Close
About
FAQ
Home
Collections
Login
USC Login
Register
0
Selected
Invert selection
Deselect all
Deselect all
Click here to refresh results
Click here to refresh results
USC
/
Digital Library
/
University of Southern California Dissertations and Theses
/
A comparative study of focus constructions
(USC Thesis Other)
A comparative study of focus constructions
PDF
Download
Share
Open document
Flip pages
Contact Us
Contact Us
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. The 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 margins, 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 corner and continuing from left to right i n 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. UMI A Bell & Howell Information Company 300 North Zeeb Road. Ann Arbor. Ml 48106-1346 USA 313/761-4700 800/521-0600 A COMPARATIVE STUDY OF FOCUS CONSTRUCTIONS by Dong-In Cho A Dissertation Presented to the FACULTY OF THE GRADATE SCHOOL UNIVERSITY OF SOUTHERN CALIFORNIA In Partial Fulfillment of the Requirements for the Degree DOCTOR OF PHILOSOPHY (Linguistics) December 1994 Copyright 1994 Dong-In Cho UMI Number: 9600964 Copyright 1994 by Cho, Dong-In All rights reserved. UMI Microform 9600964 Copyright 1995, 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 UNIVERSITY OF SOUTHERN CALIFORNIA THE GRADUATE SCHOOL UNIVERSITY PARK LOS ANGELES, CALIFORNIA 90007 This dissertation, written by Dong-In Cho under the direction of hi.?. 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 of Graduate Studies Date ...^.Y™™. 1 :..?.'...?-.?.?.'?.. DISSERTATION COMMITTEE Chairperson ^J^^rJ^.>^rr^.V^.>. ^j^t/l±U 3^^.^Jt^^Z2 Acknowledgments I would like to express my sincerest gratitude to those who enriched my intellectual journey at USC and made this dissertation possible. The great range of intellectual talent and interests among the faculties at USC is what has made these past five years such a rewarding and enriching experience. First, I would like to thank my committee members, past and present, Joseph Aoun, Hajime Hoji, Audrey Li, Maria Polinsky, and Mario Saltarelli for their encouraging discussions and insightful criticism on my often rudimentary ideas. Most of my intellectual life at USC was spent with Joseph Aoun, Hajime Hoji, and Audrey Li, without whom I cannot imagine my graduate study at USC. I would especially like to thank Hajime Hoji, the chair of my committee, for his inspiring guidance and his spending numerous hours with me over the last five years. His influence is on every page of my dissertation. I also would like to express my gratitude to Audrey Li for her kind guidance, encouragement, and for being available any time I needed her guidance. She never stopped encouraging me and made me feel that I am a real linguist. I also thank Joseph Aoun, who was my former advisor before he became Dean of Faculty, for all his keen suggestions at my numerous presentations in his office. With his well-organized and exciting lectures, he convinced me that formal syntax could be an exciting and worthwhile thing to become involved in. ii I am deeply grateful to all of my teachers outside my committee, Bernard Comrie, Jean-Roger Vergnaud, Maria Luisa Zubizarreta, Jim Gee, and Marx Seidenberg for their linguistic inputs during my study at USC. I am specially thankful to Jean-Roger Vergnaud, who provided insightful suggestions for my analysis. Linguists outside USC and visiting scholars at USC have also influenced my dissertation: Lisa Cheng, Stanley Dubinsky, Naoki Fukui, James Huang, Yuki Kuroda, Giuseppe Longobardi, and Tim Stowell. I especially thank James Huang whose detailed comments and insightful suggestions on my papers which eventually developed into Chapter III in this dissertation. This dissertation would not have been worth it without my fellow students at USC. I want to thank Nancy Antrim, Hiroshi Aoyagi, Robin Belvin, Elabbas Benmamoun, Jose Camacho, Daeho Chung, Gorka Elordieta, Kaoru Horie, Miao-Ling Hsieh, Kyoung-An Kim, Sechang Lee, Marcus Maia, Yuki Matsuda, Keiko Miyagawa, Liliana Sanchez, Patricia Scheider-Zioga, Dingxu Shi, Shu-Ing Shyu, Maki Watanabe, Shin Watanabe, and Ke Zou. Nancy deserves special mention: she spent considerable time proof-reading my papers and dissertation and she was the sources of the judgment for my English sentences. I also thank Hiroshi and Keiko for countless dissuasions on E-mail. I have also benefited from discussions with students in the Korean Linguistics Circle: Myung-kwan Park, Young-Suk Lee, Jeong Shik Lee, Keun-won Sohn, Jongbok Kim, Jeong-Ryeol Kim. Myoung-kwan Park deserves a special mention. Our conversions via long-distance calls always m turned out to be valuable. We even joined telephone company's special program for frequent and long discussions. For their friendship and care, I would like to thank Seokhoon Yoo, Taegoo Chung, Mihee Kim, Ki-Seok Lee, Young-Joon Jang, Seong Chan Kim, Sechang Lee and his wife Lanju Kim, Shinsook Lee, Hey-young Um, and Hikyoung Lee as well as Kwangmo Lee and Ikhee Kang, especially Kwangmo Lee and Sechang Lee who took good care of me when I was suffering after a car accident or too busy working on my dissertation. I am also very grateful to Yong-Seok Suh, Jong-Yuri Yoon, Jae-woong Choe, and Hyun-Kwon Yang for their constant encouragement and care. For their friendship and encouragement, I want to thank Kaoru Ohta and his wife Amy, Kuo-ming Sung and his wife Mei-ling. Very special thanks should go to Kuo-ming and Kaoru, who spent numerous hours listening to my thesis writing during the past months. My deepest gratitude goes to my former teachers at Korea University who taught me Linguistics: Professor Sung-Shik Cho, the late Professor Kwon-Ho Kim, Professor Yong-Jae Lee, Professor Kyung-Ja Park, and Professor Kiyong Lee for providing me an intellectual basis and input for my graduate study in Linguistics. I am also grateful to Professor Keeho Kim for his care. Without their encouragement and trust, this academic journey might have been much harder than it turned out to be. I find my verbal expressions are not enough to show my gratitude to them. My teaching experience at UCLA as a full-time lecturer during 1993- 1994 gave me a good opportunity to work with considerate faculties. For their support, I thank Noriko Akatsuka, John Duncan, Shoichi Iwasaki, IV Peter Lee (Chairperson), Shu-mei Shih, Masako Douglas, Sung-ock Sohn, and Kuo-ming Sung. I am especially grateful to Kuo-ming Sung for his mental and practical support when I badly needed them. My life at UCLA without him would have been unthinkable. My study at USC was made possible by a Dean Fellowship from USC for the first four years from 1989 to 1993 and by a lectureship at UCLA from 1993 to 1994. I also acknowledge with gratitude the fellowships from Korean Senior Citizens' Mutual Club (1990), the Korean American Scholarship Foundation (1991-1992), the Korean Ambassador (1992), and the Olympic Lion's Club (1992-1993). Finally, I would like to thank my family. My brothers, sisters, brother-in-laws, and sister-in-laws always kept me as one of their family members. Most of all, I would like to thank my parents, who never stop praying for me, trusting me, giving me unconditional love. This dissertation is dedicated to my parents for their bottomless love and patience. v Abstracts This thesis provides evidence supporting the claim that the DO-verbs in Japanese and Korean, namely su, ar, and ha in focus constructions are base-generated (Base-generation Hypothesis), contrary to the DO-support Hypothesis, which suggests that DO is inserted to rescue a stranded morpheme. This thesis also provides a comparative study of Japanese and Korean with respect to the category of negative items and the obligatory vs. optional presence of focus markers in Japanese and in Korean focus constructions, respectively. Verbal inflections on DO, the multiple occurrence of DO in a simplex sentence, and the co-occurrence relation between the thematic verb in the lower VP and DO in the higher VP indicates that DO in focus constructions cannot be analyzed as being inserted to support stranded morphemes like tense. We propose that the co-occurrence relation can be captured by positing the abstract verbs dO and bE, which have feature bundles. Under this analysis, dO is phonetically realized as ha in Korean and, as su with verbs in Japanese. The verb bE, however, is phonetically realized only in Japanese as ar sinceKorean does not have adjective predicates. The multiple occurrences of DO in a simplex sentence supports the Base-generation hypothesis. Under the DO-support Hypothesis, more than one INFL has to be stipulated for DO to be generated under it in a simplex sentence. Support for the Base-generation Hypothesis is also found in VP-fronting and the lack of VP-ellipsis. In the double DO vi constructions, the complement of DO containing the other DO can be fronted, indicating that DO is not generated under INFL. This thesis proposes that the obligatory presence of a focus marker in focus constructions, which has been a strong argument for the DO- Support Hypothesis (Kuroda 1965), can be attributed to the requirement that the case in the case assigner must be discharged and to the postulation that focus markers qualify the complement of DO, (that is, the focused VP), as Case assignee by providing the [+N] feature to it. Establishing that DO in the focus construction is a base-generated main verb paves the way to a unitary account for both focus constructions and light verb constructions (Grimshaw and Mester 1988). vii Table of Contents Chapter I. Introduction 1 1.1. Preliminaries 1 1.2. Outline of th e Thesis 6 Chapter II. Against the DO-Support Hypothesis 9 2.1. Introduction 9 2.2. Against the DO-Support Approach 14 2.2.1. DO in Negation Constructions 15 2.2.1.1. Ar in Japanese Negative Sentences 16 2.2.1.1.1. Evidence from Inflection 17 2.2.1.1.2. Evidence from the Co-occurrence Relation 18 2.2.1.1.3. Evidence from the Nominalizer sa 22 2.2.1.1.4. Indirect support from Prohibition Negative mal in Korean 24 2.2.1.1.5. The Japanese Neg structure and Multiple NegP/AdjP 26 2.2.1.2. Ha in Korean Negative Sentences 28 2.2.1.3. Adverbial Status of Negative ani 29 2.2.1.3.1. Apparent Problems with the Adverbial Approach 34 2.2.1.3.2. Toward the solution 37 2.2.2. DO in affirmative sentences 40 2.2.3. DO in Focus Constructions 44 2.2.4. VP-ellipsis 46 2.2.5. Multiple Occurrence of DO 48 2.2.6. VP-fronting 51 2.3. Problems and their solutions 55 2.3.1. An Obligatory Focus Marker in Focus Constructions 58 2.3.1.1. Poser (1992): Blocking of Phrasal Constructions 58 2.3.1.2. The Underspecified Complementizer Approach 62 2.3.1.3. Emphatic Markers 63 2.3.2. Obligatory vs. Optional Presence of the Focus Markers 69 2.4. Summary and Concluding Remarks 72 Chapter III. Two Types of VNs: Underspecified VN 77 3.1. Introduction 77 3.2. Dubinsky (1994): Underspecified Lexical Categories 80 3.3. Proposals 82 viii 3.3.1. Zero Derivation 84 3.3.2. Underspecification Approach: [0V, -N] feature 85 3.3.3. [+N] feature 86 3.4. The Nominal Feature of VN 87 3.4.1. Miyagawa( 1987): Unitary VN with [-V,+N] 88 3.4.2. Further Evidence for the nominal feature of VNs 89 3.5. Unitary [-V, +N] and double-o constraint 91 3.5.1. Kageyama (1982) and Miyagawa (1987) 91 3.5.2. Unincorporated VNs and double -o 94 3.6. A Verbal ([+V, -N]) approach to the VN 96 3.6.1. Adjectival modification 96 3.6.2. Assignment of Verbal Case 97 3.6.3. Adverbial modification 100 3.7. Apparent Problems with the verbal approach to verbal VNs 102 3.7.1. No Verbal Inflections 103 3.7.2. Aspect and Verbl Case Assignment 104 3.7.3. The Nominal Property of verbal VN in [NP-ACC VN] 105 3.8. Null VP-Complementizers in Light Verb Constructions 107 3.8.1. [±N] Complementizer 107 3.8.2. The null-complementizer as a VP-complementizer Ill 3.8.3. Functional Projections selected by a lexical category 114 3.9. Constituency of VNP 118 3.9.1. Focus Constructions 118 3.9.2. VP-fronting Constructions 120 3.10. Predictions 121 3.10.1. Aspectual Constructions 122 3.10.2. to-Clause and lexical specification of VNs 124 3.11. Conclusion 129 Chapter IV. LF-Incorporation 131 4.1. Introduction 131 4.2. Hoshi and Saito (1993): LF-incorporation 132 4.3. Proposal 136 4.4. Theta-grid 142 4.5. Aspectual Feature Transfer 146 4.6. Selectional Restrictions 148 4.7. Case feature Transfer 150 4.7.1. The distribution of the VN-Phrase 154 4.7.1.1. LF-incorporation Approach 155 4.7.1.2. ECP Approach to the distribution of VNP 159 4.7.1.2.1. Stowell (1981): ECP Approach and CRP 159 4.7.1.2.2. ECP and CRP in Korean 161 4.8. Conclusion 164 ix Chatpter V. Summary and Concluding Remarks 165 Bibliography 167 x Chapter I. Introduction 1.1. Preliminaries This thesis investigates focus and light verb constructions, containing the verb su or ar in Japanese and ha in Korean. The light verb constructions and focus constructions are extensively discussed within many different fields of linguistics. 1 We approach these constructions within the framework of the Principles and Parameters Theory developed in Chomsky (1981) and subsequent works. The cover term DO is used throughout this thesis for expository purposes to differentiate the verb DO in focus and negation constructions from the light verb in the Light Verb Constructions, which has the same morphology with DO. Therefore, DO includes su and ar in Japanese focus constructions and ha in Korean focus and long-form negation constructions. The canonical sentences for focus constructions and the light verb in Korean construction are demonstrated in (l). 2 lr rhe light verb constructions are extensively discussed in the literature (Grimshaw and Mester 1988; Kageyama 1982:1991; Miyagawa 1989:1990; Poser 1991; Terada 1990; Hasegawa 1991; Tsujimura 1990; Ahn 1990; Lee 1991; Kim 1991; Park 1992; Hoshi 1993; Sells 1990; Cho 1993c; Urushibara 1993; Dubinsky 1994; Huang 1994) . Focus constructions are also discussed by many linguists (Kuroda 1965; Song 1988, Kang 1988; Hoji 1991a; Cho 1992a). 2 The gloss FOC(us) is used throughout the dissertation for the nun in the VP-focus construction to differentiate it from the Topic marker which has the same morphology as the focus marker. 1 Focus Construction (1) John-i [vp chayk-ul ilk-ki]-nun ha-ess-ta John-NOM book-ACC go-COMP-FOC do-PAST-DEC 'John at least read a book.' Light Verb Construction (2) John-i ["VNP swuhak-ul kongpwu]-lul ha-ess-ta John-NOM Math-ACC study-ACC do-PAST 'John studied Math.' The goal of this thesis is to provide a unified account for both focus constructions and light verb constructions in Japanese and Korean. We first argue for the structural parallelism between these two constructions. First, we argue that DO in focus constructions is base-generated just like light verbs. Second, we claim that the so-called verbal nouns like kongpwu 'study' in (lb) are specified for [0V, -N], not for [-V, +N], contra Miyagawa (1987) . Third, we claim that light verb constructions have a null VP- complementizer corresponding to the overt VP-complementizer ki in (la). Based on the structural parallelism, we argue for the LF-incorporation of thematic verbs or verbal nouns to DO or a light verb. We claim that the LF-incorporation is responsible for the choice of DO or light verbs, the realization of the aspect of embedded verbs or verbal nouns on DO or light verbs, and the transfer of the case feature of embedded verbs or verbal nouns to DO or light verbs. Whether DO is base-generated in focus constructions in Japanese and Korean has been controversial. Some researchers claim that DO is inserted under INFL to rescue a stranded Tense morpheme (Kuroda 1965 for Japanese, and Kim 1967, Oh 1971, and Cho 1993a for Korean), some 2 others suggest that it is base-generated (Hoji 1991a; Park 1992; Poser 1992). Proponents for the DO-support Hypothesis suggest that the DO- support occurs if focus markers such as man 'only' and kkaci 'even' follow verbs. At first sight, the generalization that DO co-occurs with focus markers seems to be captured under the verb movement analysis advocated in Pollock (1989) and Chomsky (1989). Under this analysis, when verb movement is block by the head of EmpP (Emphasis Phrase) or FocP (Focus Phrase), the DO-support occurs to support the stranded Tense morpheme, as suggested in Cho (1993a). Verb movement for Korean and Japanese has been proposed in Whitman (1991). This DO-support Hypothesis, however, faces both theoretical and empirical problems. Theoretically, the bound morphemes such as focus markers in Japanese and Korean cannot be blocking elements for the verb movement. Following Lasnik's morphological filter in (2) below, Ouhalla (1989) and Benmamoun (1992) claim that only free morphemes, but not bound morphemes, block verb movement. (2) A morphologically realized affix must be a syntactic dependent of a morphologically realized category at Surface Structure. If only free morphemes block verb movement cross-linguistically as claimed in Ouhalla (1989) and Benmamoun (1992), positing focus markers as blocking elements is a weak argument, since focus markers in Japanese and Korean are bound morphemes. Empirically, as we claim in the thesis, the lack of VP-EUipsis, the presence of DO in a fronted VP, the multiple occurrence of DO in a simplex 3 sentence, and the co-occurrence relation between DO and the head of its complement VP argue against the DO-support Hypothesis for Japanese and Korean. Based on the arguments for the Base-generation Hypothesis given above, we can take the position that DO and the light verb are both base-generated. We also find parallelism between the embedded verbs in the focus constructions and verbal nouns in the light verb construction. The category of verbal nouns have been controversial. Some researchers argue for a unitary nominal feature of verbal nouns (Miyagawa 1987), while some others argue for the verbal feature of verbal nouns (Hasegawa 1991; Park 1992). Based on the impossibility of adjectival modification and Case assignment by the verbal noun, and their not being able to bear verbal inflections, this thesis claims that the verbal nouns in the contexts such as (lb) are not nouns; rather, they are underspecified for [V], and negatively specified for the [N] feature. The syntactic underspecification for verbal nouns has be argued for by Dubinsky (1994). The underspecification theory in Dubinsky (1994), however, cannot capture the fact that the verbal nouns in (lb) cannot be modified by adjectives, while our specification correctly predicts the impossibility of the adjectival modification. We also claim that there is a parallelism between the focus constructions and light verb constructions in terms of the presence of a VP-complementizer. As suggested in Lee (1991), the light verb constructions, however, have a null VP-complementizer while the focus constructions have an overt one. We provide two pieces of evidence 4 supporting this point. First, given that the [+N] categories qualify as Case assignees (Stowell 1981) and that the verbal noun in (lb) is specified for [0 V, -N], for the verbal noun phrase to qualify as a Case assignee, a [+N] category is needed. We suggest that the null complementizer is specified for [+N]. Second, given the hypothesis that we develop in this thesis, lexical categories do not directly select another lexical category; rather, they select a functional category. If this hypothesis is correct, the lexical category light verbs do not directly select the lexical category verbal noun phrase, but a functional category under which lexical categories appear. We suggest that the functional category in this case is the null VP- complementizer. Based on the structural parallelism between the focus constructions and the light verb construction, we suggest the LF-incorporation to light verbs from theta-verbs discussed in Hoshi and Saito (1993) can be extended to the focus constructions. Hoshi and Saito (1993) claim that the LF-incorporation is motivated to discharge the theta-roles of the verbal noun during the derivation between S-structure and LF (Larson 1988; Chomsky 1992). We propose that this LF-movement is motivated to satisfy a selectional restriction between DO or the light verb and the head of their complements. This LF-movement hypothesis is supported by the transfer of theta-grid, aspectual grid, and case grid features from the theta- assigning items to DO or light verbs. Our proposal of selection-motivated movement at LF has a desirable ramification in the general theory of grammar. Given that the relation of verbs and their s-selected NPs is not local due to the intervening DP (Fukui 1986; Abney 1987), the LF- 5 movement of N to the position where verbs and Ns are close can receive a straightforward account under the proposed LF-incorporation motivated by the selectional restriction. 1.2. Outline of the Thesis The rest of the thesis is organized as follows. Chapter II argues, following Hoji (1991a), that DO in focus constructions is base-generated with the categorial feature [+V], not inserted to rescue a stranded morpheme. We claim that the absence of VP-ellipsis and the presence of multiple occurrences of DO support the Base-generation Hypothesis. We further suggest that the Base-generation Hypothesis is supported by a co-occurrence relation between thematic verbs or adjectives in the lower VP and DO in the upper VP. We also propose that the negative items in Japanese and Korean are adjective and adverb, respectively, which do not constitute a blocking element for verb movement. We investigate the issue of the obligatoriness of focus particles on the VP-complementizer in the VP-focus constructions, which has been a basis for the Do Support Hypothesis. We suggest that the focus markers on the focused VPs are needed due to the requirement that the Case feature of a Case assigner has to be discharged. Chapter III discusses two types of verbal nouns. We claim that the so-called Verbal Nouns (henceforth, VN) with the same morphology have two distinct lexical features: a nominal VTVwith the lexical feature [-V, +N] 6 and a verbal VN with the lexical feature [0 V, -N]. We specifically develop the syntactic underspecification theory for these so-called verbal nouns. We also claim that the verbal noun phrase is governed by a null VP- complementizer, attempting to provide a parallel account for Light verb constructions and focus constructions. We propose the hypothesis that the lexical item does not directly select the maximal projection headed by another lexical category; rather, it selects a maximal projection headed by a functional category. Under this hypothesis verbs select functional projections like DP, IP, or CP, but not NP or VP. If this is the case, the light verb does not directly select the verbal noun phrase headed by a verbal noun; rather, it selects a functional projection, which is a null complementizer containing the verbal noun phrase in light verb constructions. Chapter IV provides a unified account for focus constructions and light verb constructions in terms of the choice of DO or light verbs, realization of the aspectual marker, and reflexive Case. We propose that the LF-incorporation of a theta-noun to the light verb or DO is motivated to satisfy a selectional restriction. We further extend the LF-incorporation hypothesis to the focus/negation constructions in Japanese and Korean. We specifically propose that the verbal noun ([0V, -N]) in light verb constructions and the theta-verb in the focus/negation constructions adjoins to DO at LF. This LF-movement is responsible for the choice of ar or su in Japanese, reflexive Case in Korean and Japanese, and theta-role transfer in Japanese and Korean. 7 Chapter V svimmarizes this thesis and presents some implications of the proposal made in the previous chapters. 8 Chapter II. Against the DO-Support Hypothesis 2.1. Introduction In this chapter, we argue that su and ar in Japanese focus constructions and ha in Korean focus and negation constructions are base-generated, not inserted to rescue a stranded morpheme under INFL. The canonical examples containing these verbs are illustrated in (1) and (2). 1 Korean (1) a. John-i ka-ki-kkaci John-NOM go-COMP-even 'John even went.' b. John-i ka-ci John-NOM go-COMP 'John did not go.' Japanese am not ha-ess-ta (Focus) HA-PAST-DEC ha-ess-ta (Negation) HA-PAST-DEC (2) a. John-ga ik-i-sae si-ta John-NOM go-COMP-even SU-PAST 'John even went.' b. John-ga ko -na -ku -ar-ta John-NOM come-NEG-COMP-AR-PAST 'John did not come.' (Focus) (Negation) In Korean, both in focus constructions, as in (la), and long-form negation constructions, as in (lb), ha appears. In Japanese, while su follows a verb 1 Throughout this dissertation, ki and ci in Korean and i in Japanese are referred to as COMP (complementizer). Following Kang (1988) and Hoji (1991a), we assume that these are VP-complementizers, which take VP as their complement. 9 in the focus constructions, ar follows the negative lexical item na in negation constructions. The main goal of this chapter is to provide evidence supporting the hypothesis that DO is base-generated. This hypothesis will be referred to as the Base-generation Hypothesis. An alternative which states that DO is inserted to support a stranded morpheme under INFL is referred to as the DO-support Hypothesis. The Base-generation Hypothesis has been argued for in Song (1988) and Hoji (1991a), contra Kuroda (1965) for Japanese, and Kim (1967), Oh (1971), and Cho (1993a) for Korean, who suggest the DO-Support Hypothesis. We provide evidence for the Base-generation Hypothesis from the co occurrence relation between DO and the head of its complement, the multiple occurrence of DO in a simplex sentence, the lack of VP-ellipsis, and the presence of DO in a fronted VP. Multiple occurrences of DO in a simplex sentence and the appearance of DO in the fronted VP support the Base-generation Hypothesis. Under the DO-support Hypothesis, more than one INFL has to be stipulated for DO to be generated under it. Under the Base-generation Hypothesis, multiple negation in Korean can be attributed to the multiple based- generating of DO and the preverbal negative adverb ani being generated as a sister of DO, as in Larsonian structure (Larson 1988). Evidence for the adverbialhood of ani is found in the attachment of the plural marker tut. The adverb ani undergoes head-movement to the verb. Evidence is found from the reduplication of the focus phrase. While other adverbs can be optionally reduplicated in the reduplication of the focus phrase, ani is 10 obligatorily reduplicated. This indicates that ani and the verb behave like a lexical unit after the incorporation. The appearance of DO in the fronted VP also supports the Base- generation Hypothesis. Given that VP-fronting does not front the Tense morpheme, the appearance of DO in the fronted VP argues against the DO- support Hypothesis. The lack of VP-ellipsis further supports the Base- generation Hypothesis in Japanese and Korean. The co-occurrence relation between DO and a head of its complement supports the Base-generation Hypothesis. The verbs ha in Korean and su co-occur with verbs, while ar co-occurs with adjectives, as the following Japanese sentences exhibit. (3) a. tabe-wa si-ta eat-FOC SU-PAST 'did eat' b. samuku-wa at-ta cold-FOC AR-PAST 'was cold' In Korean, ha co-occurs with verbs. Adjective predicates in Japanese are state verbs in Korean (Choe 1929). We capture this state of affairs for both Japanese and Korean by positing the abstract verbs dO and bE, which have their own feature bundles. The abstract verb dO co-occurs with verbs and bE occurs with adjectives, taking VPs and AdjPs as their complements, respectively. The same abstract verb dO is phonetically realized as su in Japanese and na in Korean in the same environments. The verb bE occurs only in Japanese as ar since Korean does not have 11 adjectival predicates. The co-occurrence relation suggests that the Japanese negative item ha is a lexical category adjective, as Kitagawa (1986) pointed out. 2 This is because the negative item na co-occurs with ar. Other evidence further supports the Adjectivehood of na: na bears the same verbal inflection as a verb (Kitagawa 1986); na takes -sa for nominalization as adjectives do. The consequence of this approach is that multiple NegPs in Japanese can be taken as the multiple occurrence of AdjPs; so that, we need not stipulate that the functional projection NegP iterates in Japanese. We suggest that the obligatory presence of a focus marker in focus constructions can be attributed to the requirement that VPs need Case in the complement position of a verb. We propose that the VP- complementizer -ki in Korean and -i are not specified for [+N], and, therefore, the VP complements cannot be assigned Abstract Accusative Case. We, following Stowell (1981) and Chomsky (1981), assume that only [+N] categories can be assigned Abstract Accusative Case. We also assume that focus makers can assign inherent Case just as postpositions do. Under this analysis, the VP without Case in the complement position of V, can be saved by a focus marker, which assigns inherent case to it. We, following Song (1988), Ahn (1990), and Hoji (1991a), suggest that -lul in Korean and -o in Japanese can function as emphatic markers. As pointed out by H. Hoji (p.c), these emphatic markers differ from focus markers, such as wa 'at least' and sae 'even', in that these emphatic 2 As far as I know, Kitagawa (1986) was the first to identify the negative morpheme na as an adjective. 12 markers do not have a full semantic value, and, therefore, the attachment of these two markers to VPs does not make focus constructions as natural as other regular focus markers do. In this sense, the emphatic markers can be taken as weak focus markers. We also suggest that -lul and -o are sensitive to the categories they attach to. While -lul can attach to non-[+N] categories, -o can attach only to t+N] categories. Therefore, the distribution of the emphatic marker -o is more restricted than that of-lul. This hypothesis is supported by the place where -lul can appear, but -o cannot. For one thing, -lul can be attached to verbal connectives in Korean, while -o cannot. Under this analysis the VP-focus with -lul is more acceptable than with -o, given that the VP is not [+N] category. In order to account for the optional vs. obligatory presence of focus markers in Korean and Japanese long-form negation, respectively, we, following Hoji (1991a), suggest that Korean and Japanese have abstract emphatic morphemes in negative contexts. We further propose that these abstract emphatic morphemes are covert counterparts of the overt emphatic makers -lul and -o. Under this analysis, the abstract morpheme can attach to the VP in Korean, while it cannot in Japanese. This is because the emphatic morpheme cannot attach to non-[+N] categories in Japanese. To save this situation, the overt focus marker is needed in Japanese long-form negation. This chapter is organized as follows. In Section 2.2, we find evidence for the Base-generation Hypothesis in the multiple occurrence of DO, the co-occurrence relation between DO and the head of it complements, the 13 lack of VP-ellipsis, and the appearance of DO in a fronted VP. We suggest that the Japanese negative item na belongs to the lexical category adjective. Evidence is found in the co-occurrence relation, inflection, and nominalization. We provide indirect support for the Adjectivehood of na by showing that Korean also has lexical negative items. We suggest that the negative mal and ani in Korean are a verb and an adverb, respectively. We discuss evidence for this claim in this section. In section 2.3, we account for the obligatory presence of a focus marker in focus constructions. We also treat the obligatory vs. optional presence of focus markers in Korean and Japanese long-form negations. In this section, we suggest that Poser's (1991) Blocking cannot account for the optional presence of focus markers in Korean long-form negation. We also propose that the VP-complementizers ki/ci in Korean and i/0 in Japanese are underspecified for [N], and therefore do not qualify as Case assignees. Hence, focus markers are obligatory for Case reasons. 2.2. Against the DO-Support Approach This section examines previous arguments for a DO-Support Analysis in Korean and Japanese and claims that evidence for DO-Support can be interpreted differently. Specifically, this section examines Negation and Focus contexts in Korean and Japanese, and claims that the appearance of DO in these contexts does not constitute evidence supporting a DO- Support Hypothesis in these languages. The appearance of DO in these 14 contexts has been used to argue for the DO-Support Hypothesis. This section shows that DO in Japanese and Korean can be base-generated. 2.2.1. DO in Negation Constructions In this section, we suggest that ha in Korean and ar in Japanese are base- generated. We provide evidence supporting the claim that the Japanese negative na belongs to the category of adjectives (1986). This categorial feature allows the multiple NegPs in Japanese. We also suggest that ar takes AdjP or NegP as a complement. As for Korean Neg, we discuss two types of Negs in Korean: 'prohibition' negative mal and regular negative ani. We claim that the prohibition negative mal is a verb, which heads its own projection and that ani is a complement of a verb. 3 Furthermore, we will suggest that the negative ani in Korean is an adverb. These different categories for negatives are not surprising if we consider the typological study of negation in Dahl (1979), where three types of negations are identified. These three types include 'adverb-like particles', 'auxiliary verb', and 'a morphological category on verbs'. We also suggest that ha in Korean takes VP as its complement. We will also provide additional empirical support for Song's (1988) claim that 3 We assume Larsonian structure (Larson 1988) where an adverb is a complement of a verb. This issue will be discussed in more detail in section 2.2.1.3. 15 the affirmative counterpart of the negative sentence is independently possible, questioning the DO-Support Hypothesis. 2.2.1.1. Ar in Japanese Negative Sentences In this section, we suggest that the Japanese ar is a base-generated taking AdjP as a complement, proposing that the Japanese Neg na is an adjective. 4 We further propose that the selectional relationship between ar and the negative morpheme na supports the claim that the Japanese Neg belongs to the category of adjectives. Before we examine the category of negative na in Japanese, let us compare Japanese and Korean adjectives to observe the presence of ar. Korean (4) sakwa-ka ppalkah-ess-ta apple-NOM red-PAST-DEC 'The apple was red.' Japanese (5) ringo-ga aka-kat-ta apple-NOM red-AR-PAST 'The apple was not red.' While the stative verbs (adjectival verbs) in Korean are not followed by DO, the Japanese adjectives are as seen in (5). In (5) the surface form kat is not revealing in that it does not seem to contain ar. Following H. Hoji's (p.c.) suggestion, I assume that the form 4 Kubo (1992) also suggests independently that ar in the focus construction is a base- generated stative verb. 16 ka-ta in (5) can be analyzed as ku-ar-ta. After a morphophonemic process, the form ku-ar-ta changes to kat-ta. We suggest that ku could be a Japanese counterpart of Korean ki/ci ore/a and that ku is a VP- complementizer as claimed in Hoji (1991a) for i/0. 5 2.2.1.1.1. Evidence from Inflection Returning to the category of ar, the first evidence supporting the claim that the negative morpheme belongs to the adjectival category comes from the fact that both Neg and Adjectives are followed by the same verb ar and undergo the same verbal inflections. The following pairs show this point. (6) a. tabe-na-i eat-NEG-NPAST b. tabe-na-kat-ta eat-NEG-AR-PAST (7) a. samu-i cold-NPAST b. samu-kat-ta cold-AR-PAST (Non-Past) (Past) (Non-Past) (Past) 5 Hoji (p.c.) pointed out to me that ku and i can be used as both a verb connective and VP-complementizer in Japanese. In Korean, -ale are used as verb connective and ki is used as a complementizer, but not a verb connective. In other words, in Japanese one morpheme has a dual function, while in Korean one morpheme has only one function. H. Hoji (p.c.) suggested to me that there might be a parameter which can predict the difference if we assume that Korean has a parameter in which the functional categories are more enriched than in Japanese. 17 As seen above, the adjective samu 'cold' in (7) and Negative na in (6) undergo the same pattern of inflections. Notice that the inflection on the verb in Japanese differs from that of Neg na and adjectives. Consider the inflections on the verb in (8). Japanese (8) a. tabe-ru eat-NPAST b. tabe-ta eat-PAST In (8) the verb tabe bears ru for nonpast and ta for past, as opposed to i and ta for adjectives and negative na. 2.2.1.1.2. Evidence from the Co-occurrence Relation Another argument, related to the first evidence, can be found in the co-occurrence relation between Neg/Adj and ar. Even though Japanese verbs themselves can bear inflectional morphemes, they must be followed by verbal su when they are attached to a focus marker, as seen in (9a). If adjectives or Neg are focused, ar follows, as seen in (9b) and (9c), respectively. (9) a. tabe-wa shi-ta eat-FOCUS SU-PAST 'did eat' b. samuku-wa at-ta cold-FOCUS AR-PAST 'was cold' 18 First of all, the constructions in (9) indicate that the verb co-occurs with verbal su on the one hand, while adjectives co-occur with verbal ar on the other. The intervening element wa does not block the co-occurrence relation. This co-occurrence relation can be expressed as in (10). (10) a. A (verb)—-WA—-A' (su) b. B (Adj) ----- WA—--B' (ar) This co-occurrence relation is observed between the Japanese negative morpheme na and the verb ar, as seen in (11). (11) a. tabe-na -kat-ta eat -NEG-AR-PAST 'did not eat' b. samuku-na -kat-ta cold -NEG-AR-PAST 'was not cold' The construction in (11) can be simplified as in (12). (12) a. A (Verb) Neg—-A' (ar) b. B (Adj) Neg—-A' (ar) In (12), the verb ar co-occur with the intervening element Neg as opposed to the case where the focus marker wa does not affect the choice of DO, as shown in (10). The fact that DO co-occurs with the closest lexical head X°, namely Neg, is left unexplained under the DO-Support Hypothesis. This is 19 because the selectional relation is determined before syntactic operations such as DO-support occur. This co-occurrence relation is not restricted to Neg and Focus constructions in Japanese. Light verbs in Japanese also show co occurrence relations with Verbal Nouns or Adjectival Nouns. If we take the position that DO is base-generated, a unified account can be achieved both for the focus constructions and the light verb constructions. We will discuss this issue in Chapter 4. The same is observed when the other delimiters like sae 'even' or mo 'also' replace the focus marker wa in Japanese, as seen in (13) and (14). The verb in the focused VP co-occurs with su, as seen in (13), the adjective co-occurs with ar, as seen in (14). Japanese (13) John-ga hon-o yom-i-sae/mo si-ta John-NOM book-ACC read-COMP-even/also SU-PAST 'John even read a book.' (14) ringo-ga akaku-sae/mo at-ta apple-NOM red- even / also AR-NPAST 'The apple was only red.' Korean (15) sakwa-ka ppalkah-ki- man/kkaci/to ha-ess-a apple-NOM red -KI-only / even / also HA-PAST-DEC 'The apple was only/even/also red.' In Korean, however, ha can be used for both the constructions shown in (13) and (14). Recall that ppalkah 'red' in Korean is classified as a stative verb, while aka 'red' is classified as an adjective in Japanese. The 20 generalization is that verbs in Korean and Japanese co-occur with ha and su, respectively and adjectives in Japanese co-occur with ar. We, following Hoji's suggestion (p.c), propose that this generalization can be captured by positing the abstract verbs dO and bE for both Japanese and Korean. These abstract verbs have feature bundles but not phonetic matrixes. The abstract verb dO has a different feature bundle from the one of bE. The verb dO is realized phonetically as su in Japanese and ha in Korean, while bE is phonetically realized as ar. The abstract verbs and their phonetic realization can be summarized as in (16). (16) Abstract verbs and their phonetic realization Abstract Verb dO bE co-occurring lexical item verb adjective Phonetic Realization Japanese su ar Korean ha Under this analysis, we can provide a unified analysis for both su and ha one the one hand and ar on the other. The abstract verbs dO and bE co- occur with verbs and adjectives, respectively. That the Korean counterpart of ar is missing is then accounted for since Korean does not have adjectival predicates. 6 6 The analysis concerning the co-occurrence between the main verb and dO or bE will be provided in Chapter 4. 21 In summary, this section has demonstrated the co-occurrence relation between DO and embedded predicates. The fact that ar co-occurs with adjectives or the negative lexical item na confirms that na belongs to the category of adjectives. 2.2.1.1.3. Evidence from the Nominalizer sa Nominalization by sa also indicates that Neg in Japanese belongs to the category of Adjectives. Since the nominalizing suffix sa exclusively attaches to adjective stems to turn them into abstract nouns (Kageyama 1982:217), in the same way that -ness in English does, the possibility of attaching sa to the negative na provides evidence supporting the claim that the negative in Japanese belongs to the category of adjectives. The paradigm in (17) and (18) shows this point. The following data are from H. Aoyagi (p.c.): 7 (AN: Adjectival Noun, VN: Verbal Noun) (17) a. samu (A stem) + sa-> samu-sa'coldness' b. kasiko (A stem) + sa -> kasiko-sa 'wiseness' c. taka (A stem) + sa -> taka-sa 'height' (18) a. tabe (V stem) + sa-> :|: tabe-sa eat b. Tokyo-kara (PP) + sa -> *Tokyo-kara-sa Tokyo-from 7 For some native speakers like H. Aoyagi (p.c.) the nominalized sizuka-a, namely, sizuka-sa needs two question marks, while some others like K. Ohta (p.c.) accept the same form without any question marks. We do not have an explanation for this dialectal variation. 22 c. gakusei (N) + sa -> *gakusei-sa student d. benkyoo (VN) + sa -> *benkyoo-sa study e. sizuka (AN) + sa -> sizuka-sa quiet As seen above, adjectives are the only category that can be nominalized by sa. Some compounds including the negative na may also receive sa. (19) a. nasake-na-i 'hopeless' -> nasake-na-sa 'hopelessness' hope b. azike-na-i 'tasteless' -> asike-na-sa 'tastelessness' taste c. tayori-na-i 'undependable' -> tayori-na-sa 'undependability' depend Furthermore, sa can be attached to the negative morpheme following the adjective, as seen in (20). 8 (20) Hanako-wa [kyuusyoku-ga/?-o Hanako-TOP school lunch-NOM/ACC tabe-ta-ku-na]-sa-ni gakkoo-o yasun-da eat-want-CON-Neg-COMP-because school-ACC be absent from 'Hanako was absent from school because she didn't want to eat school lunch.' No matter what analysis can be provided for the structure in (20), the crucial point is that the nominalizer sa, which is exclusively used for adjectives, can also be attached to Neg. ^h e verbal connective ku is glossed as CON. I thank H. Aoyagi for the data. 23 2.2.1.1.4. Indirect support from Prohibition Negative mal in Korean. The existence of a negative lexical item in the verbal category in Korean further supports our claim that the Japanese Neg na is an adjective, not just a functional category. In Korean, mal is employed for indicating 'prohibition'. One of the contexts, where it can appear is in the imperative, while other negative morphemes such as ani cannot be used in imperatives. Unlike Neg ani in long-form negation, mal is not followed by DO. The negative is directly followed by the imperative mood marker, as seen in (21) (21) a. ka-la go-IMP 'go' b. ka-ci mal-la go-COMP not to do-IMP 'Don't go.' (22) a. ka-ta go-DEC '(He) goes.' b. ka-ci ani ha-ta go-COMP NEG do-DEC '(He) does not go' The fact that mal is not followed by ha 'do' as in long-form negation (22) leads us to suspect that mal, unlike ani, is a verb which can directly bear inflectional morphemes. Given that ha appears when it has to support a stranded morpheme, one may claim that the absence of tense in the imperative, in general, is 24 responsible for the absence of ha in (21). Zanuttini (1991) claims that imperative constructions in English lack Tense. The sentences in (23) and (24), containing Tense, however, show a clear difference between ani and mal in that mal can directly bear verbal inflections. The negative mal can also license an NPI as ani does. (23) amwuto ka-ci ani ha-ess-eya. toy-ta anyone go-COMP NEG do-PAST-should-DEC 'Nobody should have gone.' (24) amwuto ka-ci mal-ess-eya.toy-ta anyone go-COMP NEG-PAST-should-DEC 'Nobody should have gone.' Furthermore, the sentence in (24) is a paraphrase of (23) using Neg mal. The minimal difference between (23) and (24) is the presence or absence of ha in (23) and (24), respectively. In other words, while the negative morpheme ani in (23) does not directly bear a inflectional morpheme, mal in (24) does. Also, the negative word mal can be involved in multiple prohibition, as seen in (25) (25) ka-ci mal-ci-(nun) mal-ela go-COMP not-COMP-FOC not-IMP You shouldn't not go.' Given these paradigms and the analysis that ci is a VP-complementizer, as claimed in Kang (1988), mal is a main verb which is specified for [+V, -N], having the following structure, (irrelevant nodes are omitted). 9 9 Following Han (1987), we assume that the imperative marker la is generated under C. 25 (26) CP Spec v' FP V/Neg VP ^ mal / ^ ci Spec V' ka The presence of Neg in the verbal category indirectly supports the claim that the Japanese Neg can belong to the another lexical category Adjective. Since Korean counterparts of Japanese adjectives (including Negatives) are stative verbs in Korean, the phrase headed by mal in Korean is analogous to the Japanese negative morpheme na in terms of its category. Therefore, categorizing the Japanese Neg as an Adjective is not ad hoc if we consider the Korean imperative negative morpheme mal, which is argued to be a verb [+V, -N]. This provides indirect support for the claim that Japanese NegP can be a lexical projection. 2.2.1.1.5. The Japanese Neg structure and Multiple NegP/AdjP In the preceding section, we, along the line with Kitagawa (1986), have provided the descriptions which show that the Japanese negation behaves 26 like an adjective in terms of inflections. Based on this description, we will treat Japanese negations as adjectives, which have their own projections. That is, Neg heads the AdjP, and DO heads an upper VP. Since Neg in Japanese belongs to the lexical category of Adjective, we may expect that multiple NegPs might be possible. This is borne out as seen in (27). (27) John-wa ringo-o tabe-na-ku na-ku-wa na-i John-TOP apple-ACC eat-Neg-CON Neg-CON-FOC Neg-PRES 'It is not the case that it is not true that John does not eat apples.' = 'John does not eat apples.' Since the multiple NegPs can now be taken as iterated AP, we need not stipulate the stacking of NegP in a clause in Japanese. Therefore, the multiple occurrence of Neg in Japanese can be expressed as in (28). (irreverent nodes are omitted) (28) VP T AP/NegP V Spec^AdiTNeg' ar AP/NegP Adj/Neg Spec Adj'/Neg' The discussion so far indicates that the Japanese Neg na is an adjective. Given this analysis, the hypothesis that ar is inserted cannot be maintained. This is because under the ar-support analysis, more than one 27 INFL has to be stipulated for -ar to be generated under it. If this is correct, the appearance of ar can not be attributed to the blocking effect of Adjectives or Negation. The structure in (28) with DO above the AP/NegP can be further justified if we consider the parallelism between the ar- construction discussed above on the one hand and the su-construction in Japanese and the Aa-constructions in Korean on the other, which will be discussed in the following sections. 2.2.1.2. Ha in Korean Negative Sentences The Korean counterparts of English adjectives are called (stative) verbs in that they, among others, undergo the same conjugation as other regular verbs in Korean. In this sense, Korean Neg ani is clearly not a verb in that it cannot be conjugated on its own. This can be made into a simple generalization that ani is followed by a verb, if ha is considered as a verb, as suggested in Song (1988). If we consider long-form negation, the appearance of DO seems to be triggered by the blocking effect of Neg. We, following Song (1988), take a different approach to the appearance of ha in this context. That is, we argue that the affirmative counterparts of the Korean negative constructions are freely generated. This means that ha is not inserted to support the stranded tense morpheme. Rather, the negative na is inserted to generate the negative sentence. Before we examine ha- constructions in negative contexts, some remarks are in order concerning the category of the negative ani. 28 2.2.1.3. Adverbial Status of Negative ani It is well known that Korean has two types of negation (Lee 1970; Oh 1971; Han 1987; Song 1988; Suh 1990; Yoon 1990; Lee 1992; Jung 1992, among many others): Short-form negation and Long-form negation. In the Long-form negation the negative morpheme follows VP, having ci between the VP and the negation morpheme ani. Long-form negation has the so-called /^a-support. The following two sentences in (29) illustrate Short-form negation and Long-form negation, respectively. 10 (29) a. John-i ani ka-ess-ta (Short-form) John-NOM not go-PAST-DEC 'John did not go' b. John-i ka-ci ani ha-ess-ta (Long-form) John-NOM go-CI not do-PAST-DEC 'John did not go.' •••"The negation of the stative verbs (adjectival verbs) in Korean show the same pattern. That is, only long-form negation co-occurs with ha. Korean (i) a. sakwa-ka ani ppalkah-ess-ta (short-form negation) apple-NOM NEG red-PAST-DEC 'The apple was not red.' b. sakwa-ka ppalkah-ci ani ha-ess-ta (long-form negation) apple-NOM red -CI NEG do-PAST-DEC 'Trie apple was not red. 29 Long-form negation differs from Short-form negation in that it has ci and ha. As the gloss indicates, these two types of negation do not differ in meaning 11 The Korean negative morpheme comes only before a verb, not in any other places. The ungrammaticality of the sentences in (30b), (31b), (31c), and (32b) shows this point. (30) a. John-i kay-lul John-NOM dog-ACC b *John-i ani [yp kay-lul John-NOM not dog-ACC 'John did not kick the dog.' ani cha-ess-ta not kick-PAST-DEC cha]-ess-ta kick-PAST-DEC (31) a. motun-salam-i everybody-NOM 'Everybody didn't come.' b. *ani motunsalam-i not-everybody-NOM 'Not everybody came.' c. :1: ani salam-i not man-NOM 'Lit. Not man came.' (32) a. John-i ppalli John-NOM fast 'John didn't run fast.' ani o-ess-ta (Neg + Verb) not-come-PAST-DEC o-ess-ta (Neg + QP) come-PAST-DEC o-ess-ta (Neg + Noun) come-PAST-DEC ani ttwui-ess-ta (Neg + Verb) not run-PAST-DEC b. :|: John-i ani ppalli ttwui-ess-ta (Neg + Adverb) John-NOM not fast run-PAST-DEC 'John didn't run fast.' 1 ^It does not necessarily mean that the two negations have the same scope over the Quantifier Phrase in the subject position or object position. 30 Due to these distributional facts, among others, it is generally agreed in the literature that Short-form negation ani is generated, left-adjoined to V° j as schematically illustrated in (33) (Han 1987; Suh 1990): (33) [ T p Spec [ T ' LVP Spec W tv ani V]]]] Tense ] The following test, however, shows that the base-generated adjunction of the negative morpheme to V° is not a correct hypothesis. In what follows we will take a plural maker tul in Korean as a tool to test whether the Neg ani is base-generated adjoined to V°. Before taking up this test, some remarks are in order on the property of the plural morpheme tul. The plural marker tul can appear attached to such non-canonical phrases as adverbs and postpositions, not only to NPs. The plural marker tul attached to these non-canonical phrases is claimed not to pluralize them (Choe 1929; Song 1988). Choe (1929) and Song (1988) claim that the plural marker tul on these non-canonical phrases indicates the plurality of the subject containing it. 12 In other words, the plural marker tul on the non-canonical phrases is subject-oriented, just like secondary predicates. Another property of tul is that it is clause-bound. Let us consider examples where the plural marker tul is attached to an adverb. l^The plural marker tul after the adverb ilccik 'early' is optional. This plural marker, however, cannot appear when the subject NP does not bear the plural marker. The following sentence shows this point. (i) haksayng-i ilccik-(*tul) o-ess-ta student-NOM early-PL come-PAST-DEC 'The student(s) came early.' 31 (34) a. John-i ilccik o-ess-ta John-NOM early come-PAST-DEC 'John came early.' b. *John-i ilccik-tul o-ess-ta John-NOM early-PL come-PAST-DEC 'John came early.' c. haksayng-tul-i ilccik-(tul) o-ess-ta student-PL-NOM early-PL come-PAST-DEC 'The students came early.' (34a) minimally differs from the ungrammatical sentence in (34b) in that the adverbial ilccik 'early' bears the plural marker tul. If the subject is plural, the AP is allowed to bear the plural marker, as exhibited in (34c). This plural marker cannot interrupt a lexical or incorporated unit (X^-mo vement) . (35) chayk-( :i: tul)-kapang book-(*tul)-bag 'Backpack' (36) haksayng-tul-i ilccik-(tul) o-(*tul)-ess-ta student-PL-NOM early-PL come-PL-PAST-DEC 'The students came early.' If a plural marker cannot occur inside incorporated items, the interruptability of this plural marker can be employed to test whether two morphemes are one unit or not. Given this tool, let us consider the negative construction. 13 13 Nam-Kil Kim (p.c.) readily accepts this sentence. The tendency is that older generations more readily accept this type of sentence. This might be interpreted as a trend where the negative adverb is being changed to a bound morpheme, which has to attach to a host morpheme. 32 (37) ?ani-tul tuleo-1-lay? not-PL come-FUT-Q Won't you (PL) come in?.' In (37), the plural marker can come between the negative ani and the following verb. Given this observation, I claim that the negative ani is not generated in the V° adjunction potion. I claim rather that this morpheme is an adverbial which precedes a verb. 14 In a sense, this negative morpheme ani is similar to never in that it immediately precedes a verb (Pollock 1989). (38) a. John never sold the book. b. *Never John sold the book. c. *John sold the book never. Based on the above observation, I claim that Neg ani is generated as a complement of the verb as in a Larsonian structure, as suggested to me by A. Li (p.c.). 15 The structure can drawn as in (39). (Irrelevant nodes are omitted) 14 The idea that the negative morpheme ani is an adverbial is not new. The Korean traditional grammar (Choe 1929) takes ani as an adverbial and historically ani has been used as an adverbial very productively. In old Korean, the accusative case marker ul can be attached to ani, as pointed out to me by M.K. Park (p.c). It is also noted in Park-Han (1981) that in early stage of first language acquisition, ani behaves as an adverb. The following sentences are from Park-Han (1981: 106). (i) an -ppesu tha NEG bus get on 'Let's not take a bus.' implying (Let's take a cab) In (i), the short-form of ani, namely, an immediately precedes the object NPs. 33 (39) VP Spec V' ani V 'not' If the negative morpheme ani is an adverb, not a head of NegP (cf. Cho 1993a), ani does not constitute a blocking element, yielding no DO- support. This is because the appearance of do-support can result from a blocking effect by a negative morpheme under the Minimality condition, to support the stranded tense morpheme (Pollock 1989; Benmamoun 1991; Chomsky 1988). 2.2.1.3.1. Apparent Problems with the Adverbial Approach The approach that the negative ani in Korean is an adverb, however, is not without problems. The negative morpheme ani seems to behave like a prefix to the following verb, not as an adverb, in terms of topicalization. The problem can be found in reduplicated focus phrases. As No (1988) has observed, the negative morpheme behaves like a passive morpheme, which is bound to a verb. 15 See Baker (1991) and (Baltin 1993), who claim that not in English is a lexical category adverb. Baker (1991) suggests that not is a preverbal adverb which can attach to a V as a left sister, yielding a larger V. 34 (40) a. John-i cap -hi -ki-nun cap-hi-ess-ta John-NOM catch-PASS -COMP-TOP catch-PASS-PAST-DEC b. *John-i cap-0-ki-nun cap-hi-ess-ta John-NOM catch-0-COMP-TOP catch-PASS-PAST-DEC c. :1: John-i cap -hi -ki-nun cap-0-ess-ta John-NOM catch-PASS -COMP-TOP catch-0-PAST-DEC 'John indeed got caught.' In (40), the passive morpheme cannot be deleted in the repeated phrase. In contrast, adverbs can be freely dropped in the repeated phrase. (41) John-i (cal) tali~ki-nun (cal) tali-ess-ta John-NOM well run-COMP-TOP well run-PAST-DEC 'John at least ran well.' The above paradigm shows that the passive bound morpheme hi cannot be dropped, while the adverbial can be. If all adverbs can freely be dropped in the reduplicated phrase and if the negative morpheme ani is an adverbial, we may expect that the negative morpheme ani need not be reduplicated in the same constructions. This is not borne out. (42) a. John-i ani o-ki-nun ani o-ess-ta John-NOM not come-COMP-FOC not come-PAST-DEC 'lit. Did not come, John did not come.' b. :|: John-i ani o-ki-nun o-ess-ta John-NOM not come-COMP-FOC come-PAST-DEC 'lit. Did not come, John did come.' This test indicates that the negative morpheme ani does not behave like a regular adverbial. We face contradictory results: the negative morpheme ani can or cannot be an adverb. 35 By passing we note that topicalization, however, is not a reliable test for the adverbialhood of ani. Topicalization (43) a John-i cal tali-ess-ta John-NOM well run-PAST-DEC 'John ran well.' b. John-i cal-un tali-ess-ta John-NOM well-TOP run-PAST-DEC 'John at least ran well.' (44) a John-i ani tali-ess-ta John-NOM not run-PAST-DEC 'John did not run.' b. *John-i ani-nun tali-ess-ta John-NOM not-TOP run-PAST-DEC 'John did not run.' Given that the topic marker un/nun can attach to NP, AdvP, or PP, the grammaticality of (43) is predicted. If ani is the head of an AdvP, (44b) might be incorrectly predicted to be grammatical. The ungrammaticality of (44b), however, does not constitute evidence against the adverbialhood of ani. H.Hoji pointed out to me (p.c.) that the ungrammaticality of (44b) might be related to the fact that in English the negative adverb never cannot be modified by even or also. This must be the case if we consider the adverb cenhye 'never' which is clearly an adverb and has a negative connotation. It cannot be followed by an topic marker nun, as exhibited in (45). 36 (45) a. Mary-ka cenhye ani yepputa Mary-NOM never not pretty 'Mary is never be pretty.' b. Mary-ka cenhye-nun ani yepputa Mary-NOM never-TOP not pretty 'Mary is never be pretty.' If an adverb with a counterfactual meaning cannot be followed by nun, the impossibility of topicalization of ani cannot be a valid test to decide whether ani is an adverb or not. In this and the preceding section, we have seen contradictory results: the negative morpheme ani behaves like an adverb in plural marker test but behaves like a prefix in other tests. In the following section, we try to solve the contradictory situation between the Adjectivehood and prefixhood of ani. 2.2.1.3.2. Toward the solution. The problems shown above can be solved if we assume that the adverbial negative ani undergoes head movement to the V°, as suggested in Cho (1993a). It is suggested in Cho (1993a) that the adverb ani undergoes head-movement to V°. This analysis explains why the negative cannot be immediately followed by the focus marker nun and cannot be dropped in the reduplicated phrase. 16 Now the question is how to account 16 If the scope phenomenon is strictly based on strict c-command relation (Reinhart 1976), the head-movement of ani to V° can account for the wide scope of Neg over the QP-object (see Suh 1989 for more details, who treats for QP and Neg interaction in the Aoun and Li's model (Aoun and Li 1993). 37 for the appearance of the plural marker tul. We take the position that the Neg ani can be taken as both X° and X ma x at the same time, following the spirit of Bare Phrase Structure (Chomsky 1994). 17 We assume with Roberts (1991) that the incorporation host morphologically subcategorizes for the incorporee when incorporation results in a visible amalgam of the two heads. We further assume, following Roberts (1991), that a structural slot is created for the incorporee at D-structure as a function of the lexical properties of the incorporation host. 18 This amounts to saying that the slot for the incorporee is base- generated within a incorporation host. This slot triggers the substitution of incorporee during the derivation. Under this system, as I. Roberts (1991) suggested, 1° in a language like French has the subcategorization frame [+V° ]. Roberts (1991) assumes that in this incorporation, the head of the complex formed by incorporation remains X°, and the incorporation head is proposed as X -1 . I suggest that the same can be applied to the negative morpheme ani. The adverbial generated as a complement of the V O incorporates to the V 0 . 19 Under this hypothesis, we have the following structure. 17 I thank M.K. Park (p.c) for drawing my attention to this issue. Chomsky (1994) suggests that a clitic can be X° and X ma x at the same time. 18 A similar idea is developed in Svenonius (to appear). 19 One problem with this analysis is that the movement of ani to V causes the usual lowering problem. 38 (46) For this analysis to work we need at least one of the following assumptions: (1) the plural marker can attach to X°; (2) if the plural marker tul attaches only to the XP, then we have to allow the movement in (46) is XP- movement. Given that the plural marker tul cannot be attached to a verb stem, ani cannot be X°. Note that while verb stems cannot be X° categories, adverbs can be both X° and X max . 20 Now returning to the impossibility of attaching focus marks, we attribute this fact to the obligatory head movement of the negative morpheme to V°. This analysis is again analogous to the clitic movement in that the clitic undergoes obligatory head-movement to a verb. 21 20 It is hard to test whether the plural marker attaches to XP or X°. This is because Korean is a head final language and the plural marker follows the head. I assume that the plural marker tul is attached to XP. The indirect evidence comes from the impossibility of attaching the plural marker to the verb stem. I assume that the amalgamation of a verb stem with verbal inflections has resulted from verb movement to inflection. If this is the case and if the plural marker can be attached to the verb stem, we may expect that the plural marker can appear after the verb stem. The fact that the plural marker cannot be attached to the verb stem indicates that the plural marker attaches to XP, not X°. 21 As K.-M. Sung pointed out to me (p.c), the clitic moves to the strong INFL. 39 2.2.2. DO in affirmative sentences As seen in the previous section, long-form negation is followed by the verb ha, as seen in (47). (47) John-i wuli cip-ey o-ci-lul (long-form) John-NOM our house-to come-COMP-ACC ani ha-ess-ta not do-PAST-DEC 'John did not come to my house.' One strong argument for the DO-Support Hypothesis is based on the description of the verb ha occurring in long-form negation and focus constructions, not in affirmative sentences. The following sentence is the affirmative counterpart of (47). (48) ??John-i wuli cip-ey o-ki-lul ha-ess-ta John-NOM our house-to come-COMP-ACC do-PAST-DEC 'John came to my house.' If the sentence in (48) is unacceptable, one can claim that ha is inserted to support the stranded tense morpheme, under the analysis that the Neg or focus morpheme blocks verb movement to Tense (Cho 1993a). Consequently, the DO-Support Hypothesis can be strongly supported. The consequence is that the long-form negation construction can be taken as mono-clausal with DO inserted under INFL. By contrast, if this sentence is acceptable, the long-form negative construction can be freely generated by insertion of the Neg morpheme in sentence (48), with the assumption that ci is a variant of ki, as Song (1988) suggested. Furthermore, if the sentence in (48) is grammatical, this 40 sentence has two main verbs, indicating that this sentence can be a bi- clausal construction. Since the grammaticality status of (48) can provide a clue to the negation constructions, this type of sentence has been at the core of the controversy for a long time (Oh 1971; Yang 1976; Song 1988). While some researchers such as Yang (1976) find sentences like (48) to be unacceptable and ungrammatical, some others such as Song (1988), Han (1987), and Lee (1992) find them grammatical, marking them with one question mark. If the sentence in (48) is ungrammatical, the claim that the long-form negation is derived from (48) is not just weakened, but refuted. 22 A sentence like (48) can be saved in rhetorical questions (Cho 1992b). Native speakers do not show a judgment variation for the sentence in (49). That is, the sentences containing ha without negation are taken as fully acceptable. 23 (49) John-i o-ki-lul ha-ess-ni, John-NOM come-COMP-ACC do-PAST-Q cenhwaha-ki-lul ha-ess-ni? phone-COMP-ACC do-PAST-Q? 'Lit. Did John come or call (me)?' 'John neither came nor called (me).' 22 Kuroda (1992) posits that a grammatical sentence can be derived from an ungrammatical sentence. If this approach is taken, then it is not necessary for the affirmative counterpart of the long-form negation to be grammatical. 23 We will take up the issue as to the degree of acceptability between (48) and (49) below. 41 In (49), the two VPs are contrastively emphasized with negative connotation. This sentence cannot be answered by yes /no, 'he came', or 'he called me' in any context. On the contrary, if ha 'do' is not employed in the paired questions, this sentence is interpreted as interrogative as illustrated in (50) (50) John-i o-ess-ni? cenhwaha-ess-ni? John-NOM come-PAST-Q phone-PAST-Q? 'Did John come or call (me)?' This sentence can be answered by 'he came' or 'he called.' Two alternatives, o 'come' and cenhwaha 'call', are given as a neutral choice. Since the only possible interpretation for (49) has negative presupposition, namely, 'John neither came nor called (me)', a possible approach can be that the invisible Neg triggers /m-insertion. If negative presuppositions triggered /ia-insertion, however, we would incorrectly predict that all kinds of counterfactuals trigger /m-insertion. Furthermore, the ungrammaticality of the sentence in (51a) below, in contrast to (51b) below, indicates that the postulation of an invisible Neg would be problematic. (51) a. *amwuto o-ki-lul ha-ess-ni, anybody come-COMP-ACC do-PAST-Q cenhwaha-ki-lul ha-ess-ni? phone-COMP-ACC do-PAST-Q? 'Lit. Did anybody come or call (me)?' 'Noone came or called (me).' b. amwuto o-ci-lul ani ha-ess-ta anybody come-COMP-ACC not do-PAST-Q 'Noone came.' 42 Given these problems, we rule out the possibility that the invisible Neg is involved in the appearance of ha in (49). Furthermore, there are cases where the sentences turn out to be acceptable, even though the affirmative sentence does not have a negative connotation. If adverbial phrases come before ha, as seen in (52) and (53), the /m-constructions without Neg are quite acceptable. (52) John-i wuli cip-ey o-M-lul John-NOM our house-to come-COMP-ACC cacwu ha-ess-ta frequently do-PAST-DEC 'John came to my house frequently.' (53) John-i chayk-ul ilk-ki-lul John-NOM our house-to come-COMP-ACC mayil ha-ess-ta everyday do-PAST-DEC 'John read a book everyday.' One possible account for the difference in acceptability between (53) and (48) could be that adverbs block verb movement and therefore, ha is inserted to rescue the stranded Tense morpheme. We rule out this possibility since the Japanese counterpart of (53) is ungrammatical. If adverbs can block verb movement in Korean, the same must be true in Japanese, which has the same focus constructions as in Korean. Given that the adverb does not block the head-movement of the verb and that the Aa-constructions in affirmatives is fully acceptable, we can safely say that the verb ha is base-generated. 43 Thus far we have seen that the affirmative counterpart of the long- form negation can be possible, which supports the claim that ha is a base- generated, as in Song (1988). We will later suggest that the sentences treated in this section involve emphasis and will provide more detailed explanation in the last part of this chapter. 2.2.3. DO in Focus Constructions The focused VP can also co-occur with DO both in Japanese and Korean. These focus constructions are possible when delimiters follow the verb (Kuroda 1965 for Japanese; Yang 1972 for Korean). These delimiters include mace 'even; including', mata 'every', kkaci 'even', man 'only', pakke 'nothing but', nun 'only concerned', to 'also; too, even' ya 'at least; of course', na 'rather', lato 'even as the last recourse' in Korean. The VP-focus constructions in Japanese and Korean are illustrated in (54a) and (55a). The sentences in (54b) and (55b) are regular declarative counterparts of the VP-focused constructions in (54a) and (55a), respectively. In the VP-focus constructions, the focus marker is attached to the VP, which is focused. Japanese (54) a. John-ga hon-o yom-i-wa/mo si-ta John-NOM book-ACC read-COMP-FOC/also do-PAST 'John at least read a book.' 'John also/even read a book.' b. John-ga hon-o yon-ta John-NOM book-ACC read-PAST 'John read a book.' 44 Korean (55) a. John-i chayk-ul ilk-ki-nun/to ha-ess-ta John-NOM book-ACC read-COMP-FOC/also do-PAST-DEC 'John at lease read a book.' 'John also/even read a book.' b. John-i chayk-ul ilk-ess-ta John-NOM book-ACC read-PAST-DEC 'John read a book.' Only the focused examples have DO. This state of affairs led Kuroda (1965) to propose a DO-Support Hypothesis for focus constructions, under the assumption that the Focus marker blocks verb movement, triggering su insertion to support the stranded tense morpheme. However, if we take a sentences like (52), which is repeated here as (56), the appearance of ha should be taken as base-generated, not inserted. (56) John-i wuli cip-ey o-ki-lul John-NOM our house-to come-COMP-ACC cacwu ha-ess-ta frequently do-PAST-DEC 'John came to my house frequently.' Since the sentence in (56) can be generated without the focus marker, the DO-Support Hypothesis is weakened. This construction not only argues against the DO-Support Hypothesis, but also suggests that ha in Focus constructions can be a base-generated main verb. As we will see in the next section, the absence of VP-deletion in Japanese and Korean (J&K) further confirms that DO behaves as a main verb. 45 2.2.4. VP-ellipsis This section investigates what might appear to be VP-ellipsis constructions in Korean and suggests that these constructions support our claim that DO in Japanese and Korean are base-generated main verbs. We first show that J&K do not have VP-ellipsis in the sense of English. One instance of VP-ellipsis in English is as given in (57). (57) Did John buy a book? Yes, he did [VP e]. No, he didn't [VP e] As mentioned in the introduction, the answer to a 'yes-no question' in Korean does not seem to involve VP ellipsis. (58) a. Mary-ka chayk-ul sa-ess-ni? Mary-NOM book-ACC buy-PAST-Q 'Did Mary buy a book?' b. ye, *ha-ess-eyo / sa-ess-eyo yes, do-PAST-DEC / buy-PAST-DEC Yes, (she) did.' / bought. As seen in (58b), in the answer to the yes-no question, the answer cannot be ha. Rather, the thematic verb should appear. Exactly the same fact is observed in Japanese. If VP ellipsis were possible at all in Korean, ha would be inserted to support the affixal inflectional elements under the DO- Support Hypothesis. This is not the case as seen in (58). 46 VP-focus constructions further support our claim the ha is a main verb, not the dummy support verb. Only when ha is used in the VP-focus yes-no question, ha can appear in the yes-no answer. (59) a. Mary-ka chayk-ul sa-ki-nun/kkaci ha-ess-ni? Mary-NOM book-ACC buy-COMP-FOC/even do-PAST-Q 'Is it the case that Mary bought a book?' b. ye, sa-ess-eyo Yes, buy-PAST-DEC 'Yes, (she) bought.' c. ye, (sa-ki-nun/kkaci) ha-ess-eyo yes, (buy-COMP-FOC/even) do-PAST-DEC "Yes, (she) did.' The sentence in (59a) is the VP-focus version of (58a). The question itself contains the verb ha. In this case, answering with ha is acceptable as seen in (59c). The generalization is that the presence or absence of ha in yes/no answers depends on the presence or absence of ha in the question. This paradigm can be explained only when ha in the yes-no answer is base- generated, not inserted. The same can be observed in long-form negation. (60) a. Mary-ka chayk-ul sa-ci-nun ani ha-ess-ni? Mary-NOM book-ACC buy-FOC NEG do-PAST-Q 'Is it the case that Mary did not buy a book?' b. ye, ani sa-ess-eyo Yes, NEG buy-PAST-DEC 'No, (she) did not buy (it).' c. ye, sa-ci-nun ani ha-ess-eyo yes, buy-FOC NEG do-PAST-DEC 'No, (she) did not.' 47 d ?/??ye, ani ha-ess-eyo yes, NEG do-PAST-DEC 'No, (she) did not.' The fact that ha appears in a yes-no answer as in (59c) only when the question contains ha as in (59a) suggests that Korean does not have genuine VP-ellipsis. If so, what appears to be VP-ellipsis in this language is the dropping of the VP-complement of the base-generated ha. 2.2.5. Multiple Occurrence of DO This section examines the multiple occurrence of ha in Korean and claims that this multiple occurrence can be accounted for in a base-generation approach to DO. As mentioned above, Korean has two types of negation: short-form and long-form negation. Korean also has multiple negation: more than one long-form negation or one short form negation plus one long-form negation or more. The following are examples of these combinations. 24 24The Korean double negation differs from the so-called negative concord in various languages. By negative concord, we mean the co-occurrence of two or more negative elements in a sentence, which do not negate each other. The following sentences exemplify this case. (i) I did not buy nothing (ii) U n ae nent vera (Piedmontese: Cairo Montenotte; Parry (1985)) CL NEG is NEG true 'It isn't true.' (ii) Jean ne mange pas de viande John NEG eats NEG of meat 'John doesn't eat meat.' 48 (61) a. John-i ani-ka-ess-ta (Short-form) John-NOM not-go-PAST-DEC 'John did not go' b. John-i ka-ci ani-ha-ess-ta (Long-form) John-NOM go-COMP not-do-PAST-DEC 'John did not go.' c. John-i ka-ci ani ha-ci-nun (Long+ Long) John-NOM go-COMP not do-COMP-FOC ani ha-ess-ta not do-PAST-DEC 'It is not the case that John did not go.' b. John-i ani ka-ci ani ha-ess-ta (Short + Long) John-NOM not go-COMP not do-PAST-DEC 'John did not go.' Sentences like (61c) is problematic to the DO-Support Hypothesis, as pointed out by many linguists (Suh 1990; Park 1992, among many others). If we take the DO-Support Hypothesis which states verb movement is blocked and therefore DO is inserted under INFL to support the tense morpheme, we have to say that the clause containing a single predicate has more than one tense in multiple negation constructions as in (61c) . Given this problem, the Base-generation Hypothesis is more attractive than the DO-Support Hypothesis in that we can freely generate more than one DO, as long as it takes a VP as a complement. Given this advantage, we may posit that the apparent DO-Support Construction in Korean might not be a real DO-Support Construction in the sense of English. It is rather similar to light verb constructions in that the light verb takes a VNP (Verbal Noun Phase) as a complement. 49 If we take the view that Neg is adjoined to V° (Han 1987; Suh 1990) or is a complement of V°, and that DO takes a VP complement, we can provide a unified account for both long- and short-form negations. That is, the Korean Neg is adjoined to any verb including DO and DO can be generated iteratively, which can generate all the negative sentences in (61). This approach denies the possibility that short-form and long-form negations are derived from a single basic D-structure (see Yang 1976 for the derivation from a single basic form). The multiple occurrence of negation and the combination of short-form and long-form negation such as the sentences in (61c) and (61d) are left unexplained in such an approach. The current proposal that DO in Korean is a main verb can be extended to Japanese. If the Japanese light verb su is a light verb which can iteratively appear in a sentence, multiple DO constructions can receive a natural explanation. This is borne out as the following sentence from H. Hoji (1991a, his 33b) exhibits. (Wun indicates Focus Marker). (62) naguri kakari-made gi-koso gi-wa attack by hitting-WUN DO-WUN DO-WUN si-na-katta (-ga) DO-NEG-INFL(although) 'Very Roughly: (although) (I) did not go as far as hitting (him),...' The Base-generation Hypothesis which claims that DO both in Japanese and Korean is a main verb can provide a straightforward answer to the iterative occurrence of DO in Japanese and Korean. 50 2.2.6. VP-fronting As noted in Akmajian, Steele and Wasow (1979), the process known as VP- fronting in English preposes the verb and its objects, leaving behind the auxiliary verbs. (63) a. Bill thought Mary had fixed his car, and [fixed his car] she had. b. Bill thought his car was being fixed, and [being fixed] it was. This syntactic movement can be supported by the island effects of VP, which have been reported since Ross (1967). The following examples are from Roberts (1990): (64) John said he'd win the race and... a. win the race I think that he did. b. ?? win the race I wonder whether he did. c. ?/* win the race that he did amazed me. d. ?/* win the race he went to London after he did. In (64), the phrase win the race is preposed out of a wh-island, a complex NP, or an adjunct. The deviance of the sentences in (64b) and (64d) is due to subjacency violations (Roberts 1990; Rizzi 1990), indicating that VP- fronting involves syntactic movement. What appears to be VP-fronting has been observed in Japanese (Hoji 1989). In the VP-fronting constructions, DO appears in the position where 51 \ \ \ the VP has mofted out. We, however, suggest that the presence of such constructions does not constitute evidence against our Base-generation Hypothesis. We take the position that VP-fronting is a scrambling process, as suggested in Hoji (198^). Therefore, our claim is that the fronted phase is a complement of DO, but not of INFL. Furthermore, given that the presence of VP-fronting is independent of DO-Support, the presence of VP- fronting in Japanese and Korean does not tell much about the property of DO in these languages. Both the Base-generation Hypothesis and the DO- Support Hypothesis can account for the appearance of DO in these languages. If this is so, the appearance of DO in this context does not undermine the Base-generation Hypothesis. Rather, it provides an adequate account for the appearance of DO in VP-fronting contexts. To see how the Base-generation Hypothesis can account for the VP-fronting, consider the following pair of constructions. (65) a. John-i [ku chayk-ul sa-ki-nun] ha-ess-ta John-NOM thebook-ACC buy-COMP-FOC do-PAST-DEC 'John at least bought the book.' b. [ku chayk-ul sa-ki-nun]i John-i tj ha-ess-ta thebook-FOC buy-COMP-FOC John-NOM do-PAST-DEC 'Buy the book, John did.' In this construction, the VP has fronted, leaving a trace behind. Our claim is that the fronted VP is moved out from the VP-focus constructions as in (65a). One piece of straightforward evidence supporting this movement analysis comes from the appearance of ki/ci in the fronted phrase in Korean. While the Japanese nominalizer i in the fronted phrase does not 52 tell much about the movement of VP, 25 ki/ci do. This is because ci is c- commanded/licensed by Neg in Korean before movement. In contrast, Japanese i does not have this restriction. If ci appears in the fronted VP it constitutes evidence supporting the argument that VP is fronted from the position where it is c-commanded by Neg. This is borne out as seen in (67). (66) a. John-i [o-ki/*ci-nun/to] ha-ess-ta (Aff) John-NOM come-KI/*CI-FOC/also do-PAST-DEC 'John at least/also came.' b. [o-ki/*ci-nun/to] \ John tj ha-ess-ta (Aff) come-KI/*CI-FOC/also John-NOM do-PAST-DEC ' At least/also came, John did.' (67) a. John-i [o-*ki/ci-nun/to] ani-ha-ess-ta (Neg) John-NOM come-*KI/CI-FOC/also not-do-PAST-DEC 'John at least/also did not come.' b. [o- :,: ki/ci-nun/to]j John-i tj ani-ha-ess-ta (Neg) come-*KI/NM-FOC/also John-NOM not-do-PAST-DEC 'At least/also not come, John did.' In the affirmative sentences, the underlying sentence in (66a) and the fronted version in (66b) have the same form ki. In the negative constructions in (67), both the underlying sentence and the sentence with the fronted VP contain the same form ci. Even though Japanese and Korean show VP-fronting and DO apparently shows up in these constructions, we can regard the VP-fronting as scrambling from the focus constructions. That is, the focus construction with DO is already generated before the process of VP-fronting in the proposed analysis. 25The term nominalizer for Korean ki/ci and Japanese i is a misnomer. We will claim, with Kang (1988) that these are not real nominalizers. 53 There is strong evidence supporting the Base-generation Hypothesis over the DO-support Hypothesis. We will examine the sentences containing double DO. (68) a. John-i chayk-ul ilk-ki-kkaci John-NOM book-ACC read-COMP-even ha-ki-nun ha-ess-ciman do-COMP-FOC do-PAST-even though v 'Even though John read the book...' b. [chayk-ul ilk-ki-kkaci ha-ki-nun]j book-ACC read-COMP-even do-COMP-FOC John-i tj ha-ess-ciman John-NOM do-PAST-even though 'Even though John read the book...' Under the DO-support hypothesis, the bracketed VP in (68) is fronted, leaving Tense behind. This tense is supported by /m-support. This approach, however, cannot account for the appearance of another DO in the bracketed part. If the DO-support Hypothesis is correct, DO in the bracket is also generated under INFL. If so, the DO-support hypothesis has to stipulate two INFLs in a simplex sentence, which is not allowed in current syntactic theory. Under the Base-generation hypothesis, the bracketed VP in (68a) is scrambled from the sentence in (68b) where two DO's are base-generated, strongly supporting the Base-generation Hypothesis over the DO-support Hypothesis. 54 2.3. Problems and their solutions In the proceeding sections, we have seen the arguments against the DO- Support Hypothesis. In this section we tackle the issue of the obligatoriness of a focus marker on the VP-complement in the focus constructions, discussed in Song (1988) and Hoji (1991a). The first issue concerns the obligatory presence of focus markers after the embedded clause in the focus constructions. 26 (69) John-i ka-ki-*(nun) ha-ess-ta (K) John-NOM go-COMP-FOC do-PAST-DEC 'John at least went.' (70) John-ga i-ki-*(wa) si-ta (J) John-NOM go-COMP-FOC do-PAST 'John at least went.' In (69) and (70), the focus markers nun and wa are obligatory. The question is why the focus marker in (69) is obligatory, given the arguments that ha is base-generated. This is not a trivial issue if we consider that Kuroda's (1965) DO-Support Hypothesis is based on the obligatoriness of the focus markers. In Kuroda (1965), if the focus marker is employed to focus a VP, DO-Support occurs; hence the appearance of ha and su in (69) and (70), respectively. Another issue concerns the optionality of the focus marker in long- form negation in Korean and this obligatoriness in Japanese. 26 The focus marker in (69) can be replaced by the accusative marker with some marginal acceptability, as discussed above. See the relevant sentence in (48) for this point. 55 (71) John-i ka-ci-(nun) ani ha-ess-ta (K) John-NOM go-COMP-FOC not do-PAST-DEC 'John did not go.' (72) John-ga i-ki-*(wa) si-na-katta (J) John-NOM go-COMP-FOC do-NEG-PAST 'John at least went.' Given that the presence of focus markers is obligatory in focus constructions, as seen in (69) and (70), the optionality of them in long-form negation in Korean, as seen in (71), mus t be explained. The obligatoriness of the focus marker in Japanese provides a further complication. In relation to the first question, we examine two possible approaches: Poser's (1992) Blocking Hypothesis and its alternative. We point out the problem with Poser's (1992) Blocking Hypothesis in accounting for the absence of the emphatic periphrastic forms in Japanese as opposed to the presence of them in Korean. We suggest that the obligatory presence of a focus marker in focus constructions can be attributed to the requirement that VPs need Case in the complement position of a verb. We propose that the VP- complementizer -ki in Korean and -i are not specified for [+N], and, therefore, the VP complements cannot be assigned Abstract Accusative Case. We, following Stowell (1981) and Chomsky (1981), assume that only [+N] categories can be assigned Abstract Accusative Case. We also assume that focus makers can assign inherent Case just as postpositions do. Under this analysis, the VP without Case in the complement position of V, can be saved by a focus marker, which assigns inherent case to it. 56 We, following Song (1988), Ahn (1990), and Hoji (1991a), suggest that -lul in Korean and -o in Japanese can function as emphatic markers. As pointed out by H. Hoji (p.c), these emphatic markers differ from focus markers, such as wa 'at least' and sae 'even', in that these emphatic markers do not have a full semantic value, and, therefore, the attachment of these two markers to VPs does not make focus constructions as natural as other regular focus markers do. In this sense, the emphatic markers can be taken as weak focus markers. We also suggest that -lul and -o are sensitive to the categories they attach to. While -lul can attach to non-[+N] categories, -o can attach only to [+N] categories. Therefore, the distribution of the emphatic marker -o is more restricted than that of-lul. This hypothesis is supported by the place where -lul can appear, but -o cannot. For one thing, -lul can be attached to verbal connectives in Korean, while -o cannot. Under this analysis the VP-focus with -lul is more acceptable than with -o, given that the VP is not [+N] category. In order to account for the optional vs. obligatory presence of focus markers in Korean and Japanese long-form negation, respectively, we, following Hoji (1991a), suggest that Korean and Japanese have abstract emphatic morphemes in negative contexts. We further propose that these abstract emphatic morphemes are covert counterparts of the overt emphatic makers -lul and -o. Under this analysis, the abstract morpheme can attach to the VP in Korean, while it cannot in Japanese. This is because the emphatic morpheme cannot attach to non-[+N] categories in 57 Japanese. To save this situation, the overt focus marker is needed in Japanese long-form negation. 2.3.1. An Obligatory Focus Marker in Focus Constructions In this section we examine two alternative approaches to the obligatoriness of the focus markers in focus constructions. The first approach comes from Poser's (1992) blocking hypothesis, which states that the existence of the simple form blocks the corresponding periphrastic form. The second approach concerns the underspecified complementizer Id. In the second approach, the [0N] feature of the complementizers -ki/ci in Korean and -i/0 in Japanese is responsible for the presence of the focus marker. In what follows, we see first how the blocking hypothesis can account for the obligatory presence of focus markers in focus constructions. 2.3.1.1. Poser (1992): Blocking of Phrasal Constructions In this section we examine whether the obligatory presence of a focus marker in the focus construction can be accounted for under Poser's (1992) Blocking Hypothesis. Even though this hypothesis can account for the obligatory presence of a focus marker, it incorrectly rules in the emphasis construction in Japanese. We will start with a brief summary of Poser (1992),. 58 Poser (1992), following Kageyama (1982), proposes that the periphrastic form, verb + suru, is blocked because a corresponding lexical verb already exists. Consider the following. (73) Verb Stem Derived Noun Gloss Periphrastic Gloss ir iri parching *iri suru parch mamor mamori borrowing *mamori suru protect oyog oyogi swimming *oyogi suru judge In (73), since lexical items like ir, mamor, and oyog already exist, the corresponding periphrastic forms are blocked. This blocking of phrasal constructions, however, does not occur if the focus markers (delimiters) like wa or bakari 'only' are attached to the verb, as seen in (74). This is because adding a focus marker to the verb can provide an additional meaning or function to the verb. (74) iri-wa sure 'parch' mamori-wa suru 'protect' oyogi-wa sure judge' Therefore, the constructions in (74) are not blocked since the meaning of the verbs differs from that of simplex verbs from which they are derived. Under this proposal, the periphrastic counterpart of the bare form with the same meaning is simply impossible. This blocking hypothesis can correctly rule out the focus constructions without the focus marker being attached to the embedded VP. Therefore, in the absence of a focus marker the sentence in (69) is ruled out. 59 (75) John-i wuli cip-ey o-ki-*(nun) John-NOM our house-to. come-COMP-FOC cacwu ha-ess-ta frequently do-PAST-DEC 'John came to my house frequently.' The same hypothesis can also account for the problem observed in Song (1988). Song (1988) suggests that the accusative marker -lul is obligatory in sentences like (76). 27 (76) a. ??John-i ka-M-lul ha-n-ta John-NOM go-COMP-ACC do-ASP-DEC 'John is going.' b. * John- i ka-ki ha-n-ta John-NOM go-COMP do-ASP-DEC 'John is going.' The minimal difference between (76a) and (76b) is the presence or absence of the accusative case marker after the embedded VP. The Blocking Hypothesis in Poser (1992) can correctly predict the sentence in (76b) to be ungrammatical. The grammaticality of the sentence in (76a), however, is surprising under the Blocking Hypothesis, if the accusative marker -lul functions as a pure Case marker without adding any meaning to the verb ka 'go'. This is because the Blocking should disallow the occurrence of both bare verb and the bare verb + ha for the same meaning. If we assume, following Song (1988) and Hoji (1991a), that -lul in (76a) is an emphatic maker, not a pure Case marker, the obligatoriness of 27 Although admitting that the sentence in (78a) is not acceptable, Song (1988) takes this sentence as grammatical. The sentence in (76a) is marginal to me. 60 -lul in (76a) can be accounted for. 28 Miyagawa and Ekida (1990) report that the morphological -o for accusative case maker is historically used as an emphatic marker. Recall that the periphrastic forms are possible when the focus marker wa is inserted between the verbal noun and the light verb su (Poser 1992). By adding the focus marker, the periphrastic form is given the focus force in the phrase that the focus marker is attached to, and, therefore, the focus construction is not blocked. The same can be said of the emphasis construction. Namely, the emphatic marker makes possible the periphrastic form. This blocking hypothesis, however, makes incorrect predictions in Japanese. If emphasis markers allow the periphrastic forms, then, we expect that the periphrastic form with the emphatic marker in Japanese are possible. Consider the following sentences. (77) a. * John-g a ik-i si-ta John-NOM go-COMP do-PAST 'John came.' b. John-ga ik-i-wa si-ta John-NOM go-COMP-FOC do-PAST 'John at least came.' c. * John-g a ik-i-o si-ta John-NOM go-COMP-EMP do-PAST 'John came.' The Blocking Hypothesis can correctly predict the unacceptability and acceptability of (77a) and (77b), respectively. The same hypothesis, however, incorrectly rules in the sentence in (77c). Given this problem,, we 28 We will provide additional arguments for Song (1988) and Hoji (1991a) below. 61 pursue an alternative analysis to the Blocking Hypothesis in the next section. 2.3.1.2. The Underspecified Complementizer Approach In this section, we pursue an alternative approach to the Blocking Hypothesis for the obligatoriness of the focus markers in focus constructions. We propose to attribute this state of affairs to the underspecified complementizer KI. Consider the focus constructions in (78). (78) a. John-i ka-ki-nun ha-ess-ta John-NOM go-COMP-FOC do-PAST-DEC 'John at least went.' b. *John-i ka-ki ha-ess-ta John-NOM go-COMP do-PAST-DEC 'John went.' As noted in the previous section, the focus marker is obligatory in the focus construction. To account for the obligatory presence of the focus marker, we suggest the following. First, the complementizer ki/ci in Korean and i/0 in Japanese are underspecified for [N], contra Kang (1986), who claimed that the complementizer ki has a nominal feature and can be a recipient of Case. Under our proposal, the VPs followed by the complementizer ki/ci or i/0 do not qualify as recipients for Case. Second, we propose, we suggest that the embedded VP needs Case as a complement of a Verb. We also assume that the focus particles assign inherent Case. 62 Under this analysis, the complementizers ki in Korean and i/0 in Japanese are not [+N] categories, and therefore cannot satisfy the requirement that it has to be assigned Case. This Case requirement can be met by the attachment of the focus marker. This analysis can also account for the acceptability of the emphatic constructions as in (79), with the assumption that the Korean emphatic marker -lul also assigns inherent Case. (79) John-i [wuli cip-ey o-ki]-lul John-NOM our house-to come-COMP-EMP cacwu ha-ess-ta frequently do-PAST-DEC 'John came to my house frequently.' In (79), the complementizer ki does not qualify as a Case recipient since it is not a [+N] category. The emphatic marker -lul assigns case to the VP and therefore, the sentence in (79) can be saved. We will fully discuss the emphatic markers in Japanese and Korean in the next section. 2.3.1.3. Emphatic Markers In this section we suggest that the emphatic markers -lul in Korean and -o in Japanese can assign inherent Case. We also claim that while -lul can attach to non-[+N] categories, -o can attach only to [+N] categories. Therefore, the distribution of the emphatic marker -o is more restricted than that of -lul. This hypothesis is supported by the place where -lul can 63 appear, but -o cannot. For one thing, -lul can be attached to verbal connectives in Korean, while -o cannot. Under this analysis VP-focus with -lul is more acceptable than with -o, given that the VP is not a [+N] category. (80) Emphatic Use of -lul and -o The emphatic -o in Japanese can be attached only to nominal categories while the emphatic -lul in Korean can be attached to non- nominal categories as well as to nominal categories. The acceptable status of the Japanese sentence in (81) and the unacceptable status of the Korean sentence in (82) fall under the current proposal. (81) ??John-i ka-ki-lul ha-ess-ta (K) John-NOM go-COMP-EMP do-PAST-DEC 'John went.' (82) *John-ga ik-i-o si-ta (J) John-NOM go-COMP-ACC do-PAST 'John went.' Under the current proposal, in (81), the embedded VP with ki, can be assigned inherent case by -lul, while in (82), the embedded VP cannot be assigned inherent case by -o, therefore it is ruled out. 29 In our discussion we crucially adopt the proposals by Song (1988), Ahn (1990), and Hoji (1991a), who point out that the accusative marker 29 The immediate question is why (81) is not fully acceptable. We will take up this issue below. 64 can function as an emphatic marker. To see whether the accusative marker -lul can function as an emphatic or focus marker, let us consider the following paradigm. (83) a. pi-ka rain-NOM 'It rained.' o-ess-ta fall-PAST-DEC b. pi-ka o-ki-nun rain-NOM fall-COMP-FOC 'It at least rained.' rain-NOM fall-COMP 'It rained.' d ?/*pi-ka rain-NOM 'It rained.' o-ki-lul fall-COMP-ACC ha-ess-ta do-PAST-DEC ha-ess-ta PAST-DEC ha-ess-ta PAST-DEC cinan sey tal-kan cacwu last three month often e. ?pi-ka o-ki-lul rain-NOM fall-COMP-ACC ha-teni do-RETROSPECTIVE 'Given that it rained often for last three months,.... f. pi-ka o-ki-lul ha-ess-ni? rain-NOM fall-COMP-ACC do-PAST-Q palam-i pwul-ki-lul ha-ess-ni? wind-NOM blow-COMP-ACC do-PAST-Q 'lit. Did it rain or did wind blow?' 'It did not rain nor did the wind blow.' The sentence in (83b) is the VP-focus version of the one in (83a). If we drop the focus marker, the unacceptable sentence obtains as in (83c), which is expected under either the Blocking Hypothesis or the underspecified Comp Hypothesis. If the focus marker nun is replaced by the accusative marker -lul, the unacceptable status obtains as in (83d). The acceptability status 65 of (83d), however, greatly improves if we insert an adverbial phrase between the focused VP and ha as in (83e). The embedded VP receives the emphatic interpretation in (83e). The pair question in (83f) confirms that -lul involves emphasis. The paired question in (83f) is fully acceptable with -lul after the verb. Notice that in (83f), two propositions are contrastively emphasized. The claim that two propositions are contrastively emphasized can be confirmed by the unacceptable status of the following sentence. (84) ?/*pi-ka o-ki-lul ha-ess-ni? rain-NOM fall-COMP-ACC do-PAST-Q 'Did it rain?' Without being paired, the sentence in (83f) is unacceptable, as the sentence in (84) indicates. Based on the above observations, we suggest that the accusative marker -lul can be used as an emphatic marker, supporting the hypothesis of Song (1988) and Hoji (1991a) about the emphatic use of-lul in the focus construction. If this hypothesis is correct, we have two functions of the accusative marker -lul. One function of -lul is the realization of the abstract Accusative Case, and the other is as the emphatic marker. We assume that the same is true for Japanese o. In what follows, we will examine how our proposal in (80) can account for the difference between Japanese and Korean in emphatic constructions. We will first examine Korean emphatic constructions. Consider the following pair of sentences. 66 (85) nay-ka i chayk-ul kentul-e-(??lul) po-ess-ta I-NOM thisbook-ACC touch-COMP-ACC try-PAST-DEC 'I tried reading this book.' (86) ney-ka i chayk-ul kentul-e-lul po-ess-ni? you-NOM this book-ACC touch-COMP-ACC try-PAST-Q il-e-lul po-ess-ni? read-COMP-ACC try-PAST-Q 'Have you (ever) touched this book, or have you ever read this book?" In (85), the embedded verb with -lul following it is not fully acceptable, indicating that the VP cannot be assigned Accusative. If we contrastively emphasize the VPs, the particle -lul can follow the VPs, as the sentence in (86) exhibits. Given that the sentence in (85) is not fully acceptable with the particle -lul, we assume that -lul in (85) is not a Accusative Case maker, but an emphatic marker, as discussed above. We further assume that the functional head -e in (85) is not a [+N] category, in that it can not be assigned Case, as the acceptability status of (85) suggests. Notice that the Japanese counterparts of (85) and (86) with -o after the VPs are totally unacceptable, confirming that the emphatic marker -o cannot be attached to non-[+N] categories. The proposed analysis make a correct prediction in emphatic constructions in Japanese, as the following paradigm in Japanese demonstrates. (87) a. :|: John-ga ik-i si-ta John-NOM go-COMP do-PAST 'John came.' 67 b. John-ga ik-i-mo si-ta John-NOM go-COMP-FOC do-PAST 'John at least came.' c. * John-g a ik-i-o si-ta John-NOM go-COMP-EMP do-PAST 'John came.' The sentence in (87a) is ruled out since the VP does not qualify as an abstract Case assignee due to the non-[+N] complementizer i. The sentence in (87b) is correctly ruled in since the VP can be assigned Case by mo 'also'. The sentence in (87c) with the emphatic marker -o is correctly ruled out since the emphatic marker -o cannot be attached to the VP, which is a non-[+N] category. One question arises concerning the degree of acceptability between focus constructions and emphatic constructions in Korean. If both focus markers and emphatic markers are specified for [+N], we may wonder where the difference in acceptability comes from. Consider the following Korean sentences. (88) ??John-i ka-ki-lul ha-ess-ta John-NOM go-COMP-EMP do-PAST-DEC 'John went.' (89) John-i ka-ki-nun ha-ess-ta John-NOM go-COMP-FOC do-PAST-DEC 'John at least went.' We, following H. Hoji's (p.c.) suggestion, propose that the emphatic markers -lul in Korean and -o in Japanese have weak semantic value as 68 opposed to focus markers such as nun 'concerning; at least', man 'only', and kkaci 'even' in Korean and their counterparts in Japanese. Thus far, we have argued that focus markers and emphatic markers can assign inherent Case and that while -lul can be attached to the categories other than NPs, the distribution of the Japanese emphatic marker -o is more restricted than that of Korean -lul. That is, the Japanese emphatic marker -i appears only on NPs, supporting the generalization in (80). Based on this generalization, in the next section we attempt to account for the optional vs. obligatory presence of focus markers in Korean and Japanese negative constructions. 2.3.2. Obligatory vs. Optional Presence of the Focus Markers In this section we attempt to answer the question as to the obligatory or optional presence of a focus marker in Japanese and Korean long-form negation, respectively, as demonstrated in (90) and (91). (90) John-i o-ci-(nun) ani ha-ess-ta (K) John-NOM come-COMP-FOC not do-PAST-DEC 'John did not come.' (91) John-ga kak-i-*(wa) si-na-katta (J) John-NOM go-COMP-FOC do-NEG-PAST 'John did not go.' As seen in (90), in Korean long-form negation, the focus maker nun is optional. The optionality of (90) is problematic for the analysis provided in the preceding section, where we suggest that the focus marker is needed for 69 Case reasons. Further complication arises if we consider the obligatory presence of the focus marker in the Japanese long-form negation, as in (91). To account for the optionality of the focus marker in (90), following Hoji (1991a), we suggest that Korean and Japanese have an abstract morpheme, analogous to the abstract emphatic morpheme in English (Chomsky 1957; Laka 1990). Chomsky (1957) suggests that there is a zero Aff morpheme in emphatic constructions such as (92). (92) John DID arrive. This Aff morpheme is taken as the affirmative counterpart of Negative morpheme in Laka (1990). In the long-form negation without an overt emphatic marker as in (93), we propose that the abstract emphatic morpheme follows the VP. EM in (93) implies the abstract emphatic marker. (93) John-i o-ci-EM ani ha-ess-ta (K) John-NOM come-COMP-EMP not do-PAST-DEC 'John did not come.' This abstract morpheme has the opposite semantic value of the English emphatic morpheme. We assume that this abstract morpheme co-occurs with a negation, as Hoji (1991a) suggests. We suggest that the abstract emphatic morpheme is the covert counterpart of the emphatic marker -lul and -o in. negative constructions. 70 The current proposal can account for the optional or obligatory presence of focus makers, demonstrated in (90) and (91), which are repeated here as (94) and (95). Again, EM implies an abstract Emphatic marker again. (94) John-i o-ci-EM ani ha-ess-ta (K) John-NOM come-COMP-EMP not do-PAST-DEC 'John did not come.' (95) *John-ga kak-i-EM si-na-katta (J) John-NOM go-COMP-EMP do-NEG-PAST 'John did not go.' We have claimed that the emphatic maker in Korean can be attached to non-[+N] categories, while the one in Japanese cannot. We have also suggested that the abstract emphatic markers in Korean and Japanese have the same features as their overt emphatic makers. Under this proposal, the abstract emphatic morpheme can appear after the VP in Korean, but not in Japanese. Hence the (un)grammatical status of (94) and (95). The embedded VP can be assigned inherent Case in (94) while, the embedded VP in (95) is caseless. To save this situation, the embedded VP has to be provided with the focus marker, which assigns inherent Case. Hence, the focus marker is obligatory for the sentence in (95). In this section, we have argued that the absence or presence of the abstract emphatic morpheme is responsible for the obligatory or optional presence of focus markers in Japanese and Korean. The blocking hypothesis, which counts only the surface form, cannot provide an 71 adequate explanation for the optional presence of the focus marker in the Korean focus construction. 2.4. Summary and Concluding Remarks In summary, this chapter has attempted to show that DO-verbs in Japanese and Korean, namely su, ar, and ha in focus constructions are base-generated, but not inserted to rescue a stranded morpheme such as tense. First, we have suggested that ar in Japanese takes AdjP as a VP- complement. We have provided evidence indicating that na belongs to a lexical category adjective. We found evidence from verbal inflections on ar, the co-occurrence relation between ar and the head of its complement, and from nominalization using -sa. More specifically, (1) adjectives and na bear the same verbal inflection; (2) Adjectives and na, co-occur with ar; (3) adjectives and na take -sa for nominalization. We have also provided indirect evidence that na belongs to an lexical category by showing that the Korean imperative negative ma also belongs to the lexical category verb. Illustrating the co-occurrence relation between DO and the head of its complement, we propose that the co-occurrence relation can be captured by positing the abstract verbs dO and bE. Both dO and bE have their own feature bundles. The verb dO co-occurs with verbs, while the verb bE co-occurs with adjectives, taking verbs and adjectives as heads of their complements, respectively. Under this analysis, the same abstract verb dO is phonetically realized as ha in Korean and su in Japanese in the 72 same environment. The verb bE, however, occurs only in Japanese as ar since Korean does not have adjectival predicates. Second, we have suggested that the multiple occurrence of DO supports the Base-generation hypothesis. Under the DO-support Hypothesis, more than one INFL has to be stipulated for DO to be generated under it in a simplex sentence. Under the Base-generation Hypothesis, multiple negation in Korean is attributed to the iterative base- generated DO and the preverbal negative adverb ani being generated as a complement of DO. We have claimed that the Korean negative adverb ani is generated as a complement of V° and undergoes incorporation to a verb. We have provided evidence for the adverbialhood of ani from the attachment of the plural marker tul to ani. We have also provided evidence for the incorporation oiani into the verb from the reduplication of the focus phrase. While other adverbs can be optionally reduplicated in the reduplication of the focus phrase, ani is obligatorily reduplicated. We take this fact as evidence for the incorporation of ani to verbs. We attribute this dual property of ani to its being X° and X max at the same time, following Chomsky (1994). Third, we have argued that the lack of VP-ellipsis in Japanese and Korean supports the Base-generation Hypothesis. The answer to a 'yes-no question' in Japanese and Korean does not involve VP ellipsis. We also claimed that VP-fronting supports the Base-generation Hypothesis. In the double DO constructions, the complement of DO containing the other DO can be fronted. Under the DO-support Hypothesis, where DO is inserted 73 under Tense, DO in the fronted phrase is problematic since the VP fronting does not front Tense under which DO is inserted. Fourth, we have argued that the obligatory presence of a focus marker in focus constructions can be attributed to the requirement that VPs need Case in the complement position of a verb. We propose that the VP-complementizer -Id in Korean and -i are not specified for [+N], and, therefore, the VP complements cannot be assigned Abstract Accusative Case. We, following Stowell (1981) and Chomsky (1981), assume that only [+N] categories can be assigned Abstract Accusative Case. We also assume that focus makers can assign inherent Case just as postpositions do. Under this analysis, the VP without Case in the complement position of V, can be saved by a focus marker, which assigns inherent case to it. We, following Song (1988), Ahn (1990), and Hoji (1991a), have suggested that -lul in Korean and -o in Japanese can function as emphatic markers. As pointed out by H. Hoji (p.c), these emphatic markers differ from focus markers, such as wa 'at least' and sae 'even', in that these emphatic markers do not have a full semantic value, and, therefore, the attachment of these two markers to VPs does not make focus constructions as natural as other regular focus markers do. In this sense, the emphatic markers can be taken as weak focus markers. We have also suggested that -lul and -o are sensitive to the categories they attach to. While -lul can attach to non-[+N] categories, -o can attach only to [+N] categories. Therefore, the distribution of the emphatic marker -o is more restricted than that of -lul. This hypothesis is supported by the place where -lul can appear, but -o cannot. For one 74 thing, -lul can be attached to verbal connectives in Korean, while -o cannot. Under this analysis the VP-focus with -lul is more acceptable than with -o, given that the VP is not [+N] category. In order to account for the optional vs. obligatory presence of focus markers in Korean and Japanese long-form negation, respectively, we, following Hoji (1991a), have suggested that Korean and Japanese have abstract emphatic morphemes in negative contexts. We have further proposed that these abstract emphatic morphemes are covert counterparts of the overt emphatic makers -lul and -o. Under this analysis, the abstract morpheme can attach to the VP in Korean, while it cannot in Japanese. This is because the emphatic morpheme cannot attach to non-[+N] categories in Japanese. To save this situation, the overt focus marker is needed in Japanese long-form negation. We have noted that Poser's (1992) Blocking Hypothesis, which counts only surface forms, cannot provide an account for the obligatory vs. optional focus markers in Japanese and Korean focus constructions, respectively. Under the Blocking Hypothesis, the long-form negation without a focus marker in Korean is incorrectly predicted to be unacceptable, since this hypothesis blocks formation of a periphrastic form such as verb+/za, if a corresponding lexical form already exists. Korean has a corresponding short-form negation without ha. Establishing that DO in the focus construction is not inserted but base-generated paves the way to a unitary account for both focus constructions and light verb constructions (Grimshaw and Mester 1988). 75 In the following chapters, we will in fact provide a unified account for both focus constructions and light verb constructions. 76 Chapter III. Two Types of VNs: Underspecified VN 3.1. Introduction In Chapter 2, we have argued that DO in the focus constructions in J&K is base-generated. In this chapter, we discuss the light verb construction as analogous to the focus constructions in that both constructions have verbs with incomplete theta-grids, namely DO and light verbs, respectively. We claim that VNs with the same morphology have two distinct lexical features: nominal, as in (la), and verbal, as in (lb), contra proposals for a unitary lexical feature system for all VNs, suggested in Miyagawa (1987), Sells (1990), and Kim (1991). (1) a. John-ga [suugaku-no benkyoo]-o si-ta John-NOM Math-GEN study-ACC do-PAST 'John studied Math.' b. ??John-ga [suugaku-o benkyoo]-o si-ta John-NOM MathrACC study-ACC do-PAST 'John studied Math.' This system of two distinctive lexical features for VNs has been proposed by Hasegawa (1991), Park (1992), and Dubinsky (1994). Dubinsky (1994) proposes that one of the syntactic categories of VNs is not specified for [N]. Our proposal differs from Hasegawa (1991) and Dubinsky (1994) in that the underspecified VN has the lexical feature [0V, -N]. The theory of 77 syntactic underspecification for the so-called verbal nouns (henceforth, VN) is based on the studies pursued in Burzio (1991), Kayne (1989), Van Gelderen (1992), Vanden Wyngaerd (1994), Rooryck (1994), and Dubinsky (1994) for other syntactic categories except for Dubinsky (1994), who applied the underspecification theory to VNs - 1 Following Lee (1991), we claim that the verbal noun phrase (VNP) headed by verbal VN is governed by a null VP-complementizer with the feature [+N], which enables the VNP to qualify as a Case assignee, as the construction in (2) shows. (2) John-i [swuhak-ul kongpwu-0]-lul ha-ess-ta John-NOM Math.-ACC study-COMP-ACC did-PAST-DEC 'John studied Math.' We provide two pieces of evidence supporting this point. First, given that the [+N] categories qualify as Case assignees (Stowell 1981) and that the verbal noun in (2) is specified for [0V, -N], for the verbal noun phrase to qualify as a Case assignee, a [+N] category is need. Secondly, we propose the hypothesis that the lexical item does not directly select the maximal projection headed by another lexical category; rather, it selects a maximal projection headed by a functional category. Under this hypothesis verbs select functional projections like DP, IP, or CP, but not NP or VP. If this is the case, the light verb does not directly select the verbal noun phrase 1 Milsark (1988) and Ohta (1994) also suggest that the underspecified syntactic categories can be recategorized in certain contexts. Ohta (1994) suggests a recategorization analysis for verbal Renyoo-kei Constructions in Japanese. In this system, the FP in Renyoo-kei Constructions is recategorized by a c-commanding head. In other words, the feature of FP is determined by a c-commanding head. 78 headed by a verbal noun; rather, it selects a functional projection, which is a null complementizer containing the verbal noun phrase in the light verb construction. This analysis assimilates Gerundive Constructions in English to the light verb constructions in J&K. The gerund in English behaves as a verb internally and as a nominal externally (Reuland 1983; Abney 1987 and references cited therein). The same state of affairs can be observed in the Light Verb Constructions in J&K, namely: a verbal VN can assign verbal Case, a verbal VN cannot be modified by an adjective, and the phrase headed by a verbal VN can be placed in Case positions. 2 This chapter is organized as follows. Section 3.2 reviews Dubinsky's proposal for lexical categories of VNs. Section 3.3 presents our proposals. Section 3.4 introduces Miyagawa's proposal concerning the lexical features of VN for Japanese. We provide further evidence in support of Miyagawa's lexical feature [-V, +N] for VNs. Section 3.5 presents problems with the unitary lexical feature system. Section 3.6 proposes that the (verbal) VN in [NP-ACC VN]-(ACC) has a verbal feature as opposed to the (nominal) VN in [NP-GEN VN] or in [VN]. Section 3.7 claims that the VNP is governed by a null VP-complementizer, attempting to provide a parallel account for Light verb constructions and focus constructions. Section 3.8 suggests that VN and its accusative theme form a constituency. Section 3.9 attempts to show how the proposed analysis can account for the VN 2 A verbal VN phrase cannot be placed in all Case positions. This issue will be fully discussed later in this chapter. 79 structures containing an aspectual morpheme. Section 3.10 concludes this chapter. 3.2. Dubinsky (1994): Underspecified Lexical Categories The thematically incomplete verbs (i.e., light verbs) typically co-occur with the theta-assigning nouns, VNs. The combination of VNs and light verbs can typically appear in the following three contexts. 3 (3) a. John-ga suugaku-no benkyoo-o si-ta John-NOM Math-GEN study-ACC do-PAST 'John studied Math.' b. ??John-ga suugaku-o benkyoo-o si-ta John-NOM Math-ACC study-ACC do-PAST 'John studied Math.' c. John-ga suugaku-o benkyoo si-ta John-NOM Math-ACC study do-PAST 'John studied Math.' To derive the three types of sentences in (3), Dubinsky (1994) suggests that the verbal nouns with the same morphology potentially have two different lexical categories, one with [+N] and another with no specification for [N]. Under Dubinsky's (1994) analysis, verbal nouns like benkyoo 'study' in (3) can potentially have the two category characterizations as exhibited in (4). (4) benkyoo a. [+N] b. 0 3 The term 'verbal noun' is credited to Martin (1975). 80 Two possible D-structures for the light verb constructions are shown in (5). (5) a. b. V v" 'study* suru 'study' suru In both (5a) and (5b), suru is a light verb. In (5a), the VN benkyoo 'study' is specified for [+N], while in (5b), the VN is not specified for the [N] feature. With two distinctive lexical feature for the same VN, Dubinsky (1994) can derive the three types of light verb constructions demonstrated in (3). The following sentences are from Dubinsky (1994) with minor modification. (6) a. Taroo-wa Ziroo-ni [NP yuusyoku-no teian]-o si-ta Taroo-TOP Ziro-DAT dinner-GEN propose did 'Taro proposed dinner to Ziro.' b. Taroo-wa Ziroo-ni [NP yuusyokuj-o [NP tj teian]-o si-ta Taroo-TOP Ziro-DAT dinner-GEN propose-ACC did 'Taro proposed dinner to Ziro.' c. Taroo-wa Ziroo-ni [NP yuusyokuj-o [NP tj teian] si-ta Taroo-TOP Ziro-DAT dinner-GEN propose did 'Taro proposed dinner to Ziro.' In (6a), the VN is specified for [+N], and hence Genitive Case on the theme yuusyoku 'dinner'. Dubinsky (1994) assumes that the Genitive Case assignment by an N is optional. If the genitive is assigned by the VN with the feature [+N], (6a) obtains. If the Genitive Case is not assigned by the 81 VN with the feature [+N], then, the Theme NP yuusyoku 'dinner' adjoins to NP to be assigned Accusative by the light verb su. In (6c), the VN teian 'propose' projects an unspecified XP. Therefore, the VN cannot assign Genitive Case. Hence, the theme NP adjoins to the NP to be assigned Accusative case by su. Even though the mechanism pursued in Dubinsky (1994) provides a way to derivationally relate the three types of light verb constructions in (3) or in (6), one problem arises concerning the categorial feature of VN in (6). Dubinsky basically assumes that the VNs in (6a) and (6b) have the same categorical feature. If this is correct, we predict that the VNs both in (6a) and (6b) can be modified by an adjective, which is not the case. Only the VN in (6a) can be modified by an adjective. The VNs in (6b) and (6c) cannot be modified by an adjective. Given this problem, we suggest that the lexical category of the VNs in (6a) and (6b) cannot be the same and that two lexical categories are derived in the lexicon (Kiparsky 1982; Baker 1985). In the following sections, we examine the categorial features of VNs. 3.3. Proposals In this chapter, we propose that the so-called Verbal Nouns (henceforth, VN) with the same morphology have two distinct lexical features: a nominal VN with the lexical feature [-V, +N] as in (7a) and a verbal VN with the lexical feature [0V, -N], as in (7b). 82 (7) a. John-ga [suugaku-no benkyoo]-o si-ta John-NOM Math-GEN study-ACC do-PAST 'John studied Math.' b. ??John-ga [suugaku-o benkyoo]-o si-ta John-NOM Math-ACC study-ACC do-PAST 'John studied Math.' We also claim that the verbal noun phrase (VNP) headed by verbal VN, as in (7b), is governed by a null VP-complementizer with the feature [+N], which enables the VNP to qualify as a Case assignee, as the construction in (8) exhibits. (8) John-i [swuhak-ul kongpwu-0]-lul ha-ess-ta John-NOM Math.-ACC study-COMP-ACC do-PAST-DEC 'John studied Math.' We specifically claim that VNs in the environment [NP-ACC VN]- (ACC), as in (7b), are verbal VNs with the underspecified feature [0, -N] and those in other environments are nominal VNs with the feature [-V, +N]. All the syntactic differences between these two types of VNs. (i.e. verbal vs. nominal Case assignment by VN) follow from this categorial distinction, instead of an optional incorporation of the VN to the light verb (Kageyama 1982; Terada 1990). This analysis provides an account for the light verb construction in a way analogous to the one for focus constructions in J&K developed in Chapter 2. (see Kang 1988 for Korean; Hoji 1991a for Japanese). We claim that the null VP-complementizer in the light verb construction is equivalent to the non-overt VP-complementizer in the VP-focus construction. 83 Under the current analysis, the light verb constructions containing the verbal VN is similar to the gerundive constructions in English. The verbal VN, however, differs from the verb in the gerundive constructions in English in that the verbal VN cannot bear verbal inflectional morphemes and that the VN-phrase cannot be placed in the subject position nor can it be an object of P. Following Miyagawa (1990) and Lee (1994), we suggest that Case-licensing by Aspect can solve the problems described above. In what follows, we provide the details of our proposal. We especially focus on the dual property of the verbal VNs in that they behave both like a verbal category (i.e. no adjectival modification) and like nominal categories (i.e. Case assignee) at the same time. 3.3.1. Zero Derivation We suggest that the same morphology with two different lexical features is derived by zero derivation in the lexicon, as suggested in Kiparsky (1982) for English words. Kiparsky (1982) observes that the presence of noun- verb pairs, as in (9) gives support to the arguments for zero derivation within the framework of Lexical Phonology. (9) torment ~ torment V N record, permit, conflict, transfer, rebel, convert, produce, etc. Kiparsky (1982), suggests that the words in (9) undergo V --> N zero derivation. Thus, [N torment] has a nested structure [N [V torment] ]. A 84 similar proposal was made for the English Gerund in Baker (1985) to account for the 'Ing-of in (10) , which lacks the verbal characteristic to be found in the Poss-ing in (11). (10) I enjoyed John's singing of the aria Ing-of (11) We count on Linda's solving the problem. Poss-ing Baker (1985) suggests that the affixation of-ing to the verb in (10) takes place in the lexicon, as opposed to the same operation in syntax for (10). Based on Kiparsky (1982), we suggest that verbal VNs with the feature [0V, -N] undergo zero derivation to N in the lexicon, which produces nominal VNs with the feature [-V, +N]. 3.3.2. Underspecification Approach: [0V, -N] feature It is assumed that the lexical categories defined by [-N] are within the class of Case-assigners (Stowell 1981; Chomsky 1981). These categories include verbs ([+V, -N]) and prepositions ([-V, -N]). It is also the case in J&K that verbs assign a verbal Case (Accusative Case here) and postpositions assign oblique case. Given that [-N] categories can assign Case and that the verbal VN also assigns Case (when it is followed by a light verb or aspectual morpheme), we can say that the VN is specified for [-N]. We take a [-N] category to be a potential Case assigner. That an adjective, which modifies a noun, cannot modify the verbal VN further confirms that 85 the verbal VN is not a noun. This state of affairs can be accounted for if we posit that the verbal VN is specified negatively for the [N] value. We also propose that the value for [V] is unspecified. That the verbal VN cannot bear verbal inflection can also be accounted for under the current proposal. In Japanese, only verbal categories (verbs and adjectives), which are positively specified for [V], enter into the verbal inflectional paradigm (Miyagawa 1987; lida 1987). Since the verbal VNs do not bear verbal inflectional morphology, we would say that the verbal VNs are not positively specified for [V]. However, we do not want to say that VNs are negatively specified for [V] either. This is because the [-V, -N] category, namely preposition or postposition, assigns oblique Case, which is not the Case assigned by a verbal VN. 3.3.3. [+N] feature As mentioned above, the VP headed by a verbal VN can bear Accusative Case. Given that only nominal expressions are assigned Structural Case, the VP headed by a verbal VN can be taken as [-V, +N]. If this is the case, then we are in a contradictory situation. That is, the verbal VN has to be specified for both [-N] and [+N] at the same time. As a Case assigner it has to be [-N] on the one hand and as a Case assignee, [+N]. This contradictory situation can be solved if we consider them in the same manner as English Gerundive constructions. In English Gerundive constructions, the gerund behaves as a verb internally and as a noun 86 externally (Reuland 1983; Abney 1987 and references cited therein). That is, the gerund shows the same distribution as NPs externally. The VP headed by a verbal VN shows the same property in that it can be assigned Structural Case, which is exactly what happens to NPs. Given this, following Lee (1991), we suggest that the VP is governed by a null-Comp. We further suggest that this null complementizer is a VP-complementizer, with the feature [+N]. Therefore, the Accusative Case appearing after VP is not in fact attached to VP, but to the VP-complementizer. We propose that a null-complementizer is motivated by the general requirement that a lexical head does not directly select a lexical projection, but a functional projection. Thus far we have seen the proposals in this chapter, in the following sections, we will see the details of each proposal. 3.4. The Nominal Feature of VN This section examines the nominal property of VNs. This section first presents Miyagawa's evidence supporting the claim that VNs have a nominal feature. Additional evidence will also be presented. We show that Miyagawa (1987) is basically correct in that VNs have a nominal property. We, however, do not endorse Miyagawa's (1987) claims that the lexical features of VNs are unitary. We will question Miyagawa's unitary lexical feature hypothesis in Section 4. 87 3.4.1. Miyagawa (1987): Unitary VN with [-V, +N] This section presents Miyagawa's claim that VNs are uniformly [-V, +N]. Miyagawa (1987) points out that VNs act as subject arguments and function as heads of a relative clause. Furthermore, they cannot directly bear verbal inflectional morphology. These properties are all identical to those of simple nouns. Given these facts, Miyagawa (1987) suggests that VNs are associated with the lexical features [-V, +N], the feature of a simple noun. The following examples are from Miyagawa (1987, P32). (12) Benkyoo ga hazimaru. (Subject) studying NOM start 'The studying will start.' (13) Taroo ga sukina benkyoo (Relative Clause) Tao NOM like study 'The study that Taro likes.' As seen above, the VN, benkyoo 'study' can behave as a subject argument or a head of a relative clause. VNs differ from verbs in that they do not enter the paradigm of verbal inflectional morphology as seen in (14). (14) a. benkyoo-*(si)-ta study-*(do)-PAST 'studied' b. ki-ta come-PAST 'came' 88 In (14a), the VN benkyoo 'study' cannot directly bear the past morpheme, showing that it does not behave like a verb in terms of taking verbal inflectional morphology. Notice that the regular verb ki 'come' in (14b) can directly bear the past morpheme. Given these similarities between VNs and simple nouns, Miyagawa (1987) claims that VNs have the same categorial feature as a simple noun and that the only difference between VNs and simple nouns is the capability of VNs to assign theta-roles. In what follows, we provide further evidence in support of Miyagawa's (1987) claim that VN is specified for [-V, +N]. 3.4.2. Further Evidence for the nominal feature of VNs The modification of VNs by an adjective further supports Miyagawa's claim that the VNs behave like simple nouns in terms of syntactic distribution, as seen in (15). (15) John-ga muzukasii benkyoo-o si-ta John-NOM difficult study-ACC do-PAST 'John did a difficult study.' In (15) the verbal noun benkyoo 'study' is modified by the adjective muzukasii 'difficult.' Even when the VN takes a genitive theme NP, the VN can be modified by an adjective as seen in (16). 4 > 5 4 The subject pro of the embedded clause will not be indicated throughout this paper unless it is necessary to show it. 5 The adjective muzukasii 'difficult' can also be placed before the genitive theme NP suugaku 'Math.', as seen in (i). 89 (16) John-ga [suugaku-no muzukasii benkyoo]-o si-ta John-NOM Math-GEN difficult study-ACC do-PAST 'John did a difficult study of Math.' In (16), the VN benkyoo 'study' takes the Theme suugaku 'Math.' and is modified by an adjective muzukasii 'difficult'. Furthermore, as seen in (16), the complement of the nominal VN has nominal Case, which indicates that the VN is a nominal VN. Given the well-known generalization in (17) below for the appearance of Genitive Case (Saito 1983; Kitagawa&Ross 1982), the VN must be taken as a nominal category. (17) Insertion of Genitive Case marker NP/PP N(P) Relativization of the NP in (16) is also possible. (18) John-ga si-ta [suugaku-no muzukasii benkyoo] John-NOM do-PAST Math-GEN difficult study-ACC 'The difficult study of Math that John did.' Thus far, we have seen that VNs behave like simple nouns, supporting Miyagawa's (1987) claim that VNs are identical to simple nouns in terms of the lexical feature, that is, [-V, +N]. The data shown so (i) John-ga [muzukasii suugaku-no benkyoo]-o si-ta John-NOM difficult Math-GEN study-ACC do-PAST 'John did a study of difficult Math.' The two sentences in (16) and (i), however, have two different readings. That is, while in (16) the adjective modifies the VN, in (i) it modifies the complement of the VN, namely suugaku 'Math.' 90 far, however, are limited to the patterns of [NP-GEN VN] or [VN] without its Accusative NP complement. We claim that the VN in those environments are in fact specified for [-V, +N] like simple noun. As will be exhibited in the following section, the VN in [NP-ACC VN] exhibits behaviors different from simple nouns. 3.5. Unitary [-V, +N] and double -o constraint This section examines the previous proposals in Kageyama (1982) and Miyagawa (1987) to see how the unitary [-V, +N] lexical feature hypothesis could be maintained. We also point out that the existence of an unincorporated version of [NP-ACC VN]-ACC constructions in J&K raises questions for the unitary lexical hypothesis. 3.5.1. Kageyama (1982) and Miyagawa (1987) This section briefly reviews two previous proposals concerning VN+ light verb su and sees how the impossibility of adjectival modification of VN could be treated in Kageyama (1982) and Miyagawa (1987). Let us first consider the following sentences. (19) John-ga muzukasii benkyoo-o si-ta John-NOM difficult study-ACC do-PAST 'John did a difficult study.' (20) John-ga [suugaku-no muzukasii benkyoo]-o si-ta John-NOM Math-GEN difficult study-ACC do-PAST 91 'John did a difficult study of Math.' (21) John-ga suugaku-o (*muzukasii) benkyoo-(o) si-ta John-NOM Math.-ACC difficult study-(ACC) do-PAST 'John did a difficult study of math.' Note that the modification of VN by the adjective muzukasii 'difficult' is possible if the VN does not take the Accusative Theme, as the sentences in (19) and (20) indicate. In (19) and (20), the Verbal Noun (VN) benkyoo 'study' is modified by an adjective muzukasii 'difficult', indicating that the VN has a nominal feature. In (21), however, the VN cannot be modified by the adjective. There have been two possible approaches to this phenomenon. These two approaches do not have any empirical differences in dealing with the VN plus light verb suru. One approach comes from Kageyama (1982). In Kageyama (1982), the VN benkyoo 'study' is incorporated into the following light verb su. In this approach the resulting verbal noun and light verb form a new verb. Therefore, the adjective can no longer be used to modify the verbal noun. Another approach can be found in Miyagawa (1987). In Miyagawa's (1987) approach, the light verb is inserted to provide the VN with [-V, +N] in order for it to bear verbal inflections. 6 Since the VN and the inserted suru form a verb, no adjectival modification of VN is possible. Furthermore, in Japanese, it is claimed that more than one -o in a single sentence makes the sentence marginal or unacceptable due to the 6 Miyagawa's approach is similar to Abasolo (1974:72). Abasolo (1974) calls ha a 'verbalizer', which converts a noun root into a verb root. 92 Double-o constraint (Harada 1973). 7 Therefore, the sentences with double -o's as in (22) are already taken as unacceptable even before the adjective is inserted to modify the verbal noun. (22) ??John-ga suugaku-o benkyoo-o si-ta John-NOM Math.-ACC study-ACC do-PAST 'Lit. John did a study of math.' Since sentences like (22) are taken as unacceptable, not much consideration has been given to structures like (21) in terms of adjectival modification. Thus far, we have seen that the incorporation hypothesis (Kageyama 1982) and the insertion of the light verb onto the VN (Miyagawa 1987), together with the double -o constraint make unnecessary any consideration of the impossibility for adjectival modification of VN in Japanese. In the next section, we will show that the Korean counterpart of (22) is grammatical and that even in Japanese the light verb constructions with double -o is not entirely unacceptable. 7 The marginal status of (22) indicates that the surface Double -o constraint applies to (22a). One piece of evidence in support of the surface double -o constraint can be found when one of the accusative case markers is replaced by a delimiter like sika 'only' or mo 'also', as the grammatical sentence in (ii) indicates, (i) ??John-ga suugaku-o benkyoo-o si-nai John-NOM Math.-ACC study-ACC do-not 'John studies Math.' (ii) John-ga suugaku-sika benkyoo-o si-nai John-NOM Math.-only study-ACC do-not 'John studies only Math.' In (ii) the accusative marker -o is replaced by the delimiter sika 'only' and the sentence is perfect. If (i) were under the deep double -o constraint, (ii) should be less acceptable than it is (see Poser 1981; Kuroda 1988 for more details). 93 3.5.2. Unincorporated VNs and double -o Korean light verb constructions clearly indicate that there are cases where the verbal VN is not incorporated into the following light verb and that the VN and light verb do not form a verb. Even in Japanese, despite the double-o constraint (Harada 1973; Poser 1981; Kuroda 1988), the same can be observed. The unincorporated [NP-ACC VN]-ACC constructions provide a test to check whether the unitary nominal analysis of verbal nouns is valid. The following light verb constructions in Korean suggest that the VN is not incorporated into the light verb. (23) a. John-i swuhak-ul kongpwu-lul ha-ess-ta John-NOM Math.-ACC study-ACC do-PAST-DEC 'Lit. John did a study of math.' b. John-i swuhak-ul kongpwu-lul yelsimhi ha-ess-ta John-NOM Math.-ACC study-ACC hard do-PAST 'John studied math hard.' The sentence in (23a) is the exact Korean counterpart of the Japanese sentence in (22). As opposed to the Japanese sentence with double accusative marker -o in (22), the sentence in (23) is fully acceptable, indicating that the double-o constraint is not applicable to Korean sentences. In (23), the VN is followed by an accusative marker. We assume that the accusative case marker -lul after the VN is a realization 94 of abstract Accusative Case. 8 Following Baker (1988), we also assume that NPs satisfy the Case Filter (Rouveret & Vergnaud 1980; Chomsky 1980) either by having Case or by the head of NP being incorporated. 9 Under these two assumptions, the VN in (23a) is not incorporated into the light verb ha 'do'. The insertion of an adverb between the VN and light verb, as in (23b), further suggests that the VN and light verb are not incorporated. Japanese Light Verb Constructions confirm the same point. Let us consider the sentence in (22), repeated here as (24). (24) ??John-ga suugaku-o benkyoo-o si-ta John-NOM Math.-ACC study-ACC do-PAST 'Lit. John did a study of math.' In Japanese, more than one -o in a single sentence is claimed to make the sentence marginal or unacceptable due to the (surface) Double-o constraint (Harada 1973). Sells (1990), Isoda (1991), and Hoshi & Saito (1993) however, report that Light Verb Constructions with double -o's, as in (24), are not entirely unacceptable in Japanese. Based on the description that unincorporated [NP-ACC VN]-ACC constructions are 8 Kuroda (1988) has a different view on case markers in Japanese. Kuroda (1988) claims that the case markers are not abstract Case. H. Hoji pointed out to me (p.c.) that this could be the case for J&K. In this chapter, we assume that the case markers are realizations of abstract Case. 9 Baker (1988) presents the following visibility condition (Baker 1988:117) B receives a theta role only if it is Case-indexed. A NP can be Case-indexed either by being assigned Case or by N being incorporated. 95 possible in J&K, in the next sections, we will consider the lexical property of verbal nouns. 3.6. A Verbal ([+V, -N]) approach to the VN. In this section, we suggest that the VN in [NP-ACC VN]-(ACC) is not specified for [-V, +N], as opposed to the VN which does not take an Accusative Theme. We attempt to show the verbal property of the verbal VN in two respects: a VN cannot be modified by an adjective; a VN can assign verbal Case. 3.6.1. Adjectival modification In the preceding section, we have seen that the [NP-ACC VN]-(ACC) structure is possible. We will consider now whether the verbal noun can be modified by an adjective and determine what the (im)possibility of modification says about the lexical feature of VNs. If an adjective is placed before the VN in the [NP-ACC VN]-(ACC) construction the resulting sentence is hopeless, as the ungrammatical sentence in (25b) shows. (25) a. ??John-ga suugaku-o benkyoo-o si-ta John-NOM Math.-ACC study-ACC do-PAST 'Lit. John did a study of math.' b. *John-ga suugaku-o muzukasii benkyoo-o si-ta John-NOM Math.-ACC difficult study-ACC do-PAST 'John did a difficult study of math.' 96 Note that the modification of VN by the adjective muzukasii 'difficult' is possible if the VN assigns a nominal Case, as the sentence in (26) demonstrates. (26) John-ga [suugaku-no muzukasii benkyoo]-o si-ta John-NOM Math-GEN difficult study-ACC do-PAST 'John did a difficult study of Math.' The problem with adjectival modification could be circumvented under the unitary lexical feature system by saying that since the VN incorporates into the light verb, the VN and light verb form a verb, and therefore, the adjective cannot modify the VN+light verb (Kageyama 1982). Given that the VN (which is not incorporated into the following light verb) cannot be modified by an adjective, the impossibility of adjectival modification can no longer be attributed to this incorporation. This problem leads us to an approach to the lexical feature of VN different from the incorporation approach. Adjectives can modify a noun only, but not a verb, postposition, or another adjective. In other words, adjectives modify only full-fledged nouns. The fact that an adjective cannot modify a verbal VN indicates that the verbal VN is not a full noun with the lexical feature [-V, +N]. 3.6.2. Assignment of Verbal Case The assignment of Verbal Case by the VN supports the claim that the verbal VN in [NP-ACC VN]-ACC has a verbal property, questioning the 97 unitary lexical feature hypothesis, which claims that VN is uniformly a noun with [-V, +N]. 10 The assignment of verbal Case can be found in both light verb constructions and temporal/aspectual constructions. As we have seen in the previous sections, the verbal VN in light verb constructions can assign Accusative Case. The following sentences show this point. (27) ??John-ga [suugaku-o benkyoo]-o si-ta (Japanese) John-NOM Math.-ACC study-ACC do-PAST 'John studied Math.' (28) John-i [swuhak-ul kongpwu]-lul ha-ess-ta (Korean) John-NOM Math.-ACC study-ACC do-PAST 'John studied Math.' In (27)-(28), the verbal nouns benkyoo 'study' and kongpwu 'study' assign Accusative to their complements suugaku 'Math.' and kongpwu 'study', respectively. The VN in Aspectual Constructions can also assign Accusative Case, as observed in Iida (1987), Sells (1990), and Miyagawa (1990) for Japanese. 11 When VNs are affixed by a temporal/aspectual morpheme 10 Under the unitary lexical hypothesis, which argues for the feature [-V, +N] for all VNs (Iida 1987; Miyagawa 1987; Sells 1990), the Case alternation between Accusative and Genitive on the theme NP of the VN is stipulative. Sells (1990) suggests that Aspectual verb (or morpheme) optionally assigns the feature t+Asp] to the VN with the feature [-V, +N] under government so that the VN can assign Accusative. We see two stipulations in Sells' proposal. One is optionality of the assignment of the Asp feature. Another is the category that assigns verbal case. In Sells (1990), a [-V, +N] category assigns verbal Case as long as it is governed by Asp. The following is from Sells (1990) (his 59). The feature l+asp] is optionally assigned by suru under government. 11 The same is true for Korean (see Lee 1993 for Korean). 98 such as -tyuu 'while', -go 'after', axid-izen 'before', they can assign Accusative Case. 12 - 13 The canonical examples are in (29)-(30). (29) John-ga [suugaku-o benkyoo]-chuu (Japanese) John-NOM Math-ACC study--mid 'while John is studying Math.' (30) John-i [swuhak-ul kongpwu]-cwung (Korean) John-NOM Math-ACC study-mid 'while John is studying Math.' In (29)-(30), the VNs can assign Accusative Case. The assignment of the verbal Case is possible when the VN is affixed by an temporal/aspectual morpheme. Given that nominals do not assign verbal Case, the assignment of verbal Case by the VN indicates that the VNs in (29)-(30) are not nominal categories, contra to Miyagawa (1990) and Sells (1990). English process nouns, on the other hand, never assign verbal Case, as seen in (31) below. (31) a. After the investigation *(of) the incident... b. Before the examination *(of) the eyes... 1 2 Tsujimura (1992) also discusses Case assignment by a VN affixed by temporal/aspectual morpheme. 1 3 As will be discussed later, the genitive marker no, instead of the Accusative Case marker, can appear between the VN and its theme, as seen in (i). We will discuss this issue later. (i) John-ga Isuugaku-no benkyoo]-chuu (Japanese) John-NOM Math-GEN study-mid 'while John is studying Math.' 99 By contrast, the transitive verb in gerundive constructions in English can assign verbal Case. (32) a. After investigating the incident... b. Before examining the eyes... While the process nouns in (31) behave as nouns both internally and externally in terms of case assignment, the gerundive forms in (32b) behave as a verb ([+V, -N]), at least internally, and as a noun ([-V, +N]) externally (Baker 1985). Given that the verbal category is involved in assigning Accusative Case, the appearance of the Accusative Case marker is an overt indication that the VN is a category which can assign verbal Case. Furthermore, this indicates that VN is specified for [-N] in that it can assign Case. If only the lexical items that behave as a verb internally can assign Verbal Case, the unitary lexical feature [-V, +N] certainly does not capture the fact that the VN can assign verbal Case. 3.6.3. Adverbial modification Adverbial modification does not tell much about the category of the Verbal VNs. This is because the adverb placed before the verbal VN can be taken as modifying either the light verb or the verbal noun. Consider the following sentences. 100 (33) a. John-i [swuhak-ul yelsimhi kongpwu]-lul ha-ess-ta John-NOM Math.-ACC hard study-ACC do-PAST-DEC 'John studied Math, hard.' b. John-i [swuhak-ul kongpwu]-lul yelsimhi ha-ess-ta John-NOM Math.-ACC study-ACC hard do-PAST-DEC 'John studied Math, hard.' c. John-i yelsimhi swuhak-ul kongpwu-lul ha-ess-ta John-NOM hard Math.-ACC study-ACC do-PAST-DEC 'John studied Math, hard.' In Korean and Japanese, an adverb can modify a verb as long as the former precedes the latter in the same clause. The only difference between (33a), (33b), and (33c) is the place of the adverb. In (33a) the adverb is placed before the verbal noun, while in (33b) it is after the verbal noun. In (33c), the adverb is placed before the complement of the verbal noun. Since an adverb can modify a verb as long as the former precedes the latter in the same clause, it is hard to tell whether the adverb modifies the verbal noun or a light verb. The bracketed part in (33) is not an S, but a VP containing a predicate (VN) and its theme NP in a traditional term. If the bracketed part is an S, the adverb in the embedded clause cannot modify a matrix verb, the sentences in (34) exhibit. (34) a. John-i [ pro ilccik ttena-l-kes]-ul John-NOM early leave-FUT-COMP-ACC yaksokha-ess-ta promise-PAST-DEC 'John promised to leave early.' b. John-i [ pro ttena-l-kes]-ul John-NOM leave-FUT-COMP-ACC ilccik yaksokha-ess-ta early promise-PAST-DEC 'John promised early to leave.' 101 As the English glosses imply the adverbs can be interpreted in the clause they reside in. Even though it is inconclusive whether the adverb modifies the verbal noun or not in a sentence like (33), it is clear that an adverb cannot modify a verbal noun in the environment [NP-GEN VN], as seen in (35). (35) *John-ga [suugaku-no nessin-ni benkyoo]-o site-iru John-NOM Math.-GEN diligently study-ACC doing-is 'John is studying Math hard.' The unacceptable status of (35) further confirms the nominal status of the VN in the environment [NP-GEN VN]. 3.7. Apparent Problems with the verbal approach to verbal VNs. This section demonstrates apparent problems in treating verbal VNs simply as a full-fledged verb. First of all, the VN (both nominal and verbal VN) cannot bear verbal inflections. Verbs ([+V, -N]) in J&K can bear verbal inflections. Second, it cannot assign Accusative Case unless it is followed by a light verb or affixed by a temporal/aspectual morpheme. Furthermore, while [-stative] verbs can assign Accusative Case (Kuno 1973, p349-350), the [-stative] Verbal Noun without an aspectual morpheme or a verb following it cannot assign Accusative Case. 102 3.7.1. No Verbal Inflections While full-fledged verbs can bear a verbal inflectional morpheme as seen in (36), the verbal VN cannot. 14 (36) a. *John-i swuhak-ul KONGPWU-ess-ta (Tense) John-NOM Math.-ACC study-PAST-DEC 'John studied Math.' b. John-i ttena-ess-ta John-NOM leave-PAST-DEC 'John left.' For the sentence in (36a) to be grammatical, the light verb ha has to be added to the Verbal Nouns, as seen in (37). (37) John-i swuhak-ul KONGPWU-ha-ess-ta (Tense) John-NOM Math.-ACC study-do-PAST-DEC 'John studied Math.' The fact that VN cannot bear a verbal inflectional morpheme certainly questions the possibility that the verbal VN is a full-fledged verb. 14 The same state of affairs can be observed in Japanese. Iida (1987) and Miyagawa (1987) note that the verbal suffixes such as the past tense morpheme ta and the propositional suffix reba, among others, can directly attach to the predicates in Japanese as seen in (i) and (ii). The following data are from Iida (1987). (i) tabe-ta eat-PAST (ii) tabe-reba 'tabe-if 'if PRO eat' VNs in Japanese, however, cannot bear these verbal suffixes, as the ill-formed combinations in (iii) and (iv) indicate, (iii) *benkyoo-ta 'studied (iv) *benkyoo-reba 'if PRO studied' 103 3.7.2. Aspect and Verbl Case Assignment The verbal VNs cannot assign Accusative Case unless they are followed by a verb or an aspectual marker (Iida 1987; Miyagawa 1990). This fact contrasts with that of verbs, which do not need an explicit aspectual marker to assign Accusative Case. Consider the following sentences. (38) a. ??John-ga [suugaku-o benkyoo]-o si-ta (J) John-NOM Math-ACC study-ACC do-PAST 'John studied Math.' b. John-ga [suugaku-o benkyoo]-chuu John-NOM Math-ACC study-mid 'while John is studying Math.' In (38a) and (38b), where the VN assigns Accusative Case to its complement, the VN is followed by the light verb su (si in (38a)) or by an aspectual marker chuu 'mid'. The VN, however, cannot assign Accusative Case, if it is in the object position of P (which is not an aspectual morpheme) as seen in (39b) or the VNP (verbal noun phrase) is in the subject position, as seen in (39a). object of P (39) a. *John-ga [suugaku-o benkyoo]-ni tsuite hanasita John-NOM Math.-ACC study]-about talked 'John talked about the study of Math.' subject position b. *[suugaku-o benkyoo]-ga muzukasii Math-ACC study-NOM difficult 'Studying Math, is difficult' 104 Further questions arise if we consider Aspectual Constructions. As seen in (40), if the aspectual marker follows the VN, the complement of the VN is assigned either Accusative or Genitive Case. (40) John-ga suugaku-o/no benkyoo-chuu John-NOM Math.-ACC/GEN study-mid 'while John is studying Math.' This contrasts with the construction where the VN is followed by a non- aspectual morpheme like a postposition, as in (39a), which is repeated here as (41). (41) John-ga [suugaku-*o/no benkyoo]-ni tsuite hanasita John-NOM Math.-ACC/GEN study]-about talked 'John talked about the study of Math.' In (41), the VN followed by the postposition cannot assign Accusative Case. In summary, the verbal VN behaves differently from a full-fledged verb in that it cannot bear inflectional morphology and cannot assign Accusative on its own. 3.7.3. The Nominal Property of verbal VN in [NP-ACC VN] The VP headed by the verbal VN can be assigned Accusative Case, which only nominal expressions are assigned in Japanese. The sentence in (42) illustrates this point. 105 (42) a. ??John-ga [suugaku-o benkyoo]-o si-ta John-NOM Math-ACC study-ACC do-PAST 'John studied Math.' b. John-i [swuhak-ul kongpwu]-lul ha-ess-ta John-NOM Math.-ACC study-ACC do-PAST-DEC 'John studied Math.' Given that Accusative Case is assigned to Nominal expressions, the appearance of the Accusative Case marker on the VP headed by the verbal noun indicates that the VN has a nominal property. 15 In summary, we have seen that verbal VNs exhibit a hybrid nature between nominal and verbal expressions, while nominal VNs behave exactly like a simple noun in respect to distribution. The verbal VN behaves like a verb internally in that it can assign Accusative Case and cannot be modified by an adjective. We also have seen that the verbal VN cannot be a full-fledged verb in that it cannot bear the verbal inflectional morpheme. Further, we have seen that the verbal VN phrase behaves like a NP externally in that it qualifies as a Case assignee. A further complication is observed in that the verbal noun phrase cannot be placed in the object position of a postposition or in subject position. In the next section, we attempt to provide solutions to the problems and questions that have arisen. 15 We will argue later that the Accusative Case is assigned to a null VP-complement governing the verbal VP phrase. 106 3.8. Null VP-Complementizers in Light Verb Constructions In this section, we, following Lee (1991), claim that the VNP is governed by a null complementizer. We propose that the null complementizer in the light verb construction is a VP-complementizer with a [+N] categorial feature. To show this point, we group complementizer in Korean and Japanese into two groups, based on whether they can bear case morphology or not. 16 Given these two types of complementizers, we suggest that the null-Comp in the light verb construction belongs to the group which can bear Case morphology. We find evidence for the null complementizer being a VP-complementizer from the parallelism between the VP-focus construction in J&K on the one hand and the light verb construction on the other. NPI licensing also supports the claim that the null-Comp is a VP-complementizer, not a S-complementizer. In this section, we, following Hoji's suggestion (p.c), develop the hypothesis that a lexical item selects a projection headed by a functional category, not by another lexical category. If this hypothesis is correct, the existence of a null complementizer can be further supported. 3.8.1. [±N] Complementizer Korean and Japanese have two groups of complementizers depending on whether they can bear case morphology or not. We attribute the 16 In this chapter, we provide Korean data only concerning the complementizers. Nonetheless, it is also the case that the discussion in the section holds for Japanese. 107 capability of bearing or not bearing case morphology to the feature specification of the functional category. Fukui (1990; 1994) suggests that functional categories can be specified for the feature value [+V, +N]. Grimshaw (1991) also suggests that functional categories can be specified for [±V, ±N]. Along this line, we suggest that complementizers in J&K can be specified for the feature value [±N] and that this feature specification is responsible for the case morphology attached to the complementizers. One group of complementizers in J&K cannot bear case morphology. These complementizers in Korean include -ki, -key, and -tolok. Consider the following sentences. -ko Complementizer (43) John-i [Mary-ka yepputa-ko]-(*lul) sayngkakha-ess-ta John-NOM Mary-NOM pretty-COMP-ACC thought 'John thought that Mary was pretty.' -key Complementizer (44) John-i [Mary-lul o-key]-(*lul) ha-ess-ta John-NOM Mary-ACC come-COMP-(*ACC) do-PAST-DEC 'John make Mary come.' -tolok Complementizer (45) John-i [Mary-ka o-tolok]-(*lul) pwuthakha-ess-ta John-NOM Mary-ACC come-COMP-(*ACC) ask-PAST-DEC 'John asked Mary to come.' As seen in (43)-(45), if the accusative Case marker lul attaches to these complementizers, the resulting sentences turn out to be ungrammatical. It is also the case that these complementizers cannot bear nominative case morphology either. 108 Another group of complementizers can bear case morphology. These complementizers include -ki/ci, -in, and -kes. The following sentences illustrate this point. 17 -ki complementizer (46) John-un [Mary-ka chayk-ul John-TOP Mary-NOM book-ACC pala-ess-ta hope-PAST-DEC 'John hoped that Mary bought a book.' -in Complementizer (47) John-i [Mary-ka o-ess-m]-ul al-ess-ta John-NOM Mary-ACC come-PAST-COMP-ACC find-PAST-DEC 'John found that Mary came.' -kes Complementizer 18 (48) John-i [Mary-ka o-n-kes]-lul al-ess-ta John-NOM Mary-ACC come-PAST-COMP-ACC find-PAST-DEC 'John found that Mary came.' As seen in (46)-(48), this group of complementizers can bear case morphology. Based on Stowell's (1981) suggestion that a [+N] category is a potential Case assignee, we suggest that the first group of complementizers are specified for [-N], while the second group of complementizers are [+N]. For the purpose of exposition, we refer to the first group of complementizers as [-N] complementizers, and the second group of complementizers as [+N] complementizers. 17 I assume, following Song (1988), that ci and ki complementizers are derived from the same basic form. 18 It is controversial whether kes is a noun or complementizer. Nothing in my claim hinges on this. sa-ess-ki]-lul buy-PAST-COMP-ACC 109 Our proposal that the complementizers are specified for the value [±N] can capture the traditional view or controversy on these two types of complementizers. Kim (1984) claims that [+N] complementizers are sentence nominalizers, not complementizers based on the fact that they can bear case morphology, while [-N] Comps are genuine complementizers. Under our analysis, what Kim (1984) claims to be the property of a nominalizer can be attributed to the feature [+N] on the [+N] complementizers. Given these two groups of complementizers, it is not unreasonable to say that the null complementizer in the Light Verb Construction can be specified for the value [±N]. It is also not unreasonable to say that the null complementizer happens to be in the group of complements with the feature [+N]. As seen in (49) below, the embedded VP under the null-Comp can bear the Accusative Case marker. (49) John-i [swuhak-ul kongpwu-0]-lul ha-ess-ta John-NOM Math.-ACC study-COMP-ACC do-PAST-DEC 'John studied Math.' By positing a null complementizer with the feature [+N], which can bear case morphology, we solve the contradiction of the VN behaving like a verb internally while behaving like a noun externally. In other words, positing a nominal complementizer provides a way to account for the appearance of the Accusative Case marker after the verbal VN. 110 3.8.2. The null-complementizer as a VP-complementizer This section shows the parallelism between the VP-complementizer ki/ci in focus and long-form negation on the one hand and the null- complementizer in the light verb constructions. This section specifically claims that the null-complementizer is a VP-complementizer. Consider the light verb construction in (50). (50) John-i [swuhak-ul kongpwu-0]-lul ha-ess-ta John-NOM Math.-ACC study-COMP-ACC do-PAST-DEC 'Lit. John did a study of Math.' The light verb construction in (50) is parallel to the focus constructions in (51) and the long-form negation construction in (52), which contain a VP- complement ki/ci. (51) a. John-i [ka-(*ess)-ki]-nun ha-ess-ta (Focus) John-NOM go-PAST-COMP-FOC do-PAST-DEC 'John at least went.' b. ??John-i [ka-(*ess)-ki]-lul ha-ess-ta John-NOM go-PAST-COMP-ACC do-PAST-DEC 'Did John come?.' (52) John-i [ka-(*ess)-ci] ani ha-ess-ta (Negation) John-NOM go-PAST-COMP not do-PAST-DEC 'John did not go.' The following is the parallelism between light verb constructions and Focus/Long-form negation constructions. Il l (53) Parallelism between the focus construction and light verb construction (i) the light verb ha follows the embedded VP. (ii) the light verb assigns ACC to the complement VP ((51b) and (50)). 19 (iii) the verb in focus/negation constructions and the VN in the light verb construction assign ACC to their complement. The focus or long-form negation constructions in (51) and (52) are almost identical to the light verb construction under discussion, with the exception that the verbal VN in the light verb construction is not followed by an overt VP-complementizer, but by a null complementizer, as in (50). Further parallelism between VP-focus constructions and the light verb construction in terms of the selectional relations between the light verb and the head of its complement VP/VNP, and the transfer of a Case feature to the light verb from the head of its complement VP/VNP further supports the VP-complementizer analysis for the light verb construction. I suggest another parallelism between the VP-focus and long-form negation constructions in NPI licensing. Before we discuss the parallelism between those two constructions, let us consider NPI licensing across the CP boundary in Korean. l 9 The degree of acceptability between (47) and (51b) has to be accounted for. We will bring up this issue in Chapter 4. 112 (54) ?/*John-i [Mary-ka amukesto ilk-ess-ta-ko] John-NOM Mary-NOM anything read-PAST-DEC-COMP] ani mit-ess-ta not believe-PAST-DEC 'John did not believe that Mary read anything.' As seen in (54), NPI licensing across CP is unacceptable when the NPI is a complement of the embedded verb. 20 In the long-form negative constructions and light verb constructions, NPI licensing across the bracketed embedded phrase is natural. Long-form Negation (55) John-i Mary-lul [amukesto cwu-ci] -nun John-NOM Mary-ACC anything give-COMP-FOC ani ha-ess-ta not do-PAST-DEC 'John at least did not give anything to Mary.' Light Verb Construction (56) John-i Mary-lul [amukesto ceykong] -lul John-NOM Mary-ACC anything give-ACC ani ha-ess-ta not do-PAST-DEC 'John at least did not give anything to Mary.' 20 In (54), We do not test the NPI licensing with the NPI in the subject position of the embedded clause. This is because this NPI can be taken as being placed in the matrix clause (Hoji 1991b; Yoon 1989; Saito 1983), as seen in the construction (i). (i) ?John-i amwuto [chayk-ul ilk-ess-ta-ko] John-NOM anybody book-ACC read-PAST-DEC-COMP ani mit-ess-ta not believe-PAST-DEC 'John did not believe that anyone read the book.' As we see, the acceptability greatly improves over the sentence in (54). 113 Following Hoji (1991b), Yoon (1989), and Saito (1983), we assume that the "major object" Mary in (55) and (56) is in the matrix clause and the NPI is in the embedded clause. Under this assumption, the difference in the acceptability status between (54) on the one hand and (55)-(56) on the other indicates that these two groups of sentences have two different clausal structures. That is, the bracketed part in (54) is a CP-complement while the one in (55) and (55) is a smaller category, namely, a VP- complement. This NPI test further supports a parallel analysis between the VP-focus constructions and the light verb Constructions. 3.8.3. Functional Projections selected by a lexical category In this section, we suggest another motivation for positing the null complementizer. This motivation comes from the hypothesis that the lexical item does not directly select the maximal projection headed by another lexical category. Rather it selects a maximal projection headed by a functional category, as was suggested to me by H. Hoji (p.c.). 21 Under this hypothesis verbs select functional projections like DP, IP, or CP, but not NP or VP. 22 If this is the case, the light verb does not 21 H. Hoji suggested this hypothesis to me (p.c.) without his own commitment to it. H. Hoji also pointed out to me (p.c.) that Miyagawa is independently pursuing this line of approach. 22 G. Longobardi (a lecture at USC in Fall 1994) also suggested that for NPs to be arguments they always have to be governed by the D-projection, overt or non-overt. 114 directly select the VN phrase headed by a VN; rather, it selects a functional projection containing the verbal noun phrase. 23 In fact, the claim that the lexical category does not directly select the projection headed by another lexical category is empirically supported in J&K. It is always the case that a verb in Korean does not directly select a projection headed by a verb. Consider the following sentences. (57) John-i [Mary-ka yepputa]-ko sayngkakha-ess-ta John-NOM Mary-NOM pretty-COMP think-PAST-DEC 'John thought that Mary was pretty.' (58) John-i [ Mary-ka chayk-ul ilk]-key ha-ess-ta John-NOM Mary-NOM book-ACC read-COMP do-PAST-DEC 'John had Mary read a book.' (59) John-i chayk-ul ilk-e po-ess-ta John-NOM book-ACC read-? try-PAST-DEC 'John tried reading a book.' In (57) the main verb sayngkakha 'think' selects the functional projection headed by the /20-complementizer. In (58), the causative verb ha 'do' selects the functional projection headed by the key -complementizer. In (59), the so-called auxiliary verb does not directly select a VP, but the functional projection headed by e. The morpheme e is controversially taken as a complementizer, INFL, or a verb connective. 24 Whichever is correct, the point is that the verb po 'try' cannot directly select a lexical projection, 23 As pointed out to me by H. Hoji (p.a), this hypothesis has an interesting consequence to P, which has been controversial between lexical and functional. If the hypothesis that a lexical category selects a function projection is correct and if PP can be selected by V or N, we are lead to the conclusion that P is a functional category, not a lexical category. Emonds (1985) indeed suggests that P is a functional category. 24 See Chung (1993) for an extensive discussion of this morpheme. 115 but selects a functional projection. Given this generalization, it will not be unreasonable to say that the verb ha in light verb constructions selects a projection headed by a functional element, namely, the null VP- complementizer here. If we extend the hypothesis to the focus constructions, the DO-verb ha or su does not directly select a VP headed by a lexical head, namely, a verb, but rather selects a functional projection. In the VP-focus construction, the functional projection headed by the VP-complementizer hi is selected. Based on the claim that the VNs with the same morphology have two distinct lexical features and that the verbal VN is governed by the verbal complementizer, the following structures can be constructed. 25 (Irrelevant nodes are omitted). 25 See K. Miyagawa (1993) for DP analysis of Japanese NP structures. 116 (60) Spec Spec Spec Spec Nominal VN IP VP DP NP ^ NP "~T I ~~~V' V - ^ ha D' 'do' D N' N i Verbal Noun [-V, +N] (61) Verbal VN IP Spec VP" Spec VP-CP' Spec VNP Spec NP~ r V "VP-C VN' I V ha 'do' VP-C VN Verbal Noun [0V, -N] In (60), the nominal VN has the same projection as a simple noun under the functional projection DP, whereas in (60) the verbal VN is under the VP-complementizer. Thus far, we have discussed the categorial features and structure of light verb constructions, along with the null complementizer. The structure in (61) assumes the constituency of a verbal and its accusative theme. We indeed propose that this is the case. In the following section, we bring up the issue of constituency. 117 3.9. Constituency of VNP In this section, we suggest that the verbal VN and its accusative Theme form a constituent in the embedded clause, contrary to Grimshaw and Mester (1988), who claim in their mono-clausal analysis that the complements of VNs are in the matrix clause. 3.9.1. Focus Constructions It has been observed in Kang (1988) that focus makers like nun or to place focus on any element within the VP in the focus constructions. In other words, the items focused by a focus marker in the VP-focus constructions form a constituent, namely, VP. If this is the case, we can apply the same test to the light verb constructions with a focus marker like nun or to. Before we test the constituency of the VNP in the light verb constructions, let us consider focus constructions in Kang (1988). Consider the following sentences. V-focusing (62) John-i ku chayk-ul SA-ki-nun ha-ess-ciman John-NOM thebook-ACC buy-COMP-FOC do-PAST-but ilk-ci-nun ani ha-ess-ta read-COMP-FOC NEG PAST-DEC 'John bought the book, but (he) did not read (it).' 118 Object NP-focus (63) John-i KUCHAYK-ul ilk-ki-nun ha-ess-ciman John-NOM thebook-ACC read-COMP-FOC do-PAST-but talun chayk-ul ilk-ci-nun ani ha-ess-ta other book-ACC read-COMP-FOC NEG do-PAST-DEC 'John read the book, but (he) did not read other books.' In (62) the verb sa 'buy' is contrastively focused. In (63), the object NP ku chayk 'the book' is contrastively focused. Kang (1988) also observes that the subject NP cannot be focused. (64) *JOHN-i ku chayk-ul ilk-ki-nun ha-ess-ciman John-NOM thebook-ACC read-COMP-FOC do-PAST-but Mary-ka ilk-ci-nun ani ha-ess-ta Mary-NOM read-COMP-FOC NEG do-PAST-DEC 'John read the book, but Mary did not read it.' If Kang's (1988) observation is correct, the same can be said of the light verb constructions where the focus marker is attached to the VN. Consider the following sentences. VN-focus (65) John-i kusaken-ul COSA-nun ha-ess-ciman John-NOM that affair-ACC investigation-FOC do-PAST-but poko-nun ani ha-ess-ta report-FOC not do-PAST-DEC 'John investigate the affair but he did not report (it).' Object NP-focus (66) John-i KUSAKEN-ul COSA-nun ha-ess-ciman John-NOM that affair-ACC investigation do-PAST-but talun saken-ul cosa-nun ani ha-ess-ta other affair-ACC investigation-FOC not do-PAST-DEC 'John investigate the affair but he did not investigate the other affairs' 119 As seen in (65) and (66), the VN and its complement can respectively be focused, indicating that the VNP has the same structure as the focus constructions. That is, the VN and its complement are under the VNP just as the verb and its object are under VP in the focus constructions. 3.9.2. VP-fronting Constructions VP-fronting provides another way of testing constituency. In the VP-focus constructions, while the complement of the embedded verb can be fronted, the embedded verb cannot. According to Hoji (1989), the impossibility of fronting the verb is attributed to the constituency of the embedded verb and its complement. (67) a. John-i chayk-ul ilk-ki-nun ha-ess-ta John-NOM book-ACC read-COMP-FOC do-PAST-DEC 'John at least read the book.' b. chayk-ulj John-i tj ilk-ki-nun ha-ess-ta book-ACC John-NOM read-COMP-FOC do-PAST-DEC 'The book, John at least read.' c. *ilk-ki-nunj John-i chayk-ul tj ha-ess-ta read-COMP-FOC John-NOM book-ACC do-PAST-DEC 'Read the book, John at least did.' In Hoji (1989), the ungrammaticality of (67c) is attributed to a violation of the Proper Binding Requirement. The sentence in (67c) has the following structure. 120 (68) *[ ti ilk-ki-nun]j John-i chayk-ulj tj ha-ess-ta read-COMP-FOC John-NOM book-ACC do-PAST-DEC 'Read the book, John at least did.' In (68), the trace of chayk 'book' is not properly bound, and therefore, results in an ungrammatical status. This test indicates that the verb and its complement form a constituent. The same test indicates that the VN and its complement form a constituent. (69) a. John-i swuhak-ul kongpwu-lul ha-ess-ta John-NOM Math.-ACC study-ACC do-PAST-DEC 'Lit. John did a study of math.' b. *kongpwu-lulj John-i swuhak-ul tj ha-ess-ta study-ACC John-NOM Math.-ACC do-PAST-DEC 'Lit. Study, John did Math.' As seen in (69b), the fronting of the VN yields an ungrammatical status. This ungrammatical status can be attributed to the same reason as the one in (68). That is, the VN and its complement form a constituent. 3.10. Predictions This section shows that the case alternation results from two different verbal nouns in aspectual structure and that the presence of only nominal VN, not verbal VN, in fo-constructions allows either the presence or absence of nominal Case on to-clauses. 121 3.10.1. Aspectual Constructions This section suggests that the case alternation between GEN/ACC in aspectual structures results from two different verbal nouns. Given that Accusative Case is licensed by Aspect, another advantage of the proposed analysis over the unitary categorical feature hypothesis can be found in Aspectual Constructions containing an Aspectual morpheme like go 'after', as in (70). (70) Taroo-ga Hanako-no ronbun-no/o hihan-go-ni... Taro-NOM Hanako-GEN Thesis-GEN/ACC criticism-after at 'After Taroo criticized Hanako's paper,...' In (70), the unitary lexical feature hypothesis has to make a stipulation that Aspect optionally licenses the Accusative Case assigned by VN or that Aspect is optionally assigned to the VN by the Aspect holder (Sells 1990). Our proposal correctly predicts the alternation between verbal and nominal Case as assigned by VN. That is, in the proposed system, the Genitive Case marker no is possible when the VN is specified for [-V, +N]. The Accusative Case marker is possible if the VN is specified for [0V, -N] and if the aspectual marker provides Aspect to the VN. Let us now consider the adjectival modification of VN. First consider the case where the complement of VN is followed by a genitive case marker, as in (71). We have suggested that the VNs in the environment [Theme-ACC VN]-ACC are verbal VNs. Since the VN in (71) is not in this 122 environment, the VN is taken as a nominal VN. This is borne out as seen by the adjectival modification in (71). (71) Taroo-ga [Hanako-no ronbun-no kibisii bihan]-go-ni... Taro-NOM Hanako-GEN Thesis-GEN severe criticism-after at 'After Taroo's severe criticism of Hanako's paper,...' Furthermore, an adverb cannot be placed in the NP, as seen in (72). (72) Taroo-ga [Hanako-no ronbun-no Taro-NOM Hanako-GEN Thesis-GEN (*kibisiku) hihan]-go-ni... severely criticism-after at 'After Taroo severely criticized Hanako's paper,...' Under the current analysis, if the VN takes an Accusative-marked Theme, we expect that the VN is a verbal VN. This is the case as the sentence in (73) demonstrates. (73) Taroo-ga [Hanako-no ronbun-o Taro-NOM Hanako-GEN Thesis-ACC ( :,: Mbisii) hihan-go-ni... severe criticism-after at 'After Taroo severely criticized Hanako's paper,...' In (73), as expected, the adjective kibisii 'severe' cannot modify the VN. So far, the adjectival test confirms the hypothesis of two distinct lexical features. This hypothesis, however, does not seem to account for the appearance of an adverb in a sentence like (74). 123 (74) Taroo-ga kibisiku Taro-NOM severely [Hanako-no ronbun-no hihan]-go-ni... Hanako-GEN Thesis-GEN criticism-after at 'After Taroo severely criticized Hanako's paper,...' In (74), even though the adverb kibisiku 'severely' co-occurs with the nominal VN, not verbal VN, the sentence is acceptable as opposed to (72). The only difference between (72) and (74) is the location of the adverb kibisiku 'severely'. I attribute the (un)grammaticality of (72) and (74) to the absence or presence of Aspect in the clause where the adverb appears. Namely, in (72), the NP containing the VN and adverb does not have Aspect to license the adverb. By contrast, in (74) the adverb kibisiku 'severely' is outside of NP and it is licensed by Aspect go 'after'. Following Miyagawa (1990), we assume that the Aspectual morphemes like go 'after' heads the Aspect phrase. Under this assumption, the adverb kibisiku 'severely' is under this Aspect Projection. 3.10.2. to-Clause and lexical specification of VNs The current proposal presents a simple analysis for the light verb construction where the VN takes a fo-complementizer in Japanese. As discussed above, the complementizer ko in Korean and to in Japanese cannot bear an accusative Case marker. The genitive case maker no in Japanese, however, can follow a CP headed by to. In Japanese, PPs and NPs inside an NP are typically followed by the Genitive Case maker no (Grimshaw and Mester 1988). 124 Therefore, in (75) below, the complement of the VN benkyoo, namely, suugaku 'Math.', is in the bracketed NP. (75) John-ga [NP suugaku-no benkyoo]-o si-ta John-NOM Math-GEN study-ACC do-PAST 'John studied Math.' A clausal complement can also be followed by the Genitive marker no, as seen in (76). (76) a. Taroo-ga murabito-ni [NP [ookami-ga kuru to]-no Taro-NOM villager-DAT wolf-NOM come COMP-GEN keikoku]-o si-ta warning-ACC do-PAST 'Taro warned the villagers that the wolf was coming.' b. Taroo-ga murabito-ni [ookami-ga kuru to] Taro-NOM villager-DAT wolf-NOM come COMP keikoku-o si-ta warning-ACC do-PAST 'Taro warned the villagers that the wolf was coming.' Given that the genitive case no is inserted in the environment [NP/PP.... N(P)], the appearance of no in (76a) indicates that the to-clause is in the bracketed NP. Then, we may expect that the VN is a nominal VN, but not a verbal VN. In fact, an adverbial cannot be placed before the VN in (76a), as the sentence in (77a) demonstrates. 26 (77) a. :|: Taroo-ga [ookami-ga kuru]-to-no osorosiku Taro-NOM wolf-NOM come-COMP-GEN scarily keikoku-o si-ta. warning-ACC do-PAST 'Taro scarily warned the villagers that the wolf was coming.' 26 I thank K. Miyagawa (p.c.) for examples and discussion. 125 b. Taroo-ga [ookami-ga kuru to]-no Taro-NOM wolf-NOM come COMP-GEN osorosii keikoku-o sita scary warning-ACC did 'Taro gave a scary warning that the wolf was coming.' t Furthermore, the adjective modifies the VN, as in (77b), confirming the nominal feature of the VN. Let us now consider the case where the clause is not followed by the genitive case marker. Consider the sentence in (78). (78) Taroo-ga [ookami-ga kuru to] keikoku-o sita Taro-NOM wolf-NOM come COMP warning-ACC did 'Taro gave a warning that the wolf was coming.' In (78), the bracketed clause is not followed by either the genitive case marker, or by the accusative case marker. Given our claim that VNs in the environment [NP-ACC VN]-(ACC) are taken as verbal VNs with the feature [0V, -N], it is predicted that the VN in (78) is a nominal VN. This prediction is borne out as seen in (79). (79) Taroo-ga [ookami-ga kuru to] Taro-NOM wolf-NOM come COMP osorosii keikoku-o sita scary warning-ACC did 'Taro gave a scary warning that the wolf was coming.' In (79), the adjective can modify the VN, confirming that our analysis is correct. 126 We further claim that the fo-clause is in the matrix clause. The structure in (78) gives us three possibilities. The first possibility is that the bracketed to-clause is in the matrix clause. The second possibility is that the £o-clause is in the embedded clause as an adjunct 27 . The third possibility is that the to-clause is in the embedded clause as a complement of keikoku 'warning' and is actually assigned abstract Accusative, and the abstract Case is not morphologically realized. We rule out the third possibility for the reason that the to-clause can appear in subject VNPs under the assumption that it is a complement of the VN. Accusative theme NPs cannot appear in the subject VNP, as seen in (80). (80) *[swuhak-ul kongpwu]-ga elyep-ta Math-ACC study-NOM difficult-DEC 'Studying Math, is difficult' The ungrammaticality of (80) is attributed to the VNs not being able to assign Accusative Case to its complement swuhak 'Math.' The VN is not involved in the Chain containing the Aspect, and therefore, the VN cannot assign Accusative Case. Let us now consider the sentence containing the to-clause. (81) [ookami-ga kuruto] keikoku-ga arimasita wolf-NOM come-COMP warning-NOM was 'There was a warning that the wolf was coming.' 27 Terada (1990) claims that the <o-clause is an adjunct. See Hasegawa (1991) for a complement analysis of the <o-clause. 127 As opposed to (80), the sentence in (81) is grammatical, indicating that the to-clause is not assigned Abstract Accusative Case by the VN keikoku 'warning'. The grammaticality of (81) leads us to suspect that the bracketed to-clause is in the matrix sentence. K. Miyagawa in fact pointed out to me (p.c.) tha t the insertion of a time adverbial between the to-clause and the VN supports this claim, as seen in (82). (82) [ookami-ga kuru to] kinoo keikoku-ga arimasita wolf-NOM come-COMP yesterday warning-NOM was 'Yesterday there was a warning that the wolf was coming.' We are now left with the first and second possibilities. The first and second possibilities have the structures in (84) and (83), respectively. (83) Taroo-ga [[ookami-ga kuru to] keikoku]-o sita Taro-NOM wolf-NOM come COMP warning-ACC did 'Taro gave a warning that the wolf was coming.' (84) Taroo-ga [ookami-ga kuru to] [NP keikoku]-o sita Taro-NOM wolf-NOM come COMP warning-ACC did 'Taro gave a scary warning that the wolf was coming.' The sentence in (83) cannot be correct. This is because an adjunct cannot be licensed by an nominal expression such as a nominal VN. Furthermore, an adverbial can appear before the VN, as seen in (85). (85) Taroo-ga [ookami-ga kuru to] Taro-NOM wolf-NOM come COMP osorosiku keikoku-o sita scarily warning-ACC did 'Taro scarily gave a warning that the wolf was coming.' 128 If (84) is a correct structure, the VN must be a nominal VN. This is because the VN does not have an Accusative Theme. Recall that we have claimed that the VN in the environment of [NP-ACC VN-]-(ACC) is a verbal VN being specified for [0V, -N]. 28 In other contexts, they must be nominal VNs with the feature [-V, +N]. As mentioned before, as long as an adverbial precedes its modifiee in a clause, the former can modify the latter. 29 Therefore, logically speaking, in (85), the adverbial osorosiku 'scarily' can modify either the VN or the light verb sita. Considering that an adjective cannot modify the VN in (79), the adverb in (85) modifies the light verb, not the VN. 30 Thus far, we have seen that the suggestion for [Theme-ACC VN]- ACC makes a correct prediction for the fo-clause, too. 3.11. Conclusion In summary, we have proposed two distinctive features for the VNs: one with feature [-V, +N], another with the feature [0V, -N]. Under this proposal, the so-called hybrid nature of the VNs, nominal and verbal, can be attributed to these two different syntactic features of VNs along with a null complementizer with the feature [+N] after the VN. Now, all the 28 The NP in the [NP-ACC VN1-ACC can be reinterpreted as Theme-ACC. As long as the theme is assigned ACC, the VN can be interpreted as a verbal VN. 29 We assume that adverb can modify the modifiee across the VP-complement, but not across an IP or CP boundary. 30 Since an adverb cannot modify a semantically empty verb, the light verb is not semantically empty at the level that the adverb modifies the light verb. Since the VN incorporates into the light verb at LF, the level that the adverb modifies the light verb might be LF. 129 syntactic differences between these two types of VNs. (i.e. verbal vs. nominal Case assignment by VN) follow from this categorial distinction, instead of an optional incorporation of the VN to the light verb (Kageyama 1982; Terada 1990). By defining the verbal VNs as non-nominal categories, we open the way to provide a parallel account for both focus constructions and light verb constructions. These constructions can further receive a parallel account with the gerundive construction in English. The study of the light verb construction leads us to the following question in the general theory of grammar: whether it is universal that the lexical category does not directly select another lexical category. If it is universal, then from the perspective of the current proposal the VP-shell in Larson (1988) should be recasted, so that the upper verb selects the functional projection containing the lower VP. In this respect, the functional projection PrP in Bowers (1993) might be on the right tract for the small clause in that the verb selects the functional projection PrP, not the lexical projection. 130 Chapter IV. LF-Incorporation 4.1. Introduction In this chapter, we provide a unified account for the focus constructions and light verb constructions in terms of the choice of DO in focus constructions and light verbs in light verb constructions, the realization of the aspectual marker, and reflexive Case. Based on Cho (1993b) and Hoshi and Saito (1993), we develop the LF-incorporation hypothesis for verbal nouns in which the VN incorporates into the light verb. We propose that the LF-incorporation of theta-assigning items to the light verb or DO is motivated to satisfy selectional restrictions. In English, the deverbal nouns co-occur with some light verbs, but not with other light verbs (Cattell 1984), as seen in (1) and (2). (1) a. Harry demonstrated the new technique to the class. b. Harry gave/*made a demonstration of the new technique to the class. (2) a. Bill confessed his crimes to the police b. Bill :|: gave/made a confession of his crimes to the police. In (1) and (2), give and make are used as light verbs. The choice of the light verb, however, is not free. The lexical item demonstration selects give, not make, while the lexical item confession selects make, not give. 131 We further extend the LF-incorporation hypothesis to the focus/negation constructions in Japanese and Korean. We specifically propose that the verbal noun ([0V, -N]) in light verb constructions and the embedded verb in the focus/negation constructions move to DO and light verbs, respectively, at LF. This LF-movement, I will argue, is responsible for (i) the choice of ar and su in Japanese; (ii) reflexive Case (Kang 1986; Miyagawa 1989; Kageyama 1991) in Korean and Japanese; (iii) and theta- role transfer (Grimshaw and Mester 1988; Hoshi and Saito 1993). This chapter is organized as follows. In section 4.2, we briefly review the analysis of Hoshi and Saito's (1993) LF-incorporation hypothesis in relation to theta-role assignment. In section 4.3 we present our proposal in support of LF-incorporation. In section 4.4, section 4.5, section 4.6, and section 4.7, we support our proposals with theta-grid and aspectual grid transfer, and with sectional restriction. Section 4.8 concludes this chapter. 4.2. Hoshi and Saito (1993): LF-incorporation In light verb constructions, some argument of the verbal noun appears in the projection of the verbal noun and some others appear outside the projection of the verbal noun as clausal arguments. The following sentences are from Hoshi and Saito (1993) with some modification in the gloss. 132 (3) a. Honda-ga amerika-de Honda-NOM USA-in [NP akoodo-no seisan]-o site-iru Accords-GEN production-ACC doing-is 'Honda is producing Accords in the U.S.' b. Mary-ga John-ni/-e [Nptoti-no zyooto]-o sita Mary-NOM John-to/-to land-GEN giving-ACC did 'Mary gave a piece of land to John.' In (3a), the theme of the verbal noun seisan 'production', namely akoodo 'Accord', is in the projection of the verbal noun, as indicated by the presence of the genitive marker no after akoodo 'Accord'. By contrast, the agent Honda 'Honda' and the locative phrase amerika-de 'in the U.S.A.' are outside of the projection of the VN, as indicated by the absence of the genitive marker no after these phrases. Based on the assumption that the light verb su (si in (3)) is semantically empty and that all the thematic roles originate from the verbal nouns, Grimshaw and Mester (1988) claim that clausal arguments like honda and amerika in (3a) are assigned theta-roles by the light verb to which the same theta-roles have been transferred from the verbal noun. Grimshaw and Mester (1988) state that the theta-role transfer is not arbitrary, but constrained by the following rules in the light verb constructions (Hoshi and Saito 1993; their (7)) (4) a. At least one internal theta-role of the noun must be assigned to an argument outside the NP. b. If a theta-role T is assigned outside the NP, then all theta-roles that are higher than T in the thematic hierarchy must also be assigned outside the NP. 133 In (5), however, no internal theta-role (in the sense of Williams 1981) of the verbal noun is assigned to an argument outside the NP. (5) John-ga [NP suugaku-no benkyoo]-o si-ta John-NOM Math-GEN study-ACC do-PAST 'John studied Math.' In accordance with (4a), Grimshaw and Mester (1988) regard constructions like (5) as heavy verb constructions, not light verb constructions. Namely, the verb su (si in (5)) is taken as a heavy verb. Hoshi and Saito (1993) further point out, supporting G&M's view in (5), that the sentence in (5) does not have a neutral interpretation as the light verb construction has. The interpretation for the sentence in (5) is claimed to be 'Mary did one specific action, which is 'studying Math.'. (4b) implies that if the lowest theta-role in the thematic hierarchy transfers to the light verb from the verbal noun, all other thematic roles have to transfer to the light verb. Therefore, as seen in (6) below, if the theme NP akoodo 'Accord', which is the lowest in the thematic hierarchy (Higginbotham 1985), appears as a clausal argument of the light verb, all other arguments must appear at the clausal level. (6) ??Honda-ga amerika-de akoodo-o seisan-o Honda-NOM USA-in Accords-ACC production-ACC site-iru doing-is 'Honda is producing Accords in the U.S.A.' 134 The marginal status of (6) is attributed to the surface double -o constraint, (Harada 1973; Poser 1981), which states that a simple sentence cannot have more than one o-marked phrase. The phrase in (4b) predicts that the following sentences are all ungrammatical. The following sentences are from Hoshi and Saito (1993). (7) a. *Honda-ga akoodo-o [amerika-de-no seisan]-o Honda-NOM Accords-ACC USA-in production-ACC site-iru doing-is b. *amerika-de akoodo-o [Honda-no seisan]-o USA-in Accords-ACC Honda-GEN production-ACC site-iru doing-is c. *akoodo-o [Honda-no amerika-de-o seisan]-o Honda-ACC Honda-GEN USA-in production-ACC site-iru doing-is 'Honda is producing Accords in the U.S.A.' All the sentences in (7) are in violation of (4b) in that the theta-roles higher than the thematic NP in the thematic hierarchy appear inside the NP. Hoshi and Saito (1993) capture the theta-role transfer in (4) discussed in Grimshaw and Mester (1988) by assuming that the theta-roles are discharged in a bottom-to-top fashion. In other words, the lowest theta- roles in the thematic hierarchy are discharged either by the verbal noun or by the incorporated VN+light verb at LF, and the other theta-roles are discharged top-to-bottom according to the thematic hierarchy. In Hoshi 135 and Saito (1993), the LF-incorporation of the LV to the light verb is motivated by a need to assign theta-roles at the clausal level. Hoshi and Saito (1993) assume, following Zubizarreta (1985) and Grimshaw (1990), that the subject theta-role of a noun is suppressed; hence, the sentences like the following are not light verb constructions, but a heavy verb construction 1 (8) ?Mary-ga [NP John-e -no toti-no zyooto]-o sita Mary-NOM John-to-GEN land-GEN giving-ACC did 'Mary gave a piece of land to John.' Since the subject theta-role of the noun zyooto 'giving' is suppressed, Mary does not receive a theta-role by the noun zyooto 'giving', but by the heavy verb su. Since the VN does not need to assign a theta-role at the clausal level, the LF-incorporation of the VN to su (si in (8)) is not motivated. 4.3. Proposal We propose that a lexical item with a saturated feature grid forms a chain with DO or light verbs at LF. This LF chain can be formed by the incorporation of the thematic lexical item to DO or light verb at LF (cf. Cho 1993b). 2 The movement in question is X° movement which obeys the Head 1 See, however, Terada (1990), who suggests that the subject of theta-assigning nouns are not suppressed. 2 H. Aoyagi (1993) also discusses that the movement of a honorific verb to the light verb in Japanese. 136 Movement Constraint (Travis 1984). 3 We also propose that the LF head- movement is motivated to satisfy the selectional relation between the thematic X° items and DO or light verb at LF. We assume that verbs are specified not only for theta-grid (Stowell 1981), which yields a classification of verbs in terms of transitive and intransitive, but also for a temporal and aspectual grid (Higginbotham 1985), which yields basic verb classes: stative, activities, achievements. If this is the case, DO and light verbs have empty theta-grids, temporal grids, and aspectual grids. We also suggest that the only feature that DO and light verbs have is the [+V] categorial feature. If we assume that the thematic lexical item X° forms a chain DO or light verbs, then theta-role, Case, and aspect can be discharged through this chain, supporting Hoshi and Saito (1993). We have seen in the previous section that the theta-roles of the VN can move to DO or light verbs. Based on the discussion of the focus constructions and the light verb constructions, I assume the structure in (9) for focus and the long-form negation constructions. I also assume a parallel structure (10) for light verb constructions. In these structures, the LF-movement of the thematic X° to the semantically empty verbs can be diagrammed as follows (irrelevant parts are omitted): ^HMC (Head Movement Constraint) Movement of a zero-level category B is restricted to the position of a head a that governs the maximal projection y of C, where 9-governs or L-marks y if a*C. 137 (9) (10) VP VP thematic verb verbal noun In (9), the thematic verb or adjective undergoes head movement to the head of FP and moves further up to DO. In (10), the verbal noun moves up to the ligth verb at LF. In the light verb and focus constructions, the aspectual feature appears on the upper verb, DO or light verb. This indicates that the aspectual feature on the tail of the chain is discharged on the head of the chain. The current approach can also account for the alternation between the two different DOs, su and ar, depending on the head of the lower VP or of the VNP (Verbal Noun Phrase). If we take this alternation as a selectional relation, the selection of a r and su is hierarchically reversed. The DO-verbs are selected by the hierarchically lower verb. This reversal in hierarchy is not restricted to the languages in questions. As pointed out by J.-R. Vergnaud (p.c), the English light verb constructions also show a hierarchical reversal between the selector and selectee. As Cattell (1984: 138 6) listed, each deverbal noun takes its specific light verb, as in (11) and (12). (11) GIVE: bark, belch, chuckle, cough, cry, gasp, growl, guffaw, gulp, hiccup, howl, moan, quiver, screech, shiver, shudder, sigh, snare, sneeze, sniff, snort, sob, squeak, squeal, twitch, yawn... (12) MAKE: appeal, approach, attempt, boast, bolt, bow, dash, demand, endeavor, grab, grimace, guess, joke, journey, leap, lunge, move, protest, remark, request, rush, start, try, turn. Let us consider the following examples for a clear illustration. (13) a. Nancy attempted to break the record. b. Nancy made/*gave an attempt to break the record. (14) a. Pat kicked the dog. b. Pat *made/gave the dog a kick. In (13b) and (14b), the deverbal noun attempt and kick, co-occur with the light verb make and give, respectively, not with any other light verb. The selectional relation between deverbal nouns and light verbs can be accommodated in our proposal. Under the current proposal, the deverbal nouns undergoes head-movement at LF to the light verb to check the selectional requirement. The same account can be applied to the selection 139 of DO or light verbs in Japanese. The head of the VP, AP, or VNP undergoes head-movement to DO or light verb to check the selectional requirement at LF. 4 It has to be clarified that DO or light verbs are not copies of the categorial feature of the thematic lexical items. This is because ar and su are in the same category in both focus and light verb constructions. Furthermore, since nominal VNs and verbal VNs in Japanese are specified for [-V, +N] and [0V, -N], respectively, and they select a light verb su, then the light verbs with the feature [+V] cannot be copies of the categorial feature of the VNs. The same LF-movement analysis of the thematic X° to DO or light verbs can account for the assignment of the Accusative Case by DO or light verbs. Given our LF-movement analysis for the focus and light verb constructions, we need only one stipulation. The stipulation is that the Case feature of the thematic X° can be shared by both the head and the tail of the chain. Therefore, the head of the chain, that is, DO or light verbs, has the same case features that the tail of the chain has. If our analysis is correct, Japanese is also expected to show the same phenomena in Case assignment by the light verb. The only difference between Korean and Japanese is that Japanese has the so-called double o constraint (Harada 1973; 1981; 1988), which states that a simple sentence cannot have more than one o-marked phrase. Therefore, it is expected that a Japanese sentence may have a double Accusative construction with a light verb with 4 I thank H. Hoji, who pointed out this selectional relation and J.-R. Vergnaud, who suggested the LF-movement analysis for this relation. 140 less acceptable status than its Korean counterpart. This prediction is born out as Sells (1990), Isoda (1991), and Hoshi & Saito (1993) report. The following sentences are from Hoshi and Saito (1993) (their 25). (15) a. ??Honda-ga amerika-de akoodo-o seisan-o Honda-NOM America-in Accord-ACC production-ACC site-iru doing-is 'Honda is making Accords in the U.S.A.' b. ??Mary-ga John-i/e toti-o zyooto-o sita Mary-NOM John-to land-ACC giving-ACC did 'Mary gave a piece of land to John.' As noted in Hoshi and Saito (1993), the sentences in (15) are in violation of the surface double -o constraint, which makes these sentences marginal. Since Korean does not exhibit any double -o constraint, the Korean counterpart of (15) is expected to be fully acceptable, which is the case. We also propose that the LF-incorporation hypothesis in conjunction with Accusative licensing by Aspect can also account for the distribution of VNP. Since the VNP does not contain Aspect which can license the Accusative, the distribution of VNP is restricted to places where the VN can form a chain with Aspect. Thus far, we have seen that an LF-movement analysis of the thematic X° to DO or light verbs provides a straightforward account for the state of affairs concerning the selectional relation between the theta-verb and DO or light verbs, the so-called transfer of theta-roles and Aspectual feature, and reflexive Case. In the following sections, we provide more discussion of data for the current proposals. 141 4.4. Theta-grid We assume that Hoshi and Saito (1993) are on the right track in that LF-incorporation is involved in the assignment of theta-roles. We, however, assume that the accusative theme is discharged in the projection of th e verbal noun, as the bracketing indicates. 5 (16) ??Honda-ga [akoodo-o seisan]-o site-iru Honda-NOM Accords-ACC production-ACC doing-is 'Honda is producing Accords.' We have argued for this constituency in Chapter 3, based on VP-focus and the Proper Binding Principle. In the structure in (16), the subject theta- role of the non-nominal VN is not suppressed as opposed to that of the nominal VN in (17). If this is correct, the subject theta-role is discharged at LF after the LF-incorporation of the VN to the light verb. (17) ??Honda-ga [akoodo-no seisan]-o site-iru Honda-NOM Accords-GEN production-ACC doing-is 'Honda is producing Accords.' Recall that Zubizarreta (1985) and Grimshaw (1990) claim that the subject theta-role of a noun is suppressed. Given that the subject theta-role is not suppressed in (16), we suggest that the subject theta-role is discharged at LF due to the general requirement that INFL forces the presence of a subject in the clause, 5 Huang (1994) proposes that the subject of the lower VP is pro, which is co-indexed with the matrix subject. 142 under the assumption that INFL assigns Nominative Case in Korean (1988; Han 1987; Yang 1990) and Japanese (Takezawa 1987). Formulating Shibatani's (1978) generalization in (18), Takezawa (1987) claims that having INFL as a Case assigner has as a consequence tha,t a nominative subject must be present in a sentence. (18) An independent sentence in Japanese requires at least one nominative NP. (cited from Takezawa 1987: 68) Then the VN is forced to transfer a theta-role (the highest one in the thematic hierarchy) to the light verb. Under this hypothesis, Grimshaw's generalization in (19) can be nicely captured. (19) The subject argument must always be outside of NP (Grimshaw andMester 1988:215) The generalization in (19) can rule out a sentence like (20), where the subject is inside the NP. (20) *[NP John-no suugaku-no benkyoo]-o si-ta John-GEN Math-GEN study-ACC do-PAST 'John studied Math.' We claim that the same holds for focus constructions. In the focus construction, the theme NP is in the projection of the lower verb, while the subject is outside of it. 143 (21) John-i [chayk-ul ilk-ki]-nun ha-ess-ta John-NOM book-ACC read-COMP-FOC do-PAST-DEC 'John at least read a book.' In (21), the subject theta-role is discharged after the lower verb moves to DO at LF. Some remarks are needed concerning verb reduplication with regard to the LF-incorporation discussed in Kang (1988). It has been observed in Lee (1972) and Yang (1976) that the thematic verb can be reduplicated in DO positions in focus constructions. Based on this fact, Kang (1988) claims that thematic verbs in Korean can move to DO at S-structure and that this is evidence for verb movement to DO at LF. The S-structure reduplication in focus constructions is illustrated in (22). (22) a. Mary-ka yeppu-ki-nun ha-ta Mary-NOM pretty-COMP-FOC do-DEC 'Mary is pretty.' b. Mary-ka yeppu-ki-nun yeppu-ta Mary-NOM pretty-COMP-FOC pretty-DEC 'Lit. Pretty, Mary is pretty.' Both of these sentences are focus constructions. The only difference between these two sentences is that ha in (22a) is replaced by the thematic verb in (22b). Given that (22b) is a thematic paraphrase of (22a), Kang (1988) claims that the thematic verb has moved to DO at S- structure. Kang (1988) further claims that this pair of sentences can substantiate his claim that the thematic verb moves to DO at LF. Kang's (1988) generalization, however, cannot capture the whole paradigm. The 144 negative constructions do not fall under this hypothesis, as pointed out in Song (1988: 94). (23) a. Mary-ka ani yeppu-ci-nun ani ha-ta Mary-NOM not pretty-COMP-FOC notdo-DEC 'It is not the case that Mary is not pretty.' b. Mary-ka ani yeppu-ki-nun ani yeppu-ta Mary-NOM not pretty-COMP-FOC not pretty-DEC 'Not pretty, Mary is not pretty.' '=Mary is not pretty.' In (23b), yeppu 'pretty' is duplicated. The minimal difference between (23a) and (23b) is that in (23b), ha in (23a) is replaced by the thematic verb, yeppu 'pretty'. For this pair to be evidence supporting his claim, these two sentences must have the same truth value, which is not the case as the glosses in (23) indicate. In turn, this pair does not tell much about the LF- movement of the thematic verb to DO. All it reveals is that the sentence in (23b) is not derived from (23a). We assume that reduplication is a copy process to the empty verb position in the VP-shell at S-structure (cf. Chomsky 1992). Given the incorporation hypothesis of the negative adverb ani into the main verb (Cho 1993a), the incorporated Neg and the stative verb yeppu 'pretty' undergo a copy process to the empty verb in the higher VP. By contrast, the sentence in (23a) does not involve any copy process. The thematic verb raises at LF to DO, which is base-generated, to discharge its subject theta-role. As mentioned before this LF-movement is forced by the requirement that INFL has to discharge its nominative Case (Takezawa 145 1987). This LF-incorporation hypothesis can be further supported by the transfer of an aspectual feature, which will be discussed in the next section. 4.5. Aspectual Feature Transfer The aspectual feature of the embedded clause in the focus construction has an effect on the choice of the matrix aspect. Before we examine the focus constructions with respect to Aspect, let us consider the following pair of sentences. (24) a. John-i yongkamha-(*N)-ta John-NOM brave-(ASP)-DEC 'John is brave.' b. John-i o-(N)-ta John-NOM come-(ASP)-DEC 'John came.' As seen in (24a), the Korean stative verbs cannot co-occur with an aspectual marker. Recall that the Korean counterpart of English adjectives are classified as stative verbs in Korean. Since yongkamha 'brave' in (24a) is a stative verb, the sentence in (24a) turns out to be ungrammatical with the aspectual marker nun. By contrast, the sentence in (24) which employs the non-stative verb, is grammatical with the aspectual marker. In focus constructions, as in (25) below, and long-form negation constructions, as in (26) below, the presence or absence of the aspectual marker depends upon the aspectual feature of the embedded verbs. 146 (25) a. John-i yongkamha-ki-nun ha-(*N)-ta John-NOM brave-COMP-FOC do-(ASP)-DEC 'John is at least brave.' b. John-i o-ki-nun John-NOM come-COMP-FOC 'John at least comes.' (26) a. John-i yongkamha-ci John-NOM brave-COMP 'John is not brave.' b. John-i o-ci John-NOM come-COMP 'John does not come.' ha-(N)-ta do-(ASP)-DEC ani ha-(*N)-ta not do-(ASP)-DEC ani ha-(N)-ta not do-(ASP)-DEC Given that only non-stative predicates can have the aspectual marker, we claim that the feature [-stativity] of the lower predicate has transferred to DO. The same can be observed in the light verb constructions. Korean has two types of Verbal Nouns: one with [-stative]; another with [+stative] (See Suh 1975 for this observation). The nonstative Verbal Nouns include kongpwu 'study', cosa 'investigation', etc., while the stative Verbal Nouns include phikon 'tiredness', kenkang 'healthy', etc. Given this distinction, we expect that the same observation can be made for the Light Verb Constructions. Consider the following pair of sentences: (27) John-i kongpwu-lul John-NOM study-ACC 'John is studying.' ha-(N)-ta do-ASP-DEC 147 (28) John-i phikon ha-(*N)-ta John-NOM tiredness HA-(*ASP)-DEC 'John is tired.' As seen in (27) and (28), the stativity of the verbal nouns has carried over to the following Light Verbs, which is exactly the same pattern we have observed for the focus and long-form negation constructions in (25) and (26). The transfer of Aspect from the embedded verb to the upper light verbs can be captured by the LF-incorporation hypothesis. Following Higginbotham (1985), we assume that verbs are specified for an aspectual grid. Then, the thematic verbs or verbal nouns in the lower VP in focus constructions or light verb constructions move to DO or the light verb in the upper VP at LF and discharge their Aspectual feature through the Aspectual projection above the higher VP. 6 4.6. Selectional Restrictions The choice of DO in the upper VP depends on the thematic verb in the lower VP in focus constructions in Japanese. 7 This co-occurrence relation is not restricted to the focus constructions, the light verb constructions also show the same pattern. The sentences in (29) and (30) are focus constructions in Japanese. 6 Aspectual Phrase is suggested by Miyagawa (1990) for Japanese, and by Yoon (1990), Lee (1993) for Korean. 7 I thank H. Hoji and H. Aoyagi (p.c.) for drawing my attention to this issue in Japanese and for their suggestions. 148 (29) ringo-ga aka-ku-wa at-ta apple-NOM red-COMP-FOC AR-PAST 'The apple was at least red.' (30) John-ga ik-i-wa si-ta John-NOM go-COMP-FOC SU-PAST 'John at least left.' As we noted before, there is a very clear selectional dependency between DO or light verbs and the head of their complement. That is, adjectives take ar (at in (29), while verbs take su (si in (30)) in Japanese. Light verb constructions in Japanese also show exactly the same pattern. Before we examine this pattern, some remarks are in order concerning the two types of thematic nouns in Japanese. The Japanese counterparts of the Korean Verbal Nouns can be classified into two groups: Adjectival Nouns and Verbal Nouns (Miyagawa 1987). While Verbal Nouns include benkyoo 'study', Adjectival Nouns include words such as kenkoo 'healthy' and kenmei 'wise'. These two groups of nouns take two different light verbs as seen in the following sentences: (31) John-ga kenkoo da John-NOM health(y) AR-PRES 'John is healthy.' (32) John-ga suugaku-o benkyoo-(o) si-ta John-NOM Math-ACC study-ACC SU-PAST 'John studied Math.' 149 In (31) and (32), kenkoo 'health' and benkyoo 'study' are the so-called Adjectival Noun and Verbal Noun, respectively. While the Adjectival Noun selects the light verb AR (da in (31) can be analyzed as de aru), 8 the Verbal Noun selects the light verb su (si in (32)). In (31) and (32), the selection of ar and su is hierarchically reversed. The light verbs are selected by the hierarchically lower lexical items. The LF Head-movement of the Adjectival Noun or VN to the light verb solves this problem. A similar line of approach is suggested by Longobardi concerning the sectional relation between a verb and NP under the DP analysis (a lecture at USC in Fall 1994). If the selectional requirement has to be satisfied under the sisterhood relation, the verb and the selected NP do not satisfy this sisterhood relation. Longobardi suggests that the movement of N° to D at LF satisfies the required relationship. 4.7. Case feature Transfer The case feature of the light verb also depends on the verbal noun. Let us make this issue clear by presenting two categories of light verb constructions in Korean. Non-stative VNs can assign Accusative Case and stative VNs cannot (Suh 1975). The non-stative VNs can be assigned Accusative Case by the light verb, while stative VNs cannot. The following examples exhibit this point. 8 Under the VP-Complementizer analysis, de in de aru can be analyzed as an overt VP- complementizer. I thank H. Hoji for drawing this issue to my attention. 150 (33) John-i [swuhak-ul John-NOM Math.-ACC yelsimhi ha-ess-ta hard do-PAST-DEC 'John studied Math, hard.' (34) John-i phikon-(*ul) ha-ess-ta (Cat B) John-NOM tiredness-ACC do-PAST-DEC 'John was tired.' In (33), the verbal VN assigns Accusative Case to its theme NP swuhak 'Math.' and, in turn, it is assigned the same Case it assigns. In (34), the VN phikon 'tiredness' cannot assign Accusative Case and it cannot be an assignee of Accusative Case. The same Case on the Case assigner as the one assigned by it is called 'reflexive' Case (Kageyama 1991). Contra Grimshaw and Mester (1988), we assume that the light verb ha both in (33) and (34) lacks the capability to assign Accusative Case. Based on the assumption that both light verbs in (33) and (34) are the same light verb, we develop the LF-incorporation hypothesis. If ha could assign Accusative Case, we would expect the light verb ha to assign Accusative Case to phikon 'tired(ness)'. 9 Since this is not the case, we may wonder what makes these two sentences differ in Case assignment. Since the only difference between (33) and (34) is the choice of the verbal "Stative verbal nouns take the option of being incorporated to avoid violation of the Case filter (Baker 1988). Testing by adverbial modification shows that the verbal noun and light verb are incorporated. (i) Mohn-i PHIKON maywu ha-ess-ta John-NOM tired very do-PAST-DEC 'John was very tired.' kongpwu-0]- :1: ka/lul (Cat A) study-COMP-*NOM/ACC 151 noun (that is, kongpwu 'study' in (33) and phikon 'tiredness' in (34)), we can attribute the difference in the Case assigning ability of ha in these sentences to the properties of these two verbal nouns. If we assume, following Kuno (1973), that only [-stative] verbs assign Accusative Case and that, as Suh (1975) suggests, kongpwu 'study' is a [-stative] verbal noun while phikon 'tiredness' is a [+stative] verbal noun, then the non- stative verbal noun kongpwu 'study' has [+ACC], but the stative verbal noun phikon 'tired' does not. If we assume that the Case feature of the thematic verb has been transmitted to the light verb ha, this state of affairs can be nicely captured. We attribute this reflexive Case to the LF- incorporation of the VN to the light verb, as suggested in Cho (1993b;1993c). The current LF-incorporation hypothesis can be further supported by the comparison of ergative constructions in Korean and Japanese. The presence of ergative VNs in Japanese has been argued for in Miyagawa (1989) and Tsujimura (1990). These ergative VNs do not assign Accusative Case to their theme NP and therefore the theme NP moves to the subject position to be assigned Nominative Case by INFL (Burzio 1986). 10 The sentences in (35a) and (35b) are from Grimshaw and Mester (1988) and Miyagawa (1989) with slight modifications. (35) a. ?/*Ya-ga mato-ni meityuu-o sita arrow-NOM target-DAT strike-ACC did 'The arrow struck the target' 1 0 Burzio's generalization (Burzio 1986; Chomsky 1986) A verb assigns an external thematic role iff it can assign Case. 152 b. Ya-ga mato-ni meityuu-sita arrow-NOM target-DAT strike-did 'The arrow struck the target' c. *Ya-ga mato-o meityuu-sita arrow-NOM target-AC C strike-did 'The arrow struck the target' If the ergative VN meityuu 'strike' can neither assign Accusative, as seen in (35c), nor be assigned the same Case, as seen in (35a). the ergative VN must incorporate into the light verb in (35b), to satisfy the Case filter (Rouveret & Vergnaud 1980; Chomsky 1980; Baker 1988). The Korean counterparts of (35) are constructed in (36). (36) a. hwasal-i kwanyek-ey myenchwung-ul ha-ess-ta arrow-NOM target-DAT strike-ACC do-PAST-DEC 'The arrow struck the target' b. hwasal-i kwanyek-ey myenchwung-ha-ess-ta arrow-NOM target-DAT strike-do-PAST-DEC 'The arrow struck the target' c. hwasal-i kwanyek-ul myenchwung-(ul) ha-ess-ta arrow-NOM target-ACC strike-ACC do-PAST-DEC 'The arrow struck the target' In (36), the Korean counterpart of the Japanese ergative VN in (35) can both assign Accusative and be assigned the reflexive Case at the same time. The paradigm observed in (35) and (36) can receive a straightforward account under the LF-incorporation hypothesis by proposing only one parameterization between Japanese and Korean as in (37). 153 (37) while ergative verbs in Japanese do not have the feature [+accusative], those in Korean do. Our proposal for LF-incorporation is further supported by the distribution of VNP. We will take up this issue in the following section. 4.7.1. The distribution of the VN-Phrase In this section, we attempt to provide further evidence supporting the LF- incorporation hypothesis using the distribution of VNP. We suggest that the LF-raising of VN to the light verb provides a correct account for the distribution of VNP. Specifically, we discuss the following three structures. Complement of the light verb (38) John-i [swuhak-ul kongpwu-0]-lul ha-ess-ta John-NOM Math-ACC study-COMP-ACC do-PAST-DEC 'John studied Math.' Subject position (39) *[swuhak-ul kongpwu-0]-ga elyep-ta Math-ACC study-COMP-NOM difficult-DEC 'Studying Math, is difficult' Object of postposition (40) * John- i [swuhak-ul kongpwu-0]-eytayhay John-NOM Math.-ACC study]-about malha-ess-ta talk-PAST-DEC 'John talked about studying Math.' Complement of a theta-verb (41) * John- i [swuhak-ul kongpwu-0]-lul sayngkakha~ess-ta John-NOM Math.-ACC study]-about think-PAST-DEC 'John thought about studying Math.' 154 As seen above, the VN-phrase cannot be placed in the subject position or as the object of P. We suggest that the incorporation of VN to the light verb at LF, along with the claim that Aspect is involved in licensing Accusative (Iida 1987; Miyagawa 1990; Lee 1994; Kim and Maling 1994), can account for the grammaticality of (38). No incorporation in involved in the ungrammatical sentences in (39), (40) and (41). 4.7.1.1. LF-incorporation Approach We suggest that the LF-incorporation of the verbal noun to the light verb along with the analyses and assumptions in (42) is responsible for the distribution of the VNP in J&K. To account for the distribution of VNP, we assume the following. (42) (i) VN incorporates to the light verb at LF forming a chain between the VN and light verb. (Cho 1993b; Hoshi and Saito 1993; Dubinsky 1994). (ii) [-stative] verbal VNs are potential Case assigners (cf. Suh 1975; Cho 1993b) (iii) For the Case feature to be discharged the VN has to be concatenated (Iida 1987) to or licensed by Aspect (Miyagawa 1990). (iv) The verbal Case chain formed by (iii) contains one abstract Case but the abstract Case is realized on every overt NP or [+N] CP." 1 1 This point was suggested to me by A. Li (p.c). One question about (iv) in (42) is how we can account for other languages like English and Romance languages, where the abstract Case is not realized on every overt NP. A. Li suggested the following possibility 155 LF-incorporation in (42i) is argued for in Hoshi and Saito (1993) for theta- role assignment and in the previous sections for selectional relations. It is a natural assumption that the LF-raising of the VN forms a chain with its trace. As for (42iii), if the VNs have a [istative] distinction, as suggested in Suh (1975), it is not unreasonable to say that only [-stative] VNs have the feature [+accusative]. This is because only [-stative] verbs can assign Accusative Case in Japanese (Kuno 1973) and Korean (Kang 1986). 12 > 1 3 We crucially assume that the assignment of ACC is licensed by Aspect (lida 1987; Miyagawa 1990). Even though the [-stative] verbal nouns with the [-N] feature are potential Case assigners, for the Case feature to be discharged the VN has to be concatenated to (lida 1987) or licensed by Aspect (Miyagawa 1990; Lee 1993). It is well attested to in the literature that Aspect is involved in the assignment of Case by VNs. lida (1987) claims that a VN, specified for [-V, +N], can assign Accusative Case if it is concatenated to a Aspectual morpheme in Japanese. The following senesces are from lida (1987:104) with some modification. to me (p.c). Korean type languages allow all [-NJ elements to assign Case, disregarding the content of these elements, while English and Romance languages need to look into the content of verbs, i.e. empty verbs, passive, unaccusative verbs do not assign Case. 12 Adopting Kuno (1973) for Japanese, Kang (1986) states that only [-stative] verbs can assign Accusative in Korean. 13 Some other researchers use different notions to describe the verbs which can assign Accusative Case. Kim (1990) and Hong (1991) use the notion of'agentivity' and 'determinancy', respectively. 156 (43) a. John-ga butsuri-o kenkyuu chuu da John-NOM physics-ACC research mid be 'John is doing research on physics.' b. *John-ga butsuri-o kenkyuu-wa yuumei-da John-NOM physics-ACC research-TOP famous-be 'That John is doing research on physics is famous.' The difference in acceptability between (43a) and (43b) is attributed to the lexical element attached to the VN; while the morpheme chuu 'mid' has a feature [+Aspect], the topic marker wa does not. The following Genitive counterparts of (43) indicate that the aspectual morpheme is relevant to the assignment of Accusative Case by the VN. (44) a. John-ga butsuri-no kenkyuu chuu da John-NOM physics-GEN research mid be 'John is doing research on physics.' b. John-ga butsuri-no kenkyuu-wa yuumei-da John-NOM physics-GEN research-TOP famous-be 'That John is doing research on physics is famous.' As seen in (44b), if the theme NP of the VN has Genitive Case, this sentence turns out to be acceptable. Based on this observation, Miyagawa (1990) argues that structural Cases are licensed under Aspectual phrase. 14 Kim and Maling (1994) and Lee (1994) further argue that Aspect is involved in the assignment of Accusative Case in Korean in general. 14 Lee (1993), adopting Miyagawa (1990), suggests that structural Cases in Korean are licensed under AspP, too. 157 Based on these hypotheses, let us now see how we can account for the grammaticality of the sentence in (38), repeated as (45) below. (45) John-i [swuhak-ul kongpwu-0]-lul ha-ess-ta John-NOM Math-ACC study-COMP-ACC do-PAST-DEC 'John studied Math.' In (45), the potential Case assigner VN kongpwu 'study' moves to the light verb ha, which does not have its own Case feature, but where an aspectual Feature is realized due to the LF-incorporation of the VN to the light verb. This chain now has a Case feature and is [+Aspect], which would allow the chain to license Accusative Case. This Case feature can be realized on every potential Case assignee in (45) due to (42iv). That is, the abstract Accusative Case is morphologically realized on the complement of the light verb swuhak 'Math.' and on the VP-complement of the light verb ha. This analysis can account for the ungrammaticality of the sentences in (39), (40), and (41), repeated as (46), (47), and (48) below. (46) *[swuhak-ul kongpwu-0]-ga elyep-ta Math-ACC study-COMP-NOM difficult-DEC 'Studying Math, is difficult' (47) * John- i [swuhak-ul kongpwu-0]-eytayhay John-NOM Math.-ACC study]-about malha-ess-ta talk-PAST-DEC 'John talked about the study of Math.' (48) *John-i [swuhak-ul kongpwu-0]-lul sayngkakha~ess-ta John-NOM Math.-ACC study]-ACC think-PAST-DEC 'John thought about studying Math.' 158 The current LF-incorporation analysis provides a straightforward account for the ungrammatical status of the.sentences in (46), (47) and in (48). The VNs in these sentences do not incorporate into the matrix verbs. In (46), (47) and in (48), there is no motivation for the incorporation of VN to the thematic verbs. That is, the main verbs in (46), (47) and (48) do not have a selectional relation with the matrix verb as opposed to the VN's selection of a light verb discussed above. Furthermore, the Principle of Full Interpretation (Chomsky 1986) does not motivate the movement of the VN to the theta-assigning verbs at LF. Therefore, VN and the light verb do not form a chain in (46), (47), and (48). Since the VN, the potential Case assigner, does not form a chain with a [+Aspect] element, an ungrammatical status results for those sentences. 4.7.1.2. ECP Approach to the distribution of VNP In this section, we examine the distribution of VNP in terms of ECP (Stowell 1981) and suggest that the ECP approach does not offer a correct prediction for the distribution of VNP. 4.7.1.2.1. Stowell (1981): ECP Approach and CRP Stowell (1981) proposes that ECP is responsible for the subject and object asymmetry in a clause containing an empty complementizer. Consider the following sentences from Stowell (1981). 159 (49) a. Ben knew [ [e] [the teacher was lying] ] b. Louise announced [ [e] she was angry at me ] ] (50) a. [ That the teacher was lying] was hardly obvious. b. [ That Louise was angry at me] came as no surprise. (51) a. *[ [e] [teacher was lying ]] was hardly obvious b. *[ [e] [Louise was angry at me] came as no surprise. When the empty complementizer is properly governed, as in (49), the sentence is grammatical. When the empty Comp is not properly governed, the sentences are ungrammatical, as in (51). Let us now consider the sentences where the verbal VP is the object of a postposition. Consider the sentences in (52). (52) a. *Louise talked about [ that she was angry at me ] ] b. :,: Louise talked about [ [e] she was angry at me ] ] Stowell (1981) can attribute the ungrammaticality of (52) to the Case Resistance Principle (CRP), which states that the Case assigner cannot be assigned Case. 15 Given that the clause S (CP in the current term) in (52) has the Case assigner Tense, the assignment of oblique Case to the that- clause is in violation of the CRP. 15 The Case Resistance Principle (Stowell p. 146) Case may not be assigned to a category bearing a Case-assigning feature. 160 4.7.1.2.2. ECP and CRP in Korean In this section, we suggest that the ECP account given in Stowell (1981) does not provide a correct prediction for the distribution of VNPs. We start with the VNP in subject and object positions. Object Position (53) John-i [swuhak-ul kongpwu-0]-lul ha-ess-ta John-NOM Math-ACC study-COMP-ACC do-PAST-DEC 'John studied Math.' Subject Position (54) :1: [swuhak-ul kongpwu-0]-ga elyep-ta Math-ACC study-COMP-NOM difficult-DEC 'Studying Math, is difficult' The (un)grammaticality of (53) and (54) can be accounted for by the ECP in Stowell (1981). The sentence in (53) is correctly predicted to be grammatical under the ECP account since the null complementizer is governed by the light verb. The null comp in (54), however, is not properly governed and correctly predicted to be ungrammatical. The same ECP account, however, does not provide a correct predication in the following sentence, as H. Hoji (p.c.) pointed out to me. (55) *John-i [swuhak-ul kongpwu-0]-lul sayngkakha-ess-ta John-NOM Math.-ACC study]-ACC think-PAST-DEC 'John thought about studying Math.' In (55), the null-comp is properly governed by the matrix verb, but is ungrammatical. 161 One more problem can be observed in the sentences where the VNP is the object of a postposition. If the VN-phrase is an object of P and P is a proper governer, the null Comp obeys the proper government requirement. (56) *John-i [swuhak-ul kongpwu-0]-eytayhay John-NOM Math.-ACC study]-about malha-ess-ta talk-PAST-DEC 'John talked about the study of Math.' Nonetheless, the sentence in (56) is ungrammatical. One possibility is that P is a functional head, and, therefore, cannot properly govern the null- Comp, correctly ruling out the sentence in (56). This approach, however, does not correctly predict the unacceptable status of the sentence in (56), where the VP-complementizer ki is employed. (57) *John-i [chayk-ul ilk-ki]-eytayhay John-NOM book.-ACC read-COMP]-about sayngkakha~ess-ta think-PAST-DEC 'John thought about reading a book.' Since the null-complementizer is not employed in (56), ECP is irrelevant to ruling out the sentences with the null or overt complementizer in (56) and (57). We have two possible ways to rule out these two sentences: one is LF-incorporation hypothesis argued for in the preceding sections, and the other is CRP following Stowell (1981). In the next section we will suggest that CRP cannot be generalized to Japanese and Korean and that the 162 correct analysis to account for the grammaticality of the sentences in (56) and (57) is the LF-incorporation hypothesis. If Korean and Japanese obey CRP, the ungrammatically of the sentences in (56) and (57) can also be attributed to the violation of CRP. However, the CRP account faces a problem since the CRP cannot be generalized to J&K. Even though IP with some complementizers obeys CRP, as in (58) and (59), a VP with a VP-complementizer does not obey the same constraint, as (60) demonstrates. (58) Taroo-ga [ookami-ga kuru to]-(*o) hanasi-ta (J) Taro-NOM wolf-NOM come COMP-*ACC say-PAST 'Taro said that the wolf was coming.' (59) John-i [nuktay-ga onta-ko]-(*lul) malha-ess-ta (K) Taro-NOM wolf-NOM come COMP-*ACC say-PAST 'Taro said that the wolf was coming.' (60) John-i [ku chayk-ul sa-ci]-lul ani ha-ess-ta (K) John-NOM the book-ACC buy-COMP-ACC not do-PAST-DEC 'John did not buy the book.' This CRP cannot account for the appearance of the Case marker on the VP-complementizer ci in (60). In (60), the verb in this VP can assign Accusative Case and the VP with the complementizer is assigned Accusative Case, violating the CRP. Furthermore, ki as a CP- complementizer can be assigned Accusative Case, too, as demonstrated in (61). (61) John-i [Mary-ka ttena-ess-ki]-lul John-NOM Mary-NOM buy-PAST-COMP-ACC pala-ess-ta hope-PAST-DEC 'John hoped that Mary had left' 163 Since the CRP cannot be generalized to VP-complements and even to some S-complements in Korean as shown, the CRP account does not provide a proper answer to the question of the ungrammaticality of sentences with VNP in the object position of P. Thus far, we have seen that the distribution of VNP can best be accounted for by the LF-incorporation hypothesis, and that the alternative approaches, ECP and CRP are not on the right tract. This further supports our LF-incorporation analysis of VN to the light verb. 4.8. Conclusion In summary, the LF-incorporation approach is motivated to check the selectional relation between the selector and selectee. This approach is not only supported by DO or light verb selection, but also by the selectional relationship between a verb and NP which is under DP. Since the LF- incorporation of the theta-verb moves the grids such as theta-gird, Aspect- grid, and so on, the LF-incorporation of the theta-verb to the thematically empty verb transfers theta-roles, aspect, and Case. As we have seen, this is the case. The theta-roles are discharged during the LF-incorporation (Hoshi and Saito 1993). The reflexive Case can also be accounted for under the current proposal. 164 Chatpter V. Summary and Concluding Remarks In this thesis, we have attempted to provide a unitary account for the focus constructions and light verb constructions. Many arguments are made for the structural parallelism between these two constructions. First, we claimed that expletive verbs in focus constructions are base-generated the same as the one in light verb constructions. Second, verbal nouns in light verb constructions have two distinctive categorical features. Third, the light verb constructions with the verbal VN take a null- complementizer. While arguing for these points, we suggest that negative items can be lexical items such as adjective, adverb, or verb, rather than functional items. If this is correct, in the languages in which negatives are lexical items, the negative item projects a lexical projection, not a functional projection. Given the current assumption that functional projections are located above lexical projections, the negatives being lexical forces us to reconsider the syntactic architecture. We also have claimed that lexical items do not directly select a maximal projection headed by another lexical category; rather, they select a maximal projection headed by a functional category. If this hypothesis is universal, then the VP-shell in Larson (1988) should be recast within the proposed hypothesis, so that the upper verb selects the functional projection containing the lower VP. In relation to Larsonian structure, the 165 challenging question is why English employs a phonologically null empty verb in the upper VP. The underspecified lexical category also provides another challenge to the categorial feature not only of the major categories, (lexical categories), but also of the minor categories, (functional categories). Not only lexical categories, but also functional categories can be underspecified for [N] and [V]. The syntactic underspecification in functional categories deserves pursuing. The LF-movement motivated by the selectional relation also has a non-trivial consequence. If the LF-movement is motivated by the selectional relations between the selector and selectee, the LF- incorporation of the verbal noun to the light verb need not be motivated to discharge the theta-role, as claimed by Hoshi and Saito (1993). As a matte r of fact, selection-motivated movement at LF has a wider consequences in the grammar than the motivation to discharge theta- roles. The selection-motivated movement can also treat the s-selection between a verb and noun under the DP. Furthermore, the selection of a subjunctive clause in English can be accounted for under the proposed analysis, where the selector, matrix verb, is not a semantically incomplete verb. These consequences, as well as application of our proposals to other languages remain for future research. 166 Bibliography Abasolo, Rafael. 1974. Basic Semantic Structures of Korean. Ph.D. dissertation, Georgetown University. Abney, Steven. 1987. The English Noun Phrase in Its Sentential Aspect. Ph.D. dissertation, MIT. Ahn, Hee-Don. 1990. On Light Verb Constructions in Korean and Japanese. Japanese Korean Linguistics I, 221-237. Akmajian, Adrian, Susan M. Steele, & Thomas Wasow. 1979. The Category AUX in Universal Grammar. Linguistic Inquiry 10(1), 1- 64. Aoun, Joseph., & Yen-hui Audrey Li. 1993. Syntax of Scope. MIT Press. Aoyagi, Hiroshi. 1993. On Honorific Sentences and Categories in Japanese, ms., University of Southern California. Baker, C. Lee 1991. The Syntax of Not: The Limits of Core Grammar. Linguistic Inquiry 22(3), 387-429. Baker, Mark. 1985. Syntactic Affixation and English Gerunds. In The proceedings of the West Coast Conference on Formal Linguistics IV Standford Linguistics Association. Baker, Mark. 1988. Incorporation: A Theory of Grammatical Function Changing. Chicago: The University of Chicago Press. Baltin, Mark R. 1993. Negation and Clause Structure, ms., New York University. Benmamoun, Elabbas. 1992. Functional and Inflectional Morphology Problems of Projection, Representation and Derivation. Ph.D. dissertation, University of Southern California. Bowers, John. 1993. The Syntax of Predication. Linguistic Inquiry 24(4), 591-656. Burzio, Luigi. 1986. Italian Syntax. Dordrecht: Reidel. 167 Burzio, Luigi. 1991. The Morphological Basis of Anaphora. Journal of Linguistics 27, 81-105. Cattell, Ray. 1984. Composite Predicates in English. Academic Press. Cho, Dong-In. 1992a. Functional Projections and Verb Movement. Paper read at the International Circle of Korean Linguistics held in Washington D.C. Cho, Dong-In. 1992b. Inalienable-Type Multiple Accusative Constructions in Korean and Japanese. In Patricia M. Clancy (Eds.), Japanese I Korean Linguistics (pp. 319-337). Stanford: Stanford Linguistics Association. Cho, Dong-In. 1993a. Emphasis, Negation, and Verb Movement. MIT Working Paper 20, 35-49. Cho, Dong-In. 1993b. Multiple Complement Constructions and Case Chain. Paper presented at the 5th Harvard International Symposium on Korean Linguistics. Cho, Dong-In. 1993c. Multiple Complement Constructions in Korean. Talk given at the Workshop on Theoretical East Asian Linguistics. Choe, Hyen Bae. 1929. Wuli Malpon (The Grammar of the Korean Language). Seoul, Korea: Chungumsa. Choe, Hyon Sook. 1988. Restructuring Parameters and Complex Predicates: A Transformational Approach. Ph.D. dissertation, MIT. Chomsky, Noam. 1957. Syntactic Structures. The Hague: Mouton. Chomsky, Noam. 1980. On Binding. Linguistic Inquiry 11(1), 1-46. Chomsky, Noam. 1981. Lectures on Government and Binding. Dordrecht: Foris. Chomsky, Noam. 1986. Knowledge of Language: Its Nature, Origin, and Use. Praeger. Chomsky, Noam. 1992. A Minimalist Program for Linguistic Theory. In MIT Occasional Papers in Linguistics, Chomsky, Norm. 1994. Bare Phrase Structure. MIT. 168 Chung, Taegoo. 1993. Argument Structure and Serial Verbs in Korean. Ph.D dissertation, University of Texas, Austin. Dahl, O. 1979. Typology of sentence negation. Linguistics 17, 79-106. Dubinsky, Stanley. 1994. Syntactic underspecification: A minimalist approach to light verbs. MIT Working Papers in Linguistics. Emonds, Joseph E. 1985. A unified theory of syntactic categories. Dordrecht, Holland ; Cinnaminson, U.S.A.: Foris Publications. Fukui, Naoki, & Magaret Speas. 1986. Specifiers and Projection. In Naoki Fukui, T.R. Rapoport, & E. Sagey (Eds.), MIT Working Papers in Linguistics (pp. 128-172). Cambridge, Massachusetts: MIT. Fukui, Naoki. 1986. A Theory of Category Projection and Its Applications. Ph.D. dissertation, MIT. Fukui, Naoki. 1990. The Problem of the Phrase Structure of Japanese: A Historical Survey. In Y Abe et. al. (Eds.), The Proceedings of the Nanzan University International Symposium on Japanese Teaching (Linguistic Theory Special Session). Fukui, Naoki. 1994. The principles-and-parameters approach: a comparative syntax of English and Japanese. In M Shibatani & T. Bynon (Eds.), Approaches to Language Typology. Oxford University Press. Grimshaw, Jane, & Armin Mester. 1988. Light Verbs and Theta-Marking. Linguistic Inquiry 19, 205-232. Grimshaw, Jane. 1990. Argument Structure. Cambridge, Mass: MIT Press. Grimshaw, Jane. 1991. Extended Projection, ms., Brandeis University. Han, Hak-Sung. 1987. The Configurational Structure of the Korean Language. Seoul: Hanshin Publishing Co. Harada, Sin-iti. 1973. Counter Equi-NP Deletion. Bulletin of the Research Institute ofLogopedics and Phoiatrics 7,113-148. Hasegawa, Nobuko. 1991. On Head Movement in Japanese: The Case of Verbal Nouns. Proceedings of Sophia Linguistic Society 6, 8-32. Higginbotham, James. 1985. On Semantics. Linguistic Inquiry 16, 547- 593. 169 Hoji, Hajime. 1989. VP-Proposing in Japanese. Handout presented at the J/K workshop at UCLA. Hoji, Hajime. 1991a. On the Negative Sentences and Other Related Constructions in Korean and Japanese. Handout presented at La Jolla Syntax Workshop. Hoji, Hajime. 1991b. Raising to object, ECM and the Major Object in Japanese. Talk given at the Rochester Japanese Linguistics Workshop. Hong, Ki-Sun. 1991. The Passive Construction and Case in Korean. Berkeley Linguistics Society 17. Hoshi, Hiroto, & Mamoru Saito. 1993. The Japanese Light verb Construction, ms., University of Connecticut. Huang, C.-T. James. 1994. On Lexical Structure and Syntactic Projection, ms., University of California, Irvine. Iida, Masayo. 1987. Case-Assignment by Nominals in Japanese. In Masayo Iida, Stephen Wechsler, & Draga Zee (Eds.), Working Papers in Grammatical Theory and Discourse Structure: Interactions of Morphology, Syntax, and Discourse (pp. 93-138). Stanford: CSLI. Isoda, Michio. 1991. The Light Verb Construction in Japanese. Chicago Linguistic Society 27, 261-275. Jung, Yeun-Jin. 1992. A Restrictive Theory of Functional Categories and Their Paramentric Variation. Ph.D dissertation, University of Washington. Kageyama, Taro. 1982. Word Formation in Japanese. Lingua 57, 217-258. Kageyama, Taro. 1991. Light Verb Constructions and the Syntax- Morphology Interface. In Nakajima Heizo (Eds.), Current English Linguistics in Japan (pp. 169-203). New York: Mouton. Kang, Myung Yoon. 1988. Topics in Korean Syntax: Phrase Structure, Variable Binding, Movement. Ph.D. dissertation, MIT. Kang, Young-Se. 1986. Korean Syntax and Universal Grammar. Ph.D. dissertation, Harvard University. Kayne, Richard. 1989. Notes on English agreement. CIEFL Bulletin 1, 41 - 67. 170 Kim, Jeong-Ryeol. 1991. A Lexical-Functional Grammar Account of Light Verbs. Ph.D. dissertation, University of Hawaii. Kim, Nam-Kil. 1984. The Grammar of Korean Complementation. Honolulu, Hawaii: University of Hawaii. Kim, Soon-Ham Park. 1967. A Transformational Analysis of Negation in Korean. Ph.D. dissertation, University of Michigan. Kim, Soowon, & Joan Maling. 1994. Case Assignment in the sipta- Construction. Paper presented at the 9th International Circle of Korean Linguistics. Kim, Young-joo. 1990. The Syntax and Semantics of Korean Case: The Interaction between Lexical and Syntactic Levels of Representation. Ph.D. dissertation, Harvard University. Kiparsky, Paul. 1982. Lexical Morphology and Phonology. In Linguistics in the morning calm (pp. 3-91). Seoul: Hanshin. Kitagawa, Chisato, & Claudia N. Ross. 1982. Prenominal Modification in Chinese and Japanese. Linguistic Analysis 9(1). Kitagawa, Yoshihisa. 1986. Subjects in Japanese and English. Ph.D. dissertation, University of Massachusetts. Kubo, Miori. 1992. Japanese Syntactic Structures and Their Constructional Meaning. Ph.D dissertation, MIT. Kuno, Susumo. 1973. The Structure of the Japanese Language. Cambridge, Mass.: MIT Press. Kuroda, Sige-Yuki. 1965. Generative Grammatical Studies in the Japanese Language. Ph.D. dissertation, MIT. Kuroda, Sige-Yuki. 1988. Whether We Agree or Not: A Comparative Syntax of English and Japanese. In William J. Poser (Eds.), Papers from the Second International Workshop on Japanese Syntax (pp. 103-143). Center for the Study of Language and Information. Kuroda, Sige-Yuki. 1992. Japanese Syntax and Semantics: Collected Papers. In Dordrecht/Boston/London: Kluwer Academic Publishers Group. Laka, Itziar. 1990. Negation in Syntax: On the Nature of Functional Categories and Projections. Ph.D. dissertation, MIT. 171 Larson, Richard K. 1988. On the Double Object Construction. Linguistic Inquiry 19, 335-391. Lee, Hong-bae. 1970. On negation in Korean. Language Research 6(2). Lee, Hong-bae. 1972. Problems in the Description of Korean Negation. Language Research 8(2), 60-75. Lee, Jae Hong. 1992. Postverbal Adverbs and Verb Movement in Korean. In Patricia M. Clancy (Eds.), Japanese I Korean Linguistics (pp. 429- 446). Stanford: Stanford Linguistics Association. Lee, Keon Soo. 1991. Biclausal Analysis of the So-called Double Object Constructions in Korean. Korean Linguistics 6. Lee, Young-Suk. 1993. Scrambling as Case-Driven Obligatory Movement. Ph.D. dissertation, University of Pennsylvania. Lee, Young-Suk. 1994. Accusative Case Licensing and Aspect Phrase. Paper presented at the 9th International Circle of Korean Linguistics. Martin, Samuel Elmo. 1975. A Reference Grammar of Japanese. New Haven: Yale University Press. Milsark, G. L. 1988. Single-ing. Linguistic Inquiry 19(4), 611-634. Miyagawa, Keiko. 1993. Against N'-deletion: Properties of Japanese Numerals, ms., University of Southern California. Miyagawa, Shigeru, & Fusae Ekida. 1990. Historical Development of the Accusative Case Marking in Japanese as Seen in Classical Literary Texts, ms., Ohio State University. Miyagawa, Shigeru. 1987. Lexical Categories in Japanese. Lingua 73, 144- 184. Miyagawa, Shigeru. 1989. Light Verbs and the Ergative Hypothesis. Linguistic Inquiry 21(659-688). Miyagawa, Shigeru. 1990. Case Realization and Scrambling, ms., Ohio State University. No, Yongkyoon. 1988. Negative Morphemes in Korean: Evidence for a Derivational Treatment. The Proceedings of International Circle of Korean Linguistics VI. 172 Oh, Choon-Kyu. 1971. On the Negation of Korean. Language Research 7(2). Ohta, Kaoru. 1994. Verbal Nouns in Japanese: A Recategorization Approach. Ph.D dissertation, UCLA. Ouhalla, Jamal. 1989. Sentential Negation, Relativised Minimality and the Aspectual Status of Auxiliaries. The Linguistic Review 7, 183-231. Park, Kapyong. 1992. Light Verb Constructions in Korean and Japanese. Seoul: Thae Hak Sa. Park-Han, Kyung Ja. 1981. The Development of Negation in One Korean Child. Ph.D. dissertation, University of Hawaii. Parry, M. Mair. 1985. The Dialect of Cairo Montenotte. Ph.D. dissertation, University of Wales. Pollock, Jean-Yves. 1989. Verb Movement, UG and the Structure of IP. Linguistic Inquiry 20(3), 365-424. Poser, William J. 1981. The 'Double-o Constraint': Evidence for Direct Object Relation in Japanese, ms., MIT. Poser, William J. 1991. Japanese Periphrastic Verbs and Noun Incorporation, ms., Stanford University. Poser, William J. 1992. Blocking of Phrasal Constructions by Lexical Items. In Ivan A. Sag & Anna Szabolcsi (Eds.), Lexical Matters Stanford University. Reinhart, Tanya. 1976. The Syntactic Domain of Anaphora. Ph.D. dissertation, MIT. Reuland, Eric. 1983. Governing-ing. Linguistic Inquiry 14(1), 101-136. Rizzi, Luigi. 1990. Relativized Minimality. Cambridge: The MIT Press. Roberts, Ian. 1990. Some Notes on VP-Fronting and Head Government. In Joan Mascaro and Marina Nespor (Eds.), Grammar in Progress: GROW essays for Henk van Riemsdijk Dordrecht: Foris. Roberts, Ian. 1991. Excoporation and Minimaltiy. Linguistic Inquiry 22(1), 209-218. Rooryck, Johan. 1994. On Two Types of Syntactic Underspecification: towards a feature theory shared by syntax and phonology. Paper 173 presented at the Morphology-Syntax Connection. MIT, Cambridge, MA. Ross, John R. 1967. Constraints on Variables in Syntax. Ph.D. dissertation, MIT. Rouveret, Alain, & Jean-Roger Vergnaud. 1980. Specifying Reference to the Subject: French Causatives and Conditions on Representations. Linguistic Inquiry 11(1), 97-202. Saito, Mamoru. 1983. Case and Government in Japanese. The Proceedings of the Second West Coast Conference on Formal Linguistics. Sells, Peter. 1990. More on Light Verbs and Theta-Marking. ms., Stanford University. Shibatani, Masayoshi. 1978. Mikami Akira and the notion of 'subject' in Japanese grammar. In J Hinds & I Howard (Eds.), Problems in Japanese Syntax and Semantics Tokyo: Kaitakusha. Song, Seok Choong. 1988. Explorations in Korean Syntax and Semantics. Berkeley: UC Berkeley Institute of East Asian Studies. Stowell, Timothy. 1981. Origin of Phrase Structure. Ph.D. dissertation, MIT. Suh, Jeng-Soo. 1975. Tongsa ha-uy Mwunpep (A Grammar ofha-verbs). Seoul: Hyengsel Publishing Co. Suh, Jinhee. 1990. Scope Phenomena and Aspects of Korean Syntax. Ph.D. dissertation, University of Southern California. Svenonius, Peter, to appear. C-Selection as Feature-Checking. Studia Linguistica. Takezawa, Koichi. 1987. A Configurational Approach to Case-Marking in Japanese. Ph.D. dissertation, University of Washington. Terada, Michiko. 1990. Incorporation and Argument Structure in Japanese. Ph.D. dissertation, University of Massachusetts at Amherst. Travis, Lisa d. 1984. Parameters and Effects of Word Order Variation. Ph.D. dissertation, MIT. Tsujimura, Natsuko. 1990. Ergativity of Nouns and Case Assignment. Linguistic Inquiry 21(2), 277-287. 174 Tsujimura, Natsuko. 1992. Licensing Nominal Clauses: The Case of Deverbal Nominals in Japanese. Natural Language & Linguistic Theory 10(3), 477-522. Urushibara, Saeko. 1993. Nonagentive Light Verb Constructions: Korean vs. Japanese. Harvard Studies in Korean Linguistics V, 423-435. Van Gelderen, Elly. 1992. Argument without number: the case of it and het. Linguistics 30, 381-387. Vanden Wyngaerd, Guido. 1994. PRO-legomena. In Distribution and reference of infinitival subjects Berlin: Mouton De Gruyter. Whitman, John. 1991. String Vacuous I to C. Paper presented at GLOW. Williams, Edwin. 1981. Argument Structure and Morphology. The Linguistic Review 2, 84-114. Yang, Dong-Whee. 1976. Korean Negation Revisited. Linguistic Journal of Korea 11(2), 243-48. Yang, Hyun-Kwen. 1990. Categories and Barriers in Korean. Ph.D dissertation, University of Texas, Austin. Yang, In-Seok. 1972. Korean Syntax: Case Markers, Delimiters, Complementation, and Relativization. Seoul: Paekhap Publishing Company. Yoon, Jeong-Me. 1989. ECM and Multiple Subject Constructions in Korean. In Susumu Kuno, et al (Eds.), Harvard Studies in Korean Linguistics (pp. 369-381). Seoul: Hanshin Press. Yoon, Jeong-Me. 1990. Verb Movement and the Structure of IP in Korean. Language Research 26(2). Yoon, Jong Yuri. 1990. Korean Syntax and Generalized X-bar Theory. Ph.D dissertation, University of Texas, Austin. Zanuttini, Raffaella. 1991. Syntactic Properties of Sentential Negation: A Comparative Study of Romance Languages. Ph.D. dissertation, University of Pennsylvania. Zubizarreta, Maria Luisa. 1985. The Relation between Morphophonology and Morphosyntax: The Case of Romance Causatives. Linguistic Inquiry 16, 247-289.. 175 * !
Linked assets
University of Southern California Dissertations and Theses
Conceptually similar
PDF
Ellipsis constructions in Chinese
PDF
"The unaccusative trap": L2 acquisition of English intransitive verbs.
PDF
Grammaticalization and the development of functional categories in Chinese
PDF
Form and meaning: Negation and question in Chinese
PDF
Beyond words and phrases: A unified theory of predicate composition
PDF
Andrej Belyj's "Petersburg" and James Joyce's "Ulysses": A comparative study
PDF
Correspondence and faithfulness constraints in optimality theory: A study of Korean consonantal phonology
PDF
Aspects of questions in Japanese and their theoretical implications.
PDF
Consequences of heritage language loss and maintenance and factors that affect heritage language development: Voices from second-generation Korean-American adults
PDF
A case grammar of the parker manuscript of the "Anglo-Saxon chronicle" from 734 to 891
PDF
A rhetoric of the short story: A study of the realistic narratives of Flaubert, Maupassant, Joyce, and Hyon Chin'gon
PDF
A critical edition of the Escorial manuscript of 'historia de los indios de la neuva espana' of fray toribio de benavente (motolinia) (spanish text)
PDF
Borrowing and borrowability
PDF
Arthur Machen’s treatment of the occult and a consideration of its reception in England and America
PDF
Antigone: A study in critical method
PDF
'Beowulf' and 'the hobbit': elegy into fantasy in j. R. R. Tolkien's creative technique
PDF
Communicative rights and responsibilities in an East Los Angeles barrio: An analysis of epistemic modal use
PDF
Causativity in Korean: Syntactic causative, control, and purpose
PDF
In the event of focus
PDF
Charlotte Smith: Life of a novelist, novels of a life
Asset Metadata
Creator
Cho, Dong-In
(author)
Core Title
A comparative study of focus constructions
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)
Advisor
Hoji, Hajime (
committee chair
), Li, Audrey (
committee member
), Polinsky, Maria "Masha" (
committee member
)
Permanent Link (DOI)
https://doi.org/10.25549/usctheses-c17-463208
Unique identifier
UC11352239
Identifier
9600964.pdf (filename),usctheses-c17-463208 (legacy record id)
Legacy Identifier
9600964-0.pdf
Dmrecord
463208
Document Type
Dissertation
Rights
Cho, Dong-In
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
Tags
language, linguistics