'NOT' IN THE MOOD - The University of Texas at Arlington

THE UNIVERSITY OF CHICAGO

‘NOT’ IN THE MOOD: THE SYNTAX, SEMANTICS, AND PRAGMATICS OF EVALUATIVE NEGATION

A DISSERTATION SUBMITTED TO THE FACULTY OF THE DIVISION OF THE HUMANITIES IN CANDIDACY FOR THE DEGREE OF DOCTOR OF PHILOSOPHY

DEPARTMENT OF LINGUISTICS

BY SUWON YOON

CHICAGO, ILLINOIS MARCH 2011

TABLE OF CONTENTS

List of Tables…………….………………………………………………………………………vi Abbreviations………………………………………………………………………………vii Abstract………………………………………………………………………………………viii Acknowledgements…………………………….……………………………………………….xii Chapter 1. Introduction and Theoretical Backgrounds.………………….....……………...…1 1.1. Hallmark properties of EN……………………………………………………...………….…1 1.2. Dependency on Nonveridicality: Giannakidou 1994 et seq………………………………...14 1.3. Evaluative Component on a separate dimension: Potts 2005; Giannakidou & Yoon 2009...15 1.4. Roadmap…………………………………………………………………………………….18 Chapter 2. Empirical description of Evaluative Negation...…………………………………21 2.1. EN in Verbs of Fear……………...………………………………………………………….22 2.2. EN in Verbs of Hope (New environment)...………………...………………………………26 2.3. EN in Exclamatives……………...……………………………………………………..……29 2.4. EN in Emphatic questions……………...……………………………..…………..…………37 2.5. EN in Concessive conditionals……………...………………………….…………...………38 2.6. EN in Before(without/unless)-clauses……….……........……………………………………39 2.7. EN in Until(/since)-clauses……………...…………………………………………………..44 2.8. EN in Polite requests……………...…………………………………………………………48 2.9. EN in Comparatives……………...……………………………………………………….…51 2.10. Remaining question: EN in biased questions and tag-questions..…………………...…54 2.11. Summary……………...……………………………………………………………………59

ii

Chapter 3. Previous Approaches………..…………………………….…………………….…62 3.1. Expletive approaches………………………………………………………………..….…63 3.2. Problems for expletive approaches…………………………...………………………….….69 3.3. Non-expletive approaches: EN as regular negation……..………………..…………………70 3.3.1. EN marking temporal sequence…………………………………………...………70 3.3.2. EN marking negative implicature………………………………………………....72 3.3.3. EN marking an unlikelihood…………………………………….…………...……74 3.4. Problems for non-expletive approaches…………………………...………………….……..75 3.5. EN in non-negative nonveridical contexts: problem for all current theories………..………76 3.5.1. Interlinguistic variation……………………………………………………………76 3.5.2. Intralinguistic variation……………………………………………………………78 3.6. Summary…………………………………………………………………………………….81 Chapter 4. The Licensing of EN……………………………...…………………………..……82 4.1. EN as a subcase of subjunctive marker……………………………………………..…….…84 4.2. Conceptual connection between EN and subjunctive mood I: nonveridicality…………..…87 4.3. Conceptual connection between EN and subjunctive mood II: polarity………….……........93 4.4. EN is a subcase of subjunctive marker in Korean and Japanese………...……………...…103 4.4.1. Likelihood degree of contexts…………………………………………………105 4.4.2. Non-factive complementizer ……………………………………………………108 4.4.3. Nonveridical predicates….………..……………………………………………110 4.5. Summary………………………………….……………………………………..………....113 5. The Contribution of EN.………………………..…………………………………….…114 5.1. Difference between Subjunctive and EN: positive vs. negative anticipation …………..…114

iii

5.2. The Epistemic/Buletic Model of EN……………………………………………..………...123 5.3. The Evaluative Dimension……………...………………………………………………….128 5.3.1. The Evaluative Dimension of EN…..……………………..………………..….129 5.3.2. Expressivity in EN………..…………………………………………………..….132 5.3.3. CI Logic for EN………………………...………………….…………………….137 5.3.3.1. Typological Split in the Evaluative meaning of subordinate EN….......138 5.3.3.2. EN in Exclamatives: undesirability or unlikelihood…………….….….142 5.3.3.3. EN in Polite requests: unlikelihood……………..…………………......145 5.3.3.4. EN in Temporal connectives: undesirability or unlikelihood….……....146 5.3.3.5. EN in Comparatives: undesirability or unlikelihood.……….. ………..148 5.3.3.6. Multidimensionality shown in multiple CIs.………..……… …….…..150 5.4. Summary………………………………………………………………………………...…151 Chapter 6. The Syntax of Evaluative Negation ……………………….…………….………153 6.1. Parataxis versus Hypotaxis……………………………………………………………...…154 6.1.1. Property of paratactic (root) dependencies ……………………………….......…155 6.1.2. Hypotactic Properties……………………………………… …………….…...…165 6.1.3. Combined applications of hypotactic and paratactic diagnostics…………....…..172 6.2. Syntactic Derivations…………………………………………………………….……...…176 6.3. Subjunctive mood and Evaluative feature in Korean (and Japanese)…..…….….……...…187 6.3.1. Complementizer choice as mood choice………………………….………...……187 6.3.2. Matrix predicate & Subjunctive morphology…………………….………...……194 6.3.3. Evaluative Negation with evaluative feature…………………….………...…… 200 6.3.4. A potential account: multiple Eval-Heads……………………….………...…… 202

iv

6.4. Summary and Conclusion….…………………………………………….……………...…206 Chapter 7. A case study of Rhetorical Comparatives: PI, mood, and EN…………….…..208 7.1. Debate on the negativity of comparatives…………………………………...……………..209 7.1.1. Negative analyses and their limitations………………………….........................209 7.1.2. Non-negative analyses and their limitations ……………................................….213 7.1.3. Main ideas to be proposed: a split analysis………………………………………217 7.2. Rhetorical comparatives……………………………………………...…………………….218 7.2.1. Empirical description of the rhetorical effect of an NPI in the comparative…….218 7.2.2. Analysis: RC as a product of non-referential standard and NPI use…………….228 7.2.2.1. Giannakidou 1998 ……………………...………………………...……228 7.2.2.2. Den Dikken & Giannakidou 2002 ……………………...……………..231 7.2.2.3. Putting it together ………………...……………………...……………236 7.3. RCs are not MCs…………...……………………...……………...………………………..242 7.4. Rhetoricizing the standard: NPIs, expletive negation, and the subjunctive…………….….245 7.4.1. Negative polarity items …………...……………...……………………………...245 7.4.2. Evaluative negation ……...……………...…………………………………….…249 7.4.3. Subjunctive mood …..……………………………………...…………………....254 7.5. Summary and Conclusion…..………………...…………………………...……………259 Concluding Remarks……………….……………….……………….……………….……….262 Bibliography……………….……………….……………….……………….………………...267

v

LIST OF TABLES

Table 1. Comparative Distribution of EN and Subjunctive……….……………………………….60 Table 2. Six classes of Evaluative Content of EN (tentative version) …………………...………61 Table 3. Comparative Distribution of EN and Subjunctive (in part, chapter 2.)…. ……………...83 Table 4. Complementary distribution of EN vs. Non-EN w.r.t. the likelihood-degree…………108 Table 5. Comparative Distribution of EN and Subjunctive (repeated from Table 1) ………..…120 Table 6. Interim summary: paratactic vs. hypotactic properties of various complements……...171 Table 7. Selection relation between subjunctive verbs and complementizers…………………..195 Table 8. Selection relation between indicative verbs and complementizers……………………196 Table 9. Morphosyntactic categories and polarity sensitivity (D. Levinson 2006) …………….255 Table 10. Six classes of Evaluative Content of EN (final version)…………………………….. 263

vi

ABBREVIATIONS ACC ADN ASP COMP COND CL CT DAT DECL EMPH EVAL P FEM FOC FUT GEN HON IMP IND LOC MOD NEG NF COMP NMLZ NOM OPT PERF PH PL PNP PRES PRT PST REFL REL Q S.CL SG SUBJ TOP

accusative case marker adnominal aspect marker complementizer conditional classifier contrastive topic marker dative case marker declarative sentence marker emphatic marker evaluative phrase feminine focus marker future tense genitive case honorific imperative marker indicative mood locative case modal negation non-factive complementizer nominalizer nominative case marker optative marker perfective aspect place holder plural perfective non past present tense partitive past tense reflexive relativizer question marker subject clitic singular subjunctive mood topic marker

vii

ABSTRACT

‘Not’ in the Mood: THE SYNTAX, SEMANTICS, AND PRAGMATICS OF EVALUATIVE NEGATION

November 2010 Suwon Yoon

The primary goal of the present study is to gain more insight into the phenomena of Expletive Negation. Contrary to what the traditional term “expletive negation” would make us believe, I propose an analysis of this particular type of negation as having semantic content that consists in two components: First, “expletive negation” is subject to licensing on a par with polarity items and mood markers. It thus manifests dependency to nonveridicality; Second, it also has pragmatic contribution. It triggers what we can think of as evaluative mode of negating. Therefore, I will use the abbreviation EN in this dissertation as Evaluative Negation. I will make my case by considering a broad set of data ranging from English to Korean. The Korean data will be particularly important in bringing about the argument that EN functions as a mood marker. Chapter 1 starts with the observed hallmark properties of EN and theoretical backgrounds. In chapter 2, I show the pragmatic contribution of two scalar meanings of undesirability and unlikelihood. It is further shown that the base of scale denoted by EN may vary depending on the context or the epistemic subject’s (the speaker or the matrix subject) emotional state, perhaps reflected in the tone of voice. In doing so, I propose that EN is not an imperfection or illogicality

viii

of language. Rather, EN represents another legitimate function of negation in natural language, where a negative element is adopted for the purpose of circumventing a commitment to a truthful statement and combines this with an attitude. This makes it similar to the subjunctive mood. Chapter 3 offers a critical review of two predominant camps on theories of evaluative negation – expletive approaches and non-expletive approaches. Taking more data into consideration, I provide empirical and conceptual arguments as to why prior theories require revision. First, it is shown that, contra the assumption of expletive approaches, EN does indeed have non-vacuous semantic contribution – evaluative sense that is represented as various kinds of scalar meaning. Second, the non-expletive approaches are unable to predict the variability that we find within a context and within a language. In chapter 4 I propose my analysis of EN as a mood marker. EN is a subjunctive mood marker, in particular, following the pattern of the mood choice in Greek that is regulated by nonveridicality (Giannakidou 2009). I discuss the parallels in the positive nonveridical environment in the scope of verb ‘hope’. This is unexpected under any account that we have seen, and the fact that EN requires a non-factive complementizer and a subjunctive predicate in Korean and Japanese should come as no surprise. In chapter 5, I add a pragmatic component to the semantic analysis. I argue that EN contributes an evaluative dimension. This evaluative dimension is negative anticipation, undesirability or low likelihood. I show that EN in different environments exhibits a striking parallel in interpretation as an ordering relation. I propose to capture this by means of multidimensionality of conventional implicature (Potts 2005) (as refined in Giannakidou & Yoon 2009, to appear), and show how evaluative semantics of EN exists on a separate dimension from the semantic core of utterance. Contrary to previous views that dismiss the meaning of EN,

ix

calling it expletive, my proposal captures the precise role of EN in an utterance. If my analysis is correct, it has one important implication: It allows the generalization that various subspecies of EN in language are indeed part of the grammar. In particular, they are reflexes of grammaticalization of perspective and subjective mode, on a par with subjunctive mood choice and other subjunctive phenomena like metalinguistic comparatives as argued in Giannakidou and Yoon (2009, to apper). In chapter 6, I explore the syntactic configurations of subordinate subjunctive clauses and subordinate EN-clauses in Korean (and potentially Japanese). One of the questions that will arise is whether the matrix clause and the EN-clause in Korean are hypotactically or paratactically connected. This will inevitably involve the Germanic Embedded Verb Second (EV2). The dynamic syntax of EN (and subjunctive) complements that I propose affords the following theoretical implications: First, by revealing the parallels with EV2 constructions that are in subjunctive, the syntax of EN constructions further supports the current proposal that EN is a subjunctive marker. Second, it suggests that the syntax of EN complements in Korean can be understood along the lines of Jespersen’s insight of paratactic negation in Old/Middle English. Third, it accounts for why the complementizer for subordinate EN clauses is in an analogous form with a question particle in Korean and Japanese. Finally, it furthermore supports the possibility that negation in tag-questions, which are classical paratactic configurations, can be understood as a subspecies of EN. Finally, in chapter 7, I further strengthen the connection between mood and EN by looking at rhetorical comparatives. My overall approach, if correct, has two important consequences: First, we have explanations for the crosslinguistic variation in subordinate EN between Japanese/Korean and

x

other languages. Second, it offers a plausible way of systematically characterizing EN in other environments such as exclamatives, questions, nonveridical before clauses and comparatives.

xi

ACKNOWLEDGMENTS

Special thanks go to my supervisors Jason Merchant and Anastasia Giannakidou, and the members of my thesis committee Christopher Kennedy and Marcel den Dikken. I cannot even begin to describe how deeply I appreciate their wholehearted support and their confidence in my abilities for all these years of my Ph.D. I thank Jason Merchant for understanding the importance of the new sets of data in Korean and Japanese that I came up with and having numerous meetings for discussing them with me until they grew into a full blown analysis. My works on syntax would have not existed without his surprisingly profound knowledge of East Asian literature and inspiration on the data, besides the general syntactic theories. From the beginning it was his encouragement and constant support that gave me the courage to pursue my ideas and take a firm stand. I also thank Anastasia Giannakidou for so generously offering to collaborate on a number of projects with me, which crucially show the striking parallels between the two genetically unrelated languages, Greek and Korean. I thank her for patiently following the development of my thinking, even though, I suspect, she saw the consequences of my ideas far earlier than I did. I admire her enthusiasm and positive energy which have helped me to be passionate about my works and keep being productive. I am very grateful to Christopher Kennedy for his constant support and also for trusting my abilities and offering me an opportunity to work as a researcher for the NSF Comparative Project. The weekly discussion sessions and the individual projects that were assigned for me, inspired me to pursue a number of interesting and important ideas for my own projects, including the ideas for my thesis. I truly enjoyed the times I discussed various topics with him and my colleagues. Also I am thankful that I could invite Marcel den Dikken in my thesis committee. I appreciate that his insightful

xii

comments considerably improved my analysis especially in the syntax chapter of this thesis. I also thank him for collaborating on the NPI project, offering me the opportunity to review articles for ICSH 8, inviting me to give a talk at Syntax Supper at CUNY, and also for his support and advice about the field. I thank the University of Chicago faculty, John Goldsmith, Salikoko Mufwene, Michael Silverstein, Amy Dahlstrom, Alan Yu, Karlos Arregi, Jason Riggle, Ming Xiang, and especially Jerry Sadock for all they taught me along the way. Looking back, I realize how extremely fortunate I was to be able to be surrounded and cared by these wonderful people. I would like to thank Larry Horn for his interest in my own works and co-authored works with Anastasia Giannakidou. I am very thankful for his insightful comments and it is truly honorable to me that he mentioned my works in his new books and recent papers. I also thank Josep Quer for his interest in my work and the fact that I have learned a lot about mood from his work. I thank Kyle Johnson, Peter Svenonius, Shigeru Miyagawa, Paul Portner, Anthony Kroch, Sandra Chung, Hagit Borer, Klaus von Heusinger, Ian Roberts, Ivano Caponigro, and Jaroen van Craenenbroeck, for their supports and comments on my works. Also I am grateful to Eun-Jung Yoo, Jong-Bok Kim, Dong-Whee Yang, Keun-Won Sohn, Chungmin Lee, Seungho Nam, Jongho Jun, Sang-Chul An, Ju-Hee Lee, Min-Joo Kim, Hee-Jeong Ko, Kook-Hee Gil, and EunHee Lee for the discussion and comments on my works and Korean data. I am also grateful to Osamu Sawada, Masaya Yoshida, Ichiro Yuhara, Mika Ishino, Keiko Yoshimura, Mayu Yoshihara, Satoshi Oku, Koji Kawahara, Yukio Furukawa, Yusuke Kubota, Reiko Vermeulen, and Naoko Tomioka for their discussion on Japanese data and Elena Castroviejo Miró on Catalan data.

xiii

I would like to thank my colleagues at University of Chicago. First, I benefited greatly from weekly discussions with researchers in the comparative group, Osamu Sawada, Thomas Grano, Peter Arrenga, and Jacqueline Bunting. I also thank CLS 43 officers, Malcolm Elliott, James Kirby, Osamu Sawada, and Eleni Staraki for the joyful year of planning and organizing the conference together. Also I am glad that I could share the extraordinary experience of writing up a thesis containing hundreds of pages with my cohorts in the writing group, Nikki Adams, Jacqueline Bunting, Chrisopher Straughn, and Kjersti Stensrud. They made the otherwise agonizing procedure a rather pleasant experience. I also thank my Korean colleagues, JungHyuck Lee for letting me teach Korean at the language program, and Hae-Jin Han, Young-Lee You, Eun-Hae Park, and Arum Kang for their support. Finally, I thank the ICGL 9 co-editors, Aikaterini Chatzopoulos and Alexandra Ioannidou, for understanding and being patient about my hectic schedule. I very much appreciate their support and friendship. I would like to thank the audience of GLOW 32 at Nantes, Thomas Ede Zimmermann, Louise McNally, Berit Gehrke, Daniel Gutzmann, and Anamaria Falaus, where part of my thesis was presented. I am grateful for their insightful comments and questions which helped greatly in developing the idea into a bigger project. I also thank the audience of Japanese/Korean Linguistics 18, Peter Sells, Maria Polinsky, Satoshi Tomioka, Junko Shimoyama, Takubo Yukinori, Shigeto Kawahara, Bjarke Frellesvig, Min-Joo Kim, Sung-Ock Sohn, Shin-Sook Kim, Minjeong Son, Ji-Young Shim, and thank Marcel den Dikken for organizing the conference. The comments and new data brought up during the discussion significantly improved my analysis. I would like to thank the faculty of Seoul National University, who has taught me the basics and awaken my interest in syntactic and semantic issues: my MA advisor Eun-Jung Yoo, and committee members Ki-Sun Hong and Chang Yong Sohn, and my teachers Heok-Seung

xiv

Kwon, Jae-Young Lee, Mi-jeong Song, Yong-Yae Park, and Hyopil Shin. I am grateful for my colleagues of MA program: Yun-Ah Hwang, Ee-Jin Cho, Soo-Ran Park, Yun-Kyung Cho, NaJung Kim, Young-Il Oh, and the co-editors of SNU Working Papers in English Language and Linguistics 2: Chung-gon Cho, Hye-sun Cho, Shi-Na Lee, and Jung-Ah Shin. My first step into the field was thanks to the faculty of the University of Seoul, SangSoon Lim, In-Han Jun, Jong-Sung Lim, Yung-Kyun Ryu, and my colleagues Kyoung-Chun Chang, Tae-Jin Yoon, who guided and supported me ever since. I thank my cohorts Na-Young Song and Oh-Hyang Kwon for the friendship and support. During my Ph.D. years, every summer I visited Seoul not just to see my family and friends, there was another reason that I was excited. I have a very fond memory of discussing my works and the parties afterwards were always enjoyable at SICCOG. I thank Myoung-Kwan Park, Gui-Sun Moon, Sun-Woong Kim, Young-Wha Kim, Kwang-sup Kim, Hee-Don Ahn, Hang-Jin Yoon, Jong-Yurl Yoon, Sun-Ho Hong, Rang-Hye-Yun Kim, Jinhee Suh, Sungeun Cho, Seungwan Ha, So-Young Park, and Hiroshi Aoyagi, among many others. Many of my conference trips were made possible by travel grants by Doolittle-Harrison Fellowship, CLS, Division of the Humanities, and Center for East Asian Studies. I also thank the organizers and audience of CLS 46, the 9th International Conference on Greek Linguistics (ICGL9), the 13th Sinn und Bedeutung, Purdue Linguistics Association Symposium 2007-8, the 18th International Congress of Linguists (CIL 18), the 5th Workshop on Altaic and Formal Linguistics (WAFL 5), the Northwest Linguistics Conference, the 6th Workshop in General Linguistics at UW-Madison, International Conference of Cognitive Science (ICCS 6), the 16th International Circle of Korean Linguistics (ICKL 16), LSA Annual Meeting 2007-8, the 31st Annual Penn Linguistics Colloquium (PLC), Seoul International Conference on Generative

xv

Grammar (SICOGG 7,9,10), On Linguistic Interfaces (OnLI), and the 52nd Annual Conference of the International Linguistic Association (ILA). For the financial support, I thank Mrs. Giles Whiting Foundation, The Andrew W. Mellon Foundation, the division of the Humanities and the provost at University of Chicago, for their generous financial support of Mellon Graduate Achievement Scholarship, Whiting Dissertation Write-up Fellowship, University fellowship and Provost's Summer Fellowships. I also thank the faculty of Stanford University for the Fellowship for the 2007 Linguistic Society of America (LSA) Summer Institute. I would like to express my deepest gratitude for the faculty of the department of Linguistics at Cornell University, Draga Zec, Molly Diesing, John Bowers, John Whitman, Dorit Abusch, Wayles Browne, John Hale, Abby Cohn, Wayne Harbert, Sue Hertz, Sarah Murray, Alan Nussbaum, Mats Rooth, Carol Rosen, and Michael Weiss, for having the confidence in my abilities and offering me Visiting Assistant Professorship. This opportunity is extremely meaningful to me and I am thrilled to be a part of such an excellent collection of faculty.

Finally, I thank my parents, my father Seok-Ki Yoon and my mother Nam-Soon Park, for pretty much everything—for unconditionally loving me and being very proud of me. I thank my brother and sisters for understanding my situation and being with my parents. I thank my beaux parents for the encouragements and my soul mate Seung Min for always being there for me.

xvi

Chapter 1. Introduction and Theoretical Backgrounds

The primary goal of the present study is to gain more insight into the phenomena of what has been called expletive negation.1 Contrary to the traditional term ‘expletive’ negation, I propose that this particular type of negation in a variety of contexts has semantic content that can be analyzed as two dimensions: (i) with regard to the licensing, it reveals semantic dependency on nonveridicality that is parallel to polarity items;2 (ii) with regards to the semanticopragmatic contribution, it triggers an evaluative sense, just like certain uses of subjunctive mood. In this light, the abbreviation EN in this dissertation will refer to Evaluative Negation.

1.1. Hallmark properties of EN Observing a parallel pattern between the data in Modern French (Grévisse 1980; Muller 1991; Larrivée 1994) and data in 17th century Dutch (from the Dutch writer Vondel’s work), van Helten (1883; van der Wouden 1994, 1997) provides the following generalization for EN:

Jesperson (1917, 1924) and van der Wouden (1994, 1997) use the term Paratactic Negation (PN), instead of expletive negation. The discussion on paratactic properties of EN constructions is offered in chapter 6. 2 The dependency on nonveridicality of EN has been also noted by Espinal (1997, 2000) in Romance languages. However, the way she undertakes this property for accounting for the phenomena of EN is fundamentally different from the current analysis: within her system of negative absorption, the nonveridical feature of EN is somehow recast as a negative feature which must be checked off by a Neg head. Thus her account falls under the category of negative concord analysis of EN. A detailed discussion on previous accounts will be given in chapter 3. 1

1

(1) Generalizations for EN a. EN is never obligatory. b. EN often co-occurs with subjunctive, conjunctive and other moods that are typically used to express counterfactuals, unreal hypothetical conditionals, etc. c. EN occurs after words expressing fear. d. EN may be triggered by words expressing hindrance, precaution, and prevention. e. EN is absent after words of dubitation. (see (4/5)) f. EN is found in various types of comparative constructions. (see (2)) g. EN may also occur in subordinate constructions governed by ‘conjunctive’ elements such as (French) avant que ‘before’, sans que ‘without’, à moins que ‘unless’, etc. (see (3))

(2) a. Il est autre que je ne croyais.

[French]

he is other than I not believed.Subj ‘He is different than I thought.’ b. Paris était alors plus aimable qu’il

n’est aujourd’hui.

Paris was then more amiable than-it not-is today ‘Paris was more pleasant then than it is today.’ (3) a. Avant qu’il ne fasse froid.

[French]

before that-it not gets cold ‘Before it gets cold.’ b. Le lieutenant répondit […] au salut sans qu’un muscle de sa figure ne bougeât. ‘The lieutenant answered the salute in a military way without moving a muscle in his face.’

2

(4) a. Je doute fort

que cela soit.

[French]

I doubt strongly that that be.Subj ‘I very much doubt that that can be so.’ b. Il nie

que ce soit

trouvé dans cette maison.

he denies that it be. Subj found in

that house

‘He denies that it is to be found in that house.’ (5) In twyffel, of hy met den hals syn’ schuld sou in doubt,

if he with the neck his

debt

boeten.

(Vondel) [17C Dutch]

would pay

‘In doubt if he was to pay for his quilt with his life.’

Note that most of these environments are also environments that typically trigger the subjunctive mood and have been described in the literature as nonveridical. I will also present novel data from Japanese and Korean in which EN-licensing environments are extended to “positive” nonveridical predicates like ‘hope’, thus creating a more visible parallel between EN and subjunctive than the one that we find in Romance languages, where EN is only licensed in a subset of subjunctives. In addition to general properties and distributions of EN above, there are four diagnostics that generally distinguish EN from real negation. I will illustrate these properties by exploring the data in previously studied languages as well as new data in Korean and Japanese. The parallelism in four respects will further confirm that EN in Korean and Japanese is an analogous creature to what has been analyzed as expletive negation or paratactic negation in previous literature.

3

I. No Licensing of strong NPIs Crucial evidence for the status of negation as EN comes from its inability to license a strong NPI. In order to examine this property, varied distribution of NPIs should be established first: for each NPI paradigm, its lexical semantic properties and its morphosyntactic features will determine where precisely, within the nonveridical space, the NPI will appear (Giannakidou and Yoon 2010). For the current purposes, however, the distinction between strong (or strict) and weak NPIs would suffice. First, a class of strong NPIs are ones that are licensed only by negation and antiveridical without and they must be in the scope of an antiveridical operator at LF (Giannakidou 1998, et seq.). In English, the following examples show that only either is a strong NPI while any and minimizers such as give a damn or say a word are not strong.

(6)

a. John didn’t see anybody. b. John doesn’t give a damn. c. John didn’t see Bill, either.

(7)

a. Did John see anybody? b. Did John say a word? c. *Did John see Bill, either?

(8)

a. Does John give a damn? b. If you give a damn, you will listen to me. c. *If you see Bill either, ...

4

The strength of each NPI, however, may vary across languages. In Greek, the emphatic KANENAS ‘anyone’ and minimizers are strong NPIs, just like either, and it is only acceptable with negation (Giannakidou 1998, 1999):

(9) a. KANENAN anyone

dhen idha.

[Greek]

not saw.1sg

‘I saw nobody.’ b. *An erthi if come.3sg

KANENAS... anybody

‘If anyone comes...’ c. *Irthe came.3sg

KANENAS? anybody

‘Did anyone come?’ (10) Dhen dhino djekara not

give.1sg damn

jia to ti th’apojinis. about the what will happen.2sg

I don’t give a damn about what will happen to you! (11) #/*An dhinis dhekara, tha me akousis. (If you dive a damn, you’ll listen.)

The following contrasts reveal that amwuto ‘anyone’ and minimizers in Korean are strong NPIs that can only appear in the scope of negation.

5

(12) a. Na-nun I-Top

amwuto

po-ci anh-ass-ta.

anyone

see-CI Neg-Pst-Decl

[Korean]

‘I didn’t see anyone.’ b. *Manil

amwuto

if

anyone

o-ntamyen... come-Subj

‘If anyone comes...’ c. *Amwuto anyone

o-ass-ni? com-Pst-Q

‘Did anone come?’ (13) a. John-i

kkwumccekto ha-ci

John-Nom budge an inch

anh-ass-ta.

-CI

Neg-Pst-Decl

‘John didn’t budge an inch.’ b. *John-i

kkwumccekto ha-ess-ni?

John-Nom budge an inch

-Pst-Q

‘Did John budge an inch?’ c. *John-i

kkwumccekto ha-ess-ta.

John-Nom budge an inch

-Pst-Decl

‘John budged an inch.’

On the other hand, there are broader types of NPIs such as any and minimizers in English which can be generally sanctioned by questions, imperatives, modals, protasis of conditionals, disjunctions, etc.

6

That said, it is predicted that due to the lack of negative force, EN can never licenses strong NPIs, although weak/broader types of NPIs may still be able to occur in EN-containing sentences in which case the occurrence of NPI has nothing to do with the presence of EN. Weak NPIs must have been licensed by one of the typical environments for EN which corresponds to weak NPI-licensing contexts, i.e. affective contexts such as questions, protasis of conditionals. For instance, Espinal (1997) claims that in Spanish and Catalan the identity of EN becomes clear when it fails to license a strong NPI.

(14) a. *¡Qué barbaridades no cometería nadie así!

[Spanish]

what atrocities not comit.Cond nobody like b. *Quines bestieses no devia dir mai!

[Catalan]

which nonsenses not must.Pst say never

Espinal further shows that EN, in contrast with standard negation, allows positive polarity items under its scope.

(15) a. ¡Qué barbaridades no cometería what atrocities b. Quines bestieses which

not comit.Cond no devia

nonsenses not must.Pst

alguien

así!

[Spanish]

somebody like dir algú

com ell!

say somebody like him

7

[Catalan]

This property is also shown in Korean and Japanese. As illustrated in the following data, the ENreadings in Korean (16) and Japanese (17) ‘(i) John fears/hopes that anyone might come’ become unavailable due to the presence of the strong NPI amwuto and daremo.

(16) John-un

amwuto

John-Top anyone i) EN reading:

oci-anh-ul-ci/kka

{kekcengha/kitayha}-koissta.[Korean]

come-Neg-Fut-NFComp fear/hope-Asp

*‘John {fears/hopes} that anyone might come.’

ii) Real Neg reading: √ ‘John {fears/hopes} that no one will come.’ (17) John-wa John-Top i) EN reading:

daremo

ko-nai-ka(-to)

{sinpaisi/kitaisi}-te iru.

anyone

come-Neg-NFComp

[Japanese]

fear/hope-Asp

* ‘John {fears/hopes} that anyone might come.’

ii) Real Neg reading: √ ‘John {fears/hopes} that no one will come.’

In order for the grammaticality of the sentences to be salvaged, they must be interpreted as involving real negation in (16) and (17) ‘(ii) John fears/hopes that no one will come’ so that the NPI, amwuto in Korean and daremo in Japanese, is properly licensed by real negation. The unavailability of the EN-reading (16i/17i) even under typical EN-triggering contexts (i.e., matrix predicates such as fear or hope) shows that anh or nai can never license such NPIs. Since the inability to license strong NPIs is a well-known property for EN crosslinguistically, the above fact confirms the EN status of anh and nai.

8

II. No Double Negation The most striking contrast between EN and real negation is observed when EN co-occurs with real negation under the interpretation of only one logical negation. In (18), the negative interpretation comes from the first anh (Neg1: real negation) while the second anh (Neg2: EN) is logically vacuous:

(18) John-un John-Top

Mary-ka

anh-oci-anh-ul-kka

kitayha-koissta. [Korean]

Mary-Nom

Neg1-come-Neg2-Fut-NFComp hope-Asp

‘John hopes that Mary might not come (although it is unlikely to happen).’

An analogous fact holds in Japanese negation as illustrated in (19). It is the first nai (Neg1) that operates as negation whereas the second nai (Neg1) is EN void of such negative force:

(19) John-wa

Mary-ga

John-Top Mary-Nom

ko-nai-nodewa

nai-ka(-to)

kitaisi-te iru.

come-Neg1-Comp

Neg2-NFComp hope-Asp

[Japanese]

‘John hopes that Mary will not come (although it is unlikely to happen).’

Note, however, that under normal circumstances double negation is possible in Korean (20) and in Japanese (21), acquiring a positive interpretation because two real negatives cancel each other out.

(20) Mary-ka Mary-Nom

ecey

anh-on-kes-un

anh-i-ya.

yesterday Neg1-come-NMLZ-CT

9

Neg2-be-Decl

[Korean]

‘It is not the case that Mary did not come yesterday.’ ‘Mary did come yesterday anyway.’ (21) Mary-wa Mary-Top

kinoo

ko-nai-koto-wa

na-kat-ta.

yesterday

come-Neg1-NMLZ-CT

Neg2-Pst-Decl

[Japanese]

‘It is not the case that Mary did not come yesterday.’ ‘Mary did come yesterday anyway.’

The use of such double negation tends to bring about concessive semantic effects such as ‘although she arrived very late’, ‘although she left very soon’, ‘although she unwillingly came’, etc., or emphatic effects for strengthening one’s argument. Comparing these double negation data with the above sentences, the expletiveness of the second negative (Neg2) in (20) and (21) is borne out.

III. Prosodic Marking In order to distinguish from real negation, EN tends to be prosodically marked. In Korean, for instance, EN does not receive a prosodic emphasis and oftentimes emerges in a shortened form. This property becomes clear when the above sentences with NPIs are reconsidered with an emphasis on different parts. In (22), only an EN reading is available because a prosodic focus on the preceding embedded verb OCI ‘come’ makes the following negative anh relatively weaker.

(22) *John-un John-Top

amwuto

OCI-anh-ul-kka

anyone

come-Neg-Fut-NFComp fear/hope-Asp

‘John fears/hopes that anyone will come.’

kekcengha/kitayha-koissta. [Korean]

: EN-reading

10

This renders anh expletive in this prosodic environment. However, since EN cannot license a strong NPI, the presence of the NPI amwuto makes the sentence ungrammatical. In contrast, an emphatic negative ANH in (23) manifests itself as a real negation, forcing a real negation reading. Thus an NPI is licensed here. And I assume that Japanese is analogous in this respect.

(23) John-un John-Top

amwuto

oci-ANH-ul-kka

kekcengha/kitayha-koissta.

anyone

come-Neg-Fut-NFComp

fear/hope-Decl

‘John fears/hopes that no one will come.’

: Real Neg-reading

Furthermore, EN in exclamatives tends to come about in a contracted form cyanh instead of a full form ci anh in Korean (24). This example is also important in that, as illustrated inside the parentheses, it reveals a kind of semantic effect that EN can induce, i.e., undesirability. Detailed discussion on various semantic effects of EN in different contexts will be presented in chapter 2.

(24) Ney-ka you-Nom

aisukurim-ul

ta

mekess-cyanh/#ci anh-e!

ice cream-Acc

all

ate-Comp.Neg/Comp-Neg-Decl

[Korean]

‘(I resent that) You ate up all the ice-cream!’

The prosodic weakening and shortening of anh support the current assumption on the analogous status of EN between Korean/Japanese and other languages. Observe that similar effects are

11

noted in Modern Hebrew (Eilam 2006). EN in until-clauses like (25) becomes felicitous only with appropriate intonation, which deemphasizes the negative marker lo.

(25) “Adam še-ne'ešam be-avera plilit xezkato še-hu zakai,

[Modern Hebrew]

ad še-lo huxexa ašmato ka-xok be-mišpat pumbi.” “Everyone charged with a penal offence has the right to be presumed innocent until EN proven guilty according to law in a public trial.” (Article 11 of the Universal Declaration of Human rights)

However, the prosodic marking of EN seems to be expressed in the opposite way in other languages. That is, instead of prosodic weakening above, prosodic strengthening also can be employed for distinguishing EN from real negation. Observe the following examples in English in which EN carries emphasis (thanks to Jason Merchant and Marcel den Dikken).

(26) Aww! You didn’t eat all the ice-cream!

[English]

(27) You don’t say!

IV. Lexical/morphological Distinction Since EN tends to appear in a distinct, typically shorter, morphological form from normal negation, it seems plausible to posit EN as a separate creature from real negation. For instance, EN implements a particle ne-form (as opposed to the full negation form ne…pas or pas) in French (26), lest “that–not” or ne (as opposed to not) in Old/Middle English (27) (van der Wouden and Zwarts 1993) and a particle n’ in Catalan (28) (Espinal 2000).

12

(28) Je crains qu’il I

fear

ne

vienne.

[French]

that–he not come.Subj

‘I fear that he will come.’ (29) Then fearing lest we should have fallen upon rocks,

[English]

they cast four anchors out of the stern. (30) Abans que before that

(Acts 27:29)

passi

res,

jo me

happens

anything

I

n’

meCL Neg

aniria.

[Catalan]

go.Cond.1sg

‘Before anything happens, I would leave.’

Furthermore, EN takes the form of the genitive of negation in Russian (Brown 1999a,b, Brown and Franks 1995, 1997) and in Polish (Harves 2000). Likewise the prosodic weakening and shortening facts above can be understood along similar lines – as a strategy to distinguish EN from real negation. I take these systematic contrasts between EN and real negation to argue that EN must be treated as a separate creature, casting doubt on the previous symmetric analyses between EN and regular negation. For instance, if real negation and EN are essentially equivalent, why does only the latter persistently arise in a morphophonetically weaker form across typologically distant languages? Thus I pursue an analysis in which EN is not a subtype of negation, contra Tovena’s (1996) and Abels’ (2002) assumption that EN is real negation, with the loss of negative force due to its environment. Rather, it will be shown that, just like polarity items, EN is triggered only in nonveridical environments and, just like the evaluative use of subjunctive mood, EN carries an

13

evaluative sense. These two crucial components of EN will be explained by the following theoretical frameworks in 1.2 and 1.3.

1.2. Dependency on Nonveridicality: Giannakidou 1994 et seq. As I propose that all instances of EN exhibit dependency on nonveridicality, just like polarity items and subjunctive mood (Giannakidou 2009), I briefly review here the definitions of nonveridicality. The term veridicality is defined with regard to truth and existence in Montague (1969). Giannakidou (1994 et seq.) and Zwarts (1995) propose the specific definition of veridicality in terms of truth, which correctly predicts the distributions of NPIs: NPIs may distribute only in nonveridical environments. Giannakidou recasts veridicality as a property of sentence embedding functions. As shown in the definition below, a function F is veridical if Fp entails or presupposes the truth of the proposition p. If Fp does not allow such inference to the truth of p, F is nonveridical.

(31) (Non)veridicality for propositional operators (Giannakidou 2006) i. A propositional operator F is veridical iff Fp entails or presupposes that p is true in some individual’s epistemic model ME(x); otherwise F is nonveridical. ii. A nonveridical operator F is antiveridical iff Fp entails that not p in some individual’s epistemic model: Fp → ¬ p in some ME(x).

To put it differently, whereas veridical operators are linguistic expressions which entail certainty and an individual’s commitment to the truth of a proposition in her epistemic model ME(x),

14

nonveridical operators convey uncertainty and lack of commitment. Negative operators are defined as antiveridical since not p entails the falsity of p. Notice that the main properties of nonveridical operators such as uncertainty and lack of commitment parallels to the canonical properties of subjunctive mood (which will be discussed in detail in chapter 4). In particular, it has been claimed that mood selection seems to be linked to polarity in terms of (non)veridicality – the indicative mood is related to the notion of veridicality while the subjunctive mood is related to nonveridicality in languages like Spanish, Italian, French, Greek (Giannakidou 1994, 1995, 2009; Quer 1998; Borschev et al. 2007 for Russian subjunctive). Furthermore, in this dissertation it will be shown that EN is another species with dependency on nonveridicality. The distribution of EN will be shown to track the typical nonveridical contexts in which NPIs and subjunctive mood are licensed or triggered. More importantly, the intuition behind the choice of EN is strongly reminiscent of that of subjunctive mood.

1.3. Evaluative Component on separate dimension: Potts 2005; Giannakidou & Yoon 2009 The evaluative sense of EN that is independent of the semantics of its environment will be captured by means of a refined version (Giannakidou and Yoon 2009, to appear) of multidimensionality of conventional implicatures (CI) in the sense of Potts (2005, 2007). It will be shown that EN is an “utterance modifier” that carries a “comment on the semantic core” (Potts 2005). It exhibits striking resemblance to typical CI triggers such as expressives like damn and bastard (Potts 2005, 2007), appositives (Potts 2005), honorifics (Potts 2005; Potts and Kawahara 2004), ‘even’ (Giannakidou 2007), etc. This important aspect of EN that went

15

unnoticed by previous accounts, strongly calls for the terminological transition from the traditional expletive negation to evaluative negation that I propose here. The term expletive negation is motivated by sole consideration of at-issue meaning. To begin, I will introduce Potts’s system of the multidimensionality of CI, and specific characteristics of EN concerning its CI status will be given later in chapter 5. Potts (2005) summarizes the abstract properties of CIs:

(32) a. CIs are part of the conventional meaning of words. b. CIs are commitments, and thus give rise to entailments. c. These commitments are made by the speaker of the utterance d. CIs are logically and compositionally independent of what is said.

In other words, while CI is not part of ‘what is said’, i.e. at-issue meaning (Grice 1975, 1989), it is part of the conventional meaning of the words. The expressive meaning triggered by CI is not cancelable. Lastly, CIs are speaker-oriented (Cruse 1986, Löbner 2002; Potts). Furthermore, the following propositions constitute the basic assumptions for the logic for CI in Potts’s system.

(33) a. CIs are scopeless (always have widest scope). b. CIs result in multidimensional content. c. CIs are subject to an antibackgrounding requirement. d. CIs comment upon an at-issue core.

16

First, as CIs are utterance modifiers, they take widest scope. Second, in agreement with Kratzer’s (1999) multidimensionality, the meaning of CIs is posited on a separate dimension from the basic semantics of utterance. Third, an antibackgrounding requirement means that CIs convey new information. Finally, CIs can be understood as a speaker-oriented comment on the content of the utterance. Given these assumptions, Potts develops the CI logic, which is based on Karttunen and Peters’s (1979) ideas and revise their rule-by-rule system with a new system based in type-driven translation of Klein and Sag (1985). The following composition scheme illustrates the mode of combination in the CI logic, in which the bullet function • (i.e. a separation function for independent lamda expressions) ensures that the CI meaning on another level takes part in the overall interpretation. (34) α (at-issue) • β (α) CI α (at-issue)

β (CI)

(Potts 2005: 48)

The CI logic crucially provides us with a tool for a meaning that exists on a distinct dimension. Since the evaluative meaning of EN is very similar to that of speaker-oriented adverbs like fortunately, let us look at how CIs are captured in the following sentence. It illustrates the treeadmissibility condition of the CI logic (based on Karttunen and Peters 1979) which requires that “a CI meaning always applies to an at-issue meaning to produce a CI meaning.” (Potts 2005: 48) This feature will become extremely important when the various CI meanings carried by EN are accounted for.

17

(35) a. Fortunately, Beck survived. λw.survivew(Beck): · • fortunately(λw.survivew(Beck)): b.

fortunately: < , >

λw.survivew(Beck):

(Potts 2005: 64)

The at-issue meaning here is “Beck survived” and the CI conveys “that Beck survived is fortunate.” The adverb fortunately with type < , > takes λw.survivew(Beck) with type , and gives out type . The superscript c indicates a CI type; the superscript a indicates an at-issue type. Crucially, the at-issue term λw.survivew(Beck) is percolated to the mother node and the part of the argument, and the result of the CI application is passed on to the mother node. This general composition rule is designed to separate the at-issue dimension from the CI operators. In chapter 5, I will show how the refined version of CI logic (Giannakidou and Yoon 2009, to appear) will be applied to EN.

1.4. Roadmap In chapter 2, I show the pragmatic contribution of two scalar meanings of undesirability and unlikelihood. It is further shown that the base of scale denoted by EN may vary depending on the context or the epistemic subject’s (the speaker or the matrix subject) emotional state, perhaps reflected in the tone of voice. In doing so, I propose that EN is not an imperfection or illogicality of language. Rather, EN represents another legitimate function of negation in natural language, where a negative element is adopted for the purpose of circumventing a commitment to a truthful statement and combines this with an attitude. This makes it similar to the subjunctive mood.

18

Chapter 3 offers a critical review of two predominant camps on theories of evaluative negation – expletive approaches and non-expletive approaches. Taking more data into consideration, I provide empirical and conceptual arguments as to why prior theories require revision. First, it is shown that, contra the assumption of expletive approaches, EN does indeed have non-vacuous semantic contribution – evaluative sense that is represented as various kinds of scalar meaning. Second, the non-expletive approaches are unable to predict the variability that we find within a context and within a language. In chapter 4 I propose my analysis of EN as a mood marker. EN is a subjunctive mood marker, in particular, following the pattern of the mood choice in Greek that is regulated by nonveridicality (Giannakidou 2009). I discuss the parallels in the positive nonveridical environment in the scope of verb ‘hope’. This is unexpected under any account that we have seen, and the fact that EN requires a non-factive complementizer and a subjunctive predicate in Korean and Japanese should come as no surprise. In chapter 5, I add a pragmatic component to the semantic analysis. I argue that EN contributes an evaluative dimension. This evaluative dimension is negative anticipation, undesirability, or low likelihood. I show that EN in different environments exhibits a striking parallel in interpretation as an ordering relation. I propose to capture this by means of multidimensionality of conventional implicature (Potts 2005) (as refined in Giannakidou & Yoon 2009, to appear)), and show how evaluative semantics of EN exists on a separate dimension from the semantic core of utterance. Contrary to previous views that dismiss the meaning of EN, calling it expletive, my proposal captures the precise role of EN in an utterance. If my analysis is correct, it has one important implication: It allows the generalization that various subspecies of EN in language are indeed part of the grammar. In particular, they are reflexes of

19

grammaticalization of perspective and subjective mode, on a par with subjunctive mood choice and other subjunctive phenomena like metalinguistic comparative as argued in Giannakidou and Yoon (2009, to appear). In chapter 6, I explore the syntactic configurations of subordinate subjunctive clauses and subordinate EN-clauses in Korean (and potentially Japanese). One of the questions that will arise is whether the matrix clause and the EN-clause in Korean are hypotactically or paratactically connected. This will inevitably involve the Germanic Embedded Verb Second (EV2). The dynamic syntax of EN (and subjunctive) complements that I propose affords the following theoretical implications: First, by revealing the parallels with EV2 constructions that are in subjunctive, the syntax of EN constructions further supports the current proposal that EN is subjunctive marker. Second, it suggests that the syntax of EN complements in Korean must be understood along the lines of Jespersen’s insight of paratactic negation in Old/Middle English. Third, it accounts for why the complementizer for subordinate EN clauses is in an analogous form with a question particle in Korean and Japanese. Finally, it furthermore supports the possibility that negation in tag-questions, which are classical paratactic configurations, can be understood as a subspecies of EN. The analysis can be extended to Japanese since Japanese EN has exactly the same properties. Finally, in chapter 7, I further strengthen the connection between mood and EN by looking at rhetorical comparatives. My overall approach, if correct, has two important consequences: First, we have explanations for the crosslinguistic variation in subordinate EN between Japanese/Korean and other languages. Second, it offers a plausible way of systematically characterizing EN in other environments such as exclamatives, questions, nonveridical before clauses and comparatives.

20

Chapter 2. Empirical description of Evaluative Negation

In this chapter, I present data which illustrate various aspects of the semanticopragmatic effects of evaluative negation across languages. Besides the peculiarity of lacking negative force, the status of EN has never been agreed upon also due to its broad distribution. However, an empirical overview of the distribution of EN in a wide range of environments and languages will importantly reveal the underlying systematicity in EN phenomena. In particular, I show that EN triggers various kinds of semantic effects depending on the context and the language, yet this apparent division into a number of semantics ultimately boils down to one semantic principle – the various instances of EN uniformly denote a certain ordering relation. Importantly, however, the current analysis of EN in each context that I posit here does not necessarily go against the previous insights on EN. Rather, the goal of the current proposal is to suggest a unified analysis of EN, discovering the semantic common denominator that was dormant within earlier intuitions on what EN denotes in individual contexts. Specific types of ordering relations are thus posited conforming to the insights of the previous analyses on various instances of EN. In doing so, the commonality of triggering an ordering relation will serve as the basis for a uniform semantic analysis of EN in chapter 4 and 5: since it is strongly reminiscent of the semantic characteristics of the subjunctive mood, this result suggests an important conceptual link between evaluative negation and the subjunctive mood. More empirical and theoretical motivations for the proposed link will be discussed in chapter 4, and then chapter 5 will discuss how the assumption of EN as a subspecies of subjunctive-mood marker opens the possibility of capturing the occurrences of EN in various contexts under a unified principle.

21

Some of the data presented here has been adopted from various sources in the literature. Other data has been collected from my informants. I mentioned the source for each example.

2.1. EN in Verbs of Fear To begin, I examine EN in embedded clauses under verbs of fear, and further show how the new data in Korean and Japanese contributes the crucial variability of EN that went unnoticed in the previous literature of EN. A typical environment for EN is under adversative predicates, as illustrated below in various languages (Jesperson 1917, 1924; van der Wurff 1999; van der Wouden and Zwarts 1993):

(1) a. Nature […] forbedeth that no man make himself riche.

(Chaucer) [Old/Middle English]

b. First he denied you had in him no right. (Shakespeare) c. Then fearing lest we should have fallen upon rocks, they cast four anchors out of the stern, and wished for the day. d. Je crains qu’il ne vienne.

(Acts 27:29) [French]

I fear that-he not come.Subj ‘I’m afraid he may come.’ e. Evitez qu’il ne voud parle. Prevent that-he not to-you speak ‘Don’t let him talk to you.’ (2) a. Timeo ne I-fear

that-not

veniat.

[Latin]

he come.Subj

22

‘I’m afraid he may come.’ b. Fobamai mipos fear-1sg that-not

kano

lathos.

(Ruge 1986) [modern Greek]

make-1sg error

‘I am afraid of making a mistake.’ (3) a. Tenia por que ne I.had fear that Neg

escollissin

un nou director.

(Martin, ms)[Catalan]

they.elected.Subj a new director

‘I was afraid that a new director wasn’t elected.’ b. Tenia por que no

escollissin

un nou director.

I.had fear that Neg they.elected.Subj a new director ‘I was afraid that a new director would be elected.’

Notice here that most examples with EN are inflected in subjuctive mood. More interestingly, Greek has the distinction between indicative negation and subjunctive negation (Giannakidou 1998), and in the Greek example above mipos contain the negation that accompanies the subjunctives, not the indicatives. These reveal the crucial connection between EN and mood that will be discussed further later in chapter 4. The role of EN in above contexts is similar: For instance, the following EN no seems to convey a bouletic scale like the one in (4b) which expresses that the content of the proposition ‘a man makes himself rich’ is considered to be undesirable in the attitude holder’s buletic state.

(4)

a. Nature […] forbedeth that no man make himself riche. b. ‘a man makes himself rich’

λw.fear pw(I):

Given the evaluative contents of EN, an association with the undesirability property predicts its distributional restriction only to the complement clauses of adversative predicates in languages like French, Old/Middle English, Polish, Russian, Catalan, etc. That is, if EN co-occurs with positive predicates such as espérer ‘hope’ in these languages, as in (40), a semantic conflict would arise between the undesirability conveyed by EN, as given in the parentheses, and the desirability conveyed by the predicate ‘hope’. 141

(40) *J'espére I hope

que cela

ne

that it might

se reproduise.

[French]

Neg happen again

‘I hope that it might happen again (#because it is undesirable).’

As such, the restriction of subordinate EN to contexts with a negative import in these languages is predicted by the evaluative content of EN. The crosslinguistic variation between these languages and Korean/Japanese of subordinate EN seems to originate from parameterization on the type of evaluative content of embedded EN. Thus far I have discussed the evaluative content of EN in embedded clauses and its independence from the meaning of the matrix predicates unless they induce logical contradiction.

5.3.3.2. EN in Exclamatives: undesirability or unlikelihood Now let us revisit the case of EN in exclamatives, where the evaluative content of EN may concern either likelihood or desirability. As discussed in chapter 2, previous approaches note that EN in this context marks unexpectedness and amazement (Meibauer 1990 for German; Portner and Zanuttini 1999, 2000 for Paduan Italian; Brown 1999 for Russian).

(41) Was du

nicht alles machst!

what you Neg all

[German]

do

‘The things you do!’

142

In principle the exclamative can be formed with or without EN, and the ordinary semantics with surprise of ‘The things you do!’ is maintained irrespective of the existence of EN. On a separate dimension, the evaluative content of EN nicht can be characterized as below, triggering an additional comment on the amazement in this case. Just like the EN under the verb of ‘fear’ resulting in strong undesirability, a sentence like (41) seems to end up marking stronger surprise than the one without EN by means of the dual expressions on surprise – exclamative formation the main purpose of which is to express surprise, and EN which also expresses surprise via unlikelihood presupposition. I rephrase the unlikelihood as ‘surprisingly’ in the following semantic composition.

(42) a. Surprisingly, the things you do! b. λw.do the things!w(you): · • Surprisingly (λw.do the things!w(you)):

Surprisingly:

< , >

λw.do the things!w(you):

Given the following commentary by Partee, the unexpectedness-denoting EN is shown to be a subjunctive mood marker.

(43) [Partee 2005] “even dogs and prelinguistic infants have conceptions of alternative possible state of affairs and that any indication of surprise is good evidence of this: surprise is a reaction to a mismatch between an expected state of affairs and a perceived state of affairs, and without 143

the possibility of conceiving of things as being different from how they actually are, we couldn’t be surprised.”

On the other hand, I noted that EN in exclamatives shows flexibility in that it can express a surprise like German above, or it can be employed to express the speaker’s resentment about the state of affairs in Korean (and the exact counterpart in Japanese).

(44) Ney-ka you-Nom

aisukurim-ul

ta mekess-cyanh-e!

[Korean]

ice cream-Acc all ate-Comp.Neg-Decl

‘You ate up all the ice-cream!’

In the latter case, the evaluative content of EN marks undesirability. I will not repeat here the CI logic for (44). It will be similar to the one in (42) above, with ‘undesirably’ in lieu of ‘surprisingly’. Since the form of exclamative indicates a surprise and EN encodes undesirability, (44) ends up expressing an unpleasant surprise.

(45) At-issue: ‘It is surprising that you ate up all the ice-cream.’ CI:

‘Undesirably.’

This variability in the evaluative content of EN here directly follows from the basic semantics of exclamatives – the perceived state of affairs is uttered in an exerted emotional state which can be either positive or negative.

144

5.3.3.3. EN in Polite requests: unlikelihood The meaning of EN thus far is quite predictable from the core semantic properties of the evaluative subjunctive. However, more interesting questions arise from the emergence of EN in other environments. First, recall the following example in which a speaker utilizes EN for the sake of politeness.

(46) Ceki(maliya), PH(placeholder)

hoksi

ton-com

pillyecwu-ci anh-ul-lay?

by any chance

money-a little lend-Comp

[Korean]

Neg-Fut-Q

‘The thing is … would you mind lending me some money?’

The question here is: why does EN in a request denote politeness or indirectness? If we take a closer look, there are other politeness markers such as the placeholder ceki and the lowlikelihood adverb hoksi ‘by any chance’, besides EN. Given that the employment of a lowlikelihood adverb is a legitimate politeness strategy, we can infer that the indirectness-marking EN anh here is derived from the evaluative content of unlikelihood. Hence, as given below, the CI meaning of EN here is something like ‘You can say no’ which must have derived from more fundamental meaning of unlikelihood like ‘I am aware it is unlikely that you will lend me some money.’

(47) At-issue: ‘Will you lend me some money?’ CI:

‘You can say no.’

145

5.3.3.4. EN in Temporal connectives: undesirability or unlikelihood Another mysterious case is the EN that marks a temporal sequence. As my intuition is restricted to Korean and Japanese data, I will discuss only the possible interpretation of Korean data here (assuming an analogous analysis in Japanese), and leave the applicability of the current analysis to EN in similar contexts in other languages for future research. Again the crucial hypothesis is that the meaning of temporal sequence is also derived from subjunctive meaning. Since a before-clause like (46) has been widely assumed to be counterfactual (Heinämäki 1972; Ogihara 1995; Beaver and Condoravdi 2003; Kaufmann 2005; Kaufmann and Takubo 2005) – a subtype of nonveridical contexts, it is expected to be a good environment for EN as well as the subjunctive mood.

(48) Mary defused the bomb before it exploded.

As it is assumed that what exactly is implied about ‘B’ in ‘A before B’ is determined by the context (Heinämäki 1972; Ogihara 1995; Beaver and Condoravdi 2003), the meaning of EN in the following before-clauses can be analyzed as a likelihood scale or a desirability scale depending on the context. Note that such scale exists only in the speaker’s epistemic model.

(49) Malhaci-anh-kiceney say-Neg-before

sero

alkoiss-esstenkes-ita.

each other

know-Pst-Decl

‘They knew it before saying it to each other.’ (50) Cip

han chay-to

house one.Cl-even

(J.-Y. Jung ‘Stone pillow’)

pwulsaruci anh-kiceney, saram hana-to burn

Neg-before

146

[Korean]

cwukiki-ceney …

persone one.Cl-even kill-before

‘Before he burns down any house, before he kills anyone...’(B.-J. Lee ‘government-owned ferry II’)

Whereas EN in (49) seems to denote a likelihood scalar implicature as in (51), EN in (50) gives rise to a desirability scalar implicature as in (52). In other words, ‘They knew it before EN saying it to each other’ in (49) can be paraphrased as ‘They knew it before saying it to each other, which is surprising because ‘saying it to each other’ is unlikely to have been realized yet.’ On the other hand, EN in (50) means ‘what is said in before-clause is undesirable.’ Thus CI below denotes what the speaker wants to evaluate whether it is unlikely, desirable, etc, and comment on the basic content of what is said (at-issue meaning).

(51) At-issue: ‘They knew it before (EN) saying it to each other.’ CI:

‘‘Saying it to each other’ is unlikely to have been realized yet.’

(52) At-issue: ‘Before he (EN) burns down any house, before he kills anyone...’ CI:

‘‘His burning down any house’ is undesirable.’

Furthermore, in the following sentence in Japanese (Kinoshita 1998) in which the EN nai cooccurs with the subjunctive (conditional) form nara ‘become.if’, both the subjunctive marker and EN denote the counterfactuality.

(53) Kuraku nara-nai dark

mae-ni kaette-kuru no-desuyo.

become.Subj-Neg before return

-Imp

'You must come back home before it gets dark.'

147

[Japanese]

The counterfactuality property can be reinterpreted as the speaker’s intention of unrealization of the propositional content in the real world, just like CI in (51), or it can be also due to undesirability in (52) depending on the context.

5.3.3.5. EN in Comparatives: undesirability or unlikelihood Finally, let us revisit the role of EN in comparatives. Among the various previous approaches discussed in chapter 3, the analysis closest in spirit to the current assumption – i.e. EN has an evaluative content like that of mood expressions – is found in Napoli and Nespor (1976, 1977). Crucially, they note that EN only occurs in the subjunctive mood in Italian, and EN in comparatives occurs (although not always) when it presupposes a contradiction to previous thoughts of the listener. As shown in (55), the EN non in (54b) is analyzed as marking unlikelihood.

(54) Maria è più intelligente (a) di quanto è Carlo / (b) di quanto non sia Carlo.

[Italian]

‘Mary is more intelligent (a) than Carlo is / (b) Carlo (not) is.’ (55) At-issue: ‘Mary is more intelligent than Carlo’ CI:

‘It is unlikely that Carlo’s intelligence comes close to Mary’s.’

On the other hand, I suggested how the following EN in Catalan and Spanish (Price 1990) can be accounted for by assuming the flexibility that EN here triggers a semantic effect based on a desirability scale.

148

(56) Val més que vingueu better

que no que us

quedeu

sols.

[Catalan]

that come.Subj.2pl than not that you.Cl remain.2pl alone

‘It’s better you come than you stay on your own.’ (57) At-issue: ‘It’s better you come than you stay on your own.’ CI:

‘It is undesirable that you stay on your own.’

These cases represent the rhetorical effects in comparatives. I assume that the negative rhetorical effects in (54) and (56) arise from the speaker’s presupposition of unlikelihood or undesirability. Given these two types of evaluative contents, it seems plausible to refine my earlier assumption above that EN (in chapter 2) in comparatives expresses inequality in degree. Rather, I propose that even in the case where comparatives with EN look like non-emphatic regular comparatives like (58), the task of EN is to put an emphasis on the contrast by triggering a scale in terms of likelihood or desirability depending on the context or the epistemic subject’s mind. A detailed discussion of how nonveridical elements such as NPIs, mood, and EN give rise to similar emphatic effects on comparatives will be given in chapter 7.

(58)

Maria è più

alta di quanto

(non) lo sia Giovanni.

(Donati 2000) [Italian]

Maria is more tall of how-much Neg it is Giovanni ‘Maria is taller than Giovanni is.’

When it comes to the expressive interval for EN, I noted in 5.1.2 that it ranges between [-1, 0], the negative part of the interval. This is consistent with the expressive content that EN carries, i.e. undesirability or unlikelihood, which are the ‘negative’ part of the epistemic scale of desirability

149

or likelihood. This is a welcome result since it further conforms to the basic properties of CI: ‘CIs are part of the conventional meaning of words (Potts 2005).’

5.3.3.6. Multidimensionality shown in multiple CIs Before closing the discussion on the CI of EN, let’s observe what exactly happens in the following sentence in Japanese, which contains more than one instances of EN:

(59) Mary-ga Mary-Nom

ko-nai

ka-naa!

[Japanese]

come-Neg

NFcomp-Opt(EN)

‘Oh, how I wish Mary comes (although it is unlikely to happen)!’ (60) At-issue: ‘I wish Mary comes.’ CI:

‘It is unlikely that Mary comes.’

Remarkably, this simple sentence contains two CI particles, each of which induces its own layer of meaning in addition to the ordinary semantic value. Notice that the evaluative content of the EN naa, expressing the desirability scale of the speaker, is grammaticalized as a regular optative marker in Japanese, hence its meaning is incorporated in the ordinary semantic values is the proposition ‘I wish Mary comes’. The EN nai, on the other hand, denotes the unlikelihood state which exists on a separate dimension as CI. As noted in Korean, the non-factive complementizer ka in Japanese also acts as a subjunctive mood marker encoding the uncertainty state of the proposition ‘Mary comes’, which also contributes the scalar meaning unlikelihood. Considering the historical development from negation to the subjunctive mood in Japanese that was noted in 4.3, the optative marker naa, derived from an original negative, not

150

only reveals the crucial connection from negation to the subjunctive, but more importantly it illustrates a case of grammaticalization of EN that is operative on the ordinary semantics.1 Given that meanings on different layers can be integrated into a single sentence, the conception of multidimensionality is not only useful but necessary especially for describing the precise evaluative content of the cases with multiple instances of EN and subjunctive marker.

5.4. Summary The distributional facts showed that EN in Korean and Japanese occurs in typical subjunctive contexts like polite requests, emphatic sentences, dubitatives, etc. in languages like Greek. Importantly, however, the proposal that EN is the precise analogue of the subjunctive would run into problems because in languages like Romance the subjunctive mood displays a wider distribution than EN. This crosslinguistic variation between Greek and Romance is revealed by the contrast in factive predicates which select for the indicative in Greek (Giannakidou 2009) and the subjunctive in Romance (Farkas 1992, Quer 1998, Villalta 2008). Thus section 5.1 tries to solve this puzzle on the difference between EN and subjunctive: based on the data in Korean, it was shown that EN is not simply optional, but an epistemic subject intentionally selects to employ EN in order to make a rather subtle yet significant semantic impact – potentiality with negative anticipation. In 5.2, I showed how the nonveridical semantics of the predicates that selects EN can be represented. In 5.3, I proposed that the evaluative content of EN can be represented by means of multidimensional CI logic. Nonveridicality dependency predicts that 1

Like all other instances of EN with negative interval only, the optative marker naa here seems to range over the negative interval in the sense that optative naa expresses or at least used to express unattainable wish, just like (57) above. In this light, it ranges over the negative interval in terms of likelihood. Note that the counterpart with positive interval would be desiderative subjunctive in which case EN can never appear. As noted in 4.1, this distinction is made in Greek in which ‘more probable future condition’ is encoded by the subjunctive and ‘less probable future condition’ is represented by the optative. 151

the meaning of EN is similar to that of the evaluative subjunctive whose distribution is also dependent on nonveridicality. I discussed how uncommon meanings of EN such as emphasis on unequal degree or emphasis on temporal sequence might have been derived from the basic evaluative subjunctive meaning in contexts like comparatives or nonveridical temporal clauses. I further showed how these evaluative contents of EN, modifying the whole utterance, can be captured by the CI logic. If my analysis is on the right track, it has one important implication: It allows the generalization that various subspecies of EN in language are indeed part of grammar. In particular, they are reflexes of grammaticalization of perspective and subjective mode, on a par with subjunctive mood choice. Given that EN is a grammatical element, I will discuss the syntactic aspects of how EN interacts with other mood elements in the sentence in the next chapter.

152

Chapter 6. The Syntax of Evaluative Negation

Given the proposal that EN is a subspecies of subjunctive mood marker, this chapter suggests how the semantic and pragmatic features thus far are mapped with syntax in the domains of EN and the subjunctive mood. In particular, I explore the syntactic configurations of subordinate subjunctive clauses and subordinate EN-clauses in Korean (and potentially Japanese) and compare them to those of Romance and Germanic languages. Having argued in previous chapters that EN as well as evaluative subjunctive elements uniformly trigger evaluative semantics, the main research question I will attempt to answer in this chapter is how clauses containing these evaluative elements in Korean (and Japanese) can be syntactically represented. First, I investigate, in 6.1, the configurational relation between the matrix clause and the ENclause in Korean, examining whether the two clauses are hypotactically or paratactically connected. This particular question is worth investigating since the results will offer us important implications about: (i) whether EN complements in Korean behave like Germanic Embedded Verb Second (EV2) constructions that are in subjunctive mood; (ii) whether Jespersen’s idea of paratactic negation holds in Korean EN complements; (iii) whether the fact that complementizers of EN clauses are in an analogous form of question particles in Korean and Japanese is relevant to the paratatic structure; and (iv) whether negation in tag-questions, which are classical paratactic configurations, can be understood as a subspecies of EN.1 It will be shown that EN complements in Korean reveal paratactic properties indeed, which implicates positive answers to these questions.

Though tag-questions have been analysed in other ways as well -- see, for instance, den Dikken’s (1995) analysis on heavy NP-shift across tags. 1

153

After addressing the question of what kinds of syntactic derivations are involved in sentences with subjunctive and EN clauses, I will provide, in 6.2, empirical descriptions regarding which elements constitute subjunctive mood that gives rise to an evaluative sense in Korean.

6.1. Parataxis versus Hypotaxis In the first part of the discussion, I will examine whether subordinate clauses that contain EN behave differently from (i) their exact counterparts with EN elided, (ii) subordinate clauses with regular subjunctive marker, and (iii) subordinate clauses with indicative marker. In particular, I will focus on the question of whether there is syntactic evidence for the idea that the subordinate clause in EN constructions is an independent root clause. It is important for our purposes to ask the question of whether the relation between matrix clause and EN-clause is parataxis or hypotaxis in Korean. If the connection turns out to be paratactic, it will implicate the following: For one thing, the intuition underlying the EN constructions in Korean is analogous with that of EN in Old/Middle English that is termed as Paratactic Negation by Jespersen (1940). Furthermore, if EN-clauses exhibit root properties, it will importantly support the current proposal that EN is a subjunctive marker by revealing the crucial parallelism between EN constructions in Korean and Embedded Verb Second (EV2) constructions in Germanic that are known to display root properties when in subjunctive mood.2

2

Many thanks to Marcel den Dikken for suggesting that I take Jespersen’s term “paratactic negation” seriously, and explore the link between EN constructions in Korean and EV2 constructions in Germanic. 154

6.1.1. Property of paratactic (root) dependencies Jespersen (1940: 455) termed negation in the following examples as Paratactic Negation, noting that “the clause here is in some way treated as an independent sentence, and the negative is expressed as if there had been no main sentence of that particular kind. It is well known how this develops in some languages to a fixed rule.”

(1) You may deny that you were not the meane Of my Lord Hastings late imprisonment. (2) We have forbidden … that they doe not shew any naturall worke. (3) It never occurred to me to doubt that your work … would not advance our common object in the highest degree.

Beyond the domain of above negation, the original insights of Parataxis date back to Andrés Bello that is famously defended by Davidson (1967b). More recently, the “Root Transformation (RT)” proposal by Emonds (1976) raises the issue of the problem of main and subordinate status in generative grammar by proposing that some structural configurations are only found in matrix clauses. The idea of the partactic view is that to assert that Galilleo believes that the earth is round is to assert something similar to “Galilleo believes that” and the object of believe is cataphorically related to the separate sentence “the earth is round” (Uriagereka and Torrego 2002: 253). Uriagereka and Torrego argue that both paratactic and hypotactic types of dependencies are realized in UG, comparing two non-interrogative finite connectives, the paratactic como ‘how’ and the hypotactic que ‘that’ clauses in Spanish. As paratactic clauses, they show that como

155

clauses exhibit more restricted distribution than que clauses, and syntactic dependencies across como are prohibited. Another well-known case of RT is EV2 constructions in Germanic (Hoeksema and Napoli 1993; de Haan 2001 in Frisian; Hrafnbjargarson and Wiklund 2009 in Icelandic; see also Mari and Martin 2007 in Greek). Typical paratactic properties of EV2 constructions include: a) they do not involve a complementizer; b) the verb of the second clause with paratactic nature is in the subjunctive (German EV2 constructions); and c) EV2 in German is strikingly similar to direct quotation. Furthermore, de Haan (2001) summarizes major paratactic properties of EV2 constructions in Frisian as follows:

(4) Major properties of EV2s in Frisian

(de Haan 2001)

a. general display of root phenomena b. occurrence as structural root CPs only c. obligatory occurrence outside and to the right of the matrix clause (cannot undergo topicalization nor be part of other preposings) d. no binding from outside the EV2s e. intonation unit f. independent focus domain g. (limited) iteration (matrix clause has to be a structural root) h. no extraction

156

Based on these diagnostics, I will examine subordinate EN clauses that are compared to ENelided counterparts, subordinate indicative clauses, and subordinate subjunctive clauses in Korean. First, observe that postposing of indicative complements is rather awkward. In contrast, postposing is shown to be authorized without causing oddity across subjunctive complements, EN complements, and the non-EN counterpart – the exact counterparts of the EN-complements where EN is elided.

Non-subjunctive complement (5) a. John-un John-Top

[Mary-ga

[Korean] ol-kes-ul]

hwaksinha-koiss-ta.

canonical order

Mary-Nom come-Comp-Acc certain-Asp-Decl

‘John is certain that Mary will come.’ b. ??John-un

hwaksinha-koiss-ta,

[Mary-ga

ol-kes-ul].

post-posing

Subjunctive complement (6) a. John-un John-Top

[Mary-ga

o-assumyenhako]

Mary-Nom come-SubjComp

para-koiss-ta.

canonical order

wish-Asp-Decl

‘John fears that Mary might come.’ b. John-un

para-koiss-ta,

[Mary-ga

o-assumyenhako].

post-posing

EN complement (7) a. John-un John-Top

[Mary-ga

oci-anh-ul-kka] kekcengha-koiss-ta.

canonical order

Mary-Nom come-EN-Fut-NFComp fear-Asp-Decl

‘John fears that Mary might come.’ b. John-un

kekcengha-koiss-ta,

[Mary-ga

157

oci-anh-ul-kka].

post-posing

Non-EN complement (8) a. John-un John-Top

[Mary-ga

o-l-kka]

kekcengha-koiss-ta.

canonical order

Mary-Nom come-Fut-NFComp fear-Asp-Decl

‘John fears that Mary might come.’ b. John-un

kekcengha-koiss-ta,

[Mary-ga

o-l-kka].

post-posing

The second diagnostic for root property comes from prosodic independence. As demonstrated below, an intonational break between matrix and subordinate clauses sounds natural only for subjunctive complements, EN complements, and EN-elided counterparts, but terrible for indicative complements in (9). More crucially, a rising intonation is allowed only at the end of EN complements, and EN-elided counterparts. Note that this renders them similar to a direct quote of a polar question, which is unsurprising since, as already noted, the non-factive complementizer (NFComp) kka is identical in form with a question particle.

Non-subjunctive complement (9) *John-un ## [Mary-ga John-Top

ol-kes-ul] (↑) ###

hwaksinha-koiss-ta.

Mary-Nom come-Comp-Acc

certain-Asp-Decl

‘John is certain that Mary will come.’ Subjunctive complement (10) John-un ## [Mary-ga John-Top

o-assumyenhako]

Mary-Nom come-SubjComp

###

para-koiss-ta. wish-Asp-Decl

‘John wishes that Mary might come.’

158

[Korean]

EN complement (11) John-un ## [Mary-ga John-Top

oci-anh-ul-kka] (↑) ###

Mary-Nom come-EN-Fut-NFComp

kekcengha-koiss-ta. fear-Asp-Decl

‘John fears that Mary might come.’ Non-EN complement (12) John-un ## [Mary-ga John-Top

o-l-kka]

(↑) ###

Mary-Nom come-Fut-NFComp

kekcengha-koiss-ta. fear-Asp-Decl

‘John fears that Mary might come.’

The third test is whether it allows speaker-oriented interjections: de Haan notes that speaker oriented interjections like godskes ‘gosh’ are possible only in root clauses and subjunctive EV2s, but not in dat/V-final clauses in West Frisian.

(13) Pyt hie my in boadskip stjoerd, dat godskes, hja soene trouwe. Pyt had me a message

sent

that gosh

[West Frisian]

they would marry

Pyt had sent me a message sent that gosh they would marry.

Likewise in Korean, the following speaker-oriented interjection saysangeyna ‘gosh’ can appear only in EN-complements and its EN-elided counterparts.

Non-subjunctive complement (14) John-un [Mary-ga

(*saysangeyna) tachess-tako]

John-Top Mary-Nom gosh

injured-Comp

159

hwaksinha-koiss-ta. [Korean] certain-Asp-Decl

‘John is certain that gosh Mary got injured.’ Subjunctive complement (15) John-un

[Mary-ga

(??saysangeyna) naa-ssumyenhako]

John-Top Mary-Nom gosh

recover-SubjComp

para-koiss-ta. wish-Asp-Decl

‘John wishes that gosh Mary would recover.’ EN complement (16) John-un

[Mary-ga

(saysangeyna)

John-Top Mary-Nom gosh

tachici-anh-ul-kka]

kekcengha-koiss-ta.

injured-EN-Fut-NFComp

fear-Asp-Decl

‘John fears that gosh Mary might get injured.’ Non-EN complement (17) John-un

[Mary-ga

(saysangeyna)

John-Top Mary-Nom gosh

tachici-l-kka]

kekcengha-koiss-ta.

injured-Fut-NFComp

fear-Asp-Decl

‘John fears that gosh Mary might get injured.’

More crucially, direct evidence for root property would come from testing whether a clause can occur independently. In accordance with above facts of EN that reveal paratactic properties, the following data show that only an indicative complement seems to be ungrammatical when it occurs as structural root CP.

Non-subjunctive complement (18) *Mary-ga

o-lkes-ul/tako.

[Korean]

Mary-Nom come-Comp-Acc/Comp ‘that Mary will come.’

160

Subjunctive complement (19) Mary-ga

o-assumyen.

Mary-Nom come-SubjComp ‘I wish Mary would come.’ EN complement (20) Mary-ga

oci-anh-ul-kka/ci.

Mary-Nom come-EN-Fut-NFComp ‘(I have a certain attitude toward the fact that) Mary might come.’ Non-EN complement (21) Mary-ga Mary-Nom

oci-l-kka/ci. come-Fut-NFComp

‘(I have a certain attitude toward the fact that) Mary might come.’

Furthermore, just as no complementizer can be attached in EV2 constructions in German, a structural case cannot attach to subjunctive complements, EN-complements and EN-elided counterparts.

Non-subjunctive complement (22) John-un [Mary-ga

on-tanunkes-ul]

hwaksinha-koiss-ta.

John-Top Mary-Nom come-Comp-Acc

certain-Asp-Decl

‘John is certain that Mary will come.’

161

[Korean]

Subjunctive complement (23) John-un [Mary-ga

o-assumyenhako(-*lul)]

John-Top Mary-Nom come-SubjComp-Acc

para-koiss-ta. wish-Asp-Decl

‘John wishes that Mary might come.’ EN complement (24) John-un [Mary-ga

oci-anh-ul-kka(-*lul)]

John-Top Mary-Nom come-EN-Fut-NFComp-Acc

kekcengha-koiss-ta. fear-Asp-Decl

‘John fears that Mary might come.’ Non-EN complement (25) John-un [Mary-ga

o-ul-kka(-*lul)]

kekcengha-koiss-ta.

John-Top Mary-Nom come-Fut-NFComp-Acc

fear-Asp-Decl

‘John fears that Mary might come.’

Before closing the discussion on paratactic properties, I will briefly discuss tense dependency of subjunctive complements and EN complements in Korean in order to examine whether they exhibit syntactic transparency, just as subjunctive complements have been claimed to be so in Romance. The results will tell us: (i) whether subjunctive/EN complements in Korean behave like subjunctive complements in Romance; and (ii) whether EN complements can be syntactically and semantically independent. It will be shown that EN complements in Korean have no tense dependency which enhances the assumption of paratactic property. It is widely believed that tense in subjunctive clauses is anaphoric (Anderson 1982; Pica 1984; Everaert 1984; Jakubowicz 1984; Johnson 1985). Picallo (1984, 1985) notes that the lack of independent temporal interpretation renders the assertive root clauses with subjunctive mood

162

ungrammatical in Romance languages (data taken from Quer 1998; see also Giannakidou 2009 for subjunctive as dependent tense).

(26) a. *Daniel haya llamado.

[Spanish]

Daniel call.Subj.Perf.3sg ‘Daniel has called (SUBJ).’ b. *Ahir yesterday

plogués.

[Catalan]

rain.Subj.Impf.3sg

‘Yesterday it rained (SUBJ).’

Due to the tense dependency, subjunctive Infl has been frequently assumed to be [-Tense, +Agr] in the Romance literature. In Japanese, on the other hand, Uchibori (2000) notes that subjunctive complements are not uniform with respect to licensing of temporal adverbs, and that T in a certain group of subjunctives is deficient compared to T in the indicative. It is analogous for subjunctive complements in Korean in which the tense dependency seems to hold only for certain subtypes of subjunctive complements. Unlike certain subjunctive complements, however, no tense dependency is observed in EN complements in Korean. This is supported by two facts: (i) In contrast with certain Japanese subjunctive clauses (Uchibori 2000), EN complements in Korean (and Japanese) allow past adverbs like ecey ‘yesterday’ or kucey ‘the day before yesterday’ and past form on verbs; and (ii) Unlike subjunctive clauses in Romance, tense in EN complements can be independent from that

163

of matrix clauses in Korean. The following data illustrate the availability of distinct temporal adverbs and tense forms on verbs between matrix clauses and EN complements.

(27) a. John-un

[Mary-ka

nayil

oci-anh-ul-ci/kka]

cikum kekcengha-koiss-ta.

John-Top Mary-Nom tomorrow come-Neg-Fut-NFcomp now

fear-Asp.Pres-Decl

‘John fears now that Mary might come tomorrow (although it is unlikely).’ b. John-un

[Mary-ka

ecey

oci-anh-assul-ci/kka]

cikum kekcengha-koiss-ta.

John-Top Mary-Nom yesterday come-Neg-PstPerf-NFcomp now fear-Asp.Pres-Decl ‘John fears now that Mary might have come yesterday (although it is unlikely).’ c. John-un

[Mary-ka

kucey

oci-anh-assul-ci/kka]

John-Top Mary-Nom two days ago ecey

kekcengha-koissess-ta.

yesterday

fear-Asp.Pst-Decl

come-Neg-PstPerf-NFcomp

‘John feared yesterday that Mary might have come the day before yesterday (although it was unlikely).’ d. John-un [Mary-ka

nayil

oci-anh-ul-ci/kka]

ecey

kekcengha-koissess-ta.

John-Top Mary-Nom tomorrow come-Neg-Fut-NFcomp yesterday fear-Asp.Pst-Decl ‘John feared yesterday that Mary might come tomorrow (although it was unlikely).’

In the previous literature it has been claimed that tense dependency in subjunctives is what gives rise to transparency effects such as disjoint reference between the embedded and the matrix subjects (obviation), long-distance anaphoric binding across a CP boundary, and subject raising from a finite domain (Quer 1998). Given this, it is predicted that no such transparency effects

164

would be observed in EN complements for the following reasons: (i) above data show the property of independent temporal interpretation; (ii) subjunctive complements may appear as root clauses in Korean; and (iii) no obviation effects are observed in Korean (though not illustrated for reasons of space). This is a welcome result since it is in accordance with what we have observed in this subsection. According to the results of diagnostics so far, we are leaning toward the assumption that EN complements, non-EN counterparts, and certain subjunctive complements are paratactically connected to the main clauses. A crucial puzzle, however, arises when we test the same data with diagnostics for hypotaxis below.

6.1.2. Hypotactic Properties In this subsection, I will illustrate how EN complements, non-EN counterparts, and subjunctive complements display hypotactic properties that are known to be logically incompatible with above tested paratactic properties. To begin, people have long been aware that extraction from EV2 constructions in Germanic is impossible which has raised the suspicion that EV2 constructions are not hypotactic but paratactic instead.3 In sharp contrast with abovementioned paratactic properties of EN and subjunctive complements, however, the following data show that the extraction of embedded subject kyocangi-i ‘principal-Nom’ is allowed irrespective of complement types in Korean, which implicates their hypotactic property.

It has also been pointed out that topicalisation inside a genuinely embedded clause is generally difficult in Mainland Scandinavian whereas it is unproblematic in an EV2 construction. 3

165

Non-subjunctive complement (28) Kyocang-i i,

Swuni -nun [ti sang-ul

principal-Nom Swuni-Top

cwul-kes-ul]

hwaksinha-koiss-ta.

award-Acc give-Comp-Acc

certain-Asp-Decl

‘The principali, Swuni is certain that (hei) will give an award (to her).’ Subjunctive complement (29) Kyocang-i i,

Swuni -nun [ti sang-ul

principal-Nom Swuni-Top

cwu-essumyenhako] para-koiss-ta.

award-Acc

give-SubjComp

wish-Asp-Comp

‘The principali, Swuni wishes that (hei) might give an award (to her).’ EN complement (30) Kyocang-i i,

Swuni-nun [ti sang-ul

principal-Nom Swuni-Top

cwuci-anh-ul-kka]

award-Acc give-EN-Fut-NFComp

kitayha-koiss-ta. hope-Asp-Comp

‘The principali, Swuni hopes that (hei) might give an award (to her).’ Non-EN complement (31) Kyocang-i i,

Swuni-nun [ti sang-ul

principal-Nom Swuni-Top

award-Acc

cwu-ul-kka]

kitayha-koiss-ta.

give-Fut-NFComp

hope-Asp-Comp

‘The principali, Swuni hopes that (hei) might give an award (to her).’

Furthermore, binding from outside the EV2 constructions in Germanic is known to be banned. With regard to binding, an interesting paradigm is observed in reciprocal anaphor binding in Korean. As illustrated in (32b) below, in the indicative example, a reciprocal anaphor like sero ‘each other’ cannot be bound by the object kutul ‘them’ that has been moved out of the subordinate clause. But such binding becomes available in subjunctive complements, EN-

166

complements and non-EN counterparts, which suggests their hypotactic property. A similar paradigm has been noted in Japanese (Nemeto 1993; Uchibori 1997, 2000 for Japanese).4

Non-subjunctive complement (32) a.?*Seroi-uy

sensayng-i

[John-i

kutuli-ul

pinanhay-ss-tako] malha-ss-ta.

each other’s teacher-Nom Jonh-Nom them-Acc criticize-Pst-Comp say-Pst-Decl ‘*Each other’s teacher said that John criticized them.’ b.?*Kutuli-ul them-Acc

[seroi-uy

sensayng-i

[John-i

ti pinanhay-ss-tako] malha-ss-ta.]

each other’s

teacher-Nom Jonh-Nom criticize-Pst-Comp say-Pst-Decl

‘*Them, each other’s teacher said that John criticized.’ Subjunctive complement (33) a.?*Seroi-uy sensayng-i

[kyocang-i kutuli-ul chwuchenha-yssumyenhako] parayssta.

each other’s teacher-Nom principal-Nom them-Acc recommend-Subj.Comp wish-Pst-Decl ‘*Each other’s teacher wished that the principal recommended them.’ b.?Kutuli-ul seroi-uy sensayng-i

[kyocang-i ti chwuchenha-yssumyenhako] paray-ss-ta.

them-Acc each other’s teacher-Nom principal-Nom recommend-Subj.Comp wish-Pst-Decl ‘*Them, each other’s teacher wished that the principal recommended.’

4

Note that anaphor binding facts do not tell us much in Korean as Progovac (1993) notes that languages with no overt Agr such as Chinese, Japanese and Korean, X0 reflexive can be bound across any number of [NP, NP]s or [NP, IP]s, as in Korean example below (Yang 1983): (i) Johni-un [Billj-i [Maryk-ga [Toml-uy cakii/j/k/l-ey tayhan thayto]-lul silheha-n-ta-ko] John-Top Bill-Nom Mary-Nom Tom’s self toward attitude-Acc hate-Asp-Decl-Comp sayngkakha-n-ta-ko] mit-nun-ta. think-Asp-Decl-Comp believe-Asp-Decl ‘John believes that Bill thinks that Mary hates Tom’s attitude toward self.’ 167

EN complement (34) a.?*Seroi-uy sensayng-i

[kyocang-i

kutuli-ul chwuchenhaci-anh-ul-kka] kitayhakoissta.

each other’s teacher-Nom principal-Nom them-Acc recommend-EN-Fut-NFComp hope-Asp ‘*Each other’s teacher hopes that the principal recommended them.’ b.??Kutuli-ul seroi-uy

sensayng-i [kyocang-i ti chwuchenhaci-anhul-kka] kitayhakoissta.

them-Acc each other’s teacher-Nom principal-Nom recommend-EN-Fut-NFComp hope-Asp ‘*Them, each other’s teacher hopes that the principal recommended.’ Non-EN complement (35) a.?*Seroi-uy sensayng-i

[kyocang-i

kutuli-ul chwuchenha-ul-kka]

kitayhakoissta.

each other’s teacher-Nom principal-Nom them-Acc recommend-Fut-NFComp hope-Asp ‘*Each other’s teacher hopes that the principal recommended them.’ b.??Kutuli-ul seroi-uy

sensayng-i

[kyocang-i ti chwuchenha-ul-kka]

kitayhakoissta.

them-Acc each other’s teacher-Nom principal-Nom recommend-Fut-NFComp hope-Asp ‘*Them, each other’s teacher hopes that the principal recommended.’

Another crucial property of hypotaxis seems to be the availability of multiple embedding, which is shown to be grammatical across all types of complements.

Non-subjunctive complement (36) Kica-nun

[82%-uy saramtuli-i [proi haykotangha-lkes-ul]

journalist-Top 82%-gen people-Nom

get fired-Comp-Acc

potoha-yss-ta. report-Pst-Decl

168

hwaksinha-koiss-tako] certain-Asp-Comp

‘The journalist reported that 82% of people were certain that they would get fired.’

Subjunctive complement (37) Kica-nun

[82%-uy saramtuli-i

journalist-Top 82%-gen people-Nom

[proi sungcinhay-ssumyenhako] para-koiss-tako] get promoted-SubjComp

wish-Asp-Comp

potoha-yss-ta. report-Pst-Decl ‘The journalist reported that 82% of people wished that they would get promoted.’ EN complement (38) Kica-nun

[82%-uy saramtuli-i [proi haykotanghaci-anh-ul-kka] kekcengha-koiss-tako]

journalist-Top 82%-gen people-Nom

get fired-EN-Fut-NFComp fear-Asp-Comp

potoha-yss-ta. report-Pst-Decl ‘The journalist reported that 82% of people feared that they might get fired.’ Non-EN complement (39) Kica-nun

[82%-uy saramtuli-i [proi haykotangha-ul-kka]

journalist-Top 82%-gen people-Nom

get fired-Fut-NFComp

kekcengha-koiss-tako] fear-Asp-Comp

potoha-yss-ta. report-Pst-Decl ‘The journalist reported that 82% of people feared that they might get fired.’

Before searching for a solution for the paradox of hypotaxis and parataxis of EN complements, let’s look at the case that shows ambiguous properties in Korean.

169

Recall that, in 6.1.1, we have seen that a structural case cannot attach to subjunctive complements, EN-complements and non-EN counterparts. The situation, however, is not so simple. Observe that structural case marking becomes available when a nominalizer is inserted inside subjunctive complements, EN-complements and non-EN counterparts, which renders them cases of true embedding.

Subjunctive complement (40) John-un John-Top

[Mary-ga

o-assumyen-hanun-kes(-ul)]

para-koiss-ta.

Mary-Nom come-SubjComp-Rel-NMLZ-Acc

wish-Asp-Decl

‘John wishes that Mary would come.’ EN complement (41) John-un John-Top

[Mary-ga

oci-anh-ul-kka-hanun-kes(-lul)]

kekcengha-koiss-ta.

Mary-Nom come-EN-Fut-NFComp-Rel-NMLZ-Acc fear-Asp-Decl

‘John fears that Mary might come.’ Non-EN complement (42) John-un

[Mary-ga

o-ul-kka-hanunkes(-lul)]

John-Top Mary-Nom come-Fut-NFComp-Rel-NMLZ-Acc

kekcengha-koiss-ta. fear-Asp-Decl

‘John fears that Mary might come.’

The observations so far can be summarized as in Table 6, which illustrates at a glance the dilemma of how the contrasting properties, paratactic and hypotactic properties, are shown to coexist in EN complements and subjunctive complements.

170

Table 6. Interim summary: paratactic vs. hypotactic properties of various complements Diagnostics for parataxis

EN CPs

Subjunctive CPs

Indicative CPs

[1] Does X postpose?

Yes

Yes

No

[2] Can X be prosodically

Yes

Yes

No

Yes

Yes

No

Yes

Yes

No

Yes

Yes

No

independent? [3] Can a speaker-oriented interjection occur in X? [4] Does X occur as a root CP with independent tense? [5]

Does

X

resist

case

marking? Diagnostics for hypotaxis

EN CPs

Subjunctive CPs

Indicative CPs

[6] Is extraction out of X

Yes

Yes

Yes

Yes

Yes

Yes

(reciprocal

(reciprocal

(reciprocal

possible? [7]

Is

binding

into

X

possible?

[8] Can X be embedded under

an

anaphors)

anaphors)

anaphors)

Yes

Yes

No

embedded

predicate? [9] Can X be case-marked?

Yes

(when

nominalized)

Yes

(when

nominalized)

171

Yes

(without

nominalization)

As shown in 6.2.1, subjunctive and EN complements seem to behave as root clauses by allowing postposing, independent prosody of a long pause and/or rising intonation at the end of complements, speaker-oriented interjections like ‘gosh’, independent tense marking, and occurrence as root CPs only. Surprisingly, however, 6.2.2 demonstrates that subjunctive and EN complements display hypotactic properties as well. It is shown that extraction out of subordinate clauses, reciprocal anaphor binding, multiple embedding, and case marking (after being nominalized) are permitted in subjunctive and EN complements in Korean. The results of these individual tests only show that the structure involving subjunctive and EN complements so far are quite puzzling.

6.1.3. Combined applications of hypotactic and paratactic diagnostics In order to understand what is really going on in subjunctive and EN constructions, it will be helpful to combine paratactic and hypotactic tests to apply on one example (gratia Jason Merchant). First, as already seen, the availability of extraction of embedded subject kyocang-ii ‘principal-Nom’ in (43) confirms that the syntactic relation involved in the EN complement is hypotaxis. And recall that paratactic tests such as postposing, rising tone, and speaker-oriented interjection are all shown to be grammatical in EN complements. Crucially, however, (44)-(46) illustrate that, within the sentence with the extraction of embedded subject, these paratactic properties completely disappear and postposing, rising tone, and speaker-oriented interjection become no longer applicable.

172

[6] extraction (43) Kyocang-i i,

Swuni-nun

principal-Nom Swuni-Top

[ti sang-ul award-Acc

cwuci-anh-ulkka]

kitayha-koiss-ta.

give-EN-NFComp

hope-Asp-Comp

‘The principali, Swuni hopes that (hei) might give an award (to her).’ [6+1] extraction + postposing (44) *Kyocang-i i,

Swuni-nun

principal-Nom Swuni-Top

kitayha-koiss-ta, [ti sang-ul hope-Asp-Comp

cwuci-anh-ulkka].

award-Acc give-EN-NFComp

‘The principali, Swuni hopes, (hei) might give an award (to her).’ [6+2] extraction + prosodic independence (rising tone) (45) *Kyocang-i i,

Swuni-nun [ti sang-ul

principal-Nom Swuni-Top

cwuci-anh-ulkka(↑)]

award-Acc give-EN-NFComp

kitayha-koiss-ta.

hope-Asp-Comp

‘The principali, Swuni hopes, (hei) might give an award (to her)?’ [6+3] extraction + speaker-oriented interjection (46) *Kyocang-i i,

Swuni-nun [ti saysangeyna sang-ul cwuci-anh-ulkka] kitayha-koiss-ta.

principal-Nom Swuni-Top

gosh

award-Acc give-EN-NFComp hope-Asp-Comp

‘The principali, Swuni hopes that gosh (hei) might give an award (to her).’

Furthermore, the dual embedding of the (underlined) EN complement below indicates their hypotactic structure. Just like above cases, the paratactic properties such as postposing, rising tone, and speaker-oriented interjection turn out unavailable in this examples.5

I will skip the discussion of data that involves reciprocal anaphor binding since the data would become extremely subtle to judge when they are combined with the parataxis tests. 5

173

[8] dual embedding (47) Kica-nun

[82%-uy saramtuli-i [proi haykotanghaci-anh-ulkka] kekcengha-koiss-tako]

journalist-Top 82%-gen people-Nom

get fired-EN-NFComp

fear-Asp-Comp

potoha-yss-ta. report-Pst-Decl ‘The journalist reported that 82% of people feared that they might get fired.’ [8+1] dual embedding + postposing (48) *Kica-nun

[82%-uy saramtuli -i

[t]j

kekcengha-koiss-tako]

journalist-Top 82%-gen people-Nom potoha-yss-ta,

fear-Asp-Comp

[proi haykotanghaci-anh-ulkka]j.

report-Pst-Decl

get fired-EN-NFComp

‘The journalist reported that 82% of people feared, they might get fired.’ [8+2] dual embedding + prosodic independence (rising tone) (49) *Kica-nun

[82%-uy saramtuli -i [proi haykotanghaci-anh-ulkka(↑)] kekcengha-koiss-tako]

journalist-Top 82%-gen people-Nom

get fired-EN-NFComp

fear-Asp-Comp

potoha-yss-ta. report-Pst-Decl ‘The journalist reported that 82% of people feared, they might get fired?’ [8+3] dual embedding + speaker-oriented interjection (50) *Kica-nun journalist-Top

[82%-uy saramtuli -i

[proi (mapsosa) haykotanghaci-anh-ulkka]

82%-gen people-Nom

kekcengha-koiss-tako]

potoha-yss-ta.

fear-Asp-Comp

report-Pst-Decl

gosh

174

get fired-EN-NFComp

‘The journalist reported that 82% of people feared that gosh they might get fired.’

Likewise, if EN complements receive a structural case (ACC) as in (51), the relevant structure must be hypotaxis. Again, notice that the addition of the hallmarks of parataxis makes the sentences terribly unacceptable.

[8] case marking (51) Toli-nun Toli-Top

[Swuni-ga

haykotanghaci-anh-ulkka-hanunkes]-ul kekcengha-koiss-ta.

Swuni-Nom get fired-EN-NFComp-NMLZ-Acc

fear-Asp-Comp

‘Toli feared that Swuni might get fired.’ [8+1] case marking + postposing (52) *Toli-nun Toli-Top

kekcengha-koiss-ta, [Swuni-ga fear-Asp-Comp

haykotanghaci-anh-ulkka-hanunkes]-ul.

Swuni-Nom get fired-EN-NFComp-NMLZ-Acc

‘Toli feared, Swuni might get fired.’ [8+2] case marking + prosodic independence (rising tone) (53) *Toli -nun Toli-Top

[Swuni-ga haykotanghaci-anh-ulkka-hanunkes]-ul(↑) kekcengha-koiss-ta. Swuni-Nom get fired-EN-NFComp-NMLZ-Acc

fear-Asp-Comp

‘Toli feared, Swuni might get fired?’ [8+3] case marking + speaker-oriented interjection (54) *Toli-nun [Swuni-ga mapsosa haykotanghaci-anh-ulkka-hanunkes]-ul kekcengha-koiss-ta. Toli-Top

Swuni-Nom gosh

get fired-EN-NFComp-NMLZ-Acc

‘Toli feared that gosh Swuni might get fired.’

175

fear-Asp-Comp

Though not all of the examples are illustrated here for reasons of space, it is in fact the case that the results of these combined tests to subjunctive complements will be generally analogous. The results of these combined tests implicate the following: For one thing, the conflicting facts that we observed in 6.2.1 and 6.2.2 turn out not so puzzling after all, since it was shown in 6.2.3 that they are incompatible within one sentence. I take these facts to argue that EN complements are structurally ambiguous in the sense that they can be either hypotactically or paratactically connected to their matrix clauses in Korean. In the next section, I will discuss how this structural ambiguity can be captured in the syntactic derivation within the minimalist framework.

6.2. Syntactic Derivations Section 6.1 has demonstrated the ambiguous syntactic properties in Korean EN complements (and potentially subjunctive complements). As briefly discussed in 6.2.1, the paratactic properties of apparently embedded complements are not extraordinary since similar facts are observed across languages including Germanic (Hoeksema and Napoli 1993; de Haan 2001), Romance (Uriagereka and Torrego 2002), and Balkan (Mari and Martin 2007). For instance, besides the abovementioned paratactic properties such as postposing, rising tone, and speaker-oriented interjections, EN complements in Korean behave like EV2 constructions in Germanic in the following respects: a) just as EV2 in Germanic do not involve a complementizer, EN complements in Korean cannot receive structural case without being preceded by nominalization; b) just as the verb of the second clause with paratactic nature is in the subjunctive, EN complements in Korean are shown to be in subjunctive; and c) just as EV2 in

176

German is strikingly similar to direct quotation, EN complements in Korean are shown to be strongly similar to direct quotation of a question. More crucially, the paratactic properties of EN complements in Korean are strongly reminiscent of the paradigm of como ‘how’ constructions in Spanish (Uriagereka and Torrego 2002).

(55) a. Juan explicó

la verdad de que

Juan explained the truth that

la tierra es redonda.

[Spanish]

the earth is round.

b. Juan explicó como la tierra es redonda. Juan explained how

the earth is round.

They argue that since the como-clauses are paratactically connected, their distribution is much more restricted than that of the hypotactic que ‘that’ clauses. For instance, nouns, adjectives in (56) and prepositions in (57) do not take como-clauses.

(56) a. No me gusta la idea/el hecho ?(de) que/*como I don’t like the idea/fact

that/*how …

b. Estoy harto ?(de) que/*como … I’m fed up

that/*how …

(57) a. Para que/*como so that/*how c. Desde

que/*como

since that/*how

b. Con que/*como in as much as that/*how d. Entre que/*como while that/*how

177

[Spanish]

Just like the como-clauses, the distribution of EN complements (and subjunctive complements) in Korean seems to be quite restricted. Observe the fact that the attachment of ttaymwuney ‘since’ is only available in indicative but not in subjunctive and EN complements. It is analogous for other kinds of prepositions.

Non-subjunctive complement (58) Mary-ga

onun-kes-ttaymwuney …

[Korean]

Mary-Nom come-Comp-since ‘Due to the fact that Mary will come…’ Subjunctive complement (59) Mary-ga Mary-Nom

o-assumyenhako-*ttaymwuney … come-SubjComp-Acc-since

‘Due to the case that Mary might come…’ EN complement (60) Mary-ga

oci-anh-ul-kka-*ttaymwuney …

Mary-Nom come-EN-Fut-NFComp-since ‘Due to the fact that Mary will come…’

Furthermore, just as the paratactic properties of como clauses are shown by the unavailability of predicate raising, Neg-raising, and polarity item licensing, EN complements in Korean are shown to be opaque in reciprocal anaphor binding. Neg-raising and polarity item licensing are also banned across EN complements in Korean, though not illustrated here.

178

In order to account for how the paratactic properties of EN complements arise, I am going to follow what has been assumed for the constructions with similar properties. In particular, I will adopt Uriagereka and Torrego’s (2002) analysis of como clauses. Focusing on the fact that the genitive marker de can only be associated with the clausal dependents of nominal la verdad ‘the truth’ but not with como, they argue that the structural difference of these two in (61) correspond to that of those in (62). Based on Kayne-Szabolcsi’s analysis that John’s is higher in (62b) than it is in (62a) in the overt syntax, they assume that the dependent clause in (61b) is structurally higher than that in (61a).

(61) a. … la verdad *(de) que la tierra es redonda.

[Spanish]

the truth (of) that the earth is round. b. … como (*de) la tierra es redonda. how

(of) the earth is round.

(62) a. … the sister *(of) John’s b. … John’s (*of) sister

They further note that within the minimalist system, the higher element needs a reason to move by Spell-out in contrast to the lower element that procrastinated. Thus they posit a strong feature for the structures with como in which the moved element is attracted to the strong feature, unliked the structures with la verdad. They further assume that only the D element for como structures selects for a functional category with a strong feature whereas the D heading for other structures does not. Given this, Uriagereka and Torrego’s (2002) assume the following Root Transformation for como clauses that is in the spirit of Lebeaux (1988):

179

(63) Root Transformation (Uriagereka and Torrego 2002) a.

DP

…

D

AgrP

D co-mo

Agr’

pro

Agr

…

CP

b. Juan explicó

como la tierra es redonda.

Juan explained how the earth is round.

Uriagereka and Torrego propose that the generalized root transformation is responsible for paratactic dependencies in como clauses in Spanish. More specifically, they posit that como complements may have a null pro-like expression that is licensed in discourse, just like null pronominals. A pro-clause is assumed to originally occupy the place which can be also taken by an entire clause (CP). They further assume that the pro-clause is what enters into the syntactic derivation, engaging in checking just like other syntactic formative. The structural diversion

180

arises at LF where they posit the following two options: i) pro remains as such; or ii) a separate sentence (CP above) substitutes into the pro-clause. Based on the parallel with como clauses in Spanish, it is plausible to assume the availability of dynamic syntactic derivation involving Root Transformation for subjunctive and EN constructions in Korean. This then allows us to account for why a structural case cannot attach to subjunctive complements, EN-complements and non-EN counterparts.

Non-subjunctive complement (64) John-un [Mary-ga

on-tanunkes-ul]

hwaksinha-koiss-ta.

John-Top Mary-Nom come-Comp-Acc

[Korean]

certain-Asp-Decl

‘John is certain that Mary will come.’ Subjunctive complement (65) John-un [Mary-ga

o-assumyenhako(-*lul)]

pro-Acc

John-Top Mary-Nom come-SubjComp-Acc

para-koiss-ta. wish-Asp-Decl

‘John wishes that Mary might come.’ EN complement (66) John-un [Mary-ga

oci-anh-ul-kka(-*lul)]

pro-Acc

John-Top Mary-Nom come-EN-Fut-NFComp-Acc

kekcengha-koiss-ta. fear-Asp-Decl

‘John fears that Mary might come.’ Non-EN complement (67) John-un [Mary-ga

o-ul-kka(-*lul)]

pro-Acc

John-Top Mary-Nom come-Fut-NFComp-Acc

181

kekcengha-koiss-ta. fear-Asp-Decl

‘John fears that Mary might come.’

So the idea is that the matrix verb in parataxis constructions takes a pro complement to which it assigns ACC, hence no ACC can be assigned to the clause. Following the assumption of Root Transformation, the syntactic derivation of above subjunctive complements, EN-complements and non-EN counterparts can be represented as follows:

(68) Syntactic derivation of Paratactic EN complements a. TP

TP

John-un John-Top VP

CP

pro

Mary-ga

oci-anh-ul-kka

Mary-Nom

come-EN-Fut-NFComp

T

V

182

b. EN complement John-un [Mary-ga

oci-anh-ul-kka(-*lul)]

pro-Acc

John-Top Mary-Nom come-EN-Fut-NFComp-Acc

kekcengha-koiss-ta. fear-Asp-Decl

‘John fears that Mary might come.’

This structure accounts for the reason why the above EN complements (and subjunctive complements) seem to behave as root clauses. Since the EN complement is paratactically connected to the matrix clause, it is predicated to exhibit the properties of an independent clause such allowing postposing, independent prosody of a long pause and/or rising intonation at the end of complements, speaker-oriented interjections like ‘gosh’, independent tense marking, and occurrence as root CPs only. Note that, due to the word order in Korean that the subordinate clause is center-embedded in a sentence, this paratactic structure does not follow the ‘no tangling condition’ of generalized syntactic structures. Rather, the current structure is to some extent in line with syntactic approaches assuming multidominance (cf. other recent reconceptualizations of phrase structure that rethink the traditional constraints such as proposals that counternace multidominace, McCawley 1980; Bachrach and Katzir 2006; Johnson 2009, to appear). Such paratactic CPs are apparently subject to a condition in Korean that requires that they be linearized adjacent to the pro to which they are linked. However such a condition is stated (indeed, however such adjacency conditions are to be understood in general), its effects here are clear, and are sufficient to derive the attested word order. Naturally, the remaining challenge is to understand under just what conditions such linearization constraints emerge, how exactly they

183

are to be stated, and how they are to be incorporated into a general model of grammar (see Fox and Pesetsky 2004 for one such proposal). On the other hand, recall that structural case marking becomes available when a nominalizer is inserted inside subjunctive complements, EN-complements and non-EN counterparts, which render them true embedding.

Subjunctive complement (69) John-un [Mary-ga

o-assumyen-hanun-kes(-ul)]

John-Top Mary-Nom come-SubjComp-Rel-NMLZ-Acc

para-koiss-ta. wish-Asp-Decl

‘John has a wish that Mary would come.’ EN complement (70) John-un [Mary-ga

oci-anh-ul-kka-hanun-kes(-lul)]

kekcengha-koiss-ta.

John-Top Mary-Nom come-EN-Fut-NFComp-Rel-NMLZ-Acc fear-Asp-Decl ‘John fears that Mary might come.’ Non-EN complement (71) John-un [Mary-ga

o-ul-kka-hanun-kes(-lul)]

kekcengha-koiss-ta.

John-Top Mary-Nom come-Fut-NFComp-Rel-NMLZ-Acc fear-Asp-Decl ‘John fears that Mary might come.’

This can be understood as a case where the structure is hypotaxis, just like que ‘that’ complements in Spanish. Thus these complements are expected to exhibit the properties of hypotaxis such as extraction, reciprocal anaphor binding, and multiple embedding.

184

(72) Syntactic derivation of Hypotactic EN complements a.

TP

John-un John-Top VP

CP

Mary-ga

T

V

oci-anh-ul-kka(-hanun-kes-ul)

Mary-Nom come-EN-Fut-NFComp-Rel-NMLZ-Acc b. EN complement John-un [Mary-ga

oci-anh-ul-kka-hanun-kes(-lul)]

kekcengha-koiss-ta.

John-Top Mary-Nom come-EN-Fut-NFComp-Rel-NMLZ-Acc fear-Asp-Decl ‘John fears that Mary might come.’

Due to the structural ambiguity that admits the above structure as another legitimate option, the mystery of hypotacic properties is resolved. When the CP is truly embedded as above, the extraction out of subordinate clauses, reciprocal anaphor binding, multiple embedding, and case marking (after being nominalized) are expected to be permitted in subjunctive and EN complements in Korean.

185

An important difference from the como-clause in Spanish is then that EN complements in Korean allows for not only the paratactic structures but also the hypotactic structures as well. Furthermore, in Korean the hypotactic dependent can be optionally marked by nominalizer in which case the subordinate clauses must receive ACC case. However, since EN complements do not generally receive the ACC case, whether or not the substitution has occurred is indistinguishable on the surface, which means the ambiguity that is observed in 6.2.1 and 6.2.2. When it comes to the reason why a pro-clause is impossible in hypotactic dependents, Uriagereka and Torrego treat it as a matter of pro licensing, noting the following:

(73) Uriagereka and Torrego (2002: 262) “In our terms, pro-clauses are licensed only in the Spec of an AgrP associated to a strong, point-of-view dependent Sigma head. There are no pro-clauses elsewhere, anymore than pro items in general do not appear other than associated to AgrP Specs whose head has the appropriate characteristics (in terms of strength or whatever else is relevant). The reason why parataxis is so restricted is straightforward. It requires the presence of a proform, which is itself extremely restricted, in familiar ways.”

This further implicates that, from a semantic perspective, the availability of paratactic structure of EN clauses in Korean is a welcome result. More importantly, it is strongly reminiscent of previous insights on why certain configurations are more paratactically connected to the main clause. It has been frequently noted that such configurations typically involve the case where a subject has ‘attitude toward the veridicality’ of embedded proposition (Higginbotham 1988; Hinzen 2003; Uriagereka 2008; see Dayal and Grimshaw 2009 for a pragmatic account). And

186

this is precisely the intuition that I have shown so far in EN constructions crosslinguistically. The facts of EN complements in Korean furthermore support that the concept of root transformation must be understood at the syntax-semantics/pragmatics interface.

6.3. Subjunctive mood and Evaluative feature in Korean (and Japanese) In this section, I will move onto anther question: where can the evaluative feature be posited? As is well-known, there is crosslinguistic variation as to where mood is marked. Mood can be reflected in morphology on the verb (e.g. French, Spanish, Catalan), a designated complementizer (e.g. Greek, Korean, Japanese), or both (e.g. Old/Middle English), and the innovation of the current work is adding another element to the potential mood markers, namely EN.

6.3.1. Complementizer choice as mood choice The subjunctive mood information of the embedded clause is reflected in complementizer choice in some languages (e.g. na in Greek). More interestingly, van der Wouden (1997: 196) notes that certain negative complementizers such as ne in Latin, lest in English, and mipos in Modern Greek, behave as EN themselves, thus triggering an evaluative flavor.

(74) a. Timeo I-fear

ne

veniat.

[Latin]

that-not

come.Subj.3s

‘I’m afraid he may come.’ b. Then fearing lest we should have fallen upon rocks, they cast four anchors out of the stern, and wished for the day.

(Acts 27:29)

187

c. Fobamai mipos

kano

lathos.

[modern Greek] (Ruge 1986)

fear-1sg that-not make-1sg error ‘I am afraid of making a mistake.’

Furthermore, it is noted that complementizers play an important role in EN phenomena (see Espinal 1992; Brown 1999a, b; Brown and Franks 1995, 1997; Abels 2003). To illustrate, the choice of a specific complementizer determines whether a negative element is EN or real negation in Russian (Brown 1999a, b; Abels 2003).

(75) *Ja bojus’ I fear

{kak by

/ čtoby}

how Mod

nikto

that

ne opozdal.

[Russian]

NI-who not was-late

Intended: ‘I fear that somebody might be late.’ (76) √Ja bojus’ I fear

čto

nikto

ne opozdal.

that

NI-who

not was-late

‘I’m afraid that nobody was late.’

Since there is a NPI, nikto, (75) is ungrammatical because the complementizer kak by or čtoby shows that ne is EN which is unable to license a NPI, whereas (76) is grammatical because the complementizer čto indicates that ne is real negation which licenses a NPI. Furthermore, another important contrast concerning complementizers is shown by the following data (Abels 2003).

188

(77) a. *Ne znaet not knows

li nikto

iz vas

kak èto delaetsja?

li NI-who.Nom

of you.pl

how this do-self

[Russian]

‘Don’t any of you know how to do this?’ b. *Ne znaet not knows

nikto

iz vas

kak èto delaetsja?

NI-who.Nom of you.pl

how this do-self

‘Don’t any of you know how to do this?’ c. √Ne znaet not knows

nikto

iz vas

kak èto delaetsja!

NI-who.Nom of you.PL

how this do-self

‘None of you know how to do this!’

Abels notes that the crucial contrast between (77a,b) and (77c) originates from the presence or absence of the interrogative complementizer li, assuming that a silent interrogative complementizer is employed in (77b). That is, the NPI nikto is unlicensed by ne in (77a,b) because ne is EN under the interrogative complementizer li; nikto is licensed by ne in (77c) because ne is real negation when it has declarative intonation. Abels takes these examples to suggest that the specific choice of complementizer (interrogative vs. declarative) matters in identifying the status of EN. Notice that this fact in Russian is strongly reminiscent of our earlier observations in Korean and Japanese, repeated below.

(78) NF-Complementizer for EN sentences a. John-un

Mary-ka

oci-anh-ul-{ci/kka}6

kitayha-ko issta.

[Korean]

As noted ealier, ci and kka are formal and informal variants of non-factive complementizers, respectively. 6

189

John-Top

Mary-Nom

come-Neg-Fut-NFcomp hope-Asp

‘John hopes that Mary might come.’ b. John-wa John-Top

Mary-ga

ko-nai-ka(-to)

kitaisi-te iru.

Mary-Nom

come-Neg-NFcomp

hope-Asp

[Japanese]

‘John hopes that Mary might come.’

In (78), the non-factive complementizer disambiguates the status of the negative element anh/nai as EN. As already noted, if it is replaced with the regular factive complementizer, anh/nai must be interpreted as real negation. Here an interesting parallel with Russian arises: just like li in Russian, the non-factive complementizer ci/kka in Korean and ka-to in Japanese is identical in form with a question marker. This leads us to assume that these non-factive complementizers in Korean and Japanese are another type of subjunctive mood marker, hence the possible locus of an evaluative feature. The role of the complementizer seems more crucial in these languages which lack inflectional mood morphology on verbs. Let me further explore the mood distinction on complementizers in Korean and Japanese in comparison to the mood distinction by verbal inflection in French and Catalan. For French, Siegel (2009) notes that verbs like voir ‘see’ must take the indicative mood while vouloir ‘want’ must take the subjunctive. In languages with mood morphology, the mood requirement is satisfied by inflection on the embedded verbs, as illustrated below.

(79) Je vois I see.Pres.1sg

que

tu

{vas/*ailles}

bien.

that

you

go.Pres.2sg.Ind/Subj

well

‘I see that you are well.’

190

(Siegel 2009)[French]

(80) Ils they

veulent

que tu

{*vas/ailles}

avec eux.

want.Pres.3pl

that you

go.Pres.2sg.Ind/Subj

with them

‘They want you to go with them.’

Even in languages without apparent verbal inflection of mood, such a mood selection tendency seems to hold, albeit less strictly, via another means: In languages like Korean and Japanese, the mood distinction is manifested by complementizer choice, instead of verbal inflection. Korean, for instance, exhibits a variety of complementizers that can be classified into indicative and subjunctive mood: for instance, factive complementizers like tanunkesul ‘the fact/thing that (factive complementizer)’ appear only in typical indicative mood environments, whereas subjunctive complementizers include essumeynhako

‘wishing that (desiderative

subjunctive complementizer)’, essumeyneccenahako ‘for fear that (anti-desiderative subjunctive complementizer)’, and ci/kka ‘indeterminate complementizer’ that we saw in EN sentences. As illustrated below, the verb al ‘see/know’ typically co-occurs with the indicative mood, which is represented by the factive complementizer tanunkesul. In contrast, the subjunctive verb para ‘wish’ normally occurs with the optative complementizer essumeynhako.

(81) Ney-ka

cal

you-Nom well

cinayn-{tanunkesul/*essumyenhako}

al-n-ta.

be-Comp.Ind/Comp.Subj

know-Pres-Decl

‘I know that you are well.’ (82) Kutul-un ney-ka

hamkkey ka-{*tanunkesul/essumyenhako}

they-Top you-Nom together

go-Comp.Ind/Comp.Subj

191

para-n-ta. wish-Pres-Decl

‘They wish that you would go with them.’

Second, mood distinction in languages with verbal mood inflection has been diagnosed by the following continuation possibilities (Quer 1998; Siegel 2009), which show that, unlike the subjunctive in (83a), the indicative mood in (83b) cannot be used when the speaker serves as the individual anchor (does not commit to the truth of the utterance).

(83) a. El degà no creu [que els estudiants es mereixin un premi].

(Quer 1998) [Catalan]

‘The dean does not believe that the students deserve(SUBJ) a prize.’ i. … I jo tampoc no hoc crec ‘and I do not believe it either.’ ii. … però jo crec que sí ‘but I believe they do.’ b. El degà no creu [que els estudiants es mereixen un premi]. ‘The dean does not believe that the students deserve(IND) a prize.’ i. #… I jo tampoc no hoc crec ‘and I do not believe it either.’ ii. … però jo crec que sí ‘but I believe they do.’

This test holds in the following examples in Korean in which mood distinction is shown by the complementizer selection between ci and tanunkesul. Just like the Catalan examples above, only

192

the indicative mood marked by tanunkesul in (84b) cannot be followed by a comment like “and I do not believe it either.”

(84) a. Hakcang-un ku haksayngtul-i sangpatul cakyekiissnun-ci dean-Top

the students-Nom awarded

mitci-anh-nun-ta.

deserve-Comp.Subj believe-Neg-Pres-Decl

‘The dean does not believe that the students deserve (SUBJ) a prize.’ i. …

Na-to mit-ci anh-nun-ta. ‘and I do not believe it either.’

ii. … Na-nun mit-nun-ta. ‘but I believe they do.’ b. Hakcang-un ku haksayngtul-i sangpatul cakyekiiss-tanunkesul mitci-anh-nun-ta. dean-Top

the students-Nom awarded deserve-Comp.Ind

believe-Neg-Pres-Decl

‘The dean does not believe that the students deserve (IND) a prize.’ i. # … Na-to mit-ci anh-nun-ta. ‘and I do not believe it either.’ ii. … Na-nun mit-nun-ta. ‘but I believe they do.’

[Korean]

Given the observations so far, I conclude that these complementizers in Korean seem to be legitimate counterparts to verbal mood inflections in Romance. I furthermore assume that, irrespective of the locus, these mood markers possess their own evaluative feature. Japanese is another language where complementizers perform the task of what mood morphemes do in other languages. Besides the uncertainty complementizer ka(to) that is given in

193

the EN example (84b) above, Uchibori (2000) discusses the syntactic behaviors of subjunctive complementizer yoonito. 7 Thus I will assume that ka(to) and yoonito are subjunctive complementizers containing an evaluative feature. 8

6.3.2. Matrix predicate & Subjunctive morphology In languages like Spanish, the subjunctive mood in embedded clauses is generally assumed to be lexically selected by certain matrix predicates such as “predicates of desire (e.g. ‘want’, ‘prefer’, ‘fear’), emotive factive predicates (e.g. ‘regret’, ‘be glad’, ‘be surprised’), modals (e.g. ‘it is

7

Note that Uchibori (2000) also discusses koto as subjunctive complementizer. The status of koto, however, seems controversial or ambiguous: there are three representative complemenatizers that have been extensively discussed in the literature – koto, no, and to. Watanabe (1984), for instance, notes that complementizer distinction in Japanese is based on the degree of a speaker’s certainty about the proposition uttered in the complement and the degree of evidentiary support for that proposition. He suggests that the distribution of to, koto, and no is inherently scalar: First, to denotes the weakest evidential proposition, koto denotes stronger evidence, and no denotes the strongest evidentiality. In particular, to typically appears in direct or indirect quotation or in a complement clause of an event, state, or action that tends to be irrealis, being typically selected by ‘think’, ‘hope’, ‘predict’, or ‘think wrongly’. This indicates that the proposition governed by to lacks a factive presupposition. Second, the use of koto ranges from ‘weak epistemic certainty’ to ‘high emotive’ (Givón 1980), being selected by ‘tell’, ‘know’, ‘discover’, ‘expect’, or ‘feel’. Finally, no appears in a narrow domain of cognitive utterance verbs such as be ‘certain’, ‘firmly believe’, ‘wait’, ‘remember’, ‘feel’, or ‘like’. Furthermore, Kuno (1973; see also Shibitani 1978) suggests a dichotomy between koto/no and to. On the one hand, both koto and no are “factive” complementizers which entail the speaker’s presupposition about an action, state, or event embedded under a matrix sentence. Koto entails a presupposition about an abstract concept, and no entails a presupposition about a concrete event. On the other hand, to is a “non-factive” complementizer which does not entail the speaker’s presupposition about a fact. N. A. McCawley (1978) further notes that complementizer choice in Japanese involves an invisible hierarchy of truth: No indicates a direct physical perception and is selected by verbs like ‘see’, koto indicates an indirect perception (Joseph 1976), and to is selected by counterfactual verbs like omoikomu ‘think wrongly’. Perhaps the property of indicating indirect perception is what makes koto ambiguous between indicative and subjunctive complementizer. 8 Although there are other pragmatic factors that come into play in complementizer selection in Japanese which makes the selection process dynamic and complex, pragmatic variability is not considered at this point (see Ono 2005, Suzuki 2000 for relevant discussion). 194

possible’, ‘it is necessary’), predicates expressing doubt (e.g. ‘doubt’), directives (e.g. ‘order’, ‘advise’, ‘suggest’) and causatives (e.g. ‘make’, ‘achieve’)” (Villalta 2006, 2008).

(85) Victoria quiere que Marcela

venga

al picnic.

Victoria wants that Marcela come.Pres.Subj.3sg

[Spanish]

to-the picnic.

‘Victoria wants Marcela to come to the picnic.’

Likewise in Korean, such a selection relation seems to a certain extent to hold. Desiderative subjunctive complementizers like essumyenhako ‘wishing that’ and anti-desiderative complementizers like essumyeneccena(hako) ‘for fear that’ distribute only with subjunctive predicates in (86) that typically licenses EN, as illustrated in the table 7 below.

(86) Subjunctive verbs in Korean a. volitional:

hyimangha ‘hope’, kitayha ‘hope’

b. verbs of fear:

twureweha ‘fear’, kekcengha ‘worry’

c. directives:

chwungkoha ‘advise’, ceyanha ‘suggest’

Table 7. Selection relation between subjunctive verbs and complementizers Indicative Comp

Subjunctive Comp

EN Comp

-tanunkesul

-essumyenhako/

-cianhulci

-essumyeneccena a. volitional

*

√

√

b. verbs of fear

*

√

√

c. directives

*

√

√

195

Note further that the indicative complementizer tanunkesul co-occurs with indicative predicates below that can never license EN, as summarized in the table 8.

(87) Indicative verbs in Korean a. assertive:

malha ‘say’, ilk ‘read’, cwucangha ‘claim’

b. epistemics:

mit ‘believe’

c. factive verbs:

kippu ‘glad’, al ‘know’, hwuhoyha ‘regret’

d. semifactives:

kkaytat ‘discover’, kiekha ‘remember’

(e. fiction verbs:

kkwumkkwu ‘dream’, sangsangha ‘imagine’)

Table 8. Selection relation between indicative verbs and complementizers Indicative Comp

Subjunctive Comp

EN Comp

-tanunkesul

-essumyenhako/

-cianhulci

-essumyeneccena a. assertive

√

* (√ for ’say’)

*

b. epistemics

√

*

*

c. factive verbs

√

*

*

d. semifactives

√

*

*

At this point an important question arises. Given the two elements: (i) the subjunctive-selecting predicate in the matrix clause and (ii) the subjunctive mood inflection on the embedded verb in Romance or the subjunctive complementizer in Korean and Japanese, do we need to assume one

196

common semantics that is somehow spread over the two elements, or are two separate semantics required for each element? The widely held assumption in Romance seems to be that the semantics of lexically selected subjunctive with desiderative or directive meaning originates from the semantic properties of the matrix predicate. For instance, Villalta’s (2006) analysis of Spanish subjunctive mood is closer to the former option: The comparative semantics, i.e. evaluative semantics, is posited only for the matrix predicate quiere ‘want’, and the role of the subjunctive morpheme is assumed to be an operator which introduces the alternative semantic values for the matrix predicate. Based on Rooth’s assumption on focus element like only, she assumes that the subjunctive morpheme is thus a ~-operator which has the role of evaluating the alternatives for the matrix predicate. The assumption that the main semantics of subjunctive is derived from the matrix predicate, however, has limitations in accounting for the cases where the selection of either mood is allowed and, more importantly, the subjunctive morpheme in the embedded clause seems to trigger its own evaluative content. For instance, Giorgi and Pianesi (2002) note that in Italian verbs of communication normally select for the indicative but when they select for the subjunctive, the interpretation of the matrix verb has the flavor of an evidential, triggering the source information of the content in the embedded clause, as illustrated in (88).

(88) Dicono che sia una stupida.

[Italian]

‘They say that she is (SUBJ) stupid.’

197

In favor of Villalta’s system, one may still want to argue that these particular verbs are ambiguous and the one with evidential meaning must select for the subjunctive. Observe, however, that in the following data in German and Dutch, the subjunctive is also used in an evidential way, but the subjunctive here appears in a root, non-subordinate environment. Hence the subjunctive isn’t in any obvious sense a selected form. (M. den Dikken p.c.).9

(89) a. Die Regierung von Thailand sei heute Morgen gefallen.

[German]

‘The government of Thailand would have fallen (i.e., reportedly fell) this morning.’ b. De regering van Thailand zou vanochtend zijn gevallen.

[Dutch]

Furthermore, Villalta’s assumption that subjunctive marker only does the job of triggering alternatives becomes more problematic when we look at Korean. In Korean also, verbs of communication may select for a subjunctive complementizer in which case the meaning of verb ‘say’ becomes desiderative, and the ambiguity assumption may be able to account for this example as well.

(90) Kutul-un ney-ka

hamkkey ka-essumyenhako

they-Top you-Nom together

go-Comp.Subj

malhay-ss-ta. say-Pst-Decl

‘They expressed their wish that you would go with them.’

9

Though Dutch doesn't have a subjunctive but it mimics the effects of the subjunctive in other languages with the aid of the modal "zou" in (15c) (M. den Dikken p.c.). 198

Crucially, however, the following example without a higher predicate shows that the locus of the desiderative sense in Korean can also be the subjunctive morpheme essumyenha on the complementizer essumyenhako.

(91) Ney-ka

hamkkey

you-Nom together

ka-essumyenha-y. go-Subj-Decl.

‘I wish that you would go (with me).’

This leads us to assume that, though the selection relation usually holds between the higher predicates and the lower subjunctive complementizers or verbal morphemes, this does not necessarily indicate that the semantics shared between them originates only from the matrix predicate. This point was also made by the above Dutch and German cases. As shown above, when the selection is optional between indicative and subjunctive, the subjunctive complementizers or verbal morphemes in the embedded clause plays a crucial role in giving rise to subjunctive sense such as evidential, desiderative, etc. Thus I will assume that subjunctive markers in embedded clauses, complementizers or morphology on verbs, may contain an evaluative feature that gives rise to its own evaluative semantics and it can be independent from the features in a matrix predicate. The idea is that the evaluative feature in the subjunctive predicate like ‘want’ encodes the semantics of desirability scale which typically undergoes an Agree operation with subjunctive marker in an embedded clause. In optional cases like ‘say’, on the other hand, its evaluative feature is uninterpretable until it is checked by the subjunctive marker in the embedded clause, as shown in (91). The argument for the independent evaluative feature of subjunctive makers will be further supported when we look at examples with EN.

199

6.3.3. Evaluative Negation with evaluative feature Recall that in chapter 5 I proposed that multiple occurrences of evaluative elements such as EN and subjunctive mood markers can hold their own evaluative content by assuming a refined version of multidimensionality of Potts’s (2005, 2007) CI logic that is proposed in Giannakidou and Yoon (2010, to apper). In particular, I have shown that the kinds of evaluative effects triggered by different elements may or may not correspond. In the following sentence, there are two evaluative elements, a subjunctive predicate ‘fear’ and EN, both of which give rise to an analogous evaluative effect, i.e. undesirability. If one wants to assume a Cinque (1999)-style analysis employing an Eval-head in order to describe the feature checking operation of evaluative features, one has to postulate that in (92) both features can be checked off by a common Eval-Head that regards undesirability.10

[+evalundes] (92) J’ai

peur

I have fear

[+evalundes] que cela

ne

se reproduise.

that it might

Neg

happen again

[French]

‘I am afraid that it might happen again (because it is undesirable).’

Under the current proposal that EN is a kind of subjunctive mood marker, however, Villalta’s assumption of analogous semantics between matrix predicate and subjunctive morpheme is furthermore shown to require revision. It cannot account for the data in Korean (and Japanese) in which the subjunctive predicate kitayha ‘hope’ denotes desirability whereas the EN anh 10

Note that from a pragmatic perspective the dual marking of undesirability brings about a doubly heightened attitude (i.e. strong undesirability) of the epistemic subject. 200

expresses unlikelihood. This means that, unlike (92) above, the two evaluative features with different dimensions of meaning in (93) call for two separate Eval Heads.11

[+evalunlikely] (93) John-un John-Top

[+evaldes]

Mary-ka

oci-anh-ul-ci

kitayha-koiss-ta.

Mary-Nom

come-Neg-Fut-NFcomp

hope-Asp-Decl

[Korean]

‘John hopes that Mary might come (although it is unlikely to happen).’

Likewise, the common semantics hypothesis fails to capture the cases where two instances of EN simultaneously denote different contents. Recall that multiple occurrences of EN may trigger different evaluative effects, unlikelihood and desirability.

[+evalunlikely] (94) Mary-ga Mary-Nom

[+evaldes]

ko-nai

ka-naa!

come-Neg

NFcomp-Opt

[Japanese]

‘Oh, how I wish Mary comes (although it is unlikely to happen)!’

Thus, if we have independent reason to believe that these evaluative features must be included in the domain of syntax, then we are forced to assume that each subjunctive element may possess its own evaluative feature in syntax in order to account for the above data in which each

11

The evaluative meaning of ‘unlikelihood’ here may also be attributed to the non-factive complmentizer ci. Hence one might want to posit a separate [+evalunlikely] feature for ci also, and postulate an Agree operation between them. At this point, however, I do not have independent evidence to do so. 201

occurrence of an evaluative element can have its own contents. That is, one has to posit an evaluative feature for each instance of EN that must be checked off in the syntactic derivation.

6.3.4. A potential account: multiple Eval-Heads In this subsection, I will briefly discuss the potential way to account for the syntax of the multiple occurrences of evaluative elements, although it still requires further empirical and theoretical justification for following this route which will remain for future research. As noted above, if one wants to capture the interactions amongst multiple evaluative features in subjunctive and EN sentences as part of the syntactic derivation, it is possible to assume a Cinque-Speas style analysis. That is, what has been broadly assumed to be only pragmatically-relevant features (Chafe 1986; Chafe and Nichols 1986; Sells 1987; Culy 1994) are represented in syntactic projections above IP and must be checked (Cinque 1999 for adverbs; Speas 2004 for evidentials; Kempchinsky 2009 for subjunctive mood). Speas further notes that evidentials interact with syntactic inflectional features such as person and tense, and in many languages evidential features are reflected in modal auxiliaries, adverbs or propositional attitude predicates. In 6.2.1-3, it was shown that certain lexical elements typically hold the evaluative feature across languages. Based on the assumption that EN as well as evaluative subjunctive elements is part of the grammar, uniformly triggering evaluative semantics, I have revisited the domain of what may count as subjunctive-mood markers. I have illustrated potential mood elements with an evaluative feature – EN, subjunctive morphology on verbs, and subjunctive complementizers. Then it is possible to apply a Cinque-Speas style analysis to evaluatives since the main properties of evaluatives exhibit a striking resemblance to other pragmatic features like evidentials. Just like

202

evidentials, evaluatives are analyzed to be syntactically located at the highest projections as elements with speaker-orientation are expected to be higher in syntax. It is plausible to furthermore assume that certain evaluative adverbs can convey mood information which will be discussed shortly. Despite the crosslinguistic variation between verbal mood inflection in Romance languages and complementizer selection in East Asian languages as shown above, these elements can be characterized as a representative evaluative class. As noted above, this is strongly reminiscent of Speas’ (2004) claim about evidentials that “many languages spell out evidential features with modal auxiliaries, adverbs or propositional attitude predicates”. Note that the Dutch example given earlier illustrates this directly since zou is a modal auxiliary. Hence it seems that, just like evidentials, evaluative features are spelled out via various means. In exploring the syntactic representation of evidential morphemes, Speas (2004) proposes that pragmatically-relevant features like evidential features exist in syntactic projections above IP, and they must be checked. Her analysis is built on Cinque’s (1999) analysis of adverb position and morpheme order, as shown below.

(95) Cinque (1999)’s four highest projections: [Speech Act Mood [Evaluative Mood [Evidential Mood [Epistemological Mode]]

In particular, Speas (2004: 259 (9)) summarizes the characteristics of each head as the following.

(96) Speech Act Mood: indicates the type of speech act (declarative, interrogative, etc.) Evaluative Mood: indicates speaker’s evaluation of the reported event or state as good, lucky, bad, surprising, etc.

203

Evidential Mood: indicates the nature of speaker’s evidence for truth of proposition Epistemological Mode: indicates speaker’s degree of certainty about the proposition

According to this classification, the properties of EN, subjunctive morphology on verbs, and subjunctive complementizers seem to be parallel to those of evaluative mood – they also indicate the speaker’s evaluation of the reported event or state as desirable or undesirable, as expected or unexpected, etc. Furthermore, the hierarchical order in syntactic projections in (96) above is motivated by the restricted order of adverbs in each category in English. Cinque claims that the adverbs in English given below must occur in a given order.

(97) Representative Adverbs: Speech Act Mood

frankly, confidentially

Evaluative Mood

unfortunately, luckily, surprisingly

Evidential Mood

allegedly, reportedly,

Epistemological Mode

obviously, apparently

Here the meaning induced by EN and other subjunctive-like elements seems to be analogous to that of the adverbs in the evaluative mood category such as unfortunately, luckily, and surprisingly, only with negative intervals. As Kempchinsky (2009) notes, the subjunctive morpheme in (88) above has the effect of an evidential adverb like allegedly, which is classified as speaker-oriented adverbs in Ernst (2009). Furthermore, I have shown in chapter 4 that the low-likelihood adverbs hoksilato ‘by any chance’ in Korean and mosikasuruto/hyottosuruto ‘by

204

any chance’ in Japanese typically co-occur with the low-likelihood marking EN and give extra emphasis to the epistemic state of low-likelihood. In this vein, it seems plausible to assume that certain adverbs with evaluative flavor constitute another category with an evaluative feature. For instance, this subjunctive category would include low-likelihood adverbs like perhaps, by any chance, surprisingly and undesirability adverbs like unfortunately, sadly, regrettably, etc. On the other hand, adverbs like certainly, definitely, indeed would be assumed to co-occur with the indicative mood. Hence this type of adverb is another potential locus for a syntactic evaluative feature along the lines of Cinque’s (1999) analysis of evaluative adverbs (see also Speas 2004 for evidentials; Kempchinsky 2009 for subjunctive mood). In the spirit of Cinque (1999) and Speas (2004), if there are evaluative feature bearers, the evaluative features therein can be assumed to be checked off at the evaluative mood head above IP. Furthermore, in chapter 5 I proposed that multiple occurrences of subjunctive markers denote a semantics with multiple layers of meanings. This implies that morphological realization of evaluatives within a clause may concern more than one dimension of evaluative meaning. In order to correctly capture the semantics with simultaneous occurrences of different evaluative features, one has to assume that a structure with multiple evaluative mood phrases is possible, as in (98) below. For instance, if evaluative mood feature in EvalP1 checks off the feature value of likelihood, then evaluative mood feature in EvalP2 checks off the feature value of desirability.

(98) … [EvalP1 … [EvalP2 […]]]

In 6.1, it is shown that potential evaluative feature bearers include EN, subjunctive morphology on verbs or subjunctive complementizers, and evaluative adverbs. I furthermore discuss that it is

205

possible to embed an account of evaluative negation in a set of assumptions about the syntaxsemantics/pragmatics interface in the sense of Cinque and Speas, though whether one will want to do so rests on factors that are not relevant here. For present purposes, there is no advantage of taking this account. The important contribution from the discussion in 6.1 is to make an empirical point that subjunctive markers in subordinate clauses may indeed trigger their own evaluative sense. In the next section, I will move onto another issue concerning the syntax of subjunctive and EN complements.

6.4. Summary and Conclusion In chapter 6, I have explored the syntactic configurations of subordinate subjunctive clauses and subordinate EN-clauses in Korean (and potentially Japanese) that are compared to those of Romance and Germanic languages. In 6.1 and 6.2, I investigated the configurational relation between the matrix clause and the EN-clause in Korean, examining whether the two clauses are hypotactically or paratactically connected. The proposed dynamic syntax of EN (and subjunctive) complements affords the following theoretical implications: First, by revealing the parallels with EV2 constructions that are in subjunctive, the syntax of EN constructions further supports the main idea of current proposal that EN is a subjunctive marker. Second, it suggests that the syntax of EN complements in Korean must be understood along the lines of Jespersen’s insight of paratactic negation in Old/Middle English. Third, it accounts for why the complementizer in subordinate EN clauses is in an identical form with a direct question particle in Korean and Japanese. That is, speakers who assume this paratactic structure, rather than the hypotactic one, might perceive the EN complement as a kind of indirect question that is pragmatically squeezed into another sentence without being property embedded. Finally, it supports the possibility that

206

negation in tag-questions, which are classical paratactic configurations, can be understood as a subspecies of EN. Furthermore, with regard to the relation between EN complments and tagquestions in general, the above observation that EN complments in Korean also reveal hypotactic properties like extraction does not necessarily threaten their connection. As already noted, similar observations are in fact made for tag-questions: den Dikken (1995), for instance, does not treat them as paratactically related to the main clause, showing that heavy NP-shift is available across a tag-question in English which is not straightforward if tags are not an integral part of the main clause. In 6.3, I tackled the question of how clauses containing evaluative elements in Korean can be syntactically represented. I offered empirical descriptions of which elements constitute subjunctive mood that gives rise to an evaluative sense in Korean. I furthermore discussed a potential account that multiple evaluative elements with different meanings can be captured by positing multiple Eval-Heads along the lines of Cinque and Speas, but whether we need to follow this route remains for future research.

207

Chapter 7. A case study of Rhetorical Comparatives: PI, mood, and EN

In chapter 7, I offer a case study concentrating on the semanticopragmatic properties of Rhetorical Comparatives (RCs). It is shown that rhetorical effects can be triggered by negative polarity items, EN, and the subjunctive mood. This result supports an important insight that negative polarity items and the subjunctive are of similar nature (Giannakidou 1994, 1995, 2009; Quer 1998; Borschev et al. 2007), and goes one step further to suggest another important link between EN and the subjunctive mood. This analysis implicates that the three components are closely connected under the principle of nonveridicality. The notion of RCs in the sense that I suggest here can give us a plausible foundation for the analysis of rhetorical effects in other environments, for instance, rhetorical questions. Section 7.1 starts with the discussion on previous analyses of comparatives: While the semantic approaches differ on whether there is a negative operator (¬: Jespersen 1917; Ross 1969; McConnell-Ginet 1973; Seuren 1973; Klein 1980; Stassen 1984; Larson 1988) or a nonnegative inequality operator (>: von Stechow 1984; Rullmann 1995; Kennedy 1997a; Beck, Oda, and Sugisaki 2004) in the comparative, various issues remain unresolved surrounding the “negativity” of comparatives. This chapter proposes a novel dichotomy of comparatives between ‘rhetorical’ comparatives (RCs) and regular ‘degree’ comparatives (DCs), and shows that only RCs convey “negativity” in a way parallel to negativity in rhetorical questions; while regular DCs merely establish an ordering between two objects. The variation concerning negativity in turn squares neatly with the presupposition toward the content of the standard (negative presupposition in RCs and no presupposition in DCs).

208

7.1. Introduction: debate on the negativity of comparatives 7.1.1. Negative analyses and their limitations The question of whether or not comparative clauses have a negative operator has divided the literature into two parties. Earlier researchers argue that comparatives have an underlying syntactic negative operator in a standard than-clause (Jespersen 1917; Ross 1969; McConnellGinet 1973; Seuren 1973; Klein 1980; Stassen 1984; Larson 1988). As illustrated in Ross’s deep structure (1), the negative analyses assume only one degree d to the extent that John is tall while denying the fact that anyone in the standard clause is tall to the same extent d.

(1)

John is taller than anyone.

Negative Analysis

∃d John is tall to extent d AND NOT [anyone else is tall to extent d]

(Ross 1969)

The idea of positing an abstract negative operator for than-clauses has been motivated by the following empirical evidence (Joly 1967; Seuren 1973): i) As shown in (2), nor is observed in lieu of than in various English dialects; ii) Old English Þon-ne ‘by which not’ is the etymology for than; and iii) As shown in (3), a negative particle ne is obligatory in a than-clause in French (Seuren 1973) as well as in many other Romance languages such as Catalan and Spanish (Price 1990), which are similar to Cockney English as in (4) where never is used in comparatives.

(2)

He is richer nor you’ll ever be.

(3)

Jean est plus grand que je ne pensais. Jean is taller

[English dialects]

than I Neg thought

‘Jean is taller than I thought.’

209

[French]

(4)

She did a better job than what I never thought she would.

[Cockney English]

Further potential evidence for the negative analyses comes from NPI facts in comparatives. As shown in anyone in (1) and ever in (2) above, negative approaches generally deal with comparative sentences that contain NPIs in than-clauses and extend their analysis to the ones without NPIs. The logic goes as follows: consider that negative gradable adjectives such as short/difficult/unjust/impossible allow for NPIs to occur with them as shown in (5a) whereas neutral gradable adjectives such as tall/easy/just/possible do not as shown in (5b). However, a neutral gradable adjective like tall still licenses an NPI such as anyone in a comparative as shown in (5c). For this reason, negative analyses argue that NPI-licensing is indicative of the negative status of comparative than-clauses.

(5)

a. It is difficult to ever get a straight answer from him. b. *It is easy to ever get a straight answer from him. c. John is taller than anyone.

The negative analyses, however, encounter a number of empirical and theoretical problems. First, the assumption that NPIs are licensed by the syntactic negative operator seems problematic since the monotonicity status, i.e. downward or upward entailingness, or (non-)veridicality, of comparative clauses has been controversial in the literature and never completely understood (see Hoeksema 1983; Larson 1988; Rullmann 1995; Hendriks 1995; Schwarzschild & Wilkinson 2002; Heim 2006). Contra the traditional assumption that comparative clauses are downward entailing (i.e. NPI-licensing) contexts, some recent researchers claim that they are upward

210

entailing (Larson 1988; Schwarzschild & Wilkinson 2002) or ambiguous (Rullmann 1994; Hendriks 1995; Heim 2006). On this matter, Heim notes that Schwarzschild & Wilkinson’s upward entailment (UE) analysis of comparative clauses is due to the fact that the than-clause takes widest scope. That is, inferences in (6a) and (6b) hold because the DPs are effectively interpreted with widest scope, but it does not mean that than-clauses are always UE. It has been noted that than-clauses are DE if they contain DE operators like negation or universal quantifiers.

(6)

a. John is taller than some professional basketball players are. Therefore, he is taller than some professional athletes are. b. John is taller than most of his ancestors were. Therefore, he is taller than some of his ancestors were.

Given that a comparative clause itself may vary in terms of DE or UE depending on the quantifiers it contains (Rullmann 1994 for Dutch comparatives; Hendriks 1995; Heim 2006), monotonicity does not seem to be relevant in characterizing DCs versus RCs. Second, NPIs that need negation in languages like Greek and Korean do not get licensed in the comparative (Giannakidou & Yoon 2010). Based on the fact that strict NPIs such as Greek KANENAS ‘anyone’, Korean amwuto ‘anyone’ and Greek and Korean minimizers like ‘budge an inch’ are uniformly ungrammatical in comparatives, Giannakidou & Yoon argue that regular comparative clauses do not contain a negative operator. They further show that the same argument applies to English since strict NPIs like either is ungrammatical in comparative sentence: “*John is taller than Bill (is) either.”

211

Third, as Price (1990; cf. Gaatone 1971; Napoli & Nespor 1977) notes, negative analyses assuming a semantics like (1) above cannot account for the fact that a negative particle ne also appears in equatives, for instance, in French. (Further discussion will follow in 7.4.2 regarding the appearance of a negative particle in comparatives):

(7)

Il est aussi bon he is as

qu’ils

good as they

ne puissant

l’être.

[French]

Neg can.Subj 3sg be

‘He is as good as they could be.’

Fourth, in more recent analyses of comparatives, it is no longer believed that there is only one specific degree d and that an underlying negative operator indicates the failure of the comparative standard to reach the same degree d, since regular (inequality) degree comparisons are normally made with at least two salient objects which display at least two distinct degrees. If there are two degrees that the speaker perceives when employing a comparative construction, both degrees must be encoded in the semantics of comparatives. In particular, a semantics that assumes only one degree d faces a significant challenge in the following cases, where two distinct degrees are forced to be present: (i) as von Stechow (1984) notes, Russell’s ambiguity in which both instances of she refer to the subject Mary with differing degrees in (8);1 (ii) the differential reading which requires an extra semantic device for incorporating the exact amount of difference, while denoting a relation between degrees of height between two objects at the same time as in (9); and (iii) the subcomparative where each degree is computed over two separate dimensions of measurements as in (10) (Kennedy 1997b). 1

The sentence is ambiguous between two readings with consistent and inconsistent degrees of

what the two instances of she denote. 212

(8)

Mary believes she is taller than she is.

(9)

John is 3 cm taller than Tom.

[Russell’s ambiguity] [Differential

comparative] (10) This door is taller than that window is wide.

[Subcomparative]

Although not entirely impossible, the underlying negative operator in negative approaches is difficult to provide a straightforward semantics for cases like these without causing a significant number of theoretical complications (C. Kennedy p.c.): In addition to the basic semantics like the one in (1) above, it would require a separate set of semantic tools for being able to calculate the precise degree in standard or the degree difference between subject and standard. Such complex semantics is undesirably stipulative and implausible from a child language acquisition standpoint. Finally, it will be shown in section 7.2 that, although negative polarity items seem to quite frequently occur in comparative clauses, their appearances are in fact not costless in terms of semanticopragmatic effects.

7.1.2. Non-negative analyses and their limitations Presumably for the aforementioned reasons, recent researchers no longer assume an underlying negative operator in comparatives, but instead posit a degree operator indicating an inequality relation (>) between two existing degrees such that d′ is greater than d′′, as illustrated in (11) (von Stechow 1984; Rullmann 1995; Kennedy 1997a; Kennedy & McNally 2002; Schwarzschild 2005; Beck 2006; Heim 2006 among many others).

213

(11)

Kim is taller than Lee (is).

Non-negative Analysis

max{d′| tall(kim) ≥ d′} > max{d′′ | tall(lee) ≥ d′′}

The non-negative analyses are advantageous in affording a simpler and more straightforward semantics, without having to resort to an implicit negative element. Furthermore, the aforementioned problematic cases for the negative analyses such as Russell’s ambiguity, differential readings, or subcomparatives are easily explained away within the non-negative analyses where a degree operator encodes an ordering relation between two explicit maximal degrees (d′, d′′). Despite these merits, however, the non-negative analyses are also incomplete for the following reasons. First, the abovementioned controversy on the DEness or UEness of comparative clauses (Hoeksema 1983; Rullmann 1995; Hendriks 1995; Larson 1988; Schwarzschild & Wilkinson 2002; Heim 2006) casts doubt on the non-negative theories, just as it did for the negative theories: Given that the debate has been concluded that the comparative is not inherently monotonic, but depends on the kind of quantifier it contains (Rullmann 1995; Hendriks 1995; Heim 2006), how come NPIs appear in the following comparatives which do not contain any DE operators like negation or universal quantifier?

(12)

a. Jack is taller than ANYbody else (is). b. Jack is richer than you’ll EVer be. c. This work is more than I can STANd. d. Grace’s chicken was more than I could be BOthered eating. e. Grace said the sky would sooner fall than she would budge an inch.

214

f. Jack does volunteer works more often than he lifts a finger to help his wife. g. Jack would waste money on gambling more happily than he’d give a penny to a charity. h. Jack helps other people more willingly than he pays the least bit of attention to his own family.

Here two things must be noted: first, we detect indeed a negativity that doesn't follow from the ordering alone. Second, the negativity seems to come form the NPIs themselves, which are all emphatic. Minimizers are emphatic anyway at all times, but with any and ever it is important to note the use of emphasis because they don't always come with it. These two properties, emphasis and negative effect, if properly analyzed, will be shown to be the key factors for producing the rhetorical effect, I will argue. Another potential problem for the non-negative analyses is that they do not offer any explanation for the following questions which have been the original motivations for the negative analyses of comparatives: (i) Why do certain languages or dialects contain an overtly negative element within the comparative complementizer than? (Recall that some English dialects employ overt negative nor, instead of than); (ii) Why do languages like Old English, French, Spanish, and Catalan have a negative particle ne or no in comparatives? One might wonder whether these occasional negatives necessarily deny the inequality operator (>) analyses, given that from a truth conditional perspective there is not an interesting difference between the negation analysis and the max analysis: since the relation in the max analysis logically entails that d′′ is not as great as d′, they are just different ways of representing the same truth conditions. However, the differences are certainly interesting if we think about

215

bits of the actual linguistic representation as giving rise to one kind of meaning versus the other because then, depending on what the syntactic/morphological/semantic features of the different bits are, and what kinds of Logical Forms they allow, we can make different predictions (C. Kennedy p.c.). For instance, saying ‘(i) Jack is tall to the degree d and everyone else is not as tall’ or ‘(ii) the maximal degree d1 to which Jack is tall is grater than the maximal degree d2 to which everyone else is tall’ makes no truth-conditional difference, which must be the reason why the representational difference didn’t receive much attention in the previous literature. The current study, however, focuses on the former meaning, arguing that the emergence of negation in the comparative clause across languages is not coincidental but necessary in order to convey a negative implicature toward the content of the comparative clause. In this spirit, the appearance of a negative particle in comparatives will be analyzed in section 7.4.2, where I assume that the comparatives with a negative element are only a special subtype of comparatives giving rise to rhetoricizing effects, which may have been accidently spread to regular comparatives and sometimes even to equatives in certain languages and dialects. As such, previous approaches leave behind a number of unresolved issues especially on the negativity of comparative clauses. As von Stechow (1984) notes, different theories of comparatives better explain different areas of comparative data. This suggests that we need a more refined theory of the kinds of comparatives when it comes to capturing the negativity of comparatives that is shown in the examples with NPIs above.

7.1.3. Main ideas to be proposed: a split analysis In exploring the properties of distinct kinds of comparatives at the level of the semanticspragmatics interface, I identify a novel subtype of comparatives, namely ‘rhetorical’

216

comparatives (RCs), in contrast to regular ‘degree’ comparatives (DCs). These, I argue, convey more “negativity” than the mere DCs do, and they display a number of striking parallels to rhetorical questions (RQs), hence my calling them “rhetorical”. The hallmark property for both RCs and RQs is, of course, the rhetorical flavour. Just as rhetorical force in questions is induced by NPIs in English, Serbian/Croatian, etc. (see Borkin 1971; Lawler 1971; Progovac 1994; den Dikken & Giannakidou 2002), I propose that the presence of emphatic and strong NPIs in comparatives is the main ingredient for contributing to rhetorical force as well. Furthermore, it will be shown how the separation of rhetorical comparatives from regular comparatives opens the possibility of resolving various issues surrounding the semantic and pragmatic aspects of comparatives. The purpose of this article is the following: First, in section 7.2, I use the distribution of polarity items as a departure point for developing a semantic analysis of rhetorical comparatives (RCs) that supports a divide between two types of comparatives. Our second goal in this article, in section 7.3, is to show that RCs are different from metalinguistic comparatives (MCs: Bresnan 1973; McCawley 1988; Embick 2007; Giannakidou & Stavrou 2008; Morzycki 2008; Giannakidou & Yoon 2009, to appear). Third, section 7.4 demonstrates that there are three strategies for rhetoricizing comparatives that are largely predictable from their common property, i.e. non-veridicality. Finally, the result reinforces the larger claim advanced by Giannakidou (2009) that there is a fundamental connection between polarity phenomena and mood, and our innovation here is to suggest another important connection between expletive negation and mood, under the principle of nonveridicality.

217

7.2. Rhetorical comparatives A primary motivation for the current proposal is the following. Comparative data are diverse with respect to negativity and cannot be captured by prior uniform analyses. I identify here a species of degree comparison that is more “negative” than the regular ordering (>) of the comparative makes us expect.

7.2.1. Empirical description of the rhetorical effect of an NPI in the comparative RCs are illustrated in (13), repeated from (12):

(13)

Rhetorical Comparatives

a. Jack is taller than ANYbody else (is). b. Jack is richer than you’ll EVer be. c. This work is more than I can STANd. d. Grace’s chicken was more than I could be BOthered eating. e. Grace said the sky would sooner fall than she would budge an inch. f. Jack does volunteer works more often than he lifts a finger to help his wife. g. Jack would waste money on gambling more happily than he’d give a penny to a charity. h. Jack helps other people more willingly than he pays the least bit of attention to his own family.

218

In these examples, the presence of emphatic/strong NPIs gives rise to rhetorical effects. Pragmatic reasons to use RCs are to convey: (i) a negative implicature toward the content of the standard clause and/or (ii) a presupposition of a large difference, as given below of each sentence:

(14)

Presuppositions in Rhetorical Comparatives

a. Jack is taller than ANYbody else is. i) Negative implicature: Everybody else is not going to be as tall as John is. ii) Large difference presupposition: There is a significantly large difference in degree of height between Jack and everyone else.

b. Jack is richer than you’ll EVer be. i) Negative implicature: You will never be as rich as John. ii) Large difference presupposition: There is a significantly large difference in degree of wealth between Jack and you in any foreseeable future.

c. Jack does volunteer works more often than he lifts a finger to help his wife. i) Negative implicature: Jack very rarely (or never) lifts a finger to help his wife. ii) Large difference presupposition: There is a significantly large difference in frequency between Jack doing volunteer works and him helping his wife.

d. Jack would waste money on gambling more happily than he’d give a penny to a charity. i) Negative implicature: You would never give a penny to charity.

219

ii) Large difference presupposition: There is a significantly large difference in degree of happiness that Jack would feel between wasting money on gambling and giving money to a charity.

The negativity of RCs originates from the negative implicature or negative bias. But what kind of implicature? In order to formulate the semantics of this novel subcategory of comparatives, it will be instructive to look at a strikingly similar phenomenon, namely rhetorical questions (RQs). As in (15), questions with strong NPIs/minimizers are RQs which strongly prefer a negative answer (Sadock 1971, 1974; Linebarger 1980, 1987; Kadmon & Landman 1990; Progovac 1994; Krifka 1995; Lee 1995; Guitérrez-Rexach 1996; Han 1997; Han & Siegel 1998; Giannakidou 1998, 1999; Guerzoni 2001; den Dikken & Giannakidou 2002; cf. Ladusaw 2004 for RQs as biased assertions, and van Rooy 2003 for RQs as constrained questions with information entropy).

(15)

Who could sleep a wink with this racket?

RQ

Just like in RCs, the negativity in RQ like (15) comes in the form of the negative implicature. This, however, is implicature not entailment since a positive answer like “I would” is possible in which case the implicature is cancelled (den Dikken & Giannakidou 2002). Then I propose that just like RQs, the hallmark property of RCs is the containment of a strong NPI in comparative clauses. I furthermore assume that NPIs with emphatic intonation in (13a-d) (Giannakidou 1998, 2000 shows that strong NPIs crosslinguistically have emphatic

220

intonation) and minimizer NPIs in English in (13e-h) (Krifka 1995) fall under the category of strong NPIs.2 On the other hand, just as questions with unstressed weak NPIs like any or ever can be regular information-seeking questions (Borkin 1971; Heim 1984), the following comparatives with any or ever could also be DCs that are purely informative of the relative ordering relation between two maximal degrees.

(16)

a. Jack is taller than anyone else.

DCs

b. Jack is richer than I ever was.

Note also that, besides the lexical strength or prosodic emphasis on NPIs, the presence of a modal plays an additional role in giving rise to rhetorical effects in comparatives. When a weak type of NPI ever is accompanied with a modal will, as in (17a), the resulting presupposition is equivalent to one with a strong NPI or an emphatic NPI, as in (17b).

(17)

a. Jack is richer than you’ll ever be.

RCs

b. Jack is richer than you EVer were.

This is strongly reminiscent of den Dikken & Giannakidou’s (2002) observation that the following question (18b) with the weak NPI any plus the modal would is a RQ presupposing a

2

In the category of strong NPIs that induce rhetorical effects here, NPIs that need negation such as English either, Greek KANENAS, or Korean amwuto are excluded. These NPIs should be categorized into superstrong NPIs (Zwarts 1993) that can never occur in comparative clauses without being licensed by overt negation. 221

negative answer, whereas no such presupposition exists in the absence of a modal in an information question (18a).

(18)

a. Which student read any of the papers?

[information question]

b. Which student would read any of the papers? [negative answer preferred]

This data provides a clear answer to the question of what exactly is responsible for the rhetorical flavor. Just like in RQs, I will propose that it is the contribution of the NPI itself that primarily creates rhetorical effects in RCs and the presence of a modal reinforces rhetorical effects when an NPI is too weak to trigger a rhetorical flavor by itself. More discussion on the semantic role of modals will follow in 2.3 and later in 4.3. Thus far I have shown that emphatic/strong NPIs contribute rhetorical effects in questions and comparatives. A question arises here: how precisely do these NPIs give rise to RCs? The idea that I advance here is: what makes RCs special is the fact that they contain a ‘non-referential’ standard, and that strong/emphatic NPIs play a crucial role in altering a referential standard into a non-referential standard. Our conception of the non-referential standard refers to cases where the domain of what the standard denotes goes beyond its regular borderline of presupposed subset by an excessive size adjustment, and makes it difficult to make proper reference. This anti-anaphoric standard is what creates a non-veridical context in RCs. And there seems to be two ways to introduce a non-deictic standard. The first strategy is to employ an enormously extended domain for the standard, which in turn renders itself referentially deficient. Such maximal domain extension can be accomplished either by prosodic emphasis on regular domain extenders such as any, ever (Kadmon &

222

Landman 1990, 1993; Krifka 1990, 1992, 1995; Giannakidou 1998, 2000), or by means of a weak NPI plus a modal (den Dikken & Giannakidou 2002). Domain extension has been discussed in questions with NPIs, where a weak NPI like ever is used in an information-seeking question in order to turn a biased question into an unbiased one (Krifka 1990, 1992, 1995; Kadmon & Landman 1993; van Rooy 2003). Van Rooy notes that other things being equal, general questions are normally preferred to specific ones. (19a) asks about recent visits; (19b) is a more general question, concerning ‘in your life’.

(19)

a. Have you been to China? b. Have you ever been to China?

Kadmon & Landman note that in contrast with a weaker expectation implicated in (20a), (20b) gives rise to a strong expectation that Sue does not have any potatoes whatsoever.

(20)

a. Does Sue have some potatoes? b. Does Sue have any potatoes?

They further argue that in a context where ‘potatoes’ should mean ‘cooking potatoes’, the presence of any reflects the expectation that Sue does not have any cooking potatoes, but it has been noted that Kadmon & Landman’s claim on domain widening (and strengthening) holds only for stressed ANY (Rohrbaugh 1993; Krifka 1995; van Rooy 2003). Once we take into account emphatic ANY, however, the size of the domain for potential referents can be divided into three categories depending on the selection of polarity items. First,

223

(20a) is an information-seeking question with a specific narrow domain, triggered by a PPI some. Second, (20b) is also an information-seeking question where a general wide domain of potatoes, triggered by a NPI any, is calculated under normal circumstances. Thus, as Kadmon & Landman claim, it only concerns the presence of prototypical cooking potatoes, irrespective of the existence of other kinds of potatoes. Finally, I diverge from van Rooy by assuming that prosodic emphasis on weak NPIs like ANY reinforces the domain extension from a regular one, ranging over typical references, to a maximally wider one, even further than what a context may delimit. Thus (21) below constitutes a clear case of rhetorical question, where the domain of potatoes is extended to the maximal possible level, escaping out of the borderlines for prototypical potatoes.

(21)

Does Sue have ANY potatoes?

RQ

Hence it comes to include even the least likely ones such as stale potatoes, rotten potatoes, potatoes for industrial use, or even ones of negligible quantity such as an extremely thin sliver of potato, or ones of questionable boundaries like potato powder or potato extract. This represents a ridiculously extended domain triggered by emphatic ANY, which is the main source for the rhetorical flavor. In a similar respect, den Dikken & Giannakidou discuss the domain extension of (22b) in which modification by the hell has the effect of extending the domain of quantification of who to a wider one including familiar and novel values (Heim 1982).

(22)

a. Who talked to Ariadne? b. Who the hell talked to Ariadne?

224

Whereas regular wh-phrases quantify over a presupposed subset of the domain, the domain of quantification of wh-the-hell widens to the entire domain, becoming “the open set including all persons in the universe, and all possible values are available for x, even less likely or prototypical ones” (den Dikken & Giannakidou 2002: 43). Then their intuition on wh-the-hell seems to be close to the domain of emphatic ANY. Thus the extreme domain extension produces rhetorical effects such that no one was supposed to talk to Ariadne. The insight on rhetorical effects induced by radical domain extension in interrogatives is directly applicable to comparatives. Just like the above questions, comparatives with regular any vs. emphatic ANY display a similar contrast, and it seems analogous to another weak NPI, ever vs. emphatic EVER. In the following DC (23b), the domain for the denotation of “you ever were” receives a ‘regular extension’, covering the entire domain that is accessible in a given context.

(23)

a. John is richer than anyone is.

DCs

b. John is richer than you ever were.

Even after the extension, it is still possible to refer to the maximal degree to which the wealth of the addressee peaked at some point in the past of his life. Being within appropriate boundaries of the domain is what makes the standard a referential and veridical context in DCs. However, in the following RC with emphatic EVER (24a), the domain for the standard is excessively extended beyond its boundaries and therefore it becomes difficult to make reference to any particular temporal point when the addressee is d′′ rich. On the other hand, though the non-emphatic ever in (24b) only triggers a regular domain extension, its association with the

225

modal will boosts the effects of domain extension to the level of surpassing the highest endpoint of a scale associated with the addressee’s wealth.3 I thus propose that this ‘extreme extension’ of domain is a mode of establishing a non-referential domain for the standard, giving rise to rhetorical effects.

(24)

a. John is richer than you EVER were.

RCs

b. John is richer than you’ll ever be.

The second strategy for a vague standard is to go in the opposite direction, i.e. to exploit an ‘excessive shrinking’ of the domain of the standard. Just like surpassing the maximal endpoint of a regular domain, playing down the minimal endpoint significantly decreases referentiality as well. This is typically shown in RCs with minimizers like lift a finger, bat an eye, sleep a wink, budge an inch, give a penny, or the least bit, as in (25). This extremely shrunk domain is another mode of establishing a non-referential domain for RCs.

(25)

a. Grace said the sky would sooner fall than she would budge an inch. b. Grace buys expensive presents for her assistant more often than she lifts a finger to help her husband. c. Jack would waste money on gambling more happily than he’d give a penny to the charity. d. Jack helps other people more willingly than he pays the least bit of attention to his own family.

3

In terms of non-exhaustivity, it is possible to attribute this to the fact that, just like weak NPIs, modals introduce alternatives (Kratzer 1981, 1991). 226

In accounting for why RQs are triggered by these minimizers, prior analyses diverge: i) Krifka (1995; see also Guerzoni 2001, van Rooy 2003) attributes the rhetorical force to the fact that the semantic meaning of strong NPIs like lift a finger denotes the minimal element of a scale; ii) Karttunen & Peters (1979) argue that it is because these items share a presupposition with even; iii) however, since the association of ‘even’ with the end of a scale is questioned by Kay (1990), van Rooy (2003) argues that the rhetorical effect of the use of some NPIs in questions is reducible to domain widening in the sense of Kadmon & Landman (1990). Our conception of excessive domain adjustment suggests a systematic refinement of these original ideas on the source of rhetorical effects, and the contribution of the current proposal is that such strengthening effects come not only from domain extension but also from domain shrinking. The bidirectional domain adjustment of these NPIs is a natural way of significantly reducing their referentiality, either maximizing them to be non-specific or minimizing them to be non-existent. As such, our hybrid analysis predicts the regularity in the procedure of how strong and emphatic NPIs induce rhetorical effects in comparatives as well as in questions. Given that strong NPIs consistently reveal such a property, I can generalize that non-referentiality is not delimited to a special set of lexical items in certain languages but rather a universal property for creating negative polarity in natural language. And this is the direction that I am pursuing. Furthermore, I assume that the negative rhetorical effects follow from the direct link between non-referentiality and negative polarity that has been proposed by variation approaches for polarity source (Giannakidou 1998, 2001, 2009; cf. Haspelmath 1997, Farkas 2002).

227

7.2.2. Analysis: RC as a product of non-referential standard and NPI use For the analysis of RCs, there will be three main ingredients: First, NPI is analyzed as containing a deficient, dependent, non-deictic variable (Giannakidou 1998) which triggers the undefined standard of RCs in 2.2.1. Second, negative implicature in RCs will be recast as a presupposition of negative attitude (in the sense of den Dikken & Giannakidou 2002) in 2.2.2. Finally, rhetorical effects in RCs will be attributed to the semantics of RC with an undefined standard, and the presupposition of large difference is accounted for by the domain adjusting nature of NPIs.

7.2.2.1. Giannakidou 1998: NPIs containing non-referential, non-deictic variable If the presence of strong/emphatic NPIs (and modals) renders comparatives rhetorical, as opposed to DCs, the next question is why such NPIs induce a rhetorical flavor. In order to see why, an understanding of the nature of polarity items is called for. In the polarity literature, there is an ongoing debate on the sources of polarity between scalarity only approaches (Kadmon & Landman 1993; Lee & Horn 1994; Krifka 1995; Lahiri 1998; Chierchia 2006; Farkas 2006) which assume only a scalar focus particle (even) as the polarity source, and variation approaches (Giannakidou 1998, 2001, 2009; cf. Haspelmath 1997, Farkas 2002) which also posit ‘non-referentiality’ as another source for polarity, in addition to scalarity. I follow the variation assumption for polarity sources that has been strongly supported with empirical evidence across languages, and it will be shown how non-referentiality plays a key role in establishing core properties of RCs. To begin, note that there seems to be a direct link between non-referentiality and negative polarity. It has been argued that in a number of languages certain expressions become NPIs because they are referentially deficient, even though they are not scalar. For instance, ku in

228

Salish is a determiner that “merely fails to positively assert the existence of an entity” (Matthewson 1998: 179). Another representative example of an NPI triggered by nonreferentiality is wh-the-hell items in English. Den Dikken & Giannakidou (2002) argue that the negative polarity of who the hell is due to the fact that it cannot be anaphoric to a previously introduced discourse referent, as illustrated in (26).

(26)

Someonei bought that book. John knows {whoi/*who the hell}.

They further note that what is responsible for minimizers such as budge an inch, sleep a wink, lift a finger acting as NPIs is the indefinite part since a finger doesn’t refer to a finger and budge some inches or lift two fingers loses its polarity. Giannakidou (1998: 70) defines these items as a dependent variable ∃xd which is still an existential quantifier but defective because the variable xd cannot make reference to a discourse referent.

(27)

An existential quantifier is dependent iff the variable ∃xd it contributes does not introduce a discourse referent in the main context.

The dependency in reference, however, does not mean that there is no reference: she notes that in Greek, kanenan gets to introduce a discourse referent as in (28).

(28)

An dhis kanenan1,

pes tu1

na me perimeni.

If you see anybody, tell him to wait for me.

229

[Greek]

However, this kind of variable is ‘defective’ in the sense that the assignment function g cannot receive a value in a main context, hence it may receive values only in an embedded domain. As given in (29), Giannakidou (2009) proposes that a defective reference yields a polarity-sensitive expression if it is uninterpretable as a free variable.

(29)

Non-deictic variables A variable x is non-deictic iff x cannot be interpreted as a free variable.

She further notes that such anaphoric dependency can be rescued: (i) if it is in the scope of a negative operator where the introduction of a discourse referent is not forced; or (ii) if it is under a nonveridical operator which is an embedding operator, hence it is not forced to introduce a discourse referent in the main context or to be existentially closed at the text level. If non-referentiality is the main reason why lexical items like ku in Salish, Greek NPIs, and the above minimizers become NPIs, do other NPIs also display non-referentiality? The answer I will suggest is positive at least for emphatic and strong NPIs. I furthermore propose that rhetorical effects in RCs are natural consequences of containing expressions with extremely decreased referentiality in comparative clauses. In section 7.2.1, it was shown how emphatic and strong NPIs render the content of the whole clause exceedingly non-deictic by domain adjustment. In other words, a creation of anti-anaphoric standard in RCs is a speaker’s strategy to express negative implicature toward the content of the standard clause. On the other hand, I will argue that comparative clauses with weak NPIs such as non-emphatic any or ever may remain as DCs; though these weak NPIs induce domain widening of standard to some extent, the standard is still referential and lack non-deictic variable.

230

7.2.2.2. Den Dikken & Giannakidou 2002: negative attitude triggered by NPIs and modals Given that the non-referential standard of comparatives is contributed primarily by the nondeictic properties of polarity items and secondarily from modals which become relevant only when the strength of the NPIs is weak, the next question to ponder upon is: How exactly do such items give rise to rhetorical effects? And what kind of implicature is conveyed in RCs? 4 In accounting for lexical properties of the NPI wh-the-hell items, den Dikken & Giannakidou (2002) suggest that it contributes a negative presupposition as in (30), which expresses a negative attitude of the speaker.

(30)

Presupposition of negative attitude of wh-the-hell

In the actual world w: If ∃x [P(x) (w) ∧ Q(x) (w)] à SHOULD ¬ Q(x) (w), for all possible values of x. (where x is the variable of wh-the-hell, P is the property denoted by the wh-the-hell phrase, and Q is the property denoted by the VP) (den Dikken & Giannakidou 2002)

They define this presupposition as a modalized conditional statement such that “if any x such as x did what is expressed by the VP, then x should not have done it.” This presupposition implies two important properties of NPIs. For one thing, they note that the presupposition conveys “uncertainty” about the actual existence of a value for x that can be associated with the property of VP, which makes wh-the-hell equivalent to any. The uncertainty property crucially indicates

4

In fact the radical domain extension alone can explain the modalization effect of den Dikken and Giannakidou. They argued for domain opening in the RQ anyway, and domain opening does come with possibilities (thus modality). 231

that, just like emphatic and minimizer NPIs, wh-the-hell is non-referential, hence containing a non-deictic variable in the sense of Giannakidou (1998) that is discussed in the previous section. Furthermore, they note that the use of wh-the-hell conveys the proposition that “if there is indeed an x that did what the VP says, then x did something that should not have happened (den Dikken & Giannakidou 2002: 43).” For instance, in (31), the speaker presupposes that if there is anyone who talked to Ariadne, that should not have happened.

(31)

Who the hell talked to Ariadne?

This presupposition conveyed by wh-the-hell items is defined as “negative attitude” as given below.

(32)

Negative attitude of (31) If there is a person x in w, and x talked to Ariadne in w: x should not have talked to Ariadne in w.

(den Dikken & Giannakidou 2002)

The negative attitude is what makes (31) a negative rhetorical question which is generally assumed to presuppose a negative answer (Sadock 1971, 1974; Linebarger 1980, 1987; Kadmon & Landman 1990; Progovac 1994; Krifka 1995; Lee 1995; Guitérrez-Rexach 1996; Han 1997; Han & Siegel 1998; Giannakidou 1998, 1999; Guerzoni 2001). Importantly, they further note that negative rhetorical effects become more visible with modals, illustrating how the negative answer arises in (33).

232

(33)

Who the hell would buy the book?

Since the role of modal would is introducing a set of possible worlds, i.e. its modal base K (Kratzer 1981), they treat (33) as a case of explicit quantification over possible worlds and assume domain extension for the NPI. Given this, they propose the following negative attitude for a sentence with a modal.

(34)

Presupposition of negative attitude of wh-the-hell with a modal ∀w∈K [P(x) (w) à Q(x) (w)], for all possible values of x. (where x is the variable of wh-the-hell, P is the property denoted by the wh-the-hell phrase, and Q is the property denoted by the VP) (den Dikken & Giannakidou 2002)

(35)

Negative attitude of (33) ∀w∈K: If there is a person x in w, then x does not buy that book in w. (den Dikken & Giannakidou 2002)

Here modality is expressed as the negative attitude that is relativized to the worlds in the modal base. In particular, they assume that the necessity modal would translates into a universal quantifier, which implies that the negative presupposition must hold in every world in the modal base K. Given this, they account for how negative rhetorical reading is attained. In the case where the world variable is bound by the universal quantifier, the situation eventually becomes the one in which “in no world w would anybody buy that book” which corresponds with the negative rhetorical reading.

233

Thus far it was shown that NPIs trigger domain extension and convey a presupposition of negative attitude, but the negative rhetorical readings arise only when accompanied with a modal. Importantly, the negative attitude in RQs here will give us a plausible foundation for the analysis of negative rhetorical reading in RCs. I assume that the negative attitude in above rhetorical questions is strongly similar to the negative rhetorical effects in the comparative that are triggered by strong NPIs with non-deictic variables and modals. Now recall that RCs convey negative implicature and large difference presupposition, as repeated below.

(36)

a. Jack is richer than you’ll EVer be. i) Negative implicature: You will never be as rich as John. ii) Large difference presupposition: There is a significantly large difference in degree of wealth between Jack and you in any foreseeable future.

b. Jack does volunteer works more often than he lifts a finger to help his wife. i) Negative implicature: Jack very rarely (or never) lifts a finger to help his wife. ii) Large difference presupposition: There is a significantly large difference in frequency between Jack doing volunteer works and him helping his wife.

The negative implicature here can be recast as “negative attitude” in RQs in the sense of den Dikken & Giannakidou, except that in the case of RCs the negative attitude is on the basis of the large difference between the subject and the standard. For instance, in (36a) the speaker has a negative attitude about the financial status of the addressee. Since modality is involved here, the

234

negative attitude is relativized to the worlds in the modal base of necessity. This yields the following negative rhetorical reading:

(37)

Negative attitude of (36a) ∀w∈K: If there is a degree d” to which you will get rich, d” is unimpressive or unimportant since it is significantly smaller than d’ in w.

Unlike wh-the-hell which triggers negative rhetorical reading only in combination with modality, however, I assume that a comparative with a strong NPI like lifts a finger (and without a modal) falls under the category of RCs because it also conveys a negative attitude of the speaker and yields a negative rhetorical reading. The negative rhetorical reading of (36b) seems somewhat similar to the one in (37): it implies that in no world w would the frequency of Jack helping his wife, d”, would be impressive or important since it will be significantly smaller than the frequency of Jack’s doing volunteer works, d’. Then the next question is: How can the latter case without modality induce such negative rhetorical reading? The answer I suggest is that the source of negative rhetorical reading is different for RCs with minimizer NPIs – a presupposition of large difference. A detailed discussion is given in the following section.5

5

This can be understood along the line of the fact that negative rhetorical questions can be formed when questions contain polarity items like wh-the-hell with modality as well as strong NPIs without modality such as “Did he lift a finger to help you?” 235

7.2.2.3. Putting it together: negative rhetorical effects and large difference In this section, I will propose that the NPI-containing than-clause does not have a well defined standard, because of the non-referentiality of the NPI. This is why these cannot be regular comparatives. I further propose that because of the domain extending or shrinking nature of the NPI, we have a high or low value for the degree in standard, and this creates the large difference inference, which is the only sensible inference available in the sentences. To begin, recall that the negative polarity of minimizers like give a penny is due to the indefinite part, and give ten pennies is no longer a NPI because it could be presuppositional and familiar, hence deictic. This predicts that in the following comparative with gives ten pennies, rhetorical effects disappear.

(38)

Grace buys expensive clothes more often than she gives ten pennies to the homeless.

The above sentence must be categorized as a regular DC, rather than a RC, since what the standard clause denotes is perfectly detectable in the context. The variable for the standard degree of frequency d′′ to which she gives ten pennies to the homeless is a legitimate free variable since it is specific enough to be valued in a given context. That is, it could be linked to discourse familiar values. Now consider the following comparative with a non-emphatic weak NPI anyone, which also lacks rhetorical effects:

(39)

Jack is taller than anyone.

236

Based on the observations so far, I propose that what is distinctive in RCs is their dependence on the non-referentiality of the standard, which is indeed a direct contribution of strong/emphatic NPIs. If the non-deictic standard fails to make reference to a specific discourse referent, it is impossible to posit two distinct degrees in the semantics of RCs. In the semantics of comparatives by Kennedy & McNally (2002), the presupposed and familiar standard in DCs like (40) allows for introducing a discourse referent, hence enabling the computation of the maximal degree of what the content of the standard denotes. Thus I follow the standard non-negative analysis (41) for the semantics of DCs.

(40)

Kim is taller than Lee (is).

Non-negative Analysis

max{d′| tall(kim) ≥ d′} > max{d′′ | tall(lee) ≥ d′′}

(41)

a. [[A tall]] = λd′′λx.tall(s) ≥ d′ b. [[than wh Lee is t tall]] = max {d′′ | tall(Lee) ≥ d′′} c. [[DegP –er [than wh Lee is t tall ]]] = λG λx.∃d[d > max{d′′ | tall(Lee) ≥ d′′} ∧G (d′)(x)] d. [[AP [DegP –er than wh Lee is t tall ] tall ]] = λx.∃d′ [d′ > max{d′′ | tall(Lee) ≥ d′′} ∧ tall(d′) ≥ d′] e. [[(40)]] = 1 iff ∃d′ [d′ > max{d′′ | tall(Lee) ≥ d′′} ∧ tall(Kim) ≥ d′]

237

On the other hand, I propose that the semantics of RCs consists of the following two parts: first, building upon Kennedy & McNally (2002), the semantics in (43a) represents the basic inequality relation of (42) which is truth-conditionally equivalent to that of DCs; second, since the thanclause contains a dependent variable, dd′′, the precise semantics in the standard is referentially undefined, as shown in (43b). This means that the maximal degree of what ANYbody else is tall is difficult to be gauged because the domain has been overly extended.

(42)

John is taller than ANYbody (else) is.

(43)

a. [[A tall]] = λd′′λx.tall(s) ≥ d′ b. [[than wh ANYbody else is t tall]] max {dd′′ | tall(ANYbody else) ≥ dd′′}: undefined

As such, a semantics of RCs is similar to that of DCs except that the former contains a deficient variable, which renders the standard undefined in the context. Note further that these semantics for RCs and DCs tell us that the notion of downward entailment (DE) does not offer much help in characterizing comparatives (which turned out to be ambiguous depending on what kind of quantifier they contain as discussed above). Rather, I suggest that the relevant concept should be (non-)veridicality, which is directly linked to (non-)referentiality. In terms of veridicality, the comparative clause in (40) above is veridical since it is possible to refer to one particular maximal degree to which Lee is tall in a given context, and therefore it expresses certainty about, or commitment to, the truth of a sentence. In contrast, the comparative clause of RC in (42) is nonveridical because it contains a defective variable for which the assignment function g cannot

238

assign a value in a main context. This means that the comparative clause can express neither certainty about nor commitment to the truth of a sentence. Based on the definition of non-deictic variable (Giannakidou 2009) repeated below, I assume that the precise degree for dd′′ is unattainable in the context since it is difficult to make reference to every member in the ridiculously extended domain of ANYbody (else).

(44)

Non-deictic variables A variable x is non-deictic iff x cannot be interpreted as a free variable.

And the defective variable can only be valued by being embedded by the negative attitude that is discussed in (37) above. This is consistent with our observation that negative rhetorical effects in RCs are triggered by emphatic/strong NPIs that render the content of the whole clause extremely non-referential. In other words, employment of an exaggerated standard in RCs is a speaker’s strategy for conveying a heightened emotional state, i.e. the negative attitude, toward the content of the standard clause, as seen in 2.2.2.6 This seems to support the idea that the semantics of RCs will be in a way similar to what the negative analyses of comparatives posit, repeated below in (45).

(45)

John is taller than anyone.

Negative Analysis

∃e John is tall to extent e AND NOT [anyone else is tall to extent e]

6

(Ross 1969)

See also Hoeksema (1984) which showed the phrasal comparative any to be a free choice item, and Giannakidou and Yoon (2010) who generalized the analysis.

239

However, there are reasons to be cautious about the semantics with a negative operator. For one thing, I have shown that negativity is not involved in all comparatives but only in a special type of them, namely RCs. Second, even though some kind of negativity exists in RCs, it should not be posited at the level of the syntax of a comparative (as evidenced by Giannakidou and Yoon 2010), contra what has been proposed by the negative analyses. Rather, I propose that it is located at the level of negative implicature that a RC carries. Furthermore, as already noted in the literature on polarity items in RQs (Krifka 1990, 1992, 1995; van Rooy 2003), I assume that the rhetorical effect of a sentence like (45) can be achieved only when the NPI is either morphologically strong (anybody else) or prosodically emphatic (ANYbody). Given this, the refinements that I suggest are the following: Negativity in the comparative is triggered only if a standard clause is non-referential, i.e. only for RCs: as noted in 2.2.1, the vague standard is precisely due to the non-referentiality of the NPI. This accounts for why these cannot be regular comparatives, and also why a presupposition of large difference arises in RCs. Because of the bidirectional nature of the NPI – excessive domain extension or shrinking, we have a markedly high or low value, and this creates the large difference inference discussed in 7.2.1. This assumption is in line with Giannakidou’s (2007) analysis of NPIs like even and minimizers, showing how RQs emphasize the minimal amount. Based on the discussion so far, I propose the rhetorical effects in RCs as below:

(46)

Rhetorical effects in RCs:

(i) Non-referential standard: There is no referential standard. The d′′ cannot be defined within an appropriate domain in context.

240

(ii) Large difference in degree: Because of the strong NPI, you tend to look low or high end of the scale. This creates the large difference in degree.

Before closing the discussion, I present a piece of evidence for the non-referential standard of RCs: Differential measure phrases such as 2 pounds are available only for DCs in which the standard d′′ is anchored to a specific weight of Lee in the context. As illustrated in (47a), however, a RC with a measure phrase is ungrammatical because the maximal weight that Lee will ever reach is unfixed and immeasurable. Thus the vague standard in RCs is incompatible with differential measurements that require the presence of two specific degrees.

(47)

a. *Kim is (exactly) 2 pounds heavier than Lee’ll ever be.

RC

b. Kim is (exactly) 2 pounds heavier than Lee is.

DC

Given this, various properties of RCs can be attributed to strong non-referentiality of the standard, which entails non-veridicality. In 7.2.2, I have shown the following: First, RCs are comparatives with a rhetorical flavor with respect to the content of the standard of comparison; Second, rhetorical effects are attained via comparison to a referentially defective standard; Third, a defective degree variable in a nonreferential standard is salvaged by a negative implicature of RCs; Finally, non-referentiality can be achieved by employing an item that fails to introduce a familiar/presupposed referent or a specific referent within an appropriate domain in context.

241

7.3. RCs are not MCs As a reviewer raises the question of what is the precise difference between RCs and metalinguistic comparatives (MCs: Bresnan 1973; McCawley 1988; Embick 2007; Giannakidou & Stavrou 2008; Morzycki 2008), before moving on, I would like to note a few facts about comparative subtypes: DCs, RCs and MCs. Metalinguistic comparatives as in (48) are defined as deviant comparatives in that they express disapproval as in (49), compared to regular comparatives.

(48) a. Your problems are financial more than legal. b. Your problems are financial rather than legal. (49) “It is more appropriate for me to say that your problems are financial, than to say that your problems are legal.”

(McCawley 1988)

Based on the following contrasts, I argue that RCs are a subtype of DCs only with a nonreferential standard, and an essentially different creature from MCs. In contrast with MCs, DCs and RCs do share properties in morphological, syntactic, and semantic aspects. Just to name a few, first, it is well-known that in MCs it is impossible to have a synthetic comparative form -er (McCawley 1988; Giannakidou & Stavrou 2008; Morzycki 2008), but DCs and RCs are possible with -er.

(50) a. b. (51) a.

Serena is more smart than industrious.

MCs

*Serena is smarter than industrious. *Serena is more rich than Erik.

DCs

242

b. (52) a. b.

Serena is richer than Erik. *Serena is more rich than you’ll ever be.

RCs

Serena is richer than you’ll ever be.

Second, displacement of more to the right is possible in MCs (Giannakidou & Stavrou 2008; Morzycki 2008), but is impossible in DCs and RCs.

(53) Your problems are financial more than legal.

MCs

(54) *Your problems are serious more than mine.

DCs

(55) *Your problems are serious more than mine will ever be.

RCs

Third, MCs are possible with adjectives that are not ordinarily gradable (McCawley 1988; Morzycki 2008), whereas DCs and RCs are impossible.

(56)

Your problems are more financial than legal.

(McCawley 1988) MCs

(57)

*Your problems are more financial than mine.

DCs

(58)

*Your problems are more financial than mine will ever be.

RCs

Finally, MCs are impossible with comparison of deviation (Giannakidou & Stavrou 2008; Morzycki 2008) while DCs and RCs are possible. MCs like (59) can have only the metalinguistic meaning that it is more appropriate for me to say that this table is wide than to say that this table is long, while both DCs and RCs compare two degrees on different measurements.

243

(59)

This table is more wide than long.

(metalinguistic) MCs

(60)

This table is wider than it is long.

(Kennedy 1997b, 2007)(COD) DCs

(61)

The sky would sooner fall than he would budge an inch.

(COD) RCs

According to Giannakidou & Stavrou (2008; Giannakidou & Yoon 2009, to appear), the comparative adverbial MOREML in MC does not compare degrees of (the denotation of) the adjective, but degrees of appropriateness of statements, or preference. The metalinguistic meaning is attitudinal, and expresses a contrast between two propositions.

(62)

[[MOREML]] = λpλq.∃d[R(a)(p)(d) ∧ d> max(λd’[R(a)(q)(d’)])] (GS 2008) where R is a gradable propositional attitude supplied by the context: either epistemic, or preference (desiderative or volitional); a is the individual anchor (Farkas 1992; Giannakidou 1998, 1999): typically, the speaker in this case.

It is defined that the metalinguistic MORE takes two propositional arguments: p (the proposition of the main clause), and q (the proposition of the than-clause) and compares the two propositions in terms of the degree that the speaker believes, prefers, or is willing to assert them. Given that MCs are more about comparison of degrees of appropriateness of statements, or preference on a separate dimension at an ‘epistemic’ level, I assume that just like DCs, RCs are about comparison of degrees of the denotation of the adjective, only with vague standards. I have shown in 7.2.2 that in RCs the non-referentiality of the standard is the main ingredient that is precisely responsible for triggering a rhetorical flavor that is similar to a negative assertion.

244

Now that we know RCs and MCs exist on different levels, we can answer another question raised by a reviewer: how are NPIs licensed in each comparative type? We have seen that what licenses NPIs in RCs is a non-referential standard that gives rise to non-veridicality in the semantics. However, when MCs license NPIs, the licensing force must come from a certain epistemic component that expresses strong inappropriateness of statements, or dispreference. In this respect, NPIs in MCs seem to be indirectly rescued on an attitudinal level (see Giannakidou 2009 for ‘rescuing of NPIs’), whereas NPIs in RCs are legitimately licensed on a semantic level.

7.4. Rhetoricizing the standard: NPIs, expletive negation, and the subjunctive In this section, I extend the analysis and propose that there are three strategies for rhetoricizing the standard in RCs in natural language. The first strategy is to employ strong/emphatic NPIs, as we have seen so far. The second strategy is to incorporate expletive negation in comparatives which typically gives rise to emphatic effects elsewhere. Finally, though English does not provide direct evidence for this, the third strategy is to select for the subjunctive mood.

7.4.1. Negative polarity items Polarity item licensing in questions and comparatives shares similar problems. Standard analyses have difficulty explaining why questions license NPIs since questions are neither necessarily negative nor downward entailing in any straightforward way (Klima 1964; van Rooy 2003). An important problem is that questions, typically with explicit negation, license positive polarity items such as rather or pretty, as shown in (63). Then an immediate question arises: Do all questions freely allow polarity items?

245

(63)

a. Would(n’t) you rather stay here? b. Aren’t you pretty tired?

To answer this question, we need to look beyond the basic meaning of questions. It is generally believed that the semantics of each question is equivalent to a set of possible answers, including positive and negative answers. Then, if it is a true information-seeking question, say “are you tired?”, its meaning should be a set with multiple values: {yes, I am; maybe; … ; no, I’m not}. However, if a speaker has a certain expectation of a positive or negative answer, this kind of presupposition of an expected answer has the power to license positive or negative polarity items, respectively (see Borkin 1971; van Rooy 2003; Ladusaw 2004 for detailed discussion). Thus it has been established that when it comes to polarity item licensing in questions, the crucial factor is the speaker’s presupposition regarding the expected answer, rather than the broad meaning of the question itself. Now turning to the long-standing debate on polarity item licensing in comparatives, we find a fascinating resemblance with the above debate on questions. The problems are: NPIs in clausal comparatives seem to occur quite freely (e.g. Hoeksema 1984), and it is not obvious how comparatives are negative, DE, or nonveridical. As discussed in section 7.2.1, the current consensus seems to be that comparative clause itself may vary in terms of DE or UE depending on the quantifiers it contains (Rullmann 1994 for Dutch comparatives; Hendriks 1995; Heim 2006). Hence monotonicity does not seem to be relevant in characterizing DCs versus RCs. One important consequence of the current proposal, however, is to open the possibility of resolving the variation issues surrounding NPI/PPI-licensing in comparatives at the semanticspragmatics interface. First, in terms of the polarity item licenser, the current system requires a

246

conceptual transition from the traditional monotonicity into the more recent notion of (non)veridicality (Zwarts 1995; Giannakidou 1997, 1998, 2009). Giannakidou (1997) defines veridicality as in (64), stating that a propositional operator F is veridical iff from the truth of Fp we can infer that p is true according to some individual x in his or her epistemic model. This inference can be either an entailment or a presupposition of the sentence where F occurs. On the other hand, F is non-veridical when an inference to the truth of p under F is not possible, and hence non-veridicality indicates a state of unknown or undefined truth value. And a subtype of non-veridicality is anti-veridicality which entails the falsity of p.

(64)

DEFINITION 1. (Non)veridicality for propositional operators

i. A propositional operator F is veridical iff Fp entails or presupposes that p is true in some individual’s epistemic model ME(x); otherwise F is nonveridical. ii. A nonveridical operator F is antiveridical iff Fp entails that not p in some individual’s epistemic model: Fp à ¬ p in some ME(x).

Given this, let us see how distributions of polarity items are explained under the current system. For instance, (65a) provided by the UE analyses, licenses a PPI some because it is a DC with a specific degree that the speaker can refer to. The lack of polarity item in (65b) also indicates that it is a DC.

(65) a. John is taller than some professional basketball players are. b. John is taller than most of his ancestors were.

247

I assume these DCs with PPIs or no NPIs to be ‘veridical’ in the sense that the comparison is based on a definite standard. And the following comparatives with non-emphatic weak NPIs in (66) are DCs as well because the domain for the standard, receiving a regular extension, is still referential.

(66)

a. John is taller than anyone is. b. He is richer than you ever were.

On the other hand, RCs are non-veridical environments which license strong/emphatic NPIs, as shown in Rullmann’s (1995) examples:

(67)

a. He told me more jokes than I cared to write down. b. He said the sky would sooner fall than he would budge an inch. c. John would sooner roast in Hell than give a penny to the Nader campaign. d. Mary buys expensive presents for her assistant more often than she lifts a finger to help her husband.

As already noted, the current proposal on non-referential standards in rhetorical comparatives is partly in line with the previous negative approaches to comparatives (Jespersen 1917; Ross 1969; McConnell-Ginet 1973; Seuren 1973; Klein 1980; Stassen 1984; Larson 1988). A striking parallel between our proposal and the negative analyses is that RCs are characterized as typically involving NPIs and sometimes modals, just as most data presented by the previous negative approaches uniformly contain NPIs in than-clauses as in (68-70), repeated from section 7.2.1.

248

(68)

John is taller than anyone. ∃e John is tall to extent e AND NOT [anyone else is tall to extent e]

(69)

He is richer nor you’ll ever be.

[English dialects]

(70)

She did a better job than what I never thought she would.

[Cockney English]

I further assume that their intuition on anyone in (68) was based on emphatic ANYONE (our assumption here is in line with the insight for questions in Rohrbaugh 1993, Krifka 1995, van Rooy 2003). Instead of their syntactic negative operator, however, all these examples with NPIs are now categorized as rhetorical subtypes in which NPIs are licensed by the non-veridicality driven from a non-referential standard. Furthermore, our analysis crucially diverges from the negative approaches in that such negativity is not assumed for regular DCs, which predicts the distribution of PPIs therein.

7.4.2. Expletive negation Given the current proposal on RCs, we are now able to account for the reason why a negative element such as nor, ne, or never is observed in comparatives of some languages or dialects. Presenting the data in (71-73), the negative approaches in section 7.2.1 claim that these negatives directly indicate the presence of an underlying negative operator in comparatives (Jespersen 1917; Ross 1969; McConnell-Ginet 1973; Seuren 1973; Klein 1980; Stassen 1984; Larson 1988).

(71)

He is richer nor you’ll ever be.

[English dialects]

249

(72)

Jean est plus grand que je ne pensais. Jean is taller

[French]

than I Neg thought

‘Jean is taller than I thought.’ (73)

She did a better job than what I never thought she would.

[Cockney English]

However, we have seen in section 7.2.1 that uniformly positing a negative operator in comparatives is problematic in many respects. Another interesting view on the presence of a negative in comparatives is that it is a marker of comparison. For instance, Price (1990) claims that since languages like French do not have a unique marker of inequality (like than in English), the marker of comparison is not que but rather the negative ne. However, Campos & Sachs (1995) raise a question of why ne is needed as a comparative marker if one still has plus or aussi as markers, and they note that a more serious problem for this view is that languages like Italian and Spanish that do have a unique comparative marker still exhibit such negative particles in comparatives. If it is neither a syntactic negative operator nor a comparative marker, why do some languages contain a negative in comparatives? Based on the contrast with the following contentful negation in English, it has been suggested that the above cases are expletive negation (EN) (von Stechow 1984; Rullmann 1995; Beck, Oda & Sugisaki 2004).7

(74)

a. * I need more salt than I don’t need pepper. b. * John is taller than I didn’t think.

7

Furthermore, Donati (2000) notes that negation in comparative is syntactically distinguished from real negation in that it does not draw up a negative island in Italian. 250

Such negation is traditionally termed expletive negation because it is void of a logical negative force at syntax and semantic interpretation, but it is not strictly ‘explicit’ since it certainly conveys pragmatic emphatic effects. Furthermore, the emphatic effects seem very similar to rhetorical effects that we find in RCs: The occurrence of negatives in comparatives might seem extraordinary at first glance, but it is not surprising at all if we look at the distribution of such negative elements in other environments. As illustrated in (75), EN frequently appears in exclamatives, questions, and subordinate clauses of certain predicates, uniformly encoding a speaker’s negative presupposition towards the content of a proposition (van der Wouden & Zwarts 1993; van der Wouden 1994; Brown & Franks 1997; Utrecht 1998; Brown 1999; Portner & Zanuttini 1999, 2000 among many others).

(75)

Expletive Negation in various environments

a. exclamatives No

ga-lo

magnà tuto!

Neg

has-S.Cl

eaten

[Paduan Italian] (Portner & Zanuttini 2000)

everything

‘He’s eaten everything!’ b. interrogatives Ne dopustil li kto-nibud’

ošibki?

Neg allow

mistake.Gen

Q who-any

[Russian] (Brown 1999)

‘Could someone have made a mistake?’ c. subordinate clauses of adversative predicates Þan I have no doubt Þat it ne schal wel kun

telle Þee of hem. [Old/Middle English]

than I have no doubt that it Neg will fully be-able tell

251

you of them

‘Then I have no doubt that it will be able to tell you about them.’ (Cloud of Unknowing 92. 6-7; van der Wouden & Zwarts 1993)

As noted in chapter 2, according to Portner & Zanuttini, a negative presupposition in (75a) is posited on a likelihood scale such that it is less likely that he’s eaten everything, hence the negative no marks a speaker’s strong surprise, and similar emphatic effects arise from EN in (75b) and (75c). Furthermore, in previous chapters, I have shown that what has been called expletive negation is in fact quite meaningful since crosslinguistically it triggers an emphatic effect in various environments by conveying a negative presupposition. If the negative elements observed in the above comparative data are the same creature, then we can conclude that EN in comparatives is adopted to convey emphatic effects with regard to the contrast between two objects that are being compared. And the analysis I am pursuing offers the following solution: The negative nor in (76), repeated below, is not semantically pleonastic, but rather encode negative rhetorical force toward the content of than-clause. Consequently, EN offers another important strategy for rhetoricizing comparatives, together with the employment of strong/emphatic NPIs.

(76)

He is richer nor you’ll ever be.

[English dialects]

More crucially, analogous effects triggered by EN are observed in rhetorical questions in English (77a) and Russian (77b), and in a dubious question in Russian (77c) (Brown & Franks 1995, 1997), repeated from chapter 2.

252

(77)

a. Who didn’t sleep with Mary?! b. Nu, ne

govoril li ja tebe?

well Neg told

[Russian]

Q I you

‘Well, didn’t I tell you?!’ c. Ne Neg

dopustil li kto-nibuď

ošibki?

allowed Q who-any

mistake.Gen

‘Maybe someone made a mistake?’

A view very close in spirit is found in earlier claims regarding EN (van der Wouden 1994; Espinal 2000) where EN is treated as a kind of NPI since both EN and NPIs display similar distributions such as in exclamatives, questions, and clauses under certain negative predicates. Though I do not attempt to directly argue that EN is a type of NPI, it is worthwhile to note that they have further similarities besides the distributional facts: i) As I argue, EN gives rise to emphatic effects the way that NPIs do in questions and comparatives; ii) Both EN and NPIs are never obligatory; and iii) It will be shown in the following subsection that they typically cooccur with the subjunctive mood. Before closing the discussion on EN, however, it should be noted that the rhetorical force cannot account for all instances of EN in comparatives. As shown in the following sentence (Espinal 2000), EN seems to appear in regular comparatives without necessarily having rhetorical effects.

253

(78)

Maria è più

alta di quanto

(non) lo sia Giovanni.

Maria is more tall of how-much neg

(Donati 2000) [Italian]

it is Giovanni

‘Maria is taller than Giovanni is.’

There seems to be two possible explanations for the emergence of EN in regular comparatives. First, the appearance of negative markers irrespective of comparative subtypes in certain dialects or languages can be thought of as a remnant of the manifestation of emphatic effects which has originated in emphatic comparatives and been stretched to regular degree comparatives via a historical spread or analogy. Another possibility is to extend the flexibility on the kinds of scales that EN triggers. For instance, EN in regular degree comparatives like (78) is assumed to mark the inequality relation between the two degrees. Then again, the dual marking on one semantic relation, i.e. the inequality relation in degree, brings the consequence of putting an emphasis on that semantics.

7.4.3. Subjunctive mood In addition to the employment of NPIs and EN, I suggest that adoption of the subjunctive mood is the third component contributing to the rhetoricization of comparatives. The idea that mood is relevant to the distinction between RCs and DCs relies on an important insight that mood selection seems to be linked to polarity in terms of (non-)veridicality – the indicative mood corresponds to the notion of veridicality while the subjunctive mood corresponds to nonveridicality in languages like Spanish, Italian, French, Greek (Giannakidou 1994, 1995, 2009; Quer 1998; Borschev et al. 2007 for Russian subjunctive). It has been discussed in chapter 4 that, just as polarity items are dependent variables that are referentially deficient and cannot be

254

interpreted deictically, Giannakidou (2009) notes that the subjunctive mood represents a nondeictic time, hence an instance of a polarity dependency of the temporal kind. The assumption relies on the pronominal theory of tense (Partee 1973, 1984; Kratzer 1998) and her earlier idea that the emergence of some polarity items is due to referential deficiency. Furthermore, the assumption that, just like NPI-licensing, the subjunctive mood introduces RCs is consistent with the following empirical links between the subjunctive mood and NPIs that have been proposed in the literature. Nathan & Epro (1984:522) note that “many of the constructions that trigger NPIs in English also license the subjunctive mood in Romance languages.” Furthermore, the close connections between indicative mood and PPIs and between subjunctive mood and NPIs are proposed by a recent study of crosslinguistic polarity properties by D. Levinson (2006), as given in Table 9 (other categories are omitted).

Table 9. Morphosyntactic categories and polarity sensitivity (D. Levinson 2006) Category

Positive polarity

Negative polarity

Mood

Realis (Indicative)

Irrealis (Subjunctive)

On the other hand, it is important for our purposes to recall that a number of important empirical connections between (expletive) negation and the subjunctive have been already noticed in the literature, as discussed in chapter 4. Although the exact nature of this dependency requires further investigation, it will provide at least empirical support for the current proposal that both EN and the subjunctive will trigger an analogous semanticopragmatic contribution for RCs. More crucially, a tight connection between the two is revealed in comparatives. Though both EN and the subjunctive mood are optional in Italian, they have a strong tendency to co-

255

occur (Napoli & Nespor 1976, 1977; Price 1990). Given the crosslinguistically persistent parallels between the subjunctive and NPIs, and between the subjunctive and (expletive) negation, we can infer that selection of the subjunctive mood would be an additional strategy for rhetoricizing comparatives, creating a non-veridical context. This is a welcome result because we have seen in chapter 4 that the use of the subjunctive mood as a rhetorical device is in fact written in the grammars of languages with morphological mood markers: Such rhetorical effects are triggered by the present and imperfect subjunctive mood, as noted in Latin grammars, which is called deliberative subjunctive because it is used to deliberate or doubt about something. Similar effects are conveyed by deliberative subjunctive in rhetorical questions and emphatic negation subjunctive in Greek, and it is analogous in French. Then we can expect that certain languages display a corresponding phenomenon in comparatives as well, and this is indeed the case. A selection of the subjunctive mood seems relevant to emphatic effects: In Italian, Napoli & Nespor (1976, 1977; von Stechow 1984) note that there is a semantic difference between comparatives with the indicative (79a) and with the subjunctive and EN (79b) in that only the latter entails a contradictive presupposition.

(79)

Maria è più intelligente (a) di quanto è Carlo / (b) di quanto non sia Carlo.

[Italian]

‘Mary is more intelligent (a) than Carlo is / (b) Carlo (not) is.’

In similar spirit, Battaglia & Pernicone (1965) note that the subjunctive mood in Italian emphasizes the hypothetical nature of the comparison (Price 1990: 153). Furthermore, in Latin and Greek, comparatives with the subjunctive are characterized as imaginative comparison (Hale 1892).

256

(80)

a. Noli timere quasi assem elephanto des.

[Latin]

‘Don’t be afraid as if you were giving a penny to an elephant.’ b. Serviam tibi tam quasi emeris me argento. ‘I shall serve you as though you had bought me with silver.’

Given that mood also plays a role in rhetoricizing comparatives, I suggest an extension from grammatical mood to notional subjunctive mood for capturing the asymmetries in comparatives in other languages. Here the crucial concept that I depend on is ‘notional mood’ (Jespersen 1924; Portner 1997; Giorgi & Pianesi 1998). Notional mood is distinguished from grammatical mood in the sense of Portner (1997: 182) who states “the firmest definition it (notional mood) can be given is as concerning aspects of meaning (broadly construed) which contribute to the conversational force of a clause or which constrain the attitude someone has toward what it expresses. Nothing more uniform is possible because those aspects involve both semantic and pragmatic factors, and only these factors have theoretical status.” Thus our assumption on notional mood is a broad conception incorporating the semanticopragmatic aspects of mood as well as traditional grammatical mood. The extension of subjunctive-like properties to languages without grammatical subjunctive mood distinction is in part supported by Sigurðsson’s (1990) observation regarding Faroese (a language related to Icelandic). In this language, despite the lack of subjunctive mood marker, the same verbs as Icelandic subjunctive verbs show subjunctive like properties. Also, Progovac (1994) notes that volitional verbs in Serbian/Croatian act like subjunctives even without morphological subjunctive marking.

257

The proposed dichotomy of comparatives is therefore applicable to languages with impoverished mood morphology. For instance, though RCs in English do not offer any explicit grammatical subjunctive marker, I suggest that the fact that modals contribute to rhetorical force as in (88) must be along the lines of what the subjunctive mood does in other languages.

(81)

a. Jack is richer than you’ll ever be.

RCs

b. This work is more than I can stand. c. Grace’s chicken was more than I could be bothered eating.

The idea is based on the semantic framework in which notional mood is treated as a manifestation of modality (Farkas 1985, 1992a, 1992b; Portner 1994, 1997; Giorgi & Pianesi 1998; Roussou 2000). Portner (1997: 192), for instance, illustrates the mood-indicating modals with may, noting that “mood indicating may does not carry modal force of its own, but simply indicates that its clause has a particular conversational use or is in a certain kind of semantic context” (i.e. mood):

(82)

Mood-Indicating May a. In a formal style, occurs inverted in matrix clauses ((83a)) to express a wish. b. Occurs embedded under certain operators which express desires ((83b–c)) or epistemic possibility ((83d)).

(83)

a. May you have a pleasant journey! b. Jack wishes that you may be happy. c. I pray that God may bless you. (from Palmer 1990)

258

d. It is possible that Sue may win the race.

(Portner 1997: (41)-(42))

He further argues that may in (90d) lacks modal force because it is already expressed in the matrix clause (‘is possible’), hence the redundant use of may in the embedded clause indicates properties of modal context for its clause. In section 7.4, in examining what contributes to triggering RCs, I have suggested that the notion of NPIs, EN, and subjunctive mood must be understood as a family of phenomena that are uniformly associated with non-veridicality in natural language.

7.5. Summary and Conclusion In exploring the behaviours of comparatives, the main goal of this chapter was to show that the negativity-related properties in comparatives are quite complex, certainly much more than expected by the unitary negative (Jespersen 1917; Ross 1969; Seuren 1973; Klein 1980; Larson 1988) or non-negative analyses (von Stechow 1984; Rullmann 1995; Kennedy 1997a; Beck, Oda, & Sugisaki 2004). In trying to go beyond the puzzle of negative-like versus non-negative properties in comparative clauses, I proposed that, instead of talking about a unitary phenomenon, it is preferable to delimit the negative-like properties to a subtype of them, RCs. I addressed the question of what the meaning of RCs is and how exactly this meaning is responsible for the properties of RCs, as opposed to DCs. A great part of the discussion was devoted to showing that the containment of strong/emphatic NPIs introduces an extremely fuzzy standard for RCs. I suggested that such fuzziness can be understood as the conception of unfamiliarity (Heim 1982), non-specificity, and non-exhaustivity (Krifka 1995), and showed how the bidirectional domain surpassing of strong/emphatic NPIs triggers such properties.

259

Essentially, these properties contribute to extreme non-referentiality, hence non-veridicality of the standard. This is consistent with the fact that RCs are unavailable under comparatives with differential measurements. The driving idea in this chapter is to show that a negative presupposition concerning the content of the standard in RCs follows directly from a comparison to a non-referential vague standard. I furthermore attempt to explore the analogy between what gives rise to rhetorical flavor in English as well as other languages, and I have identified three such components – negative polarity items, expletive negation, and subjunctive mood. The analysis I advanced here has a number of potentially revealing implications for our understanding of the notion of nonveridicality, suggesting the crucial link between polarity, mood, and negation. This result supports earlier observations that the subjunctive behaves like negative polarity items, marking non-veridicality (Giannakidou 1994, 1995, 2009; Quer 1998; Borschev et al. 2007), and my innovation in this chapter is to suggest that expletive negation is another component belonging to such a category. By positing distinct semanticopragmatic properties for DCs and RCs, the current proposal leads us to predict the systematic asymmetries between comparatives. It thus becomes clear why, as von Stechow (1984) notes, the previous negative approaches are descriptively adequate for data with NPIs while the non-negative approaches are not. The notion of RCs in the sense that I suggest here, importantly, can give us a plausible foundation for the analysis of rhetorical effects in other environments. Another likely domain in which to explore the consequences of this analysis is interrogatives. Although I have shown that RQs share a number of properties with RCs, how far the non-veridicality dependency analysis of polarity items, mood, and negation can be extended to questions remains to be established by future research. If the connection between rhetorical flavor and the other two components turns

260

out to be robust in other environments also, then we can conclude that the conception of nonveridicality can be reducible to at least three types of deficiency: i) deficiency in pronominal referentiality, marked by NPIs; ii) deficiency in tense, marked by the subjunctive; and iii) deficiency in presuppositional polarity, marked by EN.

261

Concluding Remarks

In exploring the phenomena of Evaluative Negation (EN) from a crosslinguistic and crosscontextual perspective, the main goal of this thesis is to show that what has been called expletive negation is in fact semantically meaningful, and syntactically operative. I propose an analysis of EN based on an investigation of a wide range of data, including previously studied data and new data in Korean and Japanese. The contributions made in this dissertation can be summarized as follows. First, this result supports earlier observations (Tovena 1996; Abels 2002, 2005; Portner and Zanuttini 1999, 2000) that EN triggers certain semantic or pragmatic effects, and goes one step further and suggests that EN occurring in a variety of contexts is in fact of a uniform nature. I have identified distinct paradigms of EN across languages that are subject to typical contexts. As shown in table 1 below which is the revised version from the tentative table 2 at the end of chapter 2, the various evaluative contents that EN appears to convey (e.g. scalar implicature in terms of probability, certainty, desirability, directness of speech, temporality, or degree) are now reduced to two sources of evaluative content, unlikelihood or undesirability. Thus EN is shown to be another legitimate function of negation in natural language, where a negative element is adopted for the purpose of circumventing a commitment to the truthful statement and combine this with an attitude. This makes it similar to subjunctive mood.

262

Table 10. Six classes of Evaluative Content of EN (final version) Classes of Evaluative Content EN

Source of

Environments

Evaluative Content

class 1

Amazement

Unlikelihood

exclamatives

class 2

Uncertainty/Doubt:

Unlikelihood

subordinate clauses of doubt, dubitative questions

class 3

Undesirability

Undesirability

subordinate questions,

clauses,

emphatic

exclamatives,

until-

clauses, concessive conditionals class 4

Politeness

Unlikelihood

requests, invitations/proposals

class 5

Temporal Sequence

Unlikelihood or

before/without/unless-clauses

Undesirability class 6

Inequality in degree

Unlikelihood or

comparatives

Undesirability

Second, I demonstrated that EN should be analyzed as a separate creature from regular negation or negative concord items. Rather, I propose my analysis of EN as a mood marker. EN is a subjunctive mood marker, in particular, following the pattern of the mood choice in Greek that is regulated by nonveridicality (Giannakidou 2009). I discuss the parallels in the positive nonveridical environment in the scope of verb ‘hope’. This is unexpected under any account that we have seen, and the fact that EN requires a non-factive complementizer and a subjunctive predicate in Korean and Japanese should come as no surprise.

263

Third, I add a pragmatic component to the semantic analysis. I argue that EN contributes an evaluative dimension. This evaluative dimension is negative anticipation, undesirability, or low likelihood. I show that EN in different environments exhibits a striking parallel in interpretation as an ordering relation. I propose to capture this by means of multidimensionality of conventional implicature (Potts 2005) (as refined in Giannakidou & Yoon 2009, to appear), and show how evaluative semantics of EN exists on a separate dimension from the semantic core of utterance. Contrary to previous views that dismiss the meaning of EN, calling it expletive, my proposal captures the precise role of EN in an utterance. If my analysis is correct, it has one important implication: It allows the generalization that various subspecies of EN in language are indeed part of the grammar. In particular, they are reflexes of grammaticalization of perspective and subjective mode, on a par with subjunctive mood choice and other subjunctive phenomena like metalinguistic comparative as argued in Giannakidou and Yoon (2009, to appear) Fourth, I have proposed that EN is dependent on nonveridicality and the evaluative contents of must be represented by means of multidimensional CI logic. Nonveridicality dependency predicts that the evaluative contents of EN are similar to that of the evaluative subjunctive whose distribution is also dependent on nonveridicality. I discuss how uncommon meanings of EN such as unequal degree or temporal sequence might have been derived from the basic evaluative subjunctive meaning in contexts like comparatives or nonveridical temporal clauses. I further show how these semantics of EN, modifying the whole utterance, can be captured by the CI logic of Potts (2005, 2007) and Giannakidou and Yoon (2009, to appear). The evaluative sense that is posited for EN and subjunctive markers, affords us a precise interpretation that can represent the complex minds of a speaker. The current analysis further implies that various subspecies of EN in language are indeed part of the grammar. In particular,

264

they are reflexes of grammaticalization of perspective and subjective mode, on a par with subjunctive mood choice. Fifth, I explore the syntactic configurations of subordinate subjunctive clauses and subordinate EN-clauses in Korean (and potentially Japanese). One of the questions that will arise is whether the matrix clause and the EN-clause in Korean are hypotactically or paratactically connected. This will inevitably involve the Germanic Embedded Verb Second (EV2). The dynamic syntax of EN (and subjunctive) complements that I propose affords the following theoretical implications: First, by revealing the parallels with EV2 constructions that are in subjunctive, the syntax of EN constructions further supports the current proposal that EN is subjunctive marker. Second, it suggests that the syntax of EN complements in Korean must be understood along the lines of Jespersen’s insight of paratactic negation in Old/Middle English. Third, it accounts for why the complementizer for subordinate EN clauses is in an analogous form with a question particle in Korean and Japanese. Finally, it furthermore supports the possibility that negation in tag-questions, which are classical paratactic configurations, can be understood as a subspecies of EN. The analysis can be extended to Japanese since Japanese EN has exactly the same properties. Finally, I offer a case study of rhetorical comparatives (RC). In exploring the semanticopragmatic properties of RCs, it is shown that rhetorical effects can be triggered by negative polarity items, evaluative negation, and the subjunctive mood. This result supports an important insight that negative polarity items and the subjunctive are of similar nature (Giannakidou 1994, 1995, 2009; Quer 1998; Borschev et al. 2007), and goes one step further to suggest another important link between evaluative negation and the subjunctive mood. This

265

analysis implicates that the three components are closely connected under the principle of nonveridicality. My overall approach, if correct, has two important consequences: First, we have explanations for the crosslinguistic variation in subordinate EN between Japanese/Korean and other languages. Second, it offers a plausible way of systematically characterizing EN in other environments such as exclamatives, questions, nonveridical before clauses and comparatives.

266

Bibliography Abels, K. 2002. Expletive (?) Negation. In J. T. Bloomington (eds.), Proceedings of FASL 10, ID: Michigan Slavic Publications. Abels, K. 2005. “Expletive negation” in Russian: A conspiracy theory. Journal of Slavic Linguistics 13, 5-74. Allan, K. 2006. Clause-type, primary illocution, and mood-like operators in English. Language Science 28, 1-50. Anderson, A. R. 1951. A Note on Subjunctive and Counterfactual Conditionals. Analysis 11, 3538. Bachrach, A. and R. Katzir. 2006. Right-Node Raising and Delayed Spellout. Interphases. Battaglia, S. and V. Pernicone. 1965. La grammatica italiana. Turin, Loescher. Beck, S. 2006. Positively Comparative. H.-M. Gaertner, S. Beck, R. Eckart, R. Musan and B. Stiebels (eds.). Bhatt, R. 1998. Argument-Adjunct Asymmetries in Rhetorical Questions. NELS 29. Bolinger, D. 1968. Postposed main phrases: An English rule for the romance subjunctive. Canadian Journal of Linguistics 10, 125-197. Bolinger, D. 1972. Degree words. De Hague: Mouton. Borkin, A. 1971. Polarity Items in Questions. Papers from the 7th Regional Meeting of Chicago Linguistic Society, 53-62. Borschev, V., E. Paducheva, B. Partee, Y. Testelets, and I. Yanovich. 2007. Russian genitives, non-referentiality, and the property-type hypothesis. Formal Approaches to Slavic Linguistics. Bresnan, J. 1973. Syntax of the comparative clause construction in English. Linguistic Inquiry 4(3): 275–343. Brown, S. 1999. The syntax of negation in Russian: a minimalist approach. CSLI. Brown, S. and S. Franks. 1995. Asymmetries in the scope of Russian Negation. Journal of Slavic Linguistics 3, 239-287. Brown, S. and S. Franks. 1997. The Syntax of Pleonastic Negation in Russian. In W. Browne (eds.), Formal Approaches to Slavic Linguistics – The Cornell Meeting, Ann Arbor: Slavica Publishers, 135-164.

267

Campos, H. and H. Sachs. 1995. A review of Susan Price’s ‘Comparative Constructions in Spanish and French Syntax’. Hispanic Linguistics 6-7: 547-572. Chafe, W. L., 1986. Evidentiality in English conversation and academic writing. In: Chafe, W.L., Nichols, J. (eds.), Evidentiality: The Linguistic Encoding of Epistemology. Ablex Publishing Corp, Norwood, NJ. Chatzopoulou, K. to appear. Negation in Greek: a diachronic study. Ph.D. dissertation. University of Chicago. Chierchia, G. 2006. Broaden your Views. Implicatures of Domain Widening and the Spontaneous Logicality of Language. Linguistic Inquiry 37.4: 535-590. Cinque, G. 1999. Adverbs and functional heads: a cross-linguistic perspective. New York, Oxford University Press. Culy, C., 1994. Aspects of logophoric marking. Linguistics 32, 1055-1094. Curme, G. 1931. A Grammar of the English Language in three Volumes, Vol. III. Syntax. Boston: D. C. Heath and Company. Dayal, V. and J. Grimshaw. 2009. Subordination at the interface. Department of Linguistics, Rutgers University. De Haan, G. J. 2001. More is Going on Upstairs than Downstairs: Embedded Root Phenomena in West Frisian, Journal of Comparative Germanic Linguistics 4: 3–38, Den Dikken, M. 1995. Extraposition as intraposition, and the syntax of English tag questions. Ms., Vrije Universiteit Amsterdam. Den Dikken, M. and Giannakidou, A. 2002. From hell to polarity: aggressively non-D-linked Wh-phrases as polarity items. Linguistic Inquiry 33: 31–61. Donati, C. 2000. A note on negation in comparison. Quaderni del Dipartimento di Linguistica – Università di Firenze 10: 55-68. Eilam, A. and T. Scheffler. 2006. Until and Expetive negation in Modern Hebrew. Swarthmore Workshop on Negation and Polarity, April 14-15. Embick, D. 2007. Blocking effects and analytic/synthetic alternations. Natural Language and Linguistic Theory 25: 1–37. Ernst, T. 2009. Speaker-oriented adverbs. Natural Language and Linguistic Theory 27(3), 497544.

268

Espinal, M. T. 1997. Non-negative negation and wh-exclamatives. In D. Forget, P. Hirschbühler, F. Martineau, M. L. Rivero (eds.), Negation and polarity. Syntax and Semantics. Amsterdam: John Benjamins, 75-93. Espinal, M. T. 2000. Expletive negation, negative concord and feature checking. Catalan Working Papers in Linguistics 8: 47-69. Espinal, M. T. 2000. EN, Negative Concord and Feature Checking. CatWPL 8, 47-69. Farkas, D. F. 1985. Intensional Descriptions and the Romance Subjunctive Mood. Garland, New York. Farkas, D. F. 1992a. On the semantics of subjunctive complements. In P. Hirschbűhler et al. (eds.), Romance Languages and Modern Linguistic Theory. John Benjamins, Amsterdam and Philadelphia, 69–104. Farkas, D. F. 1992b. Mood choice in complement clauses. In I. Kenesei and E. Pleh (eds.), Approaches to Hungarian vol. 4: The structure of Hungarian, JATE, Szeged, 77–103. Farkas, D. F. 2002. Specificity Distinctions. Journal of Semantics 19: 1-31. Farkas, D. 2003. Assertion, belief and mood choice. Ms., University of Santa Cruz. Fischer, O. 1992. Syntax, In N. Blake (eds.), The Cambridge History of the English Language II, 1066-1476, Cambridge: Cambridge University Press. Frellesvig, B. 2008. On the verb morphology of Old Japanese. In B. Frellesvig, M. Shibatani and J. C. Smith (eds.), Current issues in the history and structure of Japanese, Tokyo: Kurosio Publishers, 219-52. Fox, D. and D. Pesetsky. 2004. Cyclic Linearization of Syntactic Structure. Theoretical Linguistics 31.1-2, 1-46. Gaatone, D. 1971. Etude descriptive du systéme de la negation en français contemporain. Geneva, Librairie Droz. Giannakidou, A. 1994. The semantic licensing of NPIs and the Modern Greek subjunctive. Language and Cognition 4: 55–68. Giannakidou, A. 1995. Subjunctive, habituality and negative polarity. Semantics and Linguistic Theory, vol. 5. CLC Publications, Cornell University, Ithaca, NY, 132-150. Giannakidou, A. 1997. The Landscape of Polarity Items. PhD thesis, University of Groningen. Giannakidou, A. 1998. Polarity Sensitivity as (Non) veridical Dependency. John Benjamins, Amsterdam/Philadelphia.

269

Giannakidou, A. 1999. Affective dependencies. Linguistics and Philosophy 22, 367-421. Giannakidou, A. 2007. The landscape of EVEN. Natural Language and Linguistic Theory 25: 39-81. Giannakidou, A. 2009. The dependency of the subjunctive revisited: temporal semantics and polarity. In J. Quer (eds.), Lingua, special volume on Mood. Giannakidou, A. and M. Stavrou. 2008. Metalinguistic comparatives and negation in Greek. In D. Hill (eds.), MITWPL. Giannakidou, A. and S. Yoon. to appear. The subjective mode of comparison: metalinguistic comparatives in Greek and Korean, Natural Language and Linguistic Theory. Giannakidou, A. and S. Yoon. 2010. No NPI licensing in clausal comparatives. Proceedings of the 46th Chicago Linguistic Society (CLS 46). ed. by T. Grinsell, A. Baker, J. Thomas, R. Baglini, J. Keane. University of Chicago. Giannakidou, A. and S. Yoon. 2009. ‘Metalinguistic comparatives in Greek and Korean: Attitude semantics, expressive content, and negative polarity items’. In Proceedings of Sinn und Bedeutung 18. Giannakidou, A. and S. Yoon. 2008. Comparative and Negation: three types of comparatives in Greek

and Korean. Proceedings of the 6th International Conference of Cognitive Science (ICCS 6). The Korean Society for Cognitive Science. pp. 386-389. Seoul: Sigma Press.

Giannakidou, A. and F. Zwarts. 1999. Aspectual properties of temporal connectives. In A. Mozer (eds), Greek Linguistics ‘97: Proceedings of the 3rd International Conference on Greek Linguistics. Ellinika Grammata, Athens. 104-113. Giorgi, A. and F. Pianesi. 1997. Tense and Aspect: from semantics to morphosyntax. Oxford: Oxford University Press. Giorgi, A. 2006. A Syntactic Way to Subjunctive, University of Venice Working papers in Linguistics, vol 16. Givón, T. 1994. Irrealis and the subjunctive. Studies in Language 18: 265-337. Givón, T. 1980. The Binding Hierarchy and the Typology of Complements, Studies in Language, 4.3: 333-377. Guerzoni, E. 2001. Even-NPIs in questions. The Proceedings of NELS 32, Amherst, MA: GLSA Publishers, 153–170. Gunlogson, C. 2003, True to Form: Rising and Falling Declaratives as Questions in English, Outstanding Dissertations in Linguistics, Routledge, New York. 270

Hale, W. G. 1892. Mode and Tense in the Subjunctive ‘Comparative Clause’ in Latin. The American Journal of Philology 13.1: 62-70. Han, C.-H. 1997. Deriving the Interpretation of Rhetorical Questions. In E. Curtis, J. Lyle, and G. Webster (eds.), Proceedings of WCCFL 16, Palo Alto, CA, CSLI. Haspelmath, M. 1997. Indefinite pronouns. Oxford Studies in Typology and Linguistic Theory, Oxford: Oxford University Press. Haverkate, H. 2002. The Synatx, Semantics, and Pragmatics of Spanish Mood. John Benjamins. Heim, I. 1982. The Semantics of definite and indefinite noun phrases. Ph.D. thesis, University of Massachusetts/Amherst. Heim, I. 1984. A note on negative polarity and downward entaillingness. Proceedings of NELS 14, GLSA, University of Massachusetts, Amherst, 98-107. Heim, I. 1992. Presupposition projection and the semantics of attitude verbs. Journal of Semantics 9, 183-221. Heim, I. 2006. Remarks on comparative clauses as generalized quantifiers. Ms. Hendriks, P. 1995. Comparatives in categorial grammar. Ph.D. thesis, Rijksuniversitet Groningen. Higginbotham, J. 1988, Contexts, Models, and Meanings: A Note on the Data of 874 Semantics, in R. Kempson (ed.), Mental Representations, Cambridge, pp. 29–48. Hinzen, W. 2003. Truth's Fabric. Mind & Language 18(2): 194-219. Hoeksema, J. & D.-J. Napoli. 1993. Paratactic and Subordinative So. Journal of Linguistics 29: 291-314. Hoeksema, J. 1983. Negative Polarity and the Comparative. Natural Language and Linguistic Theory 1: 403–434. Hoeksema, J. 1984. Categorial Morphology. Ph.D. thesis, Groningen University. Hooper, J. 1975. On assertive predicates. In J. Kimball (eds.), Syntax and semantics 4, 91-124. New York/London: Academic Press. Hrafnbjargarson, G. H. and A.-L. Wiklund. 2009. General Embedded V2: Icelandic A, B, C, etc. Working Papers in Scandinavian Syntax 84:21-51.

271

Jabłónska, P. 2003. Quirky n-words in Polish: NPIs, Negative Quantifiers or neither? In A. Dahl, K. Bentzen, and P. Svenonius (eds.), Proceedings of the 19th Scandinavian Conference of Linguistics 31.1. Jespersen, O. 1940. A Modern English Grammar on Historical Principles, Part V: Syntax, vol. 4. 1st edn., Copenhagen: Ejnar Munksgaard. Jespersen, O. 1924. The philosophy of grammar. London: Allen & Unwin. Jesperson, O. 1917. Negation in English and other languages. Copenhagen: Host. Johnson, K. to appear. Towards Deriving Differences in How Wh Movement and QR are Pronounced, Lingua. Johnson, K. 2009. PEPPER and PF Movement: Reactions to Yamashita, in Proceedings of the 2007 Workshop in Greek Syntax and Semantics at MIT, MIT Working Papers in Linguistics 57, 409–425. Joly, A. 1967. Negation and the Comparative Particle in English. Cahiers de psychomécanique du langage 9, Québec: Les presses de l’Université Laval. Kadmon, N. and F. Landman. 1990. Polarity sensitivity any and free choice any. In M. Stokhof and L. Torenvliet (eds.), Proceedings of the 7th Amsterdam Colloquium, ITLI, Amsterdam, 227-251. Kadmon, N. and F. Landman. 1993. Any. Linguistics and Philosophy 16: 353–422. Kaufmann, S. and Y. Takubo. 2005. Non-veridical uses of Japanese expressions of temporal precedence. In McGloin, N. H. and J. Mori (eds.), Proceedings of the Fifteenth Conference in Japanese/Korean Linguistics (JK 15), pages 358-369. Stanford, CA: CSLI Publications. Karttunen, L. and S. Peters. 1979. Conventional Implicature. In C.-K. Oh and D. Dinneen (eds.), Syntax and Semantics, Presupposition, Academic Press, New York, Vol. 11: 1-56. Kay, P. 1990. Even. Linguistics and Philosophy 13: 59-111. Kennedy, C. and L. McNally. 2002. Scale Structure and the Semantic Typology of Gradable Predicates. Language 81.2: 345-381 Kennedy, C. 1997a. Projecting the Adjective. Ph.D. thesis, University of California, Santa Cruz. Kennedy, C. 1997b. Comparison and polar opposition. Proceedings of SALT 7, CLC Publications, Ithaca. Kempchinsky, P. 2009. What Can the Subjunctive Disjoint Reference Effect Tell Us About the Subjunctive? Lingua Vol. 119: 1788-1810.

272

Klein, E. 1980. A semantics for positive and comparative adjectives. Linguistics and Philosophy 4, 1-45. Klein, E. 1991. Comparatives. In A. von Stechow and D. Wunderlich (eds.), Semantik/semantics. An international handbook of contemporary research, Berlin: de Gruyter, 673-691. Klima, E. 1964. Negation in English. In J. Fodor and J. Katz (eds.), The structure of language, Prentice-Hall. Kratzer, A. 1981. The Notional Category of Modality. In H. J. Eikmeyer, and H. Rieser (eds.), Words, Worlds, and Contexts. Berlin, 38-74. Kratzer, A. 1998. More analogies between pronouns and tenses. In D. Strolovitch and A. Lawson (eds.), Semantics and Linguistic Theory, vol. 8., CLC, Cornell University Press. Krifka, M. 1990. Polarity phenomena and alternative semantics. In M. Stokhof and L. Torenvliet (eds.), Proceedings of the Seventh Amsterdam Colloquium, ITLI, Amsterdam, 277-301. Krifka, M. 1992. Some remarks on polarity items. In D. Zaefferer (eds.), Semantic Universals and Universal Semantics, Floris Publications, Berlin. Krifka, M. 1995. The Semantics and Pragmatics of Polarity Items. Linguistic Analysis 25: 209257. Kuno, S. 1973. The Structure of the Japanese Language. The MIT Press. Ladusaw, W. A. 1979. Polarity Sensitivity as Inherent Scope Relations. Ph.D. thesis, University of Texas at Austin, Distributed by IULC, Bloomington, Indiana, 1980, Also published by Garland Press: Outstanding Dissertations in Linguistics, New York. Ladusaw, W. 2004. Biased questions. talk given at UCSC, Santa Cruz. Larson, R. 1988. Scope of Comparatives. Linguistics & Philosophy 11: 1-26. Lasersohn, P. 1999. Pragmatic Halos. Language 75 (3): 522-551. Lasersohn, P. 2005. Context dependence, disagreement, and predicates of personal taste. Linguistics and Philosophy 28: 643-86. Lasersohn, P. 2008. Quantification and Perspective in Relativist Semantics. University of Chicago Compositionality Workshop, May 9. Lasersohn, P. 2009. Relative Truth, Speaker Commitment, and Control of Implicit Arguments. Synthese 166 (2).

273

Lawler, J. M. 1971. Any Questions? Papers from the 7th Regional Meeting of Chicago Linguistic Society, 163-173. Lebeaux, D. 1988. Language acquisition and the form of the grammar, Ph.D. thesis, University of. Massachusetts. Lee, Y.-S. and L. Horn. 1994. Any as indefinite + even. Ms., Yale University. Lee, F. A. 1995. Negative Polarity Licensing in Wh-questions: The Case for Two Licensers. Paper presented at the 69th Annual Meeting of the Linguistic Society of America. Levinson, D. 2006. Polarity Sensitivity in Inflectional Morphology, BLS 32. Levy, I. H. 1981. Manyoshu: a translation of Japan's premier anthology of classical poetry. Volume 1, Princeton: Princeton University Press. Linebarger, M. 1980. The grammar of negative polarity. Ph.D. dissertation, MIT. Linebarger, M. 1987. Negative Polarity and Grammatical Representation. Linguistics and Philosophy 10: 325-387. Martin, T. ms. Non-negative negations in Catalan. A problem for compositionality? NYU. Mari, A. and Martin, F. 2007. Tense, ability and Actuality Entailments. in Dekker, P. et al (eds), Proceedings of the Sixteenth Amsterdam Colloquium. Matthewson, L. 1998. Determiner Systems and Quantificational Strategies: Evidence from Salish. The Hague: Holland Academic Graphics. McCawley, J. 1988. The Syntactic Phenomena of English. Chicago: University of Chicago Press. McCawley, J. D. 1980. An un-syntax. In Syntax and semantics 13: Current approaches to syntax, ed. by Edith Moravcsik and Jessica Wirth, 167-93. McCawley, N.A. 1978. Another Look at No, Koto, and To: Epistemology and Complementizer Choice in Japanese. In Problems in Japanese Syntax and Semantics ed. by J. Hinds and Irwin Howard. Kaitakusha. McConnell-Ginet, S. 1973. Comparative constructions in English: A syntactic and semantic analysis. PhD Thesis, University of Rochester. Meibauer, J. 1990. Sentence mood, lexical category filling, and non-propositional nicht in German. Linguistische Berichte 130, 441-463. Mithun, M. 1999. The Languages of Native North America. Cambridge University Press, Cambridge.

274

Montague, R. 1969. On the nature of certain philosophical entities. The Monist 53: 159–194. Reprinted in R. Thomason (eds.), Formal Philosophy: Selected Papers of Richard Montague, Yale University Press, New Haven, 247–270. Morzycki, M. 2008. Metalinguistic Comparison in an Alternative Semantics for Imprecision. In M. Abdurrahman, A. Schardl, and M. Walkow (eds.), Proceedings of NELS 38, GLSA Publications, Amherst, Mass. Muller, C. 1978. La négation explétive dans les constructions comlétives, Langue Français 39, 76-103. Muller, C. 1991. La négation en français: syntaxe, sémantique et elements de comparaison avec les qutres langues romanes, Geneve. Napoli, D. J. and M. Nespor. 1976. Negatives in comparatives. Language 52 (4): 811-38. Napoli, D. J. and M. Nespor. 1977. Superficially illogical “non”. In M. P. Hagiwara (eds.), Negatives in comparatives. Nathan, G. S. and M. W. Epro. 1984. Negative polarity and the Romance subjunctive. In P. Baldi (eds.), Papers from the XIIth Linguistic Symposium on Romance Languages. Nemoto, N. 1993. Chains and Case Positions: A study from Scrambling in Japanese. Ph.D. dissertation, University of Connecticut. Ogihara, T. 1998. The ambiguity of the te iru form in Japanese. Journal of East Asian Linguistics 7, 87-120. Ono, H. 2005. On the use of the speaker (narrator)’s mental attitude to determine the selection of the Japanese complementizer no or koto: a reply to Suzuki. Journal of Pragmatics. Palmer, F. R. 2001. Mood and Modality. Second ed, Cambridge University Press, Cambridge. Palmer, F. R. 2003. Modality in English: theoretical, descriptive and typological issues. In R. Facchineti, M. Krug, R. R. Palmer (eds.), Modality in Contemporary English, Mouton de Gruyter, Berlin/New York, 1-17. Partee, B. H. 2005. Reflections of a Formal Semanticist as of Feb 2005. Manuscript, University of Massachusetts, Amherst. Picallo, C. 1985. Opaque Domains. PhD thesis, CUNY. Portner, P. 1992. Situation theory and the semantics of propositional expressions. Amherst: University of Massachusetts, PhD dissertation.

275

Portner, P. 1994. A Uniform Semantics for Aspectual -ing. In M. Gonzàlez (eds.), The Proceedings of NELS 24, GLSA, Amherst, 507-517. Portner, P. 1997. The semantics of mood, complementation, and conversational force. Natural Language Semantics 5, 167-212. Portner, P. 1999. The semantics of mood. Glot International 4 (1), 3-8. Portner, P. and R. Zanuttini. 1999. The force of negation in wh exclamatives and interrogatives. In R. L. Horn, and Y. Kato (eds.), Negation and Polarity. Oxford: Oxford University Press, 193-231. Portner, P. and R. Zanuttini. 2000. The Characterization of Exclamative Clauses in Paduan. Language. Potts, C. 2005. The Logic of Conventional Implicatures. OUP. Potts, C. 2007. The expressive dimension.Theoretical Linguistics 33(2). Price, S. 1990. Comparative Constructions in Spanish and French Syntax. Routledge. Progovac, L. 1994. Negative and Positive Polarity. Cambridge. Progovac, L. 1993. Negative Polarity: Entailment and Binding. Linguistics and Philosophy 16.2. Quer, J. 1998. Mood at the Interface. PhD thesis, University of Utrecht. Quer, J. 2001. Interpreting Mood. Probus 13, 81-111. Quer, J. 2009. Twist of mood: the distribution and interpretation of the indicative and the subjunctive. Lingua, special issue on Mood. Rohrbaugh, G. 1993. Focus and the Licensing of Negative Polarity Items. Ms., University of Texas at Austin. Ross, J. R. 1969. A Proposed Rule of Tree-Pruning. In D. A. Reibel and S. A. Schane (eds.), Modern Studies in English: Readings in Transformational Grammar. Roussou, A. 2000. On the left periphery: modal particles and complementizers. Journal of Greek Linguistics 1: 63-93. Rullmann, H. 1995. Maximality in the Semantics of WH-Constructions. Ph.D. thesis, UMass, Amherst.

276

Sadock, J. M. 1971. Queclaratives. In D. Adams, M. A. Cambell, V. Cohen, J. Levins, E. Maxwell, C. Nygren, and J. Reighard (eds.), Papers from the Seventh Regional Meeting of the Chicago Linguistic Society 7: 223-232. Sadock, J. M. 1974. Towards a Linguistic Theory of Speech Acts. Academic Press, New York, San Francisco, London. Sawada, O. 2009. Pragmatic aspects of implicit comparison: an economy-based approach. Journal of Pragmatics 41: 1079-1103. Sawada, O. 2010. Pragmatic Aspects of Scalar Modifiers. Ph.D. thesis, University of Chicago. Schwarzschild, R. 2005. Semantics of Degree and Amount. lecture note at LSA Institute. Schwarzschild, R. and K. Wilkinson. 2002. Quantifiers in Comparatives: A Semantics of Degree based on Intervals. Natural Language Semantics 10:1-41. Searle, J. 1969. Speech Acts. Cambridge University Press, Cambridge. Searle, J. and D. Vanderveken. 1985. The Foundations of Illocutionary Logic. Cambridge University Press, Cambridge. Seuren, P. 1973. The Comparative. Generative Grammar in Europe. Siegel, L. 2009. Mood selection in Romance and Balkan. Lingua 119.12: 1859-1882 Speas, M. 2004. Evidentiality, logophoricity and the syntactic representation of pragmatic features. Lingua, 114, 255-276. Stalnaker, R. 1968. A theory of conditionals. In N. Rescher (eds.), Studies in logical theory, American Philosophical Quarterly, Monograph: 2, Oxford: Basil Blackwell, 98-112. Stalnaker, R. 1984. Inquiry. Cambridge, MA: MIT Press. Stassen, L. 1984. The comparative compared. Journal of Semantics 3: 143-182. Suzuki, S. 2000. Japanese complementizers: interactions between basic characteristics and contextual factors. Journal of Pragmatics 32, 585–621. Terrell, T. D. and J. Hooper. 1974. A semantically based analysis of Mood in Spanish. Hispania, 57, 484-494. Tovena, L. M. 1996. An expletive negation which is not so redundant. Grammatical Theory and Romance Languages. Tovena, L. M. 1998. The Fine Structure of Polarity Sensitivity, New York, London.

277

Travis, C. 2003. The semantics of the Spanish subjunctive: its use in the natural semantic metalanguage. Cognitive Linguistics 14, 47-67. Uchibori, A. 2000. The Syntax of Subjunctive Complements: Evidence from Japanese. Ph.D. Dissertation, UConnecticut Distributed by MITWPL. Uriagereka, J. 2008. Syntactic anchors. Cambridge University Press. Uriagereka, J. and E. Torrego. 2002. Parataxis. Derivations: exploring the dynamics of syntax. Routledge Leading Linguists. Utrecht, D. 1998. EN in comparatives. Going Romance, XIIth Symposium on Romance Linguistics. Van der Wouden, T. and F. Zwarts. 1993. A Semantic Analysis of Negative Concord. SALT III. Van der Wouden, T. 1994. Polarity and 'illogical negation'. In M. Kanazawa and C. J. Pinon (eds.), Dymanics, Polarity, and Quantification, Stanford: CSLI, 17-45. Van der Wurff, W. 1999. On EN with adversative predicates in the history of English. In I. T.-B. van Ostade, G. Tottie, and W. Van der Wurff (eds.), Negation in the History of English, Mouton de Gruyter. Van Rooy, R. 2003. Negative polarity items in questions: Strength as relevance. Journal of Semantics 20: 239–274. Villalta, E. 2006. Context dependence in the interpretation of questions and subjunctives, Neuphilologische Fakultat, Universitat Tubingen: PhD Dissertation. Villalta, E. 2008. Mood and gradability: an investigation of the subjunctive mood in Spanish, Linguistics and Philosophy 31, 467-522. Von Stechow, A. 1984. Comparing Theories of Comparison. Journal of Semantics 3: 1-77. Watanabe, Yasuko. 1984. A study of complementization in Japanese: To, Koto, and no. Research on Language & Social Interaction 17:4, 351-368. Wehmeier, K. F. 2005. In the mood. Journal of Philosophical Logic. Yang, D.-W. 1983. The Extended Binding Theory of Anaphors. Language Research 19, S. 169192. Yoon, S. to appear. A Structural Asymmetry in Intervention Effects. Lingua.

278

Yoon, S. 2009a. Expressivity of Non-truthconditional Negation”. Paper presented at The 32nd GLOW Colloquium: semantics Nantes, France.

workshop, "Modes of Composition”. April 15-18, Université de Nantes,

Yoon, S. 2009b. Expletive Negation in Japanese and Korean. Japanese/Korean Linguistics, vol. 18. ed. by Marcel den Dikken and William McClure. Stanford: CSLI Publications. Yoon, S. 2009c. Mood in Abstract Complementizer: Altaic vs. non-Altaic languages. 2009. Proceedings of the fifth Workshop on Altaic and Formal Linguistics (WAFL 5), ed. by Ryosuke Shibagaki and Reiko Vermeulen. pp. 373-384, Cambridge, MA: MITWPL. Yoon, S. 2009d (to appear). Metalinguistic Comparatives: from Greek to Korean. Proceedings of the 9th International Conference on Greek Linguistics. ed. by A. Chatzopoulos, A. Ioannidou, and S. Yoon. University of Chicago.

Yoon, S. 2008a (to appear). Rhetorical Comparatives: polarity items, expletive negation, and subjunctive mood. Journal of Pragmatics. Yoon, S. 2008b. Syntax and Semantics of Negation. Current Issues in Unity and Diversity of Languages, Congress Book of the 18th International Congress of Linguists (CIL 18), pp. 2541-2558, John Benjamins Publishers. Zwarts, F. 1995. Nonveridical contexts. Linguistic Analysis 25: 286–312.

279