (15) and (17), as one can see from comparing (98) with (24), which is repeated below for convenience. As for the question of why such “hybrid” RCs might be attested in Korean/Mandarin-type languages, that is most likely because their proper names denote singleton sets rather than individuals.

26

As mentioned in Sect. 6.2, corresponding Mandarin data exhibit the same pattern.

6.5 Supplementary Versus Non-restrictive RC Distinction

283

(98) Derivation of (96a) under the present analysis: [DPd/r __ [Dd/r [Dd/r [DPp2 [RC uysain…] [DPp1 …[nP Con]]]]]]

(24) Surface structure of (15) under the present analysis: DPd/r, e Dd/r , e Dd/r, RC,

DPp2, DPp1,

juyou heren xietong de Obama

In sum, then, the more refined analysis we have offered in this section not only correctly predicts the absence of certain types of RCs in Korean/Mandarin-type languages but also explains why their RCs behave the way do, which is clearly different from the behavior of English-type RCs. To my knowledge, this is an outcome no existing analysis has been able to yield, so is considered yet another welcome result of the present analysis.

6.6

What Determines the Interpretive Possibilities of APs?

As mentioned in Chap. 1, there is a rich literature reporting the correlation between the surface position of an ADJ and its possible interpretations. In particular, ever since Bolinger’s (1967) seminal work, numerous linguists have shown that there is a systematic semantic difference between prenominal ADJs and their postnominal counterparts in Indo-European languages. And linguists like Cinque (2010) have even claimed that in Romance and Germanic languages, the correlation between the pre- versus postnominal position and the possible construal of an ADJ is absolute: in Germanic languages, prenominal positions are always ambiguous between ‘direct’ and ‘indirect’ N modificational meanings whereas postnominal positions are not. On the other hand, in Romance languages, postnominal ADJs may be ambiguous but prenominal ones are not. In Chap. 4, I showed, by citing Laenzlinger (2005), that there is indeed reason to think that such a correlation holds for certain Romance languages (e.g., French, Spanish, and Italian). I also demonstrated how the split-DP analysis I proposed can deliver correct surface orders for data drawn from such languages which contain a DEM and a focused element. But I did not address to what extent such a correlation may actually hold even among Romance and Germanic languages. Additionally, I did not address how languages that do not allow for postnominal N modifiers such as Korean may encode such syntax-semantics mapping.

284

6 Extensions, Refinements, and Prospects

While I cannot fully address these questions within the confines of this book, I would like to point out that even among Romance languages, the surface position of an ADJ relative to the nominal it modifies and its interpretive possibilities do not always correlate with each other. For example, as Pahom (2013) shows, Romanian ADJs may receive both direct and indirect modificational interpretations in the sense of Cinque (2010) regardless of whether they occur prenominally or postnominally. To illustrate this by comparing Romanian with Italian: while Italian ADJs may receive either I-level or S-level interpretations when occurring postnominally but only receive I-level interpretations when occurring prenominally, as shown in (99), their Romanian counterparts are equally ambiguous in either position, as shown in (100). (99) Italian (Cinque 2010: 7): a. Postnominal ADJ: ambiguous Le stelle invisibili de Andromeda sono moltissime. The stars invisible of Andromeda are very.many ‘Andromeda’s stars, which are generally invisible, are very many’ (I-level) ‘Andromeda’s generally visible stars, which happen to be invisible now, are very many.’ (S-level) b. Prenominal ADJ: unambiguous Le invisibili stelle de Andromeda esercitano un grande fascino The invisible stars of Andromeda have a great fascination ‘Andromeda’s stars, which are generally visible, have a great fascination.’ (I-level) #‘Andromeda’s generally visible stars, which happen to be invisible now, have a great fascination.’ (S-level) (100) Romanian (Pahom 2013: slide 9): a. Postnominal ADJ: ambiguous Stelele invizibile ale Andromedei sânt fascinante. fascinating Stars-the invisible of Andromeda-GEN are ‘Andromeda’s generally invisible stars are fascinating.’ (I-level) ‘Andromeda’s stars, which are generally visible, but which are invisible now, are fascinating.’ (S-level) b. Prenominal ADJ: ambiguous Invizibilele stele ale Andromedei sânt fascinante Invisible-the stars of Andromeda-GEN are fascinating ‘Andromeda’s generally invisible stars are fascinating’ (I-level) ‘Andromeda’s stars, which are generally visible, but which happen to be invisible now, are fascinating’ (S-level)

Similarly, while postnominal ADJs in Italian are ambiguous between what Cinque calls ‘restrictive’ and ‘non-restrictive’ interpretations, yet their prenominal counterparts are invariably construed as non-restrictive, Romanian ADJs may

6.6 What Determines the Interpretive Possibilities of APs?

285

receive either interpretation regardless of their position relative to the head N. To see this, compare (101)27 and (102).28 (101) Italian (Cinque 2010: 8): a. Postnominal: ambiguous Le lezioni noiose di Ferri se le ricordano tutti. The classes boring of Ferri SE LE remember all ‘Everybody remembers Ferri’s classes, all of which were boring.’ ‘Everybody remembers just those classes by Ferri that were boring.’ b. Prenominal: unambiguous lezioni di Ferri se le ricordano tutti. Le noiose The boring classes of Ferri SE LE remember all ‘Everybody remembers Ferri’s classes, all of which were boring.’ #‘Everybody remembers just those classes by Ferri that were boring.’

(NRS) (RS)

(NRS) (RS)

(102) Romanian (Pahom 2013: slide 11): a. Postnominal: ambiguous Lec iile plictisitoare ale doamnei Popescu i le amintesc to i Lessons-the boring of Miss-GEN Popescu SE LE remember all ‘Everyone remembers Ms. Popescu’s lessons, all of which were boring.’ (NRS) ‘Everyone remembers just those classes by Ms. Popescu that were boring.’ (RS) b. Prenominal: ambiguous Plictisitoarele lec ii ale doamnei Popescu i le amintesc to i Boring-the lesson of Miss-GEN Popescu SE LE remember all ‘Everyone remembers Ms. Popescu’s lessons, all of which were boring.’ (NRS) ‘Everyone remembers just those classes by Ms. Popescu that were boring.’ (RS)

The contrast between Italian and Romanian ADJs illustrated above raises the question of why Romanian may behave differently from other Romance languages. Obviously, we cannot do justice to this question because answering it properly will

27

In presenting these data, I have retained Cinque’s (2010) terminology but in view of the way we use the technical terms in this book, what he calls ‘non-restrictive’ corresponds to what we call ‘attributive’ or ‘reference-modifying’, so the Italian and Romanian data in (101)–(102) should be understood with these terminological differences in mind. 28 Some authors have reported that in some cases, there are semantic differences between prenominal and postnominal ADJs in Romanian. For example, Giusti (2005: 27, footnote 5) claims that when occurring prenominally, the ADJ sarăc ‘pitiful/poor’ receives only the evaluative interpretation and when occurring postnominally, it only receives a descriptive interpretation, as shown in (i). But I take what Pahom (2013) claims at face value and analyze her data accordingly, accepting her grammaticality judgements and intuitions as they are. (i) a.

sar cul b iat ‘the pitiful boy’ b. b iatul sar c ‘the poor boy’

(A > N) (N > A)

286

6 Extensions, Refinements, and Prospects

require a thorough historical study pertaining to the language and the countries in which it is spoken, taking into account the language contact factors that must have played a non-trivial role in shaping Modern Romanian. But even without conducting such research, one thing that is evident is that while Italian has a free-standing definite article, Romanian has a suffixal one (Cornilescu and Nicolae 2011).29 Given this, it seems plausible that what holds the key to the differences between the two languages is the free versus bound morpheme status of their definite articles. To be more specific, Romanian exhibits certain movements that Italian does not because its suffixal definite article has to be part of a free-standing morpheme regardless of whether it is an N or an ADJ (Cornilescu and Nicolae 2011) and its host has to undergo movement to the Spec position of what I call DPd/r in compliance with PELI or any other relevant constraint (compare Giusti 200530). Viewed this way, then, prenominal ADJs in Romanian are derived via different routes than their counterparts in other Romance languages: while prenominal ADJs in Italian are presumably pronounced in their base-positions without undergoing movement on their own (Cinque 2010), prenominal ADJs in Romanian surface in a higher D-space most likely to fulfill PELI and thereby value the relevant D features including [+definite] in the local domain of what I call DPd/r. And since the same results can be achieved by having the ‘N + definite article’ cluster move to Dd/r instead and doing so would not engender truth-conditional semantic differences (Cornilescu and Nicolae 2011),31 all else being equal, the raising of an ADJ would have all the interpretive possibilities that the raising of an N would or vice versa. Consequently, at least in the cases at hand, prenominally occurring ADJs in Romanian may receive discourse-linked construals (e.g., S-level interpretations) as well as more lexical or non-D-linked construals (e.g., I-level interpretations). If correct, this way of looking at things provides additional support for Svenonius’ (2008) claim that cross-linguistic variation on the ordering of DP-internal constituents is due largely to the morphological properties of the languages at hand (e.g., whether a functional element in a language is a bound or a free morpheme, or ADJs in that language bear agreement morphology or not). It also leads us to see that in languages where APs may occur both prenominally and postnominally (e.g. Italian, Romanian), their interpretive possibilities are determined not only by whether they occur before or after their head N but also where in

29 According to Cornilescu and Nicolae (2011), the suffixal definite article in Romanian arose from the DEM in Latin ille getting reanalyzed as a head (eventually occupying the D0 position). 30 Giusti (2005) takes a similar stance on the matters at hand here but the details of her idea differ from mine. She claims that prenominal ADJs in Romanian merge directly at [Spec, DP] and their merge is triggered by the [+Topic] feature they carry, which blocks N-to-D movement; she considers [Spec, DP] (i.e., the highest position inside a full-fledged DP) a topic position. While I find this idea both attractive and plausible, I leave a more complete comparison between her analysis and the present analysis with respect to Romanian data to future research. 31 Giusti (2005) claims that the prenominal position in Romanian is thematic and anaphoric whereas the postnominal position is rhematic and deictic. But importantly, such differences are information-structural rather than truth-conditional.

6.6 What Determines the Interpretive Possibilities of APs?

287

the DP-space they occur—that is, whether they are pronounced in what I call the Low Field of a DP or a higher field thereof. Notably, the way in which putative Korean ADJs are interpreted leads us to the same conclusion, despite the fact that Korean only allows for prenominal N modifiers and the details of the relevant data are rather different from what we see with the Romanian cases. To demonstrate this by first looking at so-called subsective ADJs in Korean: as I showed in Chaps. 3, 4, and 5, UN-APs in Korean are able to receive ‘direct’ as well as ‘indirect’ modificational interpretations but it is not the case that they may receive either interpretation under any or all circumstances. By way of illustration, when uttered out of the blue, the UN-AP ttwienan ‘remarkable’ always receives a subsective reading as shown in (103), and for a non-subsective reading to obtain, the discourse context has to be enriched quite a bit, as shown in (104): ttwienan in this discourse may not receive the intended non-subsective reading if the (b) sentence is not uttered immediately after the (a) sentence. And this shows that the discourse context plays an extremely important role in interpreting UN-APs. (103) Context: Out of the blue ttwiena-n uysa remarkable-ADN doctor ‘a doctor who is remarkable as a doctor’ *‘a doctor who is remarkable as a chess player’

(subsective) (non-subsective)

(104) Context: Speaker utters (b) immediately after uttering (a). a. I-pen cangki sihap-eyse etten uysa-ka chamuro This-time chess tournament-LOC some doctor-NOM really ttwiena-te-la. remarkable-RTRO-DECL.INFML Lit.: ‘At the most recent chess tournament, some doctor was really remarkable (at chess).’ b. Kurentey, [DP ku ?(cangki-ey) ttwiena-n uysa]-nun And [ that (chess-GOAL) remarkable-ADN doctor]-TOP Mina-uy namca-chinkwu-i-rako ha-te-la. M.-GEN male-friend-COP-QUOT do-RTRO-DECL.INFML ‘And I was told that that remarkable doctor (i.e., the one remarkable at chess) is Mina’s boyfriend.’ (non-subsective)

Interestingly, data containing superlative intersective UN-APs seem to present the opposite picture: superlative UN-APs exhibit more inherent ambiguity than subsective ones in the sense that when uttered out of the blue, they may receive either an ‘absolute’ or a ‘relative/comparative’ construal, as shown in (105), but their ambiguity disappears if they occur inside an object DP and the subject of the sentence they occur in has a (contrastive) topic-marking as opposed to nominative

288

6 Extensions, Refinements, and Prospects

case marking. To see this, compare (106) and (107).32 These examples show that if the sentential subject is topic-marked, only an absolute reading is possible for a superlative ADJ in Korean but if the sentential subject bears a nominative case, which may also be treated as a focus marker since it induces an ‘all new information’ or an ‘exhaustive listing’ reading (e.g., Kim 1990; Choi 1996),33 both absolute and relative readings are available for the ADJ.34 (105) Context: out of the blue. [kacang noph-un] san [most high-ADN] mountain ‘the highest mountain on Earth/in some region’ ‘the highest mountain under discussion’

(absolute) (relative)

35

(106) Sentence with a nominative case or FOC marked subject:35 Seyra-ka [kacang noph-un] san-ul olla-ss-ta. [most high-ADN] mountain-ACC climb-PST-DECL S.-NOM Reading 1: ‘Seyra climbed a mountain that is higher than any other mountain that is contextually relevant.’ (absolute) Reading 2: ‘It is Seyra who climbed a mountain that is higher than any other mountain that is contextually relevant.’ (absolute) Reading 3: ‘Seyra climbed a higher mountain than anyone else did (e.g., she climbed the highest mountain among her friends).’ (relative) Reading 4: ‘It is Seyra who climbed a higher mountain than anyone else did (e.g., she climbed the highest mountain among her friends).’ (relative)

32

As cited in Shimoyama 2014 (footnote 9), the contrast between data like (106) and (107) has also been noted by Lyu (2004) though, unlike here, not in comparison to (105) or to the interpretive behavior of other types of ADJ in Korean. 33 The effects that nominative case marking on a DP brings about in Korean are similar to what has been reported about Japanese. See, a.o., Kuno 1973, Kuroda 1965, 1972. 34 According to Shimoyama (2014, footnote 9), Japanese exhibits the same phenomenon. In the literature, it is also well-documented that focus plays a role in interpreting English sentences with adnominal superlative ADJs although the phenomena discussed are slightly different than what we see in Korean and Japanese. See, a.o., Ross 1964, Szabolcsi 1986, Heim 1999, and Sharvit and Stateva 2002. For a recent discussion of the relevant phenomena in English and Polish, see Tomaszewicz 2015. 35 The English translations for this sentence are modeled after Szabolcsi 1986, in conjunction with the fact that Korean NOM gives rise to some sort of FOC reading for major subjects.

6.6 What Determines the Interpretive Possibilities of APs?

289

(107) Sentence with a topic-marked subject: Seyra-nun [kacang noph-un] san-ul olla-ss-ta. S.-TOP [most high-ADN] mountain-ACC climb-PST-DECL Reading 1: ‘As for Seyra, she climbed a mountain that is higher than any other mountain that is contextually relevant.’ (absolute) *Reading 2: ‘As for Seyra, she climbed a higher mountain than anyone else did (e.g., she climbed the highest mountain among her friends).’ (relative)

I take the above set of facts to suggest that when it comes to subsective ADJs, their default interpretation is a direct N modificational reading, and this is because, all else being equal, what supplies the value for the property argument P inside their denotation should be their syntactic sister, as sketched in (108) for (103). If, on the other hand, the discourse context is such that their hosting DP is D-linked and a certain property is made salient as in (104), then that property may supply the value for the property argument inside their lexical entries. (108) Computation of the meaning of (103) under the present analysis: a. DPp, AP,

NP,

ttwiena-n uysa b. [[AP]] = [[ttwienan]] = via lexical entry = P. x.P(x) = 1 & x is remarkable as a member of {z: P(z) = 1} [[NP]] = [[uysa]] = via lexical entry = y.y is a doctor [[DPp]] = [[ttwienan uysa]] = [[AP]]([[NP]]) = via FA = [ P. x.P(x) = 1 & x is remarkable as a member of {z: P(z) = 1}]( y.y is a doctor) = via -reduction = x.[ y.y is a doctor](x) = 1 & x is remarkable as a member of {z: [ y.y is a doctor](z) = 1} = via -reduction = x.x is a doctor and x is remarkable as a member of {z: z is a doctor}

As for superlative ADJs, their possible interpretations may vary depending on what constitutes their comparison class C in the sense of Heim (1999), and since C is also contextually determined, the situation looks a bit more complex in cases like (105)–(107) than in (103)–(104): when the superlative ADJ kacang nophun ‘the highest’ is uttered in neutral contexts like (105) or focus contexts like (106), the comparison class which feeds the meaning of the superlative morpheme inside the AP36 can be comprised of contextually relevant mountains that are tall or the 36

For the semantics of the superlative morpheme, following standard practice, I adopt Heim’s (1999) treatment of English -est given in (i), where e ranges over individuals and d ranges over degrees; recent work on the semantics of superlative ADJs offer a revised version of it, which is more cross-categorial (e.g., Kotek et al. 2011; Tomaszewicz 2015) but for present purposes, adopting Heim’s original lexical entry will suffice. -est = C e,t . D d,et . xe. d[D(d)(x) C, y [y C Presupposition: x

y C [y x d [D(d)(y)]].

¬(D(d)(y))]]

290

6 Extensions, Refinements, and Prospects

mountains climbed by contextually relevant individuals including Seyra which are not necessarily tall but which can be ranked with respect to their heights; the former case will yield an absolute reading and the latter case will yield a relative/ comparative reading. When the sentential subject functions as a discourse topic and the ADJ itself still occurs inside the object DP as in (107), C may be comprised only of contextually relevant mountains that are tall. And this is because at the sentential-level, the subject has already been picked out from a group of individuals the predicate’s meaning may potentially hold true of, so there is no reason to compare it against other individuals with respect to the sentential predicate’s meaning—that is, to express how high a mountain the individual denoted by the subject climbed, compared to other relevant individuals. In sum, then, for information structural reasons, when a sentence instantiates ‘topic-comment’ or ‘theme-rheme’ structure as in (107), the superlative ADJ occurring inside the sentential predicate is interpreted dependently on its head N’s meaning. Consequently, it receives a direct modificational construal; when a sentence instantiates ‘rheme-theme’ structure like (106) does, the superlative ADJ occurring in it may be interpreted more or less head-N-independently (and also DP-externally) and as a result, the ADJ receives an indirect modificational construal.37 If valid, what I have presented here shows (yet again) that what determines the interpretation of an adnominal ADJ is not just its surface position but also what kind of nominal it modifies and what kind of discourse context it occurs in. Therefore, a truly adequate analysis of the syntax and semantics of N modifiers has to look at things from a more global perspective than just paying attention to whether an ADJ occurs before or after an N, or how morpho-syntactically complex it is, especially when the language at hand allows for only prenominal or postnominal N modifiers.

6.7

How Many FocPs Inside a DP?

We have thus far assumed that there is just one FocP inside a DP (see Chaps. 4 and 5), and authors like Laenzlinger (2005) and Svenonius (2008) seem to subscribe to the same view although the exact location of the FocP they posit differs

37

Though not exactly in the context of making a connection between the interpretation of superlative ADJs and information structure, several authors have shown that in languages like English, Japanese, and Polish, the superlative morpheme scopes outside of the DP in which it occurs only when relative readings are derived (e.g., Szabolcsi 1986; Heim 1985, 1999; Shimoyama 2014; Tomaszewicz 2015), so what I have suggested here is likely to carry over to other languages as well.

6.7 How Many FocPs Inside a DP?

291

from what we assume. It is entirely conceivable, however, that there are actually multiple focus positions inside a referential DP, especially when one considers the way focus is encoded at the sentential-level. Therefore, in this section, I would like to briefly address the question of how many FocPs are inside a DP structure. In the literature, it is well-established that there is more than one focus position in the clausal domain. To illustrate, in English finite sentences, a contrastive focus/ topic may occur in the left periphery, as shown in (109), and in Hungarian finite sentences, items that instantiate what Kiss (1998) calls ‘identificational focus’ occur in pre-verbal position, as shown in (110). And when we look at languages like Korean, we see that the same finite sentence may host both a contrastive focus/topic and an identificational focus, as shown in (111); note that in (111), a contrastive focus/topic item occurs at the beginning of the sentence and an identificational focus item occurs pre-verbally. (109) Illustration of English contrastive focus in the clausal domain: YOUR BOOK, you should give to Paul (not mine). [Rizzi 1997: ex. (2)] (110) Illustration of Hungarian identificational focus in the clausal domain: Mari Pétert hívta fel. ‘It was Peter that Mary called up. [Kiss 1998: ex. (31b)] (111) Illustration of Korean contrastive focus/topic and identificational focus in the clausal domain: Context: I have a friend named John. Recently, both John and I bought some candy and cookies to give away to some children, and I am letting my friend know who gave what to whom. sathang-un, na-nun Mina-eykey cwu-ess-ko candy-CONT.TOP, I-TOP M.-to give-PST-CONN kwaca-nun , na-nun Chelswu-eykey cwu-ess-ta. C.-to give-PST-DECL cookie-CONT.TOP, I-TOP ‘As for candy, I gave it to Mina, and as for cookies, I gave them to Chelswu.’ Kurentey Con-un, sathang-un, Swu-eykey cwu-ko give-CONN But John-top, candy-CONT.TOP, Sue-to kwaca-nun , Pil-eykey cwu-ess-ta. give-PST-DECL cookie-CONT.TOP, Bill-to ‘But as for John, he gave the candy to Sue and the cookies to Bill.’

292

6 Extensions, Refinements, and Prospects

Given the commonly-held view that there is a parallel between nominal structure and clausal structure (e.g., Abney 1987; Cinque 1999; Grimshaw 1990; Kayne 1994; Szabolcsi 1983, 1994)—a view that has been significantly substantiated by recent cartographic research (e.g., Rizzi 1997; Giusti 2005; Cinque 2008), it seems only logical to suspect that the parallel may extend to focus phenomena as well. If such an extension is indeed possible, then two distinct focus positions may be postulated inside a DP, one in the DP-internal region and the other in the DP-external region, as schematized in (112). In this schema, FocP1 corresponds to the contrastive focus position Rizzi (1997) postulates between FinP and CP on the basis of sentences like (109), and FocP2 corresponds to the identificational focus position Kiss (1998) postulates in pre-verbal position on the basis of sentences like (110). (112) Hypothesis about two distinct FocPs inside DP structure: [DPd/r FocP1 [DPp FocP2 ]]

What is depicted in (112) seems plausible but careful examination will be necessary to adopt it. At the moment, I have three reasons to not adopt it: One reason is that I have not as yet seen empirical evidence that there are two distinct focus positions inside DP. Another reason is that, as far as I can see, DP-internally occurring focused elements (i.e., those that occur in the periphery of what I call DPp) do not necessarily carry the [+exhaustive] feature that Kiss (1998) attributes to what she calls identificational focus. Thirdly, all DP-internally occurring focused elements can potentially receive a [+contrastive] interpretation, regardless of whether they contribute all new information or something that is presupposed. To demonstrate this, let me first say a few more words about Kiss’s theory of focus. Kiss posits that there are two types of focus in the clausal domain, namely, what she calls ‘information focus’ and what she calls ‘identificational focus’. Information focus occurs in contexts where the discourse participants do not have a closed set of individuals or properties that are known to them, as illustrated by the question-answer (Q&A) pairs in (113)–(114). Identificational focus is instantiated when there is a closed set of individuals that are known to the discourse participants and the speaker picks out one unique (plural) individual from that set. Therefore, unlike information focus, identificational focus carries both [+exhaustive] and [+contrastive] features (Kiss 1998: 268), and this is evidenced by the fact that in contexts like (113)–(114), utterances made by the hearer may be continued by the additive sentences that are found inside the parentheses, but in contexts like (115)–(116), they cannot be, as indicated by #.

6.7 How Many FocPs Inside a DP?

293

Q&A pairs where the answers carry information focus: (113) A: What did you do yesterday? B: I went to the movies. ( I also took a long nap.) (114) A: What does John like? B: Horror movies. ( He also likes hiking.) Q&A pairs where the answers carry identificational focus: (115) A: Bill danced with Mary. B: No, it was Sam who danced with Mary. (#And Peter also danced with her.) B´: No, only Sam danced with Mary. (#And Peter also danced with her.) (116) A: John sent the letter to Laura. B: No, it was Suzi that he sent the letter to. (#And he also sent it to Mary.) B´: No, he sent the letter only to Suzi. (#And he also sent it to Mary.)

When we look at the nominal domain, at first glance, it seems that adnominal ADJs can bear both types of focus. To illustrate, the focused ADJs in (117B) and (118B) occur as part of an answer to a ‘what’ question in a manner similar to what is given (113)–(114), and the focused ADJ in (119B) and (120B) represent the properties that are picked out from a set of properties that are already known to the discourse participants in a manner analogous to the sentences in (115B)–(116B). (117) Context: At a furniture store, between a sales associate and a customer. A: What are you looking for? B: I’m looking for a ROUND coffee table. (118) Context: A conversation between two co-workers who have John as their boss. A: What kind of task did John give you? B: A very TOUGH one. (119) Context: At a furniture store, between a sales associate and a customer. A: Which between these two coffee tables do you want? B: The ROUND one. (120) Context: A conversation between two co-workers. A: Between an easy task and a tough one, which do you want? B: A TOUGH one.

However, there is reason to think that stressed adnominal ADJs actually do not carry [+exhaustive] semantics. To see this, consider first (121) and (122). These discourses show that adnominal ADJs bearing a focal stress can be followed by an utterance which attributes additional properties to the individual at issue.

294

6 Extensions, Refinements, and Prospects

(121) A: What kind of coffee table do you want? B: A ROUND one. I also want it to be wooden and white as well. (122) A: What kind of mentor was John? B: He was a very CAring mentor. He was also very supportive as well.

Consider now (123)–(124). These discourses show that even ADJs which appear to be contrastive-focused may be followed by an utterance that attributes additional properties to the individual at issue. (123) A: You want a ROUND table, right? B: No, I want a SQUARE one (, not a ROUND one). And I also want it to be white. (124) A: Was John a very indifferent mentor? B: No, he was a very CAring one. But he was also very harsh and critical at times as well.

Finally, discourse (125) shows that even DEMs that bear [+contrastive] focus may not necessarily be [+exhaustive], unlike the identificational focus marked DPs in (115)–(116). (125) A: Do you want THIS table or THAT table? B: I want THAT one. And I also want the one over there as well.

Taken together, this set of facts shows that DP-internal focus phenomena have slightly different properties than clausal-level focus phenomena. Moreover, postulating just one focus position inside DP structure, more specifically in what I call the Low Field region, has been shown to capture the relevant facts, as we have demonstrated in Chaps. 4 and 5, using data drawn from English, Spanish, French, Italian, and Korean. Given this, we can conclude that there is no need to posit more than one FocP per DP even if one adopts a relatively elaborate decompositional split-DP structure as we do here. Having said that, a fuller investigation of this topic and validation of what I have claimed here has to be left for future research.

6.8

Summary

In this chapter, I have demonstrated how the new analysis of N modifiers presented in Chaps. 4 and 5 has applicability beyond the empirical domain of a few languages or a few linguistic phenomena. Most notably, in this chapter, I have offered a new formal semantic analysis of Mandarin and Korean integrated RCs which occur modifying proper names or pronouns, and in so doing, I answered the long-standing question of why East Asian languages have “hybrid” N modifiers which seem to

6.8 Summary

295

exhibit properties of both restrictive and non-restrictive RCs. Next, I proposed a way to deal with cross-linguistic variation on the syntax and semantics of DEMs by invoking a limited set of binary semantic features while also maintaining a relatively uniform internal syntax for DemPs regardless of whether they perform a purely anaphoric function or what I call a cognitive indexical function, or whether they occur in an article-less language like Korean or in an article-possessing language like English. This was followed by pointing out the differences between English-type languages and Korean-type languages with regard to the possible (sub)types of RCs they can have, and in this context, I showed how such a typological difference may fall out from the present analysis. Finally, I have taken up two remaining questions, namely, how the interpretative possibilities of adnominal APs are determined in any given language, and how many focus positions are to be posited inside a DP structure. And our discussion has shown that while we cannot offer fully satisfactory answers to these questions as yet, the proposed decompositional split-DP analysis gives us a way to begin answering them, and whatever their ultimate answers may be, it appears that they will be compatible with the assumptions and claims made in the present analysis.

References Abney, Steven. 1987. The English noun phrase in its sentential aspect. Cambridge, MA: MIT dissertation. Anderson, Stephen R., and Edward L. Keenan. 1985. Deixis. In Language typology and syntactic description, vol. 3: Grammatical categories and the lexicon, ed. Timothy Shoppen, 259–308. Cambridge, MA: Cambridge University Press. Bernstein, Judy B. 1997. Demonstratives and reinforcers in Romance and Germanic languages. Lingua 102: 87–113. Bolinger, Dwight. 1967. Adjectives in English: Attribution and predication. Lingua 18: 1–34. Bowdle, Brian F., and Gregory Ward. 1995. Generic demonstratives. In Proceedings of the 21st annual meeting of the Berkeley linguistics society, ed. Ahlers, Jocelyn, Leela Bilmes, Joshua S. Guenter, Barbara A. Kaiser, and Ju Namkung, 32–43. Berkeley, CA: Berkeley Linguistics Society. Campos, Héctor, and Melita Stavrou. 2004. Polydefinite constructions in Modern Greek and in Aromanian. In Balkan syntax and semantics, ed. Olga Mišeska Tomić, 137–144. Amsterdam: John Benjamins. Chao, Yuan Ren. 1968. A grammar of spoken Chinese. Berkeley: University of California Press. Choi, Hye-Won. 1996. Optimizing structure in context: Scrambling and information structure. Stanford, CA: Stanford University dissertation. Chomsky, Noam. 1981. Lectures on government and binding. Dordrecht: Foris. Cinque, Guglielmo. 1999. Adverbs and functional heads: A cross-linguistic perspective. Oxford, UK: Oxford University Press. Cinque, Guglielmo (ed.). 2006. Restructuring and functional heads: The cartography of syntactic structures, vol. 4. New York: Oxford University Press. Cinque, Guglielmo. 2008. Two types of nonrestrictive relatives. In Empirical issues in syntax and semantics, vol. 7, ed. Oliver Bonami, and Patricia Cabredo Hofherr, 99–137. Paris: CNRS.

296

6 Extensions, Refinements, and Prospects

Cinque, Guglielmo. 2010. The syntax of adjectives. A comparative study. Cambridge, MA: MIT Press. Constant, Noah. 2011. Appositives after all. Ms., University of Massachusetts-Amherst. Cornilescu, Alexandra, and Alexandru Nicolae. 2011. On the syntax of Romanian definite phrases. In The noun phrase in Romance and Germanic. Structure, variation and change, ed. Petra Sleeman, and Harry Perridon, 193–222. Amsterdam: John Benjamins. Del Gobbo, Francesca. 2003. Appositives at the interface. Irvine, CA: University of California-Irvine dissertation. Del Gobbo, Francesca. 2004. On prenominal relative clauses and appositive adjectives. In WCCFL 23 proceedings, ed. Ann Kelleher, Angelo J. Rodriguez, Benjamin Schmeiser, and Vineeta Chand, 182–194. Somerville, MA: Cascadilla Press. Del Gobbo, Francesca. 2005. Chinese relative clauses: Restrictive, descriptive or appositive? In Contributions to the XXX Incontro di Grammatica Generativa, ed. Laura Brugè, Giuliana Giusti, Nicola Munaro, Walter Schweikert, and Giuseppina Turano, 287–305. Venezia: Cafoscarina. Del Gobbo, Francesca. 2010. On Chinese appositive relative clauses. Journal of East Asian Linguistics 19: 385–417. Delsing, Lars-Olof. 1993. The internal structure of noun phrases in the Scandinavian languages. Lund, Sweden: University of Lund dissertation. Dimitrova-Vulchanova, Mila, and Giuliana Giusti. 1998. Fragments of Balkan nominal structure. In Possessors, predicates and movement in the determiner phrase, ed. Artemis Alexiadou, and Chris Wilde, 333–360. Amsterdam: John Benjamins. Donaldson, Bruce. C. 1993. A grammar of Afrikaans. Berlin: Mouton de Gruyter. Elbourne, Paul. 2008. Demonstratives as individual concepts. Linguistics and Philosophy 31 (4): 409–466. Giusti, Giuliana. 2005. At the left periphery of the Romanian noun phrase. In On space and time in language, ed. Martine Coene, and Liliane Tasmowski, 23–49. Cluj-Napoca: Clusium. Grimshaw, Jane. 1990. Argument structure. Cambridge, MA: MIT Press. Gundel, Jeanette, Nancy Hedberg, and Ron Zacharski. 1993. Cognitive status and the form of referring expressions in discourse. Language 69: 274–307. Hawkins, John A. 1991. On (in)definite articles: Implicatures and (un)grammaticality prediction. Journal of Linguistics 27: 405–442. Heim, Irene. 1985. Notes on comparatives and related matters. Ms., University of Texas-Austin. Heim, Irene. 1999. Notes on superlatives. Ms., MIT. Heim, Irene, and Angelika Kratzer. 1998. Semantics in generative grammar. Malden, MA: Blackwell. Hu, Xiaoliang. 2008. A contrastive study of English and Chinese relative clauses. Hunan, China: Hunan Normal University master’s thesis. Huang, C.-T. James. 1982. Logical relations in Chinese and the theory of grammar. Cambridge, CA: MIT dissertation. Ionin, Tania. 2006. This is definitely specific: Specificity and definiteness in article systems. Natural Language Semantics 14: 175–234. Ionin, Tania, Soondo Baek, Eunah Kim, Heejeong Ko, and Kenneth Wexler. 2012. That’s not so different from the: Definite and demonstrative descriptions in second language acquisition. Second Language Research 28: 69–101. Jhang, Juchin. 2012. Multiple prenominal adjectival modifiers’ ordering restrictions in Mandarin Chinese. Ms., Texas Tech University. Jo, Mi-Jeung. 2000. Nominal functional categories in Korean: A comparative study with languages with DP. Studies in Generative Grammar 10: 427–451. Julien, Marit. 2005. Nominal phrases from a Scandinavian perspective. Amsterdam: John Benjamins.

References

297

Kang, Arum. 2012. Semantics and pragmatics of definiteness in Korean: The case of ku. In Proceedings of workshop on Altaic formal linguistics, vol. 8, 187–198. Cambridge, MA: MWPIL. Kang, Arum. 2015. (In)definiteness, disjunction and anti-specificity in Korean: A study in the semantics-pragmatics interface. Chicago, IL: University of Chicago dissertation. Kayne, Richard S. 1994. The antisymmetry of syntax. Cambridge, MA: MIT Press. Kim, Young-Joo. 1990. The syntax and semantics of Korean case. Cambridge, MA: Harvard University dissertation. Kiss, Katalin É. 1998. Identificational focus versus information focus. Language 74 (2): 245–273. Klein, Ewan, and Ivan Sag. 1985. Type-driven translation. Linguistics and Philosophy 8: 163–201. Kotek, Hadas, Yasutada Sudo, Edwin Howard, and Martin Hackl. 2011. Three readings of most. In Semantics and linguistic theory (SALT), vol. 21, ed. Neil Ashton, Anca Chereches, and David Lutz, 353–372. Ithaca: CLC Publications. Kuno, Susumu. 1973. The structure of the Japanese language. Cambridge, MA: MIT Press. Kuroda, Shige-yuki. 1965. Generative grammatical studies in the Japanese language. Cambridge, MA: MIT dissertation. Kuroda, Shige-yuki. 1972. Categorical and thetic judgments: Evidence from Japanese syntax. Foundations of Language 9: 1–37. Laenzlinger, Christopher. 2005. French adjective ordering: Perspectives on DP-internal movement types. Lingua 115: 645–689. Lakoff, Robin. 1974. Remarks on ‘this’ and ‘that’. In Proceedings of the Chicago linguistics society, vol. 10, 345–356. Chicago: Chicago Linguistic Society. Larson, Richard K., and Naoko Takahashi. 2007. Order and interpretation in prenominal relative clauses. In Proceedings of the workshop on Altaic formal linguistics II. MIT working papers in linguistics, vol. 54, 101–120. Cambridge, MA: MITWPL. Leu, Thomas. 2008. The internal syntax of determiners. New York City, NY: New York University dissertation. Leu, Thomas. 2015. The architecture of determiners. New York: Oxford University Press. Liberman, Mark. 2008. Affective demonstratives. Language Log, http://languagelog.ldc.upenn. edu/nll/?p=674. Accessed 28 Mar 2018. Lin, Jo-Wang. 2003. On restrictive and non-restrictive relative clauses in Mandarin Chinese. Tsinghua Journal of Chinese Studies, New Series 33: 199–240. Lin, Jo-Wang, and Wei-Tian Dylan Tsai. 2015. Restricting non-restrictive relative clauses in Mandarin Chinese. In Chinese syntax in a cross-linguistic perspective, ed. Yen-Hui Audrey Li, Andrew Simpson, and Wei-Tien Dylan Tsai, 100–127. Oxford: Oxford University Press. Longobardi, Giuseppe. 1994. Reference and proper names. Linguistic Inquiry 25 (4): 609–665. Longobardi, Giuseppe. 1999. Bare nouns, proper names, and the syntax-semantics mapping: Toward a unified parametric approach. Rivista di Grammatica Generativa 24: 45–76. Lyu, H.-Y. 2004. Focus and superlative center in Korean. Presentation handout for LIN 393: Seminar on Noun Phrase Modification, University of Texas at Austin. Maclaran, Rose. 1982. The semantics and pragmatics of the English demonstratives. Ithaca, NY: Cornell University dissertation. Marušič, Franc, and Rok Žaucer. 2007. On the adjectival definite article in Slovenian. Pismo 5: 102–124. Marušič, Franc, and Rok Žaucer. 2013. On the nature of prenumeral adjectives. Paper presented at the Formal Descriptions of Slavic Languages 10, University of Leipzig, Leipzig. Marušič, Franc, and Rok Žaucer. 2014. A definite article in the AP—Evidence from colloquial Slovenian. In Nominal constructions in Slavic and beyond, ed. Lilia Schürcks, Anastasia Giannakidou, and Urtzi Etxeberria, 183–208. Berlin: Mouton de Gruyter. Matushansky, Ora. 2006. Why Rose is the rose: On the use of definite articles in proper names. In Empirical issues in syntax and semantics, vol. 6, ed. Oliver Bonami, and Patricia Cabredo Hofherr, 285–307. Paris: CNRS. Retrieved from http://www.cssp.cnrs.fr/eiss6/index_en.html. Accessed 28 Mar 2018.

298

6 Extensions, Refinements, and Prospects

Ono, Kiyoharu. 1994. Territories of information and Japanese demonstratives. The Journal of the Association of Teachers of Japanese 28 (2): 131–155. Oshima, David Y., and Eric McCready. 2017. Anaphoric demonstratives and mutual knowledge: The cases of Japanese and English. Natural Language and Linguist Theory 35: 801–837. Pahom, Olga. 2013. The relationship between syntactic position and adjectival meaning in Romanian. Paper presented at the 87th Annual Meeting of the Linguistic Society of America. Boston, MA, 3–6 January 2013. Patel-Grosz, Pritty, and Patrick G. Grosz. 2017. Revisiting pronominal typology. Linguistic Inquiry 48 (2): 259–297. Potts, Christopher. 2005. The logic of conventional implicatures. New York: Oxford University Press. (2003 University of California-Santa Cruz dissertation.) Potts, Christopher, and Florian Schwarz. 2010. Affective ‘this’. Linguistic Issues in Language Technology 5: 1–29. Prince, Ellen. 1981. On the inferencing of indefinite ‘this’ NPs. In Elements of discourse understanding, ed. B.L. Webber, I. Sag, and A. Joshi, 231–250. Cambridge, UK: Cambridge University Press. Rizzi, Luigi. 1997. The fine structure of the left periphery. In Elements of grammar, ed. Liliane Haegeman, 281–337. Dordrecht: Kluwer Academic Publishers. Roberts, Craige. 2002. Demonstratives as definites. In Information sharing: Reference and presupposition in language generation and interpretation, ed. Kees van Deemter, and Roger Kibble, 89–196. Stanford, CA: CSLI. Roehrs, Dorian. 2010. Demonstrative-reinforcer constructions. Journal of Comparative Germanic Linguistics 13: 225–268. Ross, John Robert. 1964. A partial grammar of English superlatives. Philadelphia, PA: University of Pennsylvania master’s thesis. Sharvit, Yael, and Penka Stateva. 2002. Superlative expressions, context, and focus. Linguistics and Philosophy 25: 453–504. Shi, Dingxu. 2010. Xianzhixing dingyu han miaoxiexing dingyu [Restrictive and descriptive modifiers]. Foreign Language Teaching and Research 42 (5): 323–328. Shimoyama, Junko. 2014. The size of noun modifiers and degree quantifier movement. Journal of East Asian Linguistics 23 (3): 307–331. Simpson, Andrew. 1998. Empty determiners and nominalisation in Chinese, Korean and Japanese. Paper presented at the USC Symposium on East Asian languages, University of Southern California, November. Sohn, Ho-Min. 1999. The Korean language. Cambridge, UK: Cambridge University Press. Suh, Eugenina. 2005. The nominal phrase in Korean: The role of D in a “determiner-less” language. In Toronto working papers in linguistics, vol. 25, 10–19. Toronto, Canada: Department of Linguistics, University of Toronto. Svenonius, Peter. 2008. The position of adjectives and other phrasal modifiers in the decomposition of DP. In Adjectives and adverbs: Syntax, semantics, and discourse, ed. Chris Kennedy, and Louise McNally, 16–42. New York: Oxford University Press. Szabolcsi, Anna. 1983. The possessor that ran away from home. The Linguistic Review 3 (1): 89–102. Szabolcsi, Anna. 1986. Comparative superlatives. In MIT working papers in linguistics, vol. 8, ed. Naoki Fukui, Tova Rapoport, and Elizabeth Sagey, 245–265. Cambridge, MA: MIT. Szabolcsi, Anna. 1994. The noun phrase. In Syntax and semantics, vol. 27: The syntactic structure of Hungarian, ed. Ferenc Kiefer, and Katalin É. Kiss, 179–274. San Diego: Academic Press. Tomaszewicz, Barbara. 2015. Relative readings of superlatives: Scope or focus?. In Proceedings of SALT, vol. 25, ed. Sarah D’Antonio, Mary Moroney, and Carol Rose Little, 452–470. Linguistic Society of America and Cornell Linguistics Circle. Tsai, Wei-Tien Dylan. 1994. On economizing the theory of A-bar dependencies. Cambridge, MA: MIT dissertation.

References

299

Vangsnes, Øystein A. 2004. Rolling up the Scandinavian noun phrase. Paper presented at the 27th GLOW Colloquium. Athens, 19–21 April 2004. von Heusinger, Klaus. 2011. Specificity, referentiality, discourse prominence: German indefinite demonstratives. In Proceedings of Sinn Und Bedeutung, vol. 15, ed. Reich et al., 9–30. Saabrücken, Germany: Saarland University Press. Wolter, Lynsey Kay. 2006. That’s that: The semantics and pragmatics of demonstrative noun phrases. Santa Cruz, CA: University of California-Santa Cruz dissertation. Zhang, Niina. 2001. On the absence of nonrestrictive relatives in Chinese. Ms., ZAS, Berlin.

Chapter 7

Summary and Conclusion

The purpose of this book has been two-fold. One was to offer a detailed overview of the adjectival classes in Korean, an understudied topic in both generative and non-generative linguistics literatures. The other was to propose a formal system that accounts for not only the morpho-syntactic and semantic behavior of Korean N modifiers but also the relevant properties of N modifiers in other languages as well. To achieve the first goal, I took a largely theory-neutral and descriptive approach. To achieve the second goal, I proposed a decompositional split-DP analysis, building on the works by Laenzlinger (2005) and Svenonius (2008), among others, which posits just one universal DP structure and two transformational operations (Merge and Move) but which is subject to an optimality theoretic output filter. The proposed analysis has been shown to derive various DP internal constituent orders for several (unrelated) languages. It has also shown that a successful account of a linguistic phenomenon need not be either derivational or representational; it can be both, and a derivational approach and a representational approach to grammar in fact complement each other. Related to this, we have seen that there are certain typological generalizations that hold for all languages such as the presence of an Adjective-like category (cf. Dixon and Aikhenvald 2004) and the ‘Relative ADJ > Absolute ADJ > N’ order (e.g., Sproat and Shih 1988, 1990), yet languages vary tremendously from each other, and cross-linguistic variation on adjective ordering is due largely to seemingly language-specific constraints. Moreover, grammaticality is not a binary notion or property—it is inherently gradient, as was already recognized by Chomsky (1975) and has been more solidly established in recent literature (see, e.g., Keller 2000; Sorace and Keller 2005; Fanselow et al. 2006; Lau et al. 2014)—and phonology plays an extremely important role in judging the syntactic grammaticality of linguistic data even if we ignore the role of context.

© Springer Nature Switzerland AG 2019 M.-J. Kim, The Syntax and Semantics of Noun Modifiers and the Theory of Universal Grammar, Studies in Natural Language and Linguistic Theory 96, https://doi.org/10.1007/978-3-030-05886-9_7

301

302

7 Summary and Conclusion

Given this, to capture cross-linguistic variation on constituent ordering without giving up on the idea of Universal Grammar (UG), we need to resort to a metasyntactic evaluation metric which applies to possible linguistic forms that are manufactured by a derivational mechanism.

7.1

Summary of the Main Findings and Claims

In addition to what is stated above, the main findings and claims made in this book can be summarized as follows: • Even though Korean lacks a prototypical category of Adjective, it possesses a variety of resources to express a full gamut of N modificational meanings, which range from affixal ADJs to full-fledged RCs (Chap. 2). • No adjectival class in Korean can encode modal modificational semantics but all of them can carry subsective modificational semantics, and while most can encode intersective meanings, some do not (Chap. 2). • In any language, the relative morpho-syntactic complexities of N modifiers and their phonological weights matter greatly in linearizing them, but the morpho-syntactic complexity of an N modifier and its semantic possibilities do not exactly correlate, contra the prevailing view (Sproat and Shih 1988, 1990; Cinque 2010) (Chap. 3). • What is referred to as a DP in the literature is comprised of three sub-layers, which I call DPp, DPq, and DPd/r, respectively, and they each correspond to what I refer to as the Low Field, the Middle Field, and the High Field of a DP (Chap. 4). • All DP-internal functional elements originate from what I call the Low Field but they may later raise to a higher position inside the Middle or the High Field to license relevant D features such as [+quantificational], [+referential], [+deictic], [+definite], and/or [+specific], thereby complying with what I have abbreviated as PELI, i.e., the presumably universal principle that ensures that all FPs are pronounced by having their head and/or Spec position filled by some phonologically overt constituent (cf. Dimitrova-Vulchanova and Giusti 1998) (Chap. 4). • Nominals can be classified into three types: (i) predicative, (ii) quantificational, and (iii) referential. A predicative nominal is comprised of what I call a DPp and it lacks a referential LocP and a UnitP, so it cannot contain a referential DEM, a true article, or a NUM. A quantificational nominal is comprised of what I call a DPq and it contains a UnitP but lacks a LocP, so it cannot contain a DEM or an article which carries a [+referential] meaning. A referential nominal is comprised of what I call a DPd/r and it may contain all the FPs present in the Low Field including a LocP and a UnitP, not to mention a FocP, a SortP, an nP, and a √P (Chap. 4). • All three types of nominals may contain a FocP, which is located right above a LocP, and any N modifier may freely occur in its Spec position except for adjectival Ns that form an N compound or an idiom with a head N (Chap. 4). • N modifiers fall into two kinds depending on whether their content is construed as integrated into the larger message that is being conveyed by the rest of the

7.1 Summary of the Main Findings and Claims

















303

sentence or supplementary to it, and while integrated N modifiers merge DP-internally, non-integrated or supplementary N modifiers merge DP-externally (compare Huddleston and Pullum 2005; Cinque 2010, 2013) (Chap. 4). Integrated N modifiers are further divided into restrictive and non-restrictive. Restrictive modifiers merge at or below the DPq level whereas non-restrictive ones merge at the DPd/r level. Consequently, while predicative and quantificational nominals only contain restrictive modifiers, referential nominals may contain both restrictive and non-restrictive modifiers in addition to other types of discourse-oriented N modifiers (compare Cinque 2010, 2013, a.o.) (Chap. 4). Syntactic structure is binary and movement is only leftward, but multiple Spec positions can be created via adjunction and any adjoined structure creates a pause immediately after it (compare Kayne 1994). Additionally, a pause between N modifiers has both syntactic and semantic import (cf. Potts 2003/ 2005) (Chaps. 4 and 5). Non-integrated or supplementary RCs enter into sentential structure via what I call a SpplP. The head of this FP selects for a TP and it hosts a comma in the sense of Potts (2003/2005), which is phonologically realized as a pause (compare Huddleston and Pullum 2005; Demirdache 1991) (Chaps. 4–6). In article-less, prenominal RC languages like Mandarin and Korean, English-type supplementary RCs are unattested, and this is for two reasons: one is that Sppl selects for a TP. The other is that in such languages, proper names and definite descriptions denote singleton sets and as such they can be modified by non-restrictive RCs (which may look like supplementary RCs when multiple of them co-occur because of the pauses that are created due to the adjunction structure they form) (compare Lin and Tsai 2015 and the references there) (Chap. 6). In Korean/Mandarin-type languages, there is no formal distinction between restrictive RCs and non-restrictive RCs, and this is due to the combination of three factors: (i) the absence of articles in these languages, (iii) the singleton-set-denoting semantics of their proper names and definite descriptions, which lets them modified by both restrictive and non-restrictive RCs, and (iii) the need to fulfill PELI for a DPd/r in the absence of articles, which sometimes triggers raising of a restrictive RC to [Spec, DPd/r] (Chap. 6). RCs in Korean and Mandarin are subject to certain ordering restrictions which are unexpected from both typological and theoretical standpoints and their behavior calls for an OT-based filtering system which is sensitive to discourse structure and the temporal relations that hold between the RCs under evaluation as well as their eventive versus non-eventive contents (Chaps. 5 and 6). The neutral/distal DEM ku in Korean exhibits a behavior that can be characterized by calling it a cognitive indexical, and given the parallels between ku and Slovenian article-like lexeme ta, Korean is unlikely to be the only language that has such type of DEMs (Chaps. 5 and 6). Grammar encodes various kinds of pragmatic meaning including point of view (e.g., Speas and Tenny 2003; Tenny 2006; Sato and Kishida 2009; Chou 2012; Kim 2013), and the way DEMs encode speaker meaning may impact the

304

7 Summary and Conclusion

linear order of N modifiers, incurring an apparent violation of typologically well-established AORs (Chaps. 3 and 5). • Despite the rich cross-linguistic variation on the syntax and semantics of DEMs, their behavior may be captured by assuming an underlyingly uniform syntax and a limited set of binary features which differentiates between speaker and attitude holder and which posits at least two different domains for demonstration (e.g., physical versus non-physical) (compare, a.o., Brugè 1996, 2002; Bernstein 1997, 2001; Diessel 1999; Elbourne 2008; Leu 2008, 2015; Wolter 2006) (Chap. 6).

7.2

What All This Leads Us To

Collectively, our findings show that there is an intricate interplay between phonology, morphology, syntax, and semantics in linearizing N modifiers and some of the “unusual” behaviors exhibited by the world’s languages are due in part to language-specific constraints. But this does not mean that the syntax of N modifiers is not constrained or it cannot be explained in a principled manner because in any given language, the DP structure is comprised of just a handful of FPs that may license N modifiers, as has been suggested by authors like Svenonius (2008) and Cinque (2010) although there are several non-trivial differences between their analyses and the analysis proposed here. Another important conclusion to be drawn from the present research is that the ‘NP-inner versus NP-outer’ distinction that Larson (1998, 2000) makes in discussing the syntax and semantics of N modifiers is real and the NP-outer space is reserved for discourse-oriented elements, but there is reason to posit a more fine-grained DP-internal structure than Larson does, as we have done here and as authors like Ihsane and Puskas (2001) and Laenzlinger (2005) have done. Furthermore, to capture the difference between integrated N modifiers and non-integrated ones, the DP-internal versus DP-external merge distinction needs to be made in addition to the NP-inner versus NP-outer spatial distinction. Next, it can be concluded that, at the core, all languages likely have the same underlying DP structure, yet what occupies each slot inside the DP may vary from language to language because different languages have different resources for N modification, as we have seen here with the case of Korean in particular, and as suggested by authors like Svenonius (2008). The present research also lets us draw at least three conclusions that pertain to linguistic typology: First, we can conclude that the way in which RCs behave in any given language has a lot to do with whether the language at hand is article-less or article-possessing and whether it is a prenominal- or a postnominal-RC language. Secondly, the fact that all Korean adjectival classes can encode subsective adnominal semantics lends support to Partee’s (2009a, b) claim that all adnominal adjectives are in fact either subsective or modal rather than intersective, modal,

7.2 What All This Leads Us To

305

privative, or subsective, unlike what has been assumed in the more traditional formal semantics literature (e.g., Kamp and Partee 1995; Partee 1995; cf. Morzycki 2016). Thirdly, the fact that Korean DEM ku performs an affective function1 although the affective meaning it carries is of a slightly different kind than what has been reported about English DEMs (see, a.o., Lakoff 1974; Prince 1981; Wolter 2006; Liberman 2008; Potts and Schwarz 2010; Acton and Potts 2014) leads us to conclude that affectivity is most likely a universal property of DEMs, as has been conjectured by authors like Potts and Schwarz (2010).

7.3

Understanding (Korean) N Modifiers in a New Light

Before closing this chapter, I would like to show that revisiting N modification phenomena in Korean in the light of what has been uncovered here leads us to have a better understanding of what role each type of N modifier may play in an article-less, prenominal modifier, and classifier language, and why certain N modifiers may be permitted only in certain contexts. To see this, let us first consider the discourse given in (1). Here, the object position of the sentence comprising A’s utterance is pronounced by a bare nominal, and even though bare nominals in Korean can be rendered both definite and indefinite interpretations as shown in (2), A’s remark in (1) makes the hearer respond with an information- or clarification-seeking question, namely, ‘What necklace?’. (1) Context: A and B are friends. They just got together and A says something to start the conversation. il-e peri-ess-ta. A: Na-nun onul [DP mokkeri]-rul today [ necklace]-ACC lose-CONN AUX-PST-DECL I-TOP Intended: ‘Today, I lost a necklace.’ B: Mwusun mokkeri? What necklace ‘What necklace?’ (2) Bare nominals receiving a definite or an indefinite interpretation in Korean: a. Mwun yel-e! (definite construal) Door open-IMP.INFRML ‘Open the door!’ b. Na cha-ka philyoha-y. (indefinite construal) I car-NOM need-DECL.IMFRML ‘I need a car.’ (Lit.: ‘To me, a car is needed.’)

1

About ku in Korean, some authors have noted that it performs an intensifier-like function when occurring as a part of an adverb kurehkey ‘in that way’ or ‘so’ (e.g., Suh 2002) and others have remarked that it may carry a mirative meaning (e.g. Kang 2018), but no previous work has articulated its affective use in the way I have done here, looking at the same kind of linguistic phenomenon.

306

7 Summary and Conclusion

Consider now (3) in comparison to (1). Here, the object position in A’s utterance contains a numeral classifier, and this time, B does not respond by asking what necklace A has lost—in fact doing so would incur pragmatic anomaly, as indicated below. Instead, B says that she feels sorry about what happened to A, based on what A has just said. This shows that in a language like Korean, a ‘NUM + CL’ cluster encodes indefiniteness, and given this, the presence of hankay in (3A) can be argued to be obligatory even though both (1A) and (3A) are syntactically grammatical sentences. (3) Context: Same as in (1). A: Na-nun onul [DP mokkeri han-kay]-rul il-e today [ necklace one-CL]-ACC lose-CONN I-TOP Intended: ‘Today, I lost a necklace.’ B: Eme, pro cengmal soksangha-keyss-ta! Oh.no, __ really be.upset-must-EXCL ‘Oh no, you must be really upset (about it)!’ B :#Mwusun mokkeri? What necklace ‘What necklace?’

peri-ess-ta. AUX-PST-DECL

Consider now (4). This discourse exemplifies that to encode definite meaning, Korean sometimes utilizes non-restrictive RCs: if the RC is absent in A’s utterance in (4), then her interlocutor will say something like ‘What necklace?’, as shown in (4B′), since what is conveyed by A is not informative enough. On the other hand, if the RC is present, then she will respond differently by saying something like what is given in (4B) and thereby showing empathy, and this demonstrates that even though bare nominals are prevalent in Korean and they may refer to definite entities as exemplified in (2a), there are contexts in which a nominal must co-occur with a non-restrictive RC in order to mark [+definite]. (4) Context: A and B are friends. They just got together and A says something to start the conversation. puracil-eyse ei A: Na-nun onul [DP *([RC Minswu-ka cinan-cwu-ey I-TOP today [ ([ M.-NOM last-week-LOC Brazil-from __ mokkerii]-rul il-e peri-ess-ta. sata-cwu- ]-n) necklace]-ACC lose-CONN AUX-PST-DECL buy-give-ANT]-REL) Intended: ‘Today, I lost the necklace *(that Minswu bought and brought for me from Brazil last week).’ B: Eme, pro cengmal soksangha-keyss-ta! Oh.no, __ really be.upset-must-EXCL ‘Oh no, you must be really upset (about it)!’ B : Mwusun mokkeri? What necklace ‘What necklace?’

In the literature, the distal DEM ku is considered as marking [+definite] in Korean because there are no articles in the language (e.g., Cho 1999; Ionin et al. 2012). Given this, one may think that ku may occur in (4A), replacing the

7.3 Understanding (Korean) N Modifiers in a New Light

307

non-restrictive RC, but in fact, it cannot. To see this, consider (5). This discourse shows that, upon hearing A’s utterance, her interlocutor would immediately say something like ‘What necklace?’, unlike the case with (4); furthermore, (5A) sounds infelicitous, more than just being uninformative. In view of what I proposed in Chap. 6, though, the infelicity of (5A) is expected because Korean DEMs can refer to discourse-old entities only; in (5), the DEM ku occurs in a sort of null context without any linguistic antecedent. (5) Context: A and B are friends. They just got together and A says something to start the conversation. mokkeri]-rul il-e peri-ess-ta. A:#Na-nun onul [DP ku I-TOP today [ that necklace]-ACC lose-CONN AUX-PST-DECL Intended: ‘Today, I lost the necklace.’ B: Mwusun mokkeri? What necklace ‘What necklace?’

At this juncture, I should point out that an attributive AP cannot substitute for the non-restrictive RC in (4A) either. To see this, consider (6). Here, the object position of A’s utterance contains the AP kappissan ‘expensive’, and the presence of this AP makes the object DP construed as referring to an indefinite entity rather than a definite one. Notice also that (6A) is actually judged uninformative since, in response to it, the hearer will say something like ‘Which expensive necklace?’ in a manner similar to (1), as given in (6B).2 (6) Context: A and B are friends. They just got together and A says something to start the conversation. A: Na-nun onul [DP kappissan mokkeri]-rul il-e peri-ess-ta. today [ expensive necklace]-ACC lose-CONN AUX-PST-DECL I-TOP Intended: ‘Today, I lost an expensive necklace.’ B: Etten kappissan mokkeri? Which expensive necklace ‘Which expensive necklace?’

Interestingly, however, in the same discourse context, one can felicitously utter what is given in (7A), which contains both a DEM and an AP as well as a non-restrictive RC; in this discourse, the hearer will not respond by saying something like ‘Which expensive necklace?’.

2

In (6), A’s utterance would not be met by B’s puzzled response if it contains han-kay; a ‘NUM + CL’ cluster encodes indefiniteness in Korean, as mentioned above, so if her utterance contains han-kay, then she would be understood as introducing a new discourse referent.

308

7 Summary and Conclusion

(7) Context: A and B are friends. They just got together and A says something to start the conversation. A: Na-nun onul [DP [RC Minswu-ka cinan-cwu-ey puracil-eyse ei I-TOP today [ [ M.-NOM last-week-LOC Brazil-from __ sata-cwu- ]-n, ku kappissan mokkerii]-rul il-e buy-give-ANT]-REL that expensive necklace]-ACC lose-CONN peri-ess-ta. AUX-PST-DECL Intended: ‘Today, I lost that really expensive necklace which Minswu bought and brought for me from Brazil last week.’ B: Eme, pro cengmal soksangha-keyss-ta! Oh.no, __ really be.upset-must-EXCL ‘Oh no, you must be really upset (about it)!’

Why can (7A) contain all three N modifiers? And what is each modifier’s role in this utterance? In view of what I have proposed in Chaps. 4 through 6, the presence of the episodic RC in (7A) is for two reasons: one is to make the object DP referential and the other is to comply with PELI. To be more specific, by occurring in the Spec position of the DPd/r, the RC not only licenses the relevant D features (i.e., [+referential; -deictic; +definite; +specific]) but also pronounces the highest FP that comprises the object DP. As for the presence of the DEM ku and the AP kappissan in the sentence, they are there to convey the speaker’s attitude toward the necklace at issue: uttering the RC before the DEM has the effect of introducing the referent of the DEM into the discourse, making it discourse-old, but this only partly licenses the DEM because if the AP were not there, the DEM would not be there either, as shown in (8A). And this suggests that ku occurs in (7A) to form a syntactic unit with kappissan—that is, it instantiates what I call a CogDem—and the ‘DEM + AP’ string implicates that even though the necklace at issue is not present in the discourse context, it is visible in the speaker’s mind and the degree to which the property of being expensive holds true of the necklace is very high, and thus is noteworthy. (8) Context: A and B are friends. They just got together and A says something to start the conversation. Minswu-ka cinan-cwu-ey puracil-eyse ei A: Na-nun onul [DP [RC I-TOP today [ [ M.-NOM last-week-LOC Brazil-from __ sata-cwu- ]-n (*ku) mokkerii]-rul il-e necklace]-ACC lose-CONN buy-give-ANT]-REL (that) peri-ess-ta. AUX-PST-DECL Intended: ‘Today, I lost the necklace that Minswu bought and brought for me from Brazil last week.’ B: Mwusun sori-i-ya? What word-COP-Q.INFRML ‘What do you mean?’ cengmal soksangha-keyss-ta! B :#Eme, pro Oh.no, __ really be.upset-must-EXCL ‘Oh no, you must be really upset (about it)!’

7.3 Understanding (Korean) N Modifiers in a New Light

309

In sum, then, in contexts like (7), the non-restrictive RC is present in the speaker’s utterance for grammar-proper reasons but the DEM’s and the AP’s presence in the utterance is more for pragmatic reasons. And given the order in which they occur, we can infer that, in Korean-type languages, all else being equal, an N modifier that helps identify the referent of the DP at hand (an ‘identificational modifier’ for short) occurs first and an N modifier that helps comment on the referent (a ‘speaker-oriented modifier’ for short) occurs next, and while an identificational N modifier may not be omitted, a speaker-oriented modifier can be. When viewed this way, Korean-type languages turn out to be not so different from English-type languages since in the latter cases too, N modifiers occur in the order of ‘Identificational > Speaker-oriented’. For example, supplementary RCs contribute speaker meanings, i.e., conventional implicatures in the sense of Potts (2003/2005), and they invariably occur following an integrated RC which helps identify the referent of the DP at hand, as illustrated in (9). (Note that in (9b), there is no pause after the second RC, and this is to indicate that this RC is an integrated one; if there is a pause after it, then it will be construed as a supplementary RC, just like the first RC, and in such cases, the sentence will be judged grammatical, but that is not what is intended here.) (9) Context: The speaker went to the Fishing Range over the weekend and he is now talking about what happened there, especially about some old man who sat next to him. a. The old man [who was sitting next to me], [who yelled at me, so I yelled back], turned out to be my girlfriend’s stepfather! (NRS-RC > Sppl-RC) b. *The old man, [who yelled at me, so I yelled back], [who was sitting next to (Sppl-RC > NRS-RC) me] turned out to be my girlfriend’s stepfather!

Given this parallel between English and Korean N modifiers, what I call CogDemPs and SpplPs in this book may merit similar treatments, and the behavior of N modifiers in data like (7)–(9) and the Korean/Mandarin RC ordering phenomena discussed in Chaps. 5 and 6 may in fact be governed by the same set of principles, namely, principles that are sensitive to discourse structure as well as the way in which human linguistic agents package and/or process information by using language. That said, I should point out that the grammaticality of (9a) and the ungrammaticality of (9b) are actually expected under the analysis I have proposed regarding the syntax of supplementary RCs and that of non-restrictive RCs in Chap. 4. First, (9a) is predicted to be grammatical because its surface structure can be readily derived as depicted in (10): given the non-co-occurrence of DEMs and articles in English, which suggests that English satisfies PELI by either having the Spec or the head position of an FP pronounced but not both, we assume that the non-restrictive RC inside the subject position of the sentence is base-generated at an adjoined [Spec, DPd/r] (similarly to RC2 in derivation (91) in Chap. 4)—it cannot be base-generated at the pre-existing [Spec, DPd/r] because the already occupies its head position by raising there from the head position of LocP. After the merge of the RC, the lower DPd/r raises to [Spec, DPd/r+], engendering an ‘DET > N > RC’ surface order, and this is followed by Sppl selecting for the entire TP and then its subject DP

310

7 Summary and Conclusion

raising to [Spec, SpplPd/r+], thereby c-commanding the Sppl-RC. Notably, the resulting structure accords with the fact that, in (9a), there is a pause both before and after the second RC; this RC is both preceded and followed by a pause because the first RC’s adjunction to the DPd/r creates a pause after it and the head of a SpplP inherently hosts a comma, which is pronounced as a pause. (10) Derivation of (9a) under the present analysis: SpplP+

DPd/r+2 DPd/r1

SpplP DPd/r RC

RC The old man

who was sitting next to me

Sppl

t1 Sppl ,

TP

who yelled at me, so I yelled back t2 turned out to be my girlfriend’s stepfather

Unlike (9a), (9b) does not obtain and here is why: since the RCs in (9b) are meant to perform the same functions as those in (9a), this sentence would be derived from what is given in (10). But there is no reason for the SpplP to move, modulo the difficulty identifying its target position, and even if there is a reason for it to move, the raising of the SpplP in (10) would yield (11). So either way, there is simply no way to derive a sentence like (9b). (11) Consequence of raising the SpplP in (10) to a higher position: ‘Sppl-RC > TP with a trace of the raised subject DP > Subject DP > NRS-RC’ *Who yelled at me, so I yelled back, turned out to be my girlfriend’s stepfather the old man who was sitting next to me.

7.4

Conclusion

In the light of this and other results and findings, I conclude that the research program proposed here is viable and it will receive even more support if one widens the domain of inquiry and looks at other languages from a similarly holistic perspective. One other thing I believe this research has shown is that nothing in language is cut-and-dried but everything in it is principled and structurally based,

7.4 Conclusion

311

and while some components of UG generate linguistic forms, others may evaluate them, and finally, language exists to aid humans to communicate, so syntactic structure is bound to parallel discourse structure, whether it pertains to the nominal domain or the clausal domain,3 and whether it concerns a prenominal language or a postnominal language.

References Acton, Eric K., and Christopher Potts. 2014. That straight talk: Sarah Palin and the sociolinguistics of demonstratives. Journal of Sociolinguistics 18 (1): 3–31. Adger, David, Daniel Harbour, and Laurel J. Watkins. 2011. Mirrors and microparameters: Phrase structure beyond free word order. Cambridge Studies in Linguistics. Cambridge, UK: Cambridge University Press. Bernstein, Judy B. 1997. Demonstratives and reinforcers in Romance and Germanic languages. Lingua 102: 87–113. Bernstein, Judy B. 2001. Focusing the ‘right’ way in Romance determiner phrases. Probus 13: 1–29. Bruening, Benjamin. 2001. Syntax at the edge: Cross-clausal phenomena and the syntax of Passamaquoddy. Cambridge, MA: MIT dissertation. Brugè, Laura. 1996. Demonstrative movement in Spanish: A comparative approach. In University of Venice working papers in linguistics, vol. 6 (1), 1–53. Brugè, Laura. 2002. The positions of demonstratives in the extended nominal projection. In Functional structure in DP and IP: The cartography of syntactic structures, vol. 1, ed. Cinque, Guglielmo, 15–53. New York: Oxford University Press. Cho, Hye-sun. 1999. Interpretation and function of the Korean demonstrative ku. Studies in Modern Grammar 18: 71–90. Chomsky, Noam. 1975. The logical structure of linguistic theory. New York: Plenum Press. Chou, Chao-Ting Tim. 2012. Syntax-pragmatics interface: Mandarin Chinese wh-the-hell and point-of-view operator. Syntax 15 (1): 1–24. Cinque, Guglielmo. 2013. Typological studies: Word order and relative clauses. New York: Routledge. Cinque, Guglielmo. 2010. The syntax of adjectives. A comparative study. Cambridge, MA: MIT Press. Demirdache, Hamida. 1991. Resumptive chains in restrictive relatives, appositives, and dislocation structure. Cambridge, MA: MIT dissertation. Diessel, Holger. 1999. Demonstratives: Form, function, and grammaticalization (Typological Studies in Language vol. 42). Amsterdam: John Benjamins. Dimitrova-Vulchanova, Mila, and Giuliana Giusti. 1998. Fragments of Balkan nominal structure. In Possessors, predicates and movement in the determiner phrase, ed. Artemis Alexiadou, and Chris Wilde, 333–360. Amsterdam: John Benjamins. Dixon, R.M.W., and Alexandra Y. Aikhenvald (eds.). 2004. Adjective classes: A cross-linguistic typology. Explorations in linguistic typology. Oxford: Oxford University Press. Elbourne, Paul. 2008. Demonstratives as individual concepts. Linguistics and Philosophy 31 (4): 409–466.

3

Given this conclusion, the decompositional OT-based analysis proposed here may shed new light on so-called discourse-configurational languages such as Warlpiri, Passamaquoddy, and Kiowa, which are well known for their (apparently) fluid surface constituent orders in the clausal domain. For discussion and data, see, a.o., Legate 2001, Bruening 2001, and Adger et al. 2011.

312

7 Summary and Conclusion

Fanselow, Gisbert, Caroline Féry, Matthias Schlesewsky, and Ralf Vogel (eds.). 2006. Gradience in grammar: Generative perspectives. Oxford: Oxford University Press. Huddleston, Rodney, and Geoffrey Pullum. 2005. A student’s introduction to English grammar. Cambridge, UK: Cambridge University Press. Ihsane, Tabea, and Genoveva Puskas. 2001. Specific is not definite. Generative Grammar in Geneva 2: 39–54. Ionin, Tania, Soondo Baek, Eunah Kim, Heejeong Ko, and Kenneth Wexler. 2012. That’s not so different from the: Definite and demonstrative descriptions in second language acquisition. Second Language Research 28: 69–101. Kamp, Hans, and Barbara H. Partee. 1995. Prototype theory and compositionality. Cognition 57: 29–191. Kang, Arum. 2018. Unexpected effect: The emphatic determiner with gradable NPs in Korean. The Journal of Studies in Language 33 (4): 595–615. Kayne, Richard S. 1994. The antisymmetry of syntax. Cambridge, MA: MIT Press. Keller, Frank. 2000. Gradience in grammar: Experimental and computational aspects of degrees of grammaticality. Edinburgh, UK: University of Edinburgh dissertation. Kim, Jaieun. 2013. Subject and point-of-view in Korean: The syntax-discourse interface. Seoul, Korea: Sogang University master’s thesis. Laenzlinger, Christopher. 2005. French adjective ordering: Perspectives on DP-internal movement types. Lingua 115: 645–689. Lakoff, Robin. 1974. Remarks on ‘this’ and ‘that’. In Proceedings of the Chicago linguistics society, vol. 10, 345–356. Chicago: Chicago Linguistic Society. Larson, Richard K. 1998. Events and modification in nominals. In Proceedings from semantics and linguistic theory (SALT) VIII, 145–168. Cornell University Press. Larson, Richard K. 2000. Temporal modification in nominals. Paper presented at the International Roundtable on the Syntax of Tense, University of Paris VII, France. Lau, Jey Han, Alexander, Clark, and Shalom Lappin. 2014. Measuring gradience in speakers’ grammaticality judgements. In Proceedings of the annual meeting of the cognitive science society, vol. 36, 821–826. Legate, Julie Ann. 2001. Two types of nominal split. Paper presented at the 36th Annual Meeting of the North East Linguistic Society (NELS), Amherst, MA, October 28–30. Leu, Thomas. 2008. The internal syntax of determiners. New York City, NY: New York University dissertation. Leu, Thomas. 2015. The architecture of determiners. New York: Oxford University Press. Liberman, Mark. 2008. Affective demonstratives. Language Log, http://languagelog.ldc.upenn. edu/nll/?p=674. Accessed 28 Mar 2018. Lin, Jo-Wang, and Wei-Tian Dylan Tsai. 2015. Restricting non-restrictive relative clauses in Mandarin Chinese. In Chinese syntax in a cross-linguistic perspective, ed. Yen-Hui Audrey, Li, Andrew Simpson, and Wei-Tien Dylan Tsai, 100–127. Oxford: Oxford University Press. Morzycki, Marcin. 2016. Modification. Cambridge, UK: Cambridge University Press. Partee, Barbara H. 1995. Lexical semantics and compositionality. In An invitation to cognitive science, vol. I: Language, ed. Lila, Gleitman, and Mark Liberman, 311–360. Cambridge, MA: MIT Press. Partee, Barbara H. 2009a. The dynamics of adjective meaning. In Computational linguistics and intellectual technologies: Papers from the annual international conference “Dialogue 2009”, ed. A.E. Kibrik et al., 593–597. Moscow: Russian State Humanities University. Partee, Barbara H. 2009b. Formal semantics, lexical semantics, and compositionality: The puzzle of privative adjectives. Philologia 7: 11–23. Potts, Christopher. 2005. The logic of conventional implicatures. New York: Oxford University Press. (2003 UC-Santa Cruz dissertation.) Potts, Christopher, and Florian Schwarz. 2010. Affective ‘this’. Linguistic Issues in Language Technology 5: 1–29.

References

313

Prince, Ellen. 1981. On the inferencing of indefinite ‘this’ NPs. In Elements of discourse understanding, ed. B.L. Webber, I. Sag, and A. Joshi, 231–250. Cambridge, UK: Cambridge University Press. Sato, Yosuke, and Maki Kishida. 2009. Psychological predicates and the point-of-view hyperprojection. Gengo Kenkyu [Language Research] 135: 123–150. Sorace, Antonella, and Frank Keller. 2005. Gradience in linguistic data. Lingua 115 (11): 1497–1524. Speas, Peggy, and Carol Tenny. 2003. Configurational properties of point of view roles. In Asymmetry in grammar, ed. Di Sciullo, Anna Maria, 315–344. Amsterdam: John Benjamins. Sproat, Richard, and Chinlin Shih. 1988. Prenominal adjectival ordering in English and Mandarin. In Proceedings of NELS, vol. 18, 465–489. Amherst, MA: GLSA. Sproat, Richard, and Chinlin Shih. 1990. The cross-linguistics distribution of adjectival ordering restrictions. In Interdisciplinary approaches to language: Essays in honor of S-Y. Kuroda, ed. Carol, Georgopoulos, and Roberta Ishihara, 565–593. Dordrecht: Kluwer. Suh, Kyung-Hee. 2002. An interactional account of the Korean demonstrative ku in conversation. Enewa Enehak [Language and Linguistics] 29: 137–158. Svenonius, Peter. 2008. The position of adjectives and other phrasal modifiers in the decomposition of DP. In Adjectives and adverbs: Syntax, semantics, and discourse, ed. Chris Kennedy, and Louise McNally, 16–42. New York: Oxford University Press. Tenny, Carol L. 2006. Evidentiality, experiencers, and the syntax of sentience in Japanese. Journal of East Asian Linguistics 15: 245–288. Wolter, Lynsey K. 2006. That’s that: The semantics and pragmatics of demonstrative noun phrases. Santa-Cruz, CA: University of California-Santa Cruz dissertation.

Index

A Abney, Steven, 292 Aboh, Enoch Oladé, 105, 109, 110 Acton, Eric K., and Christopher Potts, 209, 305 Adjective ordering, 2, 67, 175, 180, 181, 201, 301 languages Aromanian, 275 English, 1, 98, 104, 105, 122, 125, 130, 134–139, 142, 144, 146 French, 109–111, 113, 133, 148, 154, 283 Greek, 129 Italian, 110, 148, 150, 154, 283–286 Korean, 177, 183, 192, 193, 221, 224, 229, 231 Mandarin, 69, 84, 86, 237, 238, 240–243, 245, 246, 249, 303, 309 Norwegian, 106, 269 Romanian, 129, 284, 286 Slovenian, 237, 271, 272, 274–277, 303 Spanish, 110, 129, 146–148, 150, 152–154, 283 typological generalizations, 1, 3, 67, 69 weight-based, 80, 83, 89, 181–190, 218, 233 Adjective(s) absolute vs. relative, 2, 71, 97, 98, 288, 290 adjective hierarchy, 61 adjective classes English, 6, 9, 12 Korean—adjectival Ns, 29, 31–33, 35, 36, 41, 43, 47, 70, 81, 130, 161, 163, 302

Korean—adjectival prefixes, 24, 27–29, 31, 33, 34, 37, 49, 70, 81, 90 Korean—attributive determiners (ATT-DETs), 23, 33–36, 40, 48, 49, 56, 60, 70, 74, 75, 81, 90, 161, 164, 170, 171, 174, 188, 190, 192 Korean—CEK-XP or CEK-APs, 23, 31, 39–43, 45–48, 52, 60, 70, 72, 74, 75, 101, 161–164, 186, 189, 190 Korean—full relative clauses, 163, 165, 166, 168, 176, 178, 179, 194, 195, 197–207, 211, 212, 214, 218, 220, 221, 223–227, 229, 232, 233, 243, 244, 248, 250, 278–283, 294, 295, 301–304, 306–309 Korean—UN-XPs or UN-APs, 23, 47, 55, 59, 61, 70, 71, 75, 77, 81, 82, 99, 101, 163, 164, 170, 172–175, 183, 189, 190, 192, 287, 307, 308 “adverbial” semantics, 11, 35, 37, 40, 49, 56, 60, 70 evaluative semantics, 2, 98, 285 semantic taxonomy intersective, 27, 31, 38, 43–45, 52, 54, 56, 57, 59, 60, 62, 98, 99, 105, 131, 287, 302, 304 subsective, 27, 28, 32, 37, 38, 43–45, 57, 61, 62, 71, 105, 131, 134, 137, 287, 289, 302, 304 modal, 29, 33, 38, 43, 54, 58–63, 98, 302, 304 privative, 28, 32, 37, 38, 43, 45, 53, 57, 60–62, 302, 304 sortal, 74, 104, 130, 131, 133, 163, 164

© Springer Nature Switzerland AG 2019 M.-J. Kim, The Syntax and Semantics of Noun Modifiers and the Theory of Universal Grammar, Studies in Natural Language and Linguistic Theory 96, https://doi.org/10.1007/978-3-030-05886-9

315

316

Index

verbal vs. nominal adjective phrase (AP), 181, 184 Adjunction, 87, 121, 122, 145, 162, 166, 181, 201, 226, 303, 310 multiple spec, 121, 126, 303 Adnominalizer (ADN), 90 Agree, 124 Agreement, 109–111, 124, 133, 135, 143, 144, 148, 165, 269, 286 Alexiadou, Artemis, and Chris Wilder, 3 Alexiadou, Artemis, Liliane Haegeman, and Melita Stavrou, 1, 4, 9–11, 97, 128, 275 Alleged, 8, 11, 54, 59 An, Duk-Ho, 165 Anaphor (or anaphoric), 98, 147, 214, 229, 231, 232, 251–255, 257, 261, 263, 265, 266, 268, 286, 295 Anderson, Stephen R., and Edward L. Keenan, 250 Articles, 4, 113, 123, 124, 126, 128, 129, 148, 154, 286, 303, 306, 309 Attitude Holder (AH), 254–257, 259, 260, 262, 265, 304

Chierchia, Gennaro, 101 Cho, Hye-sun, 306 Choi, Hye-Won, 7, 201, 288 Chomsky, Noam, 3, 60, 123, 275, 301 Chou, Chao-Ting Tim, 233, 303 Choy, Hyen-Pay, 48 Cinque, Guglielmo, 1–4, 6, 10, 37, 67–70, 72, 80, 89, 97–99, 101–103, 105, 107, 109, 110, 113, 117, 120, 121, 134, 138, 146, 164, 183, 233, 241, 283, 284, 286, 292, 302, 304 Clark, Romane L., 32 Classifiers (CLs), 87, 88, 104, 164, 305 Comma, 6, 117, 122, 218, 278–280, 282, 303, 310 Complementizer, 17 Complementizer Phrase (CP), 110, 118, 280, 292 left periphery, The, 191, 233 Compounds, 72, 130, 131, 161, 162, 182, 302 Copular verbs, 90, 144, 182, 189 Cornilescu, Alexandra, and Alexandru Nicolae, 286

B Baker, Mark, 60 Beautiful, 11 Beck, David, 97 Bernstein, Judy B., 5, 98, 124, 128, 129, 146–148, 150, 152, 153, 269, 304 Bolinger, Dwight, 2, 3, 5, 6, 9, 10, 27, 44, 52, 67, 97, 133–135, 137, 183, 283 Borer, Hagit, 104 Bosque, Ignacio, and Carme Picallo, 110 Bouchard, Denis, 1, 3, 10, 67, 97, 183 Bowdle, Brian F., and Gregory Ward, 263 Brugè, Laura, 5, 98, 104, 110, 125, 128, 129, 146, 148, 304

D Davidson, Donald, 11 Dayal, Veneeta, 153 Degree adverb, 36, 37, 40–42, 48, 60, 77, 82, 83, 276 Deictic, 5, 69, 93, 123, 126, 136, 143, 144, 147, 149, 150, 219, 221, 246, 248, 250, 255, 257, 263, 266, 269, 270, 286, 302 DeLancey, Scott, 219 Del Gobbo, Francesca, 84, 238, 242, 250 Delsing, Lars-Olof, 124, 269 Demirdache, Hamida, 118, 303 Demonstratives (DEMs), 1, 17–19, 67, 78, 87, 88, 91, 93–95, 97, 102, 104–107, 113–115, 123–127, 129, 133, 135, 136, 142, 146–150, 152–154, 160, 163, 166, 168, 170, 172, 207, 208, 210, 212, 217–219, 221–224, 227, 229, 230, 232, 233, 237, 248–263, 265–277, 283, 286, 287, 292, 294, 295, 302, 303, 305–309 affective, 5, 78, 160, 209, 218, 233, 250, 252–262, 305 cognitive indexicals (CogDem), 208, 221–223, 226, 227, 229–233, 237, 265, 267, 272, 274, 276, 277, 295, 303, 308, 309 demonstration, 231, 304 distal, 207, 219, 249, 250, 252, 255, 256, 259, 261, 264, 266, 273, 303, 306

C Cabredo Hofherr, Patricia, and Ora Matushansky, 1 Campos, Héctor, and Melita Stavrou, 3, 275 Cartography, 5, 122, 233, 292 Case accusative, 30 case drop (non-overt case marking), 18, 20 case particles, 17, 18 nominative, 20, 30 nominative vs. topic-marking, 288 Categories, 4, 6, 17, 21, 37, 60, 93, 103, 105, 114, 124, 133, 267 C-command, 125, 147, 148, 152, 222, 247, 281 Chang, Suk-Jin, 17

Index languages Afrikaans, 269 English, 1, 126, 142, 208, 237–257, 261–265, 269, 305 French, 148, 150 German, 252 Greek, 129 Hungarian, 129 Italian, 148, 150 Japanese, 255 Jordanian Arabic, 129 Korean, 18, 78, 87, 208, 211, 214, 215, 217–224, 226, 227, 229–233, 250–252, 254–265, 267, 268, 272, 277, 278, 304, 306, 309 Norwegian, 106, 269 Romanian, 129 Slovenian, 7, 237, %, 272, 274–277, 295, 303 Spanish, 146–148, 150, 152, 154 Swiss German, 269 proximal, 127, 250, 253 –257, 259–261, 263, 265 reinforcers (RFs), 148, 150, 269 Demonte, Violeta, 108 Determiner Phrase (DP) complex DP, 97, 109–111, 113 decompositional DP, 6, 7, 97, 102, 104–108, 122, 123, 126, 127, 129–141, 143–146, 150, 153, 160, 163, 170, 180, 232, 294, 295, 301, 311 DPd/r (deictic/referential), 5, 108, 123, 124, 126–131, 133, 135, 136, 139, 143–145, 149, 150, 154, 162, 163, 166, 172, 174, 201–204, 222, 223, 226, 227, 229–233, 239, 241, 245, 246, 248, 250, 267, 274, 275, 277, 280, 286, 302, 303, 308–310 DPdeixis, 110, 123 DPdetermination, 110, 123 DPp (predicative), 5, 123, 124, 126, 127, 129–131, 133, 135, 136, 138, 139, 142, 144, 146–150, 153, 163, 180, 182, 183, 186, 192, 241, 245, 250, 282, 302 DPq (quantificational), 5, 123, 124, 126, 128, 129, 131, 133–136, 138, 140, 142, 154, 166, 241, 302 indefinite DP, 4, 98, 123, 127, 135, 171, 180, 251, 254, 258, 272, 274 multiple determiner (determiner doubling or determiner spreading or polydefiniteness), 3, 275 NP-internal vs. NP-external space, 137, 138, 304

317 predicative nominal, 3, 115, 118, 120, 125, 127, 135, 139, 153, 180, 280, 302 quantificational nominal, 120, 121, 138–140, 302 referential nominal, 115, 120, 153, 269, 302 specific DP, 55, 72, 135, 136, 143, 144, 149, 150, 164–166, 168, 211, 246, 251, 253, 254, 274, 275, 286, 302, 308 split-DP (structure), 6, 7, 110, 122, 131, 133, 146, 154, 160, 170, 180, 208, 232, 245, 250, 283, 294, 301 tri-partite DP structure, 5, 123, 131, 159, 166 high field, 5, 120, 121, 123, 131, 133, 135, 159, 302 low field, 5, 120, 121, 123, 124, 131–133, 136, 154, 159, 180, 270, 287, 294, 302 middle field, 5, 120, 121, 123, 124, 131, 133, 159, 302 Diessel, Holger, 304 Dimitrova-Vulchanova, Mila, and Giuliana Giusti, 128, 241, 302 Discourse, 7, 11, 12, 45, 62, 78, 80, 91, 93, 98, 102, 108, 127, 133, 147, 160, 165, 167, 172, 192, 194, 195, 197–202, 204, 205, 207–209, 211, 216, 218, 220–224, 226, 227, 231, 233, 239, 240, 251, 254, 255, 258, 259, 261–263, 265, 266, 274, 278, 286, 287, 289, 290, 292, 293, 303–309, 311 Dixon, R. M. W., 97, 121 Dixon, R.M.W, and Alexandra Y. Aikhenvald, 301 Doran, Ryan B., and Gregory Ward, 127, 128 Dryer, Matthew, 1, 67, 124 E Elbourne, Paul, 124, 256, 304 F Fake, 8, 62 Fanselow, Gisbert, Caroline Féry, Matthias Schlesewsky, and Ralf Vogel, 301 Feature checking/inheritance/valuation/ licensing, 110, 111, 135, 139, 143, 144, 147–150, 153, 163, 166, 168, 172, 246 Features (semantic/formal) or feature-specification, 5, 60, 117, 121, 123, 124, 126, 128, 135, 154, 222, 224, 227, 237, 250, 254, 255, 257, 258, 263, 265–268, 270, 274, 286, 292, 295, 302, 304, 308

318 Finite Phrase (FinP), 203, 292 Focus (FOC), 5, 7, 18, 75, 82, 91, 94, 102, 107, 108, 111, 113, 123, 125, 126, 147–150, 152–154, 159, 161, 168, 182, 187, 190, 218, 219, 232, 233, 237, 274, 283, 288, 289, 291–295 additive focus, 219, 292 contrastive focus, 18, 147, 168, 170, 210, 213, 218, 219, 232, 274, 287, 291, 292, 294 exhaustive focus, 288, 292–294 identificational focus, 291, 292, 294 Focus Phrase (FocP), 5, 107, 111, 113, 125, 126, 128, 132, 134, 137, 138, 146–150, 152, 153, 160, 161, 168, 169, 171, 173, 174, 233, 290–292, 294, 302 Force Phrase (ForceP), 110, 203 Former, 8, 11 Functional Projection (FP), 3, 69, 98, 102–105, 109–111, 113, 117, 119, 121, 124, 127, 133, 135, 150, 152, 153, 161, 218, 222, 233, 246, 267, 277, 302–304, 308, 309 G Giusti, Giuliana, 1, 97, 98, 110, 128, 129, 133, 149, 241, 285, 286, 292, 302 Givón, Talmy, 97 Grammaticality judgments, 76, 87, 174, 187, 204, 238 gradience/gradient, 201, 232, 301 Greenberg, Joseph H., 1, 67, 97, 103, 104 Grimshaw, Jane, 181, 233, 292 Gundel, Jeanette, Nancy Hedberg, and Ron Zacharski, 253 Guttièrez-Rexach, Javier, and Enrique Mallen, 108, 110 H Hawkins, John A., 1, 67, 87, 255 Heim, Irene, 244, 247, 288–290 Heim, Irene, and Angelika Kratzer, 53, 244, 245, 247 Hetzron, Robert, 98 Heycock, Caroline, and Roberto Zamparelli, 135 Huddleston, Rodney, 224 Huddleston, Rodney, and Geoffrey Pullum, 116, 119, 120, 224, 303 Hwang, Shin-Ja Joo, 20, 24 I Ihsane, Tabea, and Genoveva Puskas, 110, 304 Intonation (or Intonational Break), 75, 79, 81, 87, 94, 117, 121, 145, 161, 162, 166, 233

Index Invisible, 44, 99 Ionin, Tania, 251 Ionin, Tania, Soondo Baek, Eunah Kim, Heejeong Ko, and Kenneth Wexler, 251, 306 Isac, Daniela, 98 J Jacobs, Roderick, and Peter Rosenbaum, 3 Jo, Mi-Jeung, 252 Joo, Kum-Jeong, 87 Julien, Marit, 124, 269 K Kamp, Hans, 32, 61 Kamp, Hans, and Barbara H. Partee, 7, 305 Kang, Arum, 219, 266, 305 Kang, Soon Haeng, 5, 20, 23, 37, 45, 48, 52, 68, 70, 71, 78, 83, 89, 101, 164 Kaplan, David, 231 Kariaeva, Natalia, 110 Kayne, Richard S, 3, 110, 121, 292, 303 Keller, Frank, 181, 301 Kennedy, Christopher, and McNally Louise, 42 Kim, Jaieun, 233, 303 Kim, Young-Joo, 288 Kind Phrase (KiP), 108 Kiss, Katalin É, 291, 292 Koopman, Hilda, and Anna Szabolcsi, 105, 122 Ko, Youngkeun, 54 Kuno, Susumu, 255, 288 Kuroda, Shige-yuki, 288 L Laenzlinger, Christopher, 1, 3, 5, 97, 98, 102, 108–111, 113, 117, 122, 126, 127, 133, 146, 283, 290, 301, 304 Lakoff, George, 3 Lakoff, Robin, 252, 263, 305 Larson, Richard, K., 1, 3, 5, 6, 37, 44, 67, 97, 99, 101, 105, 113, 134, 137, 146, 176, 178, 183, 238, 304 Larson, Richard K., and Franc Marušič, 44 Larson, Richard K., and Naoko Takahashi, 84–86, 97, 176, 178, 238 Larson, Richard K., and Sungeun Cho, 38 Lee, Chungmin, 87 Lee, Hanjung, 7, 18 Lee, Hyo Sang, 54 Leu, Thomas, 124, 237, 269–272, 277, 304 Liberman, Mark, 209, 218, 252, 305 Lin, Jo-Wang, 237, 242–244, 246, 247, 250

Index Lin, Jo-Wang and Wei-Tian Dylan Tsai, 237, 241–243, 246, 247, 250, 303 Locative Phrase (LocP), 5, 124, 125, 128, 129, 135–139, 143, 146, 148–150, 152, 159, 161, 163, 168, 170–173, 180, 193, 222, 225, 227, 229–233, 249, 267, 270, 274, 302, 309 Longobardi, Giuseppe, 244 M Maclaran, Rose, 251 Maling, Joan, and Soowon Kim, 4, 24 Marantz, Alec, 104, 105 Martin, Samuel E., 4, 18, 20, 24, 46 Marušič, Franc, and Rok Žaucer, 237, 271, 272, 274–276 Matushansky, Ora, 244, 245 Meaning Postulate (MP), 8, 27, 28, 38, 45 Merge, 4, 6, 98, 100, 107, 109, 113, 116, 120, 123, 124, 126, 130, 131, 137, 138, 142, 144, 148, 152, 153, 161–164, 166, 169, 170, 172, 174, 182, 208, 222, 224, 227, 229, 230, 232, 241, 245, 252, 267, 274, 277, 279, 286, 301, 303, 304, 309 DP-internal vs. DP-external, 117, 120, 201, 224, 277, 279–281, 290, 292, 302, 304 Minkoff, Seth, 233 Mirative, 219, 305 Modification absolute vs. relative/comparative, 2, 71, 97, 287, 290, 301 appositive, 119, 133, 172, 210, 211, 222, 227 attributive, 98, 99 attributive vs. predicative, 3, 9, 10, 13, 29, 40, 68, 70, 77, 81, 93, 161–163, 170, 276, 285, 307 complex (morpho-syntactically) vs. simplex, 2, 4, 33, 40, 59, 61, 68, 70, 71, 75–78, 80, 82, 83, 90, 94, 98, 100, 101, 160, 164, 183, 290, 302 direct vs. indirect, 3, 5, 6, 70–72, 89, 97, 99, 100, 134, 138, 283, 287, 290 discourse-linked (or discourse-prominent), 108, 286 generic vs. non-generic, 3, 10, 44, 55, 72, 101, 137, 140 identificational, 309 individual-level vs. stage-level, 70, 84, 86, 98, 176, 178, 179, 238, 284, 286 intensional vs. extensional, 3, 61, 101 intersective vs. non-intersective, 3, 7–11, 31, 38, 43–45, 52, 54, 56, 57, 59, 60, 62, 70, 71, 98, 99, 105, 131, 287, 302, 304

319 N-dependent vs. N-independent, 3, 8, 32, 101, 104 NP-inner vs. NP-outer, 6, 101, 134, 137, 176, 304 possessive, 5, 18, 19, 22, 38, 125, 133, 146 reference-modifying vs. referent-modifying, 3, 10, 12, 70, 72, 99, 100, 134, 135, 138, 142, 144, 146 restrictive vs. non-restrictive, 6, 9, 11, 98, 108, 118–120, 129, 131, 133, 142, 146, 148, 152, 163, 165, 166, 193, 207, 210, 219, 222, 226, 233, 239, 241–243, 245–247, 250, 277–279, 282, 284, 295, 303, 306, 307, 309 speaker-oriented, 309 specific vs. non-specific, 55, 126, 135, 136, 139, 144, 149, 150, 164, 180, 211, 254 subsective vs. non-subsective, 3, 7, 8, 10, 27, 28, 32, 37, 38, 43, 45, 52, 57, 60, 61, 71, 105, 131, 134, 137, 287, 289, 302, 304 temporary- vs. enduring-property-denoting, 44, 52, 99, 134, 140 Mok, Jung-Soo, 20, 33, 35 Montague, Richard, 32, 61 Morzycki, Marcin, 9, 42, 57, 305 Move focus movement, 108, 124, 138, 145, 148–150, 152, 153, 162, 168, 169, 171, 188, 190, 192 leftward movement, 121, 123, 303 remnant movement, 149, 150, 153, 168, 281 roll-up movement, 105, 107, 109, 123, 143 snowball movement, 109, 111, 123 N Nowak, Anita, 62 Number Phrase (NumP), 98, 105–107, 127 Numeral, 1, 67, 88, 98, 104, 114, 121, 125, 128, 133, 135, 138, 144, 251, 306, 307 O Optimality Theory (OT), 7, 160, 180, 201, 206, 301 OT constraints or constraint ranking, 179–181, 183–188, 192, 202, 233 Oshima, David Y., and Ellen (Eric) McCready, 253, 254, 255, 263 P Parameter, 147 Parsons, Terence, 32 Partee, Barbara, 7, 53, 57, 61–63, 115, 304–305

320 Pause, 75, 87, 92, 95, 117, 119, 121, 122, 133, 145, 162, 164, 165, 169, 172, 181, 192, 207, 218, 224, 227, 232, 242, 243, 274, 278, 279, 282, 303, 309 Pearson, Matt, 111 Pereltsvaig, Asya, 62 Plank, Frans, 3 Point of View (POV), 233, 255, 257, 259, 260, 262, 263, 303 Postnominal, 3, 9, 10, 18, 44, 111, 118, 122, 134, 139, 147, 148, 278, 280, 282–284, 286, 290, 304, 311 Potts, Christopher, 6, 117, 209, 224, 233, 241, 252, 263, 303, 309 Potts, Christopher, and Florian Schwarz, 209, 252, 263, 305 Prenominal, 3, 4, 18, 19, 111, 118, 134, 139, 147, 278, 281, 283, 284, 286, 287, 303–305, 311 Prince, Alan, and Paul Smolensky, 180 Prince, Ellen, 252, 305 rinciple of Economy of Lexical Insertion (PELI), 129, 135, 136, 150, 163, 166, 172, 241, 246, 250, 282, 286, 302, 303, 308, 309 Probe vs. Goal, 123 Proper name, 6, 43, 51, 57, 210, 211, 241, 242, 244, 245, 248, 250, 282, 294, 303 Prosody (or Prosodic Unit), 218, 221 R Ramaglia, Francesca, 1, 3 Relative Clauses (RCs) integrated, 92, 116, 118, 120, 122, 131, 143, 222, 224, 227, 241–248, 250, 294, 302–304, 309 languages English, 1, 117, 118, 122, 142, 144, 146, 280, 281, 309, 310 Korean, 5, 19, 20, 54, 57, 59, 78, 80, 83, 84, 86, 118, 129, 163–165, 175, 176, 179, 194–201, 203–207, 224, 243, 250, 277–281, 283, 303, 306, 307, 309 Mandarin, 2, 6, 7, 84, 86, 129, 237, 241–246, 248, 250, 278–283, 303, 309 reduced relative clauses (RRCs), 4, 20, 68, 98, 101, 275 restrictive vs. non-restrictive, 6, 57, 60, 118–120, 129, 131, 133, 143, 146, 163, 165, 166, 168, 169, 193, 207, 210, 219, 222, 226, 233, 239, 241–243, 245–250, 277, 279–287, 295, 303, 306, 307, 309 supplementary, 6, 92, 116, 117, 120, 132, 144, 210, 224, 226, 233, 237, 241, 242, 250, 277–283, 303, 309

Index Rijkhoff, Jan, 104, 125, 132 Rizzi, Luigi, 110, 122, 203, 233, 291, 292 Roehrs, Dorian, 269 S Scalar, 219, 221 Scott, Gary-John, 3, 97, 98, 102, 105, 109 Scrambling, 17, 152, 172, 173, 229, 232 Sekerina, Anna Maria, 62 Sells, Peter, 233 Sharvit, Yael, and Penka Stateva, 288 Shimoyama, Junko, 288, 290 Siegel, Muffy Emily Ann, 53 Siewierska, Anna Maria, 62 Simpson, Andrew, 252 Singleton set, 244, 282, 303 Sino-Korean (SK), 21, 33, 39, 60, 90 Slavic NP-split phenomenon, 62 Smith, Carlota, 3 Smolensky, Paul, 180 Sohn, Ho-Min, 17, 21, 33, 39, 46, 48, 250 Sorace, Antonella, and Frank Keller, 301, 181 Speas, Margaret, and Carol Tenny, 233, 303 Sproat, Richard, and Chinlin Shih, 2, 5, 23, 67–69, 80, 97, 101, 107, 134, 138, 146, 183, 301, 302 Stirling, Lesley, 233 Suh, Eugenina, 252 Suh, Jeong-Soo, 4, 20, 24, 33 Suh, Kyung-Hee, 305 Svenonius, Peter, 1, 3, 5, 19, 74, 82, 97, 102–107, 113, 117, 122, 124, 127, 129, 131, 146, 150, 286, 290, 301, 304 Syllogism failure, 38 Szabolcsi, Anna, 105, 129, 288, 290, 292 T Tenny, Carol, 233, 303 Tense/Aspect/Mood (TAM) (marking), 17, 20, 47, 54, 55 retrospective, 200, 205 Tomaszewicz, Barbara, 288–290 Topic, 194, 202, 213, 230, 286, 288, 290, 291 discourse topic, 194–198, 203, 230, 231, 286, 290 topic phrase, 203, 233 Truswell, Robert, 108 U Unit Phrase (UnitP), 5, 104, 105, 125, 133, 138, 140, 144, 159, 166, 168, 169, 172, 302

Index V Vangsnes, Øystein A., 110, 269 Van Valin, Robert D. Jr., 57 Vendler, Zeno, 2, 97 Von Heusinger, Klaus, 252 W Whorf, Benjamin Lee, 2, 3, 68, 69 Williams, Edwin, 114, 115 Wolter, Lynsey K., 255, 304, 305

321 Y Yu, Hyen-Kyeng, 4, 22, 23, 34, 48 Z Zamparelli, Roberto, 108 Zribi-Hertz, Anne, 233 Zubizaretta, María Luisa, 152