University of Southern California Dissertations and Theses

On the representation and licensing of Q and Q-dependents

(USC Thesis Other)

On the representation and licensing of Q and Q-dependents

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content INFORMATION TO USERS This manuscript has been reproduced from the microfilm master. UMI films the text directly from the original or copy submitted. Thus, some thesis and dissertation copies are in typewriter face, while others may be from any type of computer printer. H ie quality of this reproduction is dependent upon the quality of the copy submitted. Broken or indistinct print, colored or poor quality illustrations and photographs, print bleedthrough, substandard margin*;, and improper alignment can adversely affect reproduction. In the unlikely event that the author did not send UMI a complete manuscript and there are missing pages, these will be noted. Also, if unauthorized copyright material had to be removed, a note will indicate the deletion. Oversize materials (e.g., maps, drawings, charts) are reproduced by sectioning the original, beginning at the upper left-hand comer and continuing from left to right in equal sections with small overlaps. Each original is also photographed in one exposure and is included in reduced form at the back of the book. Photographs included in the original manuscript have been reproduced xerographically in this copy. Higher quality 6" x 9" black and white photographic prints are available for any photographs or illustrations appearing in this copy for an additional charge. Contact UMI directly to order. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. University Microfilms international A Beil & Howell Information Company 300 North Zeeb Road. Ann Arbor. Ml 48106-1346 USA 313/761-4700 800/521-0600 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. ON THE REPRESENTATION AND LICENSING OF Q AND Q-DEPENDENTS by Daeho Chung A Dissertation Presented to the FACULTY OF THE GRADUATE SCHOOL UNIVERSITY OF SOUTHERN CALIFORNIA In Partial Fulfillment of the Requirement for the Degree DOCTOR OF PHILOSOPHY (Linguistics) August 1996 Copyright 1996, Daeho Chung Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. UMI Number: 9705084 UMI Microform 9705084 Copyright 19%, by UMI Company. All rights reserved. This microform edition is protected against unauthorized copying under Title 17, United States Code. UMI 300 North Zeeb Road Ann Arbor, MI 48103 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. UNIVERSITY OF SOUTHERN CALIFORNIA THE GRADUATE SCHOOL UMVERSrTY PARK LOS ANGELES. CALIFORNIA 90089 This dissertation, written by C h u n g , D a e h o ...................... under the direction of k . l s Dissertation Committee, and approved by all its members, has been presented to and accepted by The Graduate School, in partial fulfillment of re quirements for the degree of DOCTOR OF PHILOSOPHY Dean Date DISSERTATION COMMITTEE Chairperson Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Acknowledgments a I would like to express my deepest gratitude to all the people who have made my life at USC and this dissertation possible. Since there are so many individuals who have contributed greatly to my achievements, I will have trouble naming them all. Nevertheless. I would like to name a few individuals among the many I extend my appreciation towards. I would like to start by thanking Joseph Aoun, who has been my academic advisor during my entire stay at USC. I am immensely indebted to him academically and non-academically. His discussions with me in his office and on occasion during walks along the campus paths filled me with valuable insights and helpful feedback. The sharing of his linguistic knowledge has always impacted the ideas in my work. His enthusiasm for linguistics deeply impressed me. Even after becoming Dean of Faculties, Joseph Aoun never lost his passion for linguistics. Rather, he pressed forward to write even more productive linguistic works. I am also deeply thankful to Hajime Hoji, who has had a profound impact on my studies at USC. He was a member of all the committees that formed for my academic procedure: the screening paper committee, the qualifying paper committee, and the dissertation committee. Benefits from discussions with him and from his point-by- point comments on my papers are beyond description. His way of doing linguistics had a lot of positive influence on mine. He always stressed that we should not be Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. theory translators since there are a lot more tangible and interesting things that are waiting for us to tackle. He also advised me to understand UG based on my own native language. Even informal conversations and meetings with him enriched my life. I thank my outside dissertation committee member, Nam-Kil Kim, for his warm support and encouragement. I also thank Mrs. Kim, Dr. Kim’s wife. Samonim. my teacher’s wife, invited Korean students in the linguistics department for delicious turkey feasts every Thanksgiving. Our after-dinner conversations about our past, present, and future lives will be remembered forever. I also thank all the faculty members in the linguistics department. Special thanks to Barry Schein, who was my first and only formal semantics teacher. He showed me the 'Tarsky’s world,’ proved the deep meaning hidden in the simple sentence 'snow is white,’ and performed so many ‘events’ in the world of syntax and semantics. I am greatly indebted to his huge linguistic knowledge and to his willingness to spare his valuable time to discuss my papers with me. He generously read my first dissertation draft and gave me valuable comments. Correspondences with him through e-mail were also very fruitful for me. His replies to my messages were always quick and very useful. I would also like to thank Jean-Roger Vergnaud. Discussions with him encouraged me to extend and expand my ideas. If one were to ask him a question, he would be ready to discuss it for hours. I thank Audrey Li for her warm personality and her willingness to help me with everything. I discussed nearly every paper that I Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. wrote while I was at USC. and it gave me a lot of help. I thank Alicja Gorezka. my phonology teacher, who gave me a lot of advice with my phonology paper. I also thank Bernard Comrie, who taught me typology, Maria-Luisa Zubizarreta, who taught me Romance Syntax. I also appreciate the chance to leam from visiting scholars such as Yuki Kuroda and Giuseppe Longobardi. Much thanks to all the students in the department I enjoyed discussing linguistics and also chatting over non-linguistic matters with my friends and colleagues at USC. They were all nice and helpful. Thanks to Elabbas Benmamoun, Pat Schneider. Yongjin Kim, Tim Shi, Charlotte Reinholtz, Kevin Russel, Ke Zu, Robert Belvin. Alfredo Amaiz, Jeongdal Kim, Dami Lee, Peter Petrucci, Kyungsun Chang, Marcus Maia, Dongin Cho, Kyungan Kim, Kaoru Horie, Shin Watanabe, Liliana Sanchez. Elena Herburger, De Haan Ferdinand, Charles Paus, Abdessalam Elomari, Susan Kalt. Ingrid Shyu. Nancy Antrim, Jose Camacho, Roland Hinterholzl, Gorka Elodieta. Pablo Albizu, Maki Watanabe, Keiko Miyagawa, Yuki Matsuda, Hiroshi Aoyagi. Kyung-Kook Kim, Sechang Lee, Miao-Ling Ann Hsieh, Haewon Kim, Ibtissam Kortobi, Lina Choueri, Hong-Keun Park, Jongha Kim, Xiu-Zhi Joe Wu, Hui-Ju Grace Li, and Ayumi Ueyama. Special thanks to my senpaynim ‘senior’ Yongjin Kim, who took very good care of my family when we first came to the States. Special thanks to Abdessalam Elormari, from whom I learned a lot both academically and non- academically. He gave a lot of precious comments on my papers. He often invited my family to his home and served us wonderful dishes. He is indeed a remarkable Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. friend. I will never forget the true friendship he showed me during the 1992 riots. He willingly accompanied me to pick up my wife through the dangerous streets full of mobsters and looters. Special thanks to Nancy Antrim, who corrected the English on almost every paper that I have written. I also cherish the discussions that I had with Shin Watanabe about linguistics and non-Iinguistic matters. I also thank all the staff members in the department, especially to Laura Reiter, Cathy Stubaus, and Linda Culver. I was also very fortunate to have been given opportunities to talk with excellent scholars outside of USC also; among them are Chris Barker, Lisa Cheng, Young-Mee Yu Cho, Peter Cole, Peter Culicover, Stan Dubinsky, Naoki Fukui, Nobert Homstein. James Huang, Sharon Inkelas, Pauline Jacobson, Jonathan Kaye, Richie Kayne, Allan Hyun-Ok Kim, Soo-Won Kim, Yoshi Kitakawa, Hilda Koopman, Yuki Kuroda. Fred Landman, Howard Lasnik, Alec Marantz. Samuel Martin, Shigeru Miyagawa, Dona Napoli, Zairo Nune, David Petsesky, Mamoru Saito, Peter Sells, Tim Stowell, Anna Szabolcsi, Ken Wexler, John Whitman, and others. Talks with them produced a lot of fruit for me. While I stayed in the States, I was able to keep in touch with my old friends and make new friends both in the local area and in other states. Special thanks to UCLA friends, Kyu-Hyun Kim, Jongho Jun, Hyosang Lee, Hyunoo Lee. Seungho Nam. Yong-Yae Park, and Kyung-Hee Suh. Discussions with them broadened my views on linguistics. Meeting friends at conferences and workshops also gave me a lot of Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. insight and fun. Thanks to Hyunki Ahn, Hee-Rak Chae, Young-Jun Chang. Chan Chung, Yun-Suk Chung, Koji Hoshi, Yun-Sun Jung, Jeong-Seok B C im . Jin-Hwan Kim, Jong-Bok Kim, No-Ju Kim, Ranghyeyoon Kim, Yookyung Kim. Eun-Joo Kwak. Hisa Kitahara. Masa Koizmi, Eun-Ju Kwak. Chang-Bong Lee. Felicia Lee. Jeongshik Lee, Sookhee Lee, Young-Suk Lee, Feng-Hsi Liu, Sui-Sang Mok, Keun- Won Sohn, Sung-Ki Suh, Yuji Takano, James Yoon, Jae-Hak Yoon, Eun-Jung Yu. Jae-Il Yum. and many others. I would also like to thank unsanimtul, my dear teachers in Korea. Special thanks to Dong-Whee Yang, who invited me to the beauty of generative grammar and taught me how to work and what to work on while I was in Korea. I thank him for his guidance, which I found very helpful, judging from my experiences in the States. I also learned a lot from the professors in the English Education Department and in the English Department at Seoul National University. Among the professors in the English Education Department are Sangbeom Cheon, Juk-Ryun Hwang, Kye-Soon Lee, Maeng-Sung Lee, Yong Moon, and Myungho Shim. Among the professors in the English Department, Byungtae Cho, Choon-Hak Cho, Han-Kon Kim, In-Sook Kim, Byungkun Lee, and Namshiek Park. I also benefited greatly from the discussions with other professors in Korea. Thanks to Sung-Ho Ahn, Chung-Nam Cha, Hyun-Sook Choe, Jae-Woong Choe, Myung-Yoon Kang, Yongse Kang, Hae- Yeon Kim, Yong-Suk Kim, Young-Joo Kim, Young-Suk Kim, Young-Sun Kim. Byung-Choon Lee. Chong-Min Lee, Chung-Min Lee, Ik-Hwan Lee, Keedong Lee. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. v ii Sun-Woo Lee, Yak-Woo Park, Hyun-Kwon Yang, Jong-Ryeol Yoon, and many others. I would also like to thank my friends and colleagues in Korea: Nam-Ho Cho. Moon-Sup Han, Young-Shik Hwangbo, Sun-Woong Kim, Tae-Soo Kim, Yeon-Seung Kim, Kyung-Jae Lee, Phil-Hwan Lee, Yong-Gil Lee, Sun-Ok Song, Soo-Hyun Suh. Cha-Hoon Yi, and others. I also thank all the members of my Church, The Korean Philadelphia Presbyterian Church. They helped me feel at home here in America. Special thanks to Senior Pastor Rev. Chun-11 Cho, who always blessed and prayed for my family. Thanks to Rev. Samuel Kwang-Sam Kim, Rev. Hun-Sung Park, Rev. Jae-Il Yoo, Rev. Young- Mo Kim, Rev. 11-Young Song, Rev. Ho Kim, Rev. Ki-Jung Kim, and Rev. Jung-Hee Ryu. Thanks to Elder Joon-Ki Paek and his wife, Bible Woman. Keum-Ae Paik, who considered my family as their own. Thanks to Bible Woman Jeong-Ok Kim, who always prayed for us and treated my wife like one of her daughters. I also thank all other church members. I will never forget the kindness that they showed us and the fellowship that we shared. Thanks to Elder Sang-wan Han and his wife, who belonged to another church but invited my family very often to their dinner table and taught us the deep meaning of the Bible. Special thanks to Jeanette Cochrane, Christine Ji, John Kim, Kathy Lee, and Hong- Keun Park for proofreading my dissertation. I deeply thank my father G. Chung and my mother N. Kim. My father has always believed in me, and has given me the warmest support and encouragement to the end. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. I am sure my mother has not skipped a day without praying for me and my family. They waited a long time for the completion of my degree. It is now time to repay what I have received from them. I thank my wife Sookhee for enduring the hardships that we faced and for taking such wonderful care of our kids, Esther Haeyoung and Daniel Haemyoung. Lastly but mostly, I thank my Lord Jesus Christ, who has kept me in His grace and love. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Table of Cootents Acknowledgments....................................................................................................... ii Abstract...................................................................................................................... xii List of Abbreviations.................................................................................................xiv Chapter One. Introduction: Outline of Chapters...........................................................1 Chapter Two. Representation of Functional Categories and their Licensing Modes.................................................................... 16 2.1. Introduction....................................................................................................16 2.2. Representation of Functional Categories as a Set of Formal Features...........1 8 2.2.1. Dual Properties of Morphologically Complex Words.............................18 2.2.1.1. The Weak Projectionist Hypothesis.............................................. 20 2.2.1.2. Two Strict Projectionist Views: Incorporation vs. Checking 23 2.2.2. Functional Categories as a Set of Formal Features................................30 2.3. Two Modes of Functional Category Licensing............................................. 31 Chapter Three. Scope Licensing Modes of WH-Phrases (WPs) in situ.................... 41 3.1. Introduction.................................................................................................... 41 3.2. Examples of Dependent Licensing..................................................................44 3.2.1. Multiple WH-Fronting in East European Languages............................... 45 3.2.2. NPI Licensing in Korean......................................................................... 47 3.2.3. Multiple WPs in situ............................................................................... 51 3.3. On the Distribution of WPs in situ: Nominal/Non-NominalAsymmetries..............................................................53 3.3.1. WPs in situ in English-Type Languages.................................................. 54 3.3.2. WPs in situ in Korean-Type Languages.................................................. 57 3.4. On Movement Analyses.................................................................................. 61 3.4.1. The ECP Approach Combined with a Comp Indexing Mechanism.........61 3.4.2. Nishigauchi’s Pied-Piping Theory........................................................... 65 3.5. Two Options of WP Licensing: Binding Relation (BR) vs. Exclusive SPEC-Head Relation ( ESHR)....................................................... 75 3.5.1. Licensing ofNon-Nominal WPs in situ: ESHR.......................................75 3.5.2. Licensing of Nominal WPs in situ: Binding............................................82 3.5.3. Licensing of WPs in SPEC of C P............................................................ 87 3.6. Deriving the BR vs. ESHR Disparity: Nominal Properties of Question Clauses......................................................... 89 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 3.6.1. Basics of Binding.................................................................................... 89 3.6.2. Nominal vs. Non-Nominal WPs: Binding vs. Movement........................90 3.6.3. Question Clauses as Nominal Projections................................................91 3.6.3.1. Embedded Question Clauses as DPs............................................... 93 3.6.3.1.1. Complements of P ....................................................................93 3.6.3.1.2. Lack ofECM ........................................................................... 95 3.6.3.1.3. Center Embedding and the CRP.............................................. 98 3.6.3.1.4. Case Morphology................................................................... 100 3.6.3.1.5. Adnominal Conjugation in Question Endings........................ 101 3.6.3.1.6. Evidence in Conjunction........................................................104 3.6.3.2. Nominal Properties of Matrix Questions....................................... 107 3.6.3.2.1. Adnominal Endings in Some Question Forms in Korean 107 3.6.3.2.2. Evidence in Other Languages................................................114 3.7. Conclusion..................................................................................................118 Chapter Four. Anti-Superiority Effects (ASEs), Cyclicity, and the Economy..........121 4.1. Introduction...................................................................................................121 4.2. Previous Analyses......................................................................................... 127 4.2.1. A. Watanabe (1991)...............................................................................127 4.2.1.1. Generalization................................................................................1 27 4.2.1.2. A. Watanabe’s (1991) Explanation of the ASE..............................130 4.2.1.3. Problems......................................................................................... 135 4.2.1.3.1. Conceptual Problems..............................................................135 4.2.1.3.2. Empirical Problems............................................................... 144 4.2.2. Saito (1992)........................................................................................... 151 4.2.2.1. An Additional WP Effect and its Cyclicity.................................... 151 4.2.2.2. Free Ride on Nominal WPs............................................................154 4.2.2.3. Problems......................................................................................... 161 4.2.3. S. Watanabe (1995)................................................................................164 4.2.3.1. S. Watanabe’s (1995) WCO Account of ASEs..............................164 4.2.3.2. Problems.......................................................................................168 4.3. An Economy Based Approach to the ASE ................................................... 171 4.3.1. The Core Case of the ASE and the Economy of Derivation .................. 171 4.3.1.1. Licensing of Adverbial WPs in Terms of QBR..............................172 4.3.1.2. ASE and the Economy.................................................................. 173 4.3.2. Cyclicity of the Additional WP Effects and the Economy...................... 179 4.3.3. Exemption from the Ban on Multiple Occurrence of Adjunct WPs 181 4.3.4. Other Combinations of WPs and the Adjunction Operation.................. 183 4.3.5. Non-Nominal WP Movement Cases......................................................190 4.3.6. Some Elusive Cases...............................................................................196 4.4. Conclusion................................................................................................... 200 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Chapter Five. A Unified Interrogative Analysis of Korean WH-Phrases...............202 5.1. Introduction................................................................................................. 202 5.2. 3-QPs as Reduced Indirect Questions..........................................................207 5.2.1. C.-S.Suh (1990): Decomposition of -inka into -i and -nka................ 207 5.2.2. C.-S. Suh’s (1990) Motivations...........................................................209 5.2.3. Additional Motivations........................................................................210 5.3. V-QPs as Indirect Questions.........................................................................217 5.4. Properties of the Copula i and their Relations to 3-QPs and V-QPs............ 222 5.4.1. Post-Vowel Deletability.......................................................................222 5.4.2. Categorial Selection: NP or PP as a Complement................................ 223 5.4.3. Incorporation........................................................................................228 5.5. Some Queries.............................................................................................. 232 5.5.1. Deletability of Q-Preds.........................................................................233 5.5.2. Distinction of 3-QPs from V-QPs........................................................234 5.6. On Indefinite Quantifier Analyses of Korean WPs....................................... 239 5.6.1. Chang’s (1976) Focus Dependence Analysis....................................... 240 5.6.1.1. Focus as an Interrogative Converter............................................. 240 5.6.1.2. Difficulties with the Focus Dependence Analysis........................ 244 5.6.1.2.1. Other Non-WH-Existentials................................................. 244 5.6.1.2.2. Universals.............................................................................244 5.6.1.2.3. Lack of Focus in Declaratives ...............................................246 5.6.1.2.4. Exclusive Interrogatives........................................................ 246 5.6.1.2.5. Exclusive Indefinites.............................................................. 253 5.6.1.3. Conclusion....................................................................................254 5.6.2. Kim’s (1991) Unified Quantifier Analysis........................................... 254 5.6.2.1. Motivations for the Unified Quantifier Analysis...........................255 5.6.2.1.1. Morphological Paradigm...................................................... 256 5.6.2.1.2. Lack of Overt WH-Movement..............................................257 5.6.2.1.3. Scope Interactions.................................................................258 5.6.2.2. Difficulties with the Unified Quantifier Analysis..........................262 5.6.2.2.1. Implication of Morphological Paradigm.................................262 5.6.2.2.2. [+/-wh] Features and Checking..............................................263 5.6.2.23. WPs and QR........................................................................... 265 5.6.2.2.4. WPs vs. Other QPs................................................................ 272 5.6.2.2.5. Exclusive Interrogatives........................................................ 273 5.6.2.2.6. Problems with the Well-formedness Condition......................275 5.6.2.2.7. Problems with Question Morpheme Raising..........................278 5.6.2.3. Conclusion....................................................................................286 5.7. Concluding Remarks.................................................................................... 286 Bibliography............................................................................................................. 288 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. XII Abstract This work investigates the following inter-related issues on WH-constructions: how the question force Q of a question clause is formally represented and licensed in UG. how its dependents, i.e., WH-phrases (WPs), are scopally identified by Q. and how WPs feature-matrically are composed. Q is postulated universally as a strong abstract formal feature in a C° and it is argued to be licensed in overt syntax either by a SPEC-head relation or by a head- head relation, i.e., either by hosting an XP (WP) in its SPEC or by incorporating an X° (a [+wh] complementizer or a question conjugated predicate). The difference in the Q-licensing modes is responsible for the typological variation in question forms i.e.. WH-raising and WH-in-situ. Either mode of Q-licensing obeys the economy principle. Q-licensing by a binary operation (simple merge) vacuously satisfies the economy principle. Q-licensing by attract-a satisfies the principle, a being the highest WP or the highest predicate with a question conjugation. A heterogeneous scope licensing is proposed: licensing via a binding relation (BR) for nominal WPs (n o m WPs) and licensing via an exclusive SPEC-head relation (ESHR) for adverbial WPs (^W Ps). Q, arguably a [+N], can bind n o m WPs, but not a d v WPs, under the assumption that BR is restricted between nominal elements. Thus. n o m WPs can remain in situ, whereas a d v WPs must move to the SPEC of Q. The difference in the licensing mode is argued to be responsible for the syntactic Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. asymmetries: a d v WPs are not only bounded to Q but also compete with other WPs for Q, whereas n o m WPs show neither property. Observing that ^W Ps are neither bounded to nor compete for Q when they are c- commanded by a clause mate n o m WP, I propose a third type of scope licensing. Such a d v WPs are argued to be adjoined to the n o m WP and licensed by a quasi-binding relation (QBR), the hosting n o m WP being licensed by a BR. The QBR option should be taken over the ESHR option, whenever possible, since the former requires a shorter movement. The anti-superiority effect and its cyclic manifestation are also treated as a reflection of whether a derivation feeds or fails to feed the QBR option under the assumption that clause internal scrambling in these languages is costless. Despite the interrogative vs. quantificational ambiguity, every Korean WP is analyzed as containing a [+wh] feature to be licensed by Q. The unified interrogative analysis is possible by treating existential and universal quantifiers based on WPs as reduced indirect questions. Various facts support the interrogative analysis of the WP-based quantifiers: such existential and universal quantifiers are internally clausal like questions; existential and universal licensers contain a question particle; and the licensers function as scope barriers like question particles. with permission of the copyright owner. Further reproduction prohibited without permission. List of Abbreviations 3-QP existential quantifier V-QP universal quantifier Acc accusative case marker Adn adnominal ending ADV adverbial izer a d ywp adverbial WP Aff Affirmation AffP Affirmation Phrase ASE Anti-superiority Effect Asp aspect marker BR binding relation C complementizer Cl classifier Compl complement Con Contrastive CRP Case Resistance Principle Dat dative case marker DE declarative sentence ending ECM Exceptional Case Marking ECP Empty Category Principle ESHR exclusive SPEC-head relation Fut future tense Gen genitive case marker Hon honorific marker Imp imperative sentence ending E P I interrogative polarity item LF Logical Form LM licensing morpheme Neg negation morpheme NegP negation phrase Nmz nominalizer Nom nominative case marker NPI negative polarity item n o m \yp nominal WP OP question operator Pass passive morpheme Pi plural marker Pres present tense Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. XV Prop propositive sentence ending Pst past tense Q question force/mood Q-Pred question selecting predicate QBR quasi-binding relation QE question sentence ending QP quantifier QR Quantifier Raising Rel relative clause ending SPEC specifier SS Surface Structure Top topic marker UG Universal Grammar Voc vocative WCO Weak Crossover WP WH-phrase Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 1 Chapter One Introduction: Outline of Chapters This dissertation deals with some theoretical issues pertaining to how the interrogative force encoded in a [+wh] C° is formally represented and licensed, and how interrogative polarity items, i.e., WH-phrases (henceforth WPs), are scopally identified and feature-matrically composed. As an introduction to the ensuing chapters, this chapter briefly summarizes what will be discussed in each chapter. In chapter 2, the following two questions will be addressed: how the interrogative force is represented in the sentence grammar, and how it is activated as an interrogative function. The answers to these questions are sought from answering more general questions, such as how functional categories in general are represented and how they are licensed in the grammar. As for the first question, it will be hypothesized that functional categories are universally represented as a set of abstract formal features devoid of any phonological or semantic features, as illustrated below: Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Evidence for the abstract nature of functional categories will be provided on the basis of Kitagawa’s (1986) observation of Japanese complex predicates, Chomsky’s (1992, 1994) analysis of verb raising, Ladusaw’s (1992, 1994 UCLA colloquium) analysis of negative concord phenomena in Romance languages and some dialects of English, and Laka’s (1990) SigmaP analysis. As for the second question, i.e., the question of how functional categories are licensed, I argue that UG allows a head-head agreement as a licensing mode of the abstract features in functional categories in addition to a SPEC-head relation, which has been considered as the unique way of licensing functional categories in the literature.1 In other words, the formal feature of a functional category F is licensed in one of the following two ways: (2) a. b. FP YP SPEC-head agreement [+f] I head-head agreement 1 See, for example, Rizzi (1991) and Haegman (1992), etc., who claim that a SPEC-head relation is responsible for WH-Criterion and Neg-Criterion, etc. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. The former relation is achieved by placing in the SPEC of F an XP with a relevant feature, while the latter by incorporating an X° with a relevant feature. It will be argued that (typical) English interrogation exemplifies the relation in (2a), while Korean interrogation exemplifies the relation in (2b). Note that SPEC of a [+wh] C° is filled with a WP in English, whereas a [+wh] C° hosts a predicate with a question •y conjugation in Korean. In chapter 3, two claims will be made pertaining to the scope licensing of WPs in situ. First, every WP as an interrogative polarity item must be appropriately related to its triggering head, a [+wh] C°, in order to be scopally licensed. Although the licensing of the interrogative force encoded in a [+wh] C° is accomplished either by attracting a WP to its SPEC or attracting a predicate with a question conjugation, every WP must be scopally licensed by having an appropriate relation with a [+wh] C°. whether the WP has moved to the SPEC of the [+wh] C° or remains in situ. Otherwise, they are not properly interpreted at LF, violating the principle of full interpretation. The second claim to be made in chapter 3 is that there are two modes of scope licensing of WPs in situ depending on the categorial status of the WPs; nominal WPs are licensed by having a binding relation with a [+wh] C°, whereas non-nominal WPs are licensed by having an exclusive SPEC-head relation (to be defined) with a [+wh] It is argued in Chung (1995a) that sentential negation is similarly licensed in English and Korean under the assumption that English not is an XP and moves to SPEC o f NegP, while Korean predicates affixed with ani raise to the head of NegP. with permission of the copyright owner. Further reproduction prohibited without permission. C0.3 Thus, nominal WPs are licensed in situ without recourse to movement, while non-nominal WPs must move to its scope position at LF. The difference in the scope licensing modes is argued to be responsible for the asymmetries in the distribution of WPs: Non-nominal WPs are bounded to its scope position and compete with other WPs for that position, while nominal WPs show neither property. I also attempt to derive the binding vs. movement dichotomy in the scope licensing modes from the categorial properties of question clauses as well as WPs themselves. It will be shown that there are various indications that question clauses in English and Korean bear nominal properties. (See also Lee 1967, Kim 1974, and Fukui 1986.) The nominal feature can bind nominal WPs, but not non-nominal WPs, under the assumption that binding is restricted between two nominal elements. In chapter 4, there will be a discussion as to why multiple WH-constructions in Korean and Japanese display the so-called anti-superiority effect (henceforth ASE) (Saito 1982, 1989, 1992, A. Watanabe 1991, and S. Watanabe 1995, among others), and why it is manifested cyclically (Saito 1992). An economy based account will be proposed for the ASE and its cyclic manifestation. The ASE runs as follows. As was discussed in Saito (1982, 1989, 1992), A. Watanabe (1991), and S. Watanabe (1995) among others, a nominal WP in Japanese The difference among WPs has been characterized in various ways; argument vs. adjunct asymmetry (Chomsky 1981. Huang 1982, etc.); referential vs. non-referential asymmetry (Rizzi 1991, Cinque 1990, etc.); and nominal vs. adverbial asymmetry (Tsai 1994). I contend that the nominal vs. adverbial (or non- nominal) dichotomy best describes the facts to be discussed in this work. See some discussion in section 3.6. of chapter 3. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 5 can c-command an adverbial WP, but not vice versa. Korean behaves like Japanese in this respect. The schematic structures and examples are given below:4 (3) a .... n o m WP ... a d v WP... QE (QE=question ending) 1 I c-command b. *... “^WP ... n o m WP ... QE I I c-command (4) (=Watanbabe 1991:18, his (33)) a. ?Kimi-wa nani-o naze katta no? you-Top what-Acc why bought QE 'Why did you buy what?’ b. *Kimi-wa naze nani-o katta no? you-Top why what-Acc bought QE ‘What did you buy why?’ (5) a. ?ne-nun mwues-ul wav sa-ess-ni? you-Top what-Acc why buy-Pst-QE ‘Why did you buy what?’ * The WPs in examples (4) and (5) are adjacent. But even when they are not adjacent, the same contrast exists. Some speakers accept (b) sentences as well as (a) sentences depending on the location of pause and stress. Since it is not clear how these factors affect the judgment, our discussion will be restricted to the cases with a neutral intonation. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 6 b. *ne-nun wav mwues-ul sa-ess-ni? you-Top why what-Acc buy-Pst-QE 'What did you buy why Additionally. Saito (1992) observes that the ASE can be suspended by adding a clause mate nominal WP but not by a nominal WP in a higher clause, as schematically represented in (6): (6) a. ...n o m WP ... ^ W P ... n o m WP... b.* ...n o m WP ... [C P ... a d v WP... n o m WP ... For example, the sentence in (4b) becomes grammatical if the subject is replaced by a nominal WP as in (7a) below, but addition of a nominal WP in a higher clause does not induce a similar effect on the ASE in a lower clause as in (7b):5 (7) a. dare-ga naze nani-o katta no? who-Nom why what-Acc bought QE ‘Who bought what why?’ 5 Again there is some controversy as to the grammatical judgment of sentences like (7b). S. Watanabe (1995). for example, accepts them, although not as readily as sentences like (7a). Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 7 b. (=Saito 1992:25, his (60c)) ?*dare-ga [John-ga naze nani-o katta to] omotteru no? who-Nom J.-Nom why what-Acc bought C think QE 'Q who thinks that John bought what why.’ (7b) is as awkward as the sentence in which no nominal WP is available in the higher clause. The same contrast seems to exist in Korean as exemplified in (8): (8) a. nwu-ka wav mwues-ul sa-ess-ni? who-Nom why what-Acc buy-Pst-QE *Q who bought what why?’ b. ?* nwu-ka [John-i wav mwues-ul sa-ess-ta-ko] sayngkakha-ni? who-Nom J.-Nom why what-Acc buy-Pst-DE-C think-QE ‘Q who thinks that John bought what why.’ An economy account will be proposed for both of the above contrasts under the following three assumptions: (i) WP licensing in terms of binding is less costly than WP licensing in terms of movement (cf. Tsai 1994), (ii) non-nominal WPs in situ can be licensed by being adjoined to a nominal WP at LF as well as by moving to the SPEC of a [+wh] C° as mentioned previously (cf. Saito 1992), and (iii) clause internal scrambling, as opposed to scrambling across a clause boundary, is purely optional and costless (cf. Fukui 1993). Thus, the least costly way of licensing WPs in (4a) and (5a) Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 8 is to adjoin the non-nominal WP to the nominal WP in the same clause. The resultant complex WP. being a nominal, is licensed by having a binding relation with the nominal feature that question clauses have. Let us call this type of WP licensing "a quasi-binding relation” (henceforth, QBR). It will be argued that the deviance of (4b) and (5b) comes from the inapplication of a clause internal scrambling of the nominal WP over the adverbial WP. If this were to occur, the sentences in (b) would be exactly like those in (a) and the non- nominal WP would be licensed by the QBR option. The unacceptability of sentences such as (7b) and (8b) is similarly explained. They are unacceptable because the nominal WP in the embedded clause has not undergone scrambling over the adverbial WP. which is costless by assumption. If it had, the adverbial WP would have been licensed by being adjoined to the nominal WP without having to move all the way to the SPEC of CP. It will be claimed that there are some cases where adverbial WPs should move to the SPEC of a [+wh] C° even when there are nominal WPs which share their interrogative scope. For example, consider the following Japanese sentence in (9a) and the Korean counterpart in (9b): (9) a. dare-ea [John-ga naze ringo-o katta to] omotteru no who-Nom J.-Nom why apple-Acc bought C think QE ‘Q who thinks that John bought apples why.’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 9 b. nwu-ka [John-i wav sakwa-lul sa-ess-ta-ko] sayngkakha-ni? who-Nom J.-Nom why apple-Acc buy-Pst-DE-C think -QE *Q who thinks that John bought apples why/ These sentences are grammatical as opposed to the ones in (7b) and (8b). The only difference between these sentences and the ones in (7b) and (8b) is that the embedded object is replaced by a non-WH element in the sentences in (9). If the adverbial WP in the embedded clause were allowed to adjoin to the nominal WP in the matrix clause, then the ungrammaticality of the sentences in (7b) and (8b) would not be explained. It will be argued that the adverbial WP in the embedded clause cannot be adjoined to the nominal WP in the higher clause due to the assumption that scrambling across a clause boundary is not costless and furthermore an A-adjunction of an element across a clause boundary is not allowed. Thus the adverbial WP in (9) must move to the SPEC of the [+wh] C°. The QBR option is restricted to the adjunction of an adverbial WP to a nominal WP. but this need not be stipulated in the system since the other combinations are either unnecessary or useless: the adjunction of nominal WPs to a WP is unnecessary since they can be licensed by a binding relation, and the adjunction of an adverbial WP to another adverbial WP is useless since the hosting WP is non-nominal and has to move to Q anyway. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 10 Chapter 5 addresses internal structures of quantificational expressions that contain WPs in languages like Japanese and Korean. As discussed in Kuroda (1965) and Nishigauchi (1986) among others, Japanese WPs have different readings depending on the environment they appear in. WPs in Korean behave like Japanese WPs in this respect, as discussed in C.-S. Suh (1985,1990). (See Chang 1976, C.-M. Suh 1987, and S.-W. Kim 1989a, 1989b, 1991, for different views.). Consider the schematic structures and examples below: (10) Interrogative: [... WP ... ]-QE (QE=Question ending) a. nwu-ka sakwa-lul mek-ess {-ni/-ci/-na. etc.} ?6 who-Nom apple-Acc eat-Pst-QE 'Who ate the apple?’ b. John-i mwues-ul mek-ess{-ni/-ci/-na. etc.}? J.-Nom what-Acc eat-Pst-QE ‘What did John eat?’ (11) Existential Quantifier: W?-(i)nka/-(i)nci a. nwukwu-nka o-ess-ta. who-ENKA come-Pst-DE ‘Someone came.’ b. John-i mwues I-inka/-inci I mek-ko iss-ta. J.-Nom what-INKA eat-ing be-DE ‘John is eating something.’ 6 Korean has various question endings depending on the registers and speech styles. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (12) Universal Quantifiers (I):7 WP-(i)na/-(i)kena/-(i)tunci/-(i)lato/-(i)lcilato. etc. a. nwukwu(-na/-tunci/-lato/IciIato 1 cal sal-ki wenha-n-ta. who-ENA/-ITUNCI/-ILATO/-ILCILATO well live-Nmz want-Pre-DE 'Anybody/Everybody wants to live well off.’8 b. John-un mwues(-ina/-itunci/-ilato/ilcilato} mek-nun-ta. J.-Top what-INA/-rnJNCI/-ILATO/-ILCILATO eat-Asp-DE 'John is eating anything/everything.’ (13) Universal Quantifiers (II):... WP ..N-(u)na/-kena/-tunci/-(e)to/-(u)lcilato. etc. [John-i mwues-ul mek( -una/-kena/-tunci/-eto/-ulcilato) ] J.-Nom what-Acc eat-UNA/-KENA/-TUNCI/-ETO/-ULCILATO na-eykey malha-ela. I-to tell-Imp 'Tell me whatever John eats.' C.-S. Suh (1985, 1990) analyzes existential quantifiers as indirect questions embedded under a relative clause. For example, he proposes the following structure for (11a), where Q-Pred stands for a predicate taking a question clause as a complement and Rel stands for a relative conjugation: 7 Universal quantifiers are apparently classified into two types: cases where WP is immediately followed by universal licensing markers as in (12) and cases where a long distance relation is involved as in (13). * The so-called universal quantifiers in fact have a free choice ANY reading. See J. Suh (1990:33). Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 12 (14) [n p [c p i ej [cp 2 ej nwukwuk-<j>-nka] Q-Pred-Rel] ej](-ka) o-ess-ta. who-be-QE -Nom come-Pst-DE C.-S. Suh (1990) gave several motivations for his interrogative analysis of existential quantifiers: existential licensing morphemes, i.e., -(i)nka/-(i)nci, contain a QE like ka/ci; the empty elements can be overtly realized; WPs used to have an interrogative reading but not a quantificational reading until recently; and various interpretations of 3-QPs can be reduced to the null Q-Pred that is compatible with various Q-predicates. Some supporting evidence for the indirect question analysis will be provided in chapter 5. First, it will be observed that the existential quantifier licensing morphemes function as scope barriers. Note that the scope of a WP in a language like Korean is determined by the position of the question ending that immediately c- commands the WP. Second, the copula analysis of -/ in the existential licensing morphemes -(i)nka/-(i)nci is further confirmed by the fact that the honorific morpheme si can attach to it. Notice that the honorific morpheme otherwise only attaches to a verb stem. Another fact that shows that existential quantifiers contain a clause projected from a copula verb is found in a Kyengsang dialect. According to C.-M. Suh (1987), existential licensing morphemes in a Kyengsang dialect are -nka - nko. Note that -ka/-ko are question endings for copula verbs in the dialect. (A. Kim 1982) This indicates that existential quantifiers have a clausal structure projected from a copula verb. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 13 Furthermore, universal quantifiers are also analyzed as reduced indirect questions. One structural difference between existential quantifiers and universal quantifiers is that the former are indirect questions embedded under a relative clause, while the latter are indirect questions embedded under an adverbial clause. For example, (12b), which is repeated below for ease of reference, is analyzed with (15) below: (12) b. John-un mwues-ina/tunci mek-nun-ta. J.-Top what-[NA/ITUNCI mek-Pres-DE ‘John eats anything.’ (15) John-un [cpi e * [cp2 ej mwuesk-i {-na/-tunci} ] Q-Pred-ADV] ei mek-nun-ta. J.-Top what-be-QE eat-Pres-DE ‘John, (irrespective of) what (it) is, eats (it).’ Several motivations for the question analysis of universal quantifiers are provided. First, universal quantifier licensing morphemes contain a question ending e.g., -(i)na/- (i)kena/-(i)tuncf) or a concessive ending, e.g., -(i)lato/-(i)lcilato, etc. It does not seem to be accidental that concessive clauses are often question clauses even in English, e.g. no matter wh-, wh-ever, etc. Second, question selecting predicates can be overtly realized. Third, honorific si can appear in the licensing morpheme. Fourth, the universal licensing morphemes function as a scope barrier. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 14 It will also be observed that existential quantifiers can bear case morphology, while universal quantifiers cannot. The contrast with respect to case morphology will be attributed to the difference in the internal structures of two types of quantifiers. Remember that existential quantifiers are analyzed as indirect question clauses embedded under a relative construction, while universal quantifiers as indirect questions embedded under an adverbial clause. (See the contrast between (14) vs. (15).) Existential quantifiers are complex noun phrases, while universal quantifiers are adverbial clauses, and a separate null NP exists as a matrix element.9 It will be argued that the NP vs. non-NP distinction is responsible for the difference in case bearability. As for the semantic difference between these two types of quantifiers, it will be observed that universal quantifiers, but not existential quantifiers, can be followed by a negative relational predicate like sangkwanepsi ‘regardless off kwankyeepsi ‘irrespective off etc. Since such negative relational predicates are identity functions, regardless of what is assigned to the WP, the value of the higher clause does not change. This difference will be argued to be responsible for the universal reading or free choice ANY reading that the so-called universal quantifiers have. In the same chapter, we will examine an alternative view which analyzes Korean WPs as indefinite quantifiers (Chang 1976 and S.-W. Kim 1989a, 1989b, 1991, among others). According to this view, Korean WPs are underlyingly indefinite 9 NP and DP will be used indistinctively throughout the dissertation, unless otherwise specified. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 15 quantifiers, and the interrogative reading that they may have is derived as a result of receiving an information focus (Chang 1976) or as a result of being governed by a question particle (Kim 1989a, 1989b, 1991). We will see that, despite a few attractive points, the indefinite quantifier analyses face several non-trivial conceptual and empirical problems. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 16 Chapter Two Representation of Functional Categories and their Licensing Modes 2.1. Introduction This chapter addresses the following two issues on functional categories: how functional categories like CP and NegP are represented in the sentence grammar and how they are licensed. As for the first question, it will be hypothesized basically along the lines pursued in the minimalist approach (Chomsky 1994, 1995) that functional categories are universally represented as a set of abstract formal features, that is, devoid of any phonological or semantic features. As for the second question it will be argued that UG allows a head-head relation as a licensing mode of the abstract features in the functional categories in addition to a SPEC-head relation, which has been considered as the unique way of licensing functional categories in the literature. See, for example, Rizzi (1991) and Haegman (1992), etc., who claim that a SPEC- head relation is responsible for WH-Criterion and Neg-Criterion, etc. The organization of the chapter runs as follows. In section 1, it will be discussed how functional categories are represented. Several facts across languages will lead us to conclude that they are represented as a set of abstract formal features, devoid of any phonological or semantic features, as illustrated below: Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 17 In section 2.2., the typology of functional category licensing will be discussed. It will be maintained that there are two modes of functional category licensing: by means of a SPEC-head relation and by means of a head-head relation. The former is achieved by placing an XP element with a relevant feature in the SPEC of F, while the latter by incorporating an X° element with a relevant feature. Since the licensing expression XP or X° could be attracted from a position within YP in (1) by means of a singulary operation (move and merge), or could be introduced by means of a binary operation (merge by a generalized transformation), each mode of the functional licensing has two subtypes. Therefore, we expect the following four (sub)types of functional category licensing: a SPEC-head relation by a binary operation, a SPEC- head relation by a singulary operation, a head-head relation by a binary operation, and a head-head relation by a singulary operation, as schematically represented below: Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. SPEC-head relation head-head relation 2.2. Representation of Functional Categories as a Set of Formal Features This section briefly reviews two opposing views on the representations of functional categories, the weak projectionist hypothesis and the strict projectionist hypothesis. I follow Benmamoun (1992) in adopting the latter hypothesis as it properly accounts for various syntactic phenomena that would be otherwise unexplained. I further discuss two possible approaches in implementing the latter hypothesis, the incorporation approach (Baker 1985) and the checking approach (Chomsky 1992, 1994), and conclude that the latter approach is more appropriate for the reason that it better accounts for various syntactic facts across languages. 2.2.1. Dual Properties of Morphologically Complex Words Let us first discuss how morphologically complex words are represented in syntax. Consider the following Korean sentence as an example: Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 19 (3) nwu-ka ecey amwukes-to ani-mek-ess-ni? who-Nom yesterday anything-even Neg-eat-Pst-QE ‘Who did not eat anything yesterday?’ The underlined predicate of the sentence, i.e., ani-mek-ess-ni ‘Neg-eat-Pst-QE,’ has a complex morphological structure, composed of negation morpheme ani ‘not,’ verb stem mek ‘eat,’ past tense morpheme ess, and interrogative mood morpheme ni. Such morphologically complex objects bear two apparently contradicting characteristics. On the one hand they behave like a word, but on the other hand individual morphemes internal to them are relevant at a certain level of representation. They behave like a word, at least at the surface level, which is shown by the fact that they cannot be cleaved into individual morphemes by any syntactic operation. For example, they cannot be separated by adding an adverbial: *ani-ADV- mek-ess-ni, *ani-mek-ADV-ess-ni, etc. Individual morphemes in complex words cannot be displaced: *ani,... trmek-ess-ni, *mek,... ani-tr ess-ni, etc. On the other hand, individual morphemes in the verbal complex play a role in syntax, as Benmamoun (1992) points out. For example, a negation morpheme is relevant to the licensing of negative polarity items (henceforth NPIs), Tense to the agreement relation with time adverbials, Mood to WH-phrase (henceforth WP) licensing, etc. The syntactic relevance of the individual morphemes in the complex predicate is clear from the fact that two predicates that minimally differ with respect Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 2 0 to each of the morphemes show some syntactic difference. For example, that a negation morpheme is relevant to NPI licensing is clear from the fact that predicates with no negation morpheme cannot license NPIs, and that Tense is relevant to the agreement with time adverbials is clear from the fact that predicates with a present tense morpheme cannot cooccur with a past time adverbial, etc. Our obvious task is to simultaneously capture the dual properties of such morphologically complex objects: the single-word-like integrity at a certain level and the grammatical relevance or visibility of individual morphemes at another level. In the following subsections, two views on the treatment of such morphological complex objects that have been discussed in the literature will be reviewed. 2.2.I.I. The Weak Projectionist Hypothesis Let us first discuss the hypothesis which views every word as a lexical item, no matter how morphologically complex it is. Call this hypothesis 'the weak projectionist hypothesis,’ borrowing the terminology from Benmamoun (1991). According to this hypothesis, the verbal complex ani-mek-ess-ni ‘Neg-eat-Pst-QE’ in (2) will be introduced into syntax directly from the lexicon. It will be placed under a terminal node of a phrase structure (or projects as a head according to the current theory) just like any other lexical (i.e., X°) elements. Adopting the weak projectionist hypothesis, (3) will have a schematic representation like (4) below: Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Compl X where X=Neg-V-T-M The analysis captures the native speakers’ intuition on the lexical (one-word-like) properties of morphological complexes like ani-mek-ess-ta, as mentioned before. It does not, however, capture the fact that individual morphemes in the verbal complex play a role in syntax. At this point it is worth discussing Kitagawa’s (1986, chapters 2 and 3) lexicalist analysis of Japanese predicates, which apparently does not suffer from the problem mentioned in the preceding discussion. According to Kitagawa, Japanese complex predicates (e.g., V-sase-ru ‘V-CAUSE-Pres’ and V-ta-i ‘V-want-Pres’) and simplex predicates (e.g., V-rw ‘V-Pres’) are derived in the lexicon. He provides convincing arguments based on phonological facts (accentuation, voicing spread, and downdrift). In his system, however, the lexically derived words are cleaved into pieces through affix raising, a version of excorporation, at LF. To take a simple example, NP-ga NP- o tabe-ru ‘NP-Nom NP-Acc eat-Pres’ has the following derivation: Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. The raising is driven by complementation requirement in Kitagawa’s system. For example, ru in (5a), a sub-element of Im m , is raised to take a Vm a x as a complement. Since the individual morphemes are visible at LF, Kitagawa’s excorporation analysis does not seem to suffer from the problem that the lexicalist approach with no such LF movement faces. Kitagawa’s analysis, however, will also fail if any individual morpheme in the complex predicates ever plays a role before the derivation reaches LF. In fact individual morphemes of morphologically complex objects in Korean, which are very similar to Japanese in their morphological structure, are relevant before spell-out. For example, the negation morpheme ani, which is an affix to predicates, licenses NPIs before spell-out (Chung 1993a, 1995a, Sohn 1994, 1995). Thus, the negation morpheme must somehow be represented before spell-out, i.e., before the derivation reaches LF. This indicates that the exclusive LF raising of affixes in Kitagawa’s Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. system does not seem to work, although his observation that morphologically complex words in Japanese are pre-syntactically generated seems correct.1 2.2.I.2. Two Strict Projectionist Views: Incorporation vs. Checking A second approach to the morphologically complex objects is to view each morpheme in the complex objects to be mapped to a syntactic projection. Let us call this view ‘the strict projectionist hypothesis,’ again borrowing Benmamoun's (1991) terminology. Each morpheme being projected and being visible in syntax, the strict projectionist hypothesis does not suffer from the problem that the weak projectionist hypothesis faces. There are two views in the literature as to how to map functional categories into the phrase structure, the incorporation approach (Baker 19882 ) and the checking approach (Chomsky 1992, 1994, 1995). Let us briefly see how the two approaches represent the morphologically complex objects. According to the incorporation approach, each morpheme in the morphologically complex objects independently exists in the lexicon (just like other lexical elements) and heads its own projection in syntax. They are taken out of the lexicon (or out of 1 Another problem with Kitagawa’s analysis is that affixes raise and project, which is not allowed in the perspective of the current theory. 2 In fairness, it should be mentioned, incorporation approach had been entertained before Baker (1988). For example, Kuroda (1965) analyzes japanese causative complex predicates as being formed by raising content verbs to causative verbs. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. the numeration depending on the theory) and project their own projections in syntax. For example, functional elements like Negation ani-. Tense -ess-, and Mood -ni are chosen from the lexicon and project NegP, TenseP, and MP, respectively, just as the verb stem mek is drawn from the lexicon and projects VP. Thus, (3) will have the following structure immediately after lexical insertion:3 (6) Neg -ess SPEC VP ani- nwu-ka AD VP V’ ecey NP V I I amwukesto mek Thus, the verbal complex ani-mek-ess-ni is a syntactically derived object. It is formed as follows. Verb stem mek first raises to the immediately higher head ani, forming an amalgamation ani-mek. Next, the V-Neg amalgamation, ani-mek, raises further to Tense -ess-, forming ani-mek-ess-. Finally the amalgamation ani-mek-ess raises to 3 To have the right result, NEG ani- should be marked as a prefix, while Tense -ess and Mood -ni as suffixes. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 25 Mood -ni, forming ani-mek-ess-ni. In brief, the verbal complex is formed by means of successive syntactic incorporation.4 According to the checking hypothesis, the whole verbal complex is formed by some generative morphological process in the lexicon or in the course of being introduced into a numeration set, as Chomsky (1994, 1995) entertains. Thus, such a verbal complex is directly drawn from the lexicon or from the numeration set and projects as a head, while the related functional heads are to be checked, say. by raising the verbal complex to relevant heads. With this approach, sentence (3) will have the structure in (7) below at a certain point of derivation before the verbal complex is raised to check each of the functional heads: (7) MP TP M NegP T VP Neg SPEC VP nwu-ka AD VP V’ ecey NP V I I amwukesto ani-mek-ess-ni 4 It is logically possible that the morphologically complex words are derived by (successive) affix lowering as well as successive verb raising. This possibility, however, will not be discussed for the reason that at least the computational system in human language does not allow lowering. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 2 6 The verbal complex, which has already been formed in the lexicon or in the course of being selected as an element of a numeration set, is directly introduced into syntax and is placed under V (or projects as a head), the lowest head among relevant heads. It undergoes successive head-to-head movement, checking each of the functional categories in higher positions. Apparently the two approaches look like notational variants, not making any significant difference. Although no strong evidence is found for one approach over the other, the checking approach will be taken over the incorporation approach, basically following Kitagawa’s (1986) observation on complex predicates in Japanese, Chomsky’s (1992) treatment of French-English differences with respect to verb raising, and Ladusaw’s (1992 and colloquium at UCLA 1994) treatment of n- words in Romance languages and some dialects of English.s As mentioned in section 2.2.1.1., Kitagawa (1986) argues that Japanese complex predicates (e.g., V-sase-ru ‘V-CAUSE-Pres’ and V-ta-i ‘V-want-Pres’) and simplex predicates (e.g., V-ru ‘V-Pres’) are derived in the lexicon, based on various phonological facts such as accentuation, voicing spread, and downdrift. (See chapters 2 and 3 of his thesis.) For example, the addition of an inflectional or derivational suffix to a verb stem affects the accentuation pattern of the verb. Under the assumption that accentuation is pre-syntactically determined in Japanese, which 5 The gist o f their arguments will be given here. See the references for details. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 27 seems very plausible, the verbal complex must be formed pre-syntactically. This indicates that the complex predicates in Japanese are introduced pre-syntactically.6 Secondly. Chomsky (1992:38-89) opts for the checking approach over the incorporation approach because the former is able to avoid lowering movement in the treatment of French-English difference with respect to verb raising, (cf. Emonds 1976. Pollock 1989, etc.) Thirdly, Ladusaw (1992 and colloquium at UCLA 1994) argues for an abstract negative morpheme which must be licensed by an n-word to account for the negative concord phenomena in Romance languages and some dialects of English. N-words in some Romance languages may function as a negative quantifier or as an NPI depending on their structural position. Only the n-words that appear higher than INFL are interpreted as negative quantifiers. According to him, an n-word turns into a negative quantifier when it goes through SPEC of NegP whose head is abstractly represented. Two additional arguments for the checking approach are provided in what follows. The first argument is in relation to the import of the sigma phrase discussed in Laka (1990). According to Laka (1990), the so-called NegP and Affirmation) phrase represent two sides of a coin, this idea originating from Chomsky (1957). For example, both require cfo-support in English. Such a fact can be explained with the 6 As mentioned earlier, Kitagawa (1986) adopts an affix raising analysis in which the base structure is like the one adopted in the weak projectionist approach, but application of affix raising at LF produces a structure like the base structure adopted in the incorporation approach. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 2 8 checking approach adopted here in the following way. Suppose that a head (call it a Sigma following Laka) which is neutral with respect to Neg or Aff is drawn from the lexicon and projects. The projected sigma phrase turns into a NegP if the head (=Sigma) is licensed by a negative expression and into an AffP if licensed by an operator related to Affirmation. But with the incorporation approach, the relation between two phrases could not be captured. Note that if a negation morpheme were directly inserted under the head of NegP and an (empty) affirmative morpheme were under the head of AffP, the mutual exclusiveness between the two projections would not be structurally captured.7 Additional evidence can be found in the morphological and syntactic behaviors of suppletion form negation in Korean. In general mono- or bi-syllabic predicates in Korean may have a short form negation as well as a long form negation (H. Ahn 1990). However, verbs like iss ‘to exist’ and al ‘to know,’ though mono-syllabic, do not have short form negation: *ani-isss and *ani a! (Chung and Park 1995b). Instead, they have suppletion forms: eps ‘not to exist’ and molu ‘not to know.' To capture the idiosyncrasies of such suppletion forms, they are best analyzed as being generated pre-syntactically. It would be hardly possible to derive such suppletion forms by way of a syntactic operation, e.g. an operation which turns Neg plus iss into 7 One might think that the mutual exclusiveness could be captured in terms of the semantic properties of Neg and Aff. But a purely semantic approach seems to fail. Note that semantically negative verb may appear in the context of AfF, e.g., ...DID deny... 1 When the verb means *to stay,’ it can have the short form negation. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 29 eps and Neg plus al into molu, etc. Such suppletion forms, however, can license clause mate NPIs just like the other two forms of negation, short forms or long forms: (8) a. amwuto chavk-ul ani-sa-ess-ta. anyone book-Acc Neg-buy-Pst-DE ‘No one bought a book.’ b. amwuto chayk-ul sa-ci ani-ha-ess-ta. anyone book-Acc buy-Nmz Neg-do-Pst-DE ‘No one bought a book.’ c. amwuto i chayk-uy kakyek-ul molu-n-ta. anyone this book-Gen price-Acc not;know-Pres-DE ‘No one knows the price of the book.’ Under the assumption that NPIs are licensed by the checking relation with Neg in Korean (Chung 1993a, 1995a, Sohn 1994, 1995), Neg should be posited for the sentences including the suppletion forms, as illustrated below: (9) NPIj Neg’ VP Neg ...tj ...tj V Neg I I molu-n-ta, [+neg] NPI checking Neg Licensing Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 30 The inherently negative predicate with a suppletion form is adjoined to Neg to license the negative force, and then the negative predicate adjoined to Neg licenses the NPI that has moved to the SPEC of NegP. 2.2.2. Functional Categories as a Set of Formal Features In the previous section we have seen several indications that the checking approach is superior to the incorporation approach. In other words, functional categories do not have any phonetic features. A natural question that arises at this point is what resides in the head of functional categories. If there is nothing, how is it that functional categories are projected at all? If there is something, what is the content of them? Regardless of how abstract a functional category is, however, it must have minimum content in the lexicon, otherwise no projection can be expected. In fact, we have already recognized that functional categories have much to do with syntactic operations, e.g., Neg with NPI licensing in Korean. Therefore, they must have at least (a set of) formal features that are responsible for such syntactic operations. How about semantic features? Chomsky (1995:291) claims that some functional categories like C (the mood force) have semantic features. Functional categories themselves, however, need not have any semantic features under the assumption that functional categories are to be licensed by some element with a relevant feature. For Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 31 example, Neg is licensed by a negative expression which varies over an XP expression or an X° expression, and an interrogative C is licensed by an interrogative element like a WP as in English or a predicate with an interrogative conjugation as in Korean. If these interrogative elements carry the WH-feature, functional categories themselves need not contain those semantic features due to the economy of representation. In conclusion, functional categories are purely composed of a set of formal features, devoid of any phonological and semantic features, as schematically represented in (1), repeated below:9 ( 1) FP YP +ff +f2 2.3. Two Modes of Functional Category Licensing This section discusses how the abstract feature in the head of functional categories is licensed. It will be claimed that there are two ways to license the formal 9 In what follows, only the features relevant to discussions will be represented under functional categories unless otherwise required. For example, although T in English may include [strong D], [strong/weak V] as well as [+/-past], mostly [+/-past] will be put under T. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. features like [+neg] and [+wh], by means of a SPEC-head relation and by a head-head relation. The choice of the licensing modes depends on the projectional status of licensing expressions, e.g., negative expressions and interrogative expressions, which varies between non-heads (XP) vs. heads (X°). The former licensing mode. i.e.. licensing by a SPEC-head relation, is fulfilled when an XP negative/interrogative expression is placed in the SPEC of NegP/CP, whereas the latter, i.e., licensing by a head-head relation, is fulfilled when an X° negative/interrogative expression adjoins to the head of NegP/CP. as schematically represented below: (10) a. FP XP, F’ YP [+f\ ...(tj)... SPEC-head relation Yi F ...(ti)- head-head relation Licensing expressions, i.e., XP; in (10a) and Yj in (IOb), may or may not originate from a position within YP, as indicated with the traces in the parentheses. In other words, the licensing expressions XP or Y may be placed in their respective position either by means of a singulary operation (movement) or by means of a binary operation (simple merge). Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Natural languages seem to evince all the four submodes of the functional category licensing. Licensing of a functional category F via a SPEC-head relation has two subtypes. It can be achieved by moving an XP originated from a position within the complement of F to its SPEC, or by merging F (or its projection) and an XP. which has never been within the complement of F, by means of a binary operation. Licensing by movement is evidenced in the English WH-construction. Consider the following example: (11) Who; do you think John loves tj? WP who in (11) is moved from the complement position of the embedded verb loves to the SPEC of the matrix CP, whose interrogative mood requires a WH-expression in its checking domain. Ignoring irrelevant parts, (11) will have the following structure: (12) CP SPEC C’ whoi C IP [+wh] ...CP ...tj Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 34 The abstract [+wh] feature under C is licensed by having a SPEC-head relation with who, which has moved to the SPEC of CP. SPEC-head relation by a binary operation is evidenced arguably in English embedded questions introduced by whether (Chomsky 1995): (13) I do not know whether John left. To support our position, two things must be shown: whether in (13) is not in a head position but in a SPEC position; and it is not from within IP. That whether is in a SPEC position is argued in Kayne (1991:665-666). Unlike if, whether may introduce an infinitival clause with the PRO subject: (14) a. ...whether PRO to go b. * ... //PRO to go He attributes the contrast to a version of the PRO theorem,1 0 under the assumption that if as a lexical complementizer governs across IP, whereas no lexical complementizer is present in whether PRO to go. Now let us turn to the question whether whether is base generated in the SPEC of CP or whether it is extracted from within IP. Although Larson (1985) argues that 1 0 Kayne’s formulation of the PRO theorem is not identical to Chomsky’s (1981). although nothing hinges on that here. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. whether has moved from inside IP, to account for the scope ambiguity of disjunction. Chomsky (1994) takes whether as being placed in the SPEC of CP by a binary operation. It seems to be hard to find any strong argument for one over the other position, but Chomsky’s non-movement approach gains support from the fact that whether never remains in situ unlike other WPs in English:1 1 (15) * I do not know John left whether. One might attribute the ungrammaticality of (15) to the selectional restriction: a complement clause of the verb know must be one that starts with a WP in overt syntax. But consider sentences like (16) below: (16) *1 do not know if John left whether. The ungrammaticality of the sentence cannot be attributed to the selectional restriction mentioned above. Note that (16) would be fine without whether. Assuming the non-movement approach to whether, the [+wh] Q in (13) is licensed b y - having a SPEC-head relation with a WP that is introduced by a binary operation as schematically represented below: " Although non-nominal WPs in English are not allowed in situ generally (see chapter 3), they may remain in situ at least in all-over WP cases. (See Chomsky 1981, Kayne 1984.) In contrast, whether is never allowed in situ. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. SPEC C’ whether C IP [+wh] Licensing of the F-feature of a functional category by a head-head relation is also two fold: by a singulary operation or by a binary operation. An F-feature can be licensed by attracting an X° with a matching feature. For example, a [+wh] C° is licensed by attracting a predicate with a question conjugation in Korean. Consider the following WH-question: (18)John-i mwues-ul mek-ess-ni? J.-Nom what-Acc eat-Pst-QE ‘What did John eat?’ At a certain point of derivation, (18) will have the following structure: Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. VP T [+wh] NP V’ [+past] John-i NP V mwues-ul mek-ess-ni After the complex verb raises to M through T, the question sentence will have the following structure: (20) NP V’ John-i NP T Tj M tj Vi T [+wh] I I mek-ess-ni [+past] mwues-ul t, Assuming along with Chomsky (1992) that all the adjoined positions to a head are in the checking domain of the head, Vj in (20) is in the checking domain of M (and of T Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 38 as well). Thus the interrogative force represented by [+wh] is licensed by the predicate with a question conjugation mek-ess-ni. An F-feature can be licensed by merging an X° to F. As discussed in Chomsky (1995:289-294). an English embedded question that starts with //is an example. The Chinese WH-question particle ne or the yes/no question particle ma can be said to be merged similarly:1 2 (21) a. I asked if John left. b. I asked if on that day John left. c. I asked if quickly John left. (22) a .... NP ne? Shei kanjian-le Zhangsan ne? who see-Asp Zhangsan Qw h 'Who saw Z.S.?’ b. ...V ne? Shei qu ne? who go Q wh ‘Who will go?’ c .... PP ne? Shei song-le yi-ben shu gei Zhangsan ne? who send-Asp one-Cl book to Z.S. Qw h ‘Who sent a book to Z.S.?’ 1 2 I thank Miao-Ling Hsieh for the Chinese examples in (22) and (23). Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 39 (23) a .... NP mat Ni kanjian-le Zhangsan ma? you see-Asp Zhangsan Qy/n 'Did you see Z.S.?’ b .... V ma? Ni qu ma? you go Qy/n ‘Will you go?’ c .... PP mal Ni song-le yi-ben shu gei Zhangsan ma? you send-Asp one-Cl book to Zhangsan Qy/n ‘Did you send a book to Z.S.?’ Clearly, //in English and ne and ma in Chinese are not a conjugational element on a particular category since adjacent elements may vary in syntactic categories. There is no independent evidence that these elements have moved from a position within the embedded clause, either. Thus I assume that they are taken out of the numeration and adjoined to [+Q] C, as schematically represented:1 3 ,1 4 1 3 By [+Q] I mean the interrogative force without distinguishing [+wh] from [-wh], 1 4 The linear order between [+Q] and question expressions is irrelevant here. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. The question force in [+Q] C is licensed by having a head-head relation with i f in (24) and with ne/ma in (25). In summation, there are two modes of licensing an F-feature of a functional category. It is licensed either by having a SPEC-head relation with an XP in the SPEC or by having a head-head relation with an X° that is adjoined to the head. Each mode has two subtypes depending on how the licensing expressions are introduced to the checking domain of F, either by a singulary operation or by a binary operation. A final remark is that both modes of Q-licensing, i.e., the licensing in terms of a SPEC-head relation and the licensing in terms of a head-head relation, obey the economy principle. Q-licensing by a binary operation (simple merge) vacuously satisfies the economy principle. Q-licensing by attract-a also satisfies the principle, a being the highest WP or the highest predicate with a question conjugation. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Chapter Three Scope Licensing Modes of WH-Phrases (WPs) in situ 3.1. Introduction In the previous chapter, it was argued that there are two licensing modes of functional categories, i.e., either by a SPEC-head relation or a head-head relation. For example, a [+wh] Q is licensed either in terms of a SPEC-head relation by placing a WH-phrase (henceforth, WP) in the SPEC of CP, or in terms of a head-head relation by incorporating a [+wh] complementizer or a predicate with a question conjugation. This chapter claims the following two. First, when polarity items are triggered by a head H, every polarity item in the domain of H must be licensed by H independent of the licensing of H. For example, every WP in multiple WP questions should be scopally licensed by having a certain relation with a [+wh] Q. Otherwise, they cannot be interpreted, violating the full interpretation principle. The second claim to be made in this chapter is that there are two scope licensing modes of WPs in situ, depending on the categorial status of WPs; nominal WPs are licensed via having a binding relation with a [+wh] Q, whereas non-nominal WPs are licensed via moving to the SPEC of a [+wh] Q.1 The binding vs. movement dichotomy in the scope licensing modes will be attributed to the categorial status of question clauses as well as that of WPs themselves. 1 The difference among WPs has been characterized in various ways. See footnote 3 of chapter I. Reproduced with permission of the copyright owner. Further reproduction prohibited w ithout permission. 42 In particular, the question clauses are argued to contain a nominal feature. Under the assumption that binding relation is restricted between two nominal elements, nominal WPs entertain the binding option, but non-nominal WPs do not Therefore, non- nominal WPs must move to the SPEC of a [+wh] C° to be licensed, whereas nominal WPs can be licensed in situ by a binding relation with a [+wh] C°, which is more economical than the movement option. In section 3.2. it will be argued that the licensing of a functional category discussed in the previous chapter and the licensing of polarity items dependent on a functional category are two independent operations in grammar. Let us call the first operation head licensing and the second dependent licensing. Head licensing is required in order to license a formal feature that a functional head bears. Formal features that a head bears are licensed by having a checking relation with an element with a matching feature, as discussed in chapter 2. Dependent licensing requires that every polarity item (e.g., WPs and NPIs) be licensed by having a proper relation with its triggering head. Especially when more than one polarity item is dependent on a head, each of the items must have a proper relation with the relevant head. Section 3.3. examines the distribution of WPs in situ. The following three facts will be highlighted for later discussion. First, non-nominal WPs are bounded to their scope positions, while nominal WPs are not. (See Aoun 1986, Huang 1982, and Lasnik & Saito 1984, 1992, among others.) Thus, a syntactic island is a scope barrier for non- nominal WPs, but not for nominal WPs. Second, non-nominal WPs in the English type Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 43 languages cannot remain in situ, whereas those in the Korean type languages can (in fact must) remain in situ as long as there is no barrier between them and their scope position. Third, an interrogative scope can be shared by two or more nominal WPs but not by two or more non-nominal WPs (Aoun 1986).2 The second and third facts can be restated as follows: non-nominal WPs compete with other WPs for the scope position, while nominal WPs do not Note that the SPEC of CP is already filled with a WP in the English type languages, but not in the Korean type languages. In section 3.4., some previous approaches that try to account for the distributional facts in terms of LF movement of WPs in situ will be reviewed. It will be pointed out that the ECP approach combined with a Comp indexing mechanism has some conceptual flaw. It will also be pointed out that Nishigauchi’s subjacency theory based on the large scale pied-piped movement does not correctly capture the facts described in section 3.3. In section 3.5. an alternative theory of multiple WP licensing will be proposed to account for the nominal/non-nominal asymmetries discussed in section 3.3. It will be claimed that nominal WPs in situ can (and must for the economy reason) be licensed in situ, without undergoing movement even at LF, while non-nominal WPs cannot be licensed in this way and therefore must move to its scope position, SPEC of a [+wh] C°. In section 3.6., we will try to derive the nominal/non-nominal asymmetries from the categorial status of WPs and question clauses. It will be argued that question 2 See chapter 4 for exceptional cases. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 44 clauses are categorially nominal, which facilitates the binding option for nominal WPs. Under the assumption that licensing of WPs by means of binding is more economical than licensing by means of movement followed by a checking relation (cf. Tsai 1994), a binding option is exploited, if ever possible, as in the case of nominal WPs. Just in case a binding option is unavailable, the movement option is called for. as in the case of non-nominal WPs. Section 3.7 summarizes the chapter and discusses some theoretical consequences of the proposed theory. 3.2. Examples of Dependent Licensing Consider the following structure, where y is a functional category (or its formal feature) and ai, a 2, ..., O n are polarity items formally dependent on y: One way to license the formal feature that y bears is to raise one of the elements of the type a to the checking domain of y. For example, a [+wh] C° can be licensed by attracting a WP, as in English. Suppose that the formal feature of y in (1) is licensed by (1) 7 cti, c c 2 , a 3 , ..., a,, Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 45 attracting at. A question that arises at this point is whether other elements of the same type, i.e., a 2, a 3, ..., a„ in (1), need to be subject to any kind of licensing condition, e.g.. whether WPs in situ in multiple WH-constructions need to be licensed at all. This section argues that the answer to this question should be affirmative, utilizing several examples which point in this direction. 3.2.1. Multiple WH-Fronting in East European Languages Some East European languages like Bulgarian, Romanian, and Polish put every WP at the sentence initial position, as discussed in Rudin (1985, 1988), Wachowicz (1974), and Cheng (1991), among others. All of the WPs must be fronted in multiple WH-questions in such languages. Consider the following structures in (2) and their corresponding Bulgarian examples in (3), which are adapted from Rudin (1988): (2) a. WP;... tj... b. WPj WPj... tj... tj... c. *WPj... tj... WPj... (3) a. Koj [e vidjal Mari] who saw-3 s Mary ‘Who saw Mary?’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 46 b. Koj kogo [e vidjal e] who whom saw-3 s 'Who saw who?’ c.*Koj [e vidjal kogo] who saw-3 s whom ‘Who saw who?’ If a sentences contains only one WP, then the WP moves to the initial position of the sentence as in (3a). If there are two or more WPs, all of the WPs must be fronted as the contrast between (3b) and (3c) shows. The multiple WH-fronting in the languages mentioned above can be taken as a fact suggesting that not every instance of movement is to satisfy the morphological requirement of a head. For the sake of discussion, let us suppose that WH-fronting in these languages is to license a formal feature of a functional category, e.g., the formal feature corresponding to the interrogative force of a question sentence.3 If there is only one WP in the sentence as in (3a), then the WP moves to the initial position. This movement would license the relevant formal feature. Thus the additional movement of the second WP in (2b) is irrelevant or superfluous to the licensing of the formal feature in question. Then the very existence of the multiple WH-fronting of this kind indicates that not every instance of movement is to satisfy a morphological requirement of a head. 3 Cheng (1991, chapter 3) argues that WH-fronting in these languages does not have anything to do with question licensing, though. She claims that WPs in those languages have a null determiner and she tries to derive the multiple fronting nature of WPs from the licensing of the null determiner. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 47 Regardless of what is responsible for the multiple WH-fronting in these languages, it is clear from the above discussion that not every instance of WH-fronting in those multiple WH-fronting languages is an "altruistic" operation to license the formal features of a head. At least some instance of WH-fronting is a “self-service” operation. 3.2.2. NPI Licensing in Korean NPI licensing in Korean is another example of dependent licensing that is required independent of head licensing. In other words, every NPI should be licensed by having a proper syntactic relation with its trigger. Korean NPIs are licensed by having a SPEC-head relation with its trigger, usually negation, in overt syntax or before spell-out, as argued in Chung (1993a, 1995a) and Sohn (1994, 1995). Thus Korean NPIs should be located in the SPEC of NegP at some point of derivation before spell-out, as schematically represented below: (4) NegP NPI Neg’ SP VP Neg I 'C-head relation Consider the paradigm in (5) and examples in (6): Reproduced with permission of the copyright owner. Further reproduction prohibited w ithout permission. I 48 (5) a. .... [cp...NPI... Neg...] .... b. * ... NPI .... [cp— Neg...]... c. ... NPIj... [cp — tj... Neg...]... d. ... NPIj... [ c p ...tj... ]... Neg... e. *... [... NPI... ]... Neg... (6) a. John-i amwukesto ani-mek-ess-ta. J.-Nom anything Neg-eat-Pst-DE ‘John did not eat anything.’ b. *amwuto [c p John-i sakwa-Iul ani-mek-ess-ta]-ko sayngkakha-n-ta. anyone J.-Nom appIe-Acc Neg-eat-Pst-DE-C think-Pres-DE ‘♦Anyone thought that John did not eat an apple.’ c. amwukestoj Mary-ka [C p John-i tj ani-mek-ess-ta]-ko sayngkakha-n-ta. anything; M.-Nom J.-Nom Neg-eat-Pst-DE-C think-Pres-DE ‘Anything*, Mary thinks that John did not eat tj.’ d. amwukestoj Mary-ka [cp John-i tj mek-ess-ta]-ko sayngkakha-ci anythingj M.-Nom J.-Nom eat-Pst-DE-C think-CI ani-ha-n-ta. Neg-do-Pres-DE ‘Anything, Mary does not think that John ate.’ e. * Mary-ka [c p John-i amwukesto mek-ess-ta]-ko sayngkakha-ci M.-Nom J.-Nom anything eat-Pst-DE-C think-CI ani-ha-n-ta. Neg-do-Pres-DE ‘Mary does not think that John ate anything.’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 49 A generalization that emerges from the paradigm is that an NPI is allowed if it has ever had a clause mate relation with a negation morpheme at some point of derivation before spell-out. Let us conceive the clause mate condition between an NPI and its triggering negation morpheme as a realization of the feature checking between them.4 Sohn (1995, chapter 2) raises an important question with respect to Chomsky's (1994, 1995) position that strong features can only be located in the head of the target of movement. One problem with Chomsky’s position is that NPIs should be related to their trigger, e.g., negation, but triggers do not require an NPI in their domain. So a paradoxical situation arises: negation in a sentence with an NPI must have a strong feature to host the NPI, whereas negation in a sentence without an NPI must not. One possible way out for Chomsky, as Sohn (1995:83) mentions, would be to stipulate that Korean has two types of negation, one with a strong feature which should be checked off by an NPI and the other with no such strong feature. Sohn points out two problems with the revision of Chomsky’s position. First, there is an overgeneralization problem. When there is no strong feature on the head of 4 NPIs in Korean cannot be embedded under NP/PP while their triggering negation is outside the NP/PP: (i) a. * ... [ NP/PP— N P I...]... N eg... b. * John-i [amwuto-uy chayk]-ul ani-sa-ess-ta. J.-Nom anyone-Gen book-Acc Neg-buy-Pst-DE ‘John did not buy anyone’s book.’ This indirectly supports the requirement of the SPEC-head relation between NPIs and negation in overt syntax. See Chung (1995a) for more detail. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 50 NegP. the sentence with the negation disregards whether there is an NPI in the sentence or where it is located. Then the illicit structures in (5) could not be excluded, as far as no other principle is violated. A second problem arises when NPIs are multiply involved. Consider the structure in (7) and an example in (8) below: (7) [...NPI,... [cp...NPI2...]...Neg...] (8) (modified from Sohn 1995:85, his (117a)) *?amwuto [John-i ecey amwutevto ka-ess-ta]-ko sayngkakha-ci anyone J.-Nom yesterday anywhere go-Pst-DE-C think-CI ani-ha-n-ta. Neg-do-Pres-DE ‘No one believes that John went anywhere yesterday.’ If a strong feature resides in the head of the target, Neg, and the checking of the feature is sufficient to license constructs involving NPIs, then (7) should be predicted to be licit since NPI, would suffice to license the strong feature in Neg. But as the sentence in (8) shows, the structure in (7) is disallowed. The sentence in (8) becomes acceptable when the NPI in the embedded clause, i.e.. amwuteyto ‘anywhere,’ is replaced by a non-NPI, e.g., by hakkyo-ey ‘school-to.’ The ungrammaticality of the sentences like (8) is thus caused by the existence of the NPI which has not had a proper relation with negation before spell-out. This in turn Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 51 indicates that every NPI should be licensed by having a checking relation with its trigger. 3JJ. Multiple WPs in situ WP licensing in multiple WH-questions is another instance of dependent licensing requirement, although no strong feature is linked to the WPs in situ unlike the two cases discussed in the previous two subsections. Consider the following structure in (9) and its English example in (10), which was once explained in terms of the ECP (Huang 1982, Chomsky 1981, etc.): (9) * [cp^W Pi \jp... tj... fcte,... a d v WP...]]]? (10) * Whatj did you give tj to the man that John introduced to Mary why? The ECP effect cannot be fully subsumed under the head licensing. If satisfaction of some morphological requirement on the part of the [+wh] C° were sufficient to license WH-constructions, sentences like (10) should be fine because the [+wh] C° attracts a WP, what, to the SPEC of CP. Thus it is obvious that the ill-formedness of the English sentence is caused by the non-nominal WP, why, that appears in the island.5 5 In feet, embedding o f the WP why under an island is not a crucial factor in English since non-nominal WPs are not allowed in situ anyway, putting aside the all-over-WH-cases discussed in Chomsky (1981) and Kayne (1984), among others. Why was placed there to have a parallel structure in Korean for immediate discussion purposes. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 52 Korean adverbial WP way ‘why’ behaves similarly. It cannot appear within an island when it is scopally related to a position outside the island. Consider the structure in (11), where Vqe stands for a question conjugated predicate, and its Korean example in (12): (ID* [cp [+wh] [ip --[island - “ ‘" W P ...] ... V q e ]] (12) * ne-nun [John-i wav ssu-n] chayk-ul ilk-ni? you-Top J.-Nom why write-Adn book-Acc read-QE 'Why, do you read the book that John wrote tj?' It was postulated in the previous chapter that the raising of the predicate with a question conjugation to the [+wh] C° licenses the interrogative force of the sentence in a language like Korean, satisfying the morphological requirement of [+wh] C°. Thus, no problem arises with respect to the licensing of the abstract feature [+wh] in C°. The ill- formedness of the sentence clearly comes from the existence of a non-nominal WP way in an island, which cannot be properly related to its scope position outside the island. The ill-formedness of English sentences like (10) and of Korean sentences like (12) indicates that for a construct including WPs in situ to be licit, it is not enough to license the interrogative force of the sentence which can be accomplished by raising a WP to the SPEC of CP as in English or raising a question conjugated predicate to [+wh] Reproduced with permission of the copyright owner. Further reproduction prohibited w ithout permission. 53 C° as in Korean. Every WP in situ must be licensed by having some relation with its relevant [+wh] C°. 33. On the Distribution of WPs in situ: Nominal/Non-nominal Asymmetries Distribution of WPs in situ differs inter-categorially as well as across languages. In this section, the following three facts will be highlighted for later discussion. First non-nominal WPs are bounded to its scope position, while nominal WPs are not. (See Aoun 1986, Huang 1982, and Lasnik & Saito 1984, 1992, among others.) Thus, a syntactic island is a scope barrier for non-nominal WPs but not for nominal WPs. Second, non-nominal WPs in the English type languages cannot remain in situ, whereas those in the Korean type languages can (in fact must) remain in situ as far as there is no barrier between them and their scope position. Third, an interrogative scope can be shared by two or more nominal WPs but not by two or more non-nominal WPs (Aoun 1986). The second and third facts can be subsumed under the following generalization: non-nominal WPs compete with other WPs for the scope position, while nominal WPs do not. Note that the SPEC of CP is already filled with a WP in the English type languages, but not in the Korean type languages. Thus no non-nominal WPs are allowed in situ in the English type languages, while adverbial WPs are allowed in situ in the Korean type languages. Reproduced with permission of the copyright owner. Further reproduction prohibited w ithout permission. 3.3.1. WPs in situ in English Type Languages In a language like English, when two or more WPs share their question scope, only one WP moves to the SPEC of CP to license the [+wh] C° and the other WPs remain in situ. Not every type of WPs, however, can remain in situ. Except for the all-over-wh cases (see Chomsky 1981 and Kayne 1984, among others), non-nominal WPs cannot remain in situ. They are not allowed in situ regardless of the intervention of a syntactic island between them and their scope position and irrespective of the categorial status of the overtly moved WP, as schematically represented in (13) and exemplified in (14): (13) *WP| ... C[island )-• * * V WP ... (14) a. * Why do you believe that John ate apples why? b. * Who do you want to believe that John ate apples why?6 6 See Lasnik & Saito (1992) for the contrast «[n o n ,WP ... * dvWP ...j vs. [* dvWP ... ""W P ...] in English: (i) (=Lasnik & Saito 1992:40, their (1)) a. ‘what did you buy why b. why did you buy what Some native speakers of English take sentences like (ia) as acceptable. But 1 believe that if they are acceptable, the sentences are interpreted as if why is in the SPEC of a sluiced question. For example, sentences like (iia) are read as (iib): (ii) a. Who left for NY why? b. Who left for NY and why did he leave? Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 55 c. *Whv do you believe that John rejected [the job offer he received why]? d. * Who do you want to believe that John rejected [the job offer he received why]? No island intervenes between why in situ and its scope position in (14a, b), while there is one in (14c, d), as illustrated with the angled brackets. A non-nominal WP has overtly moved to the SPEC of the matrix CP in (14a, c), while a nominal WP has in (14b, d). None of the sentences are acceptable. Unlike adverbial WPs, nominal WPs in English are able to remain in situ. They can remain whether or not a barrier intervenes between them and their scope, as was discussed in Aoun et al (1981), Huang (1982), Lasnik & Saito (1984), and Kayne (1984), among others. The categorial status of the overtly moved WP does not matter either, as was observed by Lasnik and Saito (1992). (See footnote 6.) Consider the schematic representation in (15) and the examples in (16) and (17): (15) [c p WP, [tP-([island ) - . nom W P... (16) a. Who ate what? b. Who thinks that John ate what? c. Who met a man who wrote what? The sluicing analysis of why in this context is indirectly supported by the fact that why appears only in a peripheral position unlike its non-interrogative counterpart fo r this reason, as the contrast below illustrates: (iii) a. *Who believes why that John left for NY? b. Who believes for this reason that John left for NY? Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 56 (17) a. Why did John buy what? b. Why do you think that John bought what? c. Why did John meet a man who bought what? The SPEC of CP is filled with a nominal WP, i.e., who, in the sentences in (16). and a non-nominal WP, i.e., why, in the sentences in (17). A nominal WP is located in the matrix clause in (16a) and (17a), in the complement clause of a bridge verb in (16b) and (17b), and in an island in (16c) and (17c). All the sentences in (16) and (17) are 7 D grammatical.' 7 Apparently the sentences like the following are not subsumed under the generalization: (i) ’ Why does who leave early? Non-nominal WP why has moved to the SPEC of CP, while nominal WP who remains in subject position, conforming to the structure that the generalization incorrectly predicts to be acceptable. But we could ascribe the illicitness of the structure to a violation of the economy principle, the superiority condition. Note that Attract-a on the part of the [+wh] C° has not attracted the most local WP, who, which is generated in a position higher than why. Note also that the traditional ECP analysis could not account for the ungrammaticality, considering that a WP may be located in the embedded subject position, while why is in the SPEC of CP, as in (ii), which is substantially better than (0: (ii) Why do you think that who left early? * The grammaticality of the sentences in (17) invalidates the claim that absorption takes place exclusively among WPs that vary over individuals. Still we need to explain why a WP does not share the scope with whether and i f Consider the following example, where what can only have a matrix scope: (i) Who wondered whether John bought what? a. which person x, which thing y, x wondered whether John bought y. b.#which person x, x wondered [which thing y whether John bought y] WP whether varies over the truth values of the embedded clause while WPs like who and what vary over individuals. We may have to distinguish a [+wh] Q from a [yes/no] Q and assume that WPs other than whether are not compatible with the [yes/no] Q. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 57 In sum, nominal WPs are allowed in situ whether or not a barrier intervenes between them and their scope position and whether the SPEC of CP is filled with a non- nominal WP or a nominal WP. In contrast, non-nominal WPs cannot remain in situ (except for the all-over-the WP cases). The contrast is schematically illustrated below: (18) a. [cp WP D p ... ( U * ...) """WP... b-*[cP WP[IP...( [is la n d • • • ) ^ W P ... 3.3.2. WPs in situ in Korean Type Languages WPs in the Korean type languages remain in situ since the interrogative force in question clauses is licensed by a predicate with a question conjugation, as discussed in the previous chapter.9 Therefore, non-nominal WPs as well as nominal WPs remain in situ, as schematically represented in (19) and exemplified in (20): (19) a .... n o m W P... VQ E ? b .... > d v W P... VQ E ? 9 I suppress Takahashi’s (1993) analysis of scrambled WPs in which he argues that WPs in languages like Japanese optionally move to the SPEC of CP as a question operation. He provides several arguments including the ones based on the scope interpretation of scrambled WPs and sluicing. There is some controversy on the scope interpretation. See Kang and MQller (1994), for example. As for the sluicing argument, see Hoji (1990), who analyzes the stripping construction with case-maiicing on a par with the cleft construction. See also Nishiyama et al. (1994), who argue against Takahashi’s sluicing argument. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 58 (20) a. John-i mwues-ul mek-ess-ni? J.-Nom what-Acc eat-Pst-QE ‘What did John eat?’ b. John-i wav o-ess-ni? J.-Nom why come-Pst-QE ‘Why did John come?’ But non-nominal WPs do not enjoy as much freedom as nominal WPs do. As was observed in Chinese by Huang (1982), nominal (argument in his term) WPs can be embedded under a syntactic island, whereas non-nominal WPs cannot. The same is true of Korean WPs, as discussed in Lee (1982), Choe (1987), and Nishigauchi (1986). among others. Consider the following schematic structures in (21) and examples in (22): (21) a. ...[island...n o m W P ...]...V Q E ? b. *... [isund...a d v W P ...]...V Q E ? (22) a. John-i fnwu-ka ssu-n] chayk-ul ilk-ess-ni? J.-Nom who-Nom write-Adn book-Acc read-Pst-QE ‘Q John read the book that who wrote?’ b. * John-i [Mary-ka wav ssu-n] chayk-ul ilk-ess-ni? J.-Nom M.-Nom why write-Adn book-Acc read-Pst-QE ‘Q John read the book that Mary wrote why?’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 59 In (22a), a nominal WP nwv ‘who’ is embedded under the complex noun phrase. In (22b) a non-nominal WP way ‘why’ is also in the complex noun phrase. (22a), but not (22b), is grammatical. In addition to the island sensitiveness, there is another restriction on non-nominal WPs in situ in wh-in-situ languages. Aoun (1986:96) observes that two adverbial WPs cannot have the same scope in Chinese:1 0 (23) a. * [C P [c- Q [ip ... ^W P i... ^ W P ,...]]] b. *Jim weisheme zeme mai-Ie shu? J. why how buy-ASP book ‘Jim bought the book why, how?’ In Korean also, two or more non-nominal WPs cannot have the same scope even if no barrier intervenes between any of them and their scope position, as exemplified in (24):" 1 0 Aoun (1986) accounts for the ungrammaticality of the sentence by means of the generalized binding theory: one of the traces of the two non-nominal WPs will not be bound in its governing clause, i.e., the whole clause, due to the comp-indexing story. 1 1 The Korean counterpart of (23b) is ungrammatical, but I avoid discussing it for the reason that switching the order of the two WPs makes it acceptable, as the following contrast illustrates: (i) a. *John-i wav ettehkev chayk-ul sa-ess-ni? J.-Nom why how book-Acc buy-Pst-QE ‘Q John bought the book why how?’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 6 0 (24) *wav John-i [Mary-ka wav chayk-ul ilk-ess-ta]-ko malha-ess-ni? why J.-Nom M.-Nom why book-Acc read-Pst-DE-C say-Pst-QE ‘Why did John say that Mary read books why?’ Thus for a non-nominal WP in Korean to be licensed, at least the following two conditions should be met: no barrier intervenes between it and the scope position; and there must not be another non-nominal WP that shares the scope with it. Overall, nominal WPs in the Korean type remain in situ whether a barrier intervenes between them and their scope position. In contrast, non-nominal WPs are not allowed when a barrier intervenes between them and their scope position. Additionally two or more non-nominal WPs cannot share their interrogative scope. The three points are schematically represented below: b. ?? John-i ettehkev wav chayk-ul sa-ess-ni? J.-Nom how why book-Acc buy-Pst-QE ‘Q John bought the book how why?’ The contrast is reminiscent of the so-called anti-superiority effects, i.e., the order restriction in multiple WH-question (“ “ WP ... “‘ 'W P vs. * * dvWP... °°” WP). (See Saito 1982, 1989, 1992, A. Watanabe 1991, S. Watanabe 1995, and chapter 4 o f this thesis.) Furthermore, WP ettehkey ‘how’ may appear in syntactic islands, although not as freely as nominal WPs. See T. Chung (1991). It is not clear what causes the (semi-)nominal-like behaviors o f ettehkey. I conjecture that some nominal feature is involved in it, taking a hint from K. Lee’s (1993:369) analysis of Korean causative morpheme -key as a nominalizer -ki plus a postposition -ey ‘to;at’ A similar analysis may apply to -key in ettehkey. That is, ettehkey could be decomposed into eaeh-ki-ey. Note that etteh is an interrogative predicate meaning ‘to be in which state or to have which shape,’ or its indefinite counterpart ‘to be in a certain state or to have a certain shape.’ Our discussion on Korean “^WPs will be restricted to way ‘why,’ unless otherwise specified. See also Tsai (1994), according to whom the Chinese counterpart zenmeyang behaves like a nominal WP if it has a resultative or instrumental reading, although it behaves like a non-nominal WP if it has a manner reading. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 61 (25) a. [c p [c- Q [ip ... ([island ...) n o m W P... b. [CP [c Q [iP -..(*[island — ) ^W P ... c . * [ c p [ c Q [ i p ...a d v W P...a d v WP... 3.4. On Movement Analyses In this section we will review some earlier approaches that try to account for the facts described in the previous section in terms of WP-movement at LF. It will be pointed out that neither the ECP approach combined with a Comp indexing mechanism nor Nishighauchi’s subjacency theory based on the large scale pied-piping correctly captures the facts. 3.4.1. The ECP Approach Combined with a Comp Indexing Mechanism As was discussed in the previous section, non-nominal WPs are disallowed in situ in the English type languages. Consider the following examples: (26) a. *WhOj do you want tj to believe [that John ate apples why]? b. *How, did Mary say [that John ate apples whvl tj? Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 6 2 Regardless of the categorial status of the overtly moved WP, adverbial WPs are not allowed in English. The ECP combined with a Comp indexing mechanism was once considered to be responsible for the unacceptability of non-nominal WPs in situ in the English type languages. See Aoun (1986), Aoun et al (1981), and Lasnik & Saito (1984. 1992). among others. The general idea of the Comp indexing theory is that Comp receives the index of the WP that first moves there.1 2 The sentences in (26a, b) will have the following schematic structures before non-nominal WP why moves at LF: (27) a. CP b. CP [+wh] tj... why [+wh] ...tj... why After non-nominal why raises and adjoins to the SPEC of CP at LF, the structures will be like (28) below: 1 2 Aoun’s (1986:99) Comp indexing mechanism does not require such a restriction. In his system, when more than one element is in Comp, Comp receives a multiple index, which cannot be an antecedent for a binding theoretic reason: indexes cannot intersect. Thus that-trace effect for subject extraction follows. For the lack of the r/iaf-trace effect for non-nominal extraction Aoun assumes an index deletion mechanism at LF along the lines o f Lasnik & Saito (1984). Reproduced with permission of the copyright owner. Further reproduction prohibited w ithout permission. SPECi e whyj SPECj C IP whoj [+wh] ...tj ...tj SPECi e whyj SPECj C IP howj [+wh] ...tj... tj. Comp having index i, the index of who in (28a) or the index of how in (28b), the index of the adverbial WPj is invisible within the SPEC. Therefore, the trace of whyj is not antecedent governed. The same mechanism seems to be responsible for the fact that two or more adverbial WPs cannot share their interrogative scope in the Korean type languages, as was discussed in section 3.3.2. (cf. Aoun 1986) Consider the sentence in (24), repeated as (29) below: (29) *wav John-i [Mary-ka wav chayk-ul ilk-ess-ta]-ko malha-ess-ni? why J.-Nom M.-Nom why book-Acc read-Pst-DE-C say-Pst-QE ‘Why did John say that Mary read books why?’ Regardless of which of the two non-nominal WPs moves first, one of them will not be antecedent governed. Thus, (29) will be excluded in the same fashion as (26b) is. However, there seems to be a conceptual flaw with the attempt to account for the nominal/non-nominal asymmetry by means of the Comp indexing mechanism and Reproduced with permission of the copyright owner. Further reproduction prohibited w ithout permission. 64 binding relation (antecedent government). Note that whenever an element moves, the landing site is always a position that c-commands the launching site in the computational system due to the proper binding condition of Fiengo (1977), or the extension condition of merge in Chomsky (1994, 1995). In other words, “side-walk"' movement is generally disallowed. This tells us that the element in the landing site always c-commands its trace in the launching site. Under the assumption that the members of a chain share their index, the element in the landing site always binds the trace in the launching site. Thus if the Comp indexing mechanism is ever to be tenable, it is necessary to stipulate that the binding relation for the sake of identification of traces (antecedent government) should be different from the binding relation created as a result of a movement operation. Furthermore, the Comp indexing mechanism does not seem to be compatible with facts in X° movement. If a head moves, it adjoins to the immediately higher head as schematically represented below: (30) a* YP Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 65 In (30) head ( 3 is raised and adjoined to head a. If the index of ( 3 is not visible, and p could not antecedent govern its trace due to a similar restriction on indexing, then head movement would not be allowed at all with the structure given in (30).1 3 Thus the deviance of the sentences in (26) and (29) does not seem to lie in the identification of the trace left behind after a non-nominal WP raises, but in the licensing of the non-nominal WP itself. 3.4.2. Nishigauchi’s Pied-Piping Theory As observed by Huang (1982) in Chinese, nominal WPs can be scopally related to a position across an island, while non-nominal WPs cannot. The same is true of Korean and Japanese, as discussed in Lee (1982), Lasnik & Saito (1984), Nishigauchi (1986). and Choe (1987), among others. For example, nominal type WPs that appear in a complex NP can be scopally related to a scope position outside the island, whereas non- nominal type WPs cannot. In other words, the structure in (3 la) is licit but that in (31 b) is not, as exemplified in (32), cited from Nishigauchi (1986):1 4 1 3 One might avoid this problem by distinguishing XP adjunction from X° adjunction. If, however, the projectional status is contextually determined as in the minimalist program, the two types of adjunction cannot be distinguished any more. 1 4 Lee (1982) and Nishigauchi (1986) do not agree with Huang (1982) and Lasnik & Saito (1984) on the scope o f WPs within a WH-island. The latter linguists claim that they can have the matrix reading as well as the embedded reading, while the former linguists claim that they have the embedded reading only. See also S. Watanabe (1995), who claims that the island (in-)sensitivity depends on the types of question endings: polite endings vs. plain endings. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 6 6 (31) a. [ c p [ c Q [ip — [co m p lex NP — n 0 m WP V q e b- * [ c p [ c Q [ip— [complexNP — “ ^ W P V q e (32) a. (=Nishigauchi 1986:66, his (57)) Kimi-wa [n p [s dare-ga kai-ta] hon]-o yomi-masi-ta ka? you-Top who-Nom write-Pst book-Acc read-Hon-Pst-QE 'You read books that who wrote?’ b. (=Nishigauchi 1986:119, his (24)) * [n p [s kare-ga naze kai-ta] hon]-ga omosiroi-desu-ka? he-Nom why wrote book-Nom interesting-be-QE ‘Books that he wrote why are interesting?’ Both nominal WP dare ‘who’ in (32a) and non-nominal WP naze ‘why’ in (32b) are within a complex NP. The former, but not the latter, has a matrix scope. Nishigauchi (1986) attributes the discrepancy between nominal WPs and non- nominal WPs embedded under an syntactic island to the Subjacency Condition under the following two assumptions: the Subjacency Condition applies to LF operations as well as to SS operations; a large scale pied-piping may or may not be available depending on the categorial match or mismatch between WPs and the containing categories. WP dare ‘who’ in (32a) first moves at LF to the SPEC of a relative clause which is analyzed as the SPEC of the head noun in his system. The feature [+wh] on the WP can be percolated up to the whole complex NP since both the WP and the complex NP Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 67 share the categorial status. Now the whole complex NP, having become a WP, moves to the local domain of the [+wh] C° without violating the Subjacency Condition. In contrast, such a feature percolation followed by a pied-piped movement is unavailable for WP naze ‘why’ in (32b) since the non-nominal WP and the NP containing it do not match in their categories. Hence such a non-nominal WP must move directly to its scope position, resulting in a violation of the Subjacency Condition. Such a pied-piping theory based on the categorial match or mismatch also seems to explain the WH-island effects in languages like Japanese or Korean. WPs in a WH- island cannot have scope wider than the immediately dominating question clauses in these languages, as exemplified below:1 5 1 5 The reading in (33b) is in fact marginally available, with a heavy focus on the WP nwu ‘who.’ As far as that reading is available, the reading in (d), where both WPs have the matrix scope, also seems possible for me. I conjecture that the WPs in these cases have undergone scrambling across the embedded question clause, as illustrated in (ia) and (ib) below: (i) a. John-un nwu-kaj [t, mwues-ul kacyeo-n-ci] al-ni? J.-Top who-Nom what-Acc bring-Adn-CI know-QE b. John-un nwu-kaj mwues-ulj [tj tj kacyeo-n-ci] al-ni? J.-Top who-Nom what-Acc bring-Adn-CI know-QE One could test the validity o f the scrambling analysis by adding an adverbial element in front o f nwu-ka ‘who-Nom.’ Assuming that (at least some) adverbials do not undergo scrambling across a clause boundary, the expectation would be that no reading in (33b) or (33d) will be available, which seems to be borne out, as the following sentence indicates: (ii) John-un fkakkum nwu-ka mwues-ul kacyeo-n-ci] al-ni? J.-Top often who-Nom what-Acc bring-Adn-CI know-QE a. ‘Does John know who often brought what?’ b. # ‘ W hOj does John know whatj tj often brought tj?’ c. “ ‘Whatj does John know whoj tj often brought tj?’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 6 8 (33) John-un [nwu-ka mwues-ul kacyeo-n-ci] al-ni? J.-Top who-Nom what-Acc bring-Adn-CI know-QE a. ‘Does John know who brought what?’ b. # ‘ Whoj does John know whatj tj brought tj?’ c. # ‘Whatj does John know whoj tj brought tj?’ d. # ‘ Whoj. whatj does John know if tj brought tj?’ Neither of the WPs in the embedded clause has the matrix scope. This seems to be subsumed under the pied-piping theory combined with the category match restriction under the assumption that WPs are NPs (nominals) or PPs (adjuncts) and the question clause is a CP. The [+wh] feature in WPs cannot percolate up to the covering question clauses due to the categorial mismatch.1 6 Nishigauchi takes the pattern in elliptical answers in Japanese as one of his arguments for the pied-piping analysis.1 7 According to him, in answering the question d. # ‘ Whoj, whatj, John knows if tj often brought tj?’ When the frequency adverb kakkum ‘often’ is associated with the embedded predicate, it seems hard for any of the WPs to have a matrix reading. I absolutely do not have the reading in (33c), where the second WP has the matrix scope while the first WP has the embedded scope. Note that this is also expected with the scrambling analysis of WPs with a matrix reading since the WP in object position cannot be considered as being scrambled without the subject WP being scrambled. The last case seems to be the pure WH-island effect 1 6 I argue in section 3.6.3., however, that the question clauses have nominal properties. If this is correct then the WH-island effects in Japanese and Korean cannot be ascribed to the categorial mismatch. 1 7 Nishigauchi provides essentially two other arguments for his pied-piping analysis. One of them is based on the reading in sentences like (i) below: Reproduced with permission of the copyright owner. Further reproduction prohibited w ithout permission. 6 9 in which a WP is contained within a complex NP, Japanese speakers tend to not only provide the information corresponding to the WP, but also repeat the whole complex NP. For example, there are two possible answers to questions like (32a): (i) (=Nishigauchi 1986: 76, his (89)) ??[[Nan-ka-koku kara kita] daihyoo]-ga atumari-masi-ta ka? how;many;countries from came reps-Nom gather-Pol-Pst QE ‘Representatives who came from how many countries gathered (at the summit meeting)?’ His argument goes as follows. If nan-ka-koku ‘how many countries’ alone moves to its scope position at LF, it will have scope over the quantifier daihyoo ‘representatives,’ which is not a possible reading according to him. If, however, the WP moves within the whole complex NP and the whole complex NP moves to the scope domain, the WP will remain within the domain of the quantifier daihyoo. Thus the strange reading that the sentence has is captured. Fiengo et al (1988) defend their ECP position by decomposing expressions like nan-ka into an interrogative part and a quantifier part and only the former is relevant to WH-Iicensing. Along similar lines, Chomsky (1994, 1995) would avoid the problem with the feature raising approach, in which the quantifier part would not be raised. Another argument that he provides is that a WP within an island can bind a pronoun outside the island, as schematically represented below: (ii) ... L i^ h ... WP,...]... pronoun;... Nishigauchi claims that although no direct c-command relation obtains between the WP and the pronoun, the bound variable reading of the pronoun is available if the containing category c-commands the pronoun. But I conjecture that the pronoun in this context is not a bound variable but an E-type pronoun in the sense of Evans (1980), since the dependency relation exists even when the containing category does not c- command the pronoun: (iii) nay-ka [John-i mwues,-ul mek-ess-nun-ka] mwut-ess-ciman I-Nom J.-Nom what-Acc eat-Pst-Adn-QE ask-Pst-but (nay-ka) [(ku-ka) kukes,-ui ettehkey mek-ess-nun-ka]-nun mwut-ci ani-ha-ess-ta. (I-Nom) he-Nomit-Acc how eat-Pst-Adn-QE-Con ask-CI Neg-do-Pst-DE ‘I asked whati John ate, but I did not ask how he ate itj.' WP ntwues ‘what’ appears in the question complement clause of the first conjunct and an anaphor related to it is in the question complement clause of the second conjunct No c-command relation obtains even in the extended sense with the help of the large scale pied-piping. One might claim that the index of the WP is percolated up to the matrix clause, in which case the c-command relation might obtain. But notice that the large scale pied-piping was intended to make it possible for an island internal WP to move to the government domain of a [+wh] C°. Thus the pied-piping o f a WP beyond its scope position is directly against the original intent of the large scale pied-piping. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 70 (34) (=Nishigauchi 1986:66, his (79)) a. Austen-desu Austen-be ‘(It is) Austen.’ b. Austen-ga kai-ta hon desu Austen-Nom write-Past book be ‘(It’s) the book that Austen wrote.’ If no pied-piping is assumed, only (34a) can be the possible answer. Nishigauchi takes the fact that (34b) is a possible answer as an indication that an LF pied-piping takes place in questions like (32a). As for the fact that (34a) is also a possible answer, he simply suggests that (34a) is a truncated form of (34b).1 8 There are, however, sets of data that defy Nishigauchi’s pied-piping theory based on the categorial match or mismatch. First, nominal type WPs can appear within adjunct clauses: (35) a. [John-i nwukwu-lul manna-mvenl coh-kess-ni? J.-Nom who-Acc meet-if good-Conj-QE ‘Who do you think it will be good if John meets tj?’ '* But Kuno and Masunaga (1985) point out that sometimes both a shorter answer and a longer one are possible, and other times a shorter answer is preferred to a longer one against Nishigauchi’s expectation. In addition, as Nishigauchi (p. 100, footnote 22) admits, giving credit to Barbara Partee. what questions like (32a) solicit as an answer is primarily not the identity of books, which is clear from the fact that answers like (i) below are not appropriate: (i) Pride and Prejudice desu. is Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 71 b. [John-i nwukwu-lul manna-se] hwana-ess-ni? J.-Nom who-Acc meet-because get;angry-Pst-QE ‘Who are you angry because John met tj?’ If Nishigauchi’s restriction on the categorial match and mismatch is strictly followed, pied-piped movement of WPs in such sentences would be impossible since adjunct clause like conditionals and rationale clauses are not nominal. Thus it is expected that nominal type WPs in the adjunct clauses would violate the Subjacency Condition when they move to the scope position, the SPEC of the matrix CP. Despite the categorial mismatch the sentences in (35) are perfect. Thus either the large scale pied-piping analysis is wrong, or islands like adjunct clauses do not block movement of a nominal WP. A second problem is just opposite to the first one. If categorial match or mismatch really determines the availability or non-availability of a large scale pied-piping, then adverbial WPs are expected to undergo a pied-piped movement when they occur in an adjunct clause. But adverbial WPs are not allowed within an adjunct island either, as exemplified below: (36) a. *John-un [Bill-i caki-lul wav ttali-ese] pang-ul naka-ess-ni? J.-Top B.-Nom self-Acc why hit-because room-Acc leave-Pst-QE ‘*Whyj did John leave the room [because Bill hit him tj]?’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 72 b. * [John-i chayk-ul way ilk-umyen] sensayngnim-i cohaha-ni? J.-Nom book-Acc why read-if teacher-Nom like-QE 'Why; does the teacher like it if John reads books tj?’ Adverbial WP way ‘why’ cannot appear in an -ese ‘because’ clause as in (36a) and a conditional as in (36b). Note that way ‘why’ and the adjunct clause are of the same category, as schematically shown in (37) below:1 9 (37) [ [+wh] C [ [ a d v p i — [A D V P 2 way]...]...]] If we strictly follow Nishigauchi’s logic on the large scale pied-piping of WPs, ADVP2 way ‘why’ will first move to the SPEC of the containing adverbial clause ADVPI. rendering the whole clause a WP. Now the whole adjunct clause as a WP would move to the government domain of the [+wh] C°. Thus the sentences in (36) would be expected to be grammatical, which is not true. Another problem that the large scale pied-piping theory faces is that it mistakenly predicts sentences like (38) below to be illicit: ( 3 8 ) ne-nun [n p i[cp [pp[NP2 en** tosi]-eyse] cala-n] salamj-ul manna-ess-ni? you-Top which city-in grow-Adn man-Acc see-Pst-QE ‘Which cityj did you meet a man who grew in ej?’ 1 9 The categorial status of adverbials is not settled in the literature. My argument in the text does not hinge upon any particular theory on the adverbials, however. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 73 In (38) WP enu tosi ‘which city’ is a complement of a P, eyse ‘in,’ whose projection. PP, is embedded in a complex noun phrase, NP1, while the scope position of the WP is outside the complex noun phrase. In LF. depending on the scale of the pied-piping, either WP enu tosi ‘which city’ or PP containing the WP enu tosi-eyse ‘in which city* moves to the SPEC of the relative clause. Thus the LF of (38) would be roughly either (39) or (40) below: (39) CP in which city Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 74 (40) CP IP C which city lj V / N L N t ^ ^ P grlw P grow If PP enu tosi-eyse ‘in which city’ moves within the relative clause as in (39), a pied- piped movement is not available in the next step of movement since the PP (circled) and the containing NP (squared) do not match categorially. Thus this should be excluded just like the case where a non-nominal WP moves out of a complex NP. If WP enu tosi ‘which city,’ the circled NP, alone moves within the relative clause as in (40), the pied- piping is possible since the containing category is also an NP (squared). But the first step of the movement violates the Subjacency Condition since it crosses the barrier PP.2 0 Thus neither option should be legitimate, contrary to the fact. The discussion so far leads us to conclude that Nishigauchi’s (1986) subjacency account of the nominal/non-nominal asymmetries in the island context could not be 2 0 I follow Chomsky (1986) in assuming that every non-lexically governed XP is a barrier. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. correct In particular, his claim that a large scale pied-piping is or is not available depending on the categorial match or mismatch between WPs and containing categories cannot be factually supported. This is because nominal WPs in situ are insensitive to any kind of syntactic islands (putting aside WH-isIands, cf. footnote 15), whereas non- nominal WPs in situ are sensitive to every kind of syntactic islands. 3.5. Two Options of WP Licensing: Binding Relation (BR) vs. Exclusive SPEC- Head Relation (ESHR) In this section it will be proposed that there are two different licensing modes of WPs in situ depending on the categorial status of WPs; nominal WPs are licensed via a binding relation with Q,2 1 whereas non-nominal WPs are licensed via an exclusive SPEC-head relation (henceforth ESHR, to be defined below) with Q.2 2 3.5.1. Licensing of Non-nominal WPs in situ: ESHR Let us first discuss how non-nominal WPs in situ are licensed. It will be proposed in this section that non-nominal WPs are licensed by having an exclusive SPEC-head relation with a [+wh] C°. The ESHR and its auxiliary notions are defined below: 2 1 It will be argued in section 3.6. that the real binder is not Q but a nominal feature that a question clause has. " For the options of NPI licensing, see Chung (1995a), where I propose a similar double option for the licensing of NPIs. with permission of the copyright owner. Further reproduction prohibited w ithout permission. 76 (41) The Exclusive SPEC-head Relation (ESHR) An element a has an ESHR with a head P iff the shortest path between a and ( 3 contains one and only one maximal element, excluding a and p.2 3 - 2 4 (42) a. A path from a node a to another node P is a set of nodes the union of which links a and p. b. The shortest path from a node a to another node P is the path that contains the smallest number of nodes. There is only one position that has an ESHR with a head. Therefore for a non-nominal WP to be licensed, it competes for the position that has an ESHR with a [+wh] C0.2 3 Now let us consider how the ESHR accounts for the distribution of non-nominal WPs both in the English type languages and the Korean type languages. As was discussed in section 3.3.1., non-nominal WPs in English are allowed only in the SPEC of CP in overt syntax. No adverbial WPs can remain in situ. Consider the following structures and examples, where Q stands for a [+wh] C°: (43) a. [C pa d v WPi [cQ [IP...tj ...(n o m WPJ )... 2 3 By a maximal element I mean any XP, not distinguishing categories and segments. 2 4 To distinguish a SPEC-head relation from a head-complement relation, we may need an additional condition. Note that, if a head bears a complement but no SPEC, the complement will have an ESHR with the head. This, however, does not raise any problem with respect to the licensing of non-nominal WPs. Notice that the complement of a [+wh] C° is IP, which is never a WP. 2 5 The ESHR is a special exploitation of the idea of the Comp-indexing mechanism (Aoun et al 1981. Lasnik & Saito 1984, 1992, Aoun 1986, among others). Note that only one position is competed for in the SPEC of CP both in the ESHR and the Comp indexing mechanism. The unique position is competed for in order to license WPs themselves, not their traces, in the ESHR. It was pointed out in section 3.4.1. that the Comp-indexing for the purpose of licensing the traces of WPs has some conceptual flaw. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. b. W hy* do you think [that John ate (an apple/what}] tj? (44) a. * [ c p nomW P -, [C ’ Q [p ... tj ...- ‘" W P j ... b. ♦Who, do you want tj to believe that John ate apples why? (45) a. ♦ [ c P advW P i [c- Q [fp ... t,... ^ W P j ... b. * Whyj do you believe [that John ate apples why] tj? The adverbial WP in (43), but not the one in (44) and one of the two adverbial WPs (45), has an ESHR with Q, as the following structures illustrate: (46) CP adv WP, c C IP [+ N y h ] . . . t j f-w l ESHR (48) C P adv< W P j C ’ adv. W P j advW P j C I P [+■ -ESHR r i] ...tj...tj 'ESHR (47) CP " W P , C ’ advxj rrj nom W P j W P ; C I P [+wh] ...t, ...tj ESHR ♦ESHR Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 78 In (46) the adverbial WP and the [+wh] C° are in an ESHR since the shortest path between the two nodes contains only one maximal element, i.e., CP.2 6 In (47) and (48) “^WPj is not in an ESHR with the [+wh] C° since the shortest path between them contains two maximal elements, i.e., CP and n o m WPj in (47), and CP and “^WP, in (48). Thus (46) is licit, but (47) and (48) are not, accounting for the contrast between (43b) on the one hand and (44b) and (45b) on the other. Now consider the case where a non-nominal WP remains in situ with no WP in the SPEC of CP: (49) a. * [Cp [C Q [ip ...advW P...] b. *John ate apples why? After LF raising of the non-nominal WP to the SPEC of CP, (49) would have the same structure as (46). Thus the ESHR will obtain between the non-nominal WP and the [+wh] C°. Then why is the sentence ungrammatical? This is because the question force encoded in the [+wh] C° has not been licensed before spell-out WH-questions in English are not allowed as far as no WP resides in the SPEC of CP in overt syntax, irrespective of the categorial status of WPs in situ. 2 6 O f course the non-nominal WP why, in fact any WP in the SPEC of CP for this matter, licenses the strong feature [+wh] C°, the latter attracting the former. So if there is only one WP in the scope of a [+wh] C°, licensing of the strong feature [+wh] C° and licensing of the WP are cooperative to each other. When more than one WP is involved, only one is responsible for the licensing of [+wh] C°, and the rest of WPs must be licensed by having a proper relation with the [+wh] C°. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 79 Another illegitimate structure is the case where a non-nominal WP moves across a barrier (50) a. *[C p “^WPj [c Q [ n > ... [b r ~ t, ... b. *Why, do you meet [the man who John hit t J? Again there is no problem with the ESHR between the non-nominal WP and the [+wh] C°. But the derivation includes an illicit movement across a barrier. Let us now turn to the licensing of non-nominal WPs in situ in the Korean type languages. Unlike those in the English type languages, non-nominal WPs in the Korean type languages are allowed in situ, as far as no barrier intervenes between them and their scope position: (51) a. [c p [Q [ ip ...a d v WP... b. John-i [Mary-ka wav cwuk-ess-ta]-ko malha-ess-ni? J.-Nom M.-Nom why die-Pst-DE-C say-Pst-QE ‘Why; did John say that Mary died tj?’ Since no WP occupies the SPEC of CP in overt syntax, the non-nominal WP in (51) moves to the position. The resulting structure will be like the one below, which is similar to the one in (46) in the English type languages: Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. ESHR The shortest path between “^WP and the [+wh] C° contains only one maximal element, i.e., CP, satisfying the ESHR. Incidentally, sentences like (51b) do not have a problem with respect to the licensing of the question force encoded in [+wh] C°, although there is no overt WP in the SPEC of CP. Note that the question force in the Korean type languages is licensed by attracting a predicate with a question conjugation, as was postulated in chapter 2. Thus Korean sentences like (51) contrast with English sentences like (43) in the way the question force is licensed. When a barrier intervenes between a non-nominal WP and its scope position, then the sentence is ungrammatical: (53) a. *[C p [Q [ n > ... [b ia n d ...a d w WP ... b. * John-i [[Mary-ka wav Sam-eykey cwu-n] chayk-ul] peli-ess-ni? J.-Nom M.-Nom whyS.-to give-Adn book-Acc throw-Pst-QE ‘Why did John throw away the book that Mary gave to Sam tj?’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 81 Just as in the non-nominal movement across a barrier in the English type languages (see (50)), there is no problem with the ESHR between the non-nominal WP and the [-f-w h] C°. The derivation, however, includes an illicit movement across a barrier. Let us now consider the cases where two or more ^W Ps are involved in a question. As was discussed in section 3.3.2.. such cases are unacceptable. See the structure in (23a) and the sentence in (24), which are repeated below as (54) and (55). respectively: (54) *[C P [c Q[iP...a d v W P...a d v WP... (55) * way John-i [Mary-ka wav chayk-ul ilk-ess-ta]-ko malha-ess-ni? why J.-Nom M.-Nom why book-Acc read-Pst-DE-C say-Pst-QE "Why did John say that Mary read books why?’ Since there are two non-nominal WPs in the sentence, one of them necessarily does not have an ESHR at LF, as schematically represented below: (56) C P adv' W P i e adv. W P j ^ W P i I P ...tj... tj ... [+wh] ESHR ♦ESHR Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 82 WP, has an ESHR with the [+wh] C° since there is only one maximal element. CP. between them. In contrast, WPj does not have an ESHR with the [+wh] C° since there are two maximal elements, CP and WPj. Thus one of the two non-nominal WPs in (55) is not licensed, explaining the ungrammaticality. 3.S.2. Licensing of Nominal WPs in situ: Binding As was discussed in section 3.3.1., nominal WPs remain in situ as far as another WP fills the SPEC of CP in the English type languages. The categorial status of the WP that is in the SPEC of CP does not affect the licensing of nominal WPs in situ. In other words, nominal WPs are allowed whether the WP in the SPEC of CP is a nominal or non-nominal WP, or whether a barrier intervenes or not For example, all the following sentences are acceptable: (57) (=(16)) a. Who ate what? b. Who thinks that John ate what? c. Who met a man who wrote what? Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 83 (58)(= (17)) a. Why did John bought what? b. Why do you think that John bought what? c. Why did John meet a man who bought what? In (57) a nominal WP has moved to the SPEC of CP and another nominal WP remains in situ. The nominal WP in situ can be in the same clause in (57a), in the complement clause of a bridge verb in (57b), or even in an island in (57c). In (58) a non-nominal WP has moved to the SPEC of CP and a nominal WP remains in situ. The nominal WP in situ can be in the same clause in (58a), in the complement clause of a bridge verb in (58b), or even in an island in (58c). At this point we may ask why nominal WPs are allowed in situ in comparison with non-nominal WPs. If nominal WPs in situ were licensed by an ESHR with a [+wh] C°. like non-nominal WPs discussed in the previous section, all the sentences in (57) and (58) should be ungrammatical since the nominal WPs in situ would not have an ESHR with a [+wh] C° after they are raised at LF, as illustrated in the diagram below: Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 84 (59) WP, WP ESHR ♦ESHR WPj in (59), having moved in overt syntax, is in an ESHR with the [+wh] C°, but the nominal WP that moves at LF, i.e., WPj, would not. Thus the prediction would be that the sentences in (57) and (58) are all unacceptable, which is not true. I propose that " “"WPs, as opposed to a d v WPs, are able to be licensed in situ by being bound by a [+wh] C0.2 7 -2 8 Given this binding option of WP licensing, nominal WPs in situ are licensed as follows. All the sentences in (57) and (58) will share the following schematic structure at LF: 2 7 The proposal here is similar to Cinque’s (1990) treatment of referential WPs in situ in that binding relation is responsible for the licensing of nominal WPs in situ. One difference is that binding relation licenses the traces of referential WPs in his system, but it licenses nominal WPs themselves in my system. 2 8 For the time being, let us suppress the oddness of the binding relation, where an XP is bound by an X°. In section 3.6. I argue that the binding relation is not directly related to the [+wh] C but to an abstract nominal element that is postulated on top of CP. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. nomW P- ESHR BR W P, licenses the interrogative force of the question encoded in the [+wh] C° and at the same time is licensed by having an ESHR with a [+wh] C°. WPj is licensed by having a binding relation with a [+wh] C°. The licensing of nominal WPs in the Korean type languages will not be discussed since they do not differ from those in the English type languages in their distribution, except for the cases where there is one nominal WP and no other WP in a domain of a [+wh] C°. Consider the following sentence: (61) John-i mwues-ul mek-ess-ni? J.-Nom what-Acc eat-Pst-QE ‘What did John eat?’ The interrogative force of a question sentence encoded in the [+wh] C° is licensed by a predicate with a question conjugation in the Korean type languages, and raising of a WP is not necessary. Then there seem to be two ways of licensing the nominal WP. one by Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 86 moving to the SPEC of CP and having an ESHR with the [+wh] C° and the other by having a binding relation with the [+wh] C°. The two options are schematically represented below: (62) a. CP n o m WPi C’ b. CP IP C I IP C ..."“ WP... [+wh] ... tj... [+wh] ESHR Binding First, it could move to the SPEC of CP and have an ESHR with [+wh] since there is no WP that fills the position in overt syntax. The other way of licensing the nominal WP is by having a binding relation with the [+wh] C° without recourse to movement.2 9 A question that arises at this point is whether grammar equally allows both options for nominal WPs or whether one of the two options should be eliminated in favor of the other. Following the perspective of the current theory, I assume that one of them is to be eliminated. But the question is how it is accomplished. Tsai (1994) claims that an operator-variable construction formed in terms of a binary operation (simple merge) is 2 9 Incidentally, when a barrier intervenes a nominal WP and its scope position, the WP cannot move across the barrier without violating a movement constraint like the Subjacency or Minimal Link Condition. Thus it simply remains in situ and is licensed by having a binding relation with the [+wh] C°. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 87 more economical than the one formed in terms of a singularly operation (move and merge). He argues that this is so because the binary operation is a theoretical necessity to build up a tree, while the latter option is not Our system, however, cannot strictly exploit the same logic since the question force, a theoretical equivalent to the operator in his system, is universally postulated as a strong formal feature. I conjecture that binding is more economical since it is a process that reads off given structures rather than building up structures. 3.5.3. Licensing of WPs in SPEC of CP In this section, the scope licensing of WPs in the SPEC of CP in overt syntax will be briefly discussed just for the sake of completeness. When a WP moves to the SPEC of CP in order to check the [+wh] C°, how is it licensed by the [+wh] C°? Although the WP in the SPEC of CP is raised in order to check the strong feature of a [+wh] C°. it has an ESHR with the head; so the licensing of the strong feature [+wh] C° and the scope licensing of the WP in the SPEC of CP are cooperative to each other. Therefore, there seems to be no additional process required for the scope licensing of WPs in the SPEC of CP. Reproduced with permission of the copyright owner. Further reproduction prohibited w ithout permission. 88 If. however, more than one WP is located in the SPEC of CP as in the case of East European languages as mentioned in section 2.1.,3 0 only one of the WPs will have an ESHR with the [+wh] C°. Ruddin (1988) argues that WPs in the multiple WH-fronting languages adjoin to the right If this is correct then the leftmost WP is in an ESHR with the [+wh] C°. Thus no additional process is required for the scope licensing of the leftmost WP. The rest of WPs could not be licensed in this way since they are not in an ESHR with the [+wh] C°. Thus they are not licensed in the adjoined positions. A possibility is that they are reconstructed to their original positions. The reconstructed WPs will now be licensed just like WPs in situ in English and in Korean. One prediction would be that multiple questions could not have the form [WP ... ^ ' WP ...] in the SPEC of CP. The non-nominal WP cannot be licensed in its reconstructed (original) position since the SPEC of CP is already filled by the leftmost WP just as in English. The prediction seems to be borne out. According to Rudin (1988:476, footnote 21). non-nominal WPs are not permitted in multiple questions in languages like Bulgarian and Romanian.3 1 3 0 See Rudin (1988) for the argument that languages vary with respect to the landing site of multiply fronted WPs. According to her, every WP in Bulgarian and Romanian moves to the SPEC of CP, but only one WP moves to the SPEC of CP, and the rest of WPs are adjoined to IP in Serbo-Croatian. Polish, and Czech. If this is true, our discussion to follow is restricted to the WPs in the former type of languages. 3 1 The order [“^W P... WP ...] is also reported to be ill-formed. I conjecture that this is ruled out as a violation of the superiority condition, analogous to the ill-formedness of the following sentence: (i) ’ Why, did who buy the book t,? Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 89 3.6. Deriving the BR vs. ESHR Disparity: Nominal Properties of Question Clauses In this section, I try to derive the difference that nominal WPs and non-nominal WPs display in relation to their licensing from the categorial status of question clauses as well as WPs themselves. More specifically, under the assumption that binding is restricted to nominal elements, nominal WPs, but not non-nominal WPs. can be licensed by the categorial feature of question clauses which I argue is also nominal. Furthermore, it will be argued that non-nominal WPs move to the SPEC of a [+wh] C° to be licensed since no binding option is available due to the categorial restriction on binding relation. 3.6.1. Basics of Binding Before directly going into the discussion of the source of binding vs. movement disparity, let us briefly discuss basics of binding relation. Binding contains three factors, binder, bindee, and configurational relation between the two. I do not discuss the configurational relation, just following the general consensus that c-command, although there may be various definitions of this, is responsible for the structural relation for binding. How about the restrictions on bindees or binders? There seem to be several restrictions on them. First, only maximal elements may participate in binding. It seems that no minimal and non-maximal elements in the sense of Chomsky Reproduced with permission of the copyright owner. Further reproduction prohibited w ithout permission. 9 0 (1994), true heads in the traditional sense, may have a binding relation. When two elements have a syntactic relation and one of them is a minimal and non-maximal element, the syntactic relation the two elements have is necessarily a checking relation since a head or its syntactic feature does not seem to be visible beyond its checking domain.3 2 Secondly, I assume that the elements involved in the binding relation must be nominal categories, i.e., DPs. Binding being construed as a relation of referential dependency, two elements in a binding relation must have a locus where referentiality may reside. In general nominals may express referentiality, while non-nominals like adverbs may not. Thus I restrict binding to relations between two nominal elements. 3.6.2. Nominal vs. Non-Nominal WPs: Binding vs. Movement Let us first discuss why non-nominal WPs must have an ESHR with a [+wh] C°. Non-nominals cannot have the binding option which is restricted to nominal elements due to a binding theoretic reason, i.e., referentiality can reside exclusively in nominal elements. Thus non-nominal WPs cannot remain in situ at LF. To be licensed, they must move to the SPEC of the [+wh] C°. If they remain in situ at LF, they cannot be licensed, violating the full interpretation principle. 3 2 In Chung (1995a) it is argued that a minimal negation licenses NPIs in terms of a checking relation, while a maximal negation licenses NPIs in terms of a binding relation. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 91 Unlike non-nominals, nominals may have a binding relation. Therefore, nominal WPs are able to be licensed by a binding relation if a proper antecedent is available and it has a proper configurational relation with the WPs. Then the first question to ask at this point is what would be the proper antecedent for WPs in situ. The answer should depend on the analysis of interrogative clauses. In the next subsection I will analyze question clauses as a DP modified by a CP. so that a proper antecedent for nominal WPs in situ is made available. 3.63 . Question Clauses as Nominal Projections The question clauses are generally analyzed as CP projected from a [+wh] C°, as schematically represented below:3 3 With this structure, there seems to be no proper antecedent for a nominal WP within IP which has not moved in overt syntax. SPEC of CP is not necessarily a nominal category 3 3 O f course, no overt element will be filled in SPEC for WP-in-situ languages. (63) CP SPE( WPj C IP [+wh] ... t;... WPj .. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. since it may host a non-nominal WP like how and why in English. The head of CP. i.e.. the [+wh] C°. cannot be a proper antecedent if only maximal elements may participate in a binding relation. Therefore there is no element that properly functions as the antecedent for a nominal WP within IP. For this reason, I discard the traditional analysis of question clauses in (63). Instead I propose the following structure for question clauses: (64) DP DP CP e SPEC C’ WPj IP [+wh] ...t;... WPj. ________________ t Binding The structure in (64) has a DP on top of the CP in (63). It is very similar to a relative construction in that a CP modifies a nominal element, except that the nominal element is an empty element in (64).3 4 Now that in (64) a nominal maximal element, the lower DP, is available as a proper antecedent for a nominal WP within IP, the licensing of the nominal WP in situ 3 4 A question arises with respect to the semantic value of question clauses. If question clauses are DPs, then they should denote some tokens rather than truth values. I have no idea about this, except for pointing out that question clauses do not directly denote truth values. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 93 in terms of binding relation is possible. In the following subsections, I will provide motivations for the DP analysis of question clauses. 3.6J.1. Embedded Question Clauses as DPs There are several indications across languages that question clauses are different from declaratives in their categorial status. In this subsection I will point out various aspects which show that embedded questions are nominal.3 5 3.6 J .1.1. Complements of P Prepositions or postpostions take nominal elements as their complements. A non- nominal element cannot be a complement of a postpostion or preposition without being nominalized. Consider the following Korean sentences and their English counterparts given as the translations: (65) a. John-un [ pp [ d p sophwung kyehoikj-eytayhaye]] malha-ess-ta. J.-Top picinic plan(N)-about talk-Pst-DE ‘John talked about the picnic plan.’ 3 5 Nominal properties of Korean question clauses are already noticed in the literature. Lee (1967) analyzes question endings like -ci/-ka/-ya as sentence nominalizers. Kim (1974) equates them with “nominal complementizer” kes in their syntactic behaviors. Here I merely reanalyze them as head nouns of complex noun phrases, not nominalizers or complementizers. Fukui (1986) also claims that the question marker ka in Japanese is [+N]. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 94 b. John-un [ p p [ d p sophwung-ul kyehoikha-nun kes]-eytayhaye]] malha-ess-la. J.-Top picinic-Acc pIan(V)-Adn thing-about talk-Pst-DE 'John talked about planning a picnic.’ c. * John-un [ pp [vp sophwung-ul kyehoikhaj-eytayhaye]] malha-ess-ta. J.-Top picinic-Acc plan(V)-about talk-Pst-DE ■*John talked about plan a picnic.’ d. * John-un [pp [cpwuli-ka sophwung-ul kyehoikha-ess-ta-ko]-eytayhaye]] J.-Top we-Nom picinic-Acc plan-Pst-DE-C-about malha-ess-ta. talk-Pst-DE ‘*John talked about that we planned a picnic.’ In Korean, nominal phrases or nominalized clauses can be a complement of postpositions like -eytayhaye ‘about,’ as in (a) and (b), respectively, whereas VPs or declarative clauses cannot, as in (c) and (d), respectively. Similarly in English, DPs like the picnic plan in (65a) or gerundive phrases like planning the picnic in (65b) may be a complement of the preposition about, whereas VPs like plan a picnic in (65c) or declarative clauses like that we planned a picnic in (65d) cannot.3 6 Assuming gerundive phrases are also nominal (Abney 1987), it is clear that prepositions in English and postpostions in Korean take a nominal element as their complement. 3 6 Exceptionally preposition in seems to be able to take a that-c\ause: (i) I agree with you in that the plan will take a lot of time. It is my personal opinion that a nominal expression like the point is deleted after the preposition in this case. Reproduced with permission of the copyright owner. Further reproduction prohibited w ithout permission. 95 Now let us examine the categorial status of question clauses with this operational test in mind. Unlike declarative clauses, question clauses may be complements of postpositions or prepositions. Consider the following Korean sentences and their English counterparts given as the translations: (66) a. na-nun [cp nwu-ka yeki sa-nun-ci]-eytayhaye kwungkwumha-ta. I-Top who-Nom here live-Adn-QE-about be;wondrous-DE ‘I am wondering about [who lives here].’ b. ku-nun [c p ku-ka ettehkey kukes-ul et-ess-nun-ci]-eytayhaye na-eykey he-Top he-Nom how it-Acc get-Pst-Adn-QE-about I-Dat malha-ci ani-ha-ess-ta. tell-CI Neg-do-Pst-DE ‘He did not tell me as to [how he accomplished that].’ The bracketed parts in (66a) and (66b) are question clauses that function as complements of postpostions and prepositions. The grammaticality of the Korean and English sentences conforms to our nominal analysis of question clauses. 3.63.1.2. Lack of ECM Question clauses behave like a case barrier. More specifically, the subject of a question clause cannot receive case from a higher predicate through the so-called Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 9 6 exceptional case marking (ECM). Compare the following two sentences cited from Kuno and Robinson (1972:476, footnote 14): (67) a. I know John to be trustworthy, b. * 1 know who to be trustworthy. As in (67a), verbs like know in English may assign accusative case to the subject of the complement clause when it is a non-interrogative infinitival clause. In contrast, such an exceptional case marking is impossible when the embedded clause is an interrogative one, as in (67b). The contrast could be easily taken care of by the proposed DP analysis of question clauses. The relevant structures for (67a) and (67b) will be like (68a) and (68b). respectively: Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 97 (68) a. AgroP AgroP Agro’ Agro’ Agr VP Agr VP CP A c IP A SPEC I’ DP DP A e CP A SPEC C’ I A WP c IP The economy principle accounts for the contrast, assuming that Agro is responsible for accusative case checking.3 7 In (68a), the DP in the SPEC of IP is the most local element that is able to bear case feature. It can move to the SPEC of AgroP without violating any principle. In contrast, the WP in the SPEC of CP in (68b) cannot move to the SPEC of AgroP since the covering DP or the nominal element e, a potential case bearer, is nearer to the case position, that is, the SPEC of AgroP.3 8 3 7 See Chomsky ( 1995) for a different view. 3 8 J. Aoun (p.c.) points out to me that the non-availability of ECM in (68b) may be due to the improper movement Notice that the WP in SPEC of CP, an A’-position, raises to SPEC of AgroP, an A-position. His comment seems to be correct as far as the English data is concerned, but the same contrast with respect to ECM appears in languages like Korean, which is a WH-in-situ language- Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 3.6 J . 1.3. Center Embedding and the CRP 98 Observing a dichotomy between the [+N] categories and the [-N] categories with respect to Case assignment Stowell (1981: 146) claims that core grammar includes the following principle: (69) The Case Resistance Principle (CRP) Case may not be assigned to a category bearing a Case-assigning feature. Among the categories bearing a Case-assigning feature is the category of Tensed clauses, which Stowell analyzes as having the feature composition [+N, -V. +Tense, +/- Past]. The CRP excludes the sentences which contain a center embedded declarative clause, as exemplified below: (i) a. John-un [Mary-lul papo-lo] al-n-ta. J.-Top M.-Acc fool-as know-Pres-DE ‘John knows Mary as a fool.’ b. John-un fnwukwu-ka/*-lul papo-i-nci] al-n-ta. J.-Top who-Nom/Acc fooI-be-QE know-Pres-DE ‘John knows who is a fool.’ Provided our hypothesis with respect to the question force licensing in chapter 2, i.e., the question force is licensed by a question conjugation on the verb in Korean, the WP subject in (ib) does not move to the SPEC of CP at any point of derivation. Then the lack of ECM in (ib) remains unexplained without positing the proposed structure of question clauses. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (70) a. * 1 think that [that John lied] is a fact] b. * Although [[that the house is empty] may depress you],.. 99 These sentences are out since the bracketed tensed clauses are in a case position which is assigned Case from the immediately higher tensed clause.3 9 Unlike declarative that-clauses, question clauses may appear in this case bearing position, as exemplified below: (71) a. I think that what you get depends on what you want, b. Although what you get depends on what you want,... If the question clauses above are simply tensed CPs, then sentences should be ungrammatical, violating the CRP. If, however, the question clauses are nominal as proposed earlier, then the grammaticality naturally follows. Nominals, not being a case assigning category, may appear in a Case position. 3 9 Apparently the following sentences violate the CRP. (i) a. [That John lied] is a fact. b. Paul already knows [that John lives with his sister]. Stowell (1981:152, 159-162) avoids the problem by analyzing the tensed subject clause in (ia) as being topicalized, basically following Emonds (1976) and Roster (1978), and the tensed object clause in (ib) as being extraposed, as follows: (ii) a. [That John lied]* e, is a fact. b. Paul already knows e, [that John lives with his sister]*. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 100 3.6.3.1.4. Case Morphology Korean question clauses may bear case morphology unlike declarative clauses, as the following examples illustrate: (72) a. na-nun [John-i nwukwu-lul cohaha-nun-ci]-ka kwungkumha-ta.4 0 I.-Top J.-Nom who-Acc like-Adn-QE-Nom be;wondrous-DE ‘I wonder who John likes.’ b. na-nun [John-i nwukwu-lul cohaha-nun-ci]-lul al-n-ta. I.-Top J.-Nom who-Acc like-Adn-QE-Acc know-Pres-DE ‘I know who John likes.’ (73) a. [John-i o-ess-ta-ko](*-ka) mit-eci-n-ta. J.-Nom come-Pst-DE-C-Nom believe-Pass-Pres-DE ‘That John came is believed.’ b. na-nun [John-i o-ess-ta-ko](*-lul) sayngkakha-n-ta. I-Top J.-Nom come-Pst-DE-C-Acc think-Pres-DE ‘I think that John came.’ The question clause in (72a) takes a nominative case morpheme - t o , 4 1 and that in (72b) an accusative case morpheme -lul. In contrast, declarative clauses take neither 4 0 Traditionally, nunci is analyzed as an (embedded) interrogative ending. But I argue that it is further decomposed into an adnominal ending nun and a question ending ci. See the next subsection. 4 1 Adjectival (stative) verbs like kwungkumha ‘to be wondrous’ and calangsulep ‘to be proud’ in Korean take a nominative case marked complement, unlike English adjectival predicates for which of- Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 101 nominative case nor accusative case as shown in (73a) and (73b), respectively. The contrast between (72) and (73) clearly indicates that question clauses are nominal while declarative clauses are non-nominal. 3.6 J. 1.5. Adnominal Conjugation in Question Endings Predicates of a clause that modifies a nominal expression conjugate differently from those of declarative clauses in Korean. Such adnominal conjugation varies depending on the tense,4 2 types of verbs (stative vs. non-stative), and values of the stem final sound (vowel vs. consonant). The following chart gives us a general picture:4 3 insertion fulfills Case-filter. This is why sometimes nominative case in Korean is claimed to be a default case. See Y. Kang (1986) and M. Kang (1988). 4 2 Aspect, instead o f tense, might be a more appropriate term. Consider the following example: (i) i nonmwun-ul ilk-ko [ilk-un kes]-ul yoyakha-ela. this paper-Acc read-and read-Adn thing-Acc summarize-Imp ‘Read this paper and (then) summarize what you will have read. Although the so-called past adnominal ending -un is used, clearly the event denoted by the bracketed part is not about the one fulfilled in the past but about the one to be fulfilled in the future from the present perspective. It simply indicates that the event will takes place before the action of summarizing. 4 3 Adnominal endings for Past II and Past III can be further analyzed, te-n ‘ Retrospective;Aspect- Adn,’ ass/ess-te-n ‘Pst-Retrospective;Aspect-Adn,’ respectively. See Choi (1935:455-463) for some discussion of the retrospective aspect/tense te. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 102 (74) Adnominal Endings in Korean Non-Stative Verbs Stative Verbs [v— Cons] [v...Vow] [v— Cons] [v— Vow] I Present -nun -nun -un -n Past I -un -n None None Past n -ten -ten -ten -ten Past HI -essten -ssten -essten -ssten Future -ul -1 -ul -1 Let us take some examples in relative clauses whose predicate is non-stative and ends with a consonant: (75) a. Present [John-i mek-nun] umsik J.-Nom eat-Adn food (Adn=adnominal ending) ‘the food that John eats/is eating’ b. Past I rJohn-i mek-un] umsik J.-Nom eat-Adn food ‘the food that John ate/has eaten’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 103 c. Future [John-i mek-ul] umsik J.-Nom eat-Adn food ‘the food that John will eat’ With this conjugation in mind, let us turn to question conjugations in Korean. Interestingly enough, exactly the same pattern appears in embedded questions. Consider the following examples: (76) a. Present na-nun [John-i mwues-ul mek-nun-ci] kwungkumha-ta. I-Top J.-Top what-Acc eat-Adn-CI be;wondrous-DE ‘I wonder what John is eating/eats.’ b. Past I na-nun [John-i mwues-ul mek-un-ci] kwungkumha-ta. I-Top J.-Top what-Acc eat-Adn-CI be;wondrous-DE ‘I wonder what John ate/has eaten.’ c. Future na-nun [John-i mwues-ul mek-ul-ci] kwungkumha-ta. I-Top J.-Top what-Acc eat-Adn-CI be;wondrous-DE ‘I wonder what John will eat’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 104 The fact that predicates in question clauses take adnominal endings indicates that question clauses modify a nominal head. I analyze ci as the head noun that is modified by the question clause. In other words, Korean embedded question clauses have the following structure: Nominal WPs can be licensed by being bound by DP ci without movement. 3.63.1.6. Evidence in Conjunction Korean has two different types of conjunction morphemes depending on the categories of conjoined elements. Nominal (and marginally postpositional) phrases are conjoined by kwa/wa,4 4 whereas verbal phrases are by -ko, as the following examples illustrate: 4 4 Conjunctor -kwa follows a consonant, whereas conjunctor -wa follows a vowel. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 1 105 (78) a. nominal conjunctions na-nun [sakwa-wa pay]-lul cohaha-n-ta. I-Top apple-and pear-Acc like-Pres-DE ‘I like apples and pears.’ b. verbal conjunctions na-nun [yangchicilha-ko seyswuha]-ess-ta. I-Top brush-and wash-Pst-DE ‘I brushed my teeth and washed my face.’ Switching the conjunction morphemes makes the sentences ungrammatical. In other words, kwa/wa could not be used in verbal conjunctions and -ko could not be used in nominal conjunctions. Now let us examine how question clauses are conjoined. They are conjoined by the nominal (or postpositional) category conjunctor -(k)wa, but not by the verbal category conjunctor -ko: (79) a. John-un na-eykey [nwu-ka o-ess-nun-ci]-wa J.-Top I-to who-Nom come-Pst-Adn-QE-and [(ku-ka) ettehkey o-ess-nun-ci] mwut-ess-ta. he-Nomhow come-Pst-Adn-QE ask-Pst-DE ‘John asked me who came and how (he) came.’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 106 b. * John-un na-eykey [nwu-ka o-ess-nun-cil-ko J.-Top I-to who-Nom come-Pst-Adn-QE-and [(ku-ka) ettehkey o-ess-nunci] mwut-ess-ta. he-Nomhow come-Pst-QE ask-Pst-DE ‘John asked me who came and how (he) came.’ The fact that questions clauses are conjoined by -(k)wa clearly indicates that question clauses involve a nominal category,4 5 as the proposed structure predicts4 6 4 5 [ put aside the possibility that they are PPs. 4 6 Unlike question clauses, declarative clauses are not conjoined by -kwa/wa, but they are not conjoined by ko, either (0 a. *?Tom-un [John-i i chayk-ul ilk-ess-ta-kol-wa T.-Top [J.-Nom this book-Acc read-Pst-DE-C]-and [Mary-ka ce chayk-ul ilk-ess-ta-ko] malha-ess-ta. M.-Nom that book-Acc read-Pst-DE-C] say-Pst-DE ‘Tom said that John read this book and that Mary read that one.’ b. *Tom-un [John-i i chayk-ul ilk-ess-ta-kol-ko T.-Top [J.-Nom this book-Acc read-Pst-DE-C]-and [Mary-ka ce chayk-ul ilk-ess-ta-ko] malha-ess-ta. M.-Nom that book-Acc read-Pst-DE-C] say-Pst-DE ‘Tom said that John read this book and that Mary read that one.’ The unacceptability of (ia) is expected under our assumption that declarative clauses are not nominal. The unacceptability of (ib) needs an explanation. It may be the case that declarative clauses are neither nominal nor verbal or the so-called subordinate clausal ender -ko may be a kind of a conjunction marker. Note that it has the same form as the conjunctor -ko. Reproduced with permission of the copyright owner. Further reproduction prohibited w ithout permission. 3.6J.2. Nominal Properties of Matrix Questions 107 We have seen in the previous subsections various facts indicating that embedded question clauses in English and Korean (perhaps universally) are of a nominal category. In this section it will be observed that matrix questions also show nominal properties at least in some question forms in Korean, in a dialect in modem Japanese, in premodem Japanese, and in Sinhara (an Indo-Arian language spoken in Sri Lanka). 3.6.3.2.I. Adnominal Endings in Some Matrix Questions in Korean There are at least two cases in which an adnominal ending appears even in some matrix questions in Korean. An adnominal ending appears before a formal plain question particle -ka and in fixture tense questions in general. Let us discuss the two cases in turn. Adnominal endings appear before a formal plain style question ending -ka. although no such endings appear before its corresponding declarative ending -ta. Let us first consider the difference in the conjugation between declarative clauses and question clauses: Reproduced with permission of the copyright owner. Further reproduction prohibited w ithout permission. 108 (80) Plain Formal Endings in Declarative Clauses a. [verb Vowel]-/i-te John-i hakkyo-ey ka-n-ta. J.-Nom school-to go-Pres-DE ‘John goes to school.’ b. [v«b Consonant]-/!!//!-/# John-i sake-Iul mek-nun-ta. J.-Nom apple-Acc eat-Pres-DE ‘John eats apples.’ (81) Question Clauses a. [verb Vowel]-nun-ka John-i hakkyo-ey ka-nun-ka? J.-Nom school-to go-Pres-QE ‘Does John go to school?’ b. [verb Consonant]-/ii//!-A:a John-i sake-lul mek-nun-ka? J.-Nom apple-Acc eat-Pres-QE ‘Does John eat apples?’ Present tense morphemes in -ta declarative clauses vary between -n and -nun, depending on the stem final sound: -n follows a vowel as in (80a), whereas -nun follows a consonant as in (80b). In contrast, verbs invariably take -nun in -ka questions, as exemplified in (81a, b). Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 109 What is the morpheme -nun in question forms? I argue that it is not a variant of present tense forms in declarative clauses but the present adnominal ending for action verbs. Recall from the table (74) in section 3.6.3.1.5., which is repeated below: (74) Adnominal Endings in Korean Non-Stative Verbs Stative Verbs [v— Cons] [v-Vow] [v-Cons] [v— Vow] Present -nun -nun -un -n Past I -un -n None None Past n -ten -ten -ten -ten Past in -essten -ssten -essten -ssten Future -ul -1 -ul -1 The present adnominal form for action verbs is invariably -nun, as highlighted in the table. Conjugation for stative verbs more convincingly shows that adnominal forms are involved in -ka question forms. Notice that stative verbs do not have any overt form for the present in -ta declarative clauses as the following sentences illustrate: Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 110 (82) Declarative Clauses a. [v erb Vowel]-<j>-/a John-i khi-ka khu-ta. J.-Nom height-Nom tall-DE. ‘John is tall.’ b. [verb Consonant]-<j>-te John-i khi-ka cak-ta. J.-Nom height-Nom short-DE. ‘John is short.’ Whether the stative verb ends with a vowel, as in (82a), or with a consonant as in (82b), nothing appears for the present tense in -ta declarative sentences. Now consider the conjugation in -ka question clauses below: (83) -ka Question Clauses a. [verb Vowel]-/i-Ara John-i khi-ka khu-n-fca? J.-Nom height-Nom tall-Adn-QE. is John tall?’ b. [v e rb Consonant]-wn-Ara John-i khi-ka cak-un-kal J.-Nom height-Nom short-Adn-QE. i s John short?’ Reproduced with permission of the copyright owner. Further reproduction prohibited w ithout permission. I l l In -ka question clauses, stative verbs that end with a vowel take -n and those that end with a consonant take -un. Notice from the table in (74) that -n and -un are present adnominal endings for stative verbs. The fact that verbs take an adnominal form in -ka matrix questions indicates that matrix questions in Korean involve some kind of nominal elements just as embedded questions do. Let us now turn to the adnominal endings in future questions. In Korean declarative clauses, present and past events could be expressed by directly putting tense morphemes in the matrix predicates, at least for action verbs. Present events are indicated by putting -n/-nun after the verb stem and past events are indicated by putting -ess/-ass/-ss after the verb stem. Some of the examples are given below: (84) a. Present Tense:-n/-nun John-i o-n-ta. J.-Nom come-Pres-DE ‘John is coming.’ b. Past Tense:-ess/-ass/-sj John-i o-ass-ta. J.-Nom come-Pst-DE ‘John came.’ Present tense is indicated by -n, as in (84a), and past tense by -ass, as in (84b). Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Future events, however, cannot be expressed in this way. Instead, they take a rather complex form given in (85a) and exemplified in (85b): (85) a. [... V-(u)l\ Ares-i-ta4 7 Adn(Fut.) thing-be-DE ‘(It) is the thing that... will V ...’ b. [nayil John-i o-I] kes-i-ta. tomorrow J.-Nom come-Adn thing-be-DE (lit.) ‘(It) is the thing that John will come tomorrow.’ “ John will come tomorrow.’ The content verb takes a future adnominal ending (-(u)f) and the adnominal clause modifies a formal noun kes ‘thing,’ which is incorporated by the copula /. The construction in (85a) is not reserved only for future tense. By changing the adnominal ending in the embedded clause it may convey present and past events as well. For example, the following sentences are fine: (86) a. Present [cikum John-i o-nun] kes-i-ta. now J.-Nom come-Adn(Pres) thing-be-DE (lit.) ‘(It) is the thing that John is coming now.’ ‘John is coming now.’ 4' Copula verb i always incorporates its complement in Korean. See section 5.4.3. of chapter 5. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 113 b. Past [ecey John-i o-n] kes-i-ta. yesterday J.-Nom come-Adn(Pst) thing-be-DE (lit.) ‘(It) is the thing that John came yesterday.’ ‘John came yesterday.’ Thus, future events can be expressed exclusively by the form in (85a), whereas present and past events can be expressed in either of the two ways, as in (84) or as in (86). Now let us examine how question clauses are expressed about future events. Interestingly, future events in the question can be expressed in two ways, either by changing the declarative ending in (85a) into a question ending or by attaching an adnominal ending to the content predicate apparently without taking the complex NP followed by the copula -/. The two types are schematically represented and exemplified in (87) and (88), respectively: (87) a. [... \r-(u)t\ kes-i-ni? Adn(Fut.) thing-be-QE ‘Is (it) the thing that... will V ...’ b. [nayil John-i o-|] kes-i-ni? tomorrow J.-Nom come-Adn thing-be-QE (lit.) ‘Is (it) the thing that John will come tomorrow?’ ‘Will John come tomorrow?’ Reproduced with permission of the copyright owner. Further reproduction prohibited w ithout permission. 114 (88) a. ..M-(u)l-kkafl Adn(Fut)-QE b. John-i o-l-kka? J.-Nom come-Adn-QE ‘Will John come?’ The appearance of future adnominal forms in matrix questions again witnesses that matrix questions are also projections of a nominal element. 3.63.2.2. Evidence in Other Languages Nominal properties of matrix questions are witnessed in other languages as well. Premodem Japanese, Sinhara, and a dialect of modem Japanese are examples. According to Whitman (1995), question clauses in premodem Japanese end with an adnominal {rentaikei) marking. The following example is recited from Whitman (1995), who cites it from an old text. 4 8 QE -kka occurs after ni or /. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (89) [tir-e-ba koso] itodo sakurafa medeta-kere fall-IZ-COND the more cherry-Top wonderfuI-IZ uki.yo-ni [nani ka] fisasi-kar-u bek-i sad.world-in what-QE long-V-RT MOD-RT 'Precisely because they fall cherry blossoms are all the finer, in this woeful world what should be long-lasting?’ (IZ=izenkei (conditional), RT=rentaikei (adnominal)) Based on this pattern, Whitman claims that in premodem Japanese WH-questions ‘scope is marked by an adnominal M[musubi:DC]-marking on the predicate, while the type of the clause (interrogative) is marked by a K[kakari:DC]-marking with -ka on the WH-phrase in situ.’ Whitman further reports that Gair (1983) identifies the same pattern in Sinhara (an India-Aryan language spoken in Sri Lanka). He cites the following sentence from Gair and Sumangala (1991): (90) Siri [mokak dS] kemwe Siri what Q did-E ‘What did Siri do?’ The scope of the question is indicated by M-marking with -e, a type of nominalizing verbal suffix. permission of the copyright owner. Further reproduction prohibited without permission. 116 Without judging the correctness of Whitman’s theory, which analyzes WPs as foci in situ, what is important to us here is that even matrix questions may involve some kind of nominal elements. Some dialect in Modem Japanese (Kochi dialect) shows a phenomenon which indicates that matrix questions involve a nominal property. Although no difference between standard dialect and Kochi is shown in adnominal (relative) clauses and declarative clauses, some difference is attested in question clauses. Consider the following:4 9 (91) In Standard Dialect a. Relatives: [ej sizuka *(na)] hitoj quiet Adn man ‘a quiet person’ b. Declaratives: sizuka (*na) da. quiet Adn BE ‘(It is) quiet.’ 4 9 H. Hoji provided the examples in the text. He, however, points out the possibility that (92c) is derived from (i) by deletion: (i) sizuka na ga ka (ya)? quiet AdnGA QE BE ga in (i) corresponds to the nominal element no in the standard Japanese and kes in Korean. Thus the existence of the adnominal ending na in this context is attributable to the nominal element ga. But a new problem that arises is why such a deletion is not possible for declaratives. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 117 c. Questions: (92) In Kochi Dialect a. Relatives : b. Declaratives: c. Questions: sizuka (*na) ka? quiet Adn QE '(Is it) quiet? [e * sizuka *(na)] hito; quiet Adn man ‘a quiet person’ sizuka (*na) ya. quiet Adn BE ‘(It is) quiet.’ sizuka (na) ka? quiet Adn QE ‘(Is it) quiet? Both in standard dialect and Kochi, stative verbs like sizuka ‘to be quiet’ require an adnominal conjugation na in relative clauses, as in (91a) and (92a), and do not allow it in declarative clauses, as in (91b) and (92b). But na is allowed in Kochi question clauses, whereas it is disallowed in standard dialect. That the Kochi dialect allows the adnominal ending na in matrix questions is another indication that matrix questions can be projections of a nominal element of some sort, at least in this dialect.5 0 5 0 I do not know what causes the typological difference between languages that allow adnominal endings in the matrix questions and those that do not. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 118 3.7. Conclusion In this chapter I have argued the following. First, head licensing and dependent licensing are independently required in grammar, the former being induced by the morphological requirement in functional heads and the latter by the principle of full interpretation. Second, there are two modes of dependent licensing as shown in the asymmetries of nominal WPs in situ and non-nominal WPs in situ: nominal WPs in situ are licensed by means of binding, whereas non-nominal WPs are licensed by means of movement. Third, question clauses are projections of a nominal element. The question clause structures in English type languages and in Korean type languages are given in (93) and (94), respectively: (93) (=(64)) English DP DP CP e SPEC C’ WPj C IP [+wh] ...ti...n o m WPj Binding Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 119 (94) Korean DP CP DP IP C -ci/-ka nomwp Adn Binding The potential binder being a nominal element, nominal WPs are licensed by means of binding, which is more economical than the movement option (cf. Tsai 1994). By contrast, non-nominal WPs are licensed by means of moving to the SPEC of a [+wh] C°, the more economical binding option being unavailable. It was shown that the proposed theory of WP licensing, i.e., the two modes of dependent licensing, was able to cover a wide range of data, including the cases which other theories would fail to account for. It could accommodate the following facts. First, non-nominal WPs cannot remain in situ in English type languages since it cannot be properly licensed at LF, the SPEC of CP being already filled with a WP. Second, two or more non-nominal WPs in situ are not allowed even in Korean type languages since one of the non-nominal WPs would not be properly licensed. Third, non-nominal WPs are disallowed in islands in both types of languages since they necessarily cross a syntactic island in the course of movement to the checking domain Reproduced with permission of the copyright owner. Further reproduction prohibited w ithout permission. 120 of a [+wh] C°. Fourth, nominal WPs are allowed in islands in both types of languages since no movement is required for nominal WPs. One query that has not been discussed here is whether UG allows movements that are not driven by a morphological requirement of heads. Note that in the system proposed here, non-nominal WPs move to the local domain of its scope position. If this is true, how do we restrict movement operation? One possible answer is to say that licensing of WPs does not belong to the computational system of language, rather to a level after that but before the semantic interpretation. I leave this and many other questions open for further studies. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 121 Chapter Four Anti-Superiority Effects (ASEs), Cyclicity, and the Economy 4.1. Introduction This chapter deals with several restrictions imposed on the structural relations among WH-phrases (henceforth WPs) in multiple WH-questions in languages such as Korean and Japanese. Specifically, we will focus on the so-called anti-superiority effects (henceforth ASEs) discussed in Saito (1982, 1989, 1992), A. Watanabe (1991). and S. Watanabe (1995), among others. Let us first describe what the ASE is. As was discussed in the literature mentioned above, the structural relations among WPs decide the grammaticality of multiple WH- constructions in languages as in Korean and Japanese. The core case of the ASE is that when a nominal WP (henceforth n o m WP) and a non-nominal (adverbial) WP (henceforth ^W P) share their interrogative scope, the former can c-command the latter but not vice versa, as schematically represented in (1) and exemplified in Japanese sentences and their Korean counterparts in (2) and (3), respectively:1 (1) a. V ... n o m WP ... “^WP... b. *... “^WP ... n o m WP... (where the elements c-command to the right) 1 The cases with no c-command relation will also be discussed. See section 4 2 .1. and 4.3.5. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 122 (2) (=Watanbabe 1991:18, his (33)) a. ?Kimi-wa nani-o naze katta no? you-Top what-Acc why bought QE ‘Why did you buy what?’ b. *Kimi-wa naze nani-o katta no? you-Top why what-Acc bought QE ‘What did you buy why?’ (3) a. ?ne-nun mwues-ul (ecev) wav sa-ess-ni?2 you-Top what-Acc yesterday why buy-Pst-QE ‘Why did you buy what (yesterday)?’ b. *ne-nun wav (ecey) mwues-ul sa-ess-ni? you-Top why yesterday what-Acc buy-Pst-QE ‘What did you buy why (yesterday)?’ The restriction on the c-command relation between n o m WPs and “^WPs, i.e., the restriction that “‘ ''WPs cannot c-command n o m WPs with the same interrogative scope, is called ASE.3 2 The reason that an adverb ecey ‘yesterday’ is put between WPs is to show that adjacency is not a crucial factor. 3 According to A.Watanabe (1991:17), the ASE should not be confined to the structural relation between "““ WPs and * d v WPs. In fact he takes the contrast in (i) below as the prototypical case of ASE: (i) a. (=A.Watanabe 1991:13, his (19a)) John-wa [Mary-ga nani-o katta [ka dooka]] dare-ni tazuneta no? J.-Top M.-Nom what-Acc bought whether who-Dat asked Q ‘Who did John ask t whether Mary bought what?’ Reproduced with permission of the copyright owner. Further reproduction prohibited w ithout permission. 123 The existence of the ASE raises several interesting questions. First of all. it raises a question about the validity of our analysis of multiple WH-constructions proposed in the previous chapter. It was argued there that0 0 1 1 1 WPs in situ are licensed by having a binding relation (henceforth BR) with a nominal element that question clauses have, whereas ^W P s are licensed by having an exclusive SPEC-head relation (henceforth ESHR, see (41) of chapter 3) with a [+wh] C°. According to the theory, both (a) sentences and (b) sentences in (2) and (3) are predicted to be grammatical. WPs in (2a) and (3a) will be licensed as in (4a) and those in (2b) and (3b) will be licensed as in (4b) below: b. (=A. Watanabe 1991:14, his (2la)) ??John-wa dare-ni [Mary-ga nani-o katta [ka dooka]] tazuneta no? J.-Top who-Dat M.-Nom what-Acc bought whether asked Q ‘Who did John ask t whether Maty bought what?’ In (i) the structural relation between two “ "WPs matters, i.e., the WP external to the WH-island, dare- ni, should not c-command the island internal one, nani-o. This will not be discussed here for the reason that the claim that WPs may have an interrogative scope outside a WH-island is controversial, (cf. footnote 15 in chapter 3) Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 124 (4) a. DP DP CP b. DP DP CP [+wh] IP ESHR VP BR Move-a [+wh] D P ESHR VP n o m .WP... Move-a BR The n o m WPs in both structures are licensed in the surface position by having a BR with the nominal element that question clauses have (represented as a DP). The “^WPs are supposed to be licensed by having an ESHR with the question morpheme. Notice that each sentence contains only one ^W P. Both "“"WPs and ^W P s being licensed, all the sentences in (2) and (3) are expected to be good, contrary to the fact. Another fundamental question that arises from the ASE is why the so-called anti- superiority effects as opposed to superiority effects exist at all in the grammar of languages such as Japanese and Korean. Notice that superiority effects have been taken as a typical phenomenon attested in natural languages and have been dealt with in terms of the economy theory. It is quite unexpected that languages like Japanese and Korean have a restriction like the ASE on the combination of5 0 1 ,1 WPs and “ ^WPs. Reproduced with permission of the copyright owner. Further reproduction prohibited w ithout permission. In this chapter, I try to answer these questions in terms of the economy principle with some assumptions. First, I assume that there are two licensing modes for ^ WPs: one by having an ESHR with a [+wh] C° along the lines discussed in the previous chapter and the other by having a proper structural relation, which I call a quasi binding relation (QBR), with the nominal element of question clauses. A QBR obtains when an “^WP is properly adjoined to a n o m WP (Saito 1992) and is parasitically licensed when the "“"WP is licensed by a BR. Thus the WPs in (la) will be licensed as follows: (5) DP DP CP [+wh] IP VP "WP ,.ti. a d v n m nom "W p QRR Mnvp BR WP kx Now compare the licensing mode of this and the one in (4a). The n o m WPs in both structures are licensed in the same way, i.e., by a BR. The difference lies in the licensing of WPs. The “^WP in (4a) is licensed by an ESHR, whereas the one in Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 126 (5) is licensed by a QBR. Obviously the licensing by a QBR in (5a) has a shorter derivation than the one by an ESHR in (4a) since the distance covered by Move-a in the former is shorter than that in the latter. Thus I claim that for the economy reason, the option of licensing by a QBR is chosen over the option of licensing by an ESHR when both of the licensing modes are potentially available. Now let us turn to the illicitness of the structure in (lb) or the ungrammaticality of the sentences such as (2b) and (3b). Apparently, the option of licensing by a QBR is not available in this structure since the ^W P is higher than the n o m WP and therefore the former cannot be adjoined to the latter. The question to be asked, then, is why the structure in (lb) is not allowed in the grammar of Japanese and Korean. I claim that such a structure is illicit because it misses a less costly derivation with respect to the licensing of WPs under a certain assumption on scrambling. The assumption that I make here is that clause internal scrambling in languages like Japanese and Korean is literally optional and furthermore costless, basically following Fukui’s (1993) hypothesis on the directionality and costliness of movement This being adopted, the structure in (lb) should be excluded since the n o m WP has not scrambled over the ^W P. If the scrambling occurred, the resulting structure would be identical to the one in (la) with respect to the relative hierarchy of the two WPs and feed a less costly derivation with respect to the licensing of^W P; the ^W P could be licensed by a QBR just as in (la) without having to move all the way to the SPEC of a [+wh] C° in order to have an ESHR with it. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 127 The organization of the chapter is as follows. In section 4.2., I will critically review some previous analyses of the ASE and build up more facts surrounding the ASE for the sake of later discussion. Among the analyses entertained in the literature. A. Watanabe (1991), Saito (1992), and S. Watanabe (1995) will be discussed. In section 4.3., the main section of this chapter, I will provide an economy-based analysis as an alternative analysis of the ASE. Section 4.4. concludes the chapter. 4.2. Previous Analyses 4.2.1. A. Watanabe (1991) 4.2.1.1. Generalization A. Watanabe (1991:84) provides the following generalization or well-formedness condition on the distribution of multiple WH-construction in Japanese: (5) A multiple question is well-formed in Japanese only if at S-structure there is a WH-phrase which is not c-commanded by the WH-phrase from which the pure wh is already moved into the SPEC of CP. To comprehend the condition in (5), which is not described theory-neutrally, it is necessary to make clear several points about his theory of (multiple) WH-questions. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 128 First by the pure wh, A. Watanabe refers to an empty question operator (henceforth OP ) which he assumes is generated with WH-phrases and moves to the SPEC of CP in overt syntax on a par with WH-movement in languages like English. With the postulation of the null operator movement in a language like Japanese, he claims that WH-questions are universally licensed in overt syntax, either by raising an overt WP or by raising an OP. A second point is that, as will be discussed in the next subsection, whenever an “‘ ''WP is involved, the OP associated with the a d v WP must move to the SPEC of a [+wh] C°. Otherwise an ECP violation will result. Another point to be made is that the well- formedness condition in (5) is a licensing condition on multiple WH-questions, not on the individual WPs. Therefore, a multiple question is well-formed as long as the relation between a pair of WPs satisfies the well-formedness condition, irrespective of the relation among all other WPs in the sentence. Let us consider what the well-formedness condition in (5) will say about the following three structures where a n o m WP and an a d v WP are involved: Since an ^W P is involved in all three structures, the OP associated with it must be extracted because of the ECP factor. In (a) and (c), there is a WP, i.e., n o m WP, which is Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 129 not c-commanded by an “^WP from which OP is extracted. In contrast, in (b), there is no WP that is not c-commanded by an ^W P. Thus (a) and (c) are well-formed, whereas lb) is ill-formed. Mostly the well-formedness condition in (5) makes a correct prediction.4 The contrast between (6a) and (6b) was exemplified in (2) and (3). Structure (6c) is also allowed. Compare the following two sentences: (7) a. (=A. Watanabe 1991:93, his (198b)) dare-ga [cpJohn-ga naze kubininatta to] itta no? who-Nom J.-Nom why was;fired C said QE 'Who said [that John was fired why]?’ b. (= A. Watanabe 1991:93, his (198a)) (?)[cpJohn-ga naze kubininatta to] dare-ga itta no? J.-Nom why be;fired C who-Nom saidQE ’Who said [that John was fired why]?’ The sentence in (7a), whose structure conforms to the structure (6a), is well-formed. The sentence in (7b) is derived from (7a) by scrambling the complement clause of the bridge verb itta ‘said’ to the sentence initial position. (7b), whose schematic structure conforms to (6c), is also well-formed. The ^W P in (7b) precedes the n o m WP but does not c-command it. 4 The generalization is, however, disputed by Saito (1992). See section 4.2.2. Reproduced with permission of the copyright owner. Further reproduction prohibited w ithout permission. 4.2.1.2. A. Watanabe’s (1991) Explanation of the ASE 130 A. Watanabe (1991) ascribes the ASEs to the conspiracy of the pure question operator hypothesis and the following principles or conditions: the Relation Preservation, the Condition on a Well-Formed WH-phrase at LF, and the ECP. The Relation Preservation and the Condition on a Well-Formed WH-Phrase at LF are defined as follows:5 (8) Relation Preservation (A.Watanabel991:l 11. his (215)1 A relation established at a certain point in the derivation must be maintained throughout (9) Condition on a Well-Formed WH-Phrase at LF (A. Watanabe 1991:104, his (221)) A pure WH-operator and the associate indeterminate phrase alone must form a category in order to function as a WH-phrase. The relation that is in issue in (8) is the hierarchy among WPs. The hierarchy is defined in terms of seg(ment)-command as follows: (10) See-command (=A. Watanabe 1991:99, his (208)) a seg(ment)-commands P iff a does not dominate P and every segment that dominates a dominates P where a and P are categories. 5 A. Watanabe ( 1991:18) follows the version o f ECP proposed in Chomsky (1981). Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 131 According to the Relation Preservation in (8), a seg-command relation between two WPs established at SS must be preserved at LF. For example, if WP1 is seg- commanded by, i.e., structurally lower than, WP2 at SS, then WP1 adjoins to the OP in the SPEC of CP earlier than WP2 at LF. According to the Condition on the Well- Formed WH-Phrase at LF in (9), when an OP is extracted from a WP. the WP must move and adjoin to the OP to form a category with it before any other WP does. Let us examine how A. Watanabe explains the contrast of the sentences in (2) and (3) with these conditions and principles. The sentences are repeated below: (2) (Saito 1982=A. Watanabe 1991: 20, his (30)) a. ?Kimi-wa nani-o naze katta no? you-Top what-Acc why bought Q ‘Why did you buy what?’ b. *Kimi-wa naze nani-o katta no? you-Top why what-Acc bought Q ‘What did you buy why?’ (3) a. ne-nun mwues-ul (ecey) wav sa-ess-ni? you-Top what-Acc yesterday why bought Q ‘Why did you buy what (yesterday)?’ b. * ne-nun wav (ecey) mwues-ul sa-ess-ni? you-Top why yesterday what-Acc bought Q ‘Why did you buy what (yesterday)?’ Reproduced with permission of the copyright owner. Further reproduction prohibited w ithout permission. Relevant schematic SS structures for (a) sentences would be like (11a) and those for (b) sentences would be like (1 lb): ...,dvWP... ... n o m WP... A. Watanabe attributes the contrast between (11a) and (11b) to the question as to whether or not the seg-command relation at SS is preserved at LF. At SS, “ WP seg- commands “^WP in (1 la) and the relation in (1 lb) is the opposite. The LF structure derived from (11) will be one of the structures in (12) depending on the source of the O P*, which has moved at SS, and on the order of movement of WPs at LF: (12) a. b. CP OPi ^ X OPj X X nom WPj OPj ^ W P j OPj ^W P j OPj "“"WPj OPj c. ^ C P ^ d. CP OPj ^ X OPj X X nom WPj OPj ^ W P j OPj ^ W P j OPj no m WPj OPj Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 133 In (a) and (b) OPj in the SPEC of CP is extracted from the ^ W P , whereas in (c) and (d) it is extracted from the n o m WP. In (a) and (c), the “^WP moves first, whereas in (b) and (d), the n o m WP moves first In (11a) the " ‘ ""WP seg-commands the “^WP. Thus the well-formed LF structure will be (12a) or (12c), in which the n o m WP seg- commands the ^W P. Due to the condition on a Well-Formed WH-Phrase at LF given in (9), (12a), but not (12c), is a licit LF structure for (11a). Note that in (12a), but not in (12c), the OP and the ^ W P from which the OP is extracted form a category. Now let us consider the LF structure for (1 lb), in which the a d v WP seg-commands the n o m WP at SS. Structures in (12b) and (12d) are initial candidates. Both structures are, however, illegitimate. (12b) violates the condition on a Well-Formed WH-Phrase at LF given in (9), since the n o m WP and the OP which has been extracted from the “^WP do not form a category. In addition, (12b) violates the ECP. The ^W P would not properly antecedent govern its trace due to the Comp indexing mechanism (Aoun et al. 1981), according to which the index of the SPEC of CP would be that of the "“"WP, i.e., index i.6 (12d) is fine with respect to the condition on a Well-Formed WH-Phrase at LF given in (9) because the n o m WP and the OP which is extracted from the n o m WP 6 A. Watanabe (1991:106, his (229)) defines antecedent government as follows: (i) a antecedent governs p iff i) a and p are coindexed ii) a seg-commands P iii) no barrier intervenes. In his system, the OP must be extracted from an * d v WP, if there is any, for the ECP reason. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 134 form a category. (12d), however, violates the ECP just as (12b) does. In conclusion, a legitimate LF structure, i.e., (12a), can be derived from (11a), but no legitimate LF structure is derived from (1 lb), which explains the difference between (1 la) and (1 lb). Let us finish this subsection by discussing how A. Watanabe (1991) deals with the fact that both sentences in (7) are grammatical. Consider the sentences in (7), which are repeated below: (7) a. (=A. Watanabe 1991:93, his (198b)) dare-ga [cpJohn-ga naze kubininatta to] itta no? who-Nom J.-Nomwhy was;fired C saidQE 'Who said [that was fired why]?’ b. (= A. Watanabe 1991:93, his (198a)) (?) [cpJohn-ga naze kubininatta to] dare-ga itta no? J.-Nom why be;fired C who-Nom saidQE ‘Who said [that John was fired why]?’ The grammaticality of (7a) is explained on a par with the grammaticality of (2a) and (3a). That is, an OP is extracted from the a d v WP at SS. In LF the ^W P first moves, forming a category with the OP which has already moved to the SPEC of CP, and then the n o m WP movement follows, whereby it seg-commands the a d v WP, satisfying the Relation Preservation in (8). In (7b), no relation between the ^W P and the n o m WP has been established at SS since no seg-command relation between them exists at all. Thus Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 135 the Relation Preservation is vacuously satisfied, accounting for the grammaticality of (7b). 42.13. Problems Although quite successful in explaining the data with respect to the ASEs and other wide range of data, A. Watanabe’s (1991) account bears several non-trivial problems. In this section, I will discuss some conceptual and empirical problems that his system faces. 4.2.1 J .l. Conceptual Problems Let us first discuss conceptual problems. In addition to his resort to the ECP and syntactic levels, theoretical artifacts by now, A. Watanabe faces at least the following problems. First, suppose that there are two WPs in a multiple WH-question, WP1 and WP2, and WP1 c-commands WP2, as the following schematic structure illustrates: Reproduced with permission of the copyright owner. Further reproduction prohibited w ithout permission. 136 (13) CP [+wh] IP XP 0P-WP1 YP OP-WP2 ZP Under the circumstance in which either WPl or WP2 is able to bear an OP. Attract-a would attract the OP associated with WPl since that will license the question force with a shorter derivation than the case where the question operator is extracted from WP2. But in A. Watanabe’s system, the OP associated with WP2 moves at SS. Subsequently WP2 first moves at LF to satisfy the Relation Preservation in (8) and the Condition on a Well-Formed WH-Phrase at LF in (9), violating the economy principle (superiority).' A second problem is that the OP, which A. Watanabe (1991:49) assumes to be generated in a DP-SPEC position, is also generated in “^WPs like naze ‘why’ in Japanese and way ‘why’ in Korean which do not seem to be a DP. Furthermore, to explain the absence of the complex noun phrase island effect for n o m WPs, A. Watanabe (1991:61) claims that an OP starts from the SPEC of the DP which contains both the 7 The Condition on a Well-Formed WH-Phrase at LF in (9) could not be resorted to if OP is generated independently of WPs. See Aoun & Li (1993a, b) and Tsai (1994) for the argument that OP in Chinese is generated independently of WPs. Nevertheless, Chinese shows the same order restriction. (A. Li (p.c.)). In this work I do not endorse the OP movement approach to account for the WH-raising vs. WH-in-situ asymmetry. Rather I argued in chapter 2 that the asymmetry comes from the difference in the licensing mode of the interrogative force, i.e.. raising of a WP vs. raising of a question conjugated predicate. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 137 relative clause and the head noun along the lines of the pied-piping analysis.8 But this way of explanation of the lack of the island effect does not fare well since n o m WPs can occur inside an adjunct clause, as the following examples illustrate: (14) a. [John-i nwukwu-lul ttavli-ese] Mary-ka hwana-ess-ni? J.-Nom who-Acc hit-because M.-Nom get;angry-Pst-QE ‘Whoj did Mary get angry because John beat tj?’ b. [pro mwues-ul nwulu-myen] mwun-i yeli-ni? what-Acc push-if door-Nom open-QE ‘ What; does the door open if you push ej? Since the clauses headed by -ese ‘because’ or by -myen ‘if are clearly not nominal, the OP movement directly from the WPs is supposed to show a Subjacency effect according to the theory A. Watanabe proposes. However, n o m WPs within these clauses are perfectly allowed. Third, although the Relation Preservation demands that a relation established at a certain point of derivation be maintained throughout, the derivation from DS to SS often does not preserve a certain relation. For example, scrambling shifts the relation established earlier. A. Watanabe, in fact, considers two points of derivations, i.e., at SS and at LF. * The discrepancy between the nominal type and adverbial type with respect to the island effects is attributed to the existence and the lack of the large scale pied-piping in A. Watanabe (1991). Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 138 Fourth, the typological difference in forming WH-words that A. Watanabe (1991:52-54) supposes does not seem to be based on firm ground. He ascribes the availability of the pure question operator in Japanese but the lack of it in English to the morphological processes in forming WH-words in English and those in Japanese, although he assumes along the lines of Chomsky (1964) and Katz and Postal (1964) that the base structures are the same in these two types of languages. Being merged into the question words, i.e., into the indefinite part, the question operator in English is inoperable in the syntax. In contrast, the pure question operator in a language like Japanese is not merged into the indeterminates and the pure question operator may be syntactically active and may undergo movement The fact that indeterminates may have the readings of existential or universal quantifiers as well as the interrogative reading depending on the elements that they are associated with is taken as a piece of evidence that the associated elements are indefinites whose quantificational force is determined by the elements that are associated with them. This has been taken for granted since Kuroda(1965).9 It is not clear, however, whether the so-called indeterminates are really pure indefinites void of an interrogative force. Let us reconsider the following morphological paradigm given in A. Watanabe (1991: 51, his (109)): 9 A similar view has prevailed in Korean linguistics since Choi (1935). Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 139 (15) WH-phrases Existential Quantifiers dare-ka ‘someone’ nani-ka ‘something’ doko-ka ‘somewhere’ itsu-ka ‘sometime’ naze-ka ‘for some reason dono N-ka ‘some N’ dare ‘who’ nani ‘what’ doko ‘where’ itsu ‘when’ naze ‘why’ dono N ‘which N Universal Quantifiers Negative Polarity Items dare-mo ‘anyone’ nani-mo ‘anything’ dare-mo ‘everyone’ * nani-mo ‘everything’ doko-mo ‘everywhere’ itsu-mo ‘whenever’ *itsu-mo -------- *naze-mo -------- dono N-mo ‘any N’ doko-mo ‘anywhere’ *naze-mo -------- dono N mo ‘every N’ Although bare WH-words are considered as having interrogative readings themselves in Japanese, the facts do not reveal this. In order for a WH-word to have an interrogative reading, it should be related to a question particle like -to. Notice also that the so-called existential quantifier licenser is identical to the question particle. The identity does not seem to be a coincidence. It may well be the case that the so-called existential licenser - to is in fact a question ending.1 0 This becomes clearer when we consider the morphological paradigm in Korean given below:1 1 1 0 Kuroda (1964) also derives -to as an existential quantifier licenser (and -to as a disjunctive morpheme) from the same source of -to as a question licenser, although he does not specify exactly what it is. " Korean counterparts of Japanese WP-mo forms in Japanese have a universal (more precisely free choice) reading but not an NPI reading. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 140 (16) WH-phrases Existential Quantifiers1 2 nwukwu ‘who’ nw ukwu-<{>-nka ‘someone’ mwues ‘what’ mwues-i-nka ‘something’ eti ‘where’ eti-<j»-nka ‘somewhere’ encey ‘when’ encey-<{>-nka ‘sometime’ way ‘why’ *way-(i>nka ‘for some reason’ enuN ‘which N’ enu N-i-nka ‘some N’ elma ‘which amount’ elm a-<j>-nka ‘some amount’ Universal Quantifiers nw ukwu-<|>-na/-(j>-tunci/-<J)-la-to/-to ‘everyone’ mwues-i-na/-i-tunci/-i-la-to/* -to ‘everything’ eti-<j)-na/-<J)-tunci/-<J>-la-to/*-to ‘everywhere’ encey-<J>-na/-<j>-tunci/-(t>-la-to/*-to ‘whenever’ way *-(i)-na/-(i)-tunci/*-(i)-Ia-to/*-to----------- enu N-i-na/-i-tunci/-i-Ia-to/-to ‘every N’ elma *-< |>-na/-<j>-tiinci/-<|>-la-to/*-to ‘whichever amount’1 3 The so-called existential quantifier licenser, (i)nka, involves a question ending -nka. Similarly the so-called universal quantifier endings, (i)na/(i)tunci/(i)lato/to, etc.. involve a question ending like -na or -ci or concessive endings like -to.1 4 Notice that WH- clauses, even in languages like English, may be complements of an element that introduces a concessive clause: (17) a. No matter what you do, you should do your best 1 2 The vowel i o f the existential licenser and universal licensers which is deleted after a vowel is analyzed as a copula verb. See chapter 5. 1 3 Elma-^-na is acceptable when it means ‘approximately which amount.’ 1 4 In fact all the universal licensing endings including -i-na and -i-tunci. as well as -i-la-to or ~{e)to may be used as concessive endings. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 141 b. Irrespective of whether John comes or not, we have to leave now. The fact that the so-called existential quantifiers based on WPs contain a question particle, and the so-called universal quantifiers also involve a question particle or a concessive particle in Korean, indicates that the so-called indeterminates in Korean are in fact interrogative phrases.1 5 There is another indication that WPs in Japanese and Korean and those in English are not different in their morphological structure. Specifically, WH-words or indeterminates in Japanese or Korean do not seem to be simple indefinite nominal expressions. As was seen in (15), most of the indeterminates in Japanese start with n (e.g., nani or naze ) or d (e.g., dare, doko or dono). Hoji (1995) maintains that da or do in the question words like dare, doko, and dono has a paradigmatic relation with other demonstrative-like elements in Japanese, i.e., ko ‘this,’ so ‘that,’ and a 'that over there.’ In other words, Japanese has the following paradigm: (18) a, dare ‘who’ kore ‘this thing’ sore ‘that thing’ are ‘that thing over there’ b. doko ‘which place’ ‘where’ koko ‘this place’ soko ‘that place’ ako ‘that place over there’ 1 5 See the discussion o f chapter 5, where I analyze the so-called existential and universal quantifiers in Korean as reduced indirect questions. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 142 c. dono N ‘which N’ kono N ‘this N’ sono N ‘that N’ ano N ‘that N over there’ If d in the indeterminates that start with d is analyzed as a demonstrative or its proximate, such indeterminates in Japanese cannot be simple indefinites but must have complex structures just like English WH-words. Hoji (1995) further note® , that n in nani and naze is historically related to d. Once Japanese indeterminates are determined to be as complex as English WH-words, there would seem to be little motivation for the pure WH-operator hypothesis which was based on the assumption that the indeterminates in Japanese are simple indefinites.1 6 A close examination of morphological structures of Korean WH-words further confirms that WH-words in Korean do not differ much from their English counterparts. Consider the WH-words in Korean in (16) again, which are repeated below: 1 6 Hoji (1995), nevertheless, claims that nominal structures in Japanese are headed by N, not by D. based on the assumption that Japanese is a strictly head-final language. Thus, he posits the following structure for dare ‘who,’ for example: (i) NP DemP N I I da re In this analysis, DemP (demonstrative phrase) is just a modifier of N. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 143 (16) WH-phrases nwukwu ‘who’ mwues ‘what’ eti ‘where’ encey ‘when’ ettehkey ‘how’ way ‘why’ enuN ‘which N’ ettehanN ‘which N’ elma ‘which amount’ First notice that most of WH-words start with e- which indicates that WH-words in Korean can be further analyzed to be as complex as WH-words in English. Korean also displays the paradigmatic relation observed in Japanese at least for some WH-words: (19) a. etteh(a)-key ‘how’ ileh(a)-key ‘this way’ celeh(a)-key ‘that way’ kuleh(a)-key ‘that way’ b. encey ‘when’ icey ‘this time or now’ cecey ‘that time’ kucey ‘that time or the day before yesterday’ cf. ecey ‘yesterday’ c. ette(ha)n ‘which’ ile(ha)n ‘this kind o f cele(ha)n ‘that kind o f kule(ha)n ‘that kind o f d. mwues ‘what’ ikes ‘this (thing)’ cekes ‘that thing over there’ kukes ‘that thing’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 144 Just as Japanese has ko/so/a/do patterns, Korean seems to have i/ku/ce/e patterns.1 7 If the analysis of morphological structures for the WH-words is correct then the parametric difference assumed by A. Watanabe with respect to the pure WH-operator loses ground to be based on. A fifth question on the validity of the pure WH-operator could be found in relation to acquisition. According to A. Watanabe’s system, WPs with an interrogative reading are more complex than just indefinites under the assumption that a pure WH-operator is merged with an indeterminate, forming an interrogative phrase, whereas indeterminates have an indefinite reading when no other element is added. But WPs with an indefinite reading are reported to be learned in a far later stage of language acquisition. (Young- Joo Kim (p.c.)) 4.2.13.2. Empirical Problems Let us now turn to empirical facts that cannot be captured by A. Watanabe's (1991) condition in (5), which is repeated below: (5) A multiple question is well-formed in Japanese only if at S-structure there is a WH-phrase which is not c-commanded by the WH-phrase from which the pure wh is already moved into the target SPEC of CP. 1 7 WH-words, nwukwu ‘who,’ mwues ‘w hat’ and way ‘why’ do not start with e, though. Reproduced with permission of the copyright owner. Further reproduction prohibited w ithout permission. 145 First, according to A. Watanabe (1991), the ASE is suspended when a “ "WP is added in a higher place since the Relation Preservation in (8) is a licensing condition, i.e.. a structure is licit as long as just one pair of WPs satisfies the condition. Thus the following contrast is expected: (20) a. * ... ^W P ... n o m W P... b. V... n o m WP ... ,dvWP ... n o m WP ... The contrast certainly exists. For example, when the subject is replaced by a n o m WP in (2b) and (3b), these sentences become grammatical, as shown below: (21) a. (=(2b)) *Kimi-wa naze nani-o katta no? you-Top why what-Acc bought QE ‘What did you buy why?’ b. dare-ga naze nani-o katta no? who-Nom why what-Acc bought QE ‘Who bought what why?’ (22) a. (=(3b)) *ne-nun wav (ecey) mwues-ul sa-ess-ni? you-Top why yesterday what-Acc buy-Pst-QE ‘What did you buy why (yesterday)?’ Reproduced with permission of the copyright owner. Further reproduction prohibited w ithout permission. 146 b. nwu-ka wav (ecey) mwues-ul sa-ess-ni? who-Nom why yesterday what-Acc buy-Pst-QE ‘Who bought what why (yesterday)?’ The sentences in (b) are grammatical as opposed to the ones in (a). This is expected in his theory, as the following schematic structure illustrates: (23) CP O P 2 C’ darei OP2 IP nani3 OP2 ti I’ naze2 OP2 VP t2 VP t3 VP For the ECP reason, the OP should be extracted from the a d v WP, naze. The a d v WP should first move to form a category with the OP due to the Condition on the Well- Formed WH-Phrase at LF in (9). One of the n o m WPs will be adjoined to the category formed on the OP and naze and then the other n o m WP will be adjoined to the adjunction structure. (In (23), it is assumed that movement of nani precedes that of Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 147 dare, but their order is immaterial for this particular structure.) The seg-command relation at SS among the three WPs is that dare seg-commands nani, which in turn seg-commands naze. After WPs are raised at LF. naze is seg-commanded by dare and by nani. Thus the seg-command relation between naze and nani is not preserved, whereas the relation between naze and dare is preserved. The Relation Preservation Condition, being a licensing condition, is satisfied, correctly predicting the grammaticality of the sentences in (21b) and (22b). There is, however, a locality effect with respect to the suspension of the ASE. as Saito (1992) observes. Although the ASE is remedied by adding a n o m WP in a higher position (A. Watanabe 1991), it cannot be if the added WP is in a higher clause, as schematically illustrated in (24) below: (24) a. V... n o m WP ... ^W P ... " “"WP ..., where no clause boundary involved, b. ?*... "“"WP ... [Cp ...a d v W P... "°m W P... The contrast seems real. Consider the following pairs of sentences:1 8 (25) a. ?*Tom-wa [John-ga naze nani-o katta to] omotteru no? T.-Top J.-Nom why what-Acc bought C think QE ‘Q Tom thinks [that John bought what why]’ 1 8 It is not uncontroversial on the judgment on the sentences like (25b) and (26b). S. Watanabe (1995:66) accepts sentences like (25), but not as readily as the sentences in (21b). See the discussion in section 4.2.3. for S. Watanabe’s analysis of the ASE. Reproduced with permission of the copyright owner. Further reproduction prohibited w ithout permission. 148 b. (=Saito 1992:25, his (60c)) ?* dare-ga [John-ga naze nani-o katta to] omotteru no? who-Nom J.-Nom why what-Acc bought C think QE ‘Q who thinks [that John bought what why]’ (26) a. ?*ne-nun [John-i wav mwues-ul sa-ess-ta-ko] sayngkakha-ni? you-Top J.-Nom why what-Acc buy-Pst-DE-C think-QE ‘Q you think [that John bought what why]’ b. ?*nwu-ka [John-i wav mwues-ul sa-ess-ta-ko] sayngkakha-ni? who-Nom J.-Nom why what-Acc buy-Pst-DE-C think-QE ‘Q who thinks [that John bought what why]’ Even if the matrix subject in the sentences in (a) is replaced by a n o m WP as in (b), no amelioration seems to be attested in these sentences, as opposed to the contrast between the sentences in (a) and those in (b) in (21) and in (22). According to the theory proposed by A. Watanabe (1991), the sentences in (25b) and (26b) should be as equally acceptable as those in (21b) and (22b), since the seg- command relation among the WPs in (25b) and (26b) is identical to the one in (21b) and (22b). Note that the clause boundary in (25b) and (26b) is immaterial to the seg- command relation among WPs. Thus, A. Watanabe’s theory incorrectly predicts these sentences to be grammatical as well as those in (21b) and (22b). Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 149 A second empirical problem that A- Watanabe (1991) faces is Nishigauchi's (1986) observation that the island effect for ^W Ps becomes less serious when another n o m WP occurs in the island. Consider the following sentences:1 9 (27) a. (=Nishigauchi 1986:119, his (24)) * [ n p [ s kare-ga naze kai-ta] hon]-ga omosiroi-desu-ka? he-Nomwhy wrote book-Nom interesting-be-QE ‘Books that he wrote why are interesting?’ b. (=Nishigauchi 1986:125, his (32)) ? ( ? ) [ n p [ s dare-ga naze kai-ta] hon]-ga omosiroi-desu-ka? who-Nom why wrote book-Nom interesting-be-QE ‘Books that who wrote why are interesting?’ (28) a. * [ n p [ s ku-ka wav ssu-n] chayk]-i caymiiss-ni? he-Nom why write-Adn book-Nom interesting-be-QE ‘Books that he wrote why are interesting?’ b. ?(?) [np [ s nwu-ka way ssu-n] chayk]-i caymiiss-ni? who-Nom why write-Adn book-Nom interesting-be-QE ‘Books that who wrote why are interesting?’ A. Watanabe (1991) predicts both sentences in (27) and (28) to be ungrammatical since the OP which is detached from the ^W P nazeAvay for the ECP reason moves across 1 9 The nominal/non-nominal asymmetry with respect to island sensitivity is attributed in A. Watanabe ( 1991) to the availability o f the large scale pied-piping for “ “ WPs and the non-availability of it for “’’WPs along the lines of Nishigauchi (1986). Reproduced with permission of the copyright owner. Further reproduction prohibited w ithout permission. 150 the island, violating the Subjacency at SS or the LF trace of naze/way would violate the ECP. Third. A. Watanabe (1991) does not expect the following contrast: (29) a. *Mary-wa naze [John-ga hon-o naze katta to] omotteru no? M.-Top why J.-Nom book-Acc why bought C think QE ‘"'Why does Mary think that John bought books why?’ b. dare-ga naze [John-ga nani-o naze katta to] omotteru no? who-Nom why J.-Nom what-Acc why bought C think QE ‘Who thinks [that John bought what why] why?’ (30) a. *Mary-ka wav [John-i hon-o wav sa-ess-ta-ko] sayngkakha-ni? M.-Nom why J.-Nom book-Acc why buy-Pst-DE-C think QE ‘*Why does Mary think that John bought books why?’ b. nwu-ka wav [John-i mwues-ul wav sa-ess-ta-ko] sayngkakha-ni? who-Nom why J.-Nom what-Acc why buy-Pst-DE-C think QE ‘Who thinks [that John bought what why] why?’ All the sentences in (29) and (30) contain two “ ^WPs, one in the matrix clause and the other in the embedded clause. The only difference between the sentences in (a) and those in (b) is that an additional n o m WP is present both in the matrix and the embedded clause in the latter sentences. As was discussed in section 3.3.2. of chapter 3, two or more ^W Ps generally cannot be linked to the same scope element (Aoun 1986). A. Watanabe could expect the ungrammaticality of Japanese sentences such as (29a) and Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 151 Korean sentences such as (30a) since one of the two “‘ ''WPs would violate the ECP. Recall that “‘ ''WPs are antecedent governed by a “Comp-indexed” WP in SPEC of CP. But his theory incorrectly predicts that the Japanese sentences in (29b) and the Korean sentences in (30b) to be ungrammatical for the same reason as (a) sentences are. But the sentences in (29b) and (30b) are, in fact, grammatical. In conclusion, the generalization on the multiple WH-construction made by A. Watanabe (1991), i.e., (5), misses the cyclicity effects observed by Saito (1992), the weakened (complex NP) island effects observed by Nishigauchi (1986), and the exemption from the restriction on multiple occurrence of “‘''WPs. A. Watanabe‘s failure to account for the empirical data discussed in this section seems to be caused by his overlooking the possibility that a d v WPs are licensed not only by moving to the SPEC of a [+wh] C°, but also by having some structural relation with a n o m WP. 4.2.2. Saito (1992) 4.2.2.1. An Additional WP Effect and its Cyclicity Saito (1992) makes the following two factual claims in relation to the structural relation that “‘ ''WPs and n o m WPs must have in multiple WH-constructions. First, a n o m WP saves an otherwise illicit “‘ ''WP only when the n o m WP is in the same clause as the “‘ ''WP is. As mentioned in section 4.2.1.3., Saito observes that the illicitness of the Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 152 order ^W P ... n o in WP, where the former c-commands the latter, could be obviated by another n o m WP in the same clause, but not by a n o m WP in a higher clause. Consider the following sentences: (31) a. (=Saito1992:25, his (60c)) ?* dare-ga [John-ga naze nani-o katta to] omotteru no who-Nom J.-Nom why what-Acc bought C think QE ‘Q who thinks [that John bought what why]’ b. *John-ga [Bill-ga naze nani-o tabeta to] dare-ni itta no? J.-Nom B.-Nomwhy what-Acc ate C who-Dat saidQE ‘To whom did John say [that Bill ate what why]?’ The bracketed parts in (31) are illicit. The illicitness is not cured by adding a n o m WP in a higher clause, whether the added n o m WP c-commands the “ ^WP as in (31a) or not as in (31b). Another factual claim that Saito (1992) makes is that Nishigauchi’s (1986) observation with respect to the weakened island effect for “ ‘ ‘ ''WPs discussed in 4.2.1.3 should be qualified. Recall that the complex NP island effect for a d v WPs is drastically weakened when a n o m WP exists in the island, as exemplified in (27), repeated below: Reproduced with permission of the copyright owner. Further reproduction prohibited w ithout permission. 153 (27) a. (=Nishigauchi 1986:119, his (24)) * [ n p [ s kare-ga naze kai-ta] hon]-ga omosiroi-desu-ka? he-Nom why wrote book-Nom interesting-be-QE 'Books that he wrote why are interesting?’ b. (=Nishigauchi 1986:125, his (32)) ? ( ? ) [n p [ s dare-ga naze kai-ta] hon]-ga omosiroi-desu-ka? who-Nom why wrote book-Nom interesting-be-QE ‘Books that who wrote why are interesting?’ Saito (1992) claims that the saving '“’ "'WP must not only c-command the ^W P. but also be base generated in the same clause as the a d v WP. In other words, if a n o m WP is lower than an ^W P or if a n o m WP is base generated in a lower clause, but is relocated in a position higher than an “‘ “WP as a result of movement, the sentences show the island effects just as when there is no n o m WP. To put this schematically, structures like (32) below are disallowed, as exemplified in (33): (32) a. *...[k la n d ... “‘ “W P... "“"WP...]... i t c-command b. ...“"W P ...“"'WP... [c p ...ti... t________ I Move-a Reproduced with permission of the copyright owner. Further reproduction prohibited w ithout permission. 154 (33) a. (=Saito 1992:12, his (32)) *John-wa [n p [ ip naze nani-o katta] hito]-o sagasiteru no? J.-Top why what-Acc bought person-Acc looking;for QE ‘Q John is looking for a person that bought what why?’ b. (=Saito 1992:23, his (59b)) *kimi-wa [np [ip nani-Oj naze Mary-ni [cp John-ga tj katta to] itta] you-Top what-Acc why M-to J.-Nom bought C said hito-o sagasiteru no? person-Acc Iooking;for QE ■ Q you are looking for a person that whatj told Mary [that John bought t,] why?’ In (33a) “^WP naze ‘why’ is higher than n o ra WP nani ‘what.’ In (33b), although n o m WP nani 'what’ is higher than ^W P naze ‘why’ in overt syntax, the "“"WP is originated from the lower clause. Both sentences in (33) are ungrammatical. 4.2.2.2. Free Ride on Nominal WPs Based on the observations discussed in the previous subsection, Saito (1992) proposes that ^W P s like naze in multiple WH-constructions can be licensed by being adjoined to a clause mate "“"WP which, after hosting “^WPs, moves at LF to the SPEC of a [+wh] C°. For example, reconsider the sentences in (2) and (3), which are repeated below: Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 155 (2) (Watanabe 1991:18, his (33)) a. ?Kimi-wa nani-o naze katta no? you-Top what-Acc why bought QE ‘Why did you buy what?' b. *Kimi-wa naze nani-o katta no? you-Top why what-Acc bought QE ‘What did you buy why?’ (3) a. ?ne-nun mwues-ul (ecey) wav sa-ess-ni? you-Top what-Acc yesterday why buy-Pst-QE ‘Why did you buy what (yesterday)?’ b. *ne-nun wav (ecey) mwues-ul sa-ess-ni? you-Top why yesterday what-Acc buy-Pst-QE ‘What did you buy why (yesterday)?’ Saito (1992) assumes that objects are generated below adjuncts, taking (b) sentences as the base structures. Thus (a) sentences are viewed as the result of scrambling of object WPs to a VP adjoined position, as schematically represented below: (34) VP WHAT, VP W H Y ^ ^ VP V I BOUGHT Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 156 Saito (1992) explains the grammaticality of (a) sentences as follows. Following Mahajan (1990), Saito assumes that clause internal scrambling is taken as an A- movement whereas scrambling across a clause boundary as an A’-movement. The “‘ ''WP in (34) is first A-adjoined to the 0 0 0 1 WP, as represented below: (35) VP NP, ADVPk NPj tk WHYk WHAT, t_____ VP VP ti V I BOUGHT At this point of the derivation, the trace of the a d v WP, tk, is antecedent governed by the NP-adjoined position because WHYk c-commands tk under May’s (1985) category vs. segment distinction and no barrier intervenes between the two elements. Now the n o m WP including the a d v WP adjoined to it moves to the local domain of a [+wh] C°. as schematically represented below: Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 157 (36) CP NPi WHYk W HAT/ tj’ VP VP BOUGHT No intermediate trace for the “^WP exists in this diagram because the intermediate trace of n o m WP to which the a d v WP is adjoined, i.e., tj’, will be deleted in order to have a simple operator-variable chain (Chomsky 1989).2 0 This contrasts with the cases of movement of a d v WPs, in which no intermediate traces are deleted since the chain formed as a result of the movement is already uniform. Saito (1992) attributes the ungrammaticality of (b) sentences in (2) and (3) essentially to the ECP violation due to the rigidity condition (Saito 1982) or the Relation Preservation (A. Watanabe 1991). To repeat his point, since “^WPs in these sentences are higher than n o m WPs at SS, the former moves later than n o m WPs at LF, to preserve the relation established at SS. Since the SPEC of CP has the index of n o m WPs which move first, the trace of “^WPs would not be antecedent governed 2 0 The chain reduction does not seem to be compatible with the effects o f reconstruction to intermediate positions, e.g., the ambiguity of the anaphor interpretation in the following sentence, as discussed in Chomsky (1994, 1995): (i) [Which pictures o f himself]; did John think Tom brought tj? Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 158 under the assumption that antecedent government from an A’-adjoined position is impossible. Saito’s adjunction theory gives an account for the rigidity effects within islands. Let us reconsider the sentences in (27), which are repeated below: (27) a. (=Nishigauchi 1986:119, his (24)) * [ n p [s kare-ga naze kai-ta] hon]-ga omosiroi-desu-ka? he-Nomwhy wrote book-Nom interesting-be-QE ‘Books that he wrote why are interesting?’ b. (=Nishigauchi 1986:125, his (32)) ? ( ? ) [n p [s dare-ea naze kai-ta] hon]-ga omosiroi-desu-ka? who-Nomwhy wrote book-Nom interesting-be-QE ‘Books that who wrote why are interesting?’ Needless to say, (27a) is the case of the ECP violation. In contrast, in (27b), the a d v WP naze first adjoins to the n o m WP dare ‘who’ which is in an A-position. Then the n o m WP together with the ^W P adjoined to it moves to the SPEC of CP. Saito (1992) ascribes a little degradation of the sentence in (27b) to the Subjacency violation of the movement across a complex NP island.2 1 This contrasts with the cases where no ^W P is involved :I But I sense the deviance of the sentence is much milder than the case where an element is overtly extracted out o f a relative clause. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 159 in the island. In this case the question operator is generated outside the complex NP. violating no Subjacency condition. Now consider the sentences in (33), which are repeated below: (33) a. (=Saito 1992:12, his (32)) * John-wa [ n p [ip naze nani-o katta] hito]-o sagasiteru no? J.-Top why what-Acc bought person-Acc looking;for QE 'Q John is looking for a person that bought what why?’ b. (=Saito 1992:23, his (59b)) *kimi-wa [n p [ ip nani-Oi naze Mary-ni [cpJohn-ga tj katta to] itta] you-Top what-Acc why M-to J.-Nom bought C said hito]-o sagasiteru no? person-Acc Iooking;for QE ‘Q you are looking for [a person that whatj told Mary[that John bought tj] why]?’ According to Saito (1992), sentences like (33a) are excluded just as when no island is involved. These sentences violate the ECP due to the rigidity condition or the Relation Preservation. In sentences like (33b), a n o m WP, nani ‘what,’ undergoes a long distance scrambling across a clause boundary. As a result, it moves to an A'- position. Thus adjunction of the “‘ ''WP to the n o m WP would not help since antecedent government from an A’-adjoined position is not allowed by assumption. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 160 Saito’s theory seems to account for the fact that more than one ^W P is allowed as long as there is a rescuing n o m WP for each instance of an “^WP, which I observed in section 4.2.1.3. Let us take a look at the contrast given in (29) and (30). repeated below: (29) a. *Mary-ga naze [John-ga hon-o naze katta to] omotteruno? M.-Nomwhy J.-Nom book-Acc why bought C think QE ‘*Why does Mary think that John bought books why?’ b. dare-ga naze [John-ga nani-o naze katta to] omotteruno? who-Nomwhy J.-Nom what-Acc why bought C think QE ‘Who thinks [that John bought what why] why?’ (30) a. *Mary-ka wav [John-i chayk-ul wav sa-ess-ta-ko] sayngkakha-ni? M.-Nom why J.-Nom book-Acc why buy-Pst-DE-C think QE **Why does Mary think that John bought books why?’ b. nwu-ka wav [John-i mwues-ul wav sa-ess-ta-ko] sayngkakha-ni? who-Nom why J.-Nom what-Acc why buy-Pst-DE-C think-QE ‘Who thinks [that John bought what why] why?’ The sentences in (a) will be excluded as cases of the ECP violation. The sentences in (b), however, are grammatical since each instance of ^W Ps will A-adjoin to a n o m WP in a higher position and will antecedent govern its trace. Reproduced with permission of the copyright owner. Further reproduction prohibited w ithout permission. 161 4. 2. 2 3. Problems An immediate problem with Saito (1992) is that he fails to explain the grammaticality of A. Watanabe’s (1991) sentences like (7), which are repeated below: (7) a. (=A. Watanabe1991:93, his (198b)) dare-ga [cpJohn-ga naze kubininatta to] itta no? who-Nom J.-Nom why was;fired C said QE ‘Who said [that John was fired why]?’ b. (= A. Watanabe 1991:93, his (198a)) (?)[cpJohn-ga naze kubininatta to] dare-ga itta no? J.-Nom why be;fired C who-Nom saidQE ‘Who said [that John was fired why]?’ In the sentences in (7), an ^W P is in the embedded clause, while a n o m WP is in the matrix clause. The ^W P naze cannot be A-adjoined to the n o m WP dare since a clause boundary intervenes (and additionally dare does not c-command naze in (7b)). Having no clause mate n o m WP to adjoin to, “^WP naze in these sentences is predicted not to be licensed properly. One might say that the sentences in (7) are fine because there is an option in which the a d v WPs are licensed by moving to the local domain of the [+wh] C°, just as when the sentences without any n o m WPs are fine. However, once this is allowed, there are Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 162 at least two problems that arise. The first problem arises with his own observation regarding the cyclicity of the additional WP effect; the ASE is suspended by adding a WP in the same clause, but not by adding a WP in a higher clause. Recall the contrast between the sentences in (21) and (22) on the one hand, and those in (25) and (26) on the other, which are repeated below: (21) a. (=(2b)) *Kimi-wa naze nani-o katta no? you-Top why what-Acc bought QE ‘What did you buy why?’ b. dare-ga naze nani-o katta no? who-Nom why what-Acc bought QE ‘Who bought what why?’ (22) a. (=(3b)) *ne-nun wav (ecey) mwues-ul sa-ess-ni? you-Top why yesterday what-Acc buy-Pst-QE ‘What did you buy why (yesterday)?’ b. nwu-ka wav (ecey) mwues-ul sa-ess-ni? who-Nom why yesterday what-Acc buy-Pst-QE ‘Who bought what why (yesterday)?’ (25) a. ?*Tom-wa [John-ga naze nani-o katta to] omotteru no? T.-Top J.-Nom why what-Acc bought C think QE ‘Q Tom thinks [that John bought what why]’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 163 b. (=Saito 1992:25, his (60c)) ?*dare-ga [John-ga naze nani-o katta to] omotteruno? who-Nom J.-Nom why what-Acc bought C think QE ‘Q who thinks [that John bought what why]’ (26) a. ?*ne-nun [John-i wav mwues-ul sa-ess-ta-ko] sayngkakha-ni? you-Top J.-Nom why what-Acc buy-Pst-DE-C think-QE ‘Q who thinks [that John bought what why]’ b. ?*nwu-ka [John-i wav mwues-ul sa-ess-ta-ko] sayngkakha-ni? who-Nom J.-Nom why what-Acc buy-Pst-DE-C think-QE ‘Q who thinks [that John bought what why]’ If a d v WPs were allowed to move to the SPEC of a [+wh] C°, the sentences in (25b) and (26b) would not violate the Rigidity Condition (Saito 1982) or the Relation Preservation (A. Watanabe 1992) and would be predicted to be grammatical, just as in (21b) and (22b). The postulation of movement of “^WPs in this context bears another problem, a redundancy problem with respect to the licensing of “^WPs in (2a) and (3a). Note that there will be two ways to license the ad v WPs in these sentences. One way is to adjoin the a d v WP to the n o m WP and then move the complex to the SPEC of CP. as discussed around (35) and (36) in the previous section. The other way is to move the a d v WP to the SPEC of CP and then adjoin the n o m WP to the SPEC of CP. as schematically illustrated below: Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. nanij nazej [+wh] IP ... tj ... tj ... The very existence of the two ways of derivation for the same sentence is very doubtful, assuming that the grammar is optimal. Thus one of the derivations must be eliminated for a principled ground. But Saito’s analysis is silent about this. 4.2.3. S. Watanabe (1995) 4.2.3.I. S. Watanabe’s (1995) WCO Account of ASEs S. Watanabe (1995, chapter 2) tries to capture the ASE as an instance of Weak Crossover (henceforth WCO) phenomena along the lines of Chierchia’s (1992-1993) treatment of WP/QP interactions and Homstein’s (1994, chapter 6) extension of his idea to superiority effects in multiple WH-questions. S. Watanabe (1995, chapter 2) claims that the ASE is an instance of WCO phenomena under the assumption that whenever a d v WP naze ‘why’ and another WP share their interrogative scope, the former contains a variable bound to the other WP which becomes a generator in the sense of Chierchia (1992-1993). For example, the Reproduced with permission of the copyright owner. Further reproduction prohibited w ithout permission. 165 sentences in (2), repeated below, will have the LF structures in (38) with the variable provided: (2) (A. Watanabe 1991:18, his (33)) a. ?Kimi-wa nani-o naze katta no? you-Top what-Acc why bought QE ‘Why did you buy what?’ b. *Kimi-wa naze nani-o katta no? you-Top why what-Acc bought QE ‘What did you buy why?’ (38) a. ?Kimi-wa nanij-o [PROj naze] katta no? The structure in (38a) is licit since the variable contained in naze, which is represented as PRO, is c-commanded by the generator WP nani. In contrast, the structure in (38b) is illicit since PRO is not c-commanded by nani, a WCO c-command b. *Kimi-wa Oj naze] nanij-o katta no? *c-command violation.2 2 In other words, the difference in grammaticality between (2a) and (2b) is explained on a par with the difference in the following contrast: 2 2 S. Watanabe (1995:51) wishes to exclude the following structure as an LF for (2b): Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (39) a. Every boy; loves fhisi mother], b. *fHiSi mother] loves every bovi. 166 S. Watanabe (1995) finds further evidence for the parallelism between the ASE and the WCO in their suspension effects. In other words, he claims, based on Saito's (1992) observation in Japanese and Homstein’s (1994) observation in English, that the ASE is suspended exactly where the WCO is suspended. Saito (1992) reports that addition of a n o m WP in a lower, as well as in a higher, position saves the otherwise unacceptable structure. Let us confine our discussion to the lower WH-effects since the higher WH-effects seem to be expected by other theories. Thus, the structure given in (40b), as opposed to that in (40a), is acceptable, as exemplified in (41):2 3 (i) (=his (5lb)) *?[c p [c- [ip - [ v p naze, [vp [PRO, dare-ga] V]] ...]] ka] One possibility he suggests (p. 47) is that naze ‘why’ cannot be a generator for the reason that it cannot be d-linked in general, which is assumed to be a prerequisite for an element to be a generator. Even when naze is forced to be d-linked, S. Watanabe (1995:57), basically following Kuno and Takami (1993, chapter 3), wishes to exclude the structure for a pragmatic reason that mapping of reasons to individuals is not customary in human cognitive acts. 2 3 The sentential judgment on (41b) is not uncontroversial. For example, A. Watanabe (1991) takes such sentences to be ungrammatical. S. Watanabe (1995:59) further claims that for the structure in (40b) to be licit, the * d v WP and the quantifier should be in the same scope domain, i.e., in the same clause. Thus, the structure in (i) below is illicit, as exemplified in (ii), on a par with the ungrammaticality o f the sentences like (iii): (i) * ... * d v W P ... [CP... “ “ WP ... “""WP ...] Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 167 (40) a. *... ^W P ... " “"WP... b. ... ,dvWP ... n o m WP ... n o m WP ... (41) a. (Saito 1992, his (2b); S. Watanabe 1995: 55, his (59a)) *Naze dare-ga soko-ni itta no? why who-Nom there-to went Q ‘Why who went there?’ b. (Saito 1992, his (3b); S. Watanabe 1995: 56, his (59b)) ?Naze dare-ga doko-ni itta no? why who-Nom where-to went Q ‘Why who went where?’ The suspension of the ASE with the addition of a WH-element in a lower position is equated with the suspension of the WCO with an addition of a pronoun in a lower position, as in (42b) below, which is observed by Homstein (1994): (ii) (S. Watanabe 1995:61, his (71)) ♦Naze John-ga [dare-ga doko-ni itta to] itta no (ka oshiete kudasai) why -Nom who-Nom where-to went Comp said Q tell please ‘(Tell me) why John said that who went where.’ (iii) (Homstein 1994, S. Watanabe 1995:57, his (66)) *Hjs, mother said [that every studenti hated h|st advisor] S. Watanabe (1995:60) attributes the ungrammaticality of (ii), (and (iii) as well,) to Heim’s (1982) condition on bound variables: a bound pronoun must be in the scope domain of a quantifier to which it is bound. Notice that the quantificational (as opposed to the interrogative) scope domain o f dare in (ii) is confined to the embedded scope. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 168 (42) (S. Watanabe 1995: 57, his (62)) a. *His, mother gave every bov; a book. b. His, mother gave every bovi hisi book. 4.23.1. Problems S. Watanabe’s (1995) subsumption of the ASE under the WCO captures various similarities between them. But there are also several reasons that the two cannot be put on the same line. The first reason that they cannot be equated is that “^WPs are sensitive to intervening islands, whereas bound variables are not. Consider the schematic structure in (43) and the Korean examples in (44): (43) a. * ... n o m WP ... [ is l a n d ... ^ W P ....] b. V... QPj... [ is l a n d — pronouni....] (44) a. *nwu-ka [[e* chayk-ul way ssu-n] salamj-ul] chotayha-ess-ni? who-Nom book-Acc why write-Adn man-Acc invite-Pst-QE ‘Who invited a man who wrote a book why?’ Reproduced with permission of the copyright owner. Further reproduction prohibited w ithout permission. 169 b. motun haksavng,-i [[caJdj-uy citokyoswu-ka ej ssu-n] nonmwunj-ul all student-Nom self-Gen advisor-Nom wrote-Adn article-Acc caseyhi ilk-nun-ta.2 4 carefully read-Pres-DE •Every student carefully reads the article that his advisor wrote.’ A n o m WP nwu ‘who’ is in the matrix subject position and an a d v WP way ‘why’ is in a complex NP in (44a). The sentence is out as if there were no WP in the matrix clause. For example, even if nwu-ka in the matrix subject position is replaced by a name, e.g., John-i, the sentence is ungrammatical. According to S. Watanabe’s theory, (44a) will have the following schematic LF structure: (45) [... WPj... [island ... [PROi naze]...] Now the structure is very similar to the structure in (44b) in that a variable inside an island is bound by a quantificational expression outside the island. They both are predicted to be licit. But in fact, the structure in (43b) is allowed, whereas the one in (43 a) is not. A second fact that is not readily explained by the WCO account of the ASE is that there are cases where a d v WPs are not necessarily c-commanded by a n o m WP, as 2 4 I intentionally used a reflexive anaphor as a bound variable, instead of a pronoun, since it is controversial whether pronouns in Korean could function as bound pronouns. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 170 was observed by A. Watanabe (1991). Recall the licitness of the structure in (6c) and the grammaticality of the sentence in (7b), which are repeated below: ...a d v WP n o m WP... (7) b. (= A. Watanabe 1991:93, his (198a)) (?)[cpJohn-ga naze kubininatta to] dare-ga itta no? J.-Nom why be;fired C who-Nom said QE ‘Who said [that John was fired why]?’ According to S. Watanabe’s theory, the sentence in (6b) will have the following simplified abstract LF structure: (46) [... [PlpDj naze]...]... cjarej... no? *c-command Since the variable contained in naze is not bound by dare, this structure is predicted to be ungrammatical, which is not true.2 5 2 5 This problem could be avoided by reconstructing the scrambled CP in (7) to its original position so that dare may c-command naze. Notice, however, that the structure in (i) below is licit, as exemplified in (ii): Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 171 43. An Economy Based Approach to the ASE In this section, I argue that both the anti-superiority effect and its cyclic manifestation observed by Saito (1992) are natural consequences from the perspective of the economy principle given some plausible assumptions. 43.1. The Core Case of the ASE and the Economy of Derivation In this section I attempt to account for the core case of ASE in terms of the economy of derivation combined with the basic idea of Saito’s (1992) approach to the multiple WH-construction. “ “Wp _ *,V WP (») (?)[c p dare-ga hakubininatta to] John-ga naze itta no? who-Nom be;fired C J.-Nom why said QE ‘Q John said [that who was fired] why?’ Dare does not c-command naze in (ii), whether the CP is reconstructed or not The grammaticality of the sentence seems to indicate that reconstruction is not relevant Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 43.1.1. Licensing of Adverbial WPs in Terms of QBR 172 As was briefly introduced in section 4.1., I assume that there are two ways of licensing “^WPs. The first way is that ^W P s are licensed by means of having an ESHR with a [+wh] C° as was proposed in the previous chapter. The other way is that ^W Ps are licensed by having a structural relation, which I call a QBR (quasi-binding relation), with the nominal feature that question clauses have. To be licensed by a QBR. an “^WP must be adjoined to a "“"WP, basically like Saito’s (1992) adjunction analysis. Once an ^W P is adjoined to and covered by a n o m WP, the ^W P is parasitically licensed when the "“"WP is licensed by a BR (binding relation) with the nominal feature that question clauses have, as illustrated in (5), repeated below: (5) DP DP CP VP n o n v l WP a^WPj nom \y p QBR Move-a BR Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 173 43.1.2. ASE and the Economy The postulation of the double ways of licensing of ^W Ps, however, brings up an immediate question on the very existence of the ASE in multiple WH-constructions in languages such as Japanese and Korean. For example, reconsider the contrast in the sentences in (2) and (3). (2) (=A. Watanabe 1991:18, his (33)) a. ?Kimi-wa nani-o naze katta no? you-Top what-Acc why bought QE ‘Why did you buy what?’ b. *Kimi-wa naze nani-o katta no? you-Top why what-Acc bought QE ‘What did you buy why?’ (3) a. ?ne-nun mwues-ul (ecey) wav sa-ess-ni? you-Top what-Acc yesterday why buy-Pst-QE ‘Why did you buy what (yesterday)?’ b. *ne-nun wav (ecey) mwues-ul sa-ess-ni? you-Top why yesterday what-Acc buy-Pst-QE ‘What did you buy why (yesterday)?’ The a d v WPs in (a) sentences, i.e., naze/way, could be licensed either by an ESHR, i.e., by directly moving to the SPEC of a [+wh] C° or by a QBR, i.e., by being adjoined to Reproduced with permission of the copyright owner. Further reproduction prohibited w ithout permission. 174 the n o m WPs. nani-o/mwues-ul. In (b) sentences, the “^WPs could be licensed by an ESHR, i.e., by moving to the SPEC of the [+wh] C°, although the option of licensing by a QBR is unavailable due to the hierarchical relation of the two WPs. Thus the prediction would be that both types of sentences should be grammatical. Not only is the prediction about (b) sentences incorrect, but the postulation of the two options of licensing of “^WPs in (a) are questionable since only the most optimal derivation survives among convergent derivations according to the minimalist assumption that we basically follow. In what follows, I will try to solve the redundancy problem that arises in accounting for the licitness of the structure in (la) in terms of the economy of derivation. I also will try to answer the question of why the ASE exists in the grammar of languages like Japanese and Korean. At a certain point of the derivation the two sentences in (2) and (3) will share the following structure, assuming that adjunct phrases are generated higher than complements (Hoji 1985):2 6 (47) VP adv WP VP nom WP V 2 6 In fact, even if the hierarchical relation is reversed, our argumentation in the text is not affected except for the application of scrambling o f “ “ WP. The VP internal subject hypothesis is ignored here, since it does not make any substantial difference in the present context. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 175 To have the surface order of (2a) or (3a), where the "“"WP is higher than the ^ ' WP. the n o m WP is scrambled over the “^WP, yielding the following schematic structure: (48) VP n o m WPj v p ^W P VP Now compare the two potential derivations that make (a) sentences convergent. One is to move the “^WP to the SPEC of the [+wh] C°. The other is to adjoin the a d v WP to the n o m WP, which itself is licensed by a binding relation. The two options are schematically represented below:2 7 2 7 For the sake of convenience, heads are placed before complements. Question clauses are analyzed as DPs in chapter 3. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 176 (49) a.(=(4a)) DP DP CP b. (= (5)) DP ESHR VP B J L Move-a DP CP [+wh] IP VP nom' 'WP Q BR Move-a BR Obviously the “ ^WP movement in (49a) has a longer LF derivation than that in (49b). Notice that the ^W P moves all the way to the SPEC of CP in the former, while it is just adjoined to the “""WP, which is dominated by the immediately higher node. Thus, the option described in (49b) should be taken over the option in (49a). Now turn to the question why (b) sentences in (2) and (3) are ungrammatical. Before spell-out, the sentences will have the following structure: Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 177 (50) DP Since the "“"WP is located in a lower position, the option of adjoining the ^W P to it appears to be unavailable. Therefore, the only option that seems to be left is to move the ^ W P to the SPEC of the [+wh] C° so that it can have an ESHR with a [+wh] C°. as represented below: (51) DP DP CP [+wh] ESHR VP VP Move-a(LF) n o m WP BR Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 178 The derivation in (51) does not seem to bear any problem with respect to convergence and the economy. But my claim here is that the derivation does violate the economy principle under the assumption that clause internal scrambling is optional in languages such as Japanese and Korean and furthermore is a costless operation (Fukui 1993).2 8 Under this assumption, the ungrammatical status of (b) sentences in (2) and (3) are accounted for by the same economy measurement In other words, they are ungrammatical since the scrambling of the n o m WPs has not applied. Notice that if the scrambling operation occurred, the “^WP would be licensed by being adjoined to the n o m WP, i.e., just like the ^W P in (49b). The derivation in (49b) is shorter than that in (51). In brief, it is argued in this section that the existence of the ASE follows from the economy of derivation under the assumption that clause internal scrambling is optional and costless.2 9 2 8 Admittedly, the assumption is a mere stipulation. But I hope further study on scrambling will elucidate the aspects of the assumption. So I leave it as it is for now except for mentioning the following two aspects of scrambling. First, scrambling does not seem to take place for the purpose of feature checking. (See Sohn 1995 for this claim.) Second, the various syntactic differences with relation to the distance of scrambling (clause internal scrambling vs. scrambling across a clause boundary) may be related to the L-related (or thematic) domain of a lexical head. 2 9 A remark to be made in relation to the adjunction analysis of multiple WPs is that the costless scrambling is an operation that is available before spell-out but not after spell-out Otherwise the contrast in terms of the adjunction analysis could not hold anymore. Notice that the ungrammatical cases could be remedied simply by scrambling "“"WPs over ,d v WPs. If such a scrambling were available in the LF component after spell-out as well as before spell-out then all the ungrammatical cases discussed so far would be predicted to be good, an incorrect result But the LF scrambling seems to be banned independently. If this were allowed, phenomena like weak crossover effects or binding theoretical violations would not be attested altogether in the Japanese or Korean grammar, as far as the two relevant elements are in the same clause, which is obviously not the case. Thus, costless scrambling should be prevented from occurring in the covert component, whatever is responsible for this. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 4 J.2. Cyclicity of the Additional WP Effects and the Economy 179 As was discussed in section 4.2.2.1., the ASE may or may not be suspended b y - adding a n o m WP in a higher position (Saito 1992). It is suspended when the added WP is in the same clause, but not when it is added in a higher clause. Relevant structures and examples are repeated below for the sake of easy reference: (20) a. * ... “^WP ... n o m W P... b. V n o m WP “^WP n o m WP (21) a.(=(2b)) *Kimi-wa naze nani-o katta no? you-Top why what-Acc bought QE ‘What did you buy why?’ b. dare-ga naze nani-o katta no? who-Nom why what-Acc bought QE ‘Who bought what why?’ (22) a. (=(3b)) *ne-nun wav (ecey) mwues-ul sa-ess-ni? you-Top why yesterday what-Acc buy-Pst-QE ‘What did you buy why (yesterday)?’ b. nwu-ka wav (ecey) mwues-ul sa-ess-ni? who-Nom why yesterday what-Acc buy-Pst-QE ‘Who bought what why (yesterday)?’ Reproduced with permission of the copyright owner. Further reproduction prohibited w ithout permission. 180 (24) a. V... n o m WP ... a d v WP ... "“"WP..., where no clause boundary involved, b. ?*... n o m W P... [c p ..."‘'WP ... 1 1 0 0 1 W P... (25) a. ?*Tom-wa [John-ga naze nani-o katta to] omotteru no? T.-Top J.-Nom why what-Acc bought C think QE ‘Q Tom thinks [that John bought what why]’ b. (=Saito 1992:25, his (60c)) ?*dare-ea [John-ga naze nani-o katta to] omotteruno? who-Nom J.-Nom why what-Acc bought C think QE ‘Q who thinks [that John bought what why]’ (26) a. ?*ne-nun [John-i wav mwues-ul sa-ess-ta-ko] sayngkakha-ni? you-Top J.-Nom why what-Acc buy-Pst-DE-C think-QE ‘Q you think [that John bought what why]’ b. ?*nwu-ka [John-i wav mwues-ul sa-ess-ta-ko] sayngkakha-ni? who-Nom J.-Nom why what-Acc buy-Pst-DE-C think-QE ‘Q who thinks [that John bought what why]’ The sentences in (21b) and (22b) are acceptable, as opposed to (21a) and (22a), since the a d v WP is able to be licensed by being adjoined to the subject WP, an option preferred to the option of movement to the SPEC of the [+wh] C°. In contrast, the sentences in (25b) and (26b) are no better than (25a) and (26b) since in the embedded clause the n o m WP has not scrambled over the a d v WP. Notice the scrambling, if applied, would make the LF derivation for the identification of the “^WP shorter, compared with the option where the ^W P directly moves to the SPEC of the [+wh] C°. Reproduced with permission of the copyright owner. Further reproduction prohibited w ithout permission. 181 4 JJ. Exemption from the Ban on Multiple Occurrence of Adverbial WPs As was observed in section 3.3.2. of chapter 3, two or more ^W Ps generally cannot share their interrogative scope. A multiple occurrence of “ ^WPs is, however, sometimes allowed, as was discussed in section 4.2.2.3.2. Let us reconsider the sentences in (29) and (30), repeated below: (29) a. *Mary-ga naze [John-ga hon-o naze katta to] omotteru no? M.-Nomwhy J.-Nom book-Acc why bought C think QE '“Why does Mary think that John bought books why?’ b. dare-ga naze [John-ga nani-o naze katta to] omotteruno? who-Nom why J.-Nom what-Acc why bought C think QE ‘Who thinks [that John bought what why] why?’ (30) a. *Mary-ka wav [John-i chayk-ul wav sa-ess-ta-ko] sayngkakha-ni? M.-Nom why J.-Nom book-Acc why buy-Pst-DE-C think QE ‘“Why does Mary think that John bought books why?’ b. nwu-ka wav [John-i mwues-ul wav sa-ess-ta-ko] sayngkakha-ni? who-Nom why J.-Nom what-Acc why buy-Pst-DE-C think QE ‘Who thinks [that John bought what why] why?’ As indicated in (a) sentences, “^WP naze in Japanese or way in Korean cannot appear twice if they have the same scope.3 0 It was argued in chapter 3 that these kinds of 3 0 Of course, two or more “‘ ''WPs are allowed if they have a different scope, as exemplified below: Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 182 sentences are ungrammatical because one of the “^WPs in these sentences does not have an ESHR with the [+wh] C°. But Japanese sentences like (29b) and Korean sentences like (30b) are grammatical. The theory proposed here is able to account for the grammatical status of the sentences in (29b) and (30b). Note that the first occurrence of ^W P will be adjoined to the n o m WP in the matrix subject position and the second occurrence of “^WP to the n o m WP in the embedded object position, as schematically represented below: (52) DP CP nonv 'WPj nonv 'W P j VP ...tj ...CP QBR Move-a ...VP VP nonv ‘ WPj tic VP QBR Move-a (i) way John-i [Mary-ka wav o-ess-nun-ci] mwut-ess-ni? why J.-Nom M.-Nom why come-Pst-Adn-QE ask-Pst-QE ‘Why did John ask why Mary came?’ In (i) the first occurrence of way has the matrix scope, while the second has the embedded scope. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 183 The “‘ ''WPs in this context, being covered by a n o m WP, are licensed by a QBR. 43.4. Other Combinations of WPs and the Adjunction Operation In the previous section, it was argued that one way of licensing “‘ ''WPs was to adjoin them to a clause mate n o m WP. For the sake of completeness, let us consider in this section whether other combinations of WPs implement the adjunction operation for the licensing of WPs. Since we have dealt with the adjunction of an “^WP to a n o m WP. there remain three cases to be considered: the adjunction of a n o m WP to an a d v WP, the adjunction of an “^WP to another a d v WP, and the adjunction of a n o m WP to another n o m WP. To say the conclusion first, these operations are either unnecessary or useless. Let us first consider the case where an “^WP and a n o m WP are in the same clause and the former c-commands the latter, as schematically represented below: (53) *XP adv WP YP nonv WP ZP In the previous sections, this structure was excluded since scrambling the 0 0 1 , 1 WP over the “‘ ''WP, which feeds a shorter derivation for the purpose of licensing of WPs Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 184 involved, did not apply. Let us, however, check whether adjunction of the n o m WP to the a d v WP gives rise to an optimal derivation comparable to the case where the hierarchical order of the two WPs are reversed. Obviously the adjunction does not help. Notice that even if the adjunction takes place, the “^WP cannot be licensed in situ by a BR under the assumption that the BR is confined to nominal elements. Notice the resultant cluster of WPs remains as an WP, as represented below: (54) XP adv W P Y P "“"WPj ^ W P tj t ___________ To be licensed, the complex WP must move to the SPEC of a [+wh] C°, as schematically represented below: (55) CP 'WPj ^ W P i [+wh] IP nonv YP Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 185 Although the “^WP could be licensed by being raised to the SPEC of a [+wh] C° and having an ESHR with the [+wh] C°, the n o m WP is adjoined to the “^WP for nothing, which is an operation to be eliminated under the minimalist program. In short, such an adjunction of a n o m WP to an a d v WP is an unnecessary operation, and therefore must be eliminated for the economy reason. Secondly, let us check whether an “^WP can be licensed by being adjoined to another a d v WP. Suppose the structure in (56) below is produced at a certain point of derivation and further suppose that the lower WP, i.e., “^WPj, is adjoined to the higher WP, i.e., ^W Pj, as in (57): (56) XP (57) XP Even after the adjunction takes place, the complex construct of WPs, still being an adverbial, must move to the SPEC of a [+wh] C°, unlike the cases where an “^WP is adjoined to a n o m WP and is licensed by a QBR. Not being licensed in situ, the non- nominal complex will ultimately move to the SPEC of a [+wh] C°, as the following structure illustrates: Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 186 (58) CP I WPj ^W Pi [+wh] YP Notice that the resultant structure is not different from the one in which the ^W P, first moves to the SPEC of the [+wh] C° and subsequently “^WPj is adjoined to it. as illustrated below: (59) CP advwpj c ^W Pj ^W P, [+wh] YP Move-a Move-a Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 187 The adjoined WP will not be licensed, not having an ESHR with the [+wh] C°. whichever structure is taken. The prediction is borne out As was observed by Aoun (1986:96) in Chinese, two or more “‘ ''WPs cannot share their interrogative scope, as exemplified in the following Chinese sentence cited from Aoun (1986:96): (60) * Jim weisheme zeme mai-le shu? J. why how buy-ASPbook 'Jim bought the book why, how?’ As was discussed in section 3.3.2. of chapter 3, in Korean also, two or more non- nominal WPs cannot have the same scope even if no barrier intervenes between any of them and their scope position, as exemplified in (61): (61) *wav John-i [Maiy-ka wav chavk-ul ilk-ess-ta]-ko malha-ess-ni? why J.-Nom M.-Nom why book-Acc read-Pst-DE-C say-Pst-QE ‘Why did John say that Mary read books why?’ Unlike the combination of n o m WPs and ^W Ps, that of two ^W P s produces ill-formed sentences. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 188 Let us finally discuss the case where two " “"WPs appear in the same clause, as schematically represented in (62) below and check whether adjunction of the lower n o m WP to the higher one is possible, as in (63): (62) WPj YP (63) .. W Pj... nom WPj "“"WP, ... tj... The adjoined WP complex need not move to the SPEC of a [+wh] C since n o m WPs can be licensed by a BR, as the following structure illustrates: (64) WPj WP, The adjoined n o m WP, i.e., n o m WPj, may be said to be licensed by a QBR while the hosting n o m WP, i.e., n o m WPj, is being licensed by a BR. Compare, however, this way Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 189 of licensing and the one in which no adjunction takes place at all. When no adjunction takes place, both of the " “"WPs will be licensed by means of binding, as illustrated below: Both the licensing method illustrated in (64) and the one in (65) do not seem to have any problem with the licensing of n o m WPs per se. It seems that the latter option should be taken due to the economy of derivation and representation. Notice that the first option includes an adjunction operation and binding, whereas the latter includes two instances of binding. Assuming that binding is just a process reading off given structures, and therefore no cost is imposed on the process, multiple use of the binding option, i.e., unselective binding, will not add up any cost, either. In contrast, the option in (64) includes an adjunction process whose cost cannot be ignored in the economy calculation. Therefore, the option in (65) should be taken over the one in (64). Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 190 In sum. adjunction of a WP to another WP seems to be restricted to the case where an “^WP is adjoined to a n o ra WP. The other cases of adjunction either do not help (adjunction of an ^W P to another “^WP) or need not (and must not for the economy reason) take place (adjunction of a n o m WP to an “^WP or another n o m WP). 4.3.5. Non-Nominal WP Movement Cases So far we have seen that when an ^W P and a n o m WP share their interrogative scope, the former is parasitically licensed by being adjoined to the latter. In this section, we will see there are at least two cases where ^W Ps are licensed by having an ESHR with a [+wh] C°, even when they share their interrogative scope with a n o m WP. Let us first consider the structure in which a n o m WP and an ^ W P is separated by a clause boundary, as schematically represented in (66) and exemplified in (67) and (68) below: (66) ... n o m W P... [cp ... ^ W P ... (67) (= (7a), cited from A. Watanabe 1991:93, his (198b)) dare-ga [cpJohn-ga naze kubininatta to] itta no? who-Nom J.-Nom why was;fired C said QE ‘Who said [that John was fired why]?’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 191 (68) nwu-ka [cpJohn-i wav haykotoi-ess-ta-ko] malha-ess-ni? who-Nom J.-Nom why was;fired-Pst-DE-C said-Pst-QE ‘Who said [that John was fired why]?’ This structure contrasts with the one where another n o m WP is present in a position lower than the a d v WP. As was discussed in section 4.2.1.3.2. and section 4.3.2.2.. the latter structure is not allowed: (69) (=(25a), cited from Saito 1992:25, his (60c)) ?*dare-ea [John-ga naze nani-o katta to] omotteru no who-Nom J.-Nom why what-Acc bought C think QE ‘Q who thinks [that John bought what why]’ (70) (=(26a)) ?*nwu-ka [John-i wav mwues-ul sa-ess-ta-ko] sayngkakha-ni? who-Nom J.-Nom why what-Acc buy-Pst-DE-C think-QE ‘Q who thinks [that John bought what why]’ In section 4.3.2.2., the ungrammaticality of such sentences was attributed to the lack of scrambling of the lowest "“"WP over the “ ^WP, which would feed a shorter derivation with respect to the licensing of the ‘ “’ ''WP than the one in which the ^W P moves to the SPEC of the [+wh] C°. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 192 In (67) and (68), no n o m WP is available in the embedded clause for the “‘ ’ 'WPs to potentially adjoin to. Thus, there seem to be two options left: the “‘ ’ 'WPs should be adjoined to the "“"WPs in the higher clause or move to the SPEC of the [+wh] C°. as the following diagrams illustrate: (71) a. DP DP CP [+wh] IP nornwpj VP a d v WPj " “"WPj ...CP.. QBR Movje-a BR ...t; b. DP DP CP [+wh] IP ESHR n o m WP, VP BR Move-a If the adjunction across a clause boundary is ever possible, the derivation in (71a) should be taken over the one in (71b) since the former has a shorter process than the latter. Nevertheless, I exclude the possibility of adjoining the 8 d v WP to the n o m WP in the higher clause as a typical improper movement.3 1 Adjunction of an ad v WP to a 3 1 The cyclicity of the ASE would not be a problem even if such an adjunction across a clause boundary were allowed. Take the sentences in (69) and (70) for example. Scrambling o f the n o m WP over the * dvWP in the embedded clause would have a shorter derivation than the adjunction of the “^WP to the “ “ WP in the higher clause across a clause boundary. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 193 clause mate n o m WP may not be an improper movement since no non-L-related category has been crossed. That the a d v WP in this context is not licensed by being adjoined to the n o m WP in a higher clause is indirectly supported by the fact that such a construction cannot be embedded under an island. Remember Nishigauchi’s (1986:125) observation that the island effect for a d v WPs is drastically weakened when a n o m WP is available, as exemplified in (27) and (28) in section 4.2.1.3.2., repeated below: (27) a. (=Nishigauchi 1986:119, his (24)) * [ n p [s kare-ga naze kai-ta] hon]-ga omosiroi-desu-ka? he-Nomwhy wrote book-Nom interesting-be-QE ‘Books that he wrote why are interesting?’ b. (=Nishigauchi 1986:125, his (32)) ? (? ) [np [s dare-ga naze kai-ta] hon]-ga omosiroi-desu-ka? who-Nom why wrote book-Nom interesting-be-QE ‘Books that who wrote why are interesting?’ (28) a. * [ n p [s ku-ka way ssu-n] chayk]-i caymiiss-ni? he-Nom why write-Adn book-Nom interesting-be-QE ‘Books that he wrote why are interesting?’ b . ? ( ? ) [ n p [s nwu-ka wav ssu-n] chayk]-i caymiiss-ni? who-Nom why write-Adn book-Nom interesting-be-QE ‘Books that who wrote why are interesting?’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 194 But the weakening effect does not seem to be attested when a clause boundary intervenes between the two WPs, as schematically represented in (72) and exemplified in (73) below: (72) *[cP Q [ip ... [island ... "^W P ... [cp ... ^W P ... (73) a. *John-un Fm p Itp nwu-ka Mary-eykey [cp Sue-ka wav e, cohaha-n-ta]-ko J.-Top who-Nom M.-Dat S.-Nomwhy like-Pres-DE-C malha-n] [n p salamj-ul]] manna-ess-ni? say-Adn man-Acc meet-Pst-QE Q John met a manj that who said to Mary that Sue likes ej why?’ b. * John-un [n p [c p Mary-ka nwukwu-evkev [cp Sue-ka wav ej cohaha-n-ta]-ko J.-Top M-Nom who-Dat S-Nom why like-Pres-DE-C malha-n] [n p salamj-ul]] manna-ess-ni? say-Adn man-Acc meet-Pst-QE Q John met a manj that Mary said to whom that Sue likes ej why?’ In the sentences (73), both a n o m WP and an “^WP appear in a complex noun phrase, but they are separated by a clause boundary. If the “^WP is licensed by being adjoined to the n o m WP in this context also, then the sentences should be fine as when there is no clause boundary between the WPs. The sentences, however, are as bad as when there is no n o m WP in the higher clause. The ungrammaticality of the sentences seems to indicate that no adjunction operation is available when a clause boundary intervenes. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 195 There is another type of construction in which “‘ ’ 'WPs supposedly move to the SPEC of CP. Consider the following sentences: (74) (= (7b) cited from A. Watanabe 1991:93, his (198a)) (?)[cpJohn-ga naze kubininatta to] dare-ga itta no? J.-Nom why be;fired C who-Nom saidQE ‘Who said [that John was fired why]?’ (75) (?)[cpJohn-i way haykotoy-ess-ta-ko] nwu-ka malha-ess-ni? J.-Nom why be;fired-Pst-DE-C who-Nom say-Pst-QE ‘Who said that John was fired why?’ In these sentences a d v WP is located in the complement clause of a bridge verb which is scrambled to the initial position. Thus the relevant schematic structure will be like (76) below: Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. The “^WP cannot be adjoined to the "“"WP since neither the c-command relation nor the clause mate condition is met Thus the a d v WP must move to the SPEC of the [+wh] C°. With no barrier to intervene, such a movement is possible.3 2 43.6. Some Elusive Cases Let us now consider the cases where an “^WP c-commands a n o m WP and a clause boundary intervenes between them, as schematically represented below: (77) ... ^ W P ... [cp ... n o m WP ... 3 2 The non-barrierhood o f the scrambled CP in (76) is evidenced in the fact that overt extraction of an element out of the scrambled CP does not make any difference in grammaticality from the case in which an element is extracted out o f a non-scrambled CP. (See Saito 1985:181.) Compare the following two sentences: (i) a. [chayk-ul, [John-i [Mary-ka [Sam-i tj ilk-ess-ta-ko] malha-ess-ta-ko] book-Acc J.-Nom M.-Nom S.-Nom read-Pst-DE-C say-Pst-DE-C mit-nun-ta]] believe-Pres-DE ‘The book,, John believes that Mary said that Sam read tj.’ b. [chayk-ul, [John-i [[Sam-i tj iIk-ess-ta-ko]j Mary-ka tj malha-ess-ta-ko] book-Acc J.-Nom S.-Nom read-Pst-DE-C M.-Nom say-Pst-DE-C mit-nun-ta]] believe-Pres-DE ‘The bookj, John believes that Mary said that Sam read tj.’ Both sentences in (i) are grammatical. Notice that due to cyclicity, (ib) cannot be derived by scrambling of chayk-ul out o f the most deeply embedded clause in (ia) followed by scrambling the complement clause [Sam-i t, ilk-ess-ta-ko] to the initial position of the second most embedded clause. Thus a scrambled complement clause does not seem to bear any islandhood. Reproduced with permission of the copyright owner. Further reproduction prohibited w ithout permission. 197 According to the theory proposed here, the structure in (77) should be licit since the ^W P will be licensed by moving to the SPEC of a [+wh] C°, having an ESHR, and the n o m WP will be licensed in situ by having a BR with a nominal element that question clauses have, as represented below: (78) DP CP WHY; [+wh] ESHR CP BR Contrary to the expectation, such a structure is not allowed in Japanese and Korean, as exemplified below: (79) *naze John-ga [Mary-ga nani-o kattato] itta no? why J.-Nom M.-Nom what-Acc bought C said QE ‘Why did John say that Mary bought what?’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 198 (80) *wav John-i fMarv-ka mwues-ul sa-ess-ta-ko] malha-ess-ni? why J.-Nom M.-Nom what-Acc buy-Pst-DE-C say-Pst-QE ‘Why did John say that Mary bought what?’ Notice, however, that scrambling of the bracketed part over the ^W P makes the sentences in (79) and (80) good, as illustrated in (81) and (82) below: (81) [Mary-ga nani-o kattatojj naze John-gatj itta no? M.-Nom what-Acc bought C why J.-Nom said QE ‘Why did John say that Mary bought what?1 (82) [Mary-ka mwues-ul sa-ess-ta-ko]j way John-i tj malha-ess-ni? M.-Nom what-Acc buy-Pst-DE-C why J.-Nom say-Pst-QE ‘Why did John say that Mary bought what?’ The same contrast is attested when a n o m WP is contained in an island, as the following Korean sentences illustrate: (83) * John-i way [n p [c p nwu-ka cwu-n] os]-ul ani-ip-ni? J.-Nom why who-Nom give-Adn clothes Neg-wear-QE? ‘Q John does not wear [the clothes that who gave (to him)] why?’ (84) John-i [n p [Cp nwu-ka cwu-n] os]-uI way ani-ip-ni? J.-Nom who-Nom give-Adn clothes why Neg-wear-QE? ‘Q John does not wear [the clothes that who gave (to him)] why?’ Reproduced with permission of the copyright owner. Further reproduction prohibited w ithout permission. 199 Of course, the “‘ ’WP in (81), (82), and (84) cannot directly adjoin to the "“"WP since the latter does not c-command the former. I conjecture that the a d v WP is adjoined to the clause mate element containing the n o m WP. i.e.. the bracketed clause in (81) and (82) and the complex NP in (84), as schematically represented below:3 3 3 3 The answer pattern seems to indirectly support for the analysis. As an answer to the question in (84), (ib), but not (ia), is appropriate: (i) a. #Mary-ka, cak-ese-ya. M.-Nom small-because-DE ‘As for Mary, (it is) because it is small.’ b. [[Mary-ka cwu-n] os]-ul, cak-ese-ya. M.-Nom give-Adn clothes-Acc small-because-DE ‘As for the clothes that Mary gave (to him), (it is) because it is small.’ This contrasts with the answer patterns for the question without the * d v WP. As was discussed in Nishigauchi (1986), both a short (truncated) answer and a long answer are possible for this question: (ii) a. Mary-ya. DE b. [Mary-ka cwu-n] os-i-ya. M.-Nom give-Adn clothes-be-DE Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 200 I claim that the sentences in (79), (80), and (83) are ungrammatical because the a d v WP in those sentences is less economically licensed in comparison with the a d v WP in the sentences in (81), (82), and (84). 4.4. Conclusion In conclusion, the so-called anti-superiority effect (ASE) in languages such as Japanese and Korean is a natural consequence of economy of derivation under the following assumptions: (86) a. Scope licensing in terms of binding is less costly than scope licensing in terms of movement. b. ‘ “‘ ''WPs are licensed by having an ESHR with a [+wh] C° or by being adjoined to a ^ W P . c. Clause internal scrambling is literally optional and costless. Whenever an a d v WP has a clause mate n o m WP, the latter must c-command the former so that the a d v WP can be adjoined to the "“"WP in LF without having to move to the SPEC of a [+wh] C°. The a d v WP in such a context is parasitically licensed when the "“"WP is licensed by means of having a binding relation with the nominal element that question clauses have. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 201 In passing, I ascribe the existence of superiority effects in English type languages, and the (apparent) lack of them in the Korean type languages, to the theory that the interrogative force in question clauses is licensed by means of attracting a WP in the English type, but by means of attracting a predicate with a question conjugation in the latter type languages, as was discussed in chapter 2. Thus the [+wh] C° in the English type languages attracts the highest WP, nominal or non-nominal. In contrast, in the latter type of languages, a question conjugation applies to the predicate of the highest clause that the question scope ranges over. Thus the interrogative force of question clauses in both types of languages attracts the nearest possible WH-element, satisfying the economy condition. It is not obvious, however, why languages like English lack the option of identifying ^W P s by means of adjunction to a n o m WP. Note that in such languages a d v WPs are not allowed in situ. Additionally, it is obscure what causes the non-homogeneity of scope licensing of ^W Ps, as Joseph Aoun (p.c.) pointed out to me. For now I do not have any answers to these questions. I leave these questions open for future work. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 202 Chapter Five A Unified Interrogative Analysis of Korean WH-Phrases1 5.1. Introduction As is well known in the literature, some languages have a set of lexical items which are ambiguous between an interrogative reading and a quantifier reading, depending on the environment in which they appear. (See Kuroda 1965, Huang 1982. Li 1992, and Nishigauchi 1986, among others.) Korean belongs to this group of language, as discussed in Choi (1935), Park (1935), Chang (1976), C.-S. Suh (1985. 1990), C.-M. Suh (1987), Kang (1988), and Kim (1989a, 1989b, 1991), among others. WH-phrases (henceforth, WPs) in Korean are ambiguous between an interrogative reading and a non-interrogative quantifier reading depending on the environment they appear in. The latter reading, i.e., the non-interrogative quantifier reading, is further subdivided into two, an existential quantifier (henceforth 3-QP) reading and a universal quantifier (henceforth V-QP) reading. For example, nwukwu ‘who’ can convey an 1 A substantive part of this chapter was presented at the 1995 Harvard Workshop on Korean Linguistics at Harvard University and the 1995 Student Conference in Linguistics 7 at University of Connecticut. (The latter work is co-authored with Hong-Keun Park.) : There are some exceptional cases in which WPs are interpreted exclusively as interrogatives or exclusively as quantifiers. See section 5.6. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 203 interrogative reading ‘who,’ an 3-QP reading ‘someone,’ or a V-QP reading ‘everyone/anyone,’3 as in (1), (2), and (3), respectively:4 ,5 (1) interrogative readings a. nwu-ka o-ess{-nunka/-ci/-na}?6 who-Nom come-Pst-QE ‘Who came?’ b. na-nun [ku-ka mwues-ul kacyeo-ess {-nunka/-nunci/-na} ] molu-n-ta. I-Top he-Nom what-Acc bring-Pst-QE not;know-Pres-DE ‘I do not know what he brought.’ 3 See J. Suh (1990:33) for the argument that WPs in such a context have a free choice ANY reading rather than a universal reading. I keep using the term "universal,’ though, just for the reason that most linguists stick to the term. 4 The ambiguity of WPs will be ignored in English glosses. For example, nwukwu, though it could mean ‘who,’ ‘someone,’ or ‘everyone,’ depending on its environment, will be glossed as ‘who.’ 5 WPs in (1) may have non-interrogative readings also, (la), for example, could mean ‘Did anyone come?’ Presumably -inka, which converts a WP into an 3-QP, is deleted in such a case. Note that -inka in (2) could be deleted without changing the meaning of the sentences. 6 WP nwukwu ‘ who’ in the standard dialect is shortened to nwu when followed by nominative marker - ka but not when followed by other case markers like accusative, -lul, or dative, -eykey, or other particles like -to ‘also.’ (The short form is more productively used in Kyengsang dialects.) Case markers usually delete after 3-QPs and almost always after V-QPs. We will return to this aspect in section 5.52. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 204 (2) 3-QP readings a. nwukwu-inka o-ess-ta.7 who-INKA come-Pst-DE ‘Someone came.’ b. John-i mwues-inka mek-ko iss-ta. J.-Nom what-INKA eat-ing be-DE ‘John is eating something.’ (3) V-QP readings a. nwukwu-(ina/-ituncil cal sal-ki pala-n-ta. who-INA/ITUNCI well live-Nmz want-Pres-DE ‘Everyone wants to live well off.’ b. John-un mwues-inaAitunci mek-nun-ta. J.-Top what-INAAITUNCI eat-Pres-DE ‘John eats anything.’ Apparently interpretation of a WP depends on some other element in the sentence. At a certain level, every WP is associated with one of the particular set of morphemes which we call licensing morphemes (henceforth LMs), highlighted in (l)-(3). When properly related to LMs like -nunka, -ci, -na, -nunci, etc., which are question endings (henceforth QEs),8 WPs have an interrogative reading as in (1). When properly related 7 Vowel / of inka may be deleted after a vowel. This phonological variation, however, will be ignored in the transcription, unless otherwise specified. Similarly for -inaZ-itunci in (3). The fact that these morphemes start with the sound / is very important to our analysis, as will be clear later. See section 5.3. 8 There are many other question endings depending on the register and style of speech. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 205 to LM -(ijnka, WPs have an existential quantifier reading as in (2). 9 And when properly related to LMs like -(i)tunci, -(i)m , etc., WPs have a universal quantifier reading as in (3).1 0 Thus LMs seem to play a crucial role in determining the reading of WPs. C.-S. Suh (1990) analyzes Korean 3-QPs as indirect questions by decomposing the LM -inka into the copula verb i plus a QE -nka. 1 1 This amounts to saying that WPs in Korean 3-QPs are also interrogative elements which are associated with a QE. In this chapter I lend support to C.-S. Suh’s (1990) analysis, on the one hand, while arguing, on the other hand, that V-QPs can be similarly analyzed. In other words, I argue that every WP in Korean, interrogative or quantificational, is argued to contain a [+wh] feature to be licensed by Q. The organization of the chapter runs as follows. In section 5.2., C.-S. Suh’s (1990) interrogative analysis of 3-QPs will be introduced and supporting evidence will be provided. In section 5.3., it will be argued that V-QPs are similarly analyzed. In section 5.4., there will be a discussion as to the syntactic properties of the Korean copula verb /, to further confirm that 3-QPs and V-QPs involve a question clause projected from the 9 In addition to -inka, -inci functions as an 3-QP licensing morpheme, as exemplified below: (i) mwues-inci kkelimchikha-ta. what-INCI feel;uneasy-DE ‘Something is uneasy.’ 1 0 There are also other universal licensing morphemes, e.g. -(i)ken a , -(i)lato, -(i)lcilato, etc. 1 1 As far as I know. Park (1935) first claims that existential quantifiers involve interrogation. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 206 copula verb / which is contained in the so-called existential and universal licensers such as -(i)nka/-(i)nci/-(i)na/-(i)tunci, etc. In section 5.5., I tackle an immediate question that may arise with the across-the-board question analysis of Korean WPs, i.e., if WPs are interrogative elements across the board, then how are the reading variations derived? I suggest that it is not the LMs, i.e., QEs, that are responsible for the reading variations, but rather the semantic properties of (abstract) predicates that are required to license the question clauses projected from the QEs. It will be observed that predicates that co occur with V-QPs are restricted to negative relational predicates when they are overtly realized. The negative relational predicates semantically operate as an identity function. Thus regardless of what is chosen for the WP contained in V-QPs, the semantic value of the matrix clause is unaffected. This produces the so-called universal reading of the phrases. In section 5.6. I will discuss an alternative view of Korean WPs which characterizes Korean WPs as indefinite quantifiers. (See Chang 1976, C.-M. Suh 1987, and Kim 1989a, 1989b, 1991, among others.) It will be concluded that the indefinite quantifier view of Korean WPs across the board is incorrect despite a few attractive points that it might suggest. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 5.2. 3-QPs as Reduced indirect Questions 207 5.2.1. C.-S. Suh (1990): Decomposition of -inka into -i and -nka On the surface, Korean 3-QPs seem to function as simple nominal arguments just as their English counterparts. Consider the sentences in (2), repeated below: (2) a. nwukwu-inka o-ess-ta. who-INKA come-Pst-DE ‘Someone came.’ b. John-i mwues-inka mek-ko iss-ta. J.-Nom what-INKA eat-ing be-DE ‘John is eating something.’ WP nwukwu followed by inka reads as ‘someone’ in (2a) and mwues followed by inka reads as ‘something’ in (2b). Nwukwu-inka in (2a) and mwues-inka in (2b) apparently function as a subject and as an object, respectively. A closer examination, however, reveals that the 3-QPs based on WPs have a more complex internal structure as will be argued in the following. C.-S. Suh (1990:251) analyzes such 3-QPs as indirect questions embedded under a relative clause.1 2 1 2 All the paraphrasing examples C.-S. Suh (1990) gives have the relative clause construction, although he is not explicit about whether the indirect question clause in 3-QPs is embedded under a relative clause. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 208 According to this analysis, (2a) will have the elaborate structure in (4), where Q-Pred stands for a question selecting predicate: (4) [Npfcpie; [creej nwukwufc-i-nka] Q-Pred-Rel] ei] o-ess-ta. who-be-QE come-Pst-DE ‘[(A mam who Ij do not know) whok (hej) was] came.’ In (4) the so-called existential licensing morpheme -inka is decomposed into the copula verb / plus a QE -nka.1 3 The question clause projected from this QE, i.e., CP2, is selected by an abstract question selecting predicate indicated by Q-Pred. Three null subjects are posited: e; as the subject of the Q-Pred, ej as the subject of the copula i. and ei as the subject of the matrix verb o ‘to come.’ CPI as a relative clause modifies the empty head noun ei. Since ej and ei are a gap and the head noun in relativization, index j 1 3 C.-S. Suh (1990) gives a partial credit to C.-M. Suh (1987) for the decomposition of the LM for 3- QPs, i.e., inka, into copula verb -i plus question ending -nka. C.-M. Suh (1987:220-221) analyzes 3-QPs as having a complex structure. For example, (ia) below is analyzed as (ib): (0 a. nwukwunka-ka ttek-ul mwuk-ess-na? (C.-M. Suh 1987:220) someone-Nom rice;cake-Acc eat-Pst-QE ‘Did anyone eat the rice cake?’ b. lei nwukwu-i-nkal epka ttek-ul mwuk-ess-na? who-be-NKA Nom rice;cake-Acc eat-Pst-QE The underlined part in (ib) may raise a few questions. First, since there is no predicate that selects the question clause, it is not clear how the question clause is licensed. Second, since morphological case is never realized when an NP has no phonological matrix, it is not clear how the nominative case is realized in (ib). Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 209 and index 1 are identical. Index j and index k are also identical since they are related by the copula verb /. Therefore, ej, nwukwu*, and ei co-vary in their interpretation. 5.22 . C.-S. Suh’s (1990) Motivations C.-S. Suh (1990) motivates structure (4), based on several perspectives. Among them I repeat the following four, which are relevant to the present discussion. First, the existential licensing morpheme -inka with the initial sound i stripped, i.e., -nka, can function as a QE in obvious indirect questions, as exemplified in (5): (5) na-nun [nwu-ka o-nu-nkal kwungkumha-ta. I-Top who-Nom come-Asp-QE be;wondrous-DE ‘I am wondering who is coming.’ The bracketed part in (5) is obviously a question complement clause of Q-Pred kwungkumha ‘to be wondrous.’ Note that the question clause ends with -nka. Second, Q-Pred and other null elements in (4) can be overtly realized, as exemplified in (6): (6) |NpfrpinaVj-ka frwej nwukwut-i-nka] molu-nunl salanu-i] o-ess-ta. I-Nom who-be-QE not;know-Rel man-Nom come-Pst-DE ‘[A manj who Ii do not know who* (hej) was] came.’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 210 Q-Pred is realized as molu ‘not to know,’ its subject as nay-ka ‘I-Nom,’ and the head noun of the relative construction as salam-i ‘man-Nom.’ Third, WPs were used exclusively as interrogative terms in earlier Korean. They had not been used as quantifiers until the end of the 19th century. Instead amwu 'any* plus noun has been continuously used as an indefinite NP. Fourth, the diversity of interpretations that existential quantifiers may have can be reduced to the fact that Q-Preds, being unexpressed are compatible with various question selecting predicates. See section 5.5.2. for further discussion. 5.2 J. Additional Motivations To further motivate the indirect question analysis of 3-QPs, 1 discuss the following three aspects of 3-QPs in this section; scope barrierhood of the so-called existential licensing morphemes, honorific si insertability in 3-QPs, and formal characteristics of 3-QPs in some dialects in the Kyengsang area. Let us first notice a parallelism between QEs in obvious questions and the so- called existential licensing morphemes. They both function as scope barriers for WPs. As observed by Lee (1982), a WP cannot be scopally linked to a QE skipping a nearer QE, as schematically represented in (7a) and exemplified in (7b): Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 211 (7) a. * ...[ c p - [c p - WP1 ... W P2... QE1]... QE2J... b. John-i [nwu-ka Marv-evkev mwues-ul cwu-ess-nun-ci] al-ni? J.-Nom who-Nom M.-Dat what-Acc give-Pst-Asp-QE know-QE ‘Does John know who gave what to Mary?’ # ‘for which thing y, John knows for which person x, x gave y to Mary?’ In the structure (7a), the reading where WP1 is linked to QE1 and WP2 is linked to QE2 is not available, as the sentence in (7b) means the first reading but not the second in the English translations. In other words, the WP in the embedded object position cannot have a matrix reading when the WP in the embedded subject position has the embedded reading.1 4 Thus, question projections function as a scope barrier for WPs inside them, as opposed to other syntactic islands such as complex noun phrases, adjunct clauses, subject clauses, etc. Existential licensing morphemes, -inkal-inci, are very similar to QEs in the obvious questions in this respect. Consider the following example: 1 4 This sentence seems to have a reading where the subject WP has the matrix scope and the object WP has the embedded reading. If this is true, then could we say that the WH-isiand effect is inert in Korean? It is not necessarily guaranteed, since this reading may be obtained when the embedded subject has undergone (string vacuous) scrambling to the matrix clause. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (8) nwu-ka mwues-inka mek-ess-ni? who-Nom what-INKA eat-Pst-QE 212 ‘Who ate something?’ # ‘for which person x, for which thing y, x ate y?’ As the English translations indicate, the WP, mwues ‘what,’ cannot be linked to the matrix QE ni skipping -inka. The lack of the matrix reading of what in (8) indicates that the so-called existential licenser -inka functions as a scope barrier for WPs just like regular QEs, supporting the indirect question analysis of 3-QPs in Korean. Let us turn to the second aspect of Korean 3-QPs, i.e., honorific si insertability. Note that i in -inka can host the subject oriented honorific morpheme si as in (9) below: (9) yelepwun-cwung nwukwu-i-si-nka ka-si-eya ha-pnita. you(PI)-among who-be-Hon-QE go-Hon-must do-DE ‘Among you (PI), someone (Hon) must go.’ Note also that the honorific morpheme si otherwise attaches only to a verb stem, as shown in (10) below: (10) sensayngnim*-kkeyse* ecey* Seoul*-ey* ka-si-ess*-ta*. teacher-Nom(Hon) yesterday Seoul-to go-Hon-Pst-DE ‘(My) teacher went to Seoul yesterday.’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. The honorific morpheme si can be attached to the verb stem ka ‘to go/ but not to any of the positions marked with a star. This indicates that the underlined part in (9) involves a verbal, hence clausal, projection as indirect questions do. Finally, a piece of support for the indirect question analysis can be drawn from the structure of 3-QPs in some dialects in the Kyensang area. 3-QPs in the dialect spoken in Changwon are composed of a WP plus -nka or -n/co, as in the following example, which is cited from C.-M. Suh (1987:219): (11) Yengi-ka nwu-fnka/nkol-lul cohaha-na? Y.-Nom who-NKA/NKO-Acc like-QE ‘Does Yengi love anyone?’ What are the particles -nka/nko? To answer this question, it is worth paying attention to the types of QEs in this dialect Kim (1982) observes that there are four types of QEs in Kyengsang dialects, -ka, -ko, -na, and -no, as shown in the following examples cited from Kim (1982:108): (12) a. Minho-ka pancang-i-ka? Minho-Nom representative-be-QE(yes/no) ‘Is Minho the representative (of the class)?’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 214 b. nwu-ka pancang-i-ko? who-Nom representative-be-QE(wh) ‘Who is the representative (of the class)?’ c. Minho-ka Suni-hanthey kkoch-ul cwu-ess-na? Minho-Nom Suni-Dat flower-Acc give-Pst-QE(yes/no) ‘Did Minho give the flower to Suni?’ d. Minho-ka nwukwu-hanthev kkoch-ul cwu-ess-no? Minho-Nom who-Dat flower-Acc give-Pst-QE(wh) ‘To whom did Minho give the flower?’ -Ka is a yes/no QE for the copula verb i, -ko is a wh QE for the copula verb -na is a yes/no QE for non-copula verbs, and -no is a wh QE for non-copula verbs. With this variation in mind, let us reconsider the structure of 3-QPs in (11). The fact that the underlined part in (11) contains a QE -ka or -ko clearly indicates that 3-QPs contain a copula (abstract in this case probably for a phonological reason) and that they involve a question operation, as illustrated in (13) below: (13) Mary-ka nw u-<j)-(nka/nko}-lul cohaha-na? M.-Nom who-BE-QE(yes/no)/QE(wh)-Acc like-QE(yes/no) ‘Does Mary love anyone?’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 215 If Q-Pred and other null elements are provided as discussed in section 5.2.1.. the structure of the 3-QP in (13) will be like that in (4).1 5 To sum up, 3-QPs are analyzed as having a complex NP structure in which a question clause is embedded under a relative clause which modifies an empty head noun, as schematically represented in (14):1 6 1 5 It is still mysterious, however, that yes/no question ending -ka is used as well. 1 6 If the DP analysis of question clauses given in chapter 3 is right, then a DP should be added over CP: Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 216 (14) NP CP, NP SPEC C e, IP C SPEC I’ Rel VP I NP V’ e, CP2 V SPEC C’ Q-Pred IP C SPEC I’ QE(=[+wh]) VP I SPEC V’ ej WPk V I BE In (14) CPi as a relative clause modifies a null head noun, e,. CP2 is a question complement clause selected by a null Q-Pred. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 217 53. V-QPs as Indirect Questions V-QPs in Korean consist of a WP plus -inal-itimci, etc, as exemplified in (3). which is repeated as (15) below: (15) a. nwukwu-lina/tuncil cal sal-ki pala-n-ta. who-INA/ITUNCI well live-Nmz want-Pres-DE ‘Everyone wants to live well off.’ b. John-un mwues (-ina/-itunci1 mek-nun-ta. J.-Top what-INA/-ITUNCI eat-Pres-DE ‘John eats anything.’ I analyze V-QPs as indirect questions embedded under an adverbial clause. For instance, (15b) is analyzed as (16) below, where -ina/-itunci are decomposed into the copula verb i plus a QE -na/-tunci and a Q-Pred is postulated to license the question complement clause CP2: (16) John-un [me; [cree, mwuesr iI-na/-tunciI 1 Q-Pred-ADV] ei mek-nun-ta. J. -Top what-be-QE eat-Pres-DE ‘John, (irrespective ofT I do not care) what (it) is, eats (it).’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 218 C.-S. Suh’s (1990) as well as my first two arguments for the indirect question analysis of 3-QPs apply to V-QPs, too. Let us discuss them in turn. First, -net and -ci in -tunci can act as a QE of question complement clauses, as in (17): (17) na-nun [John-i nwukwu-lul {salangha-na/salangha-nun-ci} al-n-ta. I-Top J.-Nom who-Acc love-QE/love-Pres-QE know-Pres-DE ‘I know who John loves.’ Second, Q-Preds can be overtly realized, as in (18) below, where sangkwaneps-i 'regardless o f is inserted: (18) John-un [mwues-i-na] sangkwaneps-i cal mek-nun-ta. J.-Top what-be-QE notyelated-ADV well eat-Pres-DE ‘John eats it well regardless of what it is.’ Third, just as QEs in the obvious questions, the universal licensing morphemes, -inal-itunci, function as scope barriers, as exemplified in (19): (19) nwu-ka [ej [e * mwues-ul ha-nun] salarm-i- (na/-tunci \ 1 e; cohaha-ni? who-Nom what-Acc do-Rel man-be-QE like-QE (roughly) ‘Who likes anyone irrespective of what he does?’ # ‘for which person x, for which thing y, x likes men who do y?’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 219 As the English translations indicate, the WP in the relative clause, mwues ‘what,’ cannot have the matrix reading due to the intervening elements, -nal-hmci, which I analyze as QEs. Note in passing that -na could mean ‘only.’ In this case it does not function as a QE and therefore it does not block the matrix reading of WP mwues, as in (20): (20) nwu-ka Imwues-ul ha-nun] salam-i-na cohaha-ni? who-Nom what-Acc do-Rel man-be-only Iike-QE ‘for which person x, and for which thing y, x likes only men who do y?’ But crucially, if an overt Q-Pred is inserted, the WP necessarily has a narrow scope, that is, a V-QP reading, as in (21): (21) nwu-ka [mwues-ul ha-nun] salam-i-na sangkwaneps-i cohaha-ni? who-Nom what-Acc do-Rel man-be-QE not;related-ADV like-QE ‘Who likes anyone irrespective of what he does?’ # ‘for which person x, for which thing y, x likes men who do y?’ # ‘for which person x, for which thing y, x likes only men who do y?’ The existence of the Q-Pred requires that -na be interpreted as a QE. Therefore, the third reading, in which -na reads as ‘only,’ is unavailable. The second reading is not Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 2 2 0 available either, since the WP what cannot be linked to the matrix QE across the embedded QE. Fourth, the honorific morpheme si can be inserted, as in (22): (22) yelepwun-un nwukwu(-i)-si-tunci hwanyengha-pnita. you(Pl)-Top who-be-Hon-QE welcome-DE 'We welcome any of you.’ Recall that si attaches only to a verb in other contexts. Thus V-QPs have a clausal structure as indirect questions do. Before finishing this section, I point out that the initial sound / of the universal licensing morphemes -ina/-itunci, etc. (or the existential licensing morpheme -inka/- incf) is not an epenthetic vowel, but a copula verb. The initial sound i may or may not be overtly realized apparently depending on the sound value of the preceding sound. In general, the initial sound / of the licensing morphemes does not appear when the preceding stem ends with a vowel, e.g., nwukwu-nka, nwukwu-na/-tunci, etc. It is, therefore, very tempting to analyze the initial sound i of the universal (or existential) licensing morphemes as an epenthetic vowel that appears when the preceding stem ends with a consonant. The epenthesis analysis of the vowel /, however, is untenable in several respects. First, the vowel u is used as the epenthetic vowel in all other contexts in Korean. Second, retention of the sound is more or less acceptable, Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 2 2 1 although forms with it sound archaic and therefore less natural compared with the forms without it Third, as discussed earlier, / in the universal (and existential) licensing morphemes shows verbal properties. The analysis of / in the universal licensing morphemes as the copula verb can be indirectly supported by the fact that i does not appear when the licensing morphemes are attached to a content verb and a WP is located within the clause projected from the verb. Consider the following sentences: (23) a. [pro eti-lul ka-tuncil pro cosimha-ela. where-Acc go-QE be;careful-Imp ‘Be careful wherever you go’ b. [pro mwues-ul mek-tunci] pro cosimha-ela. what-Acc eat-QE be;careful-Imp ‘Be careful whatever you eat’ The bracketed parts in the examples already include a predicate and need not (and must not, by the economy of representation) have the copula verb i, which, I assume, appears as a default or pleonastic verb only when no other verb is available. Note that i does not appear in the universal licensing morpheme when the licensing morpheme is attached to a content verb, whether the verb stem ends with a vowel, as in (23 a), or a consonant, as in (23b). This renders untenable the epenthesis analysis of the initial Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 2 2 2 sound / of the existential and universal licensing morphemes mentioned in the previous paragraph. To sum up, I have argued in this section that so-called V-QPs can also be analyzed as indirect questions which are embedded under an adverbial clause. 5.4. Properties of the Copula / and their Relations to 3-QPs and V-QPs We have argued so far that despite the surface appearance, the so-called Korean 3- QPs and V-QPs are not simple NPs like English quantifiers of the form some N or every/all N, but have a richer structure containing a question clause. In this section, it will be argued that the apparent NP-like properties of quantifiers come from the syntactic characteristics of the copula verb contained in Korean quantifiers. Three major properties of the Korean copula verb /' and their bearings on the syntactic characteristics of 3-QPs and V-QPs will be discussed. 5.4.1. Post-Vowel Deletability The Korean copula verb / tends to be deleted if the preceding complement ends with a vowel, as was indicated earlier.1 7 Consider the following sentences: 1 7 The deletion o f i after a vowel is just a tendency. Thus retention of i even after a vowel does not make the phrase totally unacceptable, as mentioned earlier. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 223 (24) a. John-un haksavng-i-ta. J.-Top student-be-DE ‘John is a student’ b. John-un uvsa-< l> -ta. J.-Top doctor-be-DE ‘John is a doctor.’ In (24a), the complement, haksayng ‘student,’ ends with a consonant. Therefore the copula verb i remains unaffected. In (24b), the complement, uysa ‘doctor,’ ends with a vowel. Thus the copula verb i is deleted. We see the same variation in 3-QPs and V-QPs, as mentioned in section 5.3. When a WP ends with a consonant i remains unaffected as in mwues-i-nka ‘something,’ mwues-i-na/mwues-i-tunci ‘everything/anything,’ etc. When a WP ends with a vowel, / generally deletes as in nwukwu-0-nka ‘someone,’ nwukwu-tp- nalnwukwu-<fhtunci ‘everyone/anyone,’ etc. 5.4.2. Categorial Selection: NP or PP as a Complement The Korean copula verb i takes an NP or a PP as its complement An NP complement is more natural than a PP complement as exemplified below: Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 224 (25) a. John-un haksavng-i-ta. J.-Top student-be-DE ‘John is a student’ b. [John-i ecey camca-n kes-un] i pang(?-evseVi-ta.1 8 J.-Nom yesterday sleep-Adn thing-Top this room-in-be-DE ‘It is in this room that John slept yesterday.’ c. [John-i nwun-ul ttu-n kes-un] savnvek nev si (?-ev)-i-ess-ta. J.-nom eye-Acc open-Adn thing-Top dawn four o’cIock-at-be-Pst-DE ‘It was at four in the morning that John woke up.’ VPs and pure non-nominal adverbials are not allowed as complements of the copula verb i. Consider the following examples: (26) a. * [John-i ha-n kes-un] mek-(i)-ta.1 9 J.-Nom do-Adn thing-Top eat-(be)-DE ‘What John did was eating.’ b. * [ku salam-i kippeha-n kes-un] mavwu-i-ess-ta. the man-Nom feel;happy-Adn thing-Top very-be-Pst-DE ‘It was very much that the man was happy.’ '* As far as I know, no postposition in Korean ends with a consonant And PPs appear as complements of copula only in cleft-like constructions. 1 9 For this sentence to be grammatical, the defied verb has to be nominalized, e.g., mek-ki or mek-rwn kes, etc. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 225 Some derived adverbials are allowed in the cleft position. For example, adverbials formed by adding -key to a verb seem to be allowed as complement of the copula i in the cleft-like construction: (27) ??[John-i talli-n kes-un] olimphik senswu-pota ppallu-kev-i-ess-ta. J.-Nom run-Adn thing-Top Olympic player-than quickly-be-Pst-DE ‘How John ran was faster than an Olympic player.’ It was suggested in chapter 3 (footnote 11) that -key in W-key be decomposed into the nominalizer -ki and the postposition -ey meaning ‘in.’ Thus, key adverbials can be reanalyzed as PPs. Since the clefted part is a PP in (27), it could be a complement of the copula verb /. With the above facts in mind, let us now turn to the formal characteristics of 3- QPs and V-QPs. The same categorial restrictions seem to apply to 3-QPs and V-QPs. as the following examples indicate: (28) a. nwukwu-<|>-nka ‘someone’ who b. mwues-i-nka ‘something’ what c. eti(-ey)-<t>-nka ‘somewhere’ where d. encey(?-ey)-<i>-nka ‘sometime’ when Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 226 e.??etteh-key-<(>-nka how f. *w ay-< {>-nka why (29) a. nwukwu-<j>-na who b. mwues-i-na what c. eti(-ey)-<|>-na where e.???etteh-key-<|>-na ‘ how f. *w ay-<f>-na why ‘somehow’ ‘*somewhy’ ‘everyone/anyone’ ‘ everything/anything’ ‘ everywhere/anywhere ’ d. encey(?-ey)-<(y-na ‘every time’ when *everyhow/anyhow’ everywhy/anywhy’ Nominal WPs like nwukwu and mwues can most freely occur with the existential or universal licensing morphemes. WPs like eti-(ey) and encey(-ey) are PPs containing an NP and they also can host (i)nka or (i)na as in (28c, d) and (29c, d).2 1 WP ettehkey 2 0 This form is worse than ettehkey-nka in (28e), but better than way-nka in (280 and way-na in (290- 2 1 C.-S. Suh (1990:236) considers encey ‘when’ as a short form of encey-ey ‘when-at’ -ey ‘at’ is probably dropped for some phonological reason, e.g., for haploiogy. Note that -ey ‘at’ has the same sound as the vowel of the last syllable of encey. In feet -ey ‘at’ in eti-ey is also frequently deleted. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 227 "how’ can occur with them, although not perfectly, as in (28e) and (29c).2 2 The adverbial WP way 'why’ cannot occur with them. Some qualification is in order with respect to the acceptability of ettehkey-tunci ‘no matter how’ and way-nci ‘for some reason.’ It seems that the former is formed by adding a universal licensing morpheme -tunci to ettehkey ‘how,’ and the latter is formed by adding an existential licensing morpheme -nci to way ‘why.’ Both of the forms are fully acceptable like other V-QPs and 3-QPs of the same forms, e.g., mwues-i-tunci ‘what-be-QE: whatever’ and mwues-i-nci ‘what-be QE:something.’ The question is why they are fully acceptable unlike ettehkey-nka, ettehkey-na, way-nka, and way-na. which are marginally acceptable or unacceptable, as was seen in (28e), (28f), (29e), and (29f), respectively. I conjecture that these expressions are formed differently from other 3-QPs and V-QPs of the similar forms, e.g., mwues-i-tunci ‘what-be-QE’ and mwues-i- nci ‘what-be QE.’ Specifically, I speculate that ettehkey-tunci is a short form of ettehkey ha-tunci ‘how do-QE’ and way-nci from way kuleha-nci ‘why be;such;and;such-QE.’ A piece of evidence that these expressions do not involve the copula / can be drawn from the fact that / is not acceptable, not even marginally: *ettehkey-i-tunci and *way-i- nci. Recall that in other cases i is marginally acceptable even when WP ends with a vowel, e.g., eti-ey-i-nka ‘where-at-be-QE, encey(-ey)-i-nka ‘when(-at)-be-QE,’ although 2 2 C.-S. Suh (1990:245) takes both ettehkey-i-nka and way-i-nka as equally well-formed. But I sense that they are qualitatively different Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. it is better without i, eti-ey-< j> -nka ‘where-at-be-QE.’ encey(-ey)-<p -nka ‘when(-at)-be- QE.' In brief, ettehkey ‘how’ in ettehkey-tunci ‘no matter how’ and way ‘why’ in way- nci ‘for some reason’ are not complements of the copula i. Therefore the selectional restriction that applies to the complement of the copula verb does not apply to them. 5.43. Incorporation The Korean copula verb i obligatorily incorporates the head noun of its complement as witnessed by the fact that the copula i and the head noun of its complement are always put together.2 3 That incorporation takes place is evidenced in the fact that no case can be attached to the complement of the copula verb /, as in the following examples: (30) a. John-i haksayng(*-i)-i-ta. J.-Nom student-Nom-be-DE ‘John is a student.’ b. John-i uysa(*-ka)-<{>-ta. J.-Nom doctor-Nom-be-DE ‘John is a doctor.’ 2 3 When PP is a complement, the amalgamation of N-P is incorporated into the copula. Note that N, the head of the complement NP of a P is always incorporated into P. (Or P always lowers to N by a version of affix lowering.) Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 229 A nominative case, either / or ka depending on the preceding sound, cannot be attached to the complement of the copula whether the copula is overt or non-overt. This contrasts with other transitive predicates which assign accusative or nominative cases to the complement, as in the following examples:2 4 (31) a. John-i Mary-lul cohaha-n-ta. J.-Nom M.-Acc like-Pres-DE ‘John likes Mary.’ b. John-i Marv-ka coh-ta. J.-NomM.-Nom fond-DE ‘John is fond of Mary.’ (32) a. John-i ton-ul kaci-ess-ta. J.-Nom money-Acc have-Pst-DE ‘John has money.’ b. John-i ton-i iss-ta. J.-Nom money-Nom have-DE ‘John has money.’ The two verbs in (31), cohaha ‘like’ and coh ‘fond,’ have the same thematic structure, i.e., they both have thematic roles, Experiencer and Theme. Similarly the two verbs in (32), kaci ‘have’ and iss ‘possess,’ have the same thematic structure, i.e., they both have thematic roles, Possessor and Possessee. All the verbs in (31) and (32) bear two 2 4 Case markers can be dropped in these examples, which indicates that such transitive verbs allow incorporation optionally. See Saito (1982) for the similar data in Japanese. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 230 argument positions. The only difference between the verbs in (a) and those in (b) is that the former assign accusative case, but the latter do not2 5 The copula verb / is also a two place predicate, i.e., a transitive predicate, but no case is realized on the second argument as was seen in (30). This indicates that the complement is incorporated into the copula verb, which obliterates the realization of a morphological case on the complement2 6 Another indication that complements of the copula verb i are incorporated into the verb is that complements cannot be separated from the copula verb by any syntactic operation. For example, adverbs cannot intervene between the copula i and its complement or the complement cannot move by scrambling or relativization: 2 5 Kang (1986) claims that the nominative case in (b) sentences is a default case which is realized when no other case assignment takes place. 2 6 Negative copula ani ‘not to be,’ which I conjecture is shortened from ani-i ‘not-be’ due to a haplology, does not require its complement to be incorporated. A nominative case can remain in this context: (i) John-i {haksayng-i/uysa-ka} ani-ta. J.-Nom {student-Nom/doctor-Nom} not;be-DE ‘John is not {a student/a doctor}.’ The complement o f the negative copula verb, however, is hardly extractable. For example, scrambling the complement over the subject results in a very awkward sentence: (ii) *?{haksayng-i/uysa-ka}j John-i t, ani-ta. {student-Nom/doctor-Nom} J.-Nom not;be-DE ‘John is not {a student/a doctor}.’ Although (ii) is slightly better than the cases where the complement of a positive copula verb is extracted, it is still hardly acceptable. For now 1 cannot account for the low acceptability of the sentences like (ii). Maybe a parsing problem is involved due to the multiple occurrence of nominative marked phrases. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (33) a. * John-i haksavng cikum i-ta. J.-Nom student now be-DE ‘John is a student now.’ b. *hasayngj John-i tj-i-ta. student J.-Nom be-DE ‘A student, John is.’ c. * [John-i ej-i-n] haksayngj J.-Nom be-PNE student ‘a student that John is’ This contrasts with other transitive verbs whose complements are separable from verbs as shown in the following examples: (34) a. John-i Marv-lul cikum cohaha-n-ta. J.-Nom M.-Acc now like-Pres-DE ‘John likes Mary now.’ b. Mary-luli John-i tj cohaha-n-ta. M.-Acc J.-Nom like-Pres-DE ‘Mary, John likes.’ c. [John-i tj cohaha-nun] Mary J.-Nom like-Adn M. ‘Mary, who John likes’ (35) a. nay-ka sakwa-ka mavwu coh-ta. I-Nom apple-Nom very be;fond-DE ‘I am very fond of apples.’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 232 b. sakwa-ka j na-nun tj maywu coh-ta.2 7 apple-Nom I-Top very be;fond-DE * 1 am very fond of apples.’ c. [nay-ka maywu coh-un] sakwa I-Nom vexy be;fond-Adn apple ‘Apples that I am very fond o f As in (34a) and (35a), an adverb can intervene between a transitive verb and its complement And the complement can be scrambled over the first argument as in (34b)/(35b) and can be relativized as in (34c)/(35c). The obligatory incorporation property that the copula i has is related to the surface forms that 3-QPs and V-QPs take. In other words they look like simple NPs on the surface level since the copula i incorporates WPs. 5.5. Some Queries In this section, I discuss two questions that may arise as to the indirect question analysis of 3-QPs and V-QPs. First, why is it that Q-Preds can be non-overt in Korean? Second, how do we discern 3-QPs from V-QPs if they are both indirect questions? 1 1 The sentence sounds less acceptable with a nominative subject. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 5.5.1. Deletabillty of Q-Preds I would say that the first question should be raised independently of the internal structure of quantifiers because Korean grammar allows deletion of Q-Preds anyway, even for obvious indirect WH-questions. For example, in sentences like (36) below, the Q-Pred molu-ato ‘not;know-though’ can be freely deleted. (36) John-un [mwues-ul ha-nu-nka] (molu-ato) ton-ul cal pel-n-ta. J.-Top what-Acc do-Asp-QE not;know-though money-Acc well eam-Pres-DE 'John earns a lot of money, (although I do not know) what (he) is doing.’ Even in English, the deletion of Q-Preds is attested although it is less productive. Typically, Q-Preds like irrespective o f or regardless o f can be deleted if the complement clause conveys a concessive reading, as exemplified below: (37) (Regardless of) whether or not you agree, I’ll do that No matter what is responsible for the deletion of Q-Preds in obvious WH- questions, the deletion of Q-Preds in Korean quantifiers can be analogously explained once such quantifiers are analyzed as containing indirect questions. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 5.5.2. Distinction of 3-QPs from V-QPs 234 Now let us consider the second question, that is, how are 3-QPs and V-QPs distinguished if they are both indirect questions? There are at least two differences, one is related to syntactic properties, and the other to semantic interpretation. First, in general 3-QPs can bear a case particle, whereas V-QPs cannot. Case particles in 3-QPs are more or less optional as exemplified in (38): (38) a. nwukwu-i-nka(-ka) o-ess-ta. who-be-QE-Nom come-Pst-DE ‘Someone came.’ b. John-i mwues-i-nka(-lul) mek-ess-ta. J.-Nom what-be-QE-Acc eat-Pst-DE ‘John ate something.’ However, case particles in V-QPs make the sentence very marginal or unacceptable, as we see in (39): (39) a. nwukwu-<j>-(naf??-kaVtuncif*-ka)\ John-ul cohaha-n-ta. who-be- {QE-Nom/QE-Nom} J.-Acc like-Pres-DE ‘Anyone likes John.’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 235 b. [nwukwu-uy nonmwun-i-{ na/tunci} (*-ka)] chwulphan-toi-ess-ta. who-Gen paper-be-{QE/QE}-Nom publish-Pass-Pst-DE ‘Anyone’s paper was published.’ c. John-i m wues-i- (na/tunci 1 (*-lul) mek-nun-ta. J.-Nom what-be-{QE/QE}-Acc eat-Pres-DE ‘John eats anything.’ Generally, case particles are disallowed in V-QPs. Exceptionally, V-QP nwukwu-na ‘anyone’ seems to be able to bear a nominative case particle -fez as in (39a), but I suppress this peculiar case since further embedding of the WP leads to an ungrammatical sentence, as in (39b). Why do 3-QPs and V-QPs display this contrast? I attribute the contrast to the difference in their internal structures. Recall that C.-S. Suh (1990) analyzes 3-QPs as indirect questions embedded under a relative clause, and that I analyze V-QPs as indirect questions embedded under an adverbial clause. Compare, for example, the internal structure of an 3-QP in (4) with that of a V-QP in (16), which are repeated below: (4) [Np[cpiei [crc> ej nwukwu^-i-nkal Q-Pred-Rel] ei] o-ess-ta. who-be-QE come-Pst-DE ‘[(A mani who Ij do not know) who* (hej) was] came.’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 236 (16) John-un [cpie* [cP 2ej mwuesv-i (-na/tunci \ 1 Q-Pred-ADV] ei mek-nun-ta. John-Top what-be-QE eat-Pres-DE ‘John, (irrespective of/l do not care) what* (itj) is, eats (it|).’ Relative clauses modify a nominal element but adverbial clauses cannot for reasons related to subcategorization restrictions.2 8 Thus CPI and ej in (4) form a constituent, that is, an NP; however, CPI and ej in (16) do not form a constituent since CPI cannot modify a nominal element ei. Now the difference between 3-QPs and V-QPs with respect to case bearability can be explained with the assumption given in (40): (40) [n p ... a...] can get a case, i f a is overt. A nominal projection is able to receive case as long as some element in it is overtly realized. A crucial assumption here is that the realized part, a in (40), need not be the head of the nominal expression. Both in (4) and in (16), CPI cannot receive case since it is not a nominal projection. Furthermore, ei cannot receive case either, since it is not an overt element In (4), CPI plus ei, which constitute a nominal projection, NP, can receive case since part of the nominal projection, that is, CPI, is overtly realized. In 1 8 Adverbials are not allowed in the local domain of a nominal head: beautifid(*fy) scenes. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 237 contrast, in (16) the option of merging CPI and ei is not available due to the aforementioned subcategorization restrictions. Thus no case can be attached in (16). Let us now turn to the question of how we obtain the semantic difference between 3-QPs and V-QPs. A general position taken in the literature is that a WP plus -inka gives rise to an 3-QP reading and a WP plus -inal-itunci gives rise to a V-QP reading. Kuroda (1965) and Nishigauchi (1986) argue along these lines for Japanese quantifiers. But such a position is no more than a mere stipulation. The indirect question analysis of 3-QPs and V-QPs is not compatible with such an explanation. This is because, in this analysis, the so-called 3-QP licenser -inka and V- QP licensers -inal-itunci are taken as copula verbs with a question conjugation. It is dubious that a difference in question conjugations directly leads to the existential vs. universal reading difference. Thus, for the indirect question analysis of quantifiers to be viable, an alternative explanation must be provided. I suggest that the variation between an 3-QP reading and a V-QP reading depends on the choice of Q-Preds. It seems to be the case that Q-Preds for 3-QPs differ from those for V-QPs. For instance, compare (41) and (42): (41) nwukwu-i-nka (molu-nun salam-i/*sangkwaneps-i} o-ess-ta. who-be-QE not;know-Rel man-Nom/not;related-ADV come-Pst-DE ‘Someone came.’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 238 (42) John-un mwues-ina {sangkwaneps-i/* molu-nun mwulken-ul} mek-nun-ta. J.-Top what-INA not;related-ADV/not;know-ReI thing-Acc eat-Pres-DE ‘John eats anything.’ 3-QP nwukwuirtka in (41) may be followed by molu-nun salam ‘a person that I do not know’ but not by Q-Preds like sangkwaneps-i ‘irrespective of.’ On the contrary, V-QP mwuesina in (42) can be followed by sangkwaneps-i ‘irrespective o f but not by molu- nun mwulken ‘a thing that (he) does not know.’ Q-Preds for V-QPs are restricted to certain kinds of predicates such as sangkwaneps-i ‘irrespective of,’ kwankyeeps-i ‘regardless of,’ kalici ani-ha-ko ‘not discriminating,’ and so on. A common property of these predicates is that they are negative relational. These predicates behave as an identity function. Thus, no matter what is taken as the value of the WP in the complement clause of such a predicate, the value of the matrix clause is not affected. This ensures the universal-like, or more precisely, free choice ANY, reading for WPs in V-QPs.2 9 Such a treatment for reading variations between 3-QPs vs. V-QPs is able to eliminate the pure stipulation made in the literature, viz., that a WP plus -inka gives rise to an 3-QP, a WP plus -ina/itunci gives rise to a V-QP. 2 9 B. Schein (p.c.) points out to me that not all identity functions induce the free choice ANY or universal reading. For example, phrases like the thing that is itself contain an identity function (that is itself) but do not have the universal reading. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 5.6. On Indefinite Quantifier Analyses of Korean WPs 239 While I argued in the previous sections that WPs in Korean are interrogative elements across the board, the direct opposite view is logically possible. That is, it may be the case that WPs are underlyingly indefinite quantifiers across the board and the interrogative reading of WPs is derived by some mechanism. In fact, a few theories along these lines have been put forth in the literature. Chang (1976), for example, contends that WPs are underlyingly existential quantifiers and that they are changed to interrogative elements when they receive an information focus. C.- M. Suh (1987) argues along similar lines, although pitch (rather than information) focus plays a major role in his system. Let us call this approach a focus dependence analysis. Another theory that argues for the indefinite quantifier analysis of WPs is proposed by Kim (1989, 1991).3 0 According to Kim, WPs are indefinite expressions or quantifiers in the base across the board. A difference from the first approach is that interpretation of a WP depends, not on the existence of focus or lack of it, but on its relation to a question licensing ending (QE): when governed by or properly related to a QE, it has an interrogative reading but when not governed by or not properly related to a QE, it has an existential reading. Let us call this approach a unified quantifier analysis. In what follows, I will critically review the two theories that analyze Korean WPs as indefinite quantifiers and point out difficulties with these theories. 3 0 Choi (1935) first claims that Korean lacks interrogative pronouns as independent categories, although he does not provide any argument. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 5.6.1. Chang’s (1976) Focus Dependence Analysis 240 5.6.1.1. Focus as an Interrogative Converter In this subsection we briefly review the focus dependence approach to WPs. Chang (1976) claims that WPs like mmkwu ‘who’ and mwues ‘what’ are underlyingly existential quantifiers. They become interrogative elements when they are focused, as stated below: (43) SOME x -> WH x (where x is an index for person, object, time, place, etc.) [FOCUS] According to this mechanism, indefinite quantifier phrases turn into interrogative phrases when they receive a focus. This may explain the ambiguity that WPs display. As illustrated below, the meaning of a WP may be ambiguous depending on whether or not an information focus is put on the WP. (44) a. ecey NWU-ka o-ess-ni? yesterday who-Nom come-Pst-QE ‘Who came yesterday?’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 241 b. ecey nwu-ka o-ess-ni? yesterday who-Nom come-Pst-QE ‘Did someone come yesterday?’ WP nwu ‘who’ in (44a) marked with upper cases reads as an interrogative element since it is focused. On the other hand, nwu in (44b) is interpreted as an existential quantifier since it is not focused.3 1 One fact that agrees with this approach is that when a WP is located in a position in which a focus is forced, then the WP exclusively has an interrogative reading. For example, if a WP is followed by a focus affix like cocha, mace, kkaci, etc., or if a WP is a complement of a copula verb in the cleft construction, the WP has an interrogative reading but not an existential quantifier reading. Consider the following sentences: (45) John-i NWUKWU- (cocha/mace/kkaci 1 manna-ess-ni? J.-Nom WHO-EVEN meet-Pst-QE ‘For which x, x a person, did John see even x?’ # ‘For some x, x a person, did John see even x?’ (46) [John-i ej mek-un] kes-i mwues-i-ni? J.-Nom eat-Adn thing-Nom what-be-QE ‘What is it that John ate?’ # ‘Is it something that John ate?’ C.-M. Suh (1987) tries to push the same lines of arguments by observing pitch differences between interrogative elements and quantifier elements in a Kyengsang dialect Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 242 The WP nwukwu in (45) means interrogative WHO but not indefinite SOMEONE. Similarly, the WP mwues ‘what’ in (46) means WHAT but not SOMETHING. Such focused WPs are not compatible with declarative sentences, as exemplified below: (47) * John-i NWUKWU- (cocha/mace/kkaci I manna-ess-ta. J.-Nom WHO-EVEN meet-Pst-DE # ‘For some x, x a person, John saw even x.’ (48) * [John-i e * mek-un] kes-i mwues-i-ta. J.-Nom eat-Adn thing-Nom what-be-DE # ‘It was something that John ate.’ The absence of ambiguity of the sentences in (45) and (46) and the ungrammaticality of the sentences in (47) and (48) can be explained by the focus analysis. Being complements of a focus affix or being located in a focus position, the WPs in these sentences receive a focus and therefore must have an interrogative reading. Thus the WPs in (45) and (46) unambiguously have an interrogative reading and the sentences in (47) and (48) are ungrammatical because interrogative elements are incompatible with declarative sentences. One qualification is in order as to the interpretation of WPs in the focus domain. A WP in the focus domain does not necessarily have an interrogative reading. Look at the following sentences, in which a WP is in the focus domain but not exclusively: Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 243 (49) John-i enu haksavng-(cocha/mace/kkaci) manna-ess-ni? J.-Nom which student-EVEN meet-Pst-QE ‘For which x, x a student, did John see even x?’ 'For some x, x a student (not a teacher), did John see even x?’ (50) [John-i ei mek-un] kes-i enu kwail-i-ni? J.-Nom eat-Adn thing-Nom which fruit-be-QE ‘Is it some fruit (not rice) that John ate?’ ‘Which fruit is it that John ate?’ The WPs in these sentences are ambiguous. The WP enu ha/csayng in (49) could mean an interrogative ‘which student’ or existential ‘some student.’ The readings vary depending on the locus of the focus, as follows: (51) a. John-i ENU haksavng-1 cocha/mace/kkaci I manna-ess-ni? J.-Nom WHICH student-EVEN meet-Pst-QE ‘For which x, x a student, did John see even x?’ b. John-i enu HAKSAYNG- (cocha/mace/kkaci} manna-ess-ni? J.-Nom which STUDENT-EVEN meet-Pst-QE ‘For some x, x a student, did John see even x?’ When focus is placed on the WH-element, enu, an interrogative reading obtains, as in (51a). In contrast, when focus is placed on haksayng ‘student,’ then it conveys an Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 244 existential reading as in (51b). Similarly the WP enu kwail in (50) could mean 'which fruit’ or 'some fruit’ depending on the locus of the focus. Attractive as the theory may be in explaining the data mentioned above and several other phenomena, we will see in the next subsection that there are various non-trivial problems that the focus analysis faces in explaining other facts. 5.6.1.2. Difficulties with the Focus Dependence Analysis 5.6.1.2.1. Other Non-WH-Existentials According to (43), whenever an existential quantifier receives a focus, it is converted to an interrogative element. But this is not true. The so-called WH-elements may have interrogative readings, but obviously other existentials that do not contain a WH-element are never converted to interrogatives. Take the following examples: (52) a. i pang-ey koyangi han mali-ka iss-ni? this room-in cat oneCl-Nom exist-QE ‘Is there a cat in this room?’ b. i pang-ey koyangi HAN mali-ka iss-ni? this room-in cat ONE Cl-Nom exist-QE ‘Is there one cat (not many cats) in this room?’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 245 The nominal expression koyangi han mali ‘cat one-Cl’ is an existential element since it functions as the subject of an existential verb, iss ‘to exist.’ Nevertheless, it never reads as an interrogative element even when it is focused. As the English translation for (52b) indicates, it merely shows the contrast on the number of cats in the discourse. This indicates that existential-interrogative alternating expressions are restricted to a set of WH-elements. Existentials like numerals never change to interrogatives even if they are focused. Thus, even if WH-elements in Korean were indefinite existentials. they would have to be classified as existentials different from other indefinites. 5.6.1.2.2. Universals The focus dependence analysis deals with the ambiguity between an interrogative vs. existential reading of a WP but it overlooks the fact that WPs may have a universal reading as shown in the outset of this chapter. Note that the rule in (43), which deduces an interrogative reading from an existential quantifier with an appropriate information focus, mentions SOME with a focus feature but not EVERY at all. In other words, this approach is absolutely silent as to how a WP may have a universal reading. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 246 5.6.1.2 J . Lack of Focus in Declaratives A third problem with this analysis is that it does not give any principled account of the fact that no WP should be focused in declarative sentences. This is rather strange when we consider that there is no such restriction for other elements. In other words, a focus can be associated with almost any element in declarative sentences. But for WPs not to have an interrogative reading, they should somehow avoid receiving any information focus in declarative sentences. 5.6.I.2.4. Exclusive Interrogatives A fourth problem with this approach is that there exist WPs that do not have any indefinite counterpart. This is unexplained in the focus approach, where indefinite WPs and interrogative WPs are like two sides of a coin. Three types of such elements will be discussed in this section: WP way ‘why,’ WH-doublets, and definite-like WPs. All these three types of WPs have an interrogative reading but not an indefinite quantifier reading. The WH-word way ‘why’ only has an interrogative reading as illustrated below: (53) a. * John-i wav o-ess-ta. J.-Nom why come-Pst-DE ‘John came for some reason.’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 247 b. John-i wav o-ess-ni? J.-Nom why come-Pst-QE ‘Why did John come?’ # ‘Did John come for some reason?’ The WP way ‘why’ cannot be used in declarative sentences as in (53a). In interrogative sentences it only means ‘why’ but not ‘for some reason.’ If and only if interrogative WPs are derived from indefinites with a focus feature as stated in (43 ),3 2 the unique interrogative reading of way is unexplained unless there is an independent reason that way always receives a focus.3 3 3 2 Chang’ s mechanism in (43) has a unilateral condition ‘if,’ not ‘if and only if,’ though. 3 3 It is not true that way absolutely cannot have an existential reading. It could have that reading when it is ‘parasitic’ to some other WP with an existential reading: (0 John-i *(encev-inka) wav wuli cip-ev o-ese ilenmal-ul ha-ess-ta. J.-Nom when-LM why our house-at come-and this word-Acc do-Pst-DE ‘John came to my place some time for some reason and said such words.’ (0 is out without the existential quantifier encey-inka. Thus the existence of this additional WH- expression helps way to be interpreted as an existential quantifier. No adjacency between the two expressions is required: (ii) encev-inka John-i wav wuli cip-ev o-se ilenmal-ul ha-ess-ta. when-LM J.-Nom why our house-at come-and this word-Acc do-Pst-DE ‘John came to my place some time for some reason and said such words.’ However, hierarchical relation between the two is respected. Specifically way cannot be in a position higher than the additional existential quantifiers. Thus, if the order o f the two WPs is switched in (i) or in (ii), ungrammatical sentences result A clause mate condition also seems to be active. In other words a clause boundary between the two expressions makes the sentence ill-formed. Compare the following two sentences: Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 248 A second type of WPs that have exclusive interrogative readings are duplicated WPs, as discussed in Chung (1993c). Let us refer to them as WH-doublets. WH- doublets have an interrogative reading but not a quantifier reading as exemplified below:3 4 (54) a. nwukwu-nwukwu-ka o-ess-ni?3 5 who-who-Nom come-Pst-QE ‘Who(pl.) came?’ b. John-i eti-eti-ey ka-ess-ni? John-Nom where-where-at go-Pst-QE ‘Which places did John go?’ (iii) a. John-un [Mary-ka encev-inka wav caki cip-ey o-ess-ta]-ko malha-ess-ta. J.-Top M.-Nom when-LM why self house-to come-Pst-DE-C say-Pst-DE "John said that Mary visited his place once for some reason.' b. * encev-inka John-un [Mary-ka tray caki cip-ey o-ess-ta]-ko malha-ess-ta. when-LM J.-Top M.-Nom why self house-to come-Pst-DE-C say-Pst-DE ‘John once said that Mary visited his place for some reason.’ In (iiia) the WH-expressions are both in the embedded clause, and the sentence is perfect In (iiib). however, en cey-in ka is in the matrix and w ay in the embedded. The latter sentence is ungrammatical as if there is no additional WH-expression. 3 4 Not all WPs are repeatable: nw ukw u-nw ukw u ‘who-who,’ m w ues-m w ues ‘what-what’ en cey-en cey ‘when-when,’ e ti-e ti ‘where-where,’ enu-enu N ‘which-which N,’ m w usun-m w usun N ‘what-what N,’ but ?* etteh key-etteh key ‘how-how,’ *w ay-w ay ‘why-why.’ In a very restricted style of speech, for example, in stoiy telling situations, some WH-doublets may have an existential reading. But they never imply plurality, as exemplified below: (0 vevsnal-ev sokum cangswu-ka eti-eti-lul ka-nuntey, holangi-ka nathana-se, ... o!d;days-at salt seller-Nom where-where-Acc go-while tiger-Nom show;up-and,... ‘Once upon a time while a salt seller went to some place, a tiger showed up....’ 3 5 WH-doublet nw ukw u-nw ukw u is not shortened even before a nominative case marker, (cf. footnote Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 249 nwukwu-nwukwu in (54a) and eti-eti in (54b) mean ‘which men’ and ‘which places.' respectively, but they never mean ‘some men’ and ‘some places.’ There is no a priori reason that they must be focused.3 6 A third type of WPs that have an exclusive interrogative reading are WPs modified by ku ‘the/that,’ by enu ‘which,’ or by both. Consider the following examples: (55) nwu-ka ilen sayngkak-ul ha-l-kka? who-Nom such idea-Acc do-Fut-QE ‘Who would think of such a thing?’ ‘Would anyone think of such a thing?’ (56) caki-uy yenghon-ul mwues-kwa pakkwu-I swu-iss-ul-kka? self-Gen soul-Acc what-with exchange-Fut possibility-exist-Fut-QE ‘With what would a man exchange his soul?’ ‘Would a man exchange his soul with anything?’ (57) a. ku nwu-ka ilen sayngkak-ul ha-l-kka? that who-Nom such idea-Acc do-Fut-QE b. enu nwu-ka ilen sayngkak-ul ha-l-kka? which- who-Nom such idea-Acc do-Fut-QE 3 6 Under the unified inteiroptive analysis proposed in sections 2 and 3, this could be explained under the assumption that doublets are compounds. It was argued there that WPs with a quantifier reading involve a clausal structure. For example nwukwu-i-nka ‘who-be-NKA’ and nwukwu-i-na ‘who-be-INA.’ The lack of a quantifier reading of doublets follows from a plausible assumption that clauses cannot form a compound. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 250 c. ku enu nwu-ka ilen sayngkak-ul ha-l-kka? the which who-Nom such idea-Acc do-Fut-QE ‘Who would think of such a thing?’ # ‘Would anyone think of such a thing?’ (58) a. caki-uy yenghon-ul ku mwues-kwa pakkwu-1 swu-iss-ul-kka? self-Gen soul-Acc the what-with exchange-Fut possibility-exist-Fut-QE ‘With what would a man exchange his soul?’ # ‘Would a man exchange his soul with anything?’ b. caki-uy yenghon-ul enu mwues-kwa pakkwu-1 swu-iss-ul-kka? self-Gen soul-Acc which what-with exchange-Fut possibility-exist-Fut-QE ‘With what would a man exchange his soul?’ # ‘Would a man exchange his soul with anything?’ c. caki-uy yenghon-ul ku enu mwues-kwa pakkwu-1 swu self-Gen soul-Acc the which what-with exchange-Fut possibility iss-ul-kka? exist-Fut-QE ‘With what would a man exchange his soul?’ # ‘Would a man exchange his soul with anything?’ In normal questions, WPs are ambiguous between an interrogative reading and an indefinite quantifier reading as mentioned in the outset of this chapter. Thus nwu in (55), a short form of nwukwu ‘who,’ is ambiguous between WHO and SOMEONE or ANYONE. Similarly mwues in (56) is ambiguous between WHAT and SOMETHING Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 251 or ANYTHING. In contrast, when WPs are modified by ku as in (57a) and (58a). by enu as in (57b) and (58b), or by both as in (57c) and (58c),3 7 they are only interpreted as WHO/WHAT but not as SOMEONE/SOMETHING or ANYONE/ANYTHING.3 8 Such WPs usually but not necessarily appear in rhetorical questions. When they have a rhetorical reading, sentences in (57), for example, can be paraphrased as follows: (59) a. ku nwukwu-to ilen sayngkak-ul ha-ci ani-ha-n-ta. that who-TO such idea-Acc do-CI Neg-do-Pres-DE b. enu nwukwu-to ilen sayngkak-ul ha-ci ani-ha-n-ta. which who-TO such idea-Acc do-CI Neg-do-Pres-DE c. ku enu nwukwu-to ilen sayngkak-ul ha-ci ani-ha-n-ta. the which who-TO such idea-Acc do-CI Neg-do-Pres-DE ‘Nobody would think of such a thing.’ In the sentences in (59), predicates are changed to negative forms and to is added to the WPs. The particle to takes WPs and turns them into free choice ANY N. 3 7 When both appear, ku always precedes enu. 3 * Not every WP is able to bear ku or enu, as indicated in the following table: r ku enu ku enu nwukwu ‘who’ o.k. o.k. o.k. mwues ‘what’ o.k. 9* encey ‘when’ ? * * eti ‘where’ ? * * ettehkey ‘how’ ? * * way ‘why’ * ♦ * enu ‘which’ o.k. oJc. oJc. I Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 252 Definite WPs may have a genuine interrogative reading. This can be confirmed by the fact that the WPs in this context can be modified by the emphatic expression totaychey ‘on earth.’ Note that the emphatic expression generally modifies interrogative WPs. For example, sentences in (57) with the emphatic expression totaychey added are genuine interrogative sentences as shown below: (60) a. totavchev ku nwu-ka ilen sayngkak-ul ha-l-kka? on;earth the who-Nom such idea-Acc do-Fut-QE b. totavchev enu nwu-ka ilen sayngkak-ul ha-l-kka? on;earth which who-Nom such idea-Acc do-Fut-QE c. totavchev ku enu nwu-ka ilen sayngkak-ul ha-l-kka? on;earth the which who-Nom such idea-Acc do-Fut-QE ‘Who in the world would think of such a thing?’ In brief, we saw in this section that there are WPs which exclusively have an interrogative reading. Thus, the existence of such WPs will be a problem with the focus dependence analysis of WPs, which derives the interrogative reading of a WP from the corresponding indefinite quantifier, unless it is properly explained why such a WP always receives a focus. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 253 5.6.1.2.5. Exclusive Indefinites A fifth problem is the fact that WPs sometimes cannot have interrogative readings. This problem is exactly opposite in nature to the previous one. WPs read as an 3-QP when they are concatenated with so-called existential licensers -inka/-inci and as a V- QP when they are concatenated with universal licensers such as -ina/-itunci. as exemplified in (2) and (3) in the outset of the chapter.3 9 To explain the apparent lack of [+wh] reading in these sentences, there must be a theory that disallows a focus on the indefinites in these environments, a burden to the system.4 0 3 9 The focus dependence analysis and the unified quantifier analysis to be discussed in the next section merely try to account for the alternation of interrogative vs. existential quantifier reading o f bare WH- elements. But as was argued before, bare WPs are related to a (phonetically null) licensing morpheme, whereby we structurally equate bare WPs with WPs related to an overt licensing morpheme. 40 Empirically, WPs with a universal reading may be focused: (i) NWUKWU-na pep-aphey phyengtungha-ta. WHO-LM law-before equal-DE ‘Everyone is equal before law.’ Furthermore, a WP in WP-t'nto which has an existential reading may have focus, especially when followed by a contrastive focus marker -nun. (ii) NWUKWU-inka-nun sipcaka-lul ci-eya ha-n-ta. WHO-INKA-Con cross-Acc carry;on;back-must do-Pres-DE ‘SOMEONE(, whoever it is,) should be responsible.’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 254 5.6.1.3. Conclusion Undoubtedly WPs with interrogative readings have some information foci when compared to WPs with a non-interrogative reading. But the problems mentioned in the preceding subsections lead us to dismiss the focus dependence analysis. The focus, or the absence of it, on WPs is thus considered a byproduct, not a cause for the reading variation. 5.6.2. Kim’s (1991) Unified Quantifier Analysis A traditional grammarian, Choi (1935:228, 233), contends that Korean lacks interrogative pronouns as an independent category. According to Choi, the interrogative reading that WPs may have is derived from one of the two usages of indefinite pronouns; indefinites or interrogatives. A similar line of argument is maintained by Kim (1989a, 1989b, 1991: chapter 4). Kim (1991:207) contends:4 1 (61) WH-phrases are syntactically quantifier phrases in languages without WH- Movement. 4 1 Mahajan (1990; chapter 3) also contends that WPs in-situ (even in English) are QPs. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 255 Thus. WPs, being QPs. adjoin to IP by QR rather than move to SPEC of CP by WH- movement at LF. For example, the LF of the Chinese sentence in (62) is considered as (63b) rather than (63a): (62) Zhangsan kandao sheme? Zhangsan saw what ‘For which x, x a thing, Zhangsan saw x’ (63) a [c p sheme [ ip Zhangsan kandao t]] b. [ n > sheme [ n > Zhangsan kandao t]] (Kim 1991:205) In the following sections, I review Kim’s motivations for the unified quantifier analysis and discuss some problems with the analysis. 5.6.2.I. Motivations for the Unified Quantifier Analysis Kim (1991) claims that several significant consequences follow once WPs in situ are analyzed as quantifiers. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 256 5.62 .1.1. Morphological Paradigm First of all, Kim (1991) claims that the unified quantifier analysis could capture the morphological identity of WPs with interrogative uses and those with quantifier uses, as exemplified below:4 2 (64) nwu-ka o-ess-ni? who-Nom come-Pst-QE 'Who came?’ or ‘Did anybody come?’ He attributes ambiguous expressions such as nwukwu to the LF well-formedness condition stated below: (65) Well-Formedness Condition on IPI (Kim 1991:278) An interrogative polarity item (IPI) must be governed by the most local interrogative marker (Q-morpheme) at LF.4 3 By IPIs he means WPs with an interrogative reading. To distinguish the readings that a WP may have, he assumes that the question morpheme is base generated under INFL and may or may not be raised to C. When raised, the question morpheme is able to 4 2 If the WP is construed as the latter reading, -inka can be inserted after it, as mentioned earlier. Kim (1991) himself also notes this fact 4 3 The qualification ‘most local’ is needed because of WH-island effects. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 257 govern the subject, which is in the SPEC of CP.4 4 In this case an interrogative reading is produced. If the question marker remains in the base generated position, i.e., in INFL. then a quantifier reading is obtained, since the E P I in the SPEC of CP is not governed by the question morpheme in INFL. 5.6.2.1.2. Lack of Overt WH-Movement A second consequence that Kim (1991) claims can be deduced from the unified quantifier analysis is that Korean lacks overt raising of WPs. Just as quantifiers in general stay in situ at SS, WPs in Korean are not raised in overt syntax because they are quantifiers. Thus Huang’s (1982) SS/LF parameter on WH-movement is uncalled for under the quantifier analysis of WPs in-situ. They are adjoined to IP by a QR at LF like other quantifiers rather than move to the SPEC of CP. 4 4 Kim (1991) adopts a parameter as to Topic: In a language like Korean, which he assumes is a Topic prominent language. Topic should be filled with an overt element by S-structure. If we follow the well-formedness condition and the parameter on the topic prominence, WPs in the subject position of an indirect question is predicted to have an interrogative reading only. Note that embedded questions do not allow a Topic. Thus the subject must stay in the SPEC of IP. (He does not assume the VP internal subject hypothesis.) Therefore the subject, always being governed by a QE in INFL, must have an interrogative reading, which is not true: (i) na-nun [cp[ip nwu-ka o-ess-nunci] al-ko siph-ta. I-Top who-Nom come-Pst-QE know-KO want-DE ‘I want to know who came.’ “I want to know whether someone/anyone came.’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 258 5.6.2.13. Scope Interactions Thirdly, Kim (1991) takes as a piece of evidence for the unified quantifier analysis the fact about scope interaction in the following sentence: (66) nwu-ka mwues-ul kacyeo-ass-ni? EPI-Nom IPI-Acc bring-Pst-QE a. Who brought what? b. Did someone bring something? c. Who brought something? d. ?*What did someone bring? If WPs are ambiguous between interrogative use and quantifier use, the sentence (66), which contains two WPs, should logically be four way ambiguous, which is not true. As shown in the English translations, it is three-way ambiguous. Both WPs could be interpreted either as interrogative elements or as quantifiers. The third reading is the one in which the subject IPI is interpreted as an interrogative element while the object IPI as a quantifier. The fourth reading is the one where the subject IPI is interpreted as a quantifier while the object WP as an interrogative element, the opposite reading to the third one.4 5 4 5 According to Chang (1976:188), the multiple WH-question in (0 can be answered with the four sentences in (ii): Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 259 Kim tries to account for the state of facts that (66) displays with the well- formedness condition and the manipulation of the question morpheme movement According to Kim, the three readings of (66) follow from the following representations: (67) a. [cp IPIi [b> tj [v p EPIj [v p t; V]] tQE] QE] t _ l t _ J L_1 Topicalization(SS) QR(LF) QE-Raising(SS) b. [cp IPIi [ip tj [tp [vp IPIj [v p tj tvERB]] V-TNS] QE]] t i t i i r Topicalization(SS) QR(LF) V-Raising(SS) (0 nwu-ka Bill-eykey mwues-ul mek-ess-nunva-ko mwut-ess-supnikka? who-Nom B.-Dat what-Acc eat-Pst-QE-C ask-Pst-QE ‘Q who asked Bill [Q he ate what]?’ (ii) a. JOHN-i Bill-eykey mwues-ul mek-ess-nunya-ko mwut-ess-supnita. J.-Nom B.-Dat what-Acc eat-Pst-QE-C ask-Pst-DE ’John asked Bill whether he ate something.’ b. ney, nwu-ka Bill-eykey mwues-ul mek-ess-nunya-ko mwut-ess-supnita. yes who-Nom B.-Dat what-Acc eat-Pst-QE-C ask-Pst-DE ‘Yes, somebody asked Bill whether he ate something.’ c. JOHN-i Bill-eykey PPANG-ul mek-ess-nunya-ko mwut-ess-supnita. J.-Nom B.-Dat bread-Acc eat-Pst-QE-C ask-Pst-DE ‘John asked Bill whether he ate bread.’ d. nwu-ka Bill-eykey PPANG-ul mek-ess-nunya-ko mwut-ess-upnita. who-Nom B.-Dat bread-Acc eat-Pst-QE-C ask-Pst-DE ‘Somebody asked Bill whether he ate bread.’ If meaning of a question is determined by (possible) answers as in Karttunen (1976), then (i) has at least four readings. In particular, the existence of (iid), where the first WP in the matrix subject is interpreted as an existential quantifier and the second WP in the embedded object as an interrogative pronoun indicates that Kim’s generalization is either wrong or has to be qualified such that his generalization applies only to clause mate WPs. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 260 c. [cp EPIi [n > ti [ tp [v p IPIj [v p tj tvERfl]] V -T N S ] tQE] QE] Topicaliza5on(SS) •Raising(SS) QE-Ralsing(SS) Structures in (67a), (67b), and (67c) represent readings in (66a), (66b), and (66c), respectively. Due to the parameter on Topic (footnote 44), the subject IPI is in the topic position, i.e., the SPEC of CP, in all the three representations in (67). In (67a), the object IPI is adjoined to VP by QR at LF and the question morpheme is raised to C. The question morpheme in C governs both the subject IPI in the SPEC of CP and the object IPI adjoined to VP. Object E P I is governed because IP is inherently not a barrier and VP does not dominate the IPI. Being governed by the question morpheme, both of the IPIs read as interrogatives. In (67b), neither the IPI in the subject position nor the IPI in the object position is governed by the question morpheme since the subject IPI is in the SPEC of CP and the object IPI is adjoined to VP, while the question morpheme remains in situ, i.e., under INFL. IP blocks the government of subject IPI in the SPEC of CP, and TP blocks the government of the object IPI adjoined to VP.4 6 Thus the reading where both IPIs are interpreted as non-interrogative quantifiers falls out (67c) is identical to (67b) except for the raising of the question morpheme to C. Thus the subject IPI is governed by the question morpheme, while the object IPI is not 4 6 I have no idea why Kim does not postulate TP in (67a). If it were postulated, the object IPI which is adjoined to VP would not have an interrogative reading. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 261 due to the intervening TP. Thus the subject IPI reads as an interrogative pronoun, while the object IPI as a non-interrogative quantifier.4 7 Let us turn to the question of how the non-available reading in (66d) is excluded in his system. He considers the following two potential structures for the reading (Kim 1991:234 (56)): (68) a. *[Cp IPIi [ip IPIj [ip t, [v p tj V ] QE]]] b. [cp IPIj [ip tj [vp IPIj [vp tj V ] QE]]] In (68a), the subject IPI in the SPEC of CP is not governed by the question morpheme in INFL, while the object EPI which is adjoined to IP is governed by it. Thus (68a) is incorrectly predicted to be a right construction for the reading in (66d). He, however, contends that the structure is disallowed because it violates the scope rigidity condition which bans on scope alteration by an LF operation. Note that, being adjoined to IP at LF, IPIj c-commands tj, which changes the scope relation that is established at SS, where IPIj c-commands, and has scope over, IPIj. In (68b), the subject IPIj is located in the SPEC of CP due to the topic parameter and the object IPIj is adjoined to V P. D PIj is governed by Q E while IPIj is not Thus (68b) should be able to produce the reading in (66d). Kim does not provide any principled reason to exclude this structure but mentions the following: 4 7 One thing that does not agree well with this approach is the fact that V-Tns-QE always go together, which means that they are in the same position at least in the overt syntax. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 262 (69) To obtain a well-formed WH-question on the object IPI, the object IPI must be able to have a broader construal than the subject IPI that is interpreted as an existential quantifier. But in (56b)[=(68b):DC] the object IPI has only VP-scope. hence it has a narrower construal than the subject D P I. It is not difficult to see that a sentence in which a quantifier phrase has a broader scope construal than a WH-phrase is quite anomalous as a WH-question, hence it is not an acceptable WH-question universally.... Kim (1991:236) 5.6.2.2. Difficulties with the Unified Quantifier Analysis Despite the attractiveness of a few points, the unified quantifier hypothesis faces several serious difficulties. It will be pointed out that none of Kim’s motivations for the unified quantifier analysis seems to be based on solid ground. It will also be pointed out that his mechanism contains several technical and conceptual problems. 5.622.1. Implication of Morphological Paradigm The first argument that Kim (1991) provides for the unified quantifier analysis was that WPs may function as interrogative pronouns or as quantifiers. However, this argument is not logically sound. Note that this argument of his is based on the morphological identity of WPs of interrogative use and those of quantifier use. The morphological identity does not necessarily guarantee that WPs are quantifiers across the board, since the opposite possibility is left open; i.e., WPs might be interrogative Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 263 elements across the board. Thus to decide between the unified quantifier analysis and the unified interrogative analysis, independent evidence must be provided. 5.6.2.2.2. [+/-wh] Features and Checking Kim’s (1991) second argument for the quantifier hypothesis was that WPs. interrogative or non-interrogative, stay in situ at SS just as quantifiers do across languages. However, to distinguish the readings of WPs, Kim has to state that WPs with an interrogative reading are interrogative polarity items (IPIs) which must be governed by a question morpheme at LF, while WPs with an indefinite reading must not be governed by a question ending. Let us discuss what this implies and what consequences it gives rise to. Whether or not a [+/-wh] feature is lexically marked on WPs, the unified quantifier analysis runs into troubles.4 8 This is due to a general condition on feature checking as reasoned in what follows.4 9 Take a sentence where a WP resides in the embedded clause while the question morpheme is in the matrix clause, as exemplified below: (70) a. [c p , [+wh] f a ,... [C P 2 [m ... W P...]]]] 4 8 Kim (1991:279-280) explicitly claims that there is no feature like [+/-wh] on WPs in a language like Korean since they are quantifiers. But without marking the feature, the well-formedness condition would do nothing. Thus I use the feature [+/-wh] when it is relevant 4 9 Admittedly, the argument to be given is a hindsight, notions like checking being used. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 264 b. [John-un [Mary-ka nwukwu-lul salangha-n-ta-ko] mit-m]? J.-Top M-Nom who-Acc Iove-Pres-DE-C believe-QE ‘Who does John believe that Mary loves?’ ‘Does John believe that Mary loves someone/anyone?’ The question morpheme, which is in the matrix clause, cannot govern the WP at least in overt syntax since they are too far from each other. Nevertheless, (70b) is ambiguous with respect to the reading of WP nwukwu, either ‘who’ or ‘someone/anyone,’ as indicated in English translations. Under the minimalist theory, an element cannot move without participating in a certain kind of feature checking. (I put aside scrambling. See Sohn 1995.) Thus an element that moves has a morphological feature to be checked as the result of the movement. Suppose nwukwu was not marked with respect to the feature that distinguishes interrogatives from quantifiers. Then the WP would not be raised to the governing domain of the question morpheme, leaving unexplained the fact that the WP may have an interrogative reading. One step back, even if the WP were raised to check some other feature and happened to land in the government domain of a question morpheme, the landing site should not vary as far as it raises in order to check the relevant feature, again not distinguishing the two readings that WPs bear within the system Kim assumes.5 0 5 0 The argument given in this paragraph applies to any theory which assumes that movement of a WP to a certain domain produces a different object Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 265 Suppose, on the other hand, the [+/-wh] feature is marked on WPs. Then we might say that only [+wh] WPs move to the government domain of a question morpheme, capturing the ambiguity that WPs display. But these lines of explanation directly go against the original spirit of the unification of WPs of interrogative use and those of quantificational use. Note that once we begin to put such features on WPs, it amounts to admitting that there are two types ([+wh] vs. [-wh]) of lexical elements though their phonetic forms are identical by accident Thus no true unification would be reached by the unified quantifier analysis, contrary to the original intent5 1 5.62.23. WPs and QR Another difficulty with the unified quantifier analysis lies in the fact that WPs with an interrogative reading do not behave like ‘ordinary’ quantifiers. First of all, unlike other quantifiers, the so-called IPIs are not clause bounded to the scope position, as schematically illustrated below:5 2 5 1 No such problem arises in the unified interrogative analysis since WPs are uniformly related to a QE. 5 2 In fact it is controversial whether QR is clause bounded or not According to Williams (1988) and Lasnik and Saito (1989), sentences like (i) below are not ambiguous: (i) Some students think that every linguist goes to Cape Cod on Fridays. The QP in the embedded subject position cannot take scope over the QP in the matrix, which implies the clause bound property of quantifier scope. In contrast May (1985) assumes long distance QRs to capture the scope ambiguity of the WP and the QP in the following sentence: (ii) Where does Bobby think every detective will go for vacation? Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 266 (71) a. IP — [c p [ip — QP — ] t _ J Q R — b . [c p [ip - [c p [ip ••• IPI - ] -QR- EPIs may have a long distance relation with a question ending as long as no question morpheme intervenes. For example, they may appear in the complement of a bridge verb or even in a syntactic island, as exemplified below: (72) a. John-un [Mary-ka nwukwu-lul salangha-n-ta-ko] mit-ni? J.-Top M.-Nom who-Acc love-Pres-DE-C believe-QE ‘Who does John believe that Mary loves?’ b. John-un [[e; efi-ey sal-nun] salamj]-ul manna-ess-ni]? J.-Top where-at live-PNE man-Acc meet-Pst-QE ‘Q John met a man that lives where?’ The relevant schematic structures for the sentences in (72) are given in (73) below:5 3 (73) a [C P [+wh] [n>... [CP... WHO b . [c p [+wh] [n>... [np NP [c p ... WHERE However, various attempts have been made to capture the scope ambiguity in such a sentence without resorting to the long distance QR. See Aoun and Li (1989, 1993a), Sloan (1991), and Kim (1991). 5 3 For ease of exposition, the question morpheme is placed in the beginning. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 267 As for the structures like (73b), Kim (1991:280) adopts a pied-piping analysis. In other words, the whole NP including the IPI, instead of the IPI itself, raises to the government domain of the question morpheme. He did not explicitly mention the cases like the one in (73a). It may be possible to move the IPI successive cyclically: it first adjoins to the embedded E P , then to the matrix VP and finally to the matrix IP. This is illustrated below: (74) [ c p [ + w h ] [tpWHOi[ip...[vp tj [v p —[cp [n> tj [n»... tj ...] ] ] ] ] ] ] ] t It It 1 The IPI that is adjoined to the matrix IP (W HO* in (74)) is governed by the question morpheme in C given the usual distinction of the categories and segments. (See May 1985 and Chomsky 1986 among others.) Thus the interrogative reading is obtained. Another way is to adjoin the IPI to the embedded IP and then adjoin the whole embedded IP containing the adjoined IPI to the matrix IP, as illustrated below: Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 268 (75) SPEC C’ IPI C IP2 IPI WP, IP2 \NP VP SPEC CP SPEC This is, in fact, the approach that Mahajan (1991) adopts in explaining Hindi WH- movement. Mahajan also contends that WPs in situ in any language (even in English) are quantifiers and are subject to QR at LF. In order to make the QR of WPs satisfy the locality, he exploits the idea of Fiengo et al (1988). According to him, WPs are first adjoined to the embedded IP by QR. As a result of this movement, the embedded IP obtains a QP status. Now the whole embedded IP, being a QP, is adjoined to the matrix IP by a second application of QR. Mahajan’s approach also faces several non-trivial problems. First of all, the theory is still mute about QPs/WPs distinction. Note that WPs but not ordinary QPs induce Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 269 such a “pied-piped” QR of embedded IPIs. There is no a priori reason why normal quantifiers are prevented from undergoing the same fashion of pied-piped raising. In addition, it produces an undesirable consequence with respect to argument- adjunct asymmetry in island contexts. Consider the following structures: ( 7 6 ) a. [cp . [ip. ... [cp2 f o - W P...]] QE]] (SS) b. [cp i [ipi [if2 WPj [p2 ... t;... ]][n»i... [cp2 tff2 ]...]] QE] t I OR1 QR2 Suppose that CP2 is a barrier. The LF derivation in (76b) would produce an ECP violation. First, the WP, as a quantifier, is adjoined to the embedded IP (=IP2) by QR1. rendering the embedded IP a QP. The whole embedded IP is adjoined to the matrix IP (=IPI) by QR2. The trace tp2 violates the ECP since an IP is not an argument, irrespective of the categorial status of the WP adjoined to it This is because the pied- piped QPs in this context are invariably the embedded IPs, not WPs themselves. The fact is, however, that the grammaticality of sentences with the structure in (76a) depends on the categorial or thematic status of WPs themselves. In case the WP is of an argument type, then the sentence is grammatical, whereas if it is an adjunct like WHY, the sentence is ungrammatical or hardly acceptable: Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 270 (77) a. John-un [pro mwues-ul mek-umyen] paythalna-ni? J.-Top what-Acc eat-if get;stomachache-QE ‘Whatj does John get stomachache if he eats tj?’ b.* John-un [pro sakwa-lul wav mek-umyen] paythalna-ni? J.-Top apple-Acc why eat-if get;stomachache-QE ‘Why does John get stomachache if he eats apples tj?’ IPI raising in (74) or (75) is similar to QR in the sense that IPIs adjoin to IP, not substitute for SPEC of [+wh] Comp. On the other hand, it is different from QR in that such a movement is, in principle, unbounded unlike QRs in general. Thus the equation of IPI raising and QR is partially tenable at best I believe that Kim calls WH-phrases ‘interrogative polarity items,’ i.e., IPIs, due to the difference that normal quantifiers and WPs display. In other words, WPs show polarity properties in addition to quantificational properties. Given the polarity properties of IPIs, the adjunction approach Kim takes is conceptually implausible from the perspective of the current theory of grammar. In Kim’s system, question morphemes in Korean or Japanese reside under INFL in the base or raise to C if there is a WP in the topic position. Suppose a sentence contains two WPs, one in the sentence initial position, i.e., in the topic position due to his topic parameter. (See footnote 44.) The sentence structure will have the following LF structure in his system: Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Topicalizaqon(SS) QR(LF) WP2 is raised at LF to be governed by a question morpheme in Kim’s system. This could be rephrased as follows: WP2 is raised to have a local relation with a question morpheme. But there is no a priori reason for IPIs to be adjoined exclusively to an IP to have a local relation with the question morpheme but for the stipulation that they are QPs, which are supposedly adjoined to an E P by QR. In other words, the local domain of a head is not restricted to the adjoined position of its complement This kind of relation is rather peculiar from the perspective of the current theory. Note that the typical relations that a head may have is either a SPEC-head relation, a head-head relation, or a head-complement relation. In brief, Kim was right in saying that WPs are D P Is, that is, interrogative polarity items, thus they must have some structural relation with their licensing question morpheme. But his contention that WPs are QPs does not seem to be based on solid ground. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 272 5.6.2.2.4. WPs vs. Other QPs The equation of WPs with QPs yields another problem. While WPs are ambiguous between interrogative reading and non-interrogative quantifier reading, other quantifiers in general exclusively have a non-interrogative reading. Compare the following two sentences: (79) a. nwu-ka o-ess-ni? who-Nom come-Pst-QE ‘Who came?’ or ‘Did anyone come?’ b. motun salam-i o-ess-ni? all man-Nom come-Pst-QE ‘Did all men come?’ (79a) and (79b) are identical except for the fact that the subject is the WP nwu in (79a) and the universal quantifier motun salam in (79b). (79a) is ambiguous, while (79b) is not Kim (1991) tries to capture the ambiguity of (79a) by manipulating the position of QE as discussed in section 5.6.2.1.3. No other QPs read as [+wh] even in the circumstance where they could be governed by QE. Thus he has to restrict the well-formedness condition to WPs. The distinctive behavior of WPs and ordinary QPs weakens his claim that all WPs are quantifiers. Rather, WPs must be classified as a particular subset of QPs if he Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 273 adheres to the QP analysis of WPs. This amounts to saying that the so-called WPs are quantifiers but different in nature from other ‘ordinary’ QPs in the licensing process. If they were truly different from other QPs, there would be no reason to classify them as QPs. 5.6.2.2.5. Exclusive Interrogatives As discussed in section 5.6.1.2.3., WPs sometimes do not have an indefinite quantificational reading but instead have an interrogative reading. Among WPs that have an exclusive interrogative reading are the WP way ‘why,’ duplicated WPs, WPs in a focus position (i.e., WPs exclusively governed by focus elements like cocha, mace, kkaci. etc., or WPs in cleft constructions), and WPs modified by ku ‘the/that,’ or enu ‘which,’ or by both. The four sets of data are exemplified below: (80)(=(53)) a. * John-i wav o-ess-ta. J.-Nom why come-Pst-DE ‘John came for some reason.’ b. John-i wav o-ess-ni? J.-Nom why come-Pst-QE ‘Why did John come?’ # ‘Did John come for some reason?’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 274 (81) (=(54)) a. nwukwu-nwukvm-ka o-ess-ni? who-who-Nom come-Pst-QE ‘Who (pi.) came?’ # ‘Did some men come?’ b. John-i eti-eti-ev ka-ess-ni? J.-Nom where-where-at go-Pst-QE ‘Which places did John go?’ # ‘Did John go to some places?’ (82) a. (=(47)) # John-i NWUKWU- (cocha/mace/kkaci) manna-ess-ta. J.-Nom WHO-EVEN meet-Pst-DE # ‘For some x, x a person, John see even x.’ b- (=(48)) # [John-i e; mek-un] kes-i mwues-i-ta. J.-Nom eat-Adn thing-Nom what-be-DE # ‘It is something that John ate.’ (83) (=(57)) a. ku nwu-ka ilen sayngkak-ul ha-I-kka? that who-Nom such idea-Acc do-Fut-QE b. enu nwu-ka ilen sayngkak-ul ha-l-kka? which- who-Nom such idea-Acc do-Fut-QE Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 275 c. leu enu nwu-ka ilen sayngkak-ul ha-l-kka? the which who-Nom such idea-Acc do-Fut-QE ‘Who would think of such a thing?’ # ‘Would anyone think of such a thing?’ If the reading of a WP is decided exclusively by whether or not it is governed by a question morpheme at LF, as Kim argues, then it has to be shown why the WPs in (80)- (83) do not take the option of being ungovemed by the question morpheme. 5.6.2.2.6. Problems with the Well-Formedness Condition In this subsection, I point out several problems with Kim’s Well-Formedness Condition (55), which is repeated as (84) below: (84) Well-Formedness Condition on IPI (Kim 1991:278) An interrogative polarity item(IPI) must be governed by a most local interrogative marker (Q-morpheme) at LF. With (84) he tries to account for the ambiguity that sentence (56), repeated below as (85), has: Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 276 (85) nwu-ka mwues-ul kacyeo-ess-ni? IPI-Nom IPI-Acc bring-Pst-QE a. Who brought what? b. Did someone bring something? c. Who brought something? d. ?*What did someone bring? His explanation for the lack of the reading (85d) was not clear enough, as we saw at the end of section 5.6.2.1.3. What he held there was that interrogative IPIs must have a broader scope than other quantifiers. (See (69).) This is at best a generalization, not a principled explanation. Furthermore, in a language such as English, a universal quantifier may have a wide scope over WH-phrases as discussed in May (1985). For example, in the sentence Who did everyone see? the subject quantifier may have a wider scope over the WP. Another technical problem that Kim’s analysis may have lies in explaining the following sentence with both EPIs being read as non-interrogatives: (86) nwukwu-eykeyi John-i tj mwues-ul cwu-ess-ni? who-Dat J.-Nom what-Acc give-Pst-QE ‘Did John give someone something?’ The schematic LF structure for (86) will be (87) below: Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 277 (87) [ c p IPIj [n» John [t p [vpIPIj[vp tj tj tvgRB]] V-TNS] QE]] t . . 1 1 Topical ization V-raising QR The structure in (87) runs into a paradoxical situation. In order not to govern the IPI that is adjoined to VP by QR, there must be a barrier between the question morpheme and the object IPI or its possible landing site at LF. The topicalized IPI, then, must have moved across the barrier. Thus the sentence should be at best marginal for the subjacency reason, which is not true. Another potential problem arises with respect to the scope rigidity condition, which bans alteration of a scope relation established at SS by an LF operation. Consider (88a) below, which is identical to (86) except that both WPs are interrogatives this time, and its LF structure (88b): (88) a. nwukwu-eykeyj John-i tj mwues-lul cwu-ess-ni? who-Dat J.-Nom what-Acc give-Pst-QE ‘To whom did you give what?’ b. [c p IPIi [ ip John [t p [vpIPIj[vp tj tj Wera]] V-TNS] tQ g ] QE] I 1 L J V-raising QE-Raising QR t _ . | 1 Topicalization Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 278 The dative IPI resides in the SPEC of CP as a topic at SS. Thus a scope relation established in SS is that the dative IPI has wide scope over the accusative IPI. LF movement of the direct object IPI violates the rigidity condition since it forms a chain which could reverse the scope relation established at SS. Note that the LF chain formed by moving object IPI overlaps with the one formed by the movement of topic in SS. A clearer case of the rigidity condition violation will be like the sentence in (89) below, where a non-subject matrix IPI is in the topic position and another IPI is in the embedded clause: (89) nwukwu-evkevi John-i tj [Mary-ka nwukwu-lul salangha-n-ta-ko] malha-ess-ni? who-Dat J.-Nom M.-Nom who-Acc Iove-Pres-DE-C say-Pst-QE ‘To whom did John say that Mary loves who?’ When the IPI in the embedded clause has an interrogative reading, it should cut into the chain of the IPI in the matrix clause, changing the scope relation established at SS. 5.6.2.2.7. Problems with Question Morpheme Raising Kim (I991:227ff) argues that question morphemes are base generated in INFL and that they may or may not be raised to C. He crucially depends on the manipulation of the question morpheme movement to obtain the correct readings of IPIs as we saw in Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 279 section 5.6.2.1.3. But it will be shown in this section that the evidence he provides for this analysis is not only based on shaky ground but it also bears a conceptual defect. Let us discuss Kim’s (1991) argument that QE is generated under INFL. His first argument is based on the fact that QE co-occurs with -to in Japanese and -bo in Korean, which are considered complementizers: (90) (=Kim 1991:227, his (45)) a. John-wa [Maiy-ga nani-o katta ka to] tazuneta < f > . (J) J.-Top M.-Nom what-Acc bought QE C asked ‘John asked what Mary bought’ b. John-un [Mary-ka mwues-ul sa-ess-nva-ko] mwut-ess-ta. (K) J.-Top M.-Nom what-Acc buy-Pst-QE-C ask-Pst-DE ‘John asked what Mary bought’ ‘John asked if Mary bought something.’ According to Kim (1991), the question morphemes -ka in Japanese and -nya in Korean cannot be complementizers because the complementizer position is already filled with - to in Japanese and with -ko in Korean.5 4 5 4 The question ending in (90b) -nya is a matrix question ending in other contexts: (i) John-i mwues-ul mek-css-nva? J.-Nom what-Acc eat-Pst-QE 4 What did John eat?’ This seems to indicate that the embedded clause in (90b) may be an instance of matrix clauses. Note incidentally that Emonds (1976) includes as the matrix clauses not only root clauses but also coordinate Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 280 One immediate question that arises with this approach is the non-local property of the selectional restriction. Note that the question morpheme in the embedded clause is not permitted without a predicate in the matrix that selects a question clause. Thus there is a selectional relation between them. But the relation is non-local because a complementizer -ko in Korean and -to in Japanese intervenes inbetween. He tries to resolve the problem by maintaining that Comp in Korean and Japanese is transparent conjuncts and citation. This in turn indicates that -ko might just be a citation marker, direct or indirect. H. Hoji (p.c.) notes that -ha-ko does not have to select a clause. For example, it could select an inteijection: (ii) a. John-un "ak" ha-ko malha-ess-ta. J.-Top "ak" do-C say-Pst-DE, ‘John said, "ak".’ Another point worth mentioning is that with the various types of the embedded clauses the matrix verb can be neutralized to malha or just ha. (iii) a. John-un [Tom-i ttena-ess-ta-ko] malha-ess-ta. J.-Top T.-Nom leave-Pst-DE-C say-Pst-DE ‘John said that Tom left’ b. John-un [Tom-i ttena-ess-nva-kol {malha/mwut}-ess-ta. J.-Top T.-Nom leave-Pst-QE-C say/ask-Pst-DE ‘John asked whether Tom left.’ c. John-un Tom-eykey [pro ttena-la-ko] {malha/myenglyengha}-ess-ta. J.-Top T.-Nom leave-Imp-C say/command-Pst-DE ‘John commanded Tom to leave.’ d. John-un Tom-eykey [pro ttena-ca-ko] {malha/ceyanhaj-ess-ta. J.-Top T.-Nom leave-prop-C say/propose- Pst-DE ‘John proposed Tom to leave.’ This indicates that malha is equivalent to the English verb to say, which may take various sentence types in direct quotation. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 281 and inherits the property of INFL on a par with the English fact that Comp and INFL should be of the same value on the agreement feature. However, there is an indication that, at least in Korean, both -ko and the question morpheme are analyzable as complementizers. Note that a dummy verb ha can be inserted between -nya and -ko in (90b), as shown below: (91) John-un [[Mary-ka mwues-ul sa-ess-nva] ha-ko] mwut-ess-ta. J.-Top M.-Nom what-Acc buy-Pst-QE do-C ask-Pst-DE ‘John asked what Mary bought’ Thus sentences like (90b) could be analyzed as tri-clausal rather than bi-clausal. as illustrated below: (92) John-un, [proj [Mary-ka mwues-ul sa-ess-nval 6-kol mwut-ess-ta. J.-Top M.-Nom what-Acc buy-Pst-QE do-C ask-Pst-DE The verb ha can be analyzed as a proverb anaphorically related to the matrix predicate. In addition, there are cases which indicate that question morphemes are directly selected by a matrix verb with no complementizer -ko. Consider the following sentence: Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 282 (93) John-un [nwu-ka oess-na](*-ko) kwungkumhayha-n-ta. J.-Top who-Nom come-Pst-QE-C wonder-Pres-DE ‘John wonders who came.’ If question morphemes were under INFL, and the complementizer -ko were really transparent with respect to selectional restriction, it would be hard to explain sentences like (93). If only -ko, but not -na, in (93) were a complementizer, then the embedded clause would not be a CP, but an IP, which is unlikely to be the case. Thus, there must be a direct relation between the -ko complementizer and its higher verb. Verbs like mwut ‘to ask’ in (91) select the -ko complementizer, while verbs like kwungkumhayha ‘to wonder’ in (93) do not, although both predicates select an embedded question.5 5 I claim that the selectional restriction between the matrix verb and the question morpheme in sentences such as (91) is accomplished by the mediation of the proverb ha. This analysis seems more plausible than Kim’s INFL analysis because the local relation of selectional restriction can be preserved in the proposed analysis. Let us now turn to his second argument for the INFL analysis of QEs. It is based on the assumption that sentence types are heads of clauses. As it is well known, sentence types are determined by sentence particles in Korean as exemplified below: 5 5 Verbs that take -to as a complementizer are indirect speech verbs like malha ‘to tell,’ pokoha ‘to report,’ etc. and thinking verbs like mil ‘to believe,’ sayngkakha ‘to think,’ etc. See Yoon (1990) and Chung (1993b) for the classification of Korean complementizers. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 283 (94) (=Kim 1991:231 -232, his (51)) a. Chelswu-ka chayk-ul sa-ess-NI? C.-Nom book-Acc buy-Pst-QE ‘Did Chelswu buy a book?’ b. Chelswu-ka chayk-ul sa-ess-TA. C.-Nom book-Acc buy-Pst-DE ‘Chelswu bought a book.’ c. Chelswu-ya, chayk-ul sa-LA. C.-Voc book-Acc buy-Imp ‘Chelswu, buy a book.’ d. Chelswu-ya, chayk-ul sa-CA. C.-Voc book-Acc buy-Prop ‘Chelswu, let’s buy a book.’ (question) (proposition) (imperative) (1st person imperative) Questions end with ni as in (94a), propositions with to as in (94b), imperatives with la as in (94c), propositives with ca as in (94d), etc.5 6 This, however, shows nothing about the categorial status of the sentence particles. In particular, this does not show that sentence particles are INFL unless it is assumed that INFL decides the sentence types. What if the sentence particles are complementizers and complementizers determine the sentence types? Thus his claim that question morphemes reside in INFL is not established unless other independent motivations are provided. 5 6 There are various other endings that vary depending on the styles and registers Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 284 The third argument he provides is the concurrence restriction of two or more sentence particles. No two sentence particles can appear side by side: (95) (=Kim 1991:232-233, his (52a)) *Chelswu-ka chayk-ul sa-ess-NI-TA? C.-Nom book-Acc buy-Pst-QE-DE This observation is correct, but it does not show that sentence particles belong to the category INFL. It merely shows that sentence particles belong to the same category, under the assumption that two elements of the same category show a complementary distribution. The fourth argument is based on the description that some complementizers never occur in the main clause in Korean and Japanese, as in English. (96) (= Kim 1991:233, his (53)) a. *John-i chayk-ul sa-ess-upnita-ko J.-Nom book-Acc buy-Pst-DE-C b. *John-ga hon-o kat-ta-< J> -to J.-Nom book-Acc buy-Pst-DE-C c. *That John bought a book. The Korean complementizer -ko, and Japanese complementizer to cannot appear in the matrix sentence just as the English complementizer that cannot in the same situation. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (K) (J) 285 But it is not necessarily true that complementizers never appear in the matrix clauses. For example, Arabic allows complementizers in the matrix. Even in English complementizers in the matrix have to be assumed in questions. Consider the following sentence: (97) What do you do? In this sentence CP projection is required to host the WH-word in its SPEC. Once CP is projected, C (=complementizer) must also be assumed to exist since a projection without its head is unthinkable. To rule out the sentences in (96), we may have to stipulate that some complementizers appear in the matrix, but others do not. Thus none of the four arguments he provides is solid enough to show that question morphemes like -nya or -ni in Korean are generated under INFL. One last point that we can make is that the question morpheme raising story is theoretically doubtful. This is because the raising or staying in-situ of question morphemes is entirely arbitrary. Such a kind of movement is questionable from the perspective of the current theory of movement, where each instance of movement is driven by some morphological force (probably with the sole exception of scrambling). Note that the movement should not be of the kind that just facilitates the interpretation Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 286 of the so-called interrogative items, the reason being that this would have nothing to do with morphological checking of the feature that the moved element bears. 5.6.23. Conclusion The uniform quantifier hypothesis cannot stand on firm ground with the conceptual and technical problems discussed in the previous subsections, although it is attractive in a few respects. We have seen that there are several fundamental differences between QPs and WPs in their syntactic behaviors, which makes it hard to equate WPs with QPs. 5.7. Concluding Remarks Now that both 3-QPs and V-QPs are analyzed as indirect questions, WPs in such quantifiers are licensed by having some relation with a question sentence ending just like those in obvious WH-questions. This seems tenable in the sense that a set of lexical items with a particular feature, that is, WH-elements with a [+wh] feature, are identified by another set of items with a matching feature, that is, question sentence endings. Since Korean WPs, interrogative or quantificational, are uniformly analyzed as elements with a [+wh] feature, the lack of overt WH-movement in Korean cannot be ascribed to the hypothetical assumption that Korean WPs are quantifiers devoid of an Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 287 interrogative feature (Kim 1989a, 1989b, 1991). It was also shown in section 4.2.1. of chapter 4 that WH-elements in Korean are not essentially different from those in English in their morphological structure. Thus, the WH-raising vs. WH-in-situ typology cannot be attributed to any difference in their morphological composition of WH- elements, either (A. Watanabe 1991 and Tsai 1994). Rather, we claimed in chapter 2 that the raising vs. in-situ typology follows from the difference in the licensing mode of the question force, i.e., a SPEC-head relation vs. a head-head relation, not from the internal difference of WH-elements of the two types of languages. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Bibliography 288 Abney, Steven P. (1987) The English Noun Phrase in its Sentential Aspect, Ph.D. dissertation, MIT. Ahn, Byung-Hee (1968) “Cwungseykwuke-uy Sokkyekemi ‘s’-ey Tayhaye (On the Genitive Ending ‘s’ in Middle Korean),” Papers in Honor o f Dr. Swungnyeng Lee, Ulyumwunhwasa, Seoul, Korea. Ahn, Heedon (1990) Light Verbs, VP-Movement, Negation, and Clausal Architecture in Korean and English, Ph.D. dissertation, University of Wisconsin-Madison. Ahn, Sung-Ho (1990) Korean Quantification and Universal Grammar, Ph.D. dissertation, University of Connecticut. Aoun, Joseph (1986) Generalized Binding: The Syntax and Logical Form ofWH- Interrogatives, Foris Publications, Dordrecht-Holland/Riverton-U.S.A. Aoun, Joseph & Dominique Sportiche (1983) “On the Formal Theory of Government,” The Linguistic Review. Aoun, Joseph and Y.-H. Audrey Li (1989) “Scope and Constituency,” Linguistic Inquiry 20.2. Aoun, Joseph and Y.-H. Audrey Li (1993a) “WH-Elements in-situ: SS or LF?” Linguistic Inquiry 24,199-238. Aoun, Joseph and Y.-H. Audrey Li (1993b) “On the Some Differences between Chinese and Japanese WH-Elements,” Linguistic Inquiry 20,265-272. Aoun, Joseph and Y.-H. Audrey Li (1993c) Syntax o f Scope, MIT Press. Aoun, Joseph, Nobert Homstein, and Dominique Sportiche (1981) “Aspects of Wide Scope Quantification,” Journal o f Linguistic Research 1,67-95. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 289 Aoyagi, Hiroshi (1994) “On Association with Focus and Scope of Focus Particles in Japanese,” ms., Department of Linguistics, University of Southern California. Baker, C.L. (1970) “Notes on the Description of English Questions: the Role of an Abstract Question Morpheme,” Foundations o f Language 6,197-219. Baker, C.L. (1991) “The Syntax of English Not’ . The Limits of Core Grammar.” Linguistic Inquiry 22, 387-419. Baker, Mark (1988) Incorporation: A Theory o f Grammatical Function Changing, University of Chicago Press, Chicago, Illinois. Barwise, Jon and Robin Cooper (1981) “Generalized Quantifiers and Natural Language,” Linguistics and Philosophy 4,159-219. Beghelli, Fillipo (1994) “Asymmetries of Decreasing QPs,” WCCFL 13. Benmamoun, Elabbas (1992) Functional and Inflectional Morphology: Problems o f Projection, Representation and Derivation, Ph.D. dissertation, University of Southern California. Berman, S. (1989) “On Certain Differences between WH-Phrases and Indefinites,” in J. Cater et al. (eds.), NELS 20,31-74, GSLA, University of Massachusetts, Amherst. MA. Carlson, Greg N. (1980) “Polarity Any is Existential,” Linguistic Inquiry 11, 799-804. Chang, Kyungsun (1992) “Negation in Korean,” ms., Department of Linguistics, University of Southern California. Chang, Suk-Jin (1975) “Mwuntap-uy Hwayongsang (Pragmatic Aspects of Questions and Answers),” Ehakyenkwu (Language Research) 11.2., Seoul National University Press, Seoul, Korea. Cheng, Lisa L.-S. (1991) On the Typology ofWH-Questions, Ph.D. dissertation, MIT. Cheng, Lisa L.-S. and C.T. James Huang (1993) “Two Types of Donkey Sentences.” ms., Department of Linguistics, UC Irvine. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 290 Chierchia. Gennaro (1992-1993) “Questions with Quantifiers,” Natural Language Semantics 1, 181-234. Choe. Hyun-Sook (1987) “Successive-Cyclic Rightward Movement in Korean.” Kuno et al. (eds.) Harvard Studies in Korean Linguistics Vol. 2.40-56. Choe. Hyun-Sook (1987) “Syntactic Adjunction, A-Chain, and the ECP - Multiple Identical Case Construction in Korean,” in Proceedings o f the Seventeenth Annual Meeting, NELS, GLSA, University of Massachusetts, Amherst. Choe, Hyun-Sook (1988) Restructuring Parameters and Complex Predicates; A Transformational Approach, Ph.D. dissertation, MIT. Choe, Jae-Woong (1987) “LF Movement, and Pied-Piping,” Linguistic Inquiry 18.348- 353. Choi, Hyun-Pae (1935) Wuli Malpon (Our Grammar), Cengummwunhwasa, Seoul, Korea. Chomsky, Noam (1957) Syntactic Structures, Mouton, the Hague. Chomsky, Noam (1965) Aspects o f the Theory o f Syntax, MIT Press, Cambridge, Massachusetts. Chomsky, Noam (1981) Lectures on Government and Binding, Dordrecht, Foris. Chomsky, Noam (1985) Knowledge o f Language: Its Nature, Origins and Use, Praeger. New York. Chomsky, Noam (1986) Barriers, MIT Press. Chomsky, Noam (1991) “Some Notes on Economy of Derivation and Representation,” in R. Freidin (ed.) Principles and Parameters in Comparative Grammar, MIT Press, Cambridge, Massachusetts. Chomsky, Noam (1992) “A Minimalist Program for Linguistic Theory,” MIT Occasional Papers in Linguistics, No. 1, Cambridge, Massachusetts. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 291 Chomsky, Noam (1994) “Bare Phrase Structure,” MIT Occasional Papers in Linguistics No. 5, Cambridge, Massachusetts. Chomsky, Noam (1995) Minimal Syntax, MIT Press, Cambridge, Massachusetts. Chomsky, Noam and Howard Lasnik (1991) “Principles and Parameters Theory,” J. Jacobs, A, van Stechow, W. Stemefeld, and T. Vennemann (eds.) Syntax: An International Handbook o f Contemporary Research, Walter de Gruyter. Berlin. Chung, Daeho (1993a) “Negative Polarity Items in Korean,” ms., Department of Linguistics, University of Southern California. Chung, Daeho (1993b) “Case Licensing in the Korean ECM Construction,” Susumu Kuno et al. (eds.) Harvard Studies in Korean Linguistics V, Proceedings o f the 1993 Harvard International Symposium on Korean Linguistics (HAR VARD ISOKL-1993), 287-298, Department of Linguistics Harvard University, Cambridge, Massachusetts. Chung, Daeho (1993c) “WH-Phrase Licensing and Interpretation,” ms., Department of Linguistics, University of Southern California. Chung, Daeho (1994) “Question Operations in Korean,” Young-Sun Kim et al. (eds.) Studies in Generative Grammar: Festschrift to Dr. Dong-Whee Yang's 60th Birthday, 47-68. Chung, Daeho (1995a) “A Negation Typology and Distribution of NQs and NPIs,” Paper Presented at LSA Annual Meeting Jan. 5-8, 1995, New Orleans, LA. Chung, Daeho (1995b) “A Question Analysis of Korean Universal and Existential Quantifiers,” to appear in Susumu Kuno et al. (eds.) Harvard Studies in Korean Linguistics Vol. 6. Chung, Daeho and Hong-Keun Park (1995a) “An Indirect Question Analysis of Korean Quantifiers,” to appear in MTT Working Papers in Linguistics. Chung, Daeho and Hong-Keun Park (1995b) “NPIs outside Negation Scope,” to appear in Proceedings o f Japanese/Korean Linguistics Vol. 6. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 292 Chung, Taegoo (1991) “On WH-Adjunct Extraction in Korean,” Susumu Kuno et al. (eds.) Harvard Studies in Korean Linguistics IV, 241-248. Cinque, Guilemo (1990) Types o f A -Dependencies. MIT Press, Cambridge, MA. Collins, Chris (1994) “Economy of Derivation and the Generalized Proper Binding Condition,” Linguistic Inquiry 25,45-61. Diesing, Molly (1992) Indefinite, MIT Press, Cambridge, MA. Emonds, Joseph E. (1976) A Transformational Approach to English Syntax, Academic Press. Evans, Gareth (1980) “Pronouns,” Linguistic Inquiry 11, 337-362. Fiengo, Robert (1977) “On Trace Theory,” Linguistic Inquiry 77,35-61. Fiengo, Robert, C.-T. James Huang, Howard Lasnik, and Tanya Reinhart (1988) “The Syntax of WH-in-situ,” Wccfl 7. Fukui, Naoki (1986) A Theory o f Category Projection and its Application, Ph. D. dissertation, MIT. Fukui, Naoki (1988) “LF Extraction o f Naze: Some Theoretical Implications,” Natural Languages and Linguistic Theory 6 ,503-526. Fukui, Naoki (1993) “Parameters and Optionality,” Linguistic Inquiry 24,399-320. Gair, J. (1983) “Non-Configurationality, Movement, and Sinhala Focus,” Paper Presented to the Linguistics Association of Great Britain, Newcastle. Gair, J. and L. Sumangala (1991) "What to Focus in Sinhala,” Proceedings o f ESCOL 1991. Grimshaw, Jane (1977) English WH-Constructions and the Theory o f Grammar, Ph.D. dissertation, University of Massachusetts, Amherst Haegeman, L. (1992) “Negation in West Flemish and the Neg Criterion,” NELS 22. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 293 Heim, Irene (1982) The Semantics o f Definite and Indefinite Noun Phrases, Ph.D. dissertation, University of Massachusetts, Amherst Higginbotham, James (1985) “Interrogatives I,” MIT Working Papers in Linguistics. Vol. 15, MIT, Cambridge, Massachusetts. Hoji, Hajime (1985) Logical Form Constraints and Configurational Structures in Japanese, Ph.D. dissertation, University of Washington. Hoji, Hajime (1990) “Theories of Anaphora and Aspects of Japanese Syntax,” ms.. Department of Linguistics, University of Southern California. Hoji, Hajime (1995) “Demonstrative Binding and Principle B,” to appear in NELS 25. Homstein, Nobert (1994) “The Grammar of LF: From GB to Minimalism,” ms., Department of Linguistics, University of Maryland. Huang, C.-T. James (1982) Logical Relations in Chinese and the Theory o f Grammar. Ph.D. dissertation, MIT. Jackendoff, Ray (1972) Semantic Interpretation In Generative Grammar, The MIT Press, Cambridge, Massachusetts. Kang, Jung-Goo and Gereon Muller (1994) “Reconstruction vs. Copying: The Case of WH-Scope,” Paper Presented at the 5th Japanese/Korean Linguistics Conference. UCLA. Kang, Myung-Yoon (1988) Topics in Korean Syntax: Phrase Structure, Variable Binding and Movement, Ph.D. dissertation MIT. Kang, Youngsey (1986) Korean Syntax and Universal Grammar, Ph.D. dissertation, Harvard University. Karttunen, Lauri (1977) “Syntax and Semantics of Questions,” Linguistics and Philosophy 1,3-44. Kato, Yasuhiko (1985) Negative Sentences in Japanese, Sophia Linguistica, Working Papers in Linguistics 19, Sophia University, Tokyo, Japan. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 294 Katz, Jerrold and Paul M. Postal (1964) An Integrated Theory o f Linguistic Descriptions, The MIT Press, Cambridge, Massachusetts. Kayne, Richard S. (1984) Connectedness and Binary Branching, Foris, Dordrecht. Kayne, Richard S. (1991) “Romance Clitics, Verb Movement, and PRO,” Linguistic Inquiry 22,647-686. Kayne, Richard S. (1993) The Antisymmetry o f Syntax, The MIT Press, Cambridge, Massachusetts. Kim, Alan Hyun-Oak (1982) “A Universal of WH-Questions and its Parametric Variants,” Papers from Parasession on Nondeclaratives, C LS12. Kim, Nam-Kil (1974) The Grammar o f Korean Complementation, Ph.D. dissertation. University of Washington. Kim. Soowon (1989a) “WH-Phrases in Korean and Japanese are QPs,” MIT Working Papers in Linguistics 11,119-138. Kim, Soowon (1989b) “The QP Status of WH-Phrases in Korean and Japanese,” in Proceedings o f the Eighth Annual Meeting, WCCFL, CSLI, Stanford University, Stanford, California. Kim, Soowon (1991) Chain Scope and Quantification Structure, Ph. D. dissertation, Brandeis University. Kitagawa, Yoshisa (1986) Subjects in Japanese and English, Ph.D. dissertation, University of Massachusetts, Amherst. Klima, E.S. (1964) “Negation in English,” in J. A. Fodor and J. J. Katz, eds., The Structure o f Language, Prentice-Hall, Englewood Cliffs, NJ. Koizumi, Masatoshi (1991) Syntax o f Adjunct s and the Phrase Structure o f Japanese, MA thesis, the Ohio State University. Ko, Young-Keun (1987) Phyocwun Cwungsey Kwuke Mwunpeplon (Standard Middle Korean Grammar), Tap Publishing Co., Seoul, Korea. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 295 Ko, Young-Keun and Ki-Shim Nahm (eds.) (1982) Kwukeuy Thongsa. Uymilon (Syntax and Semantics of Korean), Tower Press, Seoul, Korea. Koster, J. (1978) “Why Subject Sentences Don’t Exist?” in S J. Keyser (ed.) Recent Transformational Studies in European Languages, Linguistic Inquiry Monograph No.3. Kuno, Susumu (1973) The Structure o f the Japanese Language, The MIT Press, Cambridge, Massachusetts. Kuno, Susumu and K. Masunaga (1986) “Questions with WH-Phrases in Islands,” UMOP 11,139-166. Kuno, Susumu, and J. Robinson (1972) “Multiple WH-Questions,” Linguistic Inquiry 3,463-487. Kuno, Susumu et al.(eds.) (1991) Harvard Studies in Korean Linguistics IV. Hanshin publishing Co., Seoul. Kuroda, Shige-Yuki (1965) Generative Grammatical Studies in Japanese Language, Ph.D. dissertation, MIT. Kuroda, Shige-Yuki (1988) “Whether We Agree or not,” Linguisticae Investigationes 12, 1-14. Kuroda, Shige-Yuki (1992) Japanese Syntax and Semantics, Dordrecht, Reidel. Ladusaw, William (1980) Polarity Sensitivity as Inherent Scope Relation, New York: Garland Press. Ladusaw, William (1992) “Expressing Negation,” Paper Presented at SALT 2. Laka, Miren Itziar Murgarza.(1990) Negation in Syntax: On the Nature o f Functional Categories and Projections, Ph.D. dissertation, MIT. Larson, Richard K. (1985) “On the Syntax of Disjunction Scope,” Natural Language and Linguistic Theory 3,217-264. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 296 Lasnik, Howard and Mamoru Saito (1984) “On the Nature of Proper Government," Linguistic Inquiry 14,235-289. Lasnik, Howard and Mamoru Saito (1992) Move-a; Conditions on its Application and Output, MIT Press. Lee. Hong Bae (1970) A Study o f Korean Syntax, Seoul: Pan Korean Book Corp., Seoul. Korea. Lee, Hyosang (1982) “Asymmetry in Island Constraints in Korean,” ms., Department of Linguistics, UCLA. Lee, Keedong (1993) A Korean Grammar on Semantic-Pragmatic Principles, Hankwukmwunhwasa, Seoul, Korea. Lee, Ik-Hwan (1980) Korean Particles, Complements and Questions: A Montague Grammar Approach, Hanshinmwunhwasa, Seoul, Korea. Lee, Maeng-Sung (1967) Nominalization in Korean, Ph.D. dissertation, University of Pennsylvania. Lee, Young-Suk (1993) “Licensing and Semantics of Any Revisited,” S. Kuno et al. (eds.) Harvard Studies in Korean Linguistics, Vol.5.577-592. Li, Y.-H. Audrey (1992) “Indefinite WH in Mandarin Chinese,” Journal o f East Asian Linguistics 1,125-155. Linebarger, Marcia, C. (1980) The Grammar o f Negative Polarity, Ph.D. dissertation, MIT. Mahajan, Anoop (1990) The A/A-bar Distinction and Movement Theory, Ph.D. dissertation, MIT. May, Robert (1977) The Grammar o f Quantification, Ph.D. dissertation, MIT. May, Robert (1985) Logical Form: Its Structure and Derivation, MIT Press. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 297 May, Robert (1988) “Ambiguities of Quantification and Wh: A Reply to Williams/’ Linguistic Inquiry 19,118-135. Milner, J.-C. (1979) ”Le Systeme de la Negation en Francais et L’opacite du Sujet.” Langue Francaise. Miyagawa, Shigeru (1987) “LF Affix Raising in Japanese,” Linguistic Inquiry 18,362- 367. Murasugi, Keiko (1991) Noun Phrases in Japanese and English: A Study in Syntax, Leamability and Acquisition, Ph.D. dissertation, University of Connecticut. Nam, Seungho (1993) “Another Type of Negative Polarity Items,” CSLI. Nishigauchi, Taisuke (1986) Quantification in Syntax, Ph.D. dissertation, UMass. Nishigauchi, Taisuke (1990) Quantification in the Theory o f Grammar, Kluwer. Nishiyama, Kumio, John Whitman, and Eun-Young Yi (1994) “Syntactic Movement of Overt WH-Phrases in Japanese and Korean,” ms., Department of Linguistics, Cornell University. Ohno, Yutaka (1991) “Arguments against Unselective Binding in Korean,” to appear in Susumu Kuno et al.(eds.) Harvard Studies in Korean Linguistics IV, Hanshin Publishing Co., Seoul. Park, Seung-Bin (1935) Cosenehak (Korean Linguistics), Cosenehakhoi, Seoul, Korea. Pesetsky, David (1982) Path and Categories, Ph.D. dissertation, MIT. Pesetsky, David (1987) “WH-in-situ: Movement and Unselective Binding,” E. Reuland and A. ter Meulen (eds.) The Representation o f (In)definiteness, MIT Press. Pollock, Jean-Yves (1989) “Verb Movement, Universal Grammar, and the Structure of IP,” Linguistic Inquiry 20,3655-424. Reinhart, Tanya (1983) Anaphora and Semantic Interpretation, Crom Helm, London. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 298 Reinhart Tanya (1987) “Specifier and Operator Binding,” in E. Reuland and A. ter Meulen (eds.) The Representation o f (In)definiteness, 130-167, MIT Press, Cambridge, MA. Reinhart, Tanya (1992) “Interpreting WH-in-situ,” to appear in Amsterdam Colloquium. Rizzi, Luigi (1990) Relativized Minimality, MIT Press, Cambridge, Mass. Rizzi, Luigi (1991) Residual Verb Second and the WH-Criterion, Technical Reports in Formal and Computational Linguistics, Faculte des Lettres, University of Geneva. Ross, John Robert (1967) Constraints on Variables in Syntax, Ph.D. dissertation, MIT. Rudin, Catherine (1985) “Multiple WH Movement and the Superiority Condition,” Paper Presented at LSA Annual Meeting, Minneapolis. Rudin, Catherine (1988) “On Multiple Questions and Multiple WH Fronting,” Natural Language and Linguistic Theory 6,445-501. Saito, Mamoru (1982) “Scrambling, Topicalization and Strong Crossover,” ms., Department of Linguistics and Philosophy, MIT. Saito, Mamoru (1985) Some Asymmetries in Japanese and Their Theoretical Consequences, Ph.D. dissertation, MIT. Saito, Mamoru (1989) “Scrambling as Semantically Vacuous A’-Movement,” in M.R. Baltin and A. S. Kroch (eds.) Alternative Conceptions o f Phrase Structure, The University of Chicago Press, 182-200. Saito, Mamoru (1992) “The Additional-WH Effects and the Adjunction Site Theory,” ms., Department of Linguistics, University of Connecticut. Sloan, Kelly (1991) “Quantifier-WH Interaction,” MIT Working Papers in Linguistics Vol. 15,219-237. Sohn, Keun-Won (1994) “Overt-Covert Licensing and Parametric Difference in NPIs,” ms., Department of Linguistics, University of Connecticut. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 299 Sohn, Keun-Won (1995) Negative Polarity Items, Scope, and Economy, Ph.D. dissertation, University of Connecticut Song. Seok Choong (1971) “A Note on Negation in Korean,” Linguistics 76:59-76. Song, Seok Choong (1973) “Some Negative Remarks on Negation in Korean,” Language Research 9.2:252-263. Song, Seok Choong (1988) Explorations in Korean Syntax and Semantics, Korean Research Monograph 14, Institute of East Asian Studies, University of CalifomiarBerkeley. Srivastav, Veneeta (1994) “Scope Marking as Indirect WH-Dependency,” Natural Language Semantics 2,137-170. Stowell, Tim (1981) Origins o f Phrase Structure, Ph.D. dissertation, MIT. Suh, Chung-Mok (1987) A Study on the Interrogative Sentences in Korean (in Korean). Thap Press, Seoul, Korea. Suh, Chung-Soo (1985) “Kwuke Uymwunmwun-uy Mwunceycem (Issues in Korean Questions),” Kwukehak Nonchong (Papers in Korean Linguistics), Changhakci. Seoul, Korea. Suh, Chung-Soo (1990) A Study o f Korean Grammar (in Korean), Hankwukmwunhwasa, Seoul, Korea Suh, Jinhee (1990) Scope Phenomena and Aspects o f Aspects o f Korean Syntax, Ph.D. dissertation, University of Southern California. Takakhashi, Daiko (1993) “Movement of WH-Phrases in Japanese,” Natural Language and Linguistic Theory 11,655-678. Torrego, Esther (1985) “On Empty Categories in Nominals,” ms., Department of Linguistics, UMass, Boston. Tsai, Wei-Tien (1994) On Economizing the Theory ofA-bar Dependencies, Ph.D. dissertation, MIT. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 300 Wachowicz, Kristina A. (1974) “Against the Universality of a Single WH-Question Movement,” Foundations o f Language 11, 155-166. Watanabe, Akira (1991) “WH-in-situ, Subjacency, and Chain Formation,” ms.. Department of Linguistics and Philosophy, MIT Watanabe, Shin (1995) Aspects o f Questions in Japanese and Their Theoretical Implications, Ph.D. dissertation, University of Southern Califomia. Whitman, John (1995) “Kakarimusubi from a Comparative Perspective,” Paper Presented at the 6th Japanese/Korean Linguistics Conference. Williams, Joseph M. (1975) Origins o f the English Language: A Social and Linguistic History, The Free Press. Williams, Edwin (1986) “A Reassignment of the Functions of LF,” Linguistic Inquiry 17,265-299. Williams, Edwin (1988) “Is LF Distinct form S-Structure? A Reply to May,” Linguistic Inquiry 19, 135-146. Yang, Dong-Whee (1983) “The Extended Binding Theory of Anaphors,” in Language Research 19,169-192. Yang, Dong-Whee (1992) “A Minimalist Approach to Quantification,” ms., Department of English Language and Literature, Seoul National University, Seoul, Korea. Yang, Hyun-Kwon (1990) Categories and Barriers in Korean, Ph.D. dissertation, University of Texas at Austin. Yang, In-Seok (1972) Korean Syntax: Case Markers, Delimiters, Complementation, and Relativization, Ph.D. dissertation, University of Hawaii. Yoon, Hang-Jin (1990) Functional Categories and Complementation: in English, Korean, and Turkish, Ph.D. dissertation, University of Wisconsin-Madison. Yoshimura, Noriko (1992) Scrambling and Anaphora in Japanese, Ph.D. dissertation. University of Southern California. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

Emotions: Linguistic representation and cultural conceptualization

PDF

Representation of focus and presupposition in Japanese

PDF

Paiderastia in Sophocles' "Philoctetes"

PDF

In the event of focus

PDF

The syntax of the Chinese BA-constructions and verb compounds: A morphosyntactic analysis

PDF

On the status of possessives

PDF

Language and context: toward a model of psycholinguistic competence

PDF

Frequency in sentence comprehension

PDF

The syntax of clitic doubling in modern Greek

PDF

The syntax of focus and topic in Mandarin Chinese

PDF

On the nature of particles in Japanese and its theoretical implications

PDF

The accidence and syntax in john Wallis' 1653 "grammatica linguae anglicanae": a translation and a commentary on its alleged relationship to the 1660 port-royal "grammaire generale et raisonnee"

PDF

The other world in the New World: Representations of the supernatural in short fiction of the Americas

PDF

Morphology and the lexicon: Exploring the semantics-phonology interface

PDF

A case grammar of the parker manuscript of the "Anglo-Saxon chronicle" from 734 to 891

PDF

Justus Georg Schottel and linguistic theory

PDF

Language services planning in the banking industry: An example of unplanned language policy

PDF

The life and literary and artistic activities of Robert Baldwin Ross, 1869-1918

PDF

Borrowing and borrowability

PDF

Rhetorical abstraction as a facet of expected response: A structural equation modeling analysis

Asset Metadata

Creator Chung, Daeho (author)

Core Title On the representation and licensing of Q and Q-dependents

Degree Doctor of Philosophy

Degree Program Linguistics

Publisher University of Southern California (original), University of Southern California. Libraries (digital)

Tag language, linguistics,OAI-PMH Harvest

Language English

Contributor Digitized by ProQuest (provenance)

Permanent Link (DOI) https://doi.org/10.25549/usctheses-c17-512375

Unique identifier UC11352803

Identifier 9705084.pdf (filename),usctheses-c17-512375 (legacy record id)

Legacy Identifier 9705084.pdf

Dmrecord 512375

Document Type Dissertation

Rights Chung, Daeho

Type texts

Source University of Southern California (contributing entity), University of Southern California Dissertations and Theses (collection)

Access Conditions The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the au...

Repository Name University of Southern California Digital Library

Repository Location USC Digital Library, University of Southern California, University Park Campus, Los Angeles, California 90089, USA