University of Southern California Dissertations and Theses

The syntax of cleft constructions in Japanese: A base-generation analysis

(USC Thesis Other)

The syntax of cleft constructions in Japanese: A base-generation analysis

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content THE SYNTAX OF CLEFT CONSTRUCTIONS IN JAPANESE: A BASE-GENERATION ANALYSIS by Yasuhiko Miura A Dissertation Presented to the FACULTY OF THE USC GRADUATE SCHOOL UNIVERSITY OF SOUTHERN CALIFORNIA In Partial Fulfillment of the Requirements for the Degree DOCTOR OF PHILOSOPHY (EAST ASIAN LANGUAGES AND CULTURES) May 2011 Copyright 2011 Yasuhiko Miura ii Acknowledgements It is a great pleasure to be able to express my gratitude to the people I am indebted to. First of all, I would like to thank my dissertation committee members, Hajime Hoji (chair), Audrey Li, Andrew Simpson, and Mario Saltarelli. I owe my deepest gratitude to Hajime. After I finished my master course in English Linguistics at Tohoku University, I was at a loss which way to go as a generative researcher. I came to the U.S. as an exchange student at UCLA in 1999 in the hope that I would be able to find an answer to that. During my stay, I had a chance to meet Hajime, which completely changed my life as a generative researcher. I clearly remember the day when Hajime told me enthusiastically about his research method as well as his philosophy. I was so impressed and excited that I could not sleep at that night. I decided to pursue my graduate study at USC and since then I have learned a lot from him about how we should proceed as a researcher for the establishment of generative grammar as an empirical science. Without his constant help and encouragement, I would not have been able to complete this dissertation. I would also like to thank Audrey Li, who has been on my committee since my qualifying exam, for her helpful suggestions, encouragement, and patience. I also benefited from her courses in syntax which I was able to take over the years. I am also very grateful to Andrew Simpson, who has also been on my committee since my qualifying exam, for his helpful comments and insightful courses I took. I would also like to show my gratitude to the former members of my committee: Namkil Kim and the late Jean-Roger Vergnaud, for their help and encouragement in my early stage of the graduate study. I am much indebted to the members in Hajime's research group: Yukinori Takubo, Ayumi Ueyama, Kiyoko Kataoka, Teruhiko Fukaya, J.-R. Hayashishita, Yukiko Tsuboi, Emi Mukai, Maki Irie, and Yasuo Ishii. The discussions with them both in and outside the classes at USC iii helped me lay the groundwork for my career as a researcher to pursue the generative grammar as an empirical science. I would not have been able to survive the life as a graduate student without their generous support. When I resumed my academic career at Tohoku University in 1997 after the four years of my business carrier, the primary purpose was to obtain a teaching certificate for ESL. However, the faculty of the English Linguistics Department at Tohoku University led me to the field of the generative grammar. I am especially grateful to Profs. Masaru Nakamura, Yoshiaki Kaneko and Daiko Takahashi. I am also grateful to the late S.-Y. Kuroda, who was teaching then at Tohoku University as a visiting professor, for the insightful seminar I took. I would also like to thank the graduate students at the English Linguistics Department in Tohoku University, especially, Etsuro Shima, Yoshiki Ogawa, Nobuhiro Miyosi, Yoshihito Dobashi, Hiroshi Yamashita, who patiently answers my questions and engaged in discussions with me. My visit to UCLA for about two and half years as an exchange student before I started the graduate study at USC was really fruitful thanks to the generous and knowledgeable faculty at the Department of Linguistics at UCLA. My special thanks go to Anoop Mahajan, Edward Keenan, Tim Stawell, Dominique Sportiche and Hilda Koopman. I would also like to thank the graduate students there, especially, Haiyong Liu, who helped me start the life in LA smoothly, Ivano Caponigro, Harold Torrence, Heriberto Avelino, and Eric Jackson. During the nine years of the graduate study at USC, I have received various kinds of help from so many people. I first would like to thank the faculty members of USC, especially, the faculty of the Department of Linguistics: Barry Schein, Joseph Aoun, Jim Higginbotham, Roumi Pancheva, Maria Luisa Zubizarreta, and Hagit Borer. My special thanks also go to David Bialock, who is the Director of Graduate Studies at EALC. He was always encouraging and ready to give his help when I need one, and kept track of the progress of my research. I would iv also like to express my gratitude to the staff members of EALC: Sherall Preyer and Patricia Whiting. I was able to have opportunities to teach Japanese as an Assistant Lecturer throughout the nine years of my graduate study. These were really good opportunities for my future career as well as to deepen my understanding of Japanese language. I would like to thank the lecturers of Japanese program in EALC, Yuka Kumagai, Masako Tamanaha, Yumi Matsumoto and Makiko Osaka, who patiently helped me and taught me the teaching method of Japanese. Last but not least, I would like to thank my mother, Kyoko Miura, and my aunt, Ryoko Miura, for offering me moral and financial support for the many years I was in the U.S. I would also like to thank my parents-in-law, Yoshio & Yukiko Arai, for their understanding of my decision to pursue the academic career and their financial support. My mother-in-law has just passed away before seeing me complete this dissertation, which she had longed for. Finally, I would like to thank my wife, Eri, for her support and encouragement, and our son, Ricky, for always energizing me and giving me hope. v Table of Contents Acknowledgements ................................................................................................................... ii List of Tables ............................................................................................................................ ix Abstract ..................................................................................................................................... xi Chapter 1: Introduction ....................................................................................................... 1 1.1. Cleft constructions .............................................................................................................. 1 1.2. Issues ................................................................................................................................... 5 1.2.1. The cleft construction and 'scrambling' ..................................................................... 5 1.2.1.1. Similarities between the OS-type construction and the cleft construction ..... 5 1.2.1.2. The OS-type construction in Ueyama 1998 ................................................. 10 1.2.1.2.1. Two types of the OS-type construction: Surface OS-type and Deep OS-type ................................................................................ 10 1.2.1.2.2. The OS-type construction and resumption .................................... 15 1.2.1.3. Apparent differences between the OS-type construction and the cleft construction .................................................................................................. 17 1.2.1.3.1. Structural differences between the two constructions .................... 17 1.2.1.3.2. Empirical differences between the two constructions .................... 18 1.2.2. The two types of analyses of the Japanese cleft construction ................................. 21 1.3. The goal of this work ........................................................................................................ 23 1.4. Outline .............................................................................................................................. 25 Chapter 2: Two Types of Analyses of Cleft Constructions in Japanese ............... 28 2.1. Introduction ....................................................................................................................... 28 2.2. Base-generation analyses .................................................................................................. 28 2.2.1. Hoji 1987, 1990, Ueyama & Hoji 2001, Hoji & Ueyama 2003 .............................. 28 2.2.2. Kizu 2005 ................................................................................................................ 31 2.3. Movement Analyses.......................................................................................................... 33 2.3.1. Hiraiwa and Ishihara 2002....................................................................................... 33 2.3.2. Takahashi 2006 ........................................................................................................ 35 2.4. Issues ................................................................................................................................. 36 2.4.1. Issues on the movement analyses ............................................................................ 37 2.4.1.1. Conceptual Problems of Hiraiwa & Ishihara 2002 ...................................... 37 2.4.1.1.1. PBC violation ................................................................................. 37 2.4.1.1.2. Further Consideration: Different types of A′-movement ............... 43 2.4.1.2. Empirical Problems of Takahashi 2006 ....................................................... 45 2.4.1.2.1. NSE facts in the cleft constructions ............................................... 45 2.4.1.2.2. Condition D effects ........................................................................ 49 2.4.1.3. Resumption in the cleft constructions .......................................................... 55 2.4.2. Issues on the base-generation analyses .................................................................... 57 2.4.2.1. Conceptual Problems of Kizu’s (2005) analysis .......................................... 57 vi 2.4.2.1.1. Predication relations in null operator constructions ....................... 57 2.4.2.1.2. Small clause analysis ..................................................................... 59 2.4.2.1.3. Position of the focus phrase ........................................................... 62 2.4.2.2. Issues on Hoji 1987, 1990, Hoji & Ueyama 2003 ........................................ 66 2.5. Choose among the different analyses................................................................................ 68 2.5.1. Arguments for the movement analyses ................................................................... 68 2.5.2. Between the base-generation analyses .................................................................... 70 2.6. Summary ........................................................................................................................... 70 Chapter 3: Methodological Preliminaries: EPSA Method .............................................. 72 3.1. Introduction ....................................................................................................................... 72 3.2. Computational System and the model of judgment making ............................................. 73 3.3. Prediction making in language faculty science ................................................................. 77 3.3.1. Types of hypotheses ................................................................................................ 77 3.3.2. Disconfirmability and quantitative analyzability .................................................... 83 3.3.3. *Schema-based predictions and ok Schema-based predictions .................................. 86 3.4. Experiments in language faculty science .......................................................................... 88 3.4.1. EPSA ....................................................................................................................... 88 3.4.2. Preliminary experiments .......................................................................................... 91 3.5. Research heuristics............................................................................................................ 92 3.5.1. Maximize testability ................................................................................................ 93 3.5.2. Maximize our chances of learning from errors ....................................................... 94 3.5.2.1. Some illustrations: BVA case ...................................................................... 95 3.5.2.2. Invalid hypotheses ...................................................................................... 100 3.6. Summary ......................................................................................................................... 101 Figure 1: The Model of Judgment-Making by the Informant on the acceptability of sentence α with interpretation γ(a, b) ................................................................ 75 Chapter 4: Preliminaries of the Experiments ........................................................... 104 4.1. Overview ......................................................................................................................... 104 4.2. Confirmed schematic asymmetries ................................................................................. 105 4.2.1. BVA (bound variable anaphora) ........................................................................... 105 4.2.1.1. So-NPs vs. a-NPs as bindees ...................................................................... 110 4.2.1.1.1. Experimental design .................................................................... 111 4.2.1.1.2. Results of the experiments ........................................................... 116 4.2.1.2. FD-based BVA ........................................................................................... 117 4.2.1.2.1. Experimental design (I): excluding quirky binding ..................... 117 4.2.1.2.2. Results of the experiments (I) ...................................................... 123 4.2.1.2.3. Experimental design (II): excluding ID-based BVA ................... 125 4.2.1.2.4. Results of the experiments (II) ................................................... 129 4.2.2. DR (distributive reading) ....................................................................................... 130 4.2.2.1. Avoiding a 'quirky' case ............................................................................. 136 4.2.2.1.1. Experimental design .................................................................... 136 vii 4.2.2.1.2. Results of the experiments ........................................................... 138 4.2.2.2. LF c-command ........................................................................................... 140 4.2.2.2.1. Experimental design .................................................................... 140 4.2.2.2.2. Results of the experiments ........................................................... 142 4.3. Hypotheses on Negative Sensitive Elements (NSE) ....................................................... 144 4.3.1. NSE in Japanese .................................................................................................... 144 4.3.1. Arguments for the condition on Rokuna-N ........................................................... 148 4.4. Summary ......................................................................................................................... 153 Chapter 5: The position of the focus phrase .............................................................. 155 5.1. Introduction ..................................................................................................................... 155 5.2. A proposal: a base-generation analysis of CM-Cleft ...................................................... 155 5.3. BVA in the CM-Clefts .................................................................................................... 158 5.3.1. C-command relation between the focus phrase and the presuppositional CP ....... 158 5.3.2. ok Schema-based prediction fails to be confirmed .................................................. 162 5.3.3. Modification of the analysis .................................................................................. 164 5.3.3.1. Focus position as an A'-position ............................................................... 164 5.3.3.2. Classification of informants ...................................................................... 170 5.3.4. Dependent term in the focus position .................................................................... 174 5.3.5. Kizu’s (2005) analysis revisit ................................................................................ 180 5.4. DR (distributive reading) in the CM-Clefts .................................................................... 182 5.4.1. Preliminary: DR based on c-command .................................................................. 182 5.4.2. QP in the focus position as A of DR(A, B) ........................................................... 186 5.4.2.1. Experimental design ................................................................................... 186 5.4.2.2. Results of the experiment ........................................................................... 189 5.4.3. QP in the presuppositional CP as A of DR(A, B) ................................................. 191 5.4.3.1. Experimental design ................................................................................... 191 5.4.3.2. Results of the experiment ........................................................................... 194 5.4.4. Summary of the DR-based argument .................................................................... 197 5.5. What the results tell about the structure of the CM-cleft construction?.......................... 197 5.6. Summary ......................................................................................................................... 200 Chapter 6: Base-generation analysis over Movement analysis ............................. 202 6.1. Introduction ..................................................................................................................... 202 6.2. Where do the differences arise between the two analyses? ............................................ 205 6.3. NSEs in CM-Cleft under the movement analysis ........................................................... 209 6.3.1. Predictions in the movement analysis: Takahashi 2006 ........................................ 209 6.3.2. Some possible modifications ................................................................................. 212 6.4. NSEs in CM-Cleft under the base-genaration analysis ................................................... 216 6.4.1. Predictions in the base-generation analysis ........................................................... 216 6.4.2. Experiment and its results ..................................................................................... 218 6.5. Resumption in CM-Cleft ................................................................................................. 225 6.5.1. Preliminary ............................................................................................................ 225 6.5.2. Possibility of resumption in CM-Cleft .................................................................. 229 viii 6.5.3. Resumption in the Surface OS-type ...................................................................... 232 6.6. Summary ......................................................................................................................... 234 Chapter 7: Concluding remarks ................................................................................... 237 7.1. Summary of the discussion ............................................................................................. 237 7.2. The significance of this work .......................................................................................... 242 Bibliography ......................................................................................................................... 245 ix List of Tables Table 3.1: A summary of the results of EPSA [1]-1, 2, 3 and 18 in regard to 98 Schema set (26) Table 4.1: A summary of the results of EPSA [10]-2 116 Table 4.2: A summary of the results of EPSA [1]-1 123 Table 4.3: A summary of the results of EPSA [1]-2 123 Table 4.4: A summary of the results of EPSA [1]-3 124 Table 4.5: A summary of the results of EPSA [1]-18 124 Table 4.6: A summary of the results of EPSA [1]-4 129 Table 4.7: A summary of the results of EPSA [1]-5 129 Table 4.8: A summary of the results of EPSA [1]-6 130 Table 4.9: A summary of the results of EPSA [13]-1 138 Table 4.10: A summary of the results of EPSA [13]-4 138 Table 4.11: A summary of the results of EPSA [13]-1 (adjusted) 139 Table 4.12: A summary of the results of EPSA [13]-4 (adjusted) 140 Table 4.13: A summary of the results of EPSA [13]-2 142 Table 4.14: A summary of the results of EPSA [13]-5 143 Table 5.1: A summary of the results of EPSA [25]-1 169 Table 5.2: A summary of the results of EPSA [25]-1 by group 173 Table 5.3: A summary of the results of EPSA [25]-2 178 Table 5.4: A summary of the results of EPSA [25]-2 by group 178 Table 5.5: A summary of the results of EPSA [25]-4 190 Table 5.6: A summary of the results of EPSA [25]-4 by group 190 Table 5.7: A summary of the results of EPSA [25]-5 195 x Table 5.8: A summary of the results of EPSA [25]-5 by group 196 Table 6.1: A summary of the results of EPSA [25]-3 222 Table 6.2: A summary of the results of EPSA [25]-3 (dative focus vs. nominative focus) 223 Table 6.3: A summary of the results of EPSA [25]-3 224 (dative vs. nominative focus by group) xi Abstract The aim of this thesis is to examine the cleft constructions in Japanese, specifically, the CM-Cleft in Japanese, in the hope that it will provide insight into the properties of the Computational System, which is assumed to be at the core of the language faculty. I pursue a base-generation analysis of CM-Cleft in this thesis. I adopt the EPSA (Evaluation of Predicted Schematic Asymmetries) method advocated by Hoji (2009 and subsequent works), which is a method of hypothesis testing in line with the hypothetico-deductive method and it aspires to obtain testability, reproducibility and quantitative analyzability in language faculty science. Chapter 2 provides a review of previous works on the CM-Cleft in Japanese. The analyses of the CM-Cleft in Japanese proposed in the literature are divided into two types based on how the focus phrase gets to its surface position: "base-generation analysis" and "movement analysis" I present problems posed by each analysis and issues to be addressed. The discussion in this chapter leads to the adoption of the base-generation analysis proposed in Hoji 1987 and his subsequent works. Chapter 3 introduces the EPSA method, which I will adopt throughout this thesis. I will summarize the methodology and explain some of its important concepts that are crucially relevant to this thesis. Chapter 4 provides some schematic asymmetries independently established in other works that could be regarded as having been confirmed, together with the relevant experimental designs and results. They will be the basis for the experiments on CM-Cleft conducted in the following chapters. Chapter 5 presents a base-generation analysis of CM-Cleft pursued in this thesis in line with Hoji 1987, 1990. Based on the analysis, I will make predictions in regard to BVA (Bound xii Variable Anaphora) and DR (Distributive Reading) and conduct experiments. The results of the experiments are in favor of the proposed analysis. The proposed analysis, however, cannot account for all the BVA facts, so that I will make some modifications on the proposed analysis by introducing two new hypotheses. I will then show that the predicted schematic asymmetries about BVA obtained based on these two hypotheses get confirmed. Thus, the results of the experiments provide support for the proposed analysis. Chapter 6 provide two arguments for the base-generation analysis over the movement analysis of the CM-Cleft, which are based on NSE (Neg(ation)-sensitive elements) "licensing" and possibility of resumption in CM-Cleft. I will show that the base-generation analysis proposed in this work, but not the movement analysis, can account for these phenomena, thereby giving support to the proposed analysis. Chapter 7 concludes the discussion, recapitulating the base-generation analysis of the CM-Cleft proposed in this thesis. I also discuss the significance of this work in the light of the pursuit in generative grammar. 1 Chapter 1: Introduction 1.1 Cleft constructions Cleft constructions in English have been widely discussed in the generative literature. 1 English has two types of cleft constructions, which are exemplified in (1). The type in (1a) is called (it-)cleft, while the one in (1a) is called pseudocleft. (1) a. It was the movie that Mary saw. (it-)cleft b. What Mary saw was the movie. pseudocleft In both (it-)cleft and pseudocleft, the 'dislocated' phrase (i.e., the movie in (1)) is interpreted as focus, while the CP that follows or precedes it (i.e., that Mary saw in the (it-)cleft and What Mary saw in the pseudocleft) is interpreted as presupposition. Heggie (1988) extensively discuss cleft constructions in English, both (it-)cleft and pseudocleft. She divides the analyses of (it-)cleft proposed in the literature into three types as shown in (2). (2) a. Movement analysis from a pseudocleft base-structure: Akmajian 1970, Emonds 1976 b. Base-generation analysis with internal movement: Pinkham & Hankamer 1975, Chomsky 1977, Rochemont 1986 c. No-movement analysis: Delahunty 1982 1 They include Akmajian 1970, Chomsky 1970, 1977, Pinkham & Hankamer 1975, Emonds 1976, Higgins 1979, Rochemont 1986, Haggie 1988 and Heycock & Kroch 1996 among others. 2 Heggie (1988) then proposes an analysis of (it-)cleft based on her proposed copular structures and extends the analysis to pseudocleft, providing a unified analysis of the cleft constructions in English. As for the analysis of (it-)cleft, Heggie (1988) bases her analysis on Chomsky 1997. Chomsky’s 1997 analysis of (it-)cleft is schematized in (3). (3) (based on Chomsk 1977: 92-95, (72) and (85)) it is [ S" [ TOP this book] [ S' [ COMP what that] [ IP … t …]]] Chomsky (1977:94) suggests referring to Chomsky 1974 that clefts sentences be derived from a structure in which the focused phrase is base-generated in the predicate position of the matrix sentence rather than by a movement rule. The associated proposition is then formed by wh-movement of what to COMP, which is deleted at PF. Based on (3), Heggie (1988) proposed an analysis of (it-)cleft as shown in(4). (4) (Heggie 1998:192 (8)) [ IP it [ VP be [ CP XP i [ CP Op i that i [ IP … e i … ]]]]] Haggie (1988) analyzes (it-)cleft as a copular structure, where the raising verb be takes a CP small clause complement. Haggie (1988:189) also assumes in line with Barss 1984 that the wh-movement assumed in Chomsky 1977 is a null operator movement. 2 2 Haggie (1988:189) points out that Jaegli (1982) argues that the wh-movement assumed in Chomsky 1997 is PRO to capture the fact that an overt wh-operator rarely appears in these structures. However, Haggie (1988:189) assumes in line with Barss 1984 that the operator involved is a null operator because this operator can never be arbitrary in reference. Haggie then argues that this assumption will allow us to differentiate between the sentences in (i) where an overt operator yields a very different result from a non-overt operator, as well as prevent us from having to stipulate obligatory deletion of a wh-operator in Comp. 3 Haggie (1988) then extends the analysis of (it-)cleft in (4) to pseudocleft and proposes the structure in (5). (5) (based on Heggie 1988: 373, (155)) [ CP [ CP wh i … t i … ] j [ IP be k XP i [ VP t k [ CP t i t j ]]] Pseudocleft sentences also have the copular structure with a CP small clause identical with the one in (it-)cleft except for the fact that pseudoclefts base-generate an overt wh-operator instead of a null one in (it-)cleft. Under Heggie’s (1988) analysis of cleft constructions, in both (it-)cleft and pseudocleft, the focus phrase and the presuppositional clause establish a predication relation by means of the movement of an operator, either null or overt, which has the same index as the focus phrase. Japanese has a construction that functions like English cleft constructions as exemplified in (6), which structurally resembles the pseudocrefts in English in that the presuppositional clause preceeds the focus phrase. (6) Taroo-ga atta no-wa Hanako(-ni) da/desu Taroo-Nom met NO-Top Hanako-Dat is 'It was Hanako that Taroo met.' In (6), the presuppositional clause (i.e., Taroo-ga atta no-wa) is followed by the focus phrase (i) (Haggie 1998:190, (6)) a. [How sick] is John [t]? b. *It’s [sick] [OP that [John is [t]]]. 4 Hanako(-ni) and then the copula verb da/desu. The presuppositional clause is followed by no, which is then marked with topic marker wa. 3 The focus phrase, which follows the presuppositional clause, is either case-marked or not. Following the terminology in Hoji & Ueyama 2003, I refer to the cleft sentences with the case-marker on the focus NP as CM-Cleft and those without the case-marker as Non-CM-Cleft. 4 Just like English cleft constructions, cleft constructions in Japanese have been a topic widely discussed in the generative literature in Japanese (Hoji 1987, 1990, Koizumi 1995, Kuwabara 1995, 1996, Hasegawa 1997, Hiraiwa & Ishihara 2002, Ueyama & Hoji 2003, Kizu 2005 and Takahashi 2006 among others). However, there is still no consensus about the proper syntactic analysis of the constructions. We may thus ask whether the Japanese clefts involve the same operation(s) as English clefts, and if yes, which clefts, (it-)cleft or pseudocreft. Besides the similarity between English clefts and Japanese clefts, it is observed in Hoji & Ueyama (2003) that there are similarities between scrambling constructions and cleft constructions in Japanese. We may also ask whether these similarities are just apparent or coincidental, or they are in fact based on some grammatical properties shared by the construction in question. Furthermore, the analyses of the cleft constructions in Japanese are roughly classified into two types depending on how the focus phrase is generated in the structure. Which analysis to choose is yet another important question that needs to be answered. 3 No in Japanese has several functions as shown in (i). (i) a. Prenominal no that corresponds to a genitive marker in English. b. Pronominal no that corresponds to English 'one'. c. Complementizer no that nominalizes the preceding clause. As we will see shortly, I assume that no in CM-Cleft is a complementizer that takes the presuppositional clause as its complement. 4 Hiraiwa & Ishihara (2002), on the other hand, refer to the former as Cleft and the latter as pseudo-Cleft. 5 1.2 Issues 1.2.1 The cleft construction and 'scrambling' 1.2.1.1 Similarities between the OS-type construction and the cleft construction Hoji & Ueyama (2003) discuss four constructions in Japanese in regard to resumption; cleft, scrambling, relativization and topicalization, as in Hoji 1987, 1990: chapter 5. 5 They further divide cleft constructions into C(ase-)M(arked)-Cleft and non-CM-Cleft depending on whether the focus phrase takes a case-particle or not, and classify CM-cleft, non-CM cleft, scrambling, relativization and topicalization into two groups based on the presence/absence of subjacency effects; CM-Cleft and scrambling on the one hand (Group I), and non-CM-Cleft, relativization and (non-CM-)topicalization on the other (Group II). The former exhibit subjacency effects, while the latter do not as shown in (7) and (8), respectively. (7) (Hoji & Ueyama 2003: 3, (8) without resumption) a. CM-Cleft: * [John-ga [a-no ban ec 1 houkasita otoko]-o otteiru no]-wa John-NOM that-GEN night set:fire man-ACC looking:for COMP-TOP [ ko-no gekizyoo-ni] 1 da. this-GEN theater-DAT COPULA 'It is [to this theater] 1 that John is looking for the person who set fire ec 1 that night.' 5 As for topicalization, Saito 1985 makes a distinction between PP topicalization and NP topticalization. Hoji 1990: Chapter 5 also makes the same distinction. Since the distinction is based on the presence or absence of the case-marker on the topic phrase, they can be rephrased as CM-topicalization and non-CM-topicalization, respectively, in line with the terminology employed in Hoji & Ueyama 2003 for the cleft construction. In Japanese, CM vs. non-CM distinction plays an important role; CM/non-CM-sluicing in Fukaya & Hoji 1999, and CM/non-CM-stripping and CM/ non-CM-comparative in Hoji 2003b. Topicalization discussed in Hoji & Ueyama 2003 is the non-CM-topicalization. 6 b. Scrambling: *? [Ko-no gekizyoo-ni] 1 [John-ga [a-no ban ec 1 houkasita otoko]-o otteiru this-GEN theater-DAT John-NOM that-GEN night set:fire man-ACC looking:for '[To this theater] 1 , John is looking for the person who set fire ec 1 that night.' (8) (Hoji & Ueyama 2003: 3, (9) without resumption) a. Relativization: [John-ga [ NP ec 1 itta koto-ga aru hito]-o sagasiteiru] mati 1 -ga John-NOM went fact-NOM exist person-ACC looking:for town-NOM terebi-de syookais-areteita. TV-in introduced-was 'The town 1 that John is looking for people who have been there 1 was featured in a TV program.' b. Topicalization: [Pekin-wa] 1 John-ga [ NP ec 1 itta koto-ga aru hito]-o mituketa rasii. Peking-TOP John-NOM went fact-NOM exist person-ACC found seem '[As for Peking] 1 , (it seems that) John found a person who has been there 1 .' c. Non-CM-Cleft: [John-ga [ NP [ S ec ec 1 kaita] hito]-ni aitagatteiru no]-wa [a-no hon] 1 da. John-NOM wrote person-DAT want:to:see COMP-TOP that-GEN book COPULA 'It is [that book] 1 that John wants to see the person who wrote it 1 .' Hoji & Ueyama (2003) note, citing earlier works such as Hayashishita 1997, Ueyama 1998, that all these five constructions allow resumption in a 'local' context as shown in (9). 7 (9) (Hoji & Ueyama 2003: 2, (7)) a. CM-Cleft: [Kokuren-ga kibisiku [so-re-o] 1 hihansita no]-wa [amerika-no the:United:Nations-NOM harshly that-thing-ACC criticized COMP-TOP USA-GEN booei-seisaku-o] 1 da. defense-policy-ACC COPULA 'It was [the USA's defense policy-ACC] 1 that the United Nations harshly criticized it 1 .' b. Scrambling (Hayashishita 1997, Ueyama 1998): [Toyota-ni-sae] 1 Nissan-ga [so-ko-ni] 1 syatyoo-to-no mendan-o moosiiretekita. Toyota-DAT-even Nissan-NOM that-place-DAT president-with-GEN appointment-ACC requested '[Even to Toyota] 1 , Nissan applied [to it] 1 for an appointment with the president.' c. Relativization: [John-ga [[so-ko-ni] 1 Mary-o tureteitta] mise 1 -wa moo tubureteiru. John-NOM that-place-to Mary-ACC took restaurant-TOP already went:bankrupt 'The restaurant 1 [that John took Mary there 1 ] has already gone bankrupt.' d. Topicalization [Daietto to onsen-wa] 1 [syuukansi-ga [neta-ni tumaru to] yoku diet and spa-TOP magazine-NOM topic-DAT stuck if often [so-re-o] 1 tokusyuusuru]. that-thing-ACC feature '[As for diet and hot spa] 1 , [magazines often feature [it/them] 1 [when they stuck with topics]].' e. Non-CM-Cleft: [Syuukansi-ga [neta-ni tumaru to] yoku [so-re-o] 1 tokusyuusuru no]-wa magazine-NOM topic-DAT stuck if often that-thing-ACC feature COMP-TOP 8 [daietto to onsen] 1 da. diet and spa COPULA 'It is [diet and hot spa] 1 that magazines often feature [it/them] 1 [when they stuck with topics].' However, when we replace ec with resumption in the subjacency context in (7) and (8), CM-Cleft and scrambling do not allow resumption, while Non-CM-Cleft, relativization and topicalization do, as shown in (10) and (11), respectively. Thus, the classification by Hoji & Ueyama (2003) holds in the possibility of resumption in the subjacency context, too. (10) (Hoji & Ueyama 2003: 3, (8) with resumption) a. CM-Cleft: * [John-ga [a-no ban so-ko-ni 1 houkasita otoko]-o otteiru no]-wa John-NOM that-GEN night that-place-to set:fire man-ACC looking:for COMP-TOP [ ko-no gekizyoo-ni] 1 da. this-GEN theater-DAT COPULA 'It is [to this theater] 1 that John is looking for the person who set fire to it 1 that night.' b. Scrambling: *? [Ko-no gekizyoo-ni] 1 [John-ga [a-no ban so-ko-ni 1 houkasita this-GEN theater-DAT John-NOM that-GEN night that-place-to set:fire otoko]-o otteiru man-ACC looking:for '[To this theater] 1 , John is looking for the person who set fire to it 1 that night.' 9 (11) (Hoji & Ueyama 2003: 3, (9) with resumption) a. Relativization: [John-ga [ NP so-ko-ni 1 itta koto-ga aru hito]-o sagasiteiru] mati 1 -ga John-NOM that-place-to went fact-NOM exist person-ACC looking:for town-NOM terebi-de syookais-areteita. TV-in introduced-was 'The town 1 that John is looking for people who have been there 1 was featured in a TV program.' b. Topicalization: [Pekin-wa] 1 John-ga [ NP so-ko-ni 1 itta koto-ga aru hito]-o mituketa rasii. Peking-TOP John-NOM that-place-to went fact-NOM exist person-ACC found seem '[As for Peking] 1 , (it seems that) John found a person who has been there 1 .' c. Non-CM-Cleft: [John-ga [ NP [ S ec so-re-o 1 kaita] hito]-ni aitagatteiru no]-wa [a-no hon] 1 da. John-NOM that-thing-ACC wrote person-DAT want:to:see COMP-TOP that-GEN book COPULA 'It is [that book] 1 that John wants to see the person who wrote it 1 .' Based on the observation, Hoji & Ueyama (2003) claim that the constructions in Group I (CM-Cleft and scrambling) involve an empty operator movement, while the ones in Group II (relativization, topicalization and Non-CM-Cleft) do not involve any movement. In short, in Hoji & Ueyama 2003 CM-Cleft and scrambling construction are classified into the same group in regard to the subjacency effects and the resumption possibility, and are assumed to involve the empty operator movement. Let us then present the analysis of scrambling construction in Ueyama 1998 in order to make further comparison between CM-Cleft 10 and scrambling construction. 1.2.1.2 The OS-type construction in Ueyama 1998 1.2.1.2.1 Two types of the OS-type construction: Surface OS-type and Deep OS-type Scrambling construction refers to the word order in (12b), which is regarded as 'marked' as opposed to the 'unmarked order' in (12a). (12) a. NP-NOM NP-ACC/DAT V b. NP-ACC/DAT NP-NOM V 6 Ueyama (1998) calls the scrambling construction in (12b) OS-type construction (i.e., Object-Subject word order construction) in contrast to SO-type construction (i.e., Subject-Object word order construction) in (12a). 7 In addition, Ueyama (1998) refers to the 'scrambled' object, that is, the NP-ACC/DAT which occurs before NP-NOM in the OS-type construction in (12b) as DL (i.e., a DisLocated NP). It has been known that in the OS-type construction the DL shows A-properties as well as A'-properties. Let us consider bound variable anaphora interpretation to see this point. Ueyama (1998:2) uses the term BVA to refer to any anaphoric relation between a 6 Some verbs in Japanese such as iru 'need' take a ni-marked phrase as the subject and a ga-marked phrase as the object. For such verbs, the word order in (12b) is regarded as 'unmarked order,' while (12a) as 'marked order'. 7 Ueyama (1998:23) mentions as follows; "Although the name scrambling is often used to refer to the construction in [(12b)], this is originally the name for the movement rule which derives [(12b)] from [(12a)]. In order to describe the relevant facts in an analysis-neutral way, we will call the configuration … such as in [(12b)] …. an OS-type construction." 11 distributive/quantificational expression (QP) and a singular-denoting anaphoric expression. 8 She refers to BVA between a QP and a singular-denoting anaphoric expression β as BVA (QP, β). 9 An initial condition on BVA is informally stated as in (13). 10 (13) (Ueyama 1998:30, (13)) A dependent term in BVA must be c-commanded by the (QR-)trace of a QP at LF. In SO-type constructions as in (14), BVA (dare, so-itu) is difficult to obtain, exhibiting so-called weak crossover (WCO) effects (Postal 1971). 11 In (14) the QP dare-o is raised at LF either by wh-movement or QR crossing the variable so-itu that the QP should bind. This 8 Ueyama (1998:2) uses the term BVA purely descriptively in order to use the term bound variable anaphora to refer to the relation between the two instances of the same bound variable. 9 BVA(A, B) expresses an intuition that (i) B does not have an inherent value of its own, and (ii) the value of B co-varies with the value of A. (E.g., BVA(every boy, his) in every boy loves his mother) 10 Ueyama (1998) in fact presents three sources of the BVA interpretation and the condition in (13) is only for one of the three sources of the BVA interpretation. (13) is still an informal statement and is restated by Ueyama (1998) as a necessary condition on the establishment of Formal Dependency. 11 The following contrast illustrates WCO effects in English. (i) (Ueyama 1998:29, (12)) a. Every student hit his best friend. b. *His best friend hit every student. c. ?*Who i did [his best friend] hit t i ? Ueyama (1998:30) then mentions that BVA can be established in (ii), despite the fact that t i in the θ-position does not c-command the dependent term. (ii) (Ueyama 1998:30, (14)) a. Every daughter i seems [to her father] t i to be beautiful. b. Who i t i seems [to his mother] t i to have come? In other words, raising in English does not induce WCO effects, while wh-movement in English does. This observation is often understood in a generalized form as in (iii). (iii) (Ueyama 1998:30, (15)) a. Movement to an A-position does not induce WCO effects. b. Movement to an A'-position induces WCO effects. 12 movement or resulting configulation induces WCO effects. (14) (Ueyama 1998:31, (17)) ?* [So-itu-no hahaoya]-ga dare-o aisiteru no that-guy-GEN mother-NOM who-ACC love COMP '?*Who does his mother love?' In the same light, OS-type constructions as in (15) should induce WCO effects, given that the QP Toyota-sae-o were moved to the sentence-intial position by "scrambling" crossing the variable so-ko that the QP should bind. However, contrary to the expectation, (15) does not induce WCO effects, a well known observation at least since the late 1980. (15) (Ueyama 1998:31, (18b)) Toyota-sae-o i [so-ko-o tekitaisisiteiru kaisya]-ga ec i uttaeta. Toyota-even-ACC that-place-ACC be:hostile company-NOM sued '[Even Toyota] i , [the company which is hostile to it] sued ec i .' Under the general understanding that movement to an A'-position induces WCO effects, while movement to an A-position does not induce WCO effects, the absence of WCO effects in OS-type construction as in (15) is taken as a property of A-movement. Next, in OS-type constructions as in (16), BVA(Toyota-sae, so-ko) is possibe. Given the condition on BVA in (13), the DL in (16) has to be reconstructed to the base-generated position (t i in (16)) at LF in order to get BVA(Toyota-sae, so-ko). Reconstruction effects are generally taken as a property of A'-movement. 13 (16) (Ueyama 1998:33, (23)) [So-ko-o tekitaisisiteiru kaisya] i -o Toyota-sae-ga t i uttaeta that-place-ACC be:hostile company-ACC Toyota-even-NOM sued '[The company which is hostile to it] i , even Toyota sued t i .' Thus, the OS-type construction allows BVA (QP, so-ko) in both of the configurations in (17a) and (17b). (17) a. "Absence of the WCO effects": QP-ACC/DAT ... [ ... so-ko ... ]-NOM ... V b. Reconstruction effects: [ ... so-ko ... ]-ACC/DAT ... QP-NOM ... V As noted the absence of WCO effects is a representative property of A-movement, while reconstruction effects are generally understood to be a property of A'-movement. We thus do not expect the two effects to obtain at the same time. Ueyama 2003 observes that this expectation is in fact borne out, as shown in (18). (18) (Ueyama 2003:17, (47)) ?* [So-ko i -no kaikeisi-o-sae] j [so-itu j -no kookoo-no sensei]-ga that-place-GEN accountant-ACC-even that-guy-GEN high:school-GEN teacher-NOM [subete-no zidoosyagaisya] i -ni ec j suisensita. every-GEN automobile:company-DAT recommended 14 '[Even its i accountant] j , [his j high school teacher] recommended to [every automobile company] i .' Intended (but impossible) interpretation: ∀x (x=automobile company) [it holds with even y, who is x's accountant, that y's high school teacher recommended y to x ] Based on the above observation, Ueyama (1998) argues that the OS-type construction is structurally two-way ambiguous. She calls the type in (17a) that does not show the WCO effects "Deep OS-type" and the one in (17b) that shows reconstruction effects "Surface OS-type." The structures she assigns to "Surface OS-type" and "Deep OS-type" are shown in (19) and (20), respectively. (19) (Ueyama 1998: 63, (92)) Surface OS-type: PF: NP 1 -DAT/ACC NP-NOM ... t 1 ... LF: NP-NOM NP-DAT/ACC ... (20) (Ueyama 1998: 63, (91)) Deep OS-type: PF: NP 1 -DAT/ACC NP-NOM ec 1 ... LF: NP 1 -DAT/ACC Op 1 [ NP-NOM t 1 ... In Surface OS-type the dislocated object is moved at PF across the subject, while in Deep OS-type the object is 'base-generated' in a position sister to an IP, which is considered to be an A-position. Ueyama (1998) refers to the DL in Surface OS as Surface DL and the one in Deep 15 OS as Deep DL. Deep DL is related to 'its theta-position' by means of (i) the 'base-generation' of Op in 'the theta position' with which the Deep DL is related, (ii) the LF adjunction of the Op to the IP, and (iii) the formation of the predication relation between the DL and the IP with the DL functioning as the subject. 12 1.2.1.2.2 The OS-type construction and resumption Ueyama (1998) presents (16) (repeated here as (21)) as an example of the OS-type construction that shows reconstruction effects. Thus, in (16) BVA(Toyota-sae, so-ko) is possible. (21) (=(16)) (Ueyama 1998:33, (23)) [So-ko-o tekitaisisiteiru kaisya] i -o Toyota-sae-ga t i uttaeta that-place-ACC be:hostile company-ACC Toyota-even-NOM sued '[The company which is hostile to it] i , even Toyota sued t i .' However, Hoji & Ueyama (2003) observe that when we employ resumption in (21), the intended BVA becomes impossible as shown in (22). (22) (Hoji & Ueyama 2003:6, (22a)) * [So-ko-o tekitaisisiteiru kaisya] i -o Toyota-sae-ga [so-ko-o] i uttaeta that-place-ACC be:hostile company-ACC Toyota-even-NOM that-place-ACC sued '[The company which is hostile to it] i , even Toyota sued it i .' 12 Hoji (to appear: 5) assumes that Op in (20) is an empty nominal that is represented as pro, and it acts as an empty operator. Then, the pro in Hoji to appear is a notational variant of Op in (20) as far as the empirical consequences are concerned. He also mentions that the formation of the predication relation is between the object and the λ predicate that the IP gets mapped to, much as in the case of the subject of the tough construction in English. 16 On the other hand, when we employ resumption in (15) (repeated here as (23)), the intended BVA continued to be available, as shown in (24). (23) (=(15)) (Ueyama 1998:31, (18b)) Toyota-sae-o i [so-ko-o tekitaisisiteiru kaisya]-ga ec i uttaeta. Toyota-even-ACC that-place-ACC be:hostile company-NOM sued '[Even Toyota] i , [the company which is hostile to it] sued ec i .' (24) Toyota-sae-o i [so-ko-o tekitaisisiteiru kaisya]-ga so-ko-o i uttaeta. Toyota-even-ACC that-place-ACC be:hostile company-NOM that-place-ACC sued '[Even Toyota] i , [the company which is hostile to it] sued it i .' Notice that (16) is a Surface OS-type, which allows reconstruction effects while (15) is a Deep OS-type. On the basis of the incompatibility between reconstruction effects and resumption in the OS construction, Hoji & Ueyama (2003) conclude that Surface OS-type does not allow resumption. The acceptability of examples like (24) shows that Deep OS-type, unlike the Surface OS type, allows resumption. Thus, the similarity between scrambling and CM-Cleft in regard to the resumption possibility observed in Hoji & Ueyama 2003 turns out to be between Deep OS-type and CM-Cleft. We have seen the similarities between scrambling and CM-Cleft in regard to the subjacency effects and the resumption possibility. These similarities would urge or at least tempt us to pursue a unified analysis of the two constructions. However, at the same time, there are some structural differences in these two constructions, specifically, the location of the dislocated phrase 17 and what intervene between the dislocated phrase and the rest of the sentence. Moreover, these two constructions behave differently with respect to the possibility of BVA in some circumstances. When we discuss the structure of cleft construction in Japanese, we need to take both of these similarities and differences into consideration, i.e., the analysis of cleft construction to be proposed must account for all of these properties. 1.2.1.3 Apparent differences between the OS-type construction and the cleft construction 1.2.1.3.1 Structural differences between the two constructions As mentioned above, based on the similarities between scrambling, more specifically Deep OS-type, and CM-Cleft in regard to subjacency effects and the resumption possibility, Hoji & Ueyama (2003) claim that both constructions involve an empty operator movement. Recall the structure of Deep OS-type proposed in Ueyama 1998 in (20). (20) (Ueyama 1998: 63, (91)) Deep OS-type: PF: NP 1 -DAT/ACC NP-NOM pro 1 ... LF: NP 1 -DAT/ACC Op 1 [ NP-NOM t 1 ... Ueyama (1998) assumes that the DL NP-DAT/ACC is base-generated in a position sister to an IP, which is an A-position. Predication relation is then formed between the DL and the IP, mediated by the LF adjunction of the Op to the IP. On the other hand, the structure of CM-Cleft presented in Hoji & Ueyama (2003) is the one proposed in Hoji 1987, which is shown in (25). 13 In CM-Cleft, it is commonly assumed that what precedes the topic marker wa is a CP, no being a functional head C. 13 The details of the analyses of CM-Cleft proposed in the literature will be presented in Chapter 3. 18 (25) (Hoji & Ueyama 2003:4, (12a) referring to Hoji 1987) [ Op 1 [[ ... t 1 ... ] no]]-wa/ga [NP-CM] 1 da In Deep OS-type, the predication relation is formed between the dislocated NP (DL) and the rest of the sentence, that is, the IP that the DL is a sister of. If we assume that the derivation of CM-Cleft also involves the formation of a predicate relation, it should be between the focus phrase and the CP that precedes the topic marker wa. Comparing the predication relations in the two constructions, we can immediately notice two apparent structural differences. First, in Deep OS-type, the DL that is predicated of the IP precedes the IP in the surface structure, while in CM-Cleft, the focus phrase that should be predicated of the presuppositional CP follows the CP. Second, in Deep OS-type, there is an IP boundary between the DL and the empty operator Op, while in CM-Cleft, there is a CP boundary between the focus phrase and Op. If we want to attribute the similarities between Deep OS-type and CM-Cleft in regard to subjacency effects and the resumption possibility to the involvement of the empty operator movement in the derivation of the two constructions and the resulting predication relation between the dislocated NP and the rest of the sentence, these two apparent structural differences between the two constructions would have to be accounted for. 1.2.1.3.2 Empirical differences between the two constructions Hoji & Ueyama (2003) observe two differences in behavior between Deep OS-type and CM-Cleft. First, they observe that in the case of Deep OS-type, BVA (A,B) is available when A is the DL and B is contained in the rest of the sentence, but in the case of CM-Cleft, BVA (A,B) is impossible when A is the focus phrase and B is contained in the presuppositional clause. For 19 instance, in Deep OS-type sentences presented in (15) above, repeated here as (26a), and (26b), BVA(Toyota-sae, so-ko) is possible, while it is not possible in the corresponding CM-Cleft sentences in (27a) and (27b). (26) a. (=(15)) (Ueyama 1998:31, (18b)) Toyota-sae-o i [so-ko-o tekitaisisiteiru kaisya]-ga ec i uttaeta. Toyota-even-ACC that-place-ACC be:hostile company-NOM sued '[Even Toyota] i , [the company which is hostile to it] sued ec i .' b. (Hoji & Ueyama 2003: 6, (23) withoiut resumption) Toyota-ni-sae i [so-ko-o tekitaisisiteiru kaisya]-ga ec i Nissan-o suisensita Toyota-DAT-even that-place-ACC be:hostile company-NOM Nissan-ACC recommended '[To even Toyota] i , the company which is hostile to it recommended Nissan ec i .' (27) (Hoji & Ueyama 2003: 11, (39a)) a. * [[So-ko-o tekitaisisiteiru kaisya]-ga ec 1 uttaeta no]-wa [Toyota-sae-o] 1 da. that-place-ACC be:hostile company-NOM sued COMP-TOP Toyota-even-ACC COPULA 'It was [even Toyota] 1 that [the company which is hostile to it ] sued ec 1 .' b. * [[So-ko-o tekitaisisiteiru kaisya]-ga ec 1 Nissan-o suisensita no]-wa [Toyota-ni-sae] 1 da that-place-ACC be:hostile company-NOM Nissan-ACC recommended COMP-TOP Toyota-DAT-even COPULA 'It was [to even Toyota] 1 that [the company which is hostile to it] recommended Nissan ec 1 .' Second, Hoji & Ueyama (2003) observe that Deep OS-type does not exhibit the reconstruction effects, while CM-Cleft does. As mentioned above, Hoji & Ueyama’s (2003) claim is that Surface OS-type does not allow resumption. Thus, with the use of resumption, we 20 can force an OS-type construction to be Deep OS-type. That is the account in Hoji & Ueyama 2003 of why BVA (Toyota-sae, so-ko) is impossible in (22) above, repeated here as (28a) and (28b). (28) (Hoji & Ueyama 2003:6, (22)) a. * [So-ko-o tekitaisisiteiru kaisya] 1 -o Toyota-sae-ga [so-ko-o] 1 uttaeta that-place-ACC be:hostile company-ACC Toyota-even-NOM that-place-ACC sued '[The company which is hostile to it] 1 , even Toyota sued it 1 .' b. * [So-ko-o tekitaisisiteiru kaisya] 1 -o Nissan-ga Toyota-ni-sae [so-ko-o] i suisensita that-place-ACC be:hostile company-ACC Nissan-NOM Toyota-DAT-even that-place-ACC recommended '[the company which is hostile to it ] 1 , Nissan recommended to even Toyota it 1 .' On the other hand, BVA (Toyota-sae, so-ko) is possible in the corresponding CM-Cleft as shown in (29a) and (29b). (29) (Hoji & Ueyama 2003:12, (41) with resumption) a. [Toyota-sae-ga so-ko-o 1 uttaeta no]-wa [so-ko-o tekitaisisiteiru kaisya-o] 1 da. Toyota-even-NOM that-place-ACC sued COMP-TOP that-place-ACC be:hostile company-ACC COPULA 'It was [the company which is hostile to it] 1 that even Toyota sued ec 1 .' b. [Nissan-ga Toyota-ni-sae so-ko-o 1 suisensita no]-wa Nissan-NOM Toyota-DAT-even that-place-ACC recommended COMP-TOP [so-ko-o tekitaisisiteiru kaisya-o] 1 da. that-place-ACC be:hostile company-ACC COPULA 'It was [the company which is hostile to it] 1 that Nissan recommended to even Toyota ec 1 .' 21 Hoji & Ueyama (2003) then suggest that theese differences between Deep OS-type and CM-Cleft stem from the structural position of the DL in the two constructions: the DL in the Deep OS-type is in an A-position, while that in CM-Cleft is in an A'-position. I will present the detail of their analysis in the next chapter. Again, the similarities between Deep OS-type and CM-Cleft in regard to subjacency effects and the resumption possibility is compatible with the assumption that both constructions involve an empty operator movement. If we want to pursue the similarities between the two constructions further, we have to simultaneously account for these differences in regard to BVA possibilities, or at least the analysis should not be incompatible with these facts. 1.2.2 The two types of analyses of the Japanese cleft construction There are two main types of analyses of cleft constructions in Japanese proposed in the literature. The classification is based on how the focus phrase gets to its surface position: One type base-generates the focus phrase at its surface position, which is related to the presuppositional clause by the empty operator movement within the presuppositional clause just like English cleft constructions (Hoji 1987, 1990, Hoji & Ueyama 2003, Kizu 2005, among others). The other type gets the focus phrase as well as the presuppositional clause at their surface positions by means of movement operations (Hiraiwa & Ishihara 2002, Takahashi 2006). We will refer to the former type of the analysis as "base-generation analysis" and the latter as "movement analysis." 14 One of the base-generation analyses, as proposed in Hoji 1990, is shown in (30), while the 14 As just mentioned "base-generation analysis" involves the empty operator movement within the presuppositional clause, so that it is not the case that no movement is involved in the derivation of CM-Cleft under "base-generation analyses." 22 one of the movement analyses, as proposed in Hiraiwa & Ishihara 2002, is shown in (31). 15 (30) (Hoji 1990: (225)) a. Non-CM-Cleft: [ NP [ S .. (pro i ) ... ] [ NP no i ]]-wa NP i da (An aboutness licensing is possible.) b. CM-Cleft: [ S' Op i [ S ... t i ... ] no]-wa NP i -(CASE) da (Syntactic movement is obligatory and the aboutness licensing is not possible.) (31) (cf. H&I 2002: (25)-(26)) a. Underlying structure = 'No da' in-situ focus construction: [ TopP [ FocP [ FinP/CP [ TP … XP (Focus) …] no] da] Top] b. Focus movement: [ TopP [ FocP XP (Focus) [ FinP/CP [ TP … t XP …] no] da] Top] c. Remnant Topicalization: [ TopP [ FinP/CP [ TP … t XP …] no]-wa [ FocP XP (Focus) t FinP/CP da] Top] Hiraiwa & Ishihara (2002) present the following two facts as supporting evidence for the movement analysis; (i) CM-Cleft exhibits island effects as shown in (32), and (ii) the case-marker on the focus phrase is the same as the one that an NP has if it is realized at the gap position inside the presuppositional CP. (32) (Hiraiwa & Ishihara 2002: 37, (5a)) * [John-ga [[e i e j kaita] hito i ]-o hihansita-no]-wa kono-ronbun-o j da 15 (31) is the derivation of CM-Cleft, which is Cleft in Hiraiwa & Ishihara’s (2002) terms. They only provide the structure of CM-Cleft. 23 John-Nom wrote person-Acc criticized-C-TOP this paper-Acc be '(Lit.) It is this paper j that John criticized the person who wrote e j .' They argue that the island effects are observed due to the syntactic movement of the focus phrase. The case-matching effect is argued to be a natural consequence of the movement analysis since the focus phrase is in fact an argument of the predicate. Under the base-generation analysis, the operator movement inside the presuppositional clause accounts for the island effects in (32). Also, the subject-predicate relation formed between the focus phrase and the presuppositional clause through the operator movement provides a way to account for the "case-matching" under the base-generation analysis, in line with Ueyama's (1998) analysis of so-called A-scrambling. Thus, the crucial question at this point is whether there are any phenomena that would distinguish these two analyses. If there are any phenomena about which these two analyses make different predictions, we can choose between the two analyses by investigating such phenomena. 1.3 The goal of this work In this work, I confine my discussion to the CM-Cleft constructions on the ground that they are more likely to reflect the properties of the Computational System, which is assumed to be at the core of the language faculty. The goal of generative grammar is to discover the properties of the language faculty. By focusing on the CM-Cleft constructions we would be able to have a greater chance to identify the properties of the Computational System, and thus to reveal the properties of the language faculty. I pursue a base-generation analysis of CM-Cleft in this work. I adopt the EPSA 24 (Evaluation of Predicted Schematic Asymmetries) method advocated by Hoji (2009 and subsequent works). The main general thesis of the EPSA method is that it is possible to study the language faculty as an exact science. The EPSA method is a method of hypothesis testing in line with the hypothetico-deductive method and it aspires to obtain testability, reproducibility and quantitative analyzability in language faculty science. I will investigate the CM-Cleft construction utilizing the EPSA method. In order to choose one analysis over the other, for example, the base-generation analysis over the movement analysis or the base-generation analysis proposed in this work over other base-generation analyses, we need to identify where the differences reside and what hypotheses give rise to the difference. Based on the identified hypotheses, we deduce prediction(s), whose content we may call the predicted schematic asymmetry. We then conduct experiments and compare the prediction(s) with the experimental results. The predicted schematic asymmetry contains a *Schema-based prediction, which makes our hypotheses testable in the most crucial way. Examples conforming to *Schema must be judged as completely unacceptable by informants. Thus, when an informant accepts an example conforming to *Schema, the *Schema-based prediction is disconfirmed (at least in principle), so that the relevant hypotheses need to be abandoned or modified. I will conduct experiments on the proposed base-generation analysis as well as the movement analysis and another base-generation analysis, and compare the prediction(s) under each analysis with the experimental results. The results so obtained are decisive, so that when the predicted schematic asymmetry gets confirmed, we could gain strong support for the proposed analysis. In order to form predicted schematic asymmetries in an analysis, we need to have clear understanding of the hypotheses related to the Computational System, the language-specific hypotheses and the pf-LF correspondences, which state a correspondence between a surface string and the LF representation, assumed in the analysis. Although I adopt Hoji’s 1987, 1990 analysis 25 of CM-Cleft, we need to have more articulation on the structure and hypotheses on the base-generation analysis, so that we can form predicted schematic asymmetries and conduct experiments. This is another goal of this work. 1.4 Outline I will start the discussion with the review of previous works on the cleft constructions in Japanese in Chapter 2. As mentioned above, the analyses of cleft constructions in Japanese proposed in the literature are roughly divided into two types based on how the focus phrase gets to its surface position: "base-generation analysis" and "movement analysis" I will first summarize two representative analyses of each of the base-generation analysis and movement analysis; Hoji 1987 (as well as Hoji 1990, Ueyama & Hoji 2001 and Hoji & Ueyama 2003) and Kizu 2005 for the base-generation analysis, and Hiraiwa & Ishihara 2002 and Takahashi 2006 for the movement analysis. I will then present problems posed by each analysis or issues to be addressed. The discussion in this chapter leads to the adoption of the base-generation analysis proposed in Hoji 1987 and his subsequent works. In Chapter 3, I will introduce the methodology advocated by Hoji (2009 and subsequent works), which he calls the EPSA (Evaluation of Predicted Schematic Asymmetries) method. I will adopt this methodology throughout this work. I will summarize the methodology and explain some of its important concepts that are crucially relevant to the present work. Once the methodology adopted in this work is in hand, we are ready to conduct experiments on CM-Cleft. However, we first need to ensure that we are using 'valid tools' for the experiments. In Chapter 4, I will thus present some schematic asymmetries independently established that could be regarded as having been confirmed, together with the relevant experimental designs and results. Specifically, BVA-related schematic asymmetries and 26 DR-related ones will be presented. The hypotheses or pf-LF correspondences assumed in obtaining a confirmed schematic asymmetry can be used as 'valid tools' for further experiments or more complicated experiments. In addition to these, I will also present hypotheses about NSEs (Neg(ation)-sensitive elements) proposed by Kataoka (2007), which I crucially utilize in the experiments on CM-Cleft. In Chapter 5, I will first propose the base-generation analysis pursued in this work in line with the analysis of CM-Cleft proposed by Hoji (1987, 1990). I will then compare the proposed analysis with another base-generation analysis proposed by Kizu (2005), and consider what different predictions these two analyses make. Based on the difference, I will make predictions in regard to BVA (Bound Variable Anaphora) and DR (Distributive Reading) in CM-Cleft and conduct experiments. I will then examine the results of the experiments. The results of the experiments are in favor of the proposed analysis. The proposed analysis, however, cannot account for all the BVA facts, so that I will make some modifications on the proposed analysis by introducing two new hypotheses. I will then show that the predicted schematic asymmetries about BVA obtained based on these two hypotheses get confirmed. Thus, the results of the experiments provide support for the proposed analysis. However, as a result of the introduction of two hypotheses, the base-generation analysis and the movement analysis make the same predictions with respect to BVA and DR, so that we cannot distinguish between the two analyses. In order to argue for the base-generation analysis pursued in this work, then, we need to have phenomena about which the base-generation analysis and the movement analysis make different predictions. In Chapter 6, I will discuss two such phenomena, NSE (Neg(ation)-sensitive elements) "licensing" and possibility of resumption in CM-Cleft, and show that the base-generation analysis proposed in this work, but not the movement analysis, can account for these phenomena, thereby giving support to the proposed 27 analysis. Chapter 7 concludes the discussion, recapitulating the base-generation analysis of the CM-Cleft proposed in this work. I also address how the proposed analysis capture the similarities between the CM-Cleft and Deep-OS type constructions, and how it relates to the analyses of English cleft constructions. 28 Chapter 2: Two Types of Analyses of Cleft Constructions in Japanese 2.1 Introduction The analyses of cleft constructions in Japanese proposed in the literature are roughly divided into two types based on how the focus phrase gets to its surface position: "base-generation analysis" and "movement analysis." The former type of analysis base-generates the focus phrase at its surface position, which is related to the presuppositional clause by the operator movement within the presuppositional clause. The latter type gets the focus phrase as well as the presuppositional clause at their surface positions by means of movement operations. 1 In Chapter 2, I have presented two representative analyses of each of the base-generation analysis and movement analysis; Hoji 1987 (as well as Hoji 1990, Ueyama & Hoji 2001 and Hoji & Ueyama 2003) and Kizu 2005 for the base-generation analysis, and Hiraiwa & Ishihara 2002 and Takahashi 2006 for the movement analysis. In this chapter, I will first summarize each of these four analyses and then discuss the issues each analysis poses. 2.2 Base-generation analyses 2.2.1 Hoji 1987, 1990, Ueyama & Hoji 2001, Hoji & Ueyama 2003 Hoji 1987 is a precursor of the base-generation analysis of cleft constructions in Japanese. Hoji 1990, Ueyama & Hoji 2001 and Hoji & Ueyama 2003 basically follow it except for some technical details. 2 Hoji (1987), building on Saito’s (1985) distinction between "PP topics" and "NP topics," observes that Non-CM-Cleft and CM-Cleft exhibit different behavior in regard to 1 Hajime Hoji (pc in March, 2011) points out that it is logically possible to pursue an analysis where the presuppositinal clause gets base-generated but the focus phrase gets to its surface position by movement. I won’t discuss this possibility in this work and leave it for my future research. 2 Thus, the analyses of the cleft constructions proposed in these works are the same unless otherwise mentioned. 29 subjacency effects, as shown in (1). 3 (1) (Hoji 1987: (8c) and (10b), respectively) a. (based on Saito's (1985) topic examples) [John-ga [ NP [ S e i e j atta-koto-ga-aru] nihonzin i ]-o oozei sitteiru] no wa Russell j da John-NOM have met Japanese-ACC many knows '(Lit.) It is Russell j that John knows many Japanese that have met e j .' b. * [John-ga [ NP [ S e i e j atta-koto-ga-aru] nihonzin] i -o oozei sitteiru] no wa Russell j -ni da John-NOM have met Japanese-ACC many knows 'It is with Russell j that John knows many Japanese that have met e j .' Based on this difference, Hoji (1987) assigns different structures to Non-CM-Cleft and CM-Cleft, as shown in (2a) and (2b), respectively. (2) (Hoji 1987: (13)) a. Non-CM-Cleft: [ S ......(pro) ......] no wa FOCUS da No Movement Case particles cannot be attached to the "focused" NP. "Aboutness" licenses the structure. b. CM-Cleft: [ S pro i [ S ...... t i .......]] no wa FOCUS i da Movement (i.e., scrambling, i.e., syntactic adjunction) Case particles can be attached to the "focused" NP. 3 Hoji (1987, 1990) does not employ the terms "CM-Cleft" and "Non-CM-Cleft." These terms were introduced in Ueyama & Hoji 2001. 30 chain [FOCUS i , pro i , t i ] In Hoji 1990 and on, the pro movement assumed in (2b) has been replaced with the empty operator movement. Thus, the structures of CM-Cleft are slightly revised as in (3b). (3) (Hoji 1990: (225)) 4 a. Non-CM-Cleft: [ NP [ S .. (pro i ) ... ] [ NP no i ]]-wa NP i da (An aboutness licensing is possible.) b. CM-Cleft: [ S' Op i [ S ... t i ... ] no]-wa NP i -(CASE) da (Syntactic movement is obligatory and the aboutness licensing is not possible.) Non-CM-Cleft does not involve any movement, and the structure is licensed by the "aboutness relation" (Hoji 1987:4), so that subjacency effects are irrelevant. CM-Cleft, on the other hand, involves an empty operator movement in the presuppositional clause, which induces subjacency effects in (1b). What CM-Cleft and Non-CM-Cleft have in common in Hoji 1987 is the assumption that the focus phrase is base-generated at its surface position. As for CM-Cleft, Hoji & Ueyama (2003:12) suggest that the DL ("dislocated NP") in CM-Cleft, that is the focus phrase, is in an A'-position. They then suggest, along the line of Barss 1986, that "in CM-Cleft: [Op 1 [[ ... t 1 ... ] no]]-wa/ga [ DL ...-CM ] 1 da, the A'-chain consisting solely of the DL and the A'-chain (Op 1 , t 1 ) can be composed into a single A'-chain (DL 1 , Op 1 , t 1 ), and as a result the elements contained in the DL is interpreted at the position of the trace (t 1 ) with respect to BVA." 4 In Ueyama & Hoji 2001, pro in (3) is represented as ec. Then, they mention that "[t]he ec 1 can be a so-called pro (rather than a trace)." (Ueyama & Hoji 2001:11) 31 2.2.2 Kizu 2005 Kizu (2005) pursues the base-generation analysis of the cleft constructions in line with Hoji 1990. Unlike Hoji (1990), however, Kizu (2005) treats CM-Cleft and Non-CM-Cleft alike and proposes that both involve an empty operator movement in the presuppositional clause as in (4), which is the structure of CM-Cleft in Hoji 1990. 5 More detailed structure of the cleft constructions proposed by Kizu (2005) is shown in (5). (4) (Kizu 2005: 5 (9b), referring to Hoji 1990) [ CP Op i [ IP …t i … ] no]-wa NP i -(CASE) da NM-TOP COP (5) Structure for local cleft constructions (based on Kizu 2005:55 (89), 72 (32) and 109 (17a)) 6 TopP CP IP Op i C' I' IP C VP I | … t i … no V' PP V | | P' aru XP P | de 5 Kizu (2005) assigns a different structure from (4) to long-distance cleft constructions. Kizu (2005:110) mentions that the dichotomy of distinct derivations for clefts comes not from whether the focus phrase is a bare NP or not as in Hoji 1990 but from whether clefting is local or long-distance. 6 With regard to the position of the presuppositional CP, Kizu (2005) assumes the following. (i) A presuppositional CP is base-generated either in spec of IP or a higher topic position. (Kizu 2005: 40) Thus, though I put the presuppositional CP in [Spec,TopP] in (5), it can be in [Spec,IP] under Kizu’s (2005) analysis. 32 In regard to the structure of the presuppositional clause, there is no difference between Kizu’s 2005 structure and Hoji’s 1990 one when it is the local cleft. Kizu (2005:110) assumes that a null operator Op is base-generated in an argument position in the case of the local cleft. 7 Op has an operator feature or wh-feature to be checked and undergoes movement to the spec of CP to check its feature. Kizu (2005:72) also assumes that no in the presuppositional clause is a complementizer. As for the structure for the main predicate part of cleft constructions, Kizu (2005: 55) assumes that the copula da is a complex verb, which consists of the P de and the V aru. The P de incorporates into the V at some point of derivation, or it fuses phonologically under linear adjacency. A focus phrase then appears as a complement of the P. Kizu (2005:36) further assumes that the presuppositional clause forms a predication relation with the focus phrase as proposed for English cleft constructions by Williams (1980) and Browning (1987). She mentions two peculiar facts about the predication relation in Japanese cleft constructions; (i) a presuppositional CP is not a predicate of a small clause, and (ii) the predicative CP is not in a local relation with a focus XP. In order to establish the predication relation between the focus phrase and the presuppositional CP, which are not in a local relation, Kizu (2005:36) assumes that an agreement chain is created between a null operator and a focus phrase as illustrated in (6). 7 In case of the long-distance cleft, null operator Op is base-generated in a position adjoined to the highest embedded CP. (Kizu 2005: 110) 33 (6) (Kizu 2005: 36 (52)) CP i ← → XP i (focus) Op i C' IP C i … t i … Kizu (2005:36-37) mentions that "[a] null operator in the presuppositional CP is coindexed with the C head by spec-head agreement. By hypothesis, the maximal projection of the C has the same index, and the CP forms a predication relation with the XP in focus position. We should admit to say that the predication relation between the CP and the XP in [(6)] is not like the one in (pseudo) cleft construction in English. … Under this approach, the operator in the spec of CP has agreement or feature sharing with the focus XP." 2.3 Movement Analyses 2.3.1 Hiraiwa and Ishihara 2002 Hiraiwa & Ishihara (2002) (hereafter H&I) propose the movement analysis of the cleft constructions in Japanese. Specifically, they claim that CM-Cleft (Cleft in their term) in Japanese is derived from the so-called 'No da' in-situ focus construction by the focus movement and the topicalization of the remnant CP, as shown in (7). 8 The structure they propose is 8 'No da' in-situ focus construction is exemplified in (i) (i) (H&I 2002: 38 (9)) [ CP Taro-ga kono-ringo-o tabeta no] da Taro-NOM this-apple-ACC ate C COP 'It is that Taro ate this apple.' H&I (2002:38) mention that "[t]he 'No da' in-situ focus construction is a construction where the entire matrix clause is headed by the nominalizer -no followed by the copula -da. (cf. Kuno 1973) Any phrase in the nominalized CP that has phonological prominence receives a narrow focus interpretation." Under the narrow focus interpretation, only the phrase that has phonological prominence is interpreted as focus, 34 illustrated in (8). 9 (7) (cf. H&I 2002: (25)-(26)) a. Underlying structure = 'No da' in-situ focus construction: [ TopP [ FocP [ FinP/CP [ TP … XP (Focus) …] no] da] Top] b. Focus movement: [ TopP [ FocP XP (Focus) [ FinP/CP [ TP … t XP …] no] da] Top] c. Remnant Topicalization: [ TopP [ FinP/CP [ TP … t XP …] no]-wa [ FocP XP (Focus) t FinP/CP da] Top] (8) Structure of (7c): a. Surface b. LF (after reconstruction) TopP TopP FinP/CP Top' Top' TP Fin/C FocP Top FocP Top | t XP no XP Foc' XP Foc' t Fin/CP Foc FinP/CP Foc | | da TP Fin/C da | t XP no H&I (2002: 43) assume that da is a grammaticalized focus particle/marker that heads while under the broad focus interpretation, a VP containing the phrase that has phonological prominence can be interpreted as focus. 9 I assume, as commonly assumed in the literature, that an A'-movement is reconstructed at LF, or in other words, an A'-moved phrase gets interpreted at LF at its base-generated position. 35 FocP. 10 It takes FinP/CP as its complement. The functional head Top takes FocP as its complement. It is always null in Japanese. 11 They further assumes that no is the Complementizer or the functional head Fin(iteness). In regard to the focus movement, H&I (2002: 44) assume that the focus movement to [Spec, FocP] is an instance of A'-movement, but this movement is a syntactic operation distinct from A'-scrambling, that is, it does not exhibit the 'radical reconstruction' property. 12 As for the remnant topicalization, H&I (2002: 43) assume that the nominalized FinP/CP is topicalized (with a topic marker -wa) and moved to [Spec, TopP]. 13 2.3.2 Takahashi 2006 Takahashi (2006) follows H&I (2002) in assuming that cleft constructions involve A'-movement of the focus element out of the presuppositional CP, which is followed by the remnant movement of the CP. Takahashi (2006) assumes that the remnant movement of the CP is an A-movement, departing crucially from H&I’s 2002 analysis, where the remnant movement of the CP is considered to be an A'-movement. His proposed derivation and structure of CM-Cleft are shown in (9) and (10), respectively. 10 H&I (2002:43) refer to Chris Collins (in their pc) and mention that "Chris Collins points out that a copula in fact tends to be grammaticalized into a focus particle cross-linguistically." 11 As for the positions of FocP and TopP, H&I (2002) adopt Rizzi’s (1997) articulated CP structure as in (i). (i) [ TopP [ FocP [ FinP/CP [ TP … ] Fin/C] Foc] Top] (linear order irrelevant) (cf. H&I’s (24)) H&I (2002:fn.5) mention that "Rizzi (1997) actually proposes that there is another TopP sandwiched between FinP and FocP. We will assume, however, that this position is not 'active' in Modern Japanese." 12 H&I (2002:44, fn.9) further mention that "[t]he fact that Cleft is not clause-bounded and long-distance dependency is allowed trivially shows that the alleged focus movement is not A-scrambling/movement." 13 Movement to the CP-domain is assumed to be A'-movement, so that the remnant topicalization in H&I 2002 should also be taken as A'-movement. I assume that this movement is also reconstructed at LF. See fn.9 above. 36 (9) (Takahashi 2006: 5 (12)) a. [ TP [ FP [ VP [ CP [ TP Taroo-ga Hanako-ni atta] no C ]-wa da V ] Focus] T] b. [ TP [ FP Hanako-ni [ F' [ FP [ CP t' [ C' [ TP Taroo-ga t atta] no C ]-wa da V ] Focus]] T] c. [ TP [ CP t' [ C' [ TP Taroo-ga t atta] no C ]-wa [ T' [ FP Hanako-ni [ F' [ FP t CP da V ] Focus]] T] (10) Structure of (9c) a. Surface b. LF (after reconstruction) TP TP CP T' CP T' t 1 C' FP T TP C FP T | TP C Hanako-ni 1 F' no-wa F' | Taroo-ga Hanako-ni atta Taroo-ga t 1 atta no-wa VP Foc VP Foc t CP V t CP V | | da da Takahashi (2006:5) assumes that da is a copula verb, which takes the presuppositional CP as its complement. Focus head Foc then takes the VP headed by da as its complement, projecting FP, which is in turn selected by T as its complement. The focused element first undergoes A'-movement to the specifier position of FP by way of the specifier position of CP. After the focus movement, the presuppositional CP containing the traces of the focused element undergoes A-movement to the specifier position of TP, deriving the surface form of CM-Cleft. 2.4 Issues I have summarized the base-generation analyses and the movement analyses of cleft constructions in Japanese. With the proposed structures of these analyses in hand, I now 37 discuss some issues these analyses pose. 2.4.1 Issues on the movement analyses 2.4.1.1 Conceptual Problems of Hiraiwa & Ishihara 2002 I first take up Hiraiwa & Ishihara’s (2002) movement analysis of cleft constructions in Japanese. H&I’s analysis faces some conceptual problems in regard to the Proper Binding Condition (PBC) and the property of A'-movement they assume. 2.4.1.1.1 PBC violation H&I’s (2002) structure in (7c) (repeated here as (11)) involves an unbound trace (i.e., t XP ) inside the pressupositional CP, which violates the Proper Binding Condition (PBC) in (12). (11) [ TopP [ FinP/CP [ TP … t XP …] no]-wa [ FocP XP (Focus) t FinP/CP da] Top] (12) The Proper Binding Condition (Cf. Fiengo 1974/1977 and Lasnik & Saito 1984) 14 Traces must be bound. What is intended by (12) is that "a trace must be c-commanded by its antecedent." It has been observed since Harada 1977 that Japanese exhibits the PBC effects. Consider: 14 The formulation of the PBC in Fiengo 1977 is as follows. (i) (Fiengo 1977: 45, (33)) In surface structure Sα, if [e]NPn is not properly bound by […]NPn, then Sα is not grammatical. Fiengo (1977: 45) defines proper binding as a relation that holds between a node and its trace only if the node precedes its trace. 38 (13) (Saito 1992: 83, (31)) a. [Taroo-ga [ CP Hanako-ga sono hon-o yonda to] itta] (koto) Taroo-Nom Hanako-NOM that book-Acc read COMP said fact 'Taro said that Hanako read that book' b. [Sono hon-o i [Taroo-ga [ CP Hanako-ga t i yonda to] itta]] (koto) 'That book i , Taro said that Hanako read t i ' c. * [[ CP Hanako-ga t i yonda to] j [Sono hon-o i [Taroo-ga t j itta]]] (koto) '[That Hanako read t i ] j , that book i , Taro said ' (13c) is derived from (13a) by first moving the o-marked objects out of the complement clause and then preposing the complement clauses to the sentence-initial position. The complement clauses in the sentence-initial position contain the trace of the o-marked objects, which is not c-commnded by its antecedent. (13c) is reported to be unacceptabe. The following examples show the same point. 15 (14) (Tsuboi 2006: 28, (64) and (66), respectively) a. Keisatu-wa [John-ga tyuugoku-ni nigeta to] danteisita. police-TOP John-NOM China-to escape that determined Intended: 'The police determined that John escaped to China.' b. * [John-ga t i nigeta to] j tyuugoku-ni i keisatu-wa t j danteisita. John-NOM escape that China-to police-TOP determined 15 Tsuboi (2006) shows, based on the results of the on-line experiments, that the PBC effects are not as robust as it has been claimed in the literature when the first-moved object is an o-marked one as in (13), in contrast to when it is a ni-marked one as in (14) and (15). This is as expected in light of the possibility of the so-called Major Object in Japanese, as discussed in Hoji 1991. 39 (15) (Tsuboi 2006: 28, (70) and (72), respectively) a. Yamada-sensei-ga [John-ga Toyota-ni syuusyokusita to] omoikondeita (koto) . Yamada-teacher-NOM John-NOM Toyota-DAT got:a:job that had:believed that Intended: 'Prof. Yamada had believed that John got a job at Toyota.' b. * [John-ga t i syuusyokusita to] j Toyota-ni i Yamada-sensei-ga t j omoikondeita (koto) . John-NOM got:a:job that Toyota-DAT Yamada-teacher-NOM had:believed that These unacceptabilities are what the PBC expects, which strongly suggests that the PBC is operative in Japanese as well. If the PBC is operative in Japanese, H&I’s structure of CM-Cleft in (11) should also be ruled out by the PBC. Müller (1996) observe that some unbound traces left by remnant movements are licit, evading the PBC violation. Based on his observation, Müller (1996) proposes the principle of Unambiguous Domination in (16), which serves as a constraint on remnant movements. (16) Unambiguous Domination (Müller 1996: 375 (28)) An α-trace must not be α-dominated. In (16) 'α-trace' means 'trace with a (not necessarily c-commanding) antecedent in a position of type α', and 'α-dominated' means 'dominated by a category in a position of type α'. (16) predicts that a remnant XP may not occupy a position of a certain type if it contains a trace that has an antecedent in the same type of position. As for the range of α, Müller (1996) assumes A-positions and A'-positions as commonly assumed, but he further divides A'-positions into three positions: SpecC, SpecT(opic), and XP-adjoined position (where X is I or V). 16 16 Müller (1996:377) mentions that SpecT is the specifier of a verbal functional projection TP ('topic 40 Müller’s (1996) Unambiguous Domination amounts to saying that the PBC violation arises only when we have interaction of two movements of the same type. The examples in (13c), (14b) and (15b), which were shown as examples of the violation of the PBC, are ruled out by (16) since the two movements involved are the same type of movement, that is, both are 'scrambling', assuming (i) that the derivation of a non-canonical order is derived by 'scrambling' and (ii) 'scrambling' is of just one type in terms of Muller's categorization of α. Returning to Hiraiwa & Ishihara’s (2002) analysis, focus movement and remnant FinP/CP topicalization are both assumed to be instances of A'-movement. Then, the issue is whether these two A'-movements are different types under Müller’s (1996) clasification of A'-movements. As mentioned in the footnote 11 above, H&I adopt Rizzi’s (1997) articulated CP structure in regard to the positions of FocP and TopP. This means that both FocP and TopP are in the CP domain. TopP in H&I should not be the same as TP in Müller 1996, which is assumed to be a verbal functional projection. Thus, under Müller’s (1996) clasification, the specifier positions of FocP and TopP are of the same type, SpecC. The structure of the CM-Cleft in (11) would be excluded by the PBC in (16) if we adopt Müller’s (1996) clasification of A'-movement. In this respect, Hiraiwa (2003) tries to reduce the PBC effects to the derivational mechanism of the computational system, what he calls, the EEO (Edge-Extention Operation)-based Cyclic Spell-Out as shown in (17) and (18), where phases are CP and vP, but not TP. phrase') that intervene between CP and IP, and the head of which (T) acts as a landing site for V/2 movement. 41 (17) Edge-Extention Operation (EEO) (Hiraiwa 2003: 96, (17)) An Edge-Extention Operation (EEO) is a syntactic operation that Merges β with a constituent consisting of a phase head H p and its complement α. {β {H p α}} (18) (Hiraiwa 2003: 96, (18)) a. EEO triggers Transfer/Spell-Out immediately. {β {H p α } b. β H p α => Transfer/Spell-Out t β Edge-Extension Operation According to Hiraiwa (2003:96), it follows from (18a) that at a phase head H p , Transfer/Spell-Out applies to the complement domain of H p cyclically, immediately after its edge (i.e. a specifier of H p or a new head γ) is created via Internal/ External Merge (cf. (18b)). Hiraiwa (2003:96) then concludes that the PBC effects arise when a remnant category is extracted from the domain that has already been Transferred/Spelled-Out to the interfaces, which is simply impossible. Hiraiwa (2003) assumes that Long-distance Scrambling (LDS) and Topicalization are movements to the edge of CP. Thus, if the first movement of β is LDS or topicalization, it counts as an EEO and the complement of TP is immediately Transferred/Spelled-Out. Hence, the remnant CP (=α) becomes frozen in-situ. 42 (19) (Hiraiwa 2003: 97, (19)): LDS/Topicalization [ CP2 β [ C’ C 2 [ TP T [ v*P t β v*... [ CP1(=α) t β ]]]]] Hiraiwa (2003) also assumes that Long-distance Scrambling (LDS) of β can be to [Spec, vP] and Raising-to-Object (RTO) is also to [Spec, vP]. These movements also count as an EEO since v is also a phase head. Thus VP is immediately Transferred/Spelled-Out and hence the remnant CP (=α) becomes frozen in-situ. (20) (Hiraiwa 2003: 97, (20)): LDS [ TP [ v*P β [ v’ v [ VP V [ CP(=α) t β ]]]]] (21) (Hiraiwa 2003: 97, (21)): Raising-to-Object (RTO) [ v*P β [ v*’ v* [ VP V [ CP(=α) t β ]]]] On the other hand, Raising-to-Subject (RTS) is a movement to [Spec, TP], which is not a movement to the edge of a phase head. Thus RTS does not count as an EEO and trigger Transfer/Spell-Out. Therefore, remnant movements of the vP and CP are freely allowed. (22) (Hiraiwa 2003: 98, (22)): Raising-to-Subject (RTS) a. [ TP β [ Τ [ v*P (=α) t β v* [ VP ]]]] b. [ TP β [ Τ [ VP V [ CP (=α) t β C [ TP t β ]]]] Although Hiraiwa (2003) does not discuss cleft constructions, it seems that the EEO-based Cyclic Spell-Out excludes the remnant movement of the presuppositional CP assumed in H&I 43 2002. Recall that H&I (2002) adopt Rizzi’s (1997) articulated CP structure, which includes TopP and FocP. Thus, Foc should also be a phase head. Given this, the focus movement to [Spec, FocP] would counts as an EEO and the complement of Foc is immediately Transferred/Spelled-Out. Hence, the remnant CP would become frozen in-situ. Thus, Hiraiwa’s (2003) EEO-based Cyclic Spell-Out would rule out the derivation of CM-Cleft proposed in H&I (2002). In summary, the PBC is operative in Japanese and the derivation of CM-Cleft proposed in H&I 2002 violates the PBC. Müller’s (1996) Unambiguous Domination, which rules in some unbound traces created by remnant movements, cannot save the derivation. Hiraiwa (2003) tries to reduce the PBC effects to the EEO-based Cyclic Spell-Out. However, it blocks the derivation of CM-Cleft proposed in H&I 2002. Thus, it cannot resolve the PBC problem in the derivation of CM-Cleft proposed in H&I 2002. 2.4.1.1.2 Further Consideration: Different types of A′ ′ ′ ′-movement H&I (2002) might argue that their proposed derivation of CM-Cleft would raise no problem in regard to the PBC, claiming that the focus movement and remnant FinP/CP topicalization are not the same type, so that the derivation does not violate Müller’s 1996 Unambiguous Domination. Recall that Müller (1996) assumes three types of A'-positions: SpecC, SpecT(opic), and XP-adjoined position (where X is I or V). The problem of H&I’s 2002 analysis is that the focus movement and remnant FinP/CP topicalization are both movements to the position of SpecC under Rizzi’s (1997) articulated CP structure, which H&I (2002) adopt. Since in Müller 1996 SpecT is a landing site for V/2 movement in Germanic languages and for embedded topicalization in English, it should not be a landing site for the focus movement. 44 Then the remaining position is the IP- or VP-adjoined position. In Müller 1996, however, this position is for Scrambling. H&I (2002:44) then point out, based on the 'radical reconstruction' property, which I will turn to shortly, that "what we call focus movement is a syntactic operation distinct from scrambling." Thus, the IP- or VP-adjoined position should not be a position for the focus movement, either. Since all the three positions for A'-movement assumed in Müller 1996 are problematic as a position for the focus movement, what H&I (2002) would do next is to posit a new position for the focus movement. The validity depends on how tenable/reasonable it is to posit such a new position theoretically as well as empirically. I leave it open here and move on to another, and in fact, more serious, problem posed by the focus movement assumed in H&I 2002. H&I (2002:44) mention that "focus movement in the derivation of Cleft does not exhibit the 'radical reconstruction' property, which is typical of LDS [long-distance scrambling, YM]." They present the following sentence as evidence for this. (23) (H&I 2002: 44, (28)) *[Taro-ga [Hanako-ga t i tabeta ka] siritagatteiru no]-wa nani-o i da Taro-NOM Hanako-NOM ate Q know-want C-TOP what-ACC COP ‘It is what that Taro wants to know Hanako ate.’ The example (23) is unacceptable because the focus NP nani-o is not in the c-command domain of Q-marker -ka at LF. This means that the focus NP in the cleft constructions cannot be reconstructed into the presuppositional CP at LF. H&I (2002:44) then conclude that "what we call focus movement is a syntactic operation distinct from scrambling." Reconstruction effects are one of the representative properties of A′-movement. Thus, if 45 the focus movement in H&I 2002 were indeed an A′-movement which does not exhibit the reconstruction effects, we need to introduce a new type of A'-movement that is different from the one commonly assumed. It seems costly to introduce this type of A'-movement into the current theory because it virtually has no difference from A-movement in its effects. 2.4.1.2 Empirical Problems of Takahashi 2006 Takahashi’s (2006) analysis, though it is also the movement analysis, evades the PBC problem by assuming that the focus movement (in the cleft construction) is an A'-movement, while the remnant movement of the presuppositional CP is an A-movement. Since these two movements are different types, the unbound trace created by the remnant movement of the presuppositional CP is legitimate under Müller’s 1996 Unambiguous Domination in (16). Thus, the structure of CM-Cleft proposed by Takahashi (2006) does not induce the PBC violation. However, Takahashi’s (2006) analysis of CM-Cleft faces some problems in dealing with the NSE (Neg(ation)-sensitive elements) facts and Condition D effects. 2.4.1.2.1 NSE facts in the cleft constructions I first take up the NSE (Neg(ation)-sensitive elements) facts of the CM-Cleft. Kataoka (2007) uses the term NSE, which refers to the elements generally called Negative Polarity Item (NPI) in the literature. 17 I use the tem NSE instead of NPI, following Kataoka (2007). 17 Kataoka (2007: 2-3, fn.1) mentions as follows. "The notion of 'polarity' was introduced by such works as Fauconnier 1975 and Ladusaw 1979 in order to give an account for the syntactic behavior and semantic characteristics of any(-) in English which requires negative environment, and it originally had a close connection with a particular semantic property of those items: they induce 'scale' interpretation in the terms of Fauconnier 1975, or 'downward entailment' interpretation in the terms of Ladusaw 1979, which allows an element without universal force, if it is combined with negation, to give rise to an interpretation of universal negation. If we take this into consideration, that term should not be used as a purely descriptive term, since it is possible that some elements, amari (too much) in Japanese, for 46 NSEs in English such as any(-) cannot occur outside the c-command domain of the negative element (Neg) not, as observed in (24). Based on this observation, the condition in (25) has generally been agreed upon in the literature. (24) a. He didn't [ invite anybody.] b. *Anybody didn't [invite him.] (25) (Kataoka 2007: 3, (4)) An NSE must be c-commanded by Neg at LF. (cf. Klima 1964) Japanese expressions such as XP-sika 'all but XP', rokuna-N 'good/decent N' and dare-mo (indefinite person+suffix -mo)/nani-mo (indefinite thing+suffix -mo) are instances of NSE in Japanese. They must occur with Neg -nai as shown in (26), (27) and (28), respectively. 18,19 (26) (Kataoka 2007: 1, (1)) a. Taro-wa manga-sika yoma-nai. / *yomu. Taro-TOP comics-all:but read-Neg / read (Roughly) 'Taro does not read any kind of book but comics.' b. Taro-sika manga-o yoma-nai. / *yomu. Taro-all:but comics-ACC read-Neg / read instance, though they require negation, may not induce the 'scale' interpretation in question." 18 To be precise, dare-mo and nani-mo induce an interpretation of universal negation combined with Neg. 19 Kataoka (2007: 35) mentions as follows: "The expression dare-mo/nani-mo can occur by itself or can be accompanied with a noun phrase attached by a case marker (CM), such as gakusei-CM daremo (any student). I treat the whole phrase as a constituent of an argument NP, and will refer to the phrase as (N-CM) dare-mo/nani-mo." I follow Kataoka (2007) in this respect. 47 (Roughly) 'Nobody but Taro reads comics.' (27) (Kataoka 2007: 1, (2)) a. Jiro-wa kyoositu-de rokuna-koto-o iwa-nai. / *iu. Jiro-TOP classroom-in good-thing-ACC say-Neg / say (Roughly) 'Jiro does not say any good thing in the classroom.' b. Rokuna-gakusei-ga gakkai-de happyo-o si-nai. / *suru. good-student-NOM conference-LOC presentation-ACC make-Neg /make (Roughly) 'No good students make a presentation at the conference.' (28) (Kataoka 2007: 34, (93)) a. (Gakusei-ga) dare-mo sinbun-o yoma-nai / *yomu (koto) (student-NOM) DARE-MO newspaper-ACC read-Neg / read (Comp) '(The fact that) no students read newspapers.' b. Hanako-wa tanzyoobi-ni (tomodati-o) dare-mo syootai-si-nai /*syootai-suru. Hanako-TOP birthday-on (friends-ACC) DARE-MO invite-do-Neg / invite-do 'Hanako did not invite any friend on her birthday.' It has long been assumed, without serious discussions on its validity, that the condition in (25) also holds of Japanese NSEs. Furthermore, under the condition in (25), NSEs must always be c-commanded by Neg. Hence, many researchers have taken it as the licensing condition on NSEs, with Neg being the licenser. Kataoka (2007), on the other hand, argues that not all NSEs in Japanese obey the condition in (25). She argues that among the three NSEs in Japanese just mentioned only rokuna-N 48 'good/decent N' obeys the condition (25). 20 Thus, there seems to be a consensus in the literature that at least rokuna-N must be c-commanded by Neg at LF. Let us thus consider the following examples of CM-Cleft with rokuna-N. 21 (29) a. Sono paatii-de Hanako-ga koeokaketa-no-wa rokuna otoko-ni ja-nak-katta. that party-at Hanako-Nom spoke:to-that-Top decent man-Dat be-Neg-Past '(Lit.) It wasn’t decent men that Hanako spoke to at that party.' b. * Sono paatii-de Hanako-ga koeokake-na-katta-no-wa rokuna otoko-ni datta. that party-at Hanako-Nom spoke:to-Neg-Past-that-Top decent man-Dat be '(Lit.) It was decent men that Hanako didn’t speak to at that party.' In (29) the NSE rokuna-N is situated in the focus position of CM-Cleft. Neg -nai, on the other hand, is in the matrix clause in (29a), while it is in the presuppositional CP in (29b). Recall now the structure of the CM-Cleft proposed by Takahashi (2006) shown in (10), repeated here as (30). 20 The conditions on XP-sika and dare-mo/nani-mo proposed in Kataoka 2007 are (i) and (ii), respectively. (i) (Kataoka 2007: 25, (64)) XP-sika must be in the NegP-Spec position at LF. (ii) (Kataoka 2007: 38, (106)) (N-CM) dare-mo/nani-mo must c-command Neg at LF. 21 It would be possible that (29b) is unacceptable because -nai precedes the NSE rokuna-N. However, I found the following examples not unacceptable, which suggests that the precedence is not the crucial factor here. (i) Sono paatii-de Hanako-ga koeokake-na-katta(-yo), rokuna otoko-ni. (cf. (29b)) that party-at Hanako-Nom spoke:to-Neg-Past(-Sentence Final Particle) decent man-Dat '(Lit.) Hanako didn’t speak at that party, to decent men.' For the analyses of Right Dislocation in Japanese as in (i), see; Haraguchi 1973, Simon 1989 and Murayama 1999 (Rightward movement approach), Kurogi 2007 (Double preposing approach), Kuno 1978, Abe 1999, Tanaka 2001, Yamashita 2008 and Takita to appear (Repetition + deletion approach), and Sells 1999 and Soshi and Hagiwara 2004 (Base-generation approach). The classification is taken from Takita to appear. 49 (30) [ TP [ CP t' [ C' [ TP Taroo-ga t atta] no C ]-wa [ T' [ FP Hanako-ni [ F' [ FP t CP da V ] Focus]] T] a. Surface b. LF (after reconstruction) TP TP CP T' CP T' t 1 C' FP T TP C FP T | TP C Hanako-ni 1 F' no-wa F' | Taroo-ga Hanako-ni atta Taroo-ga t 1 atta no-wa VP Foc VP Foc t CP V t CP V | | da da Under Takahashi’s 2006 analysis, focus movement is an A'-movement to the specifier position of FP. Thus, it should be reconstructed to its original position inside the presuppositional CP at LF as shown in (30b). Given this, in (29b) rokuna otoko-ni 'to decent men' is reconstructed to the position inside the presuppositional CP at LF, where it is c-commanded by Neg -nai. Therefore, Takahashi (2006) predicts that (29b) is acceptable, contrary to the fact, as long as he accepts that the licensing condition for rokuna N is at LF. The acceptability of (29a) is also problematic for Takahashi (2006). Since Neg -nai comes before the past tense marker -katta, it should be situated at the position lower than TP. Then, according to the structure of the cleft adopted in Takahashi 2006, Neg -nai cannot c-command rokuna otoko-ni 'to decent men', which is situated inside the presuppositional CP after the reconstruction at LF, because the CP is situated in the specifier position of TP. Thus, Takahashi (2006) predicts that (29a) is unacceptable, again contrary to the fact. 2.4.1.2.2 Condition D effects Next problem Takahashi’s (2006) analysis of CM-Cleft faces is Condition D effects. 50 Ueyama (1998) defines Condition D as in (31). (31) Condition D: (Ueyama 1998: 204 (64)) Nominal expressions α and β must be disjoint in reference, if α<β and α c-commands β. (32) (Ueyama 1998: 204 (65)) Nominal expressions α and β stand in the relation of α<β, iff (i) for every x, x an individual which can be expressed by β, x can be expressed by α, and, (ii) for some y, y an individual which can be expressed by α, y cannot be expressed by β. Ueyama (1998: 204) points out that "[a]ccording to the characterization of the concept 'a<b' in [(32)], the relations in [(33)] hold, for example, and we expect that those pairs of expressions should exhibit the Condition D effects." Condition D effects are exemplified in (34a), where so-ko 'that place' c-commands Toyota. (33) a. [he]<[John] b. [linguists]<[syntacticians] c. [so-ko]<[daikigyoo] ('it' < 'giant company') (34) (Ueyama 1998: 206 (69)) a. ?*So-ko i -ga [Toyota i -no ko-gaisya]-o suisensita. that-place-NOM Toyota-GEN child-company-ACC recommended 'It recommended [Toyota's subsidiary].' 51 b. Toyota i -ga [so-ko i -no ko-gaisya]-o suisensita. Toyota-NOM that-place-GEN child-company-ACC recommended 'It recommended [Toyota's subsidiary].' Now, recall again the structure of CM-Cleft proposed by Takahashi (2006) in (30) (= (10)). (30) [ TP [ CP t' [ C' [ TP Taroo-ga t atta] no C ]-wa [ T' [ FP Hanako-ni [ F' [ FP t CP da V ] Focus]] T] a. Surface b. LF (after reconstruction) TP TP CP T' CP T' t 1 C' FP T TP C FP T | TP C Hanako-ni 1 F' no-wa F' | Taroo-ga Hanako-ni atta Taroo-ga t 1 atta no-wa VP Foc VP Foc t CP V t CP V | | da da In the ditransitive constructions in Japanese, [SUB DAT ACC VERB] is assumed to be the base order (e.g., Hoji 1985, 2003a, Hayashishita 2000, among others). Then in the ditransitive sentence Taroo-ga Hanako-ni Hana-o ageta 'Taroo gave Hanako some flowers,' when the o-marked NP is clefted, the structure comes as in (35). 52 (35) [ TP [ CP t'[ C' [ TP Taroo-ga Hanako-ni t ageta] no C ]-wa [ T' [ FP Hana-o [ F' [ FP t CP da V ] Foc]] T] a. Surface b. LF (after reconstruction) TP TP CP T' CP T' t 1 C' FP T TP C FP T | TP C hana-o 1 F' no-wa F' | T-ga H-ni hana-o ageta T-ga H-ni t 1 ageta no-wa VP Foc VP Foc t CP V t CP V | | da da In (35) the o-marked NP in the focus position cannot c-command the ni-marked NP in the presuppositional clause both in the surface structure and at LF. In the surface structure in (30a), the o-marked NP is in [Spec,FP], which position cannot c-command inside the presuppositional CP. At LF after reconstruction, the o-marked NP does not c-command the ni-marked NP either, since it is located in the position lower than the ni-marked NP in their base structure. Given Condition D in (31), then, when the o-marked NP in the focus position and the ni-marked NP in the presuppositional CP is in the relation of [o-marked NP]<[ni-marked NP] (see (32)), Takahashi (2006) expects that they need not be disjoint in reference: that is, coreference is possible, since o-marked NP does not c-command ni-marked NP both in the surface structure and at LF. Let us see how Takahashi’s (2006) expectation turns out. First, aitu 'that guy' is an anaphoric epithet in Japanese, which can refer to Name as shown in (36a). When aitu c-commands Name as in (36c), it cannot refer to the Name due to Condition D in (31). In (36b), on the other hand, aitu does not c-command the Name, so it can refer to the Name as 53 expected. (36) a. Suzuki-kyoozyu-wa Yamada i -no kiboositeita kaisya-ni aitu-o i syookaisita Suzuki-prof.-TOP Yamada-GEN wished company-DAT that-guy-ACC introduced '(lit.) Prof. Suzuki introduced that guy i to the company that Yamada i had wished to work for.' b. Suzuki-kyoozyu-wa aitu i -no kiboositeita kaisya-ni Yamada-o i syookaisita Suzuki-prof.-TOP that-guy-GEN wished company-DAT Yamada -ACC introduced '(lit.) Prof. Suzuki introduced Yamada i to the company that that guy i had wished to work for.' c.*Suzuki-kyoozyu-wa aitu i -ni Yamada i -no kiboositeita kaisya-o syookaisita Suzuki-prof.-TOP that-guy-DAT Yamada-GEN wished company-ACC introduced be '(lit.) Prof. Suzuki introduced the company that Yamada i had whished to work for to that guy i .' Examples in (37) are the cleft counterparts of the examples in (36). (37) a.*?Suzuki-kyoozyu-ga Yamada i -no kiboositeita kaisya-ni t syookaisita-no-wa aitu-o i da Suzuki-prof.-NOM Yamada-GEN wished company-DAT introduced-that-TOP that-guyn-ACC be '(lit.) It was that guyn i that Prof. Suzuki introduced to the company Yamada i had wished to work for.' b. Suzuki-kyoozyu-ga aitu i -no kiboositeita kaisya-ni t syookaisita-no-wa Yamada-o i da Suzuki-prof.-NOM that-guy -GEN wished company-DAT introduced-that-TOP Yamada-ACC be '(lit.) It was Yamada i that Prof. Suzuki introduced to the company that guy i had whished to work for.' 54 c. Suzuki-kyoozyu-ga aitu i -ni t syookaisita-no-wa Yamada i -no kiboositeita kaisya-o da Suzuki-prof.-NOM that-guy-DAT introduced-that-TOP Yamada-GEN wished company-ACC be '(lit.) It was the company that Yamada i had wished to work for that Prof. Suzuki introduced to that guy i .' Under Takahashi’s (2006) structure of CM-Cleft in (30), aitu-o in (37a) does not c-command the Name Yamada in the presuppositional CP. Thus, Condition D is not relevant, so Takahashi (2006) expects that aitu can refer to Yamada in (37a). His expectation does not hold, which poses a problem for his analysis. It should be noted, however, that it is just an expectation, and it is possible that the impossibility of the coreference between aitu and Yamada in (37a) comes from factors other than Condition D. Hence, the argument in this subsection, which goes against Takahashi’s (2006) analysis, is not so fatal to him. Ueyama (1998:204) also mentions as follows: "We should note that Condition D cannot be stated without referring to a notion outside Grammar, namely, 'individuals'. This is one of the reasons why I consider that Condition D cannot be a purely linguistic condition, and hence, the Condition D effects need not be accounted for by the core portion of Grammar." If it is the case, it also weakens the argument in this subsection. 22 The facts presented here, on the other hand, do not pose any problem for H&I’s (2002) analysis of CM-Cleft, provided that the focus movement is not reconstructed and that Condition D applies at LF. Since the movement of the presuppositional CP is a regular A'-movement, it is reconstructed at LF. Then the elements inside the presuppositional CP are c-commanded by 22 Some speakers in fact find the intended coreference in examples with "Condition D violations" to be not clearly impossible, which also casts doubt on the status of Condition D as a purely linguistic condition. 55 the focus phrase in [Spec,FP], so that Condition D correctly excludes the possibility of the coreference between aitu and Yamada in (37a). However, it is again necessary to introduce a new type of A'-movement, which poses a problem for H&I’s analysis. 2.4.1.3 Resumption in the cleft constructions Resumption facts in CM-Cleft pose a problem for the movement analyses of CM-Cleft in general. It is observed in Hoji & Ueyama 1998 that CM-Cleft in Japanese allows 'resumption', as indicated in (38). (38) (Hoji 2003: 37, (90) from Hoji and Ueyama 1998) a. ?[John-ga saisyoni so-ko-o uttaeta no]-wa Toyota-o da. John-NOM first that-place-ACC sued COMP-TOP Toyota-ACC be 'It was Toyota-ACC that John sued it-ACC first.' b. ?[Mary-ga John-ni so-ko-o suisensita no]-wa Toyota-o da. Mary-NOM John-DAT that-place-ACC recommended COMP-TOP Toyota-ACC be 'It was Toyota-ACC that Mary introduced it-ACC to John.' In regard to the "resumption" in Japanese, Ueyama & Hoji (2001) claim as follows: (39) (Ueyama & Hoji 2001: 12) "Resumption" in Japanese (that has a unique grammatical property) is an overt category that appears in the launching site of a covert IP-adjunction. PF: [ . . . {pro/so-ko} 1 . . . ] LF: [ IP {pro/so-ko} 1 [ IP . . . t 1 . . . ]] Semantic Representation: λx [ . . . x . . . ] 56 Their argument for (39) is based on the reconstruction effects on BVA. In OS-type constructions as in (40), BVA (Toyota-sae, soko) is possible. This means that the o-marked NP is interpreted at the position of ec, showing what is called the reconstruction effects. (40) (Ueyama and Hoji 2001: 7, (21)) a. [So-ko-o tekitaisisiteiru kaisya] 1 -o Toyota-sae-ga ec 1 uttaeta that-place-ACC be:hostile company-ACC Toyota-even-NOM sued '[The company which is hostile to it ] 1 , even Toyota sued ec 1 .' b. [So-ko-o tekitaisisiteiru kaisya] 1 -o Nissan-ga Toyota-ni-sae ec 1 suisensita that-place-ACC be:hostile company-ACC Nissan-NOM Toyota-DAT-even recommended '[the company which is hostile to it ] 1 , Nissan recommended to even Toyota ec 1 .' When we put resumption on the ec positions of (40), BVA (Toyota-ni-sae, soko) becomes impossible as shown in (41). (41) (Ueyama & Hoji 2001: 7, (22)) a. * [So-ko-o tekitaisisiteiru kaisya] 1 -o Toyota-sae-ga [so-ko-o] 1 uttaeta b. * [So-ko-o tekitaisisiteiru kaisya] 1 -o Nissan-ga Toyota-ni-sae [so-ko-o] 1 suisensita Based on these facts, Ueyama & Hoji (2001) conclude that resumption is incompatible with the OS-type construction that exhibit so-called A'-properties, i.e., the Surface OS-type in the terms of Ueyama 1998, thereby giving support to the claim in (39). Under Ueyama & Hoji’s 2001 analysis of resumption, (39) is the ONLY configuration in which resumption "that has a unique grammatical property" appears in Japanese. Given that (39) is valid, the focus 57 movement assumed under the movement analysis of CM-Cleft does not give rise to the configuration required in (39) as it is assumed to be an instance of overt A'-movement. Therefore, resumption "that has a unique grammatical property" is predicted to be impossible in CM-Cleft under the movement analysis. The acceptability of (38) is thus problematic under the movement analyses of CM-Cleft. 23 2.4.2 Issues on the base-generation analyses I now turn to the issues concerning the base-generation analyses of the CM-Cleft, which I also pursue in this work. I first discuss some conceptual problems Kizu’s (2005) analysis faces, and then move to Hoji’s (1987) analysis. 2.4.2.1 Conceptual Problems of Kizu’s (2005) analysis 2.4.2.1.1 .Predication relations in null operator constructions In Section 3.2.2, I have presented Kizu’s (2005) analysis of cleft constructions. Her proposed structure has been shown in (5), repeated here as (42). 23 Chapter 6.5 addresses the issue of resumption in Japanese in more detail and argues for the validity of (39). 58 (42) Structure for local cleft constructions (based on Kizu 2005: 55 (89), 72 (32) and 109 (17a)) TopP CP IP Op i C' I' IP C VP I | … t i … no V' PP V | | P' aru XP P | de Kizu (2005) assumes that a predication relation must be established between the focus phrase and the presuppositional clause. Kizu (2005:39) then refers to Browning 1987 about the syntactic predication in null operator constructions in English. Browning’s (1987) assumptions are as follows. (43) (based on Kizu 2005:39) a. A CP can function as a 'complex predicate' (Williams 1980) when it contains a predicate variable. 24 [ CP Op i /WH i [ IP … t i … ]] b. The predicate has to satisfy a locality condition; the predicate and its subject must mutually c-command and no phrase can intervene between the two. Kizu’s (2005) proposed structure of cleft constructions in (42) conforms to (43a), but not to 24 "Predicate variable" is the open position in the S which makes it a (one-place) predicate. (Williams 1980:209) 59 (43b). In order to establish the predication relation between the focus phrase and the presuppositional clause without a mutual c-command relation, Kizu (2005:36) assumes that an agreement chain is created between a null operator and a focus phrase as illustrated in (6), repeated here as (44). (44) (Kizu 2005: 36 (52)) CP i ← → XP i (focus) Op i C' IP C i … t i … Kizu’s (2005:36-37) assumption is that the operator in the spec of CP has agreement or feature sharing with the focus XP. However, Kizu (2005) does not clearly explain the mechanism of agreement or feature sharing, i.e., under what structural condition(s) the proposed agreement or feature sharing can take place. As is evident from the structure in (42), the presuppositional CP and the focus phrase are neither in a local relation nor in a c-command relation with each other. This casts doubt on the assumption of such a mechanism, though it depends on how Kizu (2005) justifies the mechanism of agreement or feature sharing. 2.4.2.1.2 .Small clause analysis As a potential alternative to satisfy a locality condition in (43b), Kizu (2005) refers to Heggie 1988, which assumes that a presuppositional clause and a focus element establish a predication relation within a small clause at D-structure in English pseudoclefts. Heggie (1998) proposes that the pseudocleft sentence in (45a) is derived from the structure in (45b), where the subject AP and the predicate CP form a CP small clause. 60 (45) (Kizu 2005: 39, (57)) a. What John i is is important to himself i . b. [ VP is [ CP [ AP PRO important to himself] [ CP what John is]]] Predicate CP then moves to the spec of CP crossing the subject AP. Although the predicate and the subject are not in a local relation at S-structure, the predication relation is formed locally at D-structure. Kizu (2005:39-40) discusses the possibility of extending this CP small clause analysis of English pseudoclefts to Japanese clefts and concludes that it cannot be applied to Japanese clefts. Her argument is as follows. Kizu (2005) first presents that the underlying structure of a cleft sentence in (46a) would be (46b) under the small clause analysis. (46) (Kizu 2005: 40, (58)) a. [Taroo-ga e atta-no]wa zibun-no hahaoya-ni-da b. [ VP [ CP [ NP zibun-no hahaoya-ni] [ CP Taroo-ga e atta-no]]-da] The presuppositional CP is base-generated as a predicate of the small clause and establishes a predication relation with the focus NP at D-structure. Kizu (2005:40) points out that under this analysis we must assume that the predicate of the small clause, that is, the presuppositional CP, moves to a higher position crossing the focus NP. Kizu (2005:40) then shows that a small clause predicate does not generally undergo any movement in Japanese. She mentions that the sentence in (47a) contains a small clause, in which the subject NP has the predicate AP, as indicated in (47). 25 When we scramble out the 25 Kizu (2005) refers to Kikuchi & Takahashi 1991 for this observation. 61 predicate AP in (47a) as in (47b) and (47c), the sentences are unacceptable. (47) (Kizu 2005: 40, (59)) a. Taroo-ga [ AP [ NP zibun-no kodomo-o] [ AP hokori-ni]] omotteiru Taroo-NOM self-GEN child-ACC be:proud think 'Taro is proud of his child' b. *Taroo-ga [ AP hokori-ni] i [ AP [ NP zibun-no kodomo-o] t i ] omotteiru Taroo-NOM be:proud self-GEN child-ACC think c. *[ AP hokori-ni] i Taroo-ga [ AP [ NP zibun-no kodomo-o] t i ] omotteiru be:proud Taroo-NOM self-GEN child-ACC think Kizu (2005) thus concludes that (46b) would not be the underlying structure of (46a), so that small clause analysis for English pseudoclefts cannot be applied to Japanese cleft constructions. 26 For Kizu’s (2005) argument to hold, the generalization that a small clause predicate does not undergo any movement in Japanese must be confirmed. However, Kizu (2005) only says that a small clause predicate does not generally undergo any movement in Japanese. In fact, for me (47b) and (47c) are fairly acceptable and some informants also find them to be not totally unacceptable. Furthermore, it is not clear whether zibun-no kodomo-o hokori-ni in (47) must be assumed as a small clause. It is possible to assume instead that zibun-no kodomo-o in (47) is a major object, so that it is one of the arguments of the predicate omotteiru. Moreover, Kizu (2005) seems to assume tacitly that in order for the predication relation to 26 Kizu’s (2005:40) wording is as follows: "This fact [in (47)] cast doubt on the small clause analysis for Japanese clefts. At this point, I put this problem aside and yet stipulate that a presuppositional CP is base-generated either in spec of IP or a higher topic position, and that a focus phrase is generated as a sister to the copula verb." 62 be established in a small clause, subject and predicate must be in the configuration as in (48), where the subject precedes the predicate, both in English and Japanese. As a result of this assumption, Kizu (2005) must assume that the predicate of a small clause moves to a higher position crossing the subject. (48) [ Small clause subject [ predicate … ]] If we instead assume that the following configuration is available, where the predicate precedes the subject, no overt movement of the focus NP is needed to take place in order to get the surface word order of cleft constructions. This is what I will pursue in my proposal in this work. 27 (49) [ Small clause [ predicate … ] subject ] 2.4.2.1.3 Position of the focus phrase The next problem of Kizu’s (2005) analysis is the position of the focus phrase in cleft constructions. As shown in (42) (=(5)), under Kizu’s (2005) analysis the focus phrase appears as a complement of the P, de, whose projection PP is the complement of the verb aru. As an argument for this structure, Kizu (2005) presents a single constituent restriction on the focus phrase in cleft constructions. More specifically, Kizu (2005) argues that the focus position must be occupied by a single constituent, since the position is the complement of P, which cannot take more than one complement. Kizu (2005) then presents two arguments to show that the focus position of cleft constructions must be occupied by a single constituent. 27 See Chapter 5.2 for the details of the proposal of the CM-Cleft in this work. 63 As the first argument for a single constituent restriction, Kizu (2005:50-51) reports that a focus position in cleft constructions restricts the order between a numeral quantifier (NQ) and its associated noun phrase as in (50). These two occurrences are free in other positions as shown in (51). (50) (Kizu 2005: 51, (80)) a. [Taroo-ga e katta-no]-wa hon-o takusan dat-ta TarooNOM bought-NM-TOP book-ACC many COP-PAST 'It was many books that Taro bought.' b. * [Taroo-ga e katta-no]-wa takusan hon(-o) dat-ta TarooNOM bought-NM-TOP many book(-ACC) COP-PAST (51) (Kizu 2005: 51, (79)) a. Taroo-ga hon-o takusan katta TarooNOM book-ACC many bought 'Taro bought many books.' b. Taroo-ga takusan hon-o katta TarooNOM many book-ACC bought Kizu (2005:53) points out, referring to Kitahara 1993 and Koizumi 1995, that "it has been claimed that the NP-NQ order forms a single constituent but the other order does not. This is supported by the following coordination test." 64 (52) (Kizu 2005: 53, (84)) a. [Hon-o takusan] to [zassi-o sukosi] katta book-ACC many and magazine-ACC a:few bought '(I) bought many books and a few magazines.' b. * [Takusan hon(-o)] to [sukosi zassi(-o)] katta many book(-ACC) and a:few magazine(-ACC) bought Kizu (2005:54) mentions that the acceptability of (52a) shows that the NP-NQ sequence can be coordinated with another NP-NQ, while the unacceptability of (52b) shows that the NQ-NP phrases cannot be conjoined by to 'and'. She then argues that this supports the claim that the NP-NQ forms a constituent but the NQ-NP does not. 28 As the second argument for a single constituent restriction, Kizu (2005:54) reports that multiple foci in cleft constructions are unacceptable, as shown in (53). 29 Unlike clefting, multiple scrambling is allowed, as shown in (54). (53) (Kizu 2005: 54, (85)) * [Taroo-ga e i e j okutta-no]-wa [Hanako-ni i hon(-o) j ]-da Taro-NOM sent-NM-TOP Hanako-DAT book(-ACC)-COP '(lit.) It was [to Hanako, a book] that Taro sent.' 28 However, the example in (50a), which is reported as unacceptable by Kizu (2005), is fairly acceptable for me, and some informants told me that it is not totally unacceptable. 29 As for the (un)acceptability of (53), Kizu (2005:54) mentions as follows. "[T]he judgments for multiple foci in Japanese clefts vary in the literature. For example, Koizumi (1995), Kuwabara (1995,1996), and Hasegawa (1997) allow multiple foci in cleft constructions (with some restrictions; see Takano (2002: footnote 1)), but Inoue (1976:101) and Noda (1996:66-67) find these cases ungrammatical. I agree with the latter judgment and deal with the latter type of grammar in this book." 65 (54) (Kizu 2005: 54, (86)) Hanako-ni i hon-o j kinoo Taroo-ga t i t j okutta Hanako-DAT book-ACC yesterday TarooNOM sent 'To Hanako, a book, Taro sent (it) (to her) yesterday.' Based on these facts, Kizu (2005:54) proposes that focus phrases in clefts are not derived by scrambling but are base-generated in a predicate position that is occupied by a single constituent. Kizu (2005:56) further assumes that P de in the cleft constructions behaves like an unaccusative predicate that projects only an internal argument. 30 Since the specifier position is unavailable for the P, the focus position in (42) (=(5)) cannot be occupied by more than one element. Kizu (2005) thus reports that (53) is unacceptable. However, as Kizu (2005) herself points out (see footnote 20), many informants accept (53). This is problematic for Kiz’s (2005) analysis. Kizu (2005) says that she would account for only the unacceptable case of (53). However, if there are two groups of speakers, we need to account for both, providing a possible formal account of the difference. Such a difference must be something learnable, but it is not immediately clear how it might be. Furthermore, Kizu (2005) presents no argument for the assumption of de as a P head that takes the focus phrase as its sole complement. 31 This prevents us from examining her analysis further in a falsifiable way. The position of the focus phrase in Kizu’s (2005) proposed structure poses another problem. Recall the following example in (37), repeated here as (55), which I have presented in the discussion of Condition D effects. 30 Kizu (2005) refers to Baker 1996 in this respect. 31 Kizu (2005: footnote 41 of Chapter 2) just mentions as follows; "This idea follows Nishiyama (1998) among others. In fact, the type of category for de as part of copula is not our main concern in this book." 66 (55) *? Suzuki-kyoozyu-ga Yamada i -no kiboositeita kaisya-ni t syookaisita-no-wa aitu-o i da Suzuki-prof.-NOM Yamada-GEN wished company-DAT introduced-that-TOP that-guyn-ACC be '(lit.) It was that guyn i that Prof. Suzuki introduced to the company Yamada i had wished to work for.' Condition D in (31) states that nominal expressions α and β must be disjoint in reference if α<β and α c-commands β. In (55), aitu<Yamada holds in reference. Thus, if (55) were unacceptable due to the violation of Condition D, the focus phrase aitu should c-command Yamada in the presuppositional CP in (55). Under Kizu’s (2005) structure in (42) (=(5)), focus phrase is situated in the specifier position of PP, which is the complement of the main verb aru. Thus, the position of focus phrase in (42) (=(5)) is situated too low in the structure to c-command elements inside the presuppositional CP, which is in [Spec,TopP]. Kizu (2005), therefore, predicts that coreference between aitu and Yamada is possible, contrary to the fact. The argument here based on Condition D effects, however, is not so strong because its validity is contingent on the assumption that (55) is unacceptable due to the violation of Condition D. As I mentioned above, it is possible that the impossibility of the coreference between aitu and Yamada in (55) comes from factors other than Condition D. 2.4.2.2 Issues on Hoji 1987, 1990, Hoji & Ueyama 2003 The structure of CM-Clefts proposed in Hoji 1987 and the subsequent works has shown in (3), repeated here as (56). I will basically adopt it in my analysis of CM-Cleft. (56) (Hoji 1990: (225)) CM-Cleft: [ S' Op i [ S ... t i ... ] no]-wa NP i -(CASE) da 67 There are three issues to be addressed in regard to the structure of CM-Clefts in (56). (57) (i) Where is the focus phrase situated in CM-Clefts? (ii) What is the relation between the presuppositional CP and the focus phrase? (iii) How the focus phrase is related to 'its theta-position' inside the presuppositional CP? In regard to (57-i) Hoji & Ueyama (2003) suggest that the focus phrase is in an A'-position, but it is not clear from their proposed structure exactly where the focus phrase is situated. Regarding (57-ii) and (57-iii), as mentioned in Chapter 1.2.1.1, Hoji & Ueyama (2003) observe similarities between CM-Cleft and the Deep OS-type of 'scrambling' constructions. The structure of the Deep OS-type proposed in Ueyama 1998 is as follows. (58) (Ueyama 1998: 63, (91)) Deep OS-type: PF: NP 1 -DAT/ACC (=DL) ... NP-NOM ... ec 1 ... LF: NP 1 -DAT/ACC (=DL) Op 1 [ NP-NOM ... t 1 ... In (58) Deep DL is related to 'its theta-position' by means of (i) the 'base-generation' of Op in 'the theta position' with which the Deep DL is related, (ii) the LF adjunction of the Op to the IP, and (iii) the formation of the predication relation between the object and the IP with the object functioning as the subject. Should the same mechanism apply to CM-Cleft, where the focus phrase corresponds to the Deep DL and the presuppositional CP to the IP? There are two substantial differences between these two constructions; (i) the Deep DL comes to the left of the IP in the Deep OS-type, while the focus phrase comes to the right of the presuppositional CP in CM-Cleft, and (ii) there is an IP boundary between the Deep DL and Op in the Deep OS-type, 68 while there is a CP boundary between the focus phrase and Op in CM-Cleft. 2.5 Choose among the different analyses 2.5.1 Arguments for the movement analyses I have presented problems of the movement analyses. The structure of CM-Cleft proposed in H&I 2002 faces the PBC related problem. Takahashi’s (2005) proposed structure can evade the problem, but the NSE facts and the Condition D effects of CM-Cleft go against his analysis. Futhermore, the possibility of resumption in CM-Cleft is problematic for the movement analysis in genaral. On the other hand, the PBC problem is irrelevant to the base-generation analyses as there is no unbound trace in the proposed structures. Moreover, the NSE facts of CM-Cleft are exactly what the base-generation analyses expect. The Condition D effects of CM-Cleft is problematic for Kizu (2005), but can be accounted for under the base-generation analyses if we assume the structure in which the focus position c-commands the presuppositional CP. The possibility of resumption in CM-Cleft is also what the base-generation analyses expect. Let us next consider the arguments for the movement analyses. If the phenomena on which the arguments are based could not be accounted for by the base-generation analyses, it would pose problems for the base-generation analyses. H&I (2002) present the island effects and the fact that the focus phrase has a case-marker assigned by the predicate in the presuppositional CP as supporting facts for the movement analysis. First, CM-Cleft exhibits the island effects as shown in (59). 69 (59) (Hiraiwa & Ishihara 2002: 37, (5a)) * [John-ga [[e i e j kaita] hito i ]-o hihansita-no]-wa kono-ronbun-o j da John-Nom wrote person-Acc criticized-C-TOP this paper-Acc be '(Lit.) It is this paper j that John criticized the person who wrote e j .' In (59) the focus phrase has been extracted out of the complex NP and the sentence are unacceptable. H&I (2002) argue that the island effects are observed due to the syntactic movement of the focus phrase. Under the base-generation analyses, empty operator movement inside the presuppositional CP accounts for the island effects in (59) given that the relevant empty operator movement is clause-bounded. Thus, the island effects in CM-Cleft pose no problem for the base-generation analyses. Second, the focus phrase in CM-Cleft has a case-marker assigned by the predicate in the presuppositional CP, which H&I (2002) call the case-matching effect. H&I (2002) argue that the case-matching effect is a natural consequence of the movement analysis since the focus phrase is in fact an argument of the predicate in the base structure. Under the base-generation analysis, the subject-predicate relation that is formed between the focus phrase and the presuppositional CP through the operator movement provides a way to account for the case-matching effect in line with Ueyama's (1998) analysis of the Deep OS-type of 'scrambling' constructions shown in (58). Ueyama (1998:63), referring to Ueyama (1997), states that it is a case-marker on the DL (rather than the NP itself) that requires the empty operator movement. It is assumed in Ueyama 1998 that a case-marker is not properly interpreted without being syntactically related to the verb, and that (i) θ-marked positions are syntactically related to the verb and that (ii) the empty operator movement as in (58) forms a syntactic relation. If the same mechanism holds for CM-Cleft, the case-matching effect can also be accounted for under 70 the base-generation analysis. The above discussions show that the problems of the movement analyses pose no problem for the base-generation analyses. Moreover, the phenomena presented in support of the movement analyses can also be accounted for by the base-generation analyses. These are all in favor of the base-generation analyses of CM-Cleft over the movement analyses. 2.5.2 Between the base-generation analyses Kizu (2005) basically follows Hoji’s 1990 structure of CM-Cleft except for the position of the focus phrase as far as local clefts are concerned. As discussed in Chapter 3.4.2.1, the position of the focus phrase in Kizu 2005 poses two problems; (i) the position makes it difficult to establish a predication relation between the focus phrase in it and the presuppositional clause, and (ii) the Condition D effects cannot be accounted for if the focus phrase is situated in the position. Thus, I will pursue the base-generation analysis below based not on Kizu’s 2005 but Hoji’s 1987 (and subsequent works) analysis of CM-Cleft. 2.6 Summary In this Chapter I have summarized the base-generation analyses of cleft constructions in Japanese proposed by Hoji 1987 (as well as Hoji 1990, Ueyama & Hoji 2001 and Hoji & Ueyama 2003) and Kizu 2005, and the movement analyses proposed by Hiraiwa & Ishihara 2002 and Takahashi 2006. Technical details being aside, the base-generation analyses base-generate the focus phrase at the surface position, while the movement analyses get the focus phrase at the surface position by means of movement. I have then presented problems posed by these analyses or issues to be addressed. In regard to the movement analyses, H&I’s (2002) analysis faces the PBC related problem. 71 Though Takahashi’s (2006) analysis evades this problem, it would be quite difficult under his analysis, if not impossible, to account for the NSE facts and the Condition D effects in CM-Cleft. In addition, the possibilities of resumption are problematic for the movement analyses in general. These problems are irrelevant to or can be accounted for by the base-generation analyses, except that Kizu’s (2005) analysis has difficulty in accounting for the Condition D effects. Kizu’s (2005) analysis also makes it difficult to establish a predication relation between the focus phrase and the presuppositional clause. Thus, I will pursue the base-generation analysis based on Hoji’s 1987 (and subsequent works) analysis of CM-Cleft. Since the structure proposed in Hoji 1987 (and subsequent works) does not provide enough details, it is necessary to articulate the structure and provide supports for that. I will take up the task in the following chapters. 72 Chapter 3: Methodological Preliminaries: EPSA Method 3.1 Introduction In this work, I will adopt the methodology advocated by Hoji (2009 and subsequent works), which he calls the EPSA (Evaluation of Predicted Schematic Asymmetries) method. In this chapter, I will summarize the methodology and explain some of its important concepts that are crucially relevant to the present work. The goal of generative grammar is to discover the properties of the language faculty, which is assumed to be a part of the faculty of mind and to be most crucially responsible for our ability to produce and comprehend sentences in our language(s). 1 In regard to a research concerned with the properties of the language faculty, Hoji (2010c:1) pursues the thesis that it is possible to study the language faculty as an exact science. In order for language faculty science to be an exact science, we have to ensure that it possesses the properties in (1), which are most commonly understood to be the properties of an exact science. (1) Properties of an exact science a. Testability b. Reproducibility c. Quantitative analyzability 1 Hoji (2009:1, fn.1) points out as follows in regard to the goal of generative grammar: "It may be important to note that while it is often stated that one of the goals of generative grammar is to characterize and predict "the possible (actual and potential) occurrence of all and only the grammatical sentences of a given language," such is not a concern of the present work as it is concerned with the language faculty (more precisely, the Computational System) rather than with language (more precisely one particular language or another) although we do draw from informant judgments on sentences of a particular language." 73 Hoji (2010a) schematizes the general scientific method as schematized in (2), and he contends that language faculty science should also follow it. 2 (2) (Hoji 2010a: 2, (1)) The general scientific method (i.e., the hypothetico-deductive method): Guess — Computing Consequences — Compare with Experiment Under the hypothetico-deductive method in (2), we first form hypotheses. Then, we deduce prediction(s) from the hypotheses. We then conduct experiments and compare the prediction(s) with the experimental results. The EPSA method summarized here is a method of hypothesis testing in line with the hypothetico-deductive method, which is intended to ensure testability, reproducibility and quantitative analyzability in language faculty science. 3.2 Computational System and the model of judgment making A fundamental hypothesis in generative grammar is the existence of the Computational System at the core of the language faculty. The model of the Computational System (CS) suggested in Chomsky (1993) is schematized in Hoji 2010a as in (3). 2 Hoji (2010a:1) cites the following statements from Feynman 1965/94:150. "In general, we look for a new law by the following process. First we guess it. Then we compute the consequences of the guess to see what would be implied if this law that we guessed is right. Then we compare the result of the computation to nature, with experiment or experience, compare it directly with observation, to see if it works. If it disagrees with experiment, it is wrong. In that simple statement is the key to science. It does not make any difference how beautiful your guess is. It does not make any difference how smart you are, who made the guess, or what his name is—if it disagrees with the experiment, it is wrong. That's all there is to it." 74 (3) (Hoji 2010a: 27, (2)) The Model of the Computational System: Numeration μ=> CS => LF(μ) ⇓ PF(μ) Numeration μ: a set of items taken from the mental Lexicon LF(μ): an LF representation based on μ PF(μ): a PF representation based on μ The Computational System thus takes a numeration, which is a set of items taken from the Mental Lexicon, as input and computes a pair of mental representations, PF and LF representations; PF and LF representations are mental representations underlyingsound and meaning, respectively. 3 Hoji (2010a: 27) points out in regard to the model of the CS as follows: "Our hypotheses about the Computational System are thus meant to be about what underlies the language users' intuitions about the relation between sounds/signs and 'meanings' as reflections of properties of the Computational System. The main goal of generative grammar can therefore be understood as demonstrating the existence of the Computational System by discovering its properties." Then, the question is how we can discover properties that are reflections of the Computational System. The most commonly used data in actual research in generative grammar are based on acceptability judgments on a given sentence (under a specified interpretation). Therefore, what we deal with in an attempt to discover the properties of the Computational System is the informant's performance (i.e., language use in a broad sense). This makes it crucial that we articulate how we can extract from performance data (i.e., the informant's reactions) evidence for the hypotheses about the properties of the Computational 3 As is generally understood, "signs" in a sign language corresponds to "sound". Thus, the PF representations in a sign language are a mental representation underlying "signs". 75 System. Hoji (2010a:28) thus argues that in so far as informant judgments are assumed to be revealing about properties of the Computational System, the CS must be involved in the act of judgment-making by the informant. He then hypothesize, along with Ueyama's (2010), that when an informant makes his/her judgment, the informant comes up with a numeration μ and compares the two output representations based on μ with the 'sound' (i.e., the presented sentence α) and the relevant 'meaning' under discussion (i.e., the interpretation γ(a, b)). Hoji (2010a:29) thus adopts the model of judgment-making by the informant in Figure 1. Figure 1: The Model of Judgment-Making by the Informant on the acceptability of sentence α α α α with interpretation γ γ γ γ(a, b) (based on Ueyama (2010)): 4 - Compatibility Check - γ γ γ γ(a, b) Lexicon  α α α α ≈≈> Numeration Extractor ≈≈> μ => CS => LF(μ) => SR(μ) ⇓ PF(μ) ⇓ - Non-distinctness Check - pf(μ) ≈≈≈≈≈≈≈≈≈> β 4 The arrows in Figure 1) are interpreted as follows (Hoji 2010a: 29); The "==>" in Figure 1) indicates that a numeration is input to the Computational System (CS) and its output representations are LF and PF, and that SR and pf obtain based on LF and PF, respectively. What is intended by "≈≈>", on the other hand, is not an input/output relation, as roughly indicated in (i). (i) a. Presented Sentence α ≈≈> Numeration Extractor: ... contributes to ... b. γ(a, b) ≈≈> Numeration Extractor: ... contributes to ... c. Numeration Extractor ≈≈> numeration μ: ... forms ... d. γ(a, b) – Compatibility Check – SR(μ) ≈≈> Judgment β: ... serves as a basis for ... e. α - Non-distinctness Check - pf(μ) ≈≈> Judgment β: ... serves as a basis for ... 76 a. α: the presented sentence b. μ: numeration c. γ(a,b):the interpretation intended to be included in the 'meaning' of α involving expressions a and b d. LF(μ): the LF representation that obtains on the basis of μ e. SR(μ): the information that obtains on the basis of LF(μ) f. PF(μ): the PF representation that obtains on the basis of μ g. pf(μ): the surface phonetic string that obtains on the basis of PF(μ) h. β: the informant judgment on the acceptability of α under γ(a, b) In experiments in language faculty science, then, an informant is asked if a sentence α is acceptable under the interpretation γ(a, b). The informant tries to come up with a numeration μ, making reference to the Mental Lexicon (and also to the knowledge about "frequent sentence patterns," so to speak, that the informant must have "accumulated" through his/her linguistic experience). The CS takes the μ as input and yields LF(μ) and PF(μ) as output. Based on the LF(μ), the informant obtains SR(μ) and checks if it is compatible with γ(a, b). Likewise, based on the PF(μ), the informant obtains pf(μ) and checks if it is non-distinct from α. Compatibility between SR(μ) and γ(a, b) or the lack thereof as well as non-distinctness between pf(μ) and α or the lack thereof serve as bases for the judgment β. The crucial aspect of the model of judgment-making in Figure 1) is that what is judged by the informant is the acceptability of sentence α with interpretation γ(a, b), where γ(a, b) means a particular interpretation involving two expressions a and b, and not the simple acceptability of sentence α. Hoji (2009:69) maintains that the adoption of the model of judgment-making in Figure 1) is a consequence of (i) accepting Chomsky's (1993) model of the Computational System, (ii) committing ourselves to making our hypotheses about the Computational System empirically testable on the basis of the informants' acceptability judgments, and (iii) wanting to maximize our chances of turning the "failure" of our predictions to a discovery of properties of the Computational System. To focus on the acceptability of sentence α with interpretation γ(a, 77 b) is especially important to achieve (ii) and (iii). Even if the informant is asked if sentence α is acceptable and reports that it is totally unacceptable, we cannot necessarily attribute the unacceptability to the hypothesized grammatical reason without γ(a, b). This is because it is possible that the total unacceptability is due to parsing difficulty or the unnaturalness of the interpretation of the entire α, both of which are factors outside the Computational System. 3.3 Prediction making in language faculty science 3.3.1 Types of hypotheses Hoji (2009) takes the view that language faculty science he pursues consists of three components, taking the notions from Lakatos' 1970/1978 'scientific research programmes'. 5 (4) (Hoji 2009: 216,(5)) Three components of language faculty science: a. Hard core: the hypotheses that are adopted without direct empirical evidence and not subject to refutation or modification b. Auxiliary hypotheses: hypotheses that are subject to modification and abandonment c. Heuristics: research guidelines to follow Hoji (2009) assumes that the hypotheses in the hard core are of the following three types. (5) (Hoji 2009: 217, (6)) Three types of hypotheses in the hard core. a. hypotheses of the most general nature, not about specific properties of the Computational System 5 The term language faculty science was not introduced in Hoji 2009 yet. It was first introduced in Hoji 2010a. 78 b. hypotheses about the properties of the Computational System c. hypotheses about the properties of the Lexicon in general We have seen above that the following two hypotheses are fundamental hypotheses in generative grammar. These two hypotheses are regarded as (5a) type of hypotheses in the hard core. (6) a. The Computer System exists at the core of the language faculty. b. The input to the CS is a numeration (a set of items taken from the mental Lexicon) and its outputs are PF and LF representations (see (3) above). Since the hypothesis (6b) presupposes the existence of the mental Lexicon, (7) is regarded as another hypothesis of the (5a) type. (7) (Hoji 2009:219, (12)) The mental Lexicon exists. Furthermore, we have seen the model of judgment-making by informant proposed by Hoji (2010a) in Figure 1) above. Although the implementation of the particular model of judgment-making as in Figure 1) is yet to be open to dispute, we should at least regard the following as another (5a) type hypothesis in the hard core in so far as informant judgments are assumed to be revealing about properties of the Computational System. 79 (8) (Cf. Hoji 2009:218, (7).) The CS is embedded in the model of judgment making by the informant. In regard to the (5b) type hypotheses in the hard core, Hoji (2009) assumes the following hypothesis which seems to be adopted widely in recent years by those who adopt Chomsky's (1993) model of the CS. (9) (Hoji 2009:218, (11)) A hypothesis about the CS: There is an operation Merge, internal and external, and that is the only structure-building operation in the CS. These five hypotheses in (6)-(9) are in the hard core. Thus, we adopt them without direct empirical evidence and assume that they are not subject to refutation or modification. The hypotheses we form to deduce prediction(s) and to conduct experiments on are auxiliary hypotheses, which are subject to modification and abandonment based on the results of the experiments. Hoji (2010b:14) assumes that there are at least two types of auxiliary hypotheses: universal hypotheses and language-particular hypotheses. Hoji (2011:18) presents two types of universal hypotheses and four types of language-particular hypotheses, thus assuming six types of auxiliary hypotheses as listed in (10). (10) Types of Hypotheses: a. Universal hypotheses: (i) Hypotheses about the CS 80 (ii) Hypotheses about CS-related properties, but not just about the CS b. Language-particular hypotheses (i) Hypotheses about pf-LF correspondences (ii) Hypotheses about lexical properties that are referred to by hypotheses about the CS (iii) Hypotheses about lexical properties that are relevant to hypotheses about CS-related properties (iv) Lexical hypotheses not included above Let us illustrate these hypotheses by making reference to cases that involve BVA (Bound Variable Anaphora). 6,7 As introduced in Chapter 1, Ueyama (1998:2) uses the term BVA to refer to any anaphoric relation between a distributive/ quantificational expression (QP) and a singular-denoting anaphoric expression such that the value of the latter is not fixed to a certain individual. What we want to make prediction about BVA is when BVA is impossible and when it is not impossible. First of all, as for when BVA can arise, the following hypothesis is formed, where FD stands for Formal Dependency. (11) BVA(a, b) arises only if FD(LF(a), LF(b)). 8 As mentioned in Chapter 1 (fn.8), BVA(a, b) expresses an intuition and is not a theoretical term to express some properties of the CS. Thus, although FD and LF(x) (which stands for an LF object corresponding to x) in (11) are assumed to be theoretical terms to express properties of 6 The discussion here is based on Hoji 2011. 7 We will see more details about BVA in Chapter 5.2.1. 8 As we will see in Chapter 5.2.1, there are in fact three sources of BVA and (11) is one of the three sources. I will focus on this source of BVA here, leaving the other two aside, just for an expository purpose. 81 the CS, (11) as a whole is a universal hypothesis about CS-related properties of the (10a)-(ii) type. Next, the conditions on FD(α, β) are defined as in (12). This is regarded as a universal hypothesis about the CS since all the terms in (12) are assumed to express properties of the CS. (12) FD(α, β) only if: (i) α c-commands β at LF, and (ii) β is not marked [+Ref]. (12)-(i) refers to the c-command relation between a and b at LF. However, what the informant is presented in an experiment is merely surface strings of sentences, and the LF structure is invisible to the informant. Thus, we need to have hypotheses about (the tendency, if there is structural ambiguity, of) a correspondence between a surface string of a sentence and its LF representation. Such hypotheses are language-particular hypotheses about pf-LF correspondences of the type of (10b)-(i). (13) shows pf-LF correspondences assumed here, concerning A-ga B-o V and B-o A-ga V in Japanese. (13) A-ga B-o V corresponds to (i) at LF, while B-o A-ga V can correspond to (i) or (ii) at LF. (i) (ii) A-ga B-o B-o V A-ga ec V Furthermore, (12)-(ii) refers to the feature [+Ref] as a lexical property of β of FD(α, β), which is assumed to be a property of the CS. We need to know what lexical items in Japanese 82 are marked as [+Ref]. The hypothesis in (14) specifies that, and it is a language-particular hypothesis about lexical property that is referred to by hypothesis about the CS of the (10b)-(ii) type. (14) So-NPs in Japanese are not marked [+Ref], while a-NPs are marked [+Ref] On the other hand, the lexical property of α of FD(α, β) is not specified in (12). Thus the lexical property of α is not referred to by hypothesis about the CS. However, a of BVA(a,b) cannot be singular-denoting. Thus, we need to know what lexical items in Japanese are not singular-denoting and thus are qualified as a of BVA(a, b). Since BVA is an intuition and is not a theoretical term to express some properties of the CS, (15) is a language-particular hypothesis about lexical properties that are relevant to hypotheses about CS-related properties of the (10b)-(iii) type. (15) The following lexical items in Japanese are not singular-denoting. subete-no N 'every N' 55% izyoo-no N '55% or more N' NP-sae 'even NP' NP-sika 'nothing/no one but NP' Toyota to Nissan 'Toyota and Nissan' etc. On the other hand, b of BVA(a,b) must be singular-denoting. Thus, we need to know what lexical items in Japanese are singular-denoting and thus can qualify as b of BVA(a, b), which is shown in (16). (16) is also a language-particular hypothesis about lexical properties that are 83 relevant to hypotheses about CS-related properties of the (10b)-(iii) type. (16) The following lexical items in Japanese are singular-denoting. so-ko/a-soko 'it/that institution' so-re/a-re 'it/that thing' so-itu/a-itu 'he/that guy' so/a-no zidoosya-gaisya 'that automobile company' so/a-no daigaku-insei 'that graduate student' With the hypotheses in (11)-(15) above, we can make predictions about when BVA(a, b) is impossible and when it is not impossible, provided that a is one of the phrases listed in (15) and b is a so-NP or an a-NP. 9 We will see the actual predictions made by these predictions in Chapter 4.5.2.1 below and the experimental results in the next chapter. 3.3.2 Disconfirmability and quantitative analyzability As mentioned above, specific predictions in the language faculty science are about the acceptability status of sentence α under γ(a, b). Since our goal is to discover the properties of the language faculty, more in particular those of the Computational System, the interpretation γ(a, b) needs to be a linguistic intuition that is a reflection of properties of the CS. As noted above, Hoji (2009: 218) assumes that an operation Merge is the only structure-building operation in the CS. Given this, γ(a, b) that is crucially based on the order of Merge is a good candidate for a linguistic intuition that is a reflection of properties of the CS. Merge takes a 9 Predictions about when BVA(a, b) is impossible correspond to *Schema-based predictions, while predictions about when it is not impossible correspond to ok Schema-based predictions, which will be introduced shortly. 84 pair of syntactic objects (SO i , SO j ) and replace them by a new combined syntactic object SO ij (Chomsky 1995:226). 10 The structural relation c-command is derivative of the characterization of the operation Merge. When X is Merged with something that contains Y, X c-commands Y. Therefore, we can rephrase the last statement as follows: γ(a, b) that is crucially based on the c-command relation between an LF object corresponding to a and one corresponding to b is a good candidate for a linguistic intuition that is a reflection of properties of the CS. 11 We can try to ensure that informants' judgments are a reflection of the (hypothesized) 10 Chomsky (2001) distinguishes between two types of Merge: Internal Merge and External Merge. Chomsky 2007: 16, for example, remarks as in (i). (i) As a simple matter of logic, there are two kinds of Merge, external and internal. External Merge (EM) takes two objects, say 'eat' and 'apples', and forms the new object that corresponds to 'eat apples'. Internal Merge (IM)—often called Move—is the same, except that one of the objects is internal to the other. 11 Hoji (2009:258) calls this as the Reinhartian heuristic, referring to the two most general hypotheses put forth in Reinhart 1983 in (i) and (ii). (i) (Reinhart 1983: 25, (19)) Sentence-level semantic interpretation rules may operate on two given nodes A and B only if one of these nodes is in the domain of the other (i.e., A is in the domain of B, or B is in the domain of A, or both). (ii) (Reinhart 1983: 26, (21)) If a rule assigns node A some kind of prominence over node B, A must be a D-head of the domain which contains B. Hoji (2009:258) restates (i) and (ii) as in (iii) and (iv), respectively, taking "sentence-level semantic interpretation rules" as "CS-based rules or conditions that contribute to or regulate interpretive possibilities," which he thinks seems to be a reasonable interpretation, given the discussion in Reinhart 1983. (iii) (Hoji 2009:258, (56), restated Reinhart 1983: 25, (19)) CS-based rules or conditions that contribute to or regulate interpretive possibilities can involve A and B only if A c-commands B, or B c-commands A, or both. (iv) (Hoji 2009:259, (57), restated Reinhart 1983: 26, (21)) If B is dependent upon A in terms of how B gets interpreted, B must be c-commanded by A. Hoji (2009) takes the statements in (iii) and (iv) as applying at LF. (iii) is a hypothesis about the form of a CS-related hypothesis that pertains to an interpretation involving two elements; (iv) is an instance of a bridging statement. Hoji (2009) thus understands Reinhart 1983 as containing a heuristic like (v). (v) (Hoji 2009:259, (58)) The Reinhartian heuristic: The relation at LF that underlies γ(a, b) mentioned in a bridging statement must be based on a c-command relation between LF(a) and LF(b). 85 properties of the Computational System by invoking γ(a, b) that is crucially based on the c-command relation and designing our experiments accordingly. In order for language faculty science to be an exact science, it is most crucial to make disconfirmable predictions. Hoji (2010b:20) argues that we can attain disconfirmability on the basis of reproducibility and quantitative analyzability. We say that a prediction is disconfirmed when we have obtained the same experimental results against the prediction repeatedly. Then, how do we determine whether we have obtained the same results in an experiment? Hoji (2010b:21) argues that in order to determine whether we have obtained the same experimental results, it is necessary to "quantize" experimental results, that is, the experiments in language faculty science have to have "quantitative analyzability." According to Hoji (2010a:30), under the model of judgment-making by the informant on the acceptability of sentence α with interpretation γ(a, b) in Figure 1, if the informant fails to come up with a numeration μ that would result in pf(μ) non-distinct from α and SR(μ) compatible with the interpretation γ(a, b), the informant's judgment on α under γ(a, b) should necessarily be "complete unacceptability". This is because, in that case, the informant fails to arrive at SR(μ) compatible with the interpretation γ(a, b) as the hypothesized structural condition necessary for γ(a, b) is never met in any LF(μ) no matter what possible μ might be tried. On the other hand, even if the informant (eventually) finds such a numeration μ, that may not necessarily result in the informant reporting that α is (fully) acceptable under γ(a, b). This is because the informant judgment β can be affected by difficulty in parsing and the unnaturalness of the interpretation of the entire sentence in question, the factors outside the Computational System. Therefore, it is possible to make point-value predictions in language faculty science in terms of zero ("complete unacceptability") vs. non-zero, where we can obtain quantitative analyzability in language faculty science. 86 3.3.3.*Schema-based predictions and ok Schema-based predictions The above consideration leads to the two types of predictions in language faculty science, that is, a *Schema-based prediction and an ok Schema-based prediction. The discussions and experiments on CM-Cleft in this work make crucial reference to these notions, so that I take up these notions in some details in this subsection. A *Schema-based prediction and an ok Schema-based prediction are formulated as in (17) and (18), respectively. *Examples and ok Examples are sentences conforming to a *Schema and an ok Schema, respectively. (17) (Hoji 2010a: 31, (6)) A *Schema-based prediction: Informants judge any *Example conforming to a *Schema to be completely unacceptable under interpretation γ(a, b). (18) (Hoji 2010a: 31, (7)) An ok Schema-based prediction: 12 Informants judge ok Examples conforming to an ok Schema to be acceptable (to varying degrees) under interpretation γ(a, b). Hoji (2010a:31) then states the fundamental asymmetry as follows: " ok Schema-based predictions cannot be disconfirmed and they can only be confirmed; *Schema-based predictions, on the 12 Hoji (2010:30) points out that (18) is one of the possible formulations of an ok Schema based prediction. He presents other possible formulations as in (i), which is less stringent than (18) because the existence of just one ok Example that is judged to be acceptable would confirm (8), and as in (ii), which is more stringent. (i) (Hoji 2010:31, (8)) An ok Schema-based prediction—version 2: Informants judge some ok Example conforming to an ok Schema to be acceptable (to varying degrees) under interpretation γ(a, b). (ii) (Hoji 2010:33, (9)) An ok Schema-based prediction—version 3: Informants judge every ok Example (in an experiment) conforming to an ok Schema to be fully acceptable under interpretation γ(a, b). 87 other hand, can be disconfirmed although they cannot be confirmed because it is not possible to consider all the *Examples that would conform to a *Schema." Hoji (2010a:32) assumes that the ultimate testability of our hypotheses lies in their being subject to disconfirmation. Given this, what makes hypotheses in language faculty science testable is the *Schema-based predictions that are obtained based on these hypotheses. Now we need to ensure that the unacceptability of *Examples conforming to a *Schema is indeed due to the hypothesized property/ies of the CS. For this purpose, an ok Schema corresponding to the *Schema is constructed such that it minimally differs from the *Schema in terms of the hypothesized property/ies of the CS; it need to be demonstrated that examples confirming to such an ok Schema are acceptable (at least to some degree) under γ(a, b). The acceptability of ok Examples conforming to such an ok Schema indicates that the unacceptability of the *Examples is likely to be due to the hypothesized property/ies of the CS. Hoji (2010a:32) argues that it is necessary to introduce another type of ok Schema to complete the minimum paradigm of schematic asymmetry to be tested. This type of ok Schema is identical to the *Schema except that it does not involve interpretation γ(a, b). The acceptability of ok Examples conforming to this type of ok Schema indicates that the unacceptability of the *Examples is unlikely to be due to parsing difficulty. Thus the minimum paradigm of schematic asymmetry, i.e., a predicted schematic asymmetry, that need to be tested comes as in (19). (19) (Hoji 2010c:7) The minimum paradigm requirement: a. ok Schema (involving the same γ(a, b) as the *Schema does and the "entire interpretation" of the corresponding ok Examples is quite similar to the intended interpretation of the *Examples corresponding to the *Schema) b. *Schema (involving γ(a, b)) 88 c. ok Schema (= identical to the *Schema but without involving γ(a, b)) Hoji (2010a:32) states that a predicted schematic asymmetry of the form in (19) gets confirmed "if and only if the informants' judgments on *Examples are consistently completely unacceptable and their judgments on the corresponding ok Examples are not completely unacceptable." Thus, the *Schema-based prediction must survive a rigorous disconfirmation attempt, while at the same time the corresponding ok Schema-based predictions must be confirmed. When a predicted schematic asymmetry has passed such a test, we say that we obtain a confirmed schematic asymmetry. Hoji (2010a:32) suggests that "confirmed schematic asymmetries be regarded as "basic units of facts" for research concerned with the properties of the Computational System, i.e., as long as our research is concerned with the properties of the Computational System of the language faculty, our hypotheses should make predictions about, and be evaluated in terms of, confirmed schematic asymmetries." 3.4 Experiments in language faculty science 3.4.1 EPSA We test predicted schematic asymmetries constructed as in (19) by checking informant judgments on *Examples and corresponding ok Examples and see if the predicted schematic asymmetries are confirmed. I will briefly introduce the general design of experiments that have been conducted by Hajime Hoji and his research group. The program for the experiments has been created by Ayumi Ueyama and the method of hypothesis testing is called the EPSA (Evaluation of Predicted Schematic Asymmetries). 13 The experiments on CM-Cleft I will present in the following chapters utilize the EPSA. 13 The details of the experiments presented in this subsection are mostly based on Hoji 2010a, Chapter 3. 89 The examples constructed on the basis of the schema set in (19), a *Schema and two corresponding ok Schemata, are presented on-line to the informants, along with the specification of their intended interpretation. The examples are presented to informants in a random fashion, one at a time or three in a schema set at a time (though the order within a schema set is randomly chosen), depending upon the test type chosen by each informant. Moreover, depending on the test type of their choice, the informants either (i) choose "No" (for "not acceptable no matter what") or "Yes" (for "(more or less) acceptable"), or (ii) indicate how acceptable they find each example by clicking one of the five radio buttons as in (20). Then, what the informant has indicated is converted to numerical values as in (21); i.e., the worst score is converted to 0 and the best score to 100. 14 In the same light, the "Yes" and the "No" answers in the "Yes-or-No" test get converted to 0 and 100, respectively. 15 (20) Bad ⇔ Good o o o o o (21) 0, 25, 50, 75, 100 The informants can return to the experiment website and report their judgments on examples in the same experiment as many times as they wish; they may repeat the same "test type" as before or try a different "test type" (as to "Yes-or-No" or "Five-ranking" and also as to "one at a time" or "three at a time"). In the event that one informant has reported his/her judgment on the same experiment more than once (regardless of the "test type"), the average of that informant's 14 The survey sheet/page does not indicate the numeric values assigned to the five radio buttons. 15 Hoji (2010a:39) reports that "[a]ccording to the results we have obtained so far, the choice of the "test type" does not make a significant difference." 90 judgments on a given example is used in calculating the mean score on the example by the entire informants. According to the *Schema-based prediction in (17), the average scores of *Examples conforming to the *Schema should be close to 0. This is because the informant in that case fails to arrive at SR(μ) compatible with the interpretation γ(a, b) as the hypothesized structural condition necessary for γ(a, b) is never met in any LF(μ) no matter what possible μ might be tried. 16 On the other hand, according to the ok Schema-based prediction in (18), the average scores of ok Examples conforming to the ok Schemata should be in principle higher than 0. In reality, however, the average scores of ok Examples should be significantly higher than 0, and even close to 100. 17 As for ok Schema, the informant can come up with a numeration μ that would result in pf(μ) non-distinct from α and SR(μ) compatible with the interpretation γ(a, b), and thus ok Examples conforming to ok Schema are not completely unacceptable. When the informant finds some ok Examples not completely acceptable, it should be attributed to parsing difficulty or the unnaturalness of the interpretation of the entire α, both of which are factors outside the Computational System. It is also possible that the informant does not understand the intended interpretation properly due to ineffective instructions. 18 Researchers should try to minimize such possibilities in the course of preparing the experiment by carefully choosing the 16 We should allow some room for errors by informants due to the factors outside of the CS, such as the lack of attentiveness or patience. 17 When the score of ok Examples conforming to an ok Schemata is low, it is most likely to be due to parsing difficulty or the unnaturalness of the interpretation of the entire α. The same factor would affect the judgment of the corresponding *Schema. Thus, in order to ensure that the unacceptability of *Examples conforming to the *Schemata is in fact due to the hypothesized structural condition necessary for γ(a, b), researchers should try to come up with ok Examples which are readily acceptable (at least to the researchers themselves) so as to be able to ensure that the status of *Examples is not due to parsing difficulty or the unnaturalness of the interpretation of the entire α. 18 Hoji (2011:35) also points out that general attentiveness and patience of the informant also affect the informant's judgment. 91 relevant lexical items or preliminarily checking the effectiveness of instructions, or conducting training sessions or preliminary experiments. In the following subsection, I will explain in some details about preliminary experiments. 3.4.2 Preliminary experiments Hoji (2011:34) maintains that preliminary experiments play important roles in language faculty science. The function of preliminary experiments, he assumes, is to determine the effectiveness of various experimental devices for experiments in language faculty science. Experimental devices include hypotheses, schemata, examples, informants and instructions. We have to ensure the effectiveness of these experimental devices in conducting experiments. Hoji (2011:34) argues that by using effective experimental devices backed up by results of preliminary experiments, we can have a better chance of obtaining a confirmed schematic asymmetry in a more involved experiment. In particular, hypotheses to be used in a more involved experiment should be the ones that are backed up by confirmed schematic asymmetries obtained in less involved experiments. For example, it seems to be valid to assume that experiments (focusing) on universal properties are more involved than experiments (focusing) on language-particular properties. Thus, an experiment on the language-particular hypothesis can serve as a preliminary experiment for the one that involves a universal hypothesis as well as the language-particular hypothesis. In the same light, experiments on the OSV construction are more involved than experiments on the SOV construction since the former construction are assumed to be structurally more complicated. Likewise, experiments on the multiple OSV construction, the long-distance OSV construction and the OSV construction with resumption are certainly more involved than experiments on the simple OSV construction. Thus, an experiment on the SOV construction can serve as a preliminary 92 experiment for the one on the OSV construction as well as another experiment on the SOV construction. Then, an experiment on the OSV construction can in tern serve as a preliminary experiment for the one on the more involved OSV construction as well as another experiment on the simple OSV construction. Hoji (2011:34) also argues, making reference to Duhem's thesis, that using effective experimental devices backed up by preliminary experiments helps us to identify what is wrong more easily when we fail to obtain a confirmed schematic asymmetry in a more involved experiment. 19 More specifically, since the other devices have been shown effective by the preliminary experiments, it is most likely that the newly introduced hypothesis we are pursuing has something wrong. This leaves us only two choices to take, either abandon or modify the new hypothesis. 20 3.5 Research heuristics Recall the three components of language faculty science assumed in Hoji 2009 in (4) above: hard core, auxiliary hypotheses and heuristics. Among the three components, heuristics are research guidelines for language faculty scientists to follow. Two heuristics are proposed in Hoji 2010a, which are shown in (22). (22) Research heuristics in language faculty science: a. Maximize testability b. Maximize our chances of learning from errors 19 See (25) below. 20 In this regard, Hoji’s (2011:34) actual statement is as follows: "Equally importantly, this [i.e., using effective experimental devices backed up by preliminary experiments, YM] also reduces our chances of making "excuses" when we fail to obtain a confirmed schematic asymmetry in a more involved experiment." 93 3.5.1 Maximize testability Hoji (2010a: 42) presents the following two aspects of the "Maximize testability" heuristic. (23) (Hoji 2010a:42, (29)) The two aspects of the "Maximize testability" heuristic: a. We should work on 'phenomena' that (are likely to) lead us to confirmed schematic asymmetries. b. We should pursue hypotheses that give rise to as many *Schema-based predictions as possible, in a non-trivial manner. As noted above, confirmed schematic asymmetries are regarded as "basic units of facts" for language faculty science, whose goal is to discover the properties of the Computational System. Hoji (2010a:42) argues that, as far as the research follows the general scientific method in (2), it is necessary to have a means to identify "facts" to be accounted for by hypotheses in order to attain testability. He explains that it is because without such a means we could not tell the exact empirical content of our predictions (i.e., what we deduce from our hypotheses) and hence could not compare them with the results of our experiments. Hoji (2010a:42) thus maintains that one way to maximize testability in a research is to try to identify, and work on, 'phenomena' that (are likely to) lead us to confirmed schematic asymmetries. In regard to the second aspect of the "Maximize testability" heuristic, we have seen above that ok Schema-based predictions cannot be disconfirmed and they can only be confirmed, while *Schema-based predictions can be disconfirmed although they cannot be confirmed. Testability resides in disconfirmability, so that we can seek to attain testability of our hypotheses most crucially by making *Schema-based predictions, which can be disconfirmed. In this respect, Hoji (2010a:42) thus argues that "the maximization of testability can be pursued 94 not only by choosing to investigate certain phenomena but also by pursuing their accounts such that they lead to as many *Schema-based predictions as possible." Hoji (2010a:34) also suggests that when we modify our hypotheses in response to the failure of our predictions, we should proceed to modify our hypotheses in line with (23b). He points out that "[i]f the modification only resulted in the elimination of a *Schema-based prediction without introducing a new *Schema-based prediction, that would be like a content-reducing (i.e., degenerating) problemshift in the terms of Lakatos (1970/1978)." Thus, in order to make a progress in language faculty science, we have to make sure that we follow the heuristics in (23). 3.5.2 Maximize our chances of learning from errors As for the second heuristic in (22), "Maximize our chances of learning from errors" heuristic, Hoji (2010a:43) cites the following statement from Duhem 1909/1954: (24) (Duhem 1909/1954: chapter 6, section 2: 185): [I]f the predicted phenomenon is not produced, ... [t]he only thing the experiment teaches us is that among the propositions used to predict the phenomenon and to establish whether it would be produced, there is at least one error; but where this error lies is just what it does not tell us. In order to learn from errors, then, we need to identify where the errors lie. Hoji (2010a:43) suggests that "[w]e can try to maximize our chances of learning from errors by minimizing the number of hypotheses whose validity is to be tested in a given experiment, and more importantly, also by not using hypotheses that have been shown to be invalid in earlier 95 experiments." 3.5.2.1 Some illustrations: BVA case Recall the discussion about the types of hypotheses in Chapter 4.3.1. Five hypotheses have been introduced to make predictions about when BVA(a, b) is impossible and when it is not impossible, which are repeated below. (11) BVA(a, b) arises only if FD(LF(a), LF(b)). (12) FD(α, β) only if: (i) α c-commands β at LF, and (ii) β is not marked [+Ref]. (25) A-ga B-o V corresponds to (i) at LF, while B-o A-ga V can correspond to (i) or (ii) at LF. (i) (ii) A-ga B-o B-o V A-ga ec V (14) So-NPs in Japanese are not marked [+Ref], while a-NPs are marked [+Ref] (15) The following lexical items in Japanese are not singular-denoting. subete-no N 'every N' 55% izyoo-no N '55% or more N' NP-sae 'even NP' 96 NP-sika 'nothing/no one but NP' Toyota to Nissan 'Toyota and Nissan' etc. One schema set obtained from these five hypotheses is in (26), where crucially in the *Schema, so-NP is not c-commanded by the QP given the LF structure in (13). (26) a. ok Schema A QP-ga so-NP-no N-o/ni V BVA (QP, so-NP) b. *Schema B so-NP-no N-ga QP-o/ni V BVA (QP, so-NP) c. ok Schema C so-NP-no N-ga QP-o/ni V (With so-NP "referring to" a particular individual/object) Hoji has been conducting experiments on this schema set in EPSA[1]-1, 2, 3 and 18. 21,22 The "intended binder" in the examples in EPSA[1]-1, 2 and 3 are kanarinokazu-no NP 'most of the NPs', 55% izyoo-no NP '55% or more NPs' and NP-sae 'even NP', respectively, while those in the examples in EPSA[1]-18 are subete-no NP 'every NP' and 3-ttu-no NP 'three NP'. The following is a set of examples with kanarinokazu-no NP as the "intended binder." 21 In the following discussion, I will make reference to specific on-line experiments under the EPSA by "EPSA[n]-m," by using the actual number assigned to each experiment in the program. 22 These EPSA contain not only the schema set in (26) but also other schema sets. Here I only present the results concerning the schema set in (26) in these EPSA. 97 (27) Example group (QP = kanarino kazu-no N) a. ok Example A-1 Kanarino kazu-no tihoozititai-ga so-ko-no syokuin-o sinbunsizyoo-de hihansita. considerable number-GEN local:government-NOM that-place-GEN agent-ACC newspaper-in criticized 'Each of a considerable number of local governments criticized its agents in the newspaper.' b. *Example B-1 So-ko-no syokuin-ga kanarino kazu-no tihoozititai-o sinbunsizyoo-de hihansita. that-place-GEN agent-NOM considerable number-GEN local:government-ACC newspaper-in criticized 'Its agents criticized a considerable number of local governments in the newspaper.' c. ok Example C-1 So-ko-no syokuin-ga kanarino kazu-no tihoozititai-o sinbunsizyoo-de hihansita. that-place-GEN agent-NOM considerable number-GEN local:government-ACC newspaper-in criticized 'That place’s agents criticized a considerable number of local governments in the newspaper.' A summary of the results of EPSA [1]-1, 2, 3 and 18 in regard to the schema set in (26) is shown in Table 3.1. 23 23 Some informants have judged the same example more than once; but in such cases the values in the summary chart are based on the average score on a given example by each such informant counting it as one value. 98 Table 3.1: A summary of the results of EPSA [1]-1, 2, 3 and 18 in regard to Schema set (26) QP as A ok Schema A *Schema B ok Schema C EPSA [1]-1: 24 participants kanarinokazu-no NP 'most of the NPs' 47 values 97 47 values 10 48 values 93 EPSA [1]-2: 26 participants 55% izyoo-no NP '55% or more NPs' 52 values 98 51 values 4 51 values 96 EPSA [1]-2: 17 participants NP-sae 'even NP' 34 values 97 34 values 6 34 values 96 EPSA [1]-18: 10 participants subete-no NP 'every NP' 10 values 100 10 values 51 10 values 100 3-ttu-no NP 'three NP' 10 values 100 10 values 55 10 values 100 What is crucial in this result is that the average scores on the *Schema are very high when subete-no NP 'every NP' and 3-ttu-no NP 'three NP' are used as the "intended binder" (i.e., a of BVA(a, b)) compared to the scores for the other three QPs, which are substantially low. According to the *Schema-based prediction in (17), the average scores of *Schema should be close to 0. The high scores on the *Schema with subete-no NP 'every NP' and 3-ttu-no NP 'three NP' thus clearly disconfirm the *Schema-based prediction. This means that there is at least one error on the hypotheses in (11)-(15). The pf-LF correspondences in (13) are widely accepted in the literature of Japanese. The hypothesis in (14) has been shown valid by obtaining confirmed schematic asymmetries in other experiments conducted by Hajime Hoji, whose details are shown in the next chapter. As for the hypothesis (15), there would be no objection that the QPs listed there as well as kanarinokazu-no NP 'most of the NPs' and 3-ttu-no NP 'three NP' are not singular-denoting. Therefore, we can attribute the disconfirmation of the *Schema-based prediction to either the hypothesis in (11) or (12). Returning to the result in Table 3.1, at least for some QPs, the predicted schematic asymmetry has been confirmed. For 99 these QPs, the hypothesis in (12) seems to be valid. Thus, it seems reasonable here to assume that there is an error in the hypothesis in (11), more specifically, FD(LF(a), LF(b)) may not be a sole source of BVA(a, b). Given that the condition that must be satisfied by another source of BVA as P, we can modify the hypothesis in (11) as in (28). 24 (28) BVA(a, b) arises only if (i) FD(LF(a), LF(b)), or (ii) P is satisfied Is the modification in (28) a content-reducing (i.e., degenerating) problemshift mentioned above? If the modification only resulted in the elimination of a *Schema-based prediction without introducing a new *Schema-based prediction, that would be a content-reducing (i.e., degenerating) problemshift. The *Schema in (26) stays intact as far as the hypothesis in (28)-(i) is concerned. However, we need to be able to distinguish between (28)-(i) and (28)-(ii) to test the modified hypothesis. Otherwise, the modification would be regarded as a content-reducing (i.e., degenerating) problemshift. What is obvious in the above discussion is that the types of QPs have something to do with P. Ueyama (1998) in fact proposes the conditions on another source of BVA, which include the one on the property of QPs, and concludes that this source of BVA does not belong to the CS. Thus, if we exclude this source of BVA (i.e., (28)-(ii)) and focus on the BVA of (28)-(i), that is, FD(LF(a), LF(b)), which is assumed to be a property of the CS, we can keep the *Schema-based prediction and would be 24 This another source of BVA(a, b) is related to what Ueyama (1998) calls 'quirky binding', but we will postpone the detailed discussion of this until next chapter. In fact, Ueyama (1998) introduces one more source of BVA, so that three sources of BVA are assumed in Ueyama 1998. 100 able to increase a chance to obtain a confirmed schematic asymmetry. 25 3.5.2.2 Invalid hypotheses As mentioned above, we can try to maximize our chances of learning from errors by not using hypotheses that have been shown to be invalid in earlier experiments. If a *Schema-based prediction got disconfirmed and we were able to identify which hypothesis is not valid based on the above considerations, we should not examine further predictions on the basis of that hypothesis, combined with other new hypotheses. Suppose that we combined such an invalid hypothesis with other new hypotheses (language-particular and/or universal) and made a new prediction. When the new *Schema-based prediction got disconfirmed, we could not tell that it is due to the newly-introduced hypothesis/hypotheses because the original hypothesis is invalid, to begin with. Hoji (2010a:44) argues in this respect that "[a] general research heuristic is therefore that if a testable hypothesis is shown to be invalid in what is considered to be a "simple case", we must avoid using it for what we consider to be a "more complex case"," which he assumes is in line with the "Maximize our chances of learning from errors" heuristic. Hoji (2010a:45) then concludes as follows: "We can attain testability of our hypotheses most effectively by making reference to confirmed schematic asymmetries, and most crucially to *Schema-based predictions, provided that the ultimate testability of our hypotheses lies in their being subject to disconfirmation. Our emphasis on the importance of building our research on confirmed schematic asymmetries and most crucially on *Schema-based predictions thus derives from the "Maximize our chances of learning from errors" heuristic." 25 We will see some more details in Chapter 4.2.1.2. 101 3.6 Summary In this chapter, I have summarized the EPSA method advocated by Hoji (2009 and subsequent works) for the research concerned with the properties of the language faculty. Hoji (2010b) refers to such research as "language faculty science" and maintains that it is possible to study the language faculty as an exact science, whose characteristics, he states, are testability, reproducibility and quantitative analyzability. Hoji (2010a) contends that language faculty science as an exact science should follow the general scientific method in Feynman 1965/94 as in (2), so-called the hypothetico-deductive method. Under the hypothetico-deductive method in (2), we form hypotheses and then deduce prediction(s) from the hypotheses. We then conduct experiments and compare the prediction(s) with the experimental results. The EPSA method is a method of hypothesis testing in line with the hypothetico-deductive method, by following which we can hope to attain testability, reproducibility and quantitative analyzability in language faculty science. A fundamental hypothesis in generative grammar is the existence of the Computational System at the core of the language faculty. The main goal of generative grammar can therefore be understood as demonstrating the existence of the CS by discovering its properties. Since what we deal with in an attempt to discover the properties of the CS is one's linguistic performance, it is crucial that we articulate how we can extract from performance data evidence for the hypotheses about the properties of the CS. Hoji (2010a:29) adopts from Ueyama 2010 the model of judgment-making by informant in Figure 1), where what is judged by informant is the acceptability of sentence α with interpretation γ(a, b). Focusing on the interpretation γ(a, b) is particularly important for the purpose of making our hypotheses about the CS empirically testable on the basis of the informants' acceptability judgments. Thus, specific predictions in language faculty science are 102 about the acceptability status of sentence α under γ(a, b). Since our goal is to discover the properties of the CS, γ(a, b) needs to be a linguistic intuition that is a reflection of properties of the CS. Given that an operation Merge is the only structure-building operation in the CS, γ(a, b) that is crucially based on the c-command relation between an LF object corresponding to a and one corresponding to b is a good candidate for a linguistic intuition that is a reflection of properties of the CS. In order for language faculty science to be an exact science, it is most crucial to make disconfirmable predictions based on the hypotheses. Hoji (2010b) maintains that we attain disconfirmability on the basis of reproducibility and quantitative analyzability. Under the model of judgment-making in Figure 1), quantitative analyzability in language faculty science is in terms of zero ("complete unacceptability") vs. non-zero ("unacceptability to varying degrees "). Specifically, we form two types of predictions based on the hypotheses: *Schema-based predictions and ok Schema-based predictions. Only *Schema-based predictions can be disconfirmed. Therefore, what makes the hypotheses in language faculty science testable is the *Schema-based predictions. However, a *Schema-based prediction is not enough to conduct an experiment. We need two types of ok Schema to back up the *Schema. One ok Schema minimally differs from the *Schema in terms of the hypothesized properties of the CS, which isassumed to be acceptable under γ(a, b). The acceptability of such an ok Schema indicate that the unacceptability of the *Schema is due to the hypothesized properties of the CS. The other ok Schema is identical to the *Schema except that it does not involve γ(a, b). The acceptability of such an ok Schema would indicate that the unacceptability of the *Schema cannot be due to parsing difficulty or the unnaturalness of the interpretation of the entire sentence. A predicted schematic asymmetry so formed gets confirmed, i.e., a confirmed schematic asymmetry obtains, if and only if the informants' judgments on *Examples are consistently 103 completely unacceptable and their judgments on the corresponding ok Examples are not completely unacceptable. Confirmed schematic asymmetries so obtained are regarded as "basic units of facts" for research concerned with the properties of the CS. As long as our research is concerned with the properties of the CS, our hypotheses should make predictions about, and be evaluated in terms of, confirmed schematic asymmetries. We have seen two research heuristics in language faculty science proposed by Hoji (2010a): the "Maximize testability" heuristic and the "Maximize our chances of learning from errors" heuristic. In regard to the former heuristic, we should work on 'phenomena' that (are likely to) lead us to confirmed schematic asymmetries and pursue hypotheses that give rise to as many *Schema-based predictions as possible. As for the latter heuristic, we can maximize our chances of learning from errors by minimizing the number of hypotheses whose validity is to be tested in a given experiment and by not using hypotheses that have been shown to be invalid in earlier experiments. In what follows, I will conduct experiments on the cleft construction in Japanese by adopting the EPSA method and by following the research heuristics just outlined, hoping to reveal some of the properties of the Computational System. 104 Chapter 4: Preliminaries of the Experiments 4.1 Overview Once the methodology adopted in this work is in hand, we are now ready to conduct experiments on CM-Cleft. However, we first need to make sure that we are using 'valid tools' in our experiments. What do 'valid tools' mean? According to Hoji (2010b:14), predictions in language faculty science are based on at least two types of hypotheses: universal hypotheses and language-particular hypotheses. Suppose that we are trying to reveal properties of CM-Cleft which are reflections of the Computational System. We will form a hypothesis about a property of CM-Cleft. Then, combining the hypothesis with another hypothesis or other hypotheses, either universal or language-particular, we will deduce a prediction. We will then conduct an experiment to test the prediction. What is particularly important here is that we should do our best to ensure that the hypothesis/hypotheses to be combined with the one about the property of CM-Cleft has/have been shown to be valid (or, at least, has/have not been shown to be invalid) in prior experiments. In terms of the EPSA (Evaluation of Predicted Schematic Asymmetries) method adopted here, such (a) hypothesis/hypotheses must be the one(s) backed up by a confirmed schematic asymmetry. As mentioned in the previous chapter, confirmed schematic asymmetries are regarded as "basic units of facts" for research concerned with the properties of the Computational System. Thus, when we obtain confirmed schematic asymmetries in accordance with predictions made under a hypothesis, it can be said that we acquire new tools for our further prove into properties of the Computational System because we can assume the hypothesis to be valid in our further investigation and experiments (see Hoji 2010:44). Without using hypotheses backed up by confirmed schematic asymmetries, we won’t be able to expect to obtain robust informant judgments in the experiments. Furthermore, 105 when the newly formed predicted schematic asymmetries are fail to be confirmed, we will find difficulty in identifying the culprit. In this chapter I will present such confirmed schematic asymmetries that have independently been established on which the experiments on CM-Cleft in this work are crucially based. I will present three confirmed schematic asymmetries established in other works. The first two are BVA-related confirmed schematic asymmetries; one is on A and the other on B of BVA (A,B). The other is DR-related confirmed schematic asymmetries. I will also present hypotheses about NSEs (Neg(ation)-sensitive elements) proposed by Kataoka (2007), which I crucially utilize in the following experiments on CM-Cleft. Though Kataoka (2007) has not obtained confirmed schematic asymmetries by conducting EPSA experiments, her hypotheses seem to be quite tenable considering the data and arguments presented in her work. I will also present her data and arguments below. 4.2 Confirmed schematic asymmetries 4.2.1 BVA (bound variable anaphora) I will crucially rely on the BVA facts of CM-Cleft in my experiments. I will therefore first present independently established confirmed schematic asymmetries in regard to BVA. As mentions in Chapter 1, the term BVA is employed by Ueyama (1998) to refer to any anaphoric relation between a distributive/quantificational expression (QP) and a singular-denoting anaphoric expression such that the value of the latter is not fixed to a certain individual and is construed instead as a variable bound by the QP. She expresses BVA between a QP A and a singular-denoting anaphoric expression B as BVA(A, B). A condition on BVA has informally been stated there, which is repeated here as (1). 106 (1) (Ueyama 1998:30, (13)) A dependent term in BVA must be c-commanded by the (QR-)trace of a QP at LF. However, three types of BVA are in fact introduced in Ueyama (1998), as indicated in (2), and (1) applies to only one of the three types of BVA. (2) a. FD (Formal Dependency)-based BVA b. ID (Indexical Dependency)-based BVA c. Quirky binding Ueyama (1998:155) claims that FD-based BVA and ID-based BVA are syntactic in nature. The properties of each type are shown in (3) and (4), respectively. (3) Properties of FD-based BVA (Hoji 2003: 43, (105)) a. It is sensitive to c-command. b. Reconstruction effects are observed. c. Local disjointness effects are observed. d. The 'semantic content' of the 'bindee' makes a difference (4) Properties of ID-based BVA (based on Hoji 2003: 43, (105)) a. It is sensitive to precedence. b. Reconstruction effects are not observed. c. Local disjointness effects are not observed. d. The type of the 'binder' makes a difference. 107 Based on these properties, Ueyama (1998: 155) proposes the following structural and lexical conditions on the two types of BVA. (5) (Ueyama 1998: 115, (65)) a. Structural condition on FD: *FD(α,β) if α does not c-command β at LF. b. Lexical condition on FD: *FD(α,β) if β is a large NP. (6) (Ueyama 1998: 115, (66)) a. Structural condition on ID: *ID(α,β) if α does not precede β at PF. b. Lexical condition on ID: *ID(α,β) if α is an A-type QP. The QP α is classified into two types, an A-type and a B-type, as shown in (7). The dependent term β is also classified into two types, large NP and small NP, as shown in (8). (7) (Ueyama 1998: 124, (12)) a. A-type QPs: NP-sae 'even NP' kanarinokazu-no NP 'most of the NPs' 10 izyoo-no NP 'ten or more NPs' 55%-no NP '55% of the NPs' 108 NP1 to NP2 (to) 'NP1 and NP2' NP1 ka NP2 (ka) 'either NP1 or NP2' b. B-type QPs: do-no NP 'which NP' do-no NP-mo 'every NP' (subete-no NP 'every NP') 1 (8) (Ueyama 1998: 124, (13)) a. large NPs: so-no zidoosya-gaisya 'that automobile company' so-no daigaku-insei 'that graduate student' b. small NPs so-ko 'it/that institution' so-re 'it/that thing' (so-itu 'he/that guy') 2 The other type of BVA is quirky binding, where a so-word as a dependent term appears to be involved in BVA without being c-commanded by an apparent QP at LF, nor being preceded by a QP at PF, as exemplified in (9). 1 Ueyama (1998) puts subete-no NP 'every NP' in parentheses in (7b) "because some degree of marginality accompanies to it when it patterns with do-no NP 'which NP'. The acceptability variation among speakers is also observed. (It seems that every NP in English qualifies as a B-type QP more easily than subete-no NP in Japanese.)" (Ueyama 1998: 124, fn.5) 2 Ueyama (1998) puts so-itu 'he/that guy' in parentheses in (8b) and mentions as follows. "As is implied by the gloss 'guy', so-itu is regarded to be a rough and impolite expression: usually such a word is used by men in referring to a non-respected person, and a lady is not supposed to use such words. Unlike 'that guy' in English, however, so-itu can refer to either a male or a female." (Ueyama 1998: 126, fn.8) 109 (9) (Ueyama 1998: (80)) a. ? So-ko-no bengosi-ga Toyota to Nissan-o suisensita (node, ato-wa dareka-ni that-place-GEN attorney-NOM Toyota and Nissan-ACC recommended because rest-TOP someone-DAT Mazda-o suisensite-moraw-eba ii dake da). Mazda-ACC recommend-ask-if good only COPULA '(Since) {its/a retained} attorney recommended Toyota and Nissan (, now we have only to ask someone to recommend Mazda).' b. ? So-ko-no bengosi-ga subete-no zidoosya-gaisya-o uttaeteiru (node, that-place-GEN attorney-NOM every-GEN automobile-company-ACC sued because zidoosya-gyookai-wa daikonran-ni otiitteiru). automobile-industry-TOP disorder-DAT be:thrown:into '(Since) {its/a retained} attorney has sued every automobile company (, the automobile industry has been thrown into a state of disorder).' Ueyama (1998:213) points out that in (9) "neither FD nor ID is involved precisely because the so-word in this case is not a dependent term." She then presents the conditions of the availability of quirky binding as follows. (10) (Ueyama 1998: 214, (81)) a. The apparent QP must "refer" to a specific group of individuals. b. The construction must not be the Deep OS-type. c. There should not be other quirky binding in the relevant clause. d. The apparent QP must be in a position which is salient enough to be a "topic" of a sentence. e. The so-word must be non-individual-denoting. 110 It should be noted that Hoji (2009:175) assumes "that the only structural relation visible at LF is that of c-command and that the informant's intuitions about 'meaning' would be a reflection of, and hence would be revealing about, properties of the Computational System only if they are based on properties of LF." Since the ID-based BVA is not sensitive to c-command, it should be regarded as not belonging to the Computational System. It seems clear that quirky binding does not belong to the Computational System, either, because it makes reference to factors outside the Computational System as the conditions in (10) show. Thus, if we want to reveal properties of the Computational System based on BVA, we must ensure that a given BVA is based on FD and not on ID or quirky binding. In order to ensure this, we must consider both the binder and bindee in a given BVA. 4.2.1.1 So-NPs vs. a-NPs as bindees According to Ueyama (1998), B of BVA(A,B) must be a singular-denoting anaphoric expression such that the value of the latter is not fixed to a certain individual. Along with the so-words in (8), there are a-words in Japanese, both roughly corresponding to that NP in English. 3 (11) (Ueyama 1998: 177, (9)) a. a-re / so-re 'that thing/it' b. a-itu / so-itu 'that guy/he' c. a-soko / so-ko 'that institution/it' d. a-no hito / so-no hito 'that person' e. a-no daigaku / so-no daigaku 'that university' 3 Japanese has another type of demonstrative with the prefix ko-, which is referred to as ko-words. Ko-words roughly correspond to this NP in English. 111 A-words in (11) are also singular-denoting expressions. One may take a-re, a-itu and a-soko as small NPs, while taking a-no NPs as large NPs, just as in the case of the corresponding so-words. As a property of FD-based BVA in (3d), the semantic content of the bindee makes a difference, that is, as shown in (5b), β οf FD(α,β) must be a small NP. Then, can we use a-re, a-itu and a-soko as B of FD-based BVA(A, B)? As mentioned above, B of BVA(A,B) must be a singular-denoting anaphoric expression such that the value of the latter is not fixed to a certain individual. Ueyama (1998:182) observes that a-words must be connected to individuals that the speaker knows through his direct experience, and thus they are always referential. Thus, they do not comply with the requirement of B of BVA(A,B) that the value is not fixed to a certain individual. Thus, a-re, a-itu and a-soko should not be used as B of FD-based BVA(A, B). Hoji (EPSA [10]-2) has been conducting experiments on this difference between so-words and a-words as the dependent term of BVA and obtained confirmed schematic asymmetries. 4 The details of the experiments are presented below. 4.2.1.1.1 Experimental design The hypothesis checked by Hoji (EPSA [10]-2) in regard to the difference between so-words and a-words as the dependent term of BVA is shown in (12). (12) So-NPs can be B of BVA(A, B) but a-NPs cannot. (12) is the consequence of the hypotheses about the lexical property of a-NPs and that of 4 In what follows, I refer to the experiments which have been conducted or are being conduced in line with the EPSA method as "EPSA [10]-2," for "identification purposes," along with the name of the researcher who has designed the particular experiment in question. 112 so-NPs proposed in Ueyama 1998 as shown in (13). (13) a. A-NPs are D-indexed. b. So-NPs are either I-indexed or non-indexed. A D-indexed NP is inherently referential and hence does not require a linguistic antecedent, while non-indexed NPs and I-indexed NPs require a linguistic antecedent. Since a D-indexed NP is strictly 'referential' and it has to be understood in connection with a specific individual known to the speaker, it cannot give rise to a BVA interpretation. Given in (14) and (15) are predicted schematic asymmetries obtained on the basis of (12). In Examples corresponding to each Schema group, he uses two QPs in (16) as A of BVA(A, B). (14) Schema group 1: (EPSA [10]-2) a. ok Schema A QP-ga so-NP-no N-o V BVA (QP, so-NP) b. *Schema B QP-ga a-NP-no N-o V BVA (QP, a-NP) c. ok Schema C QP-ga a-NP-no N-o V (With a-NP "referring to" a particular individual/object) (15) Schema group 2: (EPSA [10]-2) a. ok Schema A QP-ga [[ … so-NP … ] N]-ni V BVA (QP, so-NP) 113 b. *Schema B QP-ga [[ … a-NP … ] N]-ni V BVA (QP, a-NP) c. ok Schema C QP-ga [[ … a-NP … ] N]-ni V(With a-NP "referring to" a particular individual/object) (16) a. niwari izyoo-no N '20% or more N' b. kanarino kazu-no N 'a considerable number of N' The Examples constructed on the basis of the Schema in (14) are (17) and (18), and the ones constructed on the basis of the Schema in (15) are (19) and (20). (17) Schema group 1 (QP = niwari izyoo-no N) a. ok Example A-1 Niwari izyoo-no zititai-ga so-ko-no keesatu-o hihansita. 20% more-GEN local:government-NOM that-place-GEN police-ACC criticized 'Each of 20% or more local governments criticized its police.' b. *Example B-1 Niwari izyoo-no zititai-ga a-soko-no keesatu-o hihansita. 20% more-GEN local:government-NOM that-place-GEN police-ACC criticized 'Each of 20% or more local governments criticized its police.' c. ok Example C-1 Niwari izyoo-no zititai-ga a-soko-no keesatu-o hihansita. 20% more-GEN local:government-NOM that-place-GEN police-ACC criticized 114 '20% or more local governments criticized that police.' (18) Schema group 1 (QP = kanarino kazu-no N) a. ok Example A-2 Kanarino kazu-no seizika-ga so-itu-no hisyo-o hihansita. considerable number-GEN politician-NOM that-guy-GEN secretary-ACC criticized 'Each of a considerable number of politicians criticized that guy's secretary.' b. *Example B-2 Kanarino kazu-no seizika-ga a-itu-no hisyo-o hihansita. considerable number-GEN politician-NOM that-guy-GEN secretary-ACC criticized 'Each of a considerable number of politicians criticized that guy's secretary.' c. ok Example C-2 Kanarino kazu-no seizika-ga a-itu-no hisyo-o hihansita. considerable number-GEN politician-NOM that-guy-GEN secretary-ACC criticized 'A considerable number of politicians criticized that guy's secretary.' (19) Schema group 2 (QP = ni-wari izyoo-no N) a. ok Example A-3 Niwari izyoo-no zititai-ga so-ko-o hihansita zassikisya]-ni 20% more-GEN local:government-NOM that-place-ACC criticized magazine:reporter-DAT renraku-o totta. contact-ACC made 'Each of 20% or more local governments made contact with a/the magazine reporter(s) who had criticized it.' 115 b. *Example B-3 Niwari izyoo-no zititai-ga a-soko-o hihansita zassikisya]-ni 20% more-GEN local:government-NOM that-place-ACC criticized magazine:reporter-DAT renraku-o totta. contact-ACC made 'Each of 20% or more local governments made contact with a/the magazine reporter(s) who had criticized it.' c. ok Example C-3 Niwari izyoo-no zititai-ga a-soko-o hihansita zassikisya]-ni 20% more-GEN local:government-NOM that-place-ACC criticized magazine:reporter-DAT renraku-o totta. contact-ACC made '20% or more local governments made contact with a/the magazine reporter(s) who had criticized that place.' (20) Schema group 1 (QP = kanarino kazu-no N) a. ok Example A-4 Kanarino kazu-no seizika-ga so-itu-o hihansita zassikisya]-ni considerable number-GEN politician-NOM that-guy-ACC criticized magazine:reporter-DAT renraku-o totta. contact-ACC made 'Each of a considerable number of politicians made contact with a/the magazine reporter(s) who had criticized that guy.' b. *Example B-4 Kanarino kazu-no seizika-ga a-itu-o hihansita zassikisya]-ni 116 considerable number-GEN politician-NOM that-guy-ACC criticized magazine:reporter-DAT renraku-o totta. contact-ACC made 'Each of a considerable number of politicians made contact with a/the magazine reporter(s) who had criticized that guy.' c. ok Example C-4 Kanarino kazu-no seizika-ga a-itu-o hihansita zassikisya]-ni considerable number-GEN politician-NOM that-guy-ACC criticized magazine:reporter-DAT renraku-o totta. contact-ACC made 'A considerable number of politicians made contact with a/the magazine reporter(s) who had criticized that guy.' 4.2.1.1.2 Results of the experiments Table 4.1 is a summary of the results of the experiment on the predicted schematic asymmetries in (14) and (15). Table 4.1: A summary of the results of EPSA [10]-2 (Hoji 2010b: 48) EPSA [10]-2 (Total 37 participants; 670 answers) ... as of Sep/11/2010 Schema A Schema B Schema C Schema Group 1 71 values 74 72 values 7 72 values 92 Schema Group 2 72 values 74 70 values 5 71 values 90 Total 143 values 74 142 values 6 143 values 91 117 Of the 37 informants, 16 are "linguistically naïve" and 21 are "linguistically informed". 5 "670 answers" means that there have been 670 occurrences of a reported judgment. Some informants have judged the same example more than once; but in such cases the values are based on the average score on a given example by each such informant. As mentioned in Chapter 3, according to Hoji (2009, 2010a) we shall say that we have obtained a confirmed schematic asymmetry when the informant judgments on a number of *Examples conforming to a *Schema are as predicted and robust (i.e., the *Schema-based prediction has survived a rigorous disconfirmation attempt) and when furthermore the ok Examples of the ok Schemas corresponding to the *Schema are judged to be significantly more acceptable than the *Examples (i.e., the corresponding ok Schema-based predictions have been confirmed. In Table 4.1 the average score on *Examples conforming to a *Schema among the 37 informants is 6 on the 0-100 scale and the average scores of the ok Examples of the ok Schemas corresponding to the *Schema 74 and 91. In accordance with what is stated in the previous chapter, it seems reasonable to consider that the *Schema-based prediction has survived a disconfirmation attempt and the corresponding ok Schema-based predictions have been confirmed. Hence we have obtained a confirmed schematic asymmetry. 4.2.1.2 FD-based BVA 4.2.1.2.1 Experimental design (I): excluding quirky binding Next consider what QPs can be qualified as α of FD(α,β). We first have to ensure that the QPs cannot give rise to quirky binding. A configuration of quirky binding is shown in (21) 5 When informants register for EPSA to participate in the experiments, they are asked whether they understand (i) "bound variable anaphora" or "bound readings" and (ii) "A takes wide scope over B" as they are used in linguistic discussion. Hoji (2010a:40, fn.30) classifies the informants who state that they understand at least (i) or (ii) as "linguistically informed," while those who state that they understand neither of the two as "linguistically naïve." See Chapter 5.3.3.2 for the details of the classification of the informants. 118 (see the examples of quirky binding in (9) above). (21) So-NP-no N-ga QP-o V BVA (QP, so-NP) Recall one of the conditions on the availability of quirky binding in (10a), repeated here. (10) (Ueyama 1998: 214, (81)) a. The apparent QP must "refer" to a specific group of individuals. Thus, if we use QPs that do not "refer" to a specific group of individuals, quirky binding is expected to be impossible. For such QPs, then, BVA should always be unavailable in the configuration in (21). It is pointed out in Ueyama 1998, 2003 and Hoji 2003 that it is difficult or impossible for QPs such as those in (22) to be used to refer to a specific group of individuals, in contrast to QPs such as those in (23). (22) kanarinokazu-no NP 'most of the NPs' 55% izyoo-no NP '55% or more NPs' NP-sae 'even NP' (23) subete-no NP 'every NP' 3-ttu-no NP 'three NPs' Hoji (EPSA [1]-1,2,3 and 18) checks these QPs if a so-word can be involved in BVA without 119 being c-commanded by these QPs at LF. EPSA [1]-1, 2 and 3 test the QPs listed in (22), while EPSA [1]-18 tests the QPs listed in (23). The predicted schematic asymmetries he considers are (24), (25) and (26). (24) Schema group 1: (EPSA [1]-1,2,3 and 18) a. ok Schema A QP-ga so-NP-no N-o/ni V BVA (QP, so-NP) b. *Schema B so-NP-no N-ga QP-o/ni V BVA (QP, so-NP) c. ok Schema C so-NP-no N-ga QP-o/ni V (With so-NP "referring to" a particular individual/object) (25) Schema group 2: (EPSA [1]-1,2 and 3) a. ok Schema A QP-ga [[ … so-NP … ] N]-o/ni V BVA (QP, so-NP) b. *Schema B [[ … so-NP … ] N]-ga QP-o/ni V BVA (QP, so-NP) c. ok Schema C [[ … so-NP … ] N]-ga QP-o/ni V (With so-NP "referring to" a particular individual/object) (26) Schema group 3: (EPSA [1]-1,2 and 3) a. ok Schema A QP-ga [N-ga so-NP-no N-o/ni V]-to V BVA (QP, so-NP) 120 b. *Schema B So-NP-no N-ga [N-ga QP-o/ni V]-to V BVA (QP, so-NP) c. ok Schema C So-NP-no N-ga [N-ga QP-o/ni V]-to V (With so-NP "referring to" a particular individual/object) EPSA [1]-1, 2 and 3 test all the three Schema groups in (24), (25) and (26), while EPSA [1]-18 tests the Schema group in (24). As for EPSA [1]-1, 2 and 3, I only present one set of the Examples of each Schema group with a QP kanarinokazu-no NP 'most of the NPs' and with the object being o-marked, which is in EPSA [1]-1. The only difference among EPSA [1]-1, 2 and 3 is the selection of QP. All the other parts, including the instructions, are the same. Furthermore, Hoji (EPSA [1]-1,2 and 3) has checked another set of the Examples of each Schema group, in which the object is ni-marked. Thus, a total of six example sets (altogether 18 sentences) are tested in each EPSA experiment. EPSA [1]-18, on the other hand, has one Example set with the object being o-marked for each QP. (27) Schema group 1 (QP = kanarino kazu-no N) a. ok Example A-1 Kanarino kazu-no tihoozititai-ga so-ko-no syokuin-o sinbunsizyoo-de hihansita. considerable number-GEN local:government-NOM that-place-GEN agent-ACC newspaper-in criticized 'Each of a considerable number of local governments criticized its agents in the newspaper.' b. *Example B-1 So-ko-no syokuin-ga kanarino kazu-no tihoozititai-o sinbunsizyoo-de hihansita. that-place-GEN agent-NOM considerable number-GEN local:government-ACC newspaper-in criticized 121 'Its agents criticized a considerable number of local governments in the newspaper.' c. ok Example C-1 So-ko-no syokuin-ga kanarino kazu-no tihoozititai-o sinbunsizyoo-de hihansita. that-place-GEN agent-NOM considerable number-GEN local:government-ACC newspaper-in criticized 'That place’s agents criticized a considerable number of local governments in the newspaper.' (28) Schema group 2 (QP = kanarino kazu-no N) a. ok Example A-1 Kanarino kazu-no tihoozititai-ga naganen so-ko-de hataraiteita syokuin-o considerable number-GEN local:government-NOM for:many:years that-place-for was:working agent-ACC sinbunsizyoo-de hihansita. newspaper-in criticized 'Each of a considerable number of local governments criticized the agents who had been working for it for many years in the newspaper.' b. *Example B-1 Naganen so-ko-de hataraiteita syokuin-ga kanarino kazu-no tihoozititai-o for:many:years that-place-for was:working agent-NOM considerable number-GEN local:government-ACC sinbunsizyoo-de hihansita. newspaper-in criticized 'The agents who had been working for it for many years criticized a considerable number of local governments in the newspaper.' c. ok Example C-1 Naganen so-ko-de hataraiteita syokuin-ga kanarino kazu-no tihoozititai-o for:many:years that-place-for was:working agent-NOM considerable number-GEN local:government-ACC 122 sinbunsizyoo-de hihansita. newspaper-in criticized 'The agents who had been working for that place for many years criticized a considerable number of local governments in the newspaper.' (29) Schema group 3 (QP = kanarino kazu-no N) a. ok Example A-1 Kanarino kazu-no tihoozititai-ga zaimusyoo-ga so-ko-no syokuin kumiai-o considerable number-GEN local:government-NOM finance:ministry-NOM that-place-GEN staff union-ACC hihansita-to happyoosita. criticized-that reported 'A considerable number of local governments reported that the Ministry of Finance criticized its staff union.' b. *Example B-1 So-ko-no syokuin kumiai-ga zaimusyoo-ga kanarino kazu-no tihoozititai-o that-place-GEN staff union-NOM finance:ministry-NOM considerable number-GEN local:government-ACC hihansita-to happyoosita. criticized-that reported 'Its staff unions reported that the Ministry of Finance criticized a considerable number of local governments.' c. ok Example C-1 So-ko-no syokuin kumiai-ga zaimusyoo-ga kanarino kazu-no tihoozititai-o that-place-GEN staff union-NOM finance:ministry-NOM considerable number-GEN local:government-ACC hihansita-to happyoosita. criticized-that reported 123 'That place’s staff unions reported that the Ministry of Finance criticized a considerable number of local governments.' 4.2.1.2.2 Results of the experiments (I) Table 4.2 to 4.5 are summaries of the results of the experiment on the predicted schematic asymmetries in (24), (25) and (26). Table 4.2: A summary of the results of EPSA [1]-1 (QP = kanarino kazu-no N) EPSA [1]-1 (Total 24 participants; 966 answers) ... as of Jan/10/2011 Schema A Schema B Schema C Schema Group 1 47 values 97 47 values 10 48 values 93 Schema Group 2 48 values 95 47 values 26 48 values 93 Schema Group 3 48 values 91 48 values 5 48 values 91 Total 143 values 94 142 values 14 144 values 92 Table 4.3: A summary of the results of EPSA [1]-2 (QP = 55% izyoo-no NP) EPSA [1]-2 (Total 26 participants; 1153 answers) ... as of Jan/10/2011 Schema A Schema B Schema C Schema Group 1 52 values 98 51values 4 51 values 96 Schema Group 2 52 values 98 51 values 10 52 values 95 Schema Group 3 52 values 96 51 values 9 52 values 98 Total 156 values 97 153 values 8 155 values 96 124 Table 4.4: A summary of the results of EPSA [1]-3 (QP = NP-sae) EPSA [1]-3 (Total 17 participants; 665 answers) ... as of Jan/10/2011 Schema A Schema B Schema C Schema Group 1 34 values 97 34 values 6 34 values 96 Schema Group 2 34 values 97 34 values 6 34 values 91 Schema Group 3 34 values 95 34 values 3 34 values 98 Total 102 values 96 102 values 5 102 values 95 Table 4.5: A summary of the results of EPSA [1]-18 (QP = subeteno NP and 3-ttu-no NP) EPSA [1]-18 (Total 10 participants; 117 answers) ... as of Mar/30/2011 Schema A Schema B Schema C subeteno NP 'every NP' 10 values 100 10 values 51 10 values 100 3-ttu-no NP 'three NPs' 10 values 100 10 values 55 10 values 100 Total 20 values 100 20 values 53 20 values 100 The average scores on *Examples conforming to *Schema of the QPs listed in (24) (i.e., EPSA[1]-18) are high over 50. This result shows that these QPs can give rise to quirky binding for some informants, that is to say, these QPs can be taken to refer to a specific group of individuals. We thus should avoid using these QPs in the experiments that are crucially based on FD(α,β). On the other hand, the average scores on *Examples conforming to *Schema of the QPs listed in (22) are low falling under 14, while the average score of the corresponding ok Examples are close to 100, which are significantly higher than the *Examples. Remember, however, that ideally *Examples conforming to a *Schema sould be completely unacceptable under the specified interpretation. In this light, the score on *Examples with kanarino kazu-no N seems 125 to be a little too high. On the other hand, the score on *Examples with NP-sae is very low. Thus, the QP NP-sae is most unlikely to give rise to quirky binding. This can be understood that it is most difficult for informants to take NP-sae to refer to a specific group of individuals, which is one of the conditions forquirky binding. Although the scores on the *Examples with the use of the QPs kanarino kazu-no N and 55% izyoo-no NP are higher than NP-sae, I will use the three QPs in my experiments keeping the above considerations in mind. 4.2.1.2.3 Experimental design (II): excluding ID-based BVA We have just seen that the QP NP-sae is almost impossible and the QPs kanarinokazu-no N and 55% izyoo-no NP are difficult to give rise to quirky binding. Let us turn to the other two sources of BVA: FD-based BVA and ID-based BVA. Recall here the structural conditions on FD-based BVA and ID-based BVA in (5a) and (6a), respectively. (5) a. Structural condition on FD: *FD(α,β) if α does not c-command β at LF. (6) a. Structural condition on ID: *ID(α,β) if α does not precede β at PF. From these two conditions, we can infer that when BVA is possible in a configuration where a QP (which cannot give rise to quirky binding) does not precede the dependent term at PF but it c-commands the dependent term at LF, the BVA so obtained must be FD-based. LF reconstruction provides us such a configuration. When a dislocated object (DL) of the OS-type construction contains a dependent term and 126 a QP is the subject as in (30), ID-based BVA should be unavailable since the QP does not precede the dependent term at PF. (30) a. So-NP-no N-o/ni QP-ga V (cf. (24a)) b. [[ … so-NP … ]N]-o/ni QP-ga V (cf. (25a)) Thus, if BVA is available in such a configuration, the DL must be 'reconstructed' to the position where it is c-commanded by the QP at LF. 6 This means that such an OS-type construction is the Surface OS-type. The Schemata in (30) are the ok Schemata in the next experiment (EPSA [1]-4, 5 and 6), while *Schemata are the same as the ones in the last experiment (EPSA [1]-1, 2 and 3). Thus, the predicted schematic asymmetries come as in (31) and (32). 7 (31) Schema group 1: (EPSA [1]-4, 5 and 6) a. ok Schema A so-NP-no N-o/ni QP-ga V BVA (QP, so-NP) b. *Schema B so-NP-no N-ga QP-o/ni V BVA (QP, so-NP) c. ok Schema C so-NP-no N-ga QP-o/ni V (With so-NP "referring to" a particular individual/object) 6 Ueyama (1998) does not assume LF reconstruction. Instead, she assumes PF movement. Thus, the DL of the Surface OS-type remains in the base-generated position throughout the derivation to LF, but is moved to the surface position at PF. 7 Schema group 3 of the last experiment (EPSA [1]-1,2,3) (see (26)) is not included in this experiment. 127 (32) Schema group 2: (EPSA [1]-4, 5 and 6) a. ok Schema A [[ … so-NP … ] N]-o/ni QP-ga V BVA (QP, so-NP) b. *Schema B [[ … so-NP … ] N]-ga QP-o/ni V BVA (QP, so-NP) c. ok Schema C [[ … so-NP … ] N]-ga QP-o/ni V (With so-NP "referring to" a particular individual/object) The three QPs in the last experiment, kanarino kazu-no N, 55% izyoo-no NP and NP-sae, are used in these experiments, too. Again, I only present one set of the Examples of each Schema group with the QP kanarinokazu-no NP 'most of the NPs' and the object is o-marked, which is in EPSA [1]-4. The only difference among EPSA [1]-4, 5 and 6 is the selection of QP. Furthermore, Hoji (EPSA [1]-4, 5 and 6) has checked another set of the Examples of each Schema group, in which the object is ni-marked. Thus, a total of four example sets (altogether 12 sentences) are tested in each EPSA. (33) Schema group 1 (QP = kanarino kazu-no N) a. ok Example A-1 So-ko-no syokuin-o kanarino kazu-no tihoozititai-ga sinbunsizyoo-de ihansita. that-place-GEN agent-ACC considerable number-GEN local:government-NOM newspaper-in criticized 'Each of a considerable number of local governments criticized its agents in the newspaper.' 128 b. *Example B-1 So-ko-no syokuin-ga kanarino kazu-no tihoozititai-o sinbunsizyoo-de hihansita. that-place-GEN agent-NOM considerable number-GEN local:government-ACC newspaper-in criticized 'Its agents criticized a considerable number of local governments in the newspaper.' c. ok Example C-1 So-ko-no syokuin-ga kanarino kazu-no tihoozititai-o sinbunsizyoo-de hihansita. that-place-GEN agent-NOM considerable number-GEN local:government-ACC newspaper-in criticized 'That place’s agents criticized a considerable number of local governments in the newspaper.' (34) Schema group 2 (QP = kanarino kazu-no N) a. ok Example A-1 Naganen so-ko-de hataraiteita syokuin-o kanarino kazu-no tihoozititai-ga for:many:years that-place-for was:working agent-ACC considerable number-GEN local:government-NOM sinbunsizyoo-de hihansita. newspaper-in criticized 'Each of a considerable number of local governments criticized the agents who had been working for it for many years in the newspaper.' b. *Example B-1 Naganen so-ko-de hataraiteita syokuin-ga kanarino kazu-no tihoozititai-o for:many:years that-place-for was:working agent-NOM considerable number-GEN local:government-ACC sinbunsizyoo-de hihansita. newspaper-in criticized 'The agents who had been working for it for many years criticized a considerable number of local governments in the newspaper.' 129 c. ok Example C-1 Naganen so-ko-de hataraiteita syokuin-ga kanarino kazu-no tihoozititai-o for:many:years that-place-for was:working agent-NOM considerable number-GEN local:government-ACC sinbunsizyoo-de hihansita. newspaper-in criticized 'The agents who had been working for that place for many years criticized a considerable number of local governments in the newspaper.' 4.2.1.2.4 Results of the experiments (II) Table 4.6 to 4.8 are summaries of the results of the experiments on the predicted schematic asymmetries in (31) and (32). Table 4.6: A summary of the results of EPSA [1]-4 (QP = kanarinokazu-no N) EPSA [1]-4 (Total 10 participants; 179 answers) ... as of Jan/10/2011 Schema A Schema B Schema C Schema Group 1 20 values 71 19 values 22 20 values 100 Schema Group 2 20 values 90 20 values 29 20 values 95 Total 40 values 81 39 values 25 40 values 97 Table 4.7: A summary of the results of EPSA [1]-5 (QP = 55% izyoo-no NP) EPSA [1]-4 (Total 10 participants; 179 answers) ... as of Jan/10/2011 Schema A Schema B Schema C Schema Group 1 20 values 71 19 values 22 20 values 100 Schema Group 2 20 values 90 20 values 29 20 values 95 Total 40 values 81 39 values 25 40 values 97 130 Table 4.8: A summary of the results of EPSA [1]-6 (QP = NP-sae) EPSA [1]-6 (Total 12 participants; 203 answers) ... as of Jan/10/2011 Schema A Schema B Schema C Schema Group 1 24 values 75 34 values 2 24 values 99 Schema Group 2 24 values 82 34 values 10 24 values 100 Total 48 values 79 102 values 6 102 values 99 The average scores on the ok Examples conforming to ok Schema A are about 80 for all the three QPs. This shows that it is possible for many of the informants to obtain BVA interpretation without QPs preceding the dependent term at PF. This BVA, therefore, should not be an ID-based one since the ID-based BVA requires the PF precedence of QPs. Moreover, the low scores on the *Examples conforming to the *Schema means that quirky binding is almost impossible with these QPs (though the score with kanarino kazu-no N is a little high). Therefore, though the number of the informants is not very large in EPSA[1]-4 and 6 yet, it seems safe to conclude that BVA in the ok Examples conforming to ok Schema A is an FD-based one. As mentioned above, if we want to reveal properties of the Computational System based on BVA, we must ensure that a given BVA is based on FD. We would be able to do it by using QPs NP-sae and 55% izyoo-no NP (and possibly kanarino kazu-no N) as binder and a small so-NP (such as so-ko, so-itu and so-re) as bindee. 4.2.2 DR (distributive reading) In the above discussion of BVA, we have seen that by controlling the type of QPs and dependent terms, we can ensure that the BVA is FD-based. FD-based BVA is crucially based on the structural relation of c-command. It is particularly important to have a way to sort out 131 the FD-based BVA. Recall the assumption we adopt in this thesis "that the only structural relation visible at LF is that of c-command and that the informant's intuitions about 'meaning' would be a reflection of, and hence would be revealing about, properties of the Computational System only if they are based on properties of LF" (Hoji (2009:175). Therefore, when we are trying to reveal properties of the Computational System in other constructions, we should concentrate on the c-command relation among NPs in those constructions, and the FD-based BVA provides us with a tool to do it. Scope interpretation between two scope bearing elements, more particularly a wide scope distributive reading of one over the other, is another phenomenon which has been argued in the literature to be based on the structural relation of c-command. If such is indeed the case, we have another tool to investigate properties of the Computational System. Kuroda 1969/1970 observes that in Japanese a sentence such as in (35) is unambiguous, while the scrambling construction such as in (36) allows two readings in regard to the scope interpretation between the two QPs. Hoji 1985 further demonstrates that the same holds with other types of quantificational expressions. 8 (35) (Ueyama 1998: 40, (43)) SO-type construction: Dareka-ga daremo-o aisiteiru someone-NOM everyone-ACC love 'Someone loves everyone.' ∃x[∀y[ x loves y ]] *∀y[∃x[ x loves y ]] 8 The discussion in regard to the scope interpretation in this subsection is mainly from Ueyama 1998: Chapter 2.3 and Hoji 2003a: Chapter 2.3. 132 (36) (Ueyama 1998: 40, (44)) OS-type construction: a. Dareka-o i daremo-ga t i aisiteiru someone-ACC everyone-NOM love 'Someone i , everyone loves t i .' ∃y[∀x[ x loves y ]] ∀x[∃y[ x loves y ]] b. Daremo-o i dareka-ga t i aisiteiru everyone-ACC someone-NOM love 'Everyone i , someone loves t i .' ∀y[∃x[ x loves y ]] ∃x[∀y[ x loves y ]] Thus, we obtain the following generalization. 9 (37) (Ueyama 1998: 41, (46)) (i) In the SO-type construction, 'QP1-NOM QP2-ACC/DAT V' QP1>QP2, but *QP2>QP1. (ii) In the OS-type construction, 'QP2-ACC/DAT QP1-NOM V' QP1>QP2, and QP2>QP1. Ueyama (1998) assumes the scope interpretation hypothesis in (38) in order to account for the 9 Kuroda 1969/1970 states the generalization as follows. (Kuroda 1970:138,1992: chap. 2, 97) (i) If a predicate corresponds to a sentence frame with the 'preferred' word order, the semantic order of quantifiers is given by their linear order; if a predicate corresponds to a sentence frame with 'inverted' word order, the semantic order of quantifiers is ambiguous. 133 generalization in (37). 10 (38) (Ueyama 1998: 46, (57)) Scope Interpretation Hypothesis: a. The scope of a QP is its c-commanding domain after QR. b. If QP1 c-commands QP2 before QR applies, QP1 must also c-command QP2 after QR. 11 According to (38), c-command relation between two QPs is retained after QR, which means that the positions before QR fix the scope relation between two QPs. Hoji (2003) thus formulates the scope interpretation hypothesis as in (39) confining itself to the wide scope distributive 10 Huang 1982 proposes the condition in (i) based on Chinese data. Hoji 1985:248 reformulates the condition in (i) as in (ii). (i) (Ueyama 1998: 44, (50)) Scope Interpretation Hypothesis in Huang 1982: a. The scope of a QP is its c-commanding domain after QR. b. Suppose A and B are both QPs or both Q-NPs or Q-expressions, then if A c-commands B at SS, A also c-commands B at LF. ((ib) = Huang 1982:220 (70) "The General Condition on Scope Interpretation") (ii) (Ueyama 1998: 45, (51)) Scope Interpretation Hypothesis in Hoji 1985: a. The scope of a QP is its c-commanding domain after QR. b. at LF *QP i QP j t j t i where each member c-commands the member to its right ((iib) = Hoji 1985:248 (76)) Ueyama (1998:45) points out that Hoji (1985) reformulates Huang’s hypothesis in (i) as in (ii) because (i) would predict that not only the SO-type but the OS-type construction is unambiguous. Hoji’s hypothesis in (ii), on the other hand, can accounts for the generalization in (37) successfully with the additional assumption regarding the trace deletion. With the introduction of the Deep OS-type and Surface OS-type of the OS-type construction in Ueyama 1998, however, the additional assumption regarding the trace deletion becomes unnecessary, and the hypothesis in (38) suffices to account for the generalization in (37). For more detailed discussion on this regard, see Ueyama 1998: Chapter 2.3.2. 11 Ueyama (1998: 47, fn.25) points out that "[(38b)] should be understood as a mere generalization. It has yet to be considered how the relevant constraint should be stated to achieve the effect in question under the framework outlined in section 1.3 [in Ueyama 1998]." 134 reading. DR(A, B) in (39) stands for a wide scope distributive reading of A over B. 12 (39) (Hoji 2003a: 27, (66)) DR(A, B) is available only if the A-position trace of A c-commands that of B at LF. Although the generalization in (37) has been widely accepted in the literature, it is reported that some speakers detect scope ambiguity in the SO-type construction. For example, Ueyama (1998) provides an SO-type construction such as (40), saying that it allows the second reading relatively easily. 13 (40) (Ueyama 1998: 41, (47)) SO-type construction: [Dareka-ga [uti-no subete-no sensyu-o] bikoositeiru] (toyuu koto-wa, zen'in-ga someone-NOM our-GEN all-GEN athlete-ACC shadow COMP fact-TOP everyone-NOM kiken-ni sarasareteiru toyuu koto da.) danger-DAT exposed COMP fact COPULA '(The fact that) someone is shadowing every athlete of ours (means that everyone's life is in danger.)' ∃x[∀y(y = athlete)[ x is shadowing y ]] ∀y(y = athlete)[∃x[ x is shadowing y ]] In (40) DR(subete-no sensyu-o, dareka-ga) is available without the relevant c-command or 12 Hoji (2003a: 27, fn.42) points out that "[a]s in the case of BVA(A, B), DR(A, B) is a descriptive term (representing the speaker's intuition of a particular kind) and is not part of the theory." DR(A, B) expresses an intuition that the value of B differs according to the value of A . (E.g., BVA(everyone, someone) in Someone loves everyone) 13 See also Kitagawa 1990, Kuroda 1994, Kuno et al. 1999, Hayashishita 1999, 2000b, 2004 and Hoji 2003a. 135 precedence relation, which is reminiscent of the quirky binding in BVA. Recall that we can eliminate the possibility of quirky binding by controlling the type of QPs as binders, more specifically by using QPs which can not refer to a specific group of individuals. The same seems to apply to scope ambiguity in the SO-type construction. Ueyama (1998:42) points out that "[g]enerally speaking, the ambiguity in question is easier to obtain when the second QP is interpreted to refer to a specific group of individuals. For example, the availability of ambiguity in an SO-type construction decreases if we use QPs such as 55%-no NP '55% of the NPs', 10 izyoo-no NP 'ten or more NPs', or kanarinokazu-no NP 'most of the NPs', which resist a specific group reading more strongly than QPs such as daremo 'everyone'." For example, Ueyama (1998) reports that unlike in (40), DR(10 izyoo-no kaisya-ni, 30%-no ginkoo-ga) is hard to obtain in (41). (41) (Ueyama 1998: 42, (48a)) SO-type construction: [30%-no ginkoo-ga] [10 izyoo-no kaisya-ni] huseina kasituke-o syooninsiteiru 30%-GEN bank-NOM 10 or:more-GEN company-DAT illegal loan-ACC approve '[30% of the banks] have approved illegal loans [to ten or more companies].' (i) QP-NOM > QP-DAT 30%x(x = bank)[TEN-OR-MOREy(y = company)[ x has approved illegal loans to y ]] '30% of the banks are such that it has approved illegal loans to ten or more companies.' (ii) ??/?*QP-DAT > QP-NOM TEN-OR-MOREy(y = company)[30%x(x = bank)[ x has approved illegal loans to y ]] 'There are ten or more companies such that 30% of the banks have approved illegal loans to it.' 136 If we can in fact sort out DR based on c-command from the 'quirky' case as in (40) by controlling the type of QPs, we can use DR as a tool to test the c-command relation in other constructions, such as the cleft construction. We will therefore discuss this possibility in the following subsection. 4.2.2.1 Avoiding a 'quirky' case 4.2.2.1.1 Experimental design Mukai (EPSA [13]-1 and 4) conducts experiments to test if the type of QPs affects the scope ambiguity in the SO-type construction, as suggested in Hayashishita 1997, Ueyama 1998, Hoji 2003a. The predicted schematic asymmetry she obtained based on (39) is (42). Mukai (EPSA [13]-1 and 4) tests nine QPs in (43) as A of DR(A, B). 14 (42) Schema group 1: (EPSA [13]-1 and 4) a. ok Schema A A-ga [#-CL-no N]-o V-T DR (A, #-CL-no N) b. *Schema B [#-CL-no N]-ga A-o V-T DR (A, #-CL-no N) c. ok Schema C [#-CL-no N]-ga A-o V-T DR (#-CL-no N, A) (43) a. sukunakutomo 4-ttu-no NP 'at least four NPs' NP-sae 'even NP' 14 EPSA [13]-1 and 4 both test the same schema group in (42). EPSA [13]-1 tests the QPs in (43), while EPSA [13]-4 tests the QPs in (43). Mukai assumes that the QPs in (43) are interpreted to refer to a specific group of individuals, and hence exhibit scope ambiguity. 137 55% izyoo-no NP '55% or more NPs' 2-wari izyoo-no NP '20% or more NPs' b. subete-no NP 'every NP' zen-in 'everyone' kanarinokazu-no NP 'most of the NPs' 4-ttu-no NP 'four NPs' dare-mo 'everybody' The Examples constructed on the basis of the Schema in (42) with sukunakutomo 4-ttu-no NP 'at least four NPs' as A of DR(A,B) are (44). All the other Examples are constructed just by replacing "sukunakutomo 4-ttu-no NP" with other QPs in (43). (44) Schema group 1 (A of DR(A, B) = sukunakutomo 4-ttu-no NP) a. ok Example A Sukunakutomo 4-ttu-no tihoozititai-ga san-nin-no seezika-o hihansita. at:least 4-CL-GEN local:government-NOM 3-CL-GEN politician-ACC criticized 'At least four local governments criticized three politicians.' b. *Example B San-nin-no seezika-ga sukunakutomo 4-ttu-no tihoozititai-o hihansita. 3-CL-GEN politician-NOM at:least 4-CL-GEN local:government-ACC criticized 'Three politicians criticized at least four local governments.' c. ok Example C San-nin-no seezika-ga sukunakutomo 4-ttu-no tihoozititai-o hihansita. 3-CL-GEN politician-NOM at:least 4-CL-GEN local:government-ACC criticized 138 'Three politicians criticized at least four local governments.' 4.2.2.1.2 Results of the experiments Table 4.9 and 4.10 are summaries of the results of the experiment on the predicted schematic asymmetry in (42). Table 4.9: A summary of the results of EPSA [13]-1 EPSA [13]-1 (Total 11 participants; 166 answers) ... as of Jan/20/2011 QP as A Schema A Schema B Schema C sukunakutomo 4-ttu-no NP 'at least four NPs' 11 values 97 11 values 23 11 values 100 NP-sae 'even NP' 11 values 97 11 values 28 11 values 100 55% izyoo-no NP '55% or more NPs' 11 values 97 11 values 28 11 values 100 2-wari izyoo-no NP '20% or more NPs' 11 values 96 11 values 25 10 values 97 Total 44 values 97 44 values 26 43 values 99 Table 4.10: A summary of the results of EPSA [13]-4 EPSA [13]-4 (Total 6 participants; 117 answers) ... as of Jan/20/2011 QP as A Schema A Schema B Schema C subete-no NP 'every NP' 6 values 100 6 values 56 6 values 100 zen-in 'everyone' 6 values 100 6 values 39 6 values 100 kanarinokazu-no NP 'most of the NPs' 6 values 97 6 values 22 6 values 100 4-ttu-no NP 'four NPs' 5 values 100 6 values 52 6 values 100 dare-mo 'everybody' 6 values 100 5 values 60 6 values 95 Total 29 values 99 29 values 45 30 values 99 139 The results show that all the QPs in EPSA [13]-1 and kanarinokazu-no NP 'most of the NPs' in EPSA [13]-4 are somewhat hard to be A of DR(A,B) without them c-commanding or preceding B, while it is much easier to get such scope interpretation with the rest of the QPs in [13]-4. It should be noted that the average scores on the *Examples conforming to the *Schema of the former type of QPs are not particularly low. However, Mukai (pc in January, 2011) points out that two of the 11 informants in EPSA [13]-1 and one of the six informant in EPSA [13]-4 (, who is the same person as the one of the two informants in EPSA [13]-1 just alluded to) report all the *Examples as totally acceptable (in addition to the ok Examples). Their judgments are quite unusual compared to the ones of the majority of informants though the total number of informants is still small. If we tentatively set their judgments aside as anomaly, we get the results in Table 4.11 and 4.12. Table 4.11: A summary of the results of EPSA [13]-1 (adjusted) EPSA [13]-1 (Total 9 participants) ... as of Jan/20/2011 QP as A Schema A Schema B Schema C sukunakutomo 4-ttu-no NP 9 values 97 9 values 6 9 values 100 NP-sae 9 values 97 9 values 12 9 values 100 55% izyoo-no NP 9 values 97 9 values 12 9 values 100 2-wari izyoo-no NP 9 values 95 9 values 8 9 values 96 Total 36 values 96 36 values 10 35 values 99 140 Table 4.12: A summary of the results of EPSA [13]-4 (adjusted) EPSA [13]-4 (Total 5 participants) ... as of Jan/20/2011 QP as A Schema A Schema B Schema C subete-no NP 5 values 100 5 values 47 5 values 100 zen-in 5 values 100 5 values 27 5 values 100 kanarinokazu-no NP 5 values 97 5 values 7 5 values 100 4-ttu-no NP 4 values 100 5 values 42 5 values 100 dare-mo 5 values 100 4 values 50 5 values 95 Total 24 values 99 24 values 34 25 values 99 With this exclusion, the average scores on the *Examples conforming to the *Schema of the QPs in EPSA [13]-1 and kanarinokazu-no NP 'most of the NPs' in EPSA [13]-4 are quite low. Of course, we need to collect more judgments to see if they are really anomalies or not, but we can at least conclude that the QPs in EPSA [13]-1 and kanarinokazu-no NP 'most of the NPs' in EPSA [13]-4, which seem to resist a specific-group reading, are hard to be A of DR(A,B) without them c-commanding or preceding B. Thus, we can use such QPs as A of DR(A,B) and test the c-command relation in other constructions. 4.2.2.2 LF c-command 4.2.2.2.1 Experimental design Mukai (EPSA [13]-2 and 5) also conducts experiments to test if LF c-command and not PF precedence is in fact required for DR(A,B). The predicted schematic asymmetry she obtained based on (39) is (45). 141 (45) Schema group 1: (EPSA [13]-2 and 5) a. ok Schema A [#-CL-no N]-o A-ga V-T DR (A, #-CL-no N) b. *Schema B [#-CL-no N]-ga A-o V-T DR (A, #-CL-no N) c. ok Schema C [#-CL-no N]-ga A-o V-T DR (#-CL-no N, A) The only difference from EPSA [13]-1 and EPSA [13]-4 is ok Schema A, which is the OS-type construction instead of the SO-type construction in EPSA [13]-1 and 4. *Schema B and ok Schema C are the same as EPSA [13]-1 and 4. Mukai (EPSA [13]-2 and 5) uses the same nine QPs in EPSA [13]-1 and 4 in (43) as A of DR(A, B). The Examples constructed on the basis of the Schema in (45) with sukunakutomo 4-ttu-no NP 'at least four NPs' as A of DR(A,B) are in (46). All the other Examples are constructed by replacing "sukunakutomo 4-ttu-no NP" with other QPs in (43). (46) Schema group 1 (A of DR(A, B) = sukunakutomo 4-ttu-no NP) a. ok Example A San-nin-no seezika-o sukunakutomo 4-ttu-no tihoozititai-ga hihansita. 3-CL-GEN politician-ACC at:least 4-CL-GEN local:government-NOM criticized 'At least four local governments criticized three politicians.' b. *Example B San-nin-no seezika-ga sukunakutomo 4-ttu-no tihoozititai-o hihansita. 3-CL-GEN politician-NOM at:least 4-CL-GEN local:government-ACC criticized 142 'Three politicians criticized at least four local governments.' c. ok Example C San-nin-no seezika-ga sukunakutomo 4-ttu-no tihoozititai-o hihansita. 3-CL-GEN politician-NOM at:least 4-CL-GEN local:government-ACC criticized 'Three politicians criticized at least four local governments.' 4.2.2.2.2 Results of the experiments Table 4.13 and 4.14 are summaries of the results of the experiment on the predicted schematic asymmetries in (45). Table 4.13: A summary of the results of EPSA [13]-2 EPSA [13]-2 (Total 7 participants; 123 answers) ... as of Jan/21/2011 QP as A Schema A Schema B Schema C sukunakutomo 4-ttu-no NP 7 values 94 7 values 16 7 values 100 NP-sae 6 values 100 7 values 10 7 values 100 55% izyoo-no NP 7 values 100 7 values 28 7 values 100 2-wari izyoo-no NP 7 values 96 7 values 17 7 values 100 Total 27 values 97 28 values 18 28 values 100 143 Table 4.14: A summary of the results of EPSA [13]-5 EPSA [13]-5 (Total 4 participants; 73 answers) ... as of Jan/21/2011 QP as A Schema A Schema B Schema C subete-no NP 6 values 100 6 values 75 6 values 100 zen-in 6 values 100 6 values 78 6 values 100 kanarinokazu-no NP 6 values 93 6 values 34 6 values 100 4-ttu-no NP 5 values 100 6 values 75 6 values 100 dare-mo 6 values 100 5 values 66 6 values 95 Total 29 values 98 29 values 65 30 values 100 What is crucial in these experiments is the acceptability of ok Examples A, especially with the QPs in EPSA [13]-2 and kanarinokazu-no NP 'most of the NPs' in EPSA [13]-5. We have seen in the previous experiments that these QPs are hard to be A of DR(A,B) without them c-commanding or preceding B. Thus, the acceptability of ok Examples A with these QPs does not come from the 'quirky' type of scope interpretation available in (40). The acceptability of ok Examples A with these QPs also shows that the precedence is not a relevant factor for DR. Thus, the results of these experiments further support the hypothesis in (39), repeated here. (39) (Hoji 2003a: 27, (66)) DR(A, B) is available only if the A-position trace of A c-commands that of B at LF. By using these QPs as A of DR(A,B), we can make sure that the DR so obtained must be based on c-command relation between A and B, more specifically, (the trace of) A c-commands (that of) B at LF. Thus, we can use such DR(A,B) to test the c-command relation in other 144 constructions. 4.3 Hypotheses on Negative Sensitive Elements (NSE) 4.3.1 NSE in Japanese In one of the experiments on CM-Cleft I will discuss below, I will crucially utilize the structural conditions on 'Neg(ation)-sensitive' elements (NSE) proposed by Kataoka (2007). I will therefore summarize Kataoka’s (2007) proposals in this subsection. NSEs are elements that must occur with sentential negation (henceforth, Neg), which are in general called Negative Polarity Items (NPI) in the literature. Following Kataoka (2007), I will use the term NSE instead of NPI. 15 Japanese expressions such as XP-sika 'all but XP,' rokuna-N 'good/decent N' and dare-mo (indefinite person+suffix -mo)/nani-mo (indefinite thing+suffix -mo) 'nobody/nothing' must occur with Neg -nai as shown in (47), (48) and (49)/(50)), respectively. 16,17 15 Kataoka (2007:1, fn.2) points out as follows. "The notion of negative polarity was introduced by works such as Fauconnier 1975 and Ladusaw 1979 in order to give an account for the syntactic behavior and semantic characteristics of any(-) in English which requires negative environment, and it originally had a close connection with a particular semantic property of those items: they induce a scale interpretation in the terms of Fauconnier 1975, or downward entailment interpretation in the terms of Ladusaw 1979, which allows an element without universal force, if it is combined with negation, to give rise to an interpretation of universal negation. It should be noticed that not all NSEs necessarily induce the scale interpretation: e.g., amari ('too much') and rokuna-N as in [(48)] in Japanese, and therefore the term NPI should not be used as a purely descriptive term if we take that into consideration. We henceforth use the term NSE to refer to those elements which require negation, including those which have been called NPI in English or other languages, for ease of exposition." 16 To be precise, dare-mo and nani-mo induce an interpretation of universal negation combined with Neg. 17 Kataoka (2007: 28) mentions as follows: "The expression dare-mo/nani-mo can occur by itself or can be accompanied by a noun phrase attached by a case marker (CM), such as gakusei-CM daremo (any student). I treat the whole phrase as a constituent of argument NP, and will refer to the phrase as (N-CM) dare-mo/nani-mo." Since I only use rokuna-N in the Neg-related discussion of CM-Cleft, I do not go into the internal structure of (N-CM) dare-mo/nani-mo. For the reason why Kataoka (2007) takes this view, see Kataoka 2006: 204, fn.2. 145 (47) (Kataoka 2007: 1, (1)) a. Taro-wa manga-sika yoma-nai. / *yomu. Taro-TOP comics-all:but read-Neg / read (Roughly) 'Taro does not read any kind of book but comics.' b. Taro-sika manga-o yoma-nai. / *yomu. Taro-all:but comics-ACC read-Neg / read (Roughly) 'Nobody but Taro reads comics.' (48) (Kataoka 2007: 1, (2)) a. Jiro-wa kyoositu-de rokuna-koto-o iwa-nai. / *iu. Jiro-TOP classroom-in good-thing-ACC say-Neg / say (Roughly) 'Jiro does not say any good thing in the classroom.' b. Rokuna-gakusei-ga gakkai-de happyo-o si-nai. / *suru. good-student-NOM conference-LOC presentation-ACC make-Neg /make (Roughly) 'No good students make a presentation at the conference.' (49) (Kataoka 2007: 28, (93)) a. (Gakusei-ga) dare-mo sinbun-o yoma-nai / *yomu (koto) (student-NOM) DARE-MO newspaper-ACC read-Neg / read (Comp) '(The fact that) no students read newspapers.' b. Hanako-wa tanzyoobi-ni (tomodati-o) dare-mo syootai-si-nai /*syootai-suru. Hanako-TOP birthday-on (friends-ACC) DARE-MO invite-do-Neg / invite-do 'Hanako did not invite any friend on her birthday.' 146 (50) (Kataoka 2007: 28, (94)) a. So-no-hako-no-naka-kara-wa nani-mo dete-ko-nak-atta. that-GEN-box-GEN-inside-from-TOP NANI-MO get:out-come-Neg-Past 'Nothing came out of that box.' b. Taro-wa kudamono-o nani-mo tabe-nai / *taberu. Taro-TOP fruits-ACC NANI-MO eat-Neg / eat 'Taro does not eat any fruit.' NSEs in English such as any(-) cannot occur outside the c-command domain of the negative element (not), as observed in (51). Based on this observation, the condition in (52) has been generally agreed upon in the literature. (51) a. He didn't [ invite anybody.] b. *Anybody didn't [invite him.] (52) (Kataoka 2007: 2, (4)) An NSE must be c-commanded by Neg at LF. (cf. Klima 1964) It has long been assumed that the condition in (52) also holds of Japanese NSEs without serious discussions on its validity. Furthermore, under the condition in (52), NSEs must always be c-commanded by Neg. Hence, many researchers have taken (52) as the licensing condition on NSEs. Kataoka (2007), on the other hand, argues that not all NSEs in Japanese obey the condition in (52) and there are at least three types of NSEs in terms of syntactic conditions that they 147 observe and only one of the three types observe the condition in (52). 18 More specifically, Kataoka (2007) proposes the following structural conditions on rokuna-N, XP-sika and (N-CM) dare-mo/nani-mo. (53) (Kataoka 2007: 16, (43)) Rokuna-N must be c-commanded by Neg at LF. (54) (Kataoka 2007: 21, (64)) XP-sika must be in the NegP-Spec position at LF. (55) (Kataoka 2007: 30, (106)) (N-CM) dare-mo/nani-mo must c-command Neg at LF. Thus, under Kataoka’s (2007) analysis of NSEs in Japanese, among the three NSEs only Rokuna-N obeys the same condition as English NSEs in (52). In the experiment I will discuss below to argue against the movement analyses of CM-Cleft, I will test the possibility of the NSE in the focus position when the Neg is in the matrix clause and when the Neg is in the presuppositional CP. The undelying assumption on the experiment is that the matrix Neg can c-command the focus position, while the Neg in the presuppositional CP cannot. Thus, I need to have NSEs that must be c-commanded by Neg. In Kataoka (2007) Rokuna-N is such an NSE. In order to ensure that Rokuna-N must in fact be c-commanded by Neg, I will 18 Chung & Hong (1998), Chung (2006), Sells (2006) and Sells & Kim (2006) discuss the existence of NSEs that do not obey the condition in (52) in Korean. 148 summarize below Kataoka’s (2007) arguments for the condition on rokuna-N in (53). 19 4.3.2 Arguments for the condition on Rokuna-N Kataoka (2006, 2007: Chapter 4) presents three arguments for the condition (53) based on the facts of (i) BVA interpretation, (ii) resumption and (iii) the intervention effects of clause boundaries. First, let us see Kataoka’s (2007) argument based on the BVA facts. As we have seen in Chapter 1, the position of the DL in the Deep OS-type construction is an IP-adjoined position under Ueyama 1998. In regard to the position of Neg, Kataoka (2007) assumes (56). (56) (Kataoka 2007: 6, (11)) Sentential negative element Neg (-nai) is in the sister position of the VP at LF. (e.g., Masuoka 1989, Pollock 1989) Kataoka (2006: Chapter 3.4, 2007: Chapter 3.2) demonstrates that the Deep DL cannot be in the c-command domain of Neg at LF on the basis of BVA, DR and resumption facts of Japanese, in support of (56). 20 Then, when we force rokuna-N to be a Deep DL, we predict, under the condition in (53), that the sentence is unacceptable since rokuna-N is not c-commanded by Neg. See the following examples. 19 In the Neg-related experiment in this work, I need to have NSEs that have to be c-commanded by Neg at LF. Thus, I only use rokuna-N as an NSE in the Neg-related experiment below. Though many researchers still assume that other NSEs also comply with the condition in (52), I will put these NSEs aside in my experiments since Kataoka (2007) demonstrates that they don’t comply with the condition in (52). Kataoka 2007: Chapter 5 and 7 contains arguments for the conditions on XP-sika and (N-CM) dare-mo/nani-mo, which are different from the one in (52). 20 But see fn.25 for some informant judgments against the assumption and her explanation for that. 149 (57) (Kataoka 2007: 17, (48)) a. * [Rokuna-kaisya-o] i [soko i -o sitteiru-sensei]-ga gakusei-ni syookaisi-nak-atta. good-company -ACC it-ACC know professor-NOM student-DAT introduce-Neg-Past 'No good companies, the professor who knows it introduced to students.' b. * [Rokuna-ginkoo-o] i [soko i -no bengosi]-ga Mitubisi-ni suisensi-nak-atta. good-bank -ACC it-GEN attorney-NOM Mitsubishi-DAT recommend-Neg-Past 'No good banks, its attorney recommended to Mitsubishi.' When we force BVA(rokuna-N, so-ko) in (57), the sentences become unacceptable as predicted by Kataoka (2007). 21 It should be noted that rokuna-N can be a dislocated object in the OS-construction as shown in (58). (58) (Kataoka 2007: 17, (47)) a. [Rokuna-kaisya-o] Yamada-sensei-ga gakusei-ni syookaisi-nak-atta. good-company -ACC Yamada-professor-NOM student-DAT introduce-Neg-Past 'No good companies, Prof. Yamada introduced to students.' b. [Rokuna-ginkoo-o] Toyota-ga Mitubisi-ni suisensi-nak-atta. good-bank -ACC Toyota-NOM Mitsubishi-DAT recommend-Neg-Past 'No good banks, Toyota recommended to Mitsubishi.' Second, according to Ueyama & Hoji 2001 and Hoji & Ueyama 2003, resumption forces a 21 Kataoka (2007: 17, fn.18) notes that "[t]here are some speakers who accept these examples." Kataoka (2007) tries to explain this judgmental fluctuation in the appendix and points out that this can come from the variation in the bases for the BVA interpretation. Hajime Hoji (pc in April, 2007) suggested to me that it might be possible for Deep DL to be adjoined to VP. If so, we have to identify an independent way to force Deep DL to be adjoined to IP and conduct an experiment accordingly. 150 DL in a sentence to be a Deep DL. Thus, in the same light as in BVA, when a sentence has Rokuna-N as a DL and resumption at the position related to the DL, we predict, under the condition in (53), that the sentence is unacceptable since rokuna-N, being a Deep DL, is not c-commanded by Neg. See the following examples. (59) (Kataoka 2007: 18, (49)) a. Rokuna-hon-o Taroo-ga yoma-nak-atta. good-company-ACC Taro-NOM read-Neg-Past 'No good books, Taro did not read.' b. *Rokuna-hon-o Taroo-ga sore-o yoma-nak-atta. good-company-ACC Taro-NOM it-ACC read-Neg-Past (60) (Kataoka 2007: 18, (50)) a. [Rokuna-kaisya-o] Yamada-sensei-ga gakusei-ni syookaisi-nak-atta. good-company -ACC Yamada-professor-NOM student-DAT introduce-Neg-Past 'No good companies, Prof. Yamada introduced to students.' b. *[Rokuna-kaisya-o] Yamada-sensei-ga gakusei-ni soko-o syookaisi-nak-atta. good-company -ACC Yamada-professor-NOM student-DAT there-ACC introduce-Neg-Past The only difference between the (a) and (b) examples above is the absence in (a) and the presence in (b) of the resumption (sore in (59) and soko in (60)) in the canonical object position. The (b) examples in (59) and (60) with resumption are both unacceptable as predicted by Kataoka (2007). It is noted that resumption itself is not a problem as the acceptability of (61) 151 shows. (61) (Kataoka 2007: 18, (49)) Harry Potter-o Taroo-ga sore-o yoma-nak-atta. Harry Potter-ACC Taro-NOM it-ACC read-Neg-Past 'Harry Potter, Taro did not read it.' The above discussion based on BVA and resumption shows that rokuna N complies with the condition in (53), provided that Deep DL is outside the c-command domain of Neg. Kataoka (2006: Chapter 4.3) presents further argument for the condition in (53). She observes that when a sentence is complex having rokuna-N in the main clause and Neg in the embedded clause, the sentence is unacceptable, as shown in (62). (62) (Kataoka 2006: 105, (27)-(30)) 22 a. * Rokuna hito-ga [Taroo-ga manga-o yoma-nai]-to it-ta. good person-NOM Taroo-NOM comic-ACC read-Neg-that say-Past '(Lit.) Good persons said that Taroo does not read comics.' b.* Rokuna gakusee-ga [Yamada-sensee-ga syukudai-o dasa-nai]-riyuu-o tazune-ta. good student-NOM Yamada-teacher-NOM homework-ACC assign-Neg-reason-ACC ask-Past '(Lit.) Good students ask the reason why Mr.Yamada does not assign homework.' c.* [Hon-o yoma-nai]-gakusei-ga, rokuna tooan-o kaku. book-ACC read-Neg student-NOM good answer-ACC write '(Lit.) Students that do not read books write good answers.' 22 Gloss and translation in (62) and (63) are my own. 152 d.* Konkaino goodoo setumeekai-ni syatyoo-ga syussekisi-na-kereba, this:time joint recruiting:fair-to president-NOM attend-Neg-if rokuna gakusee-ga kuru. good student-NOM come '(Lit.) If the president does not attend the next joint recruiting fair, good students will come.' On the other hand, when a sentence has rokuna-N in the embedded clause and Neg in the main clause, the sentence is acceptable though it is a little degraded, which is shown in (63). 23 (63) (Kataoka 2006: 105, (31)) a. ? Boku-ni-wa [Yamada-ga rokuna ronbun-o kaku]-to omo-e-nai. me-to-TOP Yamada-NOM good paper-ACC write-that think-POT-Neg '(Lit.) To me, it cannot be thought that Yamada writes good paper.' b. ? Kono-kurasu-ni-wa [sensee-ni rokuna koto-o iu]-yatu-ga i-nai. this-class-in-TOP teacher-to good thing-ACC tell person-NOM be-Neg '(Lit.) There is no one in this class that tells good things to his/her teacher.' When there is a clause boundary between rokuna-N and its associated Neg, sentences are never acceptable if rokuna-N is in the main clause, while sentences can be acceptable if Neg is in the main clause. When Neg is in the main clause, it c-commands rokuna-N in the embedded clause. Thus, these facts also show that rokuna-N obeys the condition in (53). Kataoka’s (2006, 2007) three arguments above strongly suggest that the condition (53) is 23 Kataoka (2006:105) mentions that sentences of the structure in (63) are acceptable only when the matrix verbs are omou 'think' and verbs that express existence such as iru 'be/exist'. She assumes that unacceptability of the sentences with other verbs is attributed to some pragmatic factors. 153 valid, that is, Rokuna-N must be c-commanded by Neg at LF. Therefore, in the Neg-related experiment on CM-Cleft in Chapter 7, I will use Rokuna-N as an instance of NSEs that must be c-commanded by Neg at LF. 4.4 Summary In this chapter I have presented some schematic asymmetries independently established in other works that could be regarded as confirmed ones, together with the relevant experimental designs and results. I have presented three such confirmed schematic asymmetries. The first two are BVA-related confirmed schematic asymmetries; one is on A and the other on B of BVA(A,B). The other is DR-related one. I have also presented hypotheses about NSEs proposed by Kataoka (2007). Though Kataoka (2007) has not obtained confirmed schematic asymmetries by conducting EPSA experiments, her hypotheses are quite tenable considering the data and arguments presented in her work. As I mentioned above, according to Hoji (2010b:14), predictions in language faculty science are based on at least two types of hypotheses: universal hypotheses and language-particular hypotheses. The hypothesis/hypotheses to be combined with the hypothesis we pursue must be the one(s) backed up by confirmed schematic asymmetries. Without using hypotheses backed up by confirmed schematic asymmetries, we won’t be able to expect to obtain robust informant judgments in the experiments. Furthermore, when the newly formed *Schema-based prediction are disconfirmed, we will have difficulty in identifying which hypothesis is wrong. Furthermore, we have seen that by controlling the type of QPs and dependent terms in BVA interpretation, we can ensure the FD-based BVA, which is crucially based on the structural relation of c-command. Similarly, we have seen that by controlling the type of QPs, 154 we can identify that the DR is most likely based on a c-command relation at LF. It is particularly important to have a way to sort out the FD-based BVA or DR based on c-command. As mentioned above, Hoji (2009:175) assumes "that the only structural relation visible at LF is that of c-command and that the informant's intuitions about 'meaning' would be a reflection of, and hence would be revealing about, properties of the Computational System only if they are based on properties of LF." Therefore, when we are trying to reveal properties of the Computational System in other constructions (CM-Cleft as in this work), we should concentrate on the c-command relation among NPs in the constructions. FD-based BVA and DR based on c-command provide us with tools to do that. With such tools in hand I will turn to experiments on CM-Cleft in the following two chapters. In Chapter 6, I will discuss two experiments based on the BVA and DR facts of CM-Cleft. The purpose of these experiments is to check the structural relation between the focus phrase and the presuppositional CP, more specifically, to check if the focus phrase c-commands the elements inside the presuppositional CP. In Chapter 7, I will address two experiments based on the NSE and resumption facts of CM-Cleft. The latter crucially refers to BVA. The purpose of these experiments is to provide arguments against the movement analysis of CM-Cleft. 155 Chapter 5: The position of the focus phrase 5.1 Introduction With the methodology and some important preliminary experiments in hand, we are finally in a position to conduct experiments on CM-Cleft. In this chapter, I will first propose the base-generation analysis pursued in this work in line with the analysis of CM-Cleft proposed by Hoji (1987, 1990). I will then compare the proposed analysis with another base-generation analysis proposed by Kizu (2005), and consider what different predictions these two analyses make. The two analyses differ in the position of the focus phrase; in the proposed analysis the focus position c-commands elements inside the presuppositional CP, while in Kizu’s (2005) analysis it does not. Based on this structural difference, I will make predictions in regard to BVA (Bound Variable Anaphora) and DR (Distributive Reading) in CM-Cleft and conduct experiments. I will then examine the results of the experiments. The results of the experiments show that the *Schema-based predictions made under Kizu’s (2005) analysis are disconfirmed, suggesting that the base-generation analysis proposed by Kizu (2005) is not tenable. The *Schema in Kizu’s (2005) analysis is an ok Schema under the proposed analysis. The proposed analysis, however, cannot account for all the BVA facts. In order to account for them, I will introduce two hypotheses; (i) the focus position in CM-Cleft is an A'-position, and (ii) a phrase in the focus position, a single member A'-chain, can be composed with another A'-chain, forming a single A'-chain, and is interpreted at the A-position of the new chain, that is, at the tail of the chain. I will show that the predicted schematic asymmetries about BVA obtained based on these two hypotheses as well as the proposed LF-structure of CM-Cleft get confirmed. Thus, the results of the experiments give support to the proposed analysis. 156 5.2 A proposal: a base-generation analysis of CM-Cleft Recall the representative base-generation analysis of CM-Cleft proposed by Hoji (1990), which was presented in Chapter 3.2.1. 1 (1) (Hoji 1990: (225)) CM-Cleft: [ S' Op i [ S ... t i ... ] no]-wa NP i -(CASE) da (Syntactic movement is obligatory and the aboutness licensing is not possible.) I pursue the base-generation analysis basically adopting this analysis of CM-Cleft. Based on (1), I propose the structure of CM-Cleft as illustrated in (2). (2) IP VP I SC V | CP NP 1 -CM da IP C | Op 1 IP no-wa … t 1 … In the structure in (2), the focus phrase NP 1 -CM forms a small clause (SC in (2)) with the 1 As shown in Chapter 2, Hoji 1987 is the precursor of the base-generation analysis of CM-Cleft, proposing a structure similar to the one in (1). The only difference between the structure in (1) and the one proposed in Hoji 1987 is that what is moved inside the presuppositional CP is an empty operator Op in the former, while it is pro in the latter. The analysis of the CM-Cleft in Hoji (to appear) assumes the same pro movement as in Ueyama’s (1998) analysis of the Deep-OS type constructions, which corresponds to the empty operator movement. The structure in (1) and the one in Hoji 1987 can therefore be understand as notational variants of each other. 157 presuppositional CP. This small clause is selected as a complement by the copula verb da. A predication relation is established between the focus phrase and the presuppositional CP. Let us provisionally assume that the focus position is an A-position as also assumed in Kizu 2005. 2 Kizu (2005) also proposes a base-generation analysis of CM-Cleft basically adopting Hoji’s (1990) analysis, which was presented in Chapter 3.2.2. Her proposed structure of CM-Cleft is repeated here as (3). 3 (3) Structure for local cleft constructions: (based on Kizu 2005: 55 (89), 72 (32) and 109 (17a)) TopP CP IP Op i C' I' IP C VP I | … t i … no V' PP V | | P' aru XP P | de In the structure in (3), the copula verb da is a complex verb, which consists of the P de and the V aru. The P de incorporates into the V at some point of derivation, or it fuses phonologically under linear adjacency. A focus phrase XP is base-generated as a complement of the P. 2 Kizu (2005) does not clearly state that the position of the focus phrase is A-position. However, as the structure in (3) shows, the focus phrase is the complement of P. Complement positions are commonly assumed as A-positions. 3 As mentioned in Chapter 2.2.2, fn 4, Kizu’s (2005) analysis crucially differs from Hoji’s (1990) in long-distance cleft constructions. As far as the local cleft is concerned, her analysis is very similar to Hoji’s (1990) except for (i) the position of the focus phrase, and (ii) the trigger of the empty operator movement and its landing site. 158 One crucial difference between the proposed structure in (2) and Kizu’s (2005) structure in (3) is the structural relation between the focus phrase and the presuppositional CP. In (2), the focus phrase and the presuppositional CP form a small clause, so that they mutually c-command each other. As a result, the focus phrase c-commands elements inside the presuppositional CP. In (3), on the other hand, the focus phrase is a complement of the P, whose projection PP is in tern a complement of the V aru. Thus, the focus phrase is deeply embedded in the structure, so that it cannot c-command elements inside the presuppositional CP. Therefore, we can distinguish between Kizu’s (2005) analysis and the analysis proposed here by examining the c-command relation between the focus phrase and some elements inside the presuppositional CP. As we have observed in Chapter 4, BVA (bound variable anaphora) interpretation, specifically, FD-based BVA, is crucially based on the c-command relation between a QP and a dependent term. Thus, I first examine BVA facts of CM-Cleft, focusing on the relation between a QP in the focus position and a dependent term inside the presuppositional CP. 5.3 BVA in the CM-Clefts 5.3.1 C-command relation between the focus phrase and the presuppositional CP Recall the conditions on FD-based BVA proposed in Ueyama (1998), which have been presented in Chapter 4. (4) (Ueyama 1998: 115, (65)) a. Structural condition on FD: *FD(α,β) if α does not c-command β at LF. 159 b. Lexical condition on FD: 4 *FD(α,β) if β is a large NP. According to these conditions, FD(α, β) is possible only if (i) α c-commands β at LF, and (ii) β is a small NP. So-ko 'it/that institution', so-re 'it/that thing' and so-itu 'he/that guy' are small NPs, in contrast to so-no-NP 'that NP', which is a large NP. In addition, B of BVA(A,B) must be a singular-denoting anaphoric expression whose value is not fixed to a certain individual, and we have seen that so-words just shown (so-ko, so-re and so-itu) are, but a-words such as a-soko, a-re and a-itu are not, qualified as B of BVA(A,B). This is because a-words are connected to individuals that the speaker knows through his direct experience, that is, a-words are inherently referential. Recall that as shown in Chapter 4.2.1.1, Ueyama (1998:226) assumes that the QP α in FD(α,β) undergoes QR, whose trace is then mapped to a bound variable, and β in FD(α,β) is also mapped to a bound variable, resulting in the sentence containing bound variable anaphora. The bound variables must be A'-bound by the QP, so that their positions must be A-position. Therefore, both the position of the QR trace of α and that of β in FD(α, β) must be A-positions. Recall also that there are three sources of BVA; FD-based BVA, ID-based BVA and quirky binding. BVA(A,B) is available without A c-commanding B under ID-based BVA and quirky binding. Since we need to focus on the c-command relation in CM-Cleft, we need to exclude ID-based BVA and quirky binding. We have seen in Chapter 4 that ID-based BVA is possible only with B-type QPs, including those listed in (5b). 4 As mentioned in Chapter 4.2.3.2.2, according to Ueyama (1998:126) the distinction between large NPs and small NPs is basically determined based on the 'amount of semantic content' on N. Since the 'amount of semantic content' is a matter of degree, it follows that it is a partition relative to each other, rather than an absolute distinction. Since the 'amount of semantic content' is subjective in nature, it is well expected that the ways of classification vary depending on speakers and contexts." 160 (5) (Ueyama 1998: 124, (12)) a. A-type QPs: NP-sae 'even NP' kanarinokazu-no NP 'most of the NPs' 10 izyoo-no NP 'ten or more NPs' 55%-no NP '55% of the NPs' NP1 to NP2 (to) 'NP1 and NP2' NP1 ka NP2 (ka) 'either NP1 or NP2' b. B-type QPs: do-no NP 'which NP' do-no NP-mo 'every NP' (subete-no NP 'every NP') 5 Thus, we can avoid ID-based BVA by using A-type QPs. On the other hand, the condition of the availability of quirky binding in regard to QPs is that the QP must refer to a specific group of individuals. 6 Other than the B-type QPs in (5b), NP1 to NP2 (to) and NP1 ka NP2 (ka) among the A-type QPs in (5a) refer (or are more easily understood to refer) to a specific group of individuals. Thus, by avoiding such QPs and instead using QPs that cannot (easily) refer to a specific group of individuals, such as NP-sae 'even NP', we can avoid quirky binding. In short, we can ensure the FD-based BVA by using QPs that do not refer to a specific group of individuals as the binder and small so-words as the dependent term. Then, under Kizu’s (2005) analysis with the pf-LF correspondence that the CM-Cleft of 5 See Chapter 4, fn.1 for why subete-no NP 'every NP' is in parentheses in Ueyama (1998). 6 See Chapter 4 (10) for other conditions of the availability of quirky binding discussed in Ueyama (1998). 161 the form NP-CM V-T-no-wa NP-CM da correspond to the LF structure in (3), we can obtain the following *Schema. (6) *Schema: [… so-NP-no NP-CM … V-T]-no-wa QP-CM da BVA(QP, so-NP) This is because the focus position, which is the A-position, does not c-command elements inside the presuppositional clause in accordance with (3). Since the case-markers on the QP and NP containing so-NP are not specified in (6), the QP in the focus position can be the subject, accusative object, dative object or postpositional phrase. Thus, one of the *Examples conforming to the *Schema in (6) is shown in (7), in which the QP in the focus position is the subject with the nominative case-marker ga. 7 (7) *Example: So-ko-no syokuin-ni rihuzinna yookyuu-o siteiru-no-wa tookyooto-sae-ga da that-place-Gen official-Dat unreasonable request-Acc doing-that-Top Tokyo-even-Nom be 'It is even the Tokyo metropolitan government that is making unreasonable requests to its officials.' I have checked the acceptability of the example in (7) and two other examples with QP kanarinokazu-no tihoozititai 'a considerable number of local governments' and 55% izyoo-no 7 The example (7) is one of the examples in EPSA[25]-1 that I have been conducting. However, in EPSA[25]-1, (7) is not a *Example but an ok Example, and as we will see shortly, its corresponding ok Schema is as in (i), in which the case-markers on the focus phrase and the NP containing the so-ward are specified. (i) [… so-ko-no NP-ni … V-T]-no-wa QP-ga da BVA(QP, so-ko) The details of the result of EPSA[25]-1 will be shown in Section 3.3. 162 tihoozititai '55% or more local governments' conforming to the schema in (6). The average score of the judgments of 13 informants is 67 on the scale of 0-100, with 0 being completely unacceptable and 100 being fully acceptable. In fact, six out of the 13 informants found (7) quite easily acceptable with BVA interpretation between QP and soko 'that place'. This result disconfirm the*Schema-based prediction as indicated in (6) that was made on the basis of the structure of CM-Cleft in (3) proposed by Kizu (2005). On the other hand, under the analysis with the pf-LF correspondence that the CM-Cleft of the form NP-CM V-T-no-wa NP-CM da correspond to the LF structure in (2), the schema in (6) is an ok Schema. That is because the focus position in (2), that is the A-position, c-commands elements inside the presuppositional clause. The result thus confirms this ok Schema-based prediction. 5.3.2 ok Schema-based prediction fails to be confirmed As just mentioned, the *Schema in (6) under Kizu’s (2005) analysis is an ok Schema under the analysis with the LF-structure in (2), in which the focus position is the A-position (repeated here in (8) as ok Schema). 8 (8) (= (6)) ok Schema: [… so-NP-no NP-CM … V-T]-no-wa QP-CM da BVA(QP, so-NP) 8 A *Schema that corresponds to (8) would be (i), in which the so-NP is in the focus position and the QP inside the presuppositional clause. The focus position cannot be c-commanded by elements inside the presuppositional CP under the structure in (2). (However, as we will see shortly, it will be an ok Schema with some specification on case-markers when we assume chain binding.) (i) *Schema: […QP-CM … V-T]-no-wa so-NP-no NP-CM da BVA(QP, so-NP) 163 Thus, the example in (7), which was once a *Example under Kizu 2005, is now an ok Example conforming to (8). It is repeated below in (9) as an ok Example. (9) ok Example (=(7), given as a *Example under Kizu 2005): So-ko-no syokuin-ni rihuzinna yookyuu-o siteiru-no-wa tookyooto-sae-ga da that-place-Gen official-Dat unreasonable request-Acc doing-that-Top Tokyo-even-Nom be 'It is even the Tokyo metropolitan government that is making unreasonable requests to its officials.' As mentioned above, the case-markers on the QP and NP containing so-NP are not specified in (8). Therefore, (10) could possibly be another ok Example under the formulation of the ok Schema as in (8). In (9) what is in the focus position is the subject with the nominative case-marker ga, while in (10) what is in the focus position is the dative object with the dative case-marker ni. (10) ok Example: So-ko-no syokuin-ga rihuzinna yookyuu-o siteiru-no-wa tookyooto-ni-sae da that-place-Gen official-Nom unreasonable request-Acc doing-that-Top Tokyo-Dat-even be 'It is even to the Tokyo metropolitan government that its officials are making unreasonable requests.' I have checked the acceptability of the example in (10) and two other examples with QP kanarinokazu-no tihoozititai 'a considerable number of local governments' and 55% izyoo-no 164 tihoozititai '55% or more local governments' conforming to the ok Schema in (8). 9 The average score of the judgments of 13 informants with BVA interpretation between QP and soko is 4 again on the scale of 0-100, with 0 being completely unacceptable and 100 being fully acceptable. Thus, the ok Schema-based prediction based on the LF-structure in (2), in which the focus position is the A-position, fails to be confirmed. 10,11 Recall that the average score of the judgments of 13 informants on the example in (9) (= (7)) and other two examples is 67. The example in (10) is lexically as well as structurally similar to (9). Therefore, it is implausible to reduce the low acceptability of (10) to parsing difficulty or the unnaturalness of the interpretation. Thus, we could conclude that there is something wrong in the analysis of CM-Cleft presented above. 5.3.3 Modification of the analysis 5.3.3.1 Focus position as an A'-position The difference between the two ok Examples in (9) and (10) is that in (9) the QP in the focus position is the subject with the nominative case-marker ga, while in (10) the QP is the dative object with the dative case-marker ni. LF-structures of (9) and (10) on the basis of (2) are expressed as in (11a) and (11b), respectively. 9 As we will see shortly, (10) is in fact not an ok Example but a *Example in EPSA[25]-1. Its corresponding *Schema is as in (i), in which the case-markers on the focus phrase and the NP containing the so-ward are specified. (See fn. 7 above) (i) [… so-ko-no NP-ga … V-T]-no-wa QP-ni da BVA(QP, so-ko) 10 Recall the discussion in Chapter 3.3.3 that ok Schema-based predictions cannot be disconfirmed since it is not possible to exhaustively check all the possible ok Examples of an ok Schema. 11 Under Kizu’s (2005) analysis, (10) is a *Example since the focus QP does not c-commands so-NP inside the presuppositional CP. Thus, the low acceptability of (10) is as expected. However, as shown above (9) is also a *Example and the prediction is disconfirmed. Thus, under Kizu’s (2005) analysis that does not refer to the case-marker difference on the focus phrase cannot account for the difference in acceptability between (9) and (10). See Section 3.5 for more discussion on Kizu (2005)’s analysis. 165 (11) a. [ SC [ CP [ IP Op i [ IP t i soko-no NP-ni ... V-T]]-no-wa] [QP i -ga]] da b. [ SC [ CP [ IP Op i [ IP soko-no NP-ga t i ... V-T]]-no-wa] [QP i -ni]] da Focusing on the differences in the c-command relation in these structures, in (11a) the position inside the presuppositional CP that is related to the focus QP (i.e., the canonical argument position of V corresponding to the focus QP, that is, the trace position of Op) c-commands the dependent term soko, while in (11a) the position inside the presuppositional CP that is related to the focus QP does not c-commands soko. It would be reasonable to seek the cause of the difference in acceptability between (9) and (10) in this structural difference between (11a) and (11b). As I briefly introduced in Chapter 3.2.1, Hoji & Ueyama (2003:12) account for this difference by assuming that the focus position in CM-Cleft is an A'-position. As pointed out above, both the position of the QR trace of α and that of β in FD(α, β) must be A-positions. If we assume that the focus position in CM-Cleft is an A'-position, then, the QP in the focus position cannot be interpreted at the position. Hoji & Ueyama (2003:12) then suggest, along the line of Barss 1986, that "in CM-Cleft: [Op 1 [[ ... t 1 ... ] no]]-wa/ga [ DL ...-CM ] 1 da, the A'-chain consisting solely of the DL [i.e., the focus phrase: YM] and the A'-chain (Op 1 , t 1 ) can be composed into a single A'-chain (DL 1 , Op 1 , t 1 ), and as a result the elements contained in the DL is interpreted at the position of the trace (t 1 ) with respect to BVA." I adopt Hoji & Ueyama’s (2003) assumptions that the focus position in CM-Cleft is an A'-position as well as their assumption about chain binding. As a necessary condition of the chain binding, I assume that a single member A'-chain can be composed with another A'-chain only if the former c-commands the latter. Thus, the following two hypotheses are introduced in this work. 166 (12) The focus position in CM-Cleft is an A'-position. (13) A single member A'-chain can be composed with another A'-chain, forming a single A'-chain, only if the single member A'-chain c-commands the other A'-chain. Under these hypotheses, in (11a) a single A'-chain (QP i , Op i , t i ) is composed and the chain, more specifically, a member of the chain in an A-position, that is, t i , c-commands the dependent term soko, resulting in the availability of BVA(QP, soko). On the other hand, in (11b) a single A'-chain (QP i , Op i , t i ) is composed, but t i , which is in an A-position, does not c-command the dependent term soko, resulting in the unavailability of BVA(QP, soko). This accounts for the difference in acceptability between (9) and (10). With the hypotheses in (12) and (13), together with the LF structure in (2) and the hypotheses with regard to BVA listed above, we can obtain a predicted schematic asymmetry as in (14) under the analysis proposed in this work, in which the ga-marked phrase is a subject and the ni-marked phrase is a dative object of the sentence. 12 (14) Schema group: a. ok Schema A [… so-NP-no NP-ni … V-T]-no-wa QP-ga da BVA(QP, so-NP) b. *Schema B [… so-NP-no NP-ga … V-T]-no-wa QP-ni da BVA(QP, so-NP) c. ok Schema C [… so-NP-no NP-ga … V-T]-no-wa QP-ni da 12 Some verbs in Japanese such as iru 'need' take an ni-marked phrase as the subject and a ga-marked phrase as the object. Such verbs need to be excluded in the schema group in (14). 167 (With so-NP referring to a particular individual/object) The predicted schematic asymmetry in (14) is what is tested in EPSA[25]-1. One ok Example conforming to the ok Schema A is (9) (= (7)), while one *Example conforming to the *Schema B is (10). In addition, there are two more examples tested in each Schema with QP kanarinokazu-no tihoozititai 'a considerable number of local governments' and 55% izyoo-no tihoozititai '55% or more local governments'. Thus, a total of nine examples are tested, which is shown in (15)-(17). (15) a. ok Examples A-1 (= (9)): BVA(tookyooto-sae, soko) So-ko-no syokuin-ni rihuzinna yookyuu-o siteiru-no-wa tookyooto-sae-ga da that-place-Gen official-Dat unreasonable request-Acc doing-that-Top Tokyo-even-Nom be 'It is even the Tokyo metropolitan government that is making unreasonable requests to its officials.' b. *Examples B-1 (= (10)): BVA(tookyooto-sae, soko) So-ko-no syokuin-ga rihuzinna yookyuu-o siteiru-no-wa tookyooto-ni-sae da that-place-Gen official-Nom unreasonable request-Acc doing-that-Top Tokyo-Dat-even be 'It is even to the Tokyo metropolitan government that its officials are making unreasonable requests.' c. ok Examples C-1 (=(15b)): with soko referring to Ministry of Finance already mentioned in the preceding context (16) a. ok Examples A-2 : BVA(kanarinokazuno tihoozititai, soko) So-ko-no syokuin-ni rihuzinna yookyuu-o siteiru-no-wa kanarino kazu-no 168 that-place-Gen official-Dat unreasonable request-Acc doing-that-Top considerable number-Gen tihoozititai-ga da local:government-Nom be 'It is a considerable number of local governments that are making unreasonable requests to its officials.' b. *Examples B-2 : BVA(kanarinokazuno tihoozititai, soko) So-ko-no syokuin-ga rihuzinna yookyuu-o siteiru-no-wa kanarino kazu-no that-place-Gen official-Nom unreasonable request-Acc doing-that-Top considerable number-Gen tihoozititai-ni da local:government-Dat be 'It is to a considerable number of local governments that its officials are making unreasonable requests.' c. ok Examples C-2 (= (16b)): with soko referring to Ministry of Finance already mentioned in the preceding context (17) a. ok Examples A-3 : BVA(55% izyoono tihoozititai, soko) So-ko-no syokuin-ni rihuzinna yookyuu-o siteiru-no-wa 55% izyoo-no that-place-Gen official-Dat unreasonable request-Acc doing-that-Top 55%:or:more-Gen tihoozititai-ga da local:government-Nom be 'It is 55% or more local governments that are making unreasonable requests to its officials.' b. *Examples B-3 : BVA(55% izyoono tihoozititai, soko) So-ko-no syokuin-ga rihuzinna yookyuu-o siteiru-no-wa 55% izyoo-no that-place-Gen official-Nom unreasonable request-Acc doing-that-Top 55%:or:more-Gen 169 tihoozititai-ni da local:government-Dat be c. ok Examples C-3 (= (17b)): with soko referring to Ministry of Finance already mentioned in the preceding context Table 5.1 summarizes the result of EPSA[25]-1 on the predicted schematic asymmetry in (14), which has now 13 informants participated. 13 Table 5.1: A summary of the results of EPSA [25]-1 EPSA [25]-1 (Total 13 participants; 194 answers) ... as of Mar/2/2011 QP OK Schema A *Schema B OK Schema C NP-sae 'even NP' 13 values 66 9 values 1 9 values 80 kanarinokazu-no NP 'a considerable number of NPs' 13 values 65 9 values 9 9 values 77 55% izyoo-no NP '55% or more NPs' 13 values 65 9 values 1 9 values 82 Total 39 values 65 39 values 4 39 values 80 The average score on the three *Examples conforming to the *Schema is very low, while the average scores of the corresponding ok Examples are significantly higher than that of the *Examples. 14 Therefore, although the total number of informants is still small, the predicted 13 I have reported the average scores of the examples conforming to ok Schema A and *Schema B above, though the former was a *Schema and the latter is an ok Schema. See examples in (7) and (10), and the discussions thereabout. 14 As we have seen in Chapter 3.3.3, *Examples conforming to a *Schema should be completely unacceptable. However, we should allow some room for errors by informants due to the factors outside of the CS, such as the lack of attentiveness or patience, though researchers should of course try their best to eliminate the effects of such extra-grammatical factors on the informant judgments. Therefore, it seems reasonable to consider a score of as low as 5 or 6, for example, on a *Schema not as disconfirmation of the *Schema-based prediction. 170 schematic asymmetry in (14) could be said to have gotten confirmed so far. 5.3.3.2 Classification of informants With the result of the EPSA[25]-1 in hand, let us recall the fundamental asymmetry between a *Schema-based prediction and an ok Schema-based prediction presented in Chapter 4.3.3, repeated here as (18) and (19), respectively. (18) (Hoji 2010a: 31, (6)) A *Schema-based prediction: Informants judge any *Example conforming to a *Schema to be completely unacceptable under interpretation γ(a, b). (19) (Hoji 2010a: 31, (7)) An ok Schema-based prediction: 15 Informants judge ok Examples conforming to an ok Schema to be acceptable (to varying degrees) under interpretation γ(a, b). As we have seen in Chapter 4, according to (18) and (19), a predicted schematic asymmetry gets confirmed if and only if the informants' judgments on *Examples are consistently completely unacceptable and their judgments on the corresponding ok Examples are not completely unacceptable. Therefore, in order to obtain a confirmed schematic asymmetry, the *Schema-based prediction must survive a rigorous disconfirmation attempt, while at the same time the corresponding ok Schema-based predictions must be confirmed. 16 15 See Chapter 3, fn.11 for other possible formulations of an ok Schema based prediction. 16 Recall also that ok Schema-based predictions cannot be disconfirmed and they can only be confirmed, while *Schema-based predictions can be disconfirmed although they cannot be confirmed (see Chapter 3.3.3). 171 In EPSA[25]-1, the average score on the *Examples conforming to the *Schema is very low. Since the number of informants is still small and only three *Examples are tested, it is too early to say that the *Schema-based prediction has survived a rigorous disconfirmation attempt, but at least the *Schema-based prediction has been surviving so far. On the other hand, the average scores of the corresponding ok Examples are more than 0 and are significantly higher than that of the *Examples. Therefore, we can say that the ok Schema-based predictions have gotten confirmed so far. Thus, the result of EPSA[25]-1 as it stands is in favor of the proposed analysis. However, I want to look a little further into the result of EPSA[25]-1. One may feel that the average scores of the ok Examples conforming to the ok Schemata, especially the ok Schema A, are not as high, as one might expect though this is not crucial to the confirmation of ok Schema-based predictions. The (un)acceptability of ok Schemata can be affected by factors outside the Computational System such as parsing difficulty and the unnaturalness of the interpretation of the entire sentence. This is related to the resourcefulness of the informants in Hoji’s (2009:80) term. Hoji (2009:81) calls the attribute of the informants as shown in (20) structural/contextual resourcefulness. (20) The informants who try all the possibilities in parsing α (hence all the possible numerations that could correspond to α) and try to consider various pragmatic contexts to see if it is indeed impossible for them to accept sentence α under γ. Hoji (2009: 81) then points out that "[i]f the informants have a great deal of structural/contextual resourcefulness, they can reduce [the degree of parsing difficulty and the unnaturalness of the interpretation of the entire sentence: YM] to minimal, and for that reason 172 their judgments β would … [be] more revealing about the properties of the Computational System than the judgments by the informants who do not have as much structural/contextual resourcefulness." When informants register for EPSA to participate in the experiments, they are asked whether they understand (i) "bound variable anaphora" or "bound readings" and (ii) "A takes wide scope over B" as they are used in linguistic discussion. Hoji (2010a:40, fn.30) classifies the informants who state that they understand at least (i) or (ii) as "linguistically informed", while the ones who state that they understand neither of the two as "linguistically naïve." It is perhaps reasonable to assume that "linguistically informed" informants have more structural/contextual resourcefulness than "linguistically naïve" informants, though the degree must difer greatly among individuals. Hoji (2010a:40, fn.31) in fact reports, based on the actual result of an experiment, that ""linguistically-naïve" informants (14 informants) tend to judge the *Examples somewhat less acceptable than "linguistically-informed" informants (20 informants) (e.g., 39 by the former and 70 by the latter on Schemata B [i.e., *Schema: YM] … ; but the former also judge ok Examples less acceptable than the latter (e.g., 63 by the former and 88 by the latter on Schemata C [i.e., ok Schema: YM] …" 17 If we classify the 13 informants in EPSA[25]-1 into "linguistically-informed" informants and "linguistically-naïve" informants based on the above factors and calculate the average 17 This is based on the result of EPSA[5]-1 conducted by Hoji with regard to the so-called "local anaphors" otagai in Japanese. One example set included in EPSA[5]-1 is shown in (i). (i) (Hoji 2010a: 37, (19)) a. ok Example Mary-wa [John to Bill-ga otagai-ni toohyoosi-ta to] omoikonde-i-ta 'Mary thought that John and Bill had voted for each other.' b. *Example John to Bill-wa [Mary-ga otagai-ni toohyoosi-ta to] omoikonde-i-ta 'John and Bill thought that Mary had voted for each other.' c. ok Example John to Bill-wa [Mary-ga karera-ni toohyoosi-ta to] omoikonde-i-ta 'John and Bill thought that Mary had voted for them.' 173 score of each group, the result of EPSA[25]-1 in Table 5.1 comes as shown in Table 5.2. Table 5.2: A summary of the results of EPSA [25]-1 by group EPSA [25]-1 (Total 13 participants; 194 answers) ... as of Mar/2/2011 OK Schema A *Schema B OK Schema C Total 39 values 65 39 values 4 39 values 80 linguistically-informed (8 informants) 24 values 88 24 values 6 24 values 92 linguistically-naïve (5 informants) 15 values 30 15 values 1 15 values 60 The average score on the *Examples conforming to the *Schema by the five "linguistically-naïve" informants is very low, but the average scores of the ok Examples conforming to the ok Schemata, especially ok Schema A, are also very low. 18 Therefore, we cannot attribute the unacceptability of the *Examples conforming to the *Schema by these five informants solely to the hypothesized property/ies of the CS. On the other hand, the average score on the *Examples conforming to the *Schema by the eight "linguistically-informed" informants is still very low, while the average scores of the corresponding ok Examples are significantly more acceptable than the *Examples. Thus, as far as the "linguistically-informed" informants are concerned, we can obtain the 'clearer' schematic asymmetry. Then, how should we deal with the judgments by the "linguistically-naïve" informants? We may do some training sessions for them or conduct preliminary experiments and see if the scores of the ok Examples will improve while that of the *Examples remains very low. We should note, however, that the problem is the low scores on ok Schemata, which might 18 The focus phrase in CM-Cleft is followed by the copula da. For this NP-CM da sequence, when the nominative case-marker ga is used, the sequence NP-ga da sounds a little unnatural compare to the other case-markers (e.g., o, ni) or postpositions (e.g., kara 'from', to 'with'). This may be one of the reasons why the score of ok Schema A is lower than that of ok Schema C, though I won’t pursue any theoretical account for this here. 174 be due to some factors outside the CS, and that the *Schema-based prediction has not been disconfirmed. Therefore, it is not fatal to the proposed analysis and the result of EPSA[25]-1 still confirms the predicted schematic asymmetry in (14). 5.3.4 Dependent term in the focus position The CM-Cleft discussed above has a QP in the focus position and a dependent term inside the presuppositional CP. We will discuss next the case in which a QP is inside the presuppositional CP, while a dependent term is in the focus position. Under the proposed structure of the CM-Cleft in (2), the focus position is never c-commanded by any elements inside the presuppositional CP. Therefore, when a dependent term is in the focus position and a QP is inside the presuppositional CP, the dependent term can never be c-commanded by the QP. However, as pointed out in the preceding section, FD(α, β) is a relation between the QR trace of α and the dependent term β, both of which are in the A-position. Thus, in order to obtain FD-based BVA, the QR trace of α and the dependent term β must be in the A-position. Under the hypothesis in (12) that the focus position in CM-Cleft is an A'-position, however, the dependent term in the focus position cannot be interpreted there. It is thus interpreted by means of chain binding hypothesized in (13). The focus phrase, which is a single member A'-chain, and the operator chain (Op i , t i ) are composed into an A'-chain (focus phrase i , Op i , t i ) and the focus phrase is interpreted at the trace position of Op, which in an A-position. Given these hypotheses, then, if we ensure the FD-based BVA by using what Ueyama (1998) calls A-type QPs that do not refer to a specific group of individuals as A of BVA(A, B), we can obtain the predicted schematic asymmetry in (21), in which the ga-marked phrase is the 175 subject and the ni-marked phrase is the dative object of the sentence. 19,20 (21) Schema group: a. ok Schema A [… QP-ga … V-T]-no-wa so-NP-no N-ni da BVA(QP, so-NP) b. *Schema B [… QP-ni … V-T]-no-wa so-NP-no N-ga da BVA(QP, so-NP) c. ok Schema C [… QP-ni … V-T]-no-wa so-NP-no N-ga da (With so-NP referring to a particular individual/object) In the ok Schema A in (21a), the position inside the presuppositional CP related to the focus NP with a dependent term is c-commanded by the QP. On the other hand, in the *Schema B in (21b), the position inside the presuppositional CP related to the focus NP containing a dependent term is not c-commanded by the QP. The predicted schematic asymmetry in (21) is what is tested in EPSA[25]-2. Three QP Tookyooto-sae 'even the Tokyo metropolitan government', kanarinokazu-no tihoozititai 'a considerable number of local governments' and 55% izyoo-no tihoozititai '55% or more local governments' are used in the examples of each schema. Thus, a total of 9 examples are tested, which are shown in (22)-(24). 19 See fn.21 above for the verbs to take the NI GA V pattern. 20 For the hypotheses related to BVA, see Chapter 4.2.1. 176 (22) a. ok Examples A-1 : BVA(tookyooto-sae, soko) Tookyooto-sae-ga rihuzinna yookyuu-o siteiru-no-wa so-ko-no syokuin-ni da Tokyo-even-Nom unreasonable request-Acc doing-that-Top that-place-Gen official-Dat be 'It is to its officials that even the Tokyo metropolitan government is making unreasonable requests.' b. *Examples B-1 : BVA(tookyooto-sae, soko) Tookyooto-ni-sae rihuzinna yookyuu-o siteiru-no-wa so-ko-no syokuin-ga da Tokyo-even-Dat unreasonable request-Acc doing-that-Top that-place-Gen official-Nom be 'It is its officials that are making unreasonable requests to even the Tokyo metropolitan government.' c. ok Examples C-1 (= (22b)): with soko referring to Ministry of Finance already mentioned in the preceding context (23) a. ok Examples A-2 : BVA(kanarinokazuno tihoozititai, soko) Kanarino kazu-no tihoozititai-ga rihuzinna yookyuu-o siteiru-no-wa considerable number-Gen local:government-Nom unreasonable request-Acc doing-that-Top so-ko-no syokuin-ni da that-place-Gen official-Dat be 'It is to its officials that a considerable number of local governments are making unreasonable requests.' b. *Examples B-1 : BVA(kanarinokazuno tihoozititai, soko) Kanarino kazu-no tihoozititai-ni rihuzinna yookyuu-o siteiru-no-wa considerable number-Gen local:government-Dat unreasonable request-Acc doing-that-Top so-ko-no syokuin-ga da that-place-Gen official-Nom be 177 'It is its officials that are making unreasonable requests to a considerable number of local governments.' c. ok Examples C-2 (= (23b)): with soko referring to Ministry of Finance already mentioned in the preceding context (24) a. ok Examples A-3 : BVA(55% izyoono tihoozititai, soko) 55% izyoo-no tihoozititai-ga rihuzinna yookyuu-o siteiru-no-wa 55%:or:more-Gen local:government-Nom unreasonable request-Acc doing-that-Top so-ko-no syokuin-ni da that-place-Gen official-Dat be 'It is to its officials that 55% or more local governments are making unreasonable requests.' b. *Examples B-3 : BVA(55% izyoono tihoozititai, soko) 55% izyoo-no tihoozititai-ni rihuzinna yookyuu-o siteiru-no-wa 55%:or:more-Gen local:government-Dat unreasonable request-Acc doing-that-Top so-ko-no syokuin-ga da that-place-Gen official-Nom be 'It is its officials that are making unreasonable requests to 55% or more local governments.' c. ok Examples C-3 (= (24b)): with soko referring to Ministry of Finance already mentioned in the preceding context Table 5.3 summarizes the result of EPSA[25]-2 on the predicted schematic asymmetry in (21), which has now 12 informants participated. 178 Table 5.3: A summary of the results of EPSA [25]-2 EPSA [25]-2 (Total 12 participants; 158 answers) ... as of Mar/2/2011 QP OK Schema A *Schema B OK Schema C NP-sae 'even NP' 11 values 75 11 values 22 12 values 77 kanarinokazu-no NP 'a considerable number of NPs' 12 values 88 11 values 29 12 values 70 55% izyoo-no NP '55% or more NPs' 12 values 84 11 values 19 12 values 67 Total 35 values 83 33 values 23 36 values 71 The average score on the three *Examples conforming to the *Schema are not particularly low, though the average scores of the corresponding ok Examples are still significantly higher than that of the *Examples. Furthermore, the average score of the ok Examples conforming to the ok Schema C is not very high. At this point, just as we did in the discussion of the result of EPSA[25]-1, let us look a little further into the result of EPSA[25]-2, paying attention to informant classification. If we classify the 12 informants in EPSA[25]-2 into "linguistically-informed" informants and "linguistically-naïve" informants, as described above, and calculate the average score of each group, the result of EPSA[25]-2 in Table 5.3 comes as shown in Table 5.4. Table 5.4: A summary of the results of EPSA [25]-2 by group EPSA [25]-2 (Total 12 participants; 158 answers) ... as of Mar/2/2011 OK Schema A *Schema B OK Schema C Total 35 values 83 33 values 23 36 values 71 linguistically-informed (8 informants) 23 values 92 21 values 32 24 values 93 linguistically-naïve (4 informants) 12 values 65 12 values 8 12 values 29 179 The average score on the *Examples conforming to the *Schema by the four "linguistically-naïve" informants is 8, but the average score of the ok Examples conforming to the ok Schema C is also very low. 21 This means that the examples conforming to the *Schema B and ok Schema C are difficult to parse or unnatural for these three informants. Therefore, we cannot attribute the unacceptability reported by the three informants on the *Examples conforming to the *Schema to the hypothesized property/ies of the CS. On the other hand, the average score on the *Examples conforming to the *Schema by the eight "linguistically-informed" informants is still low, while the average scores of the corresponding ok Examples are significantly higher than that of the *Examples. However, under the formulation of the *Schema-based prediction in (18), any *Examples conforming to a *Schema should be completely unacceptable. Thus, the average score 32 by the eight "linguistically-informed" informants would disconfirm the *Schema-based prediction. In fact among the eight "linguistically-informed" informants three informants found all the *Examples conforming to the *Schema acceptable (with the score of 62 or more). Since the judgments of the four informants disconfirm the *Schema-based prediction, it has to be regarded as a serious problem. Hajime Hoji (pc in March, 2011) suggests one possible account for the acceptability of the *Examples conforming to the *Schema in (21b). He suggests that ni-marked phrase inside the presuppositional clause can be in the Deep DL position. Given this, the LF-structure of the *Schema in (21b) comes as in (25), in which QP-ni is in the Deep DL position. 22 (25) [ QP-ni [ IP Op i [ IP t i V-T]]]-no-wa so-NP-no N i -ga da 21 As pointed out in fn.9 above, when the focus phrase has the nominative case-marker ga, the sequence NP-ga da sounds a little unnatural compare to the other case-markers or postpositions. In EPSA[25]-2, *Schema B and ok Schema C have ga-marked NP in the focus position. The low score of ok Schema C compared to ok Schema A might be due to this unnaturalness. 22 I omit the empty operator movement that corresponds to the Deep-DL. It is not clear how the two empty operator movements interact in such a structure, though. 180 In the LF-structure in (25), the trace of Op is c-commanded by QP-ni, so that BVA(QP, so-NP) is predicted to be possible. Thus, the acceptability of the *Examples conforming to the *Schema in (21b) is accounted for. However, if we do not have ways to force this DL position or to avoid this position, the assumption would only eliminate the *Schema-based prediction without introducing a new *Schema-based prediction. As in Chapter 4.5.1, such a move is a content-reducing (i.e., degenerating) problemshift in the terms of Lakatos (1970/1978). Thus, if we pursue this possibility, we need to articulate the assumption so that it can give rise to new *Schema-based predictions and test the predictions, which I have to leave for my future research. 5.3.5 Kizu’s (2005) analysis revisit We should return here to Kizu’s (2005) analysis of CM-Cleft presented in (3) above. (3) Structure for local cleft constructions: (based on Kizu 2005: 55 (89), 72 (32) and 109 (17a)) TopP CP IP Op i C' I' IP C VP I | … t i … no V' PP V | | P' aru XP P | de At the beginning of the BVA-related discussion in this subsection (Chapter 5.2.1), we have 181 shown that the *Schema-based prediction obtained on the basis of the structure proposed in Kizu 2005 was disconfirmed. However, what if Kizu (2005) also adopted the hypotheses in (12) and (13), that is, what if she also made recourse to chain binding? I repeat the hypotheses in (12) and (13) below. (12) The focus position in CM-Cleft is an A'-position. (13) A single member A'-chain can be composed with another A'-chain, forming a single A'-chain only if the single member A'-chain c-commands the other A'-chain. Under the formulation of the hypothesis on chain binding as in (13), a single member A'-chain needs to c-command another A'-chain in order for the two A'-chains to be composed into one. However, the focus phrase XP, which is now a single member A'-chain, is situated in the complement position of the PP in (3), so that it cannot c-command the A'-chain (Op i , t i ) inside the presuppositional CP. Therefore, an A'-chain (XP i , Op i , t i ) cannot be formed. If Kizu (2005) in fact pursued the same account as taken here based on chain binding, she would need to modify (13) so that chain binding is possible without c-command. As shown in Chapter 3.4.2.1, Kizu (2005) assumes that a predication relation must be established between the focus phrase and the presuppositional CP, and she suggests some mechanisms to establish this without the focus phrase c-commanding the presuppositional CP. I have pointed out some problems of her assumptions there. Basically the same problems arise here, too. Thus, the validity of Kizu’s (2005) account depends on how she could establish the predication relation and here the chain binding without c-command relation. Moreover, this modification would only eliminate the *Schema-based prediction without introducing a new *Schema-based prediction. As 182 mentioned above, such a movement is a content-reducing (i.e., degenerating) problemshift in the terms of Lakatos (1970/1978). Furthermore, it seems difficult for Kizu (2005) to assume the hypothesis in (12), that is, the focus position in CM-Cleft is an A'-position. Under Kizu’s (2005) analysis, the focus phrase is situated in the complement position of the PP. Complement positions are commonly assumed as A-positions, which is incompatible with the hypothesis in (12). In short, Kizu’s (2005) analysis has difficulty in adopting the hypothesis in (12), and even if it did, the formulation of the hypothesis in (13) would make it impossible for Kizu (2005) to make recourse to chain binding as the focus phrase does not c-command elements inside the presuppositional CP. Kizu (2005) might make a modification on (13) so that the chain binding is possible without c-command. The validity of Kizu’s (2005) account would depend on the validity of such a modification. Furthermore, such a modification would only eliminate the *Schema-based prediction without introducing a new *Schema-based prediction, which is a content-reducing (i.e., degenerating) problemshift. 5.4 DR (distributive reading) in the CM-Clefts 5.4.1 Preliminary: DR based on c-command By focusing on the FD-based BVA, which is crucially based on the c-command, we have obtained the schematic asymmetries in (14) and (21), which have so far been confirmed. Under the base-generation analysis proposed here in (2), the two hypotheses in (12) and (13) are crucial to deduce these schematic asymmetries. (12) The focus position in CM-Cleft is an A'-position. 183 (13) A single member A'-chain can be composed with another A'-chain, forming a single A'-chain, only if the single member A'-chain c-command the other A'-chain. Therefore, it is important to show the validity of these two hypotheses. As we have seen in Chapter 4.2.2, scope interpretation between two scope bearing elements, more specifically a wide scope distributive reading of one over the other (i.e., DR), is another phenomenon which is based on the structural relation of c-command as far as we can exclude the "quirky" type of DR. Thus, if the argument constructed in the last section with regard to BVA could be duplicated with regard to DR, we could obtain further support for the two hypotheses in (12) and (13). In this subsection, we will see if we can develop such an argument with regard to DR in the CM-Cleft constructions. The scope interpretation hypothesis formulated in Hoji 2003a is in (26), in which DR(A, B) stands for a wide scope distributive reading of A over B. (26)(= (48) in Chapter 4) (Hoji 2003a: 27, (66)) DR(A, B) is available only if the A-position trace of A c-commands that of B at LF. Recall the pf-LF correspondence assumed in Chapter 4, which is based on Ueyama’s (1998) analysis of scrambling construction in Japanese. 23 23 See Chapter 1.2.1.2 for the detailed analysis of scrambling construction in Japanese proposed in Ueyama 1998. 184 (27) A-ga B-o V corresponds to (i) at LF, while B-o A-ga V can correspond to (i) or (ii) at LF. (i) (ii) A-ga B-o B-o V A-ga ec V Given the pf-LF correspondence in (27), the SO-type construction is unambiguous in regard to DR between A and B, that is, only a wide scope distributive reading of A over B (i.e., DR(A, B)) is possible in the SO-type construction, while the OS-type construction (i.e., scrambling construction) allows two readings, DR(A, B) as well as DR(B, A). As has been shown in Chapter 4.2.2, however, DR(B, A) is available in some cases of A-ga B-o V. (28) is one such example presented in Ueyama 1998. (28) (Ueyama 1998: 41, (47)) SO-type construction: [Dareka-ga [uti-no subete-no sensyu-o] bikoositeiru] (toyuu koto-wa, zen'in-ga someone-NOM our-GEN all-GEN athlete-ACC shadow COMP fact-TOP everyone-NOM kiken-ni sarasareteiru toyuu koto da.) danger-DAT exposed COMP fact COPULA '(The fact that) someone is shadowing every athlete of ours (means that everyone's life is in danger.)' ∃x[∀y(y = athlete)[ x is shadowing y ]] ∀y(y = athlete)[∃x[ x is shadowing y ]] Since the sentence is an SO-type construction, its available LF is only (27)-(i), (the A-position trace of) B can never c-command (the A-position trace of) A at LF. Therefore, DR(B, A) in 185 (28) is unexpected under the scope interpretation hypothesis in (26). We have seen that the availability of this "quirky" type of DR is closely related to the type of QPs used as B of DR(B, A). The results of the EPSA[13]-1 and 4 conducted by Emi Mukai show that it is difficult to obtain the "quirky" type of DR with the QPs listed in (29), while such is not the case with the ones listed in (29b). 24 (29) a. sukunakutomo 4-ttu-no NP 'at least four NPs' NP-sae 'even NP' 55% izyoo-no NP '55% or more NPs' 2-wari izyoo-no NP '20% or more NPs' kanarinokazu-no NP 'a good number of NPs' b. subete-no NP 'every NP' zen-in 'everyone' 4-ttu-no NP 'four NPs' dare-mo 'everybody' This is reminiscent of the quirky binding of BVA, in which the QPs in (29b) that can give rise to quirky binding are classified into the QPs that refer to a specific group of individuals. Thus, by using QPs such as those listed in (29a), which cannot (or at least cannot easily) refer to a specific group of individuals as an object NP, we can maintain that in the SO-type construction of the form A-ga B-o V, DR(A, B) is possible, but DR(B, A) is not. We can thus modify the scope interpretation hypothesis in (26) as in (30). 24 See Chapter 4.2.2.1 for the detailed discussion and the experimental results. 186 (30) DR(A, B) is available only if (i) the A-position trace of A c-commands that of B at LF, or (ii) the condition on the "quirky" type of DR is satisfied for A. (31) The condition on the "quirky" type of DR: The QP A of DR(A, B) must refer to a specific group of individuals. 5.4.2 QP in the focus position as A of DR(A, B) 5.4.2.1 Experimental design Combining hypotheses on DR in (30) and (31) with the pf-LF correspondence in CM-Cleft in (32) as well as the hypotheses in (12) and (13), we can obtain a predicted schematic asymmetry in (33) in regard to scope interpretation between a QP in the focus position and one inside the presuppositional CP in CM-Cleft. (32) pf-LF correspondence: pf: [… QP-CM … V-T]-no-wa QP-CM da LF (before QR): IP VP I SC V | CP QP 1 -CM da IP C | Op 1 IP no-wa QP…t 1 … V-T (12) The focus position in CM-Cleft is an A'-position. 187 (13) A single member A'-chain can be composed with another A'-chain, forming a single A'-chain, only if the single member A'-chain c-command the other A'-chain. (33) Schema group: a. ok Schema A [#-CL-no N]-o V-T-no-wa A-ga da DR(A, #-CL-no N) b. *Schema B [#-CL-no N]-ga V-T-no-wa A-o da DR(A, #-CL-no N) c. ok Schema C [#-CL-no N]-ga V-T-no-wa A-o da DR(#-CL-no N, A) In (33a), A is in the focus position, where it can c-command [#-CL-no N] inside the presuppositional CP. Thus, it is predicted that DR(A, #-CL-no N) is possible. However, the focus position is in fact an A'-position, so that A cannot be interpreted in the focus position. Given the pf-LF correspondence in (32), the LF structures of (33a) and (33b) (= (33c)) come as in (34a) and (34b), respectively. (34) a. [ SC [ CP [ IP Op i [ IP t i [#-CL-no N]-o ... V-T]]-no-wa] A i -ga] da b. [ SC [ CP [ IP Op i [ IP [#-CL-no N]-ga t i ... V-T]]-no-wa] A i -o] da Under the hypothesis in (13), then, A in an A'-position can be combined with the chain (Op i , t i ) inside the presuppositional CP forming an A'-chain (A i , Op i , t i ). As a result, A is interpreted at the position of t i , which is an A-position. In (34a), the position of t i c-commands [#-CL-no N], so that DR(A, #-CL-no N) is still predicted to be possible. 188 In (33b), too, A cannot be interpreted in the focus position. A is combined with the chain (Op i , t i ) inside the presuppositional CP forming an A'-chain (A i , Op i , t i ) and gets interpreted at the position of t i . However, in (34b) the position of t i does not c-commands [#-CL-no N], so that DR(A, #-CL-no N) is predicted to be impossible as long as we use as A a QP such as those listed in (29a). The position of t i in (34b) is instead c-commanded by [#-CL-no N], so that DR(#-CL-no N, A) is predicted to be possible, making (33c) an ok Schema C . The predicted schematic asymmetry in (33) is what is tested in EPSA[25]-4. Three QPs from (29a) are used to construct the examples: Nagasakisi-sae 'even Nagasaki City', kanarinokazu-no tihoozititai 'a considerable number of local governments' and 55% izyoo-no tihoozititai '55% or more local governments'. Thus, a total of 9 examples are tested, which are shown in (35)-(37). (35) (A of DR(A, B) = NP-sae) a. ok Example A-1 : DR(Nagasaki-si-sae, san-nin-no seezika) San-nin-no seezika-o hihansita-no-wa Nagasaki-si-sae-ga da 3-CL-GEN politician-ACC criticized-that-Top Nagasaki-city-even-NOM be 'It was even Nagasaki City that criticized three politicians.' b. *Example B-1 : DR(Nagasaki-si-sae, san-nin-no seezika) San-nin-no seezika-ga hihansita-no-wa Nagasakisi-sae-o da 3-CL-GEN politician-NOM criticized-that-Top Nagasaki:city-even-ACC be 'It was even Nagasaki City that three politicians criticized.' c. ok Example C-1 (= (35b)): DR(san-nin-no seezika, Nagasaki-si-sae) 189 (36) (A of DR(A, B) = kanarinokazu-no NP) a. ok Example A-2 : DR(kanarinokazu-no tihoozititai, san-nin-no seezika) San-nin-no seezika-o hihansita-no-wa kanarinokazu-no tihoozititai-ga da 3-CL-GEN politician-ACC criticized-that-Top a:considerable:number-of local government-NOM be 'It was a considerable number of local governments that criticized three politicians.' b. *Example B-2 : DR(kanarinokazu-no tihoozititai, san-nin-no seezika) San-nin-no seezika-ga hihansita-no-wa kanarinokazu-no tihoozititai-o da 3-CL-GEN politician-NOM criticized-that-Top a:considerable:number-of local government-ACC be 'It was a considerable number of local governments that three politicians criticized.' c. ok Example C-2 (=(36b)): DR(san-nin-no seezika, kanarinokazu-no tihoozititai) (37) (A of DR(A, B) = 55% izyoo-no NP) a. ok Example A-3 : DR(55% izyoo-no tihoozititai, san-nin-no seezika) San-nin-no seezika-o hihansita-no-wa 55% izyoo-no tihoozititai-ga da 3-CL-GEN politician-ACC criticized-that-Top 55%:or:more-of local government-NOM be 'It was 55% or more of the local governments that criticized three politicians.' b. *Example B-3 : DR(55% izyoo-no tihoozititai, san-nin-no seezika) San-nin-no seezika-ga hihansita-no-wa 55% izyoo-no tihoozititai-o da 3-CL-GEN politician-NOM criticized-that-Top 55%:or:more-of local government-ACC be 'It was 55% or more of the local governments that three politicians criticized.' c. ok Example C-3 (= (37b)): DR(san-nin-no seezika, 55% izyoo-no tihoozititai) 5.4.2.2 Results of the experiment Table 5.5 summarizes the result of EPSA[25]-4 on the predicted schematic asymmetry in (33), which has now 12 informants participated. 190 Table 5.5: A summary of the results of EPSA [25]-4 EPSA [25]-4 (Total 12 participants; 133 answers) ... as of Mar/2/2011 QP OK Schema A *Schema B OK Schema C NP-sae 'even NP' 12 values 77 12 values 27 12 values 77 kanarinokazu-no NP 'a considerable number of NPs' 11 values 72 12 values 29 12 values 77 55% izyoo-no NP '55% or more NPs' 12 values 79 11 values 22 12 values 66 Total 35 values 76 35 values 26 36 values 73 The average score on the three *Examples conforming to the *Schema are not particularly low, though the average scores of the corresponding ok Examples are still significantly higher than that of the *Examples. Furthermore, the average score of the ok Examples conforming to the ok Schema C is not very high. At this point, just as we did in the discussion of the result of EPSA[25]-1 and 2, let us look a little further into the result of EPSA[25]-4 by classifying the informants. If we classify the 12 informants in EPSA[25]-4 into "linguistically-informed" informants and "linguistically-naïve" informants and calculate the average score of each group, the result of EPSA[25]-4 in Table 5.5 comes as shown in Table 5.6. Table 5.6: A summary of the results of EPSA [25]-4 by group EPSA [25]-4 (Total 12 participants; 133 answers) ... as of Mar/2/2011 OK Schema A *Schema B OK Schema C Total 35 values 76 35 values 26 36 values 73 linguistically-informed (8 informants) 23 values 100 23 values 33 24 values 95 linguistically-naïve (4 informants) 12 values 31 12 values 12 12 values 29 191 The average score on the *Examples conforming to the *Schema by the three "linguistically-naïve" informants is 12, but the average score of the ok Examples conforming to the ok Schema C is also very low. 25 This means that the examples conforming to the *Schema B and ok Schema C are difficult to parse or unnatural for these three informants. Therefore, we cannot attribute the unacceptability reported by the three informants on the *Examples conforming to the *Schema to the hypothesized property/ies of the CS. On the other hand, the average score on the *Examples conforming to the *Schema by the eight "linguistically-informed" informants is still low, while the average scores of the corresponding ok Examples are significantly higher than that of the *Examples. However, the average score 33 by the eight "linguistically-informed" informants is too high and it would disconfirm the *Schema-based prediction. It should be noted, however, that among the eight "linguistically-informed" informants two informant found all the *Examples conforming to the *Schema totally acceptable and one found one of the three *Examples totally acceptable, while the other five find all the *Examples quite difficult (with the score of at most 25). Since the total number of informants is still small, we need to wait to see if the judgments of these two informants can be regarded as an anomaly or has to be taken into a serious consideration. 5.4.3 QP in the presuppositional CP as A of DR(A, B) 5.4.3.1 Experimental design Next consider the case in which A of DR(A, B) is inside the presuppositional CP and B is in the focus position. This is analogous to the BVA case discussed in Chapter 5.2.4, in which A of BVA(A, B) is inside the presuppositional CP and the dependent term B is in the focus position. 25 See fn.9 above for the possible unnaturalness caused by the sequence NP-ga da. In EPSA[25]-4, ok Schema A has this sequence. 192 Again, combining hypotheses on DR in (30) and (31) with the pf-LF correspondence in CM-Cleft in (32) as well as the hypotheses in (12) and (13), we can obtain another predicted schematic asymmetry, as given in (38). (38) Schema group: a. ok Schema A A-ga V-T-no-wa [#-CL-no N]-o da DR(A, #-CL-no N) b. *Schema B A-o V-T-no-wa [#-CL-no N]-ga da DR(A, #-CL-no N) c. ok Schema C A-o V-T-no-wa [#-CL-no N]-ga da DR(#-CL-no N, A) Given the pf-LF correspondence in (32), the LF structures of (38a) and (38b) (= (38c)) come as in (39a) and (39b), respectively. (39) a. [ SC [ CP [ IP Op i [ IP A-ga t i ... V-T]]-no-wa] [#-CL-no N] i -o] da b. [ SC [ CP [ IP Op i [ IP t i A-o ... V-T]]-no-wa] [#-CL-no N] i -ga] da In (39a), [#-CL-no N]-o is in the focus position. Since the focus position is an A'-position, [#-CL-no N] cannot be interpreted in that position. [#-CL-no N] is combined with the chain (Op i , t i ) inside the presuppositional CP forming an A'-chain ([#-CL-no N] i , Op i , t i ) and gets interpreted at the position of t i . The position of t i in (39a) is c-commanded by A, so that DR(A, #-CL-no N) is predicted to be possible, giving rise to the ok Schema A in (38a). On the other hand, in (39b) the position of t i of the A'-chain ([#-CL-no N] i , Op i , t i ) is not 193 c-commanded by A, so that DR(A, #-CL-no N) is predicted to be impossible, obtaining the *Schema B in (38b). However, in (39b) the position of t i in (39b) c-command A, so that DR(#-CL-no N, A) is predicted to be possible, yielding the ok Schema C in (38c). The predicted schematic asymmetry in (38) is what is tested in EPSA[25]-5. Just like EPSA[25]-4, three QPs from (29a) are used to construct the examples: Nagasakisi-sae 'even Nagasaki City', kanarinokazu-no tihoozititai 'a considerable number of local governments' and 55% izyoo-no tihoozititai '55% or more local governments'. Thus, a total of 9 examples are tested, which are shown in (40)-(41). (40) (A of DR(A, B) = NP-sae) a. ok Example A-1 : DR(Nagasaki-si-sae, san-nin-no seezika) Nagasaki-si-sae-ga hihansita-no-wa san-nin-no seezika-o da Nagasaki-city-even-NOM criticized-that-Top 3-CL-GEN politician-ACC be 'It was three politicians that even Nagasaki City criticized.' b. *Example B-1 : DR(Nagasaki-si-sae, san-nin-no seezika) Nagasaki-si-sae-o hihansita-no-wa san-nin-no seezika-ga da Nagasaki-city-even-ACC criticized-that-Top 3-CL-GEN politician-NOM be 'It was three politicians that criticized even Nagasaki City.' c. ok Example C-1 (= (40b)): DR(san-nin-no seezika, Nagasaki-si-sae) (41) (A of DR(A, B) = kanarinokazu-no NP) a. ok Example A-2 : DR(kanarinokazu-no tihoozititai, san-nin-no seezika) Kanarinokazu-no tihoozititai-ga hihansita-no-wa san-nin-no seezika-o da a:considerable:number-of local government-NOM criticized-that-Top 3-CL-GEN politician-ACC be 194 'It was three politicians that a considerable number of local governments criticized.' b. *Example B-2 : DR(kanarinokazu-no tihoozititai, san-nin-no seezika) Kanarinokazu-no tihoozititai-o hihansita-no-wa san-nin-no seezika-ga da a:considerable:number-of local government-ACC criticized-that-Top 3-CL-GEN politician-NOM be 'It was three politicians that criticized a considerable number of local governments.' c. ok Example C-2 (= (41b)): DR(san-nin-no seezika, kanarinokazu-no tihoozititai) (42) (A of DR(A, B) = 55% izyoo-no NP) a. ok Example A-3 : DR(55% izyoo-no tihoozititai, san-nin-no seezika) 55% izyoo-no tihoozititai-ga hihansita-no-wa san-nin-no seezika-o da 55%:or:more-of local government-NOM criticized-that-Top 3-CL-GEN politician-ACC be 'It was three politicians that 55% or more of the local governments criticized.' b. *Example B-3 : DR(55% izyoo-no tihoozititai, san-nin-no seezika) 55% izyoo-no tihoozititai-o hihansita-no-wa san-nin-no seezika-ga da 55%:or:more-of local government-ACC criticized-that-Top 3-CL-GEN politician- NOM be 'It was three politicians that criticized 55% or more of the local governments.' c. ok Example C-3 (= (42b)): DR(san-nin-no seezika, 55% izyoo-no tihoozititai) 5.4.3.2 Results of the experiment Table 5.7 summarizes the result of EPSA[25]-5 on the predicted schematic asymmetry in Table 5.5), which has now 12 informants participated. Basically the same judgment pattern as in EPSA[25]-4, which is the case where the QP in the focus position is A of DR(A, B), has been obtained so far. 195 Table 5.7: A summary of the results of EPSA [25]-5 EPSA [25]-5 (Total 12 participants; 125 answers) ... as of Mar/2/2011 QP OK Schema A *Schema B OK Schema C NP-sae 'even NP' 12 values 58 12 values 11 12 values 79 kanarinokazu-no NP 'a considerable number of NPs' 12 values 68 12 values 19 12 values 70 55% izyoo-no NP '55% or more NPs' 12 values 67 11 values 27 12 values 79 Total 36 values 64 35 values 19 36 values 76 The average score on the three *Examples conforming to the *Schema are not particularly low, though the average scores of the corresponding ok Examples are still significantly higher than that of the *Examples. 26 Furthermore, the average score of the ok Examples conforming to the ok Schema C is not very high. At this point, just as we did in the above EPSA, let us look a little further into the result of EPSA[25]-5 by classifying the informants. If we classify the 12 informants in EPSA[25]-5 into "linguistically-informed" informants and "linguistically-naïve" informants and calculate the average score of each group, the result of EPSA[25]-4 in table 5.7 comes as shown in Table 5.8. 26 See fn.9 above for the possible unnaturalness caused by the sequence NP-ga da. In EPSA[25]-5, *Schema B and ok Schema C have this sequence. 196 Table 5.8: A summary of the results of EPSA [25]-5 by group EPSA [25]-5 (Total 12 participants; 125 answers) ... as of Mar/2/2011 OK Schema A *Schema B OK Schema C Total 36 values 64 35 values 19 36 values 76 linguistically-informed (8 informants) 24 values 85 23 values 29 24 values 95 linguistically-naïve (4 informants) 12 values 22 12 values 0 12 values 37 The average score on the *Examples conforming to the *Schema by the four "linguistically-naïve" informants is 0, but the average score of the ok Examples conforming to the ok Schema C is also very low. This means that the examples conforming to the *Schema B and ok Schema C are difficult to parse or unnatural for these three informants. Therefore, we cannot attribute the unacceptability reported by the three informants on the *Examples conforming to the *Schema to the hypothesized property/ies of the CS. On the other hand, the average score on the *Examples conforming to the *Schema by the eight "linguistically-informed" informants is still low, while the average scores of the corresponding ok Examples are significantly higher than that of the *Examples. However, the average score 29 by the eight "linguistically-informed" informants is too high and it would disconfirm the *Schema-based prediction. It should be noted, however, that among the eight "linguistically-informed" informants two informant found two of the *Examples conforming to the *Schema totally acceptable and two found one or two of the three *Examples fairly acceptable with the score between 50 and 75, while the other informants find all the *Examples quite difficult (with the score of at most 25). Since the total number of informants is still small, we need to wait to see if the judgments of these three informants can be regarded as an anomaly or has to be taken into a serious consideration. 197 5.4.4 Summary of the DR-based argument The base-generation analysis proposed in this work has introduced two hypotheses in (12) and (13). The last section has shown that the predicted schematic asymmetries in regard to BVA, which were obtained based on these two hypotheses and the pf-LF correspondence proposed in this work, have so far been confirmed. Since the account of the BVA facts crucially depends on the two hypotheses, it is important to show their validity. Since DR is another phenomenon which is based on the structural relation of c-command as far as we can exclude the "quirky" type of DR, if the argument constructed in the last section with regard to BVA could be duplicated with regard to DR, we could obtain further support for the two hypotheses in (12) and (13). The results of the experiments, however, are not quite in harmony with the predicted schematic asymmetries obtained for DR. Crucially, the scores on the *Examples conforming to the *Schemata are not low enough to survive the disconfirmation attempts. Since the total number of informants is still small, it would be too early to conclude that the *Schema-based predictions were disconfirmed. We need to collect more informant judgments to see the real picture and if the*Schema-based predictions will in fact be disconfirmed, we would have to articulate the experimental design of the DR-related experiments or modify the hypotheses. I need to leave it for my future research, but at least we cannot get strong support for the introduction of the two hypotheses in (12) and (13) from the DR-based argument. 5.5 What the results tell about the structure of the CM-cleft construction? The schematic asymmetries obtained in regard to BVA in CM-Cleft are (14) and (21). 198 (14) Schema group: a. ok Schema A [… so-NP-no NP-ni … V-T]-no-wa QP-ga da BVA(QP, so-NP) b. *Schema B [… so-NP-no NP-ga … V-T]-no-wa QP-ni da BVA(QP, so-NP) c. ok Schema C [… so-NP-no NP-ga … V-T]-no-wa QP-ni da (With so-NP referring to a particular individual/object) (21) Schema group: a. ok Schema A [… QP-ga … V-T]-no-wa so-NP-no NP-ni da BVA(QP, so-NP) b. *Schema B [… QP-ni … V-T]-no-wa so-NP-no NP-ga da BVA(QP, so-NP) c. ok Schema C [… QP-ni … V-T]-no-wa so-NP-no NP-ga da (With so-NP referring to a particular individual/object) These schematic asymmetries are deduced from the hypotheses about BVA, the pf-LF correspondence of CM-Cleft in (43) based on the proposed structure in this work, and crucially the two hypotheses in (12) and (13), which are repeated below. (43) pf-LF correspondence: 27 pf: [… NP-CM … V-T]-no-wa NP-CM da 27 The position of the trace of Op (t 1 ) in (43) is just one instance and the position is the canonical argument position of V, which is related to the noun in the focus position. 199 LF (before QR): IP VP I SC V | CP NP 1 -CM da IP C | Op 1 IP no-wa NP-CM…t 1 … V-T (12) The focus position in CM-Cleft is an A'-position. (13) A single member A'-chain can be composed with another A'-chain, forming a single A'-chain, only if the single member A'-chain c-commands the other A'-chain. The validity of the BVA-related hypotheses has been shown in the preceding chapter. Thus, the confirmation of the predicted schematic asymmetries in (14) and (21) gives support to the proposed structure of CM-Cleft as well as the hypotheses in (12) and (13). Although the number of informants is still small, the predicted schematic asymmetries in (14) and (21) have so far been confirmed, which is in favor of the proposed base-generation analysis. However, it should be noted here that as a result of the introduction of the two hypotheses in (12) and (13), the focus phrase in CM-Cleft is interpreted not in the focus position but in the trace position of the empty operator. Under the movement analysis, this position is where the focus phrase is base-generated and gets interpreted after the reconstruction at LF. 28 Therefore, 28 But not under Hiraiwa & Ishihara’s (2002) analysis with the assumption that the focus movement does not reconstruct. As shown in Chapter 2.4.1.1.2, H&I (2002:44) point out that "focus movement in the derivation of Cleft does not exhibit the 'radical reconstruction' property, which is typical of LDS [long-distance scrambling, YM]." The introduction of this new type of A'-movement that does not 200 as far as the BVA facts (and probably the DR facts) of CM-Cleft presented in this chapter are concerned, the movement analysis makes the same predictions as the base-generation analysis proposed here. Thus, we cannot discriminate between the movement analysis and the base-generation analysis in these respects. In order to argue for the base-generation analysis pursued in this work, then, we need to have phenomena about which the movement analysis and the base-generation analysis make different predictions and show that the predicted schematic asymmetries obtained under the base-generation analysis get confirmed, while the ones under movement analysis do not. We will discuss such phenomena in the next chapter. 5.6 Summary In this chapter, I have first proposed the structure of CM-Cleft pursued in this work, which is one of the base-generation analyses based on Hoji (1987, 1990). The analysis proposed here differs from Kizu’s (2005) base-generation analysis in the position of the focus phrase. More specifically, in the former the focus position c-commands elements inside the presuppositional CP, while in the latter it does not. FD-based BVA is crucially based on c-command. Thus, by checking the possibility of FD-based BVA in CM-Cleft, we can distinguish between the two analyses. In the discussion of BVA in CM-Cleft, I have shown that a *Schema-based prediction under Kizu’s (2005) analysis of CM-Cleft is disconfirmed. The *Schema under Kizu’s (2005) analysis is an ok Schema under the proposed analysis, so that the BVA fact that disconfirms the *Schema-based prediction under Kizu 2005 does not pose a problem to the proposed analysis. However, a *Schema-based prediction under the proposed analysis has been shown to be reconstruct, however, will face difficulty in accounting for the BVA facts provided in this chapter. 201 disconfirmed by another BVA fact. In order to account for that, I have introduced two hypotheses; (i) the focus position in CM-Cleft is an A'-position, and (ii) a phrase in the focus position, a single member A'-chain, can be composed with another A'-chain, forming a single A'-chain, and is interpreted at the A-position of the new chain, that is, at the tail of the chain. The predicted schematic asymmetries in (14) and (21) were obtained based on the modified analysis and have been tested in EPSA[25]-1 and EPSA[25]-2, respectively. Although the number of informants in each EPSA is still small, the results so far seem to be basically in harmony with the predicted schematic asymmetries, and thus give support to the proposed analysis. DR is another relation which is crucially based on c-command if we carefully exclude the "quirky" type of DR. Thus, if the argument constructed with regard to BVA could be duplicated with regard to DR, we could obtain further support for the two hypotheses in (12) and (13). Based on the two hypotheses and the pf-LF correspondence proposed in this work as well as DR-related hypotheses, we have obtained the predicted schematic asymmetries in (33) and (38), which have been tested in EPSA[25]-4 and EPSA[25]-5, respectively. The results of the experiments, however, are not quite in harmony with the predicted schematic asymmetries. Crucially, the scores on the *Examples conforming to the *Schemata are not low enough to survive the disconfirmation attempts. Since the total number of informants is still small, it would be too early to conclude that the *Schema-based predictions were disconfirmed. However, at least we cannot (yet) get strong support for the introduction of the two hypotheses in (12) and (13) from the DR-based argument. 202 Chapter 6: Base-generation analysis over Movement analysis 6.1 Introduction In the preceding chapter I have proposed the base-generation analysis, which is based on Hoji 1987 and 1990. In the proposed analysis, a CM-Cleft sentence of the form in (1a) corresponds to the LF structure in (1b). (1) pf-LF correspondence: a. pf: [… NP-CM … V-T]-no-wa NP-CM da b. LF (before QR): TP VP T SC V | CP NP 1 -CM da TP C | Op 1 TP no-wa NP-CM…t 1 … V-T In the LF-structure in (1b), the focus phrase NP 1 -CM forms a small clause (SC in (1)) with the presuppositional CP. Given the LF-structure in (1b), the focus phrase c-commands elements inside the presuppositional CP at LF. I have demonstrated on the basis of the experiments that this pf-LF correspondence together with the two hypotheses in (2) and (3) can account for the BVA facts and DR facts in CM-Cleft. (2) The focus position in CM-Cleft is an A'-position. 203 (3) A single member A'-chain can be composed with another A'-chain, forming a single A'-chain, only if the single member A'-chain c-command the other A'-chain. However, as I mentioned at the end of the preceding chapter, as a result of the introduction of the two hypotheses in (2) and (3), the focus phrase in CM-Cleft is interpreted not in the focus position but in the trace position of the empty operator movement. Under the movement analysis, this position is where the focus phrase is base-generated and gets interpreted after the reconstruction is applied at LF. 1 Recall the movement analysis proposed by Takahashi (2006), which was summarized in Chapter 2.3.2. The proposed derivation and structures are repeated below in (4) and (5), respectively. (4) (Takahashi 2006: 5 (12)) a. [ TP [ FP [ VP [ CP [ TP Taroo-ga Hanako-ni atta] no C ]-wa da V ] Focus] T] b. [ TP [ FP Hanako-ni [ F' [ FP [ CP t' [ C' [ TP Taroo-ga t atta] no C ]-wa da V ] Focus]] T] c. [ TP [ CP t' [ C' [ TP Taroo-ga t atta] no C ]-wa [ T' [ FP Hanako-ni [ F' [ FP t CP da V ] Focus]] T] 1 See Chapter 5, fn.20 for a different view in the movement analysis. 204 (5) Structure of (4c) a. Surface b. LF (after reconstruction) TP TP CP T' CP T' t 1 C' FP T TP C FP T | TP C Hanako-ni 1 F' no-wa F' | Taroo-ga Hanako-ni atta Taroo-ga t 1 atta no-wa VP Foc VP Foc t CP V t CP V | | da da Under Takahashi’s (2006) analysis, the focus movement of Hanako-ni in (4b) is A'-movement to [Spec,FP], so that the focus phrase is reconstructed to the original A-position inside the presuppositional CP at LF as shown in (5b). Therefore, the focus phrase gets interpreted at its original position inside the presuppositional CP at LF for BVA and DR interpretation. As a result, as far as the BVA facts and the DR facts of CM-Cleft presented in the preceding chapter are concerned, the movement analysis makes the same predictions as the base-generation analysis proposed in this work. Thus, we cannot discriminate between the movement analysis and the base-generation analysis in these respects. In order to argue for the base-generation analysis pursued in this work, then, we need to have phenomena about which the movement analysis and the base-generation analysis make different predictions and show that the predicted schematic asymmetries obtained under the base-generation analysis get confirmed, while the ones under movement analysis do not. In this chapter, I will discuss two such phenomena, NSE (Neg(ation)-sensitive elements) "licensing" and possibility of resumption in CM-Cleft, and show that the base-generation 205 analysis proposed in this work, but not the movement analysis, can account for these phenomena. 6.2 Where do the differences arise between the two analyses? The most crucial difference between the base-generation analysis and the movement analysis is the position of the focus phrase at LF. Under the base-generation analysis proposed in this work, the focus phrase is interpreted at the trace position of the empty operator Op inside the presuppositional CP as a result of the chain formation between the single member A'-chain consisting of the focus phrase and the A'-chain of the empty operator movement (Op, t). However, as the LF-structure of CM-Cleft in (1b) shows, the focus phrase is still in the focus position at LF. This focus phrase forms a small clause (SC in (1b)) with the presuppositional CP, so that it is outside the domain of the presuppositional CP. On the other hand, under the movement analysis proposed by Takahashi (2005), the focus phrase is reconstructed to the original position at LF, which is inside the presuppositional CP. The focus phrase is thus inside the presuppositional CP at LF. If there are some phenomena that crucially refer to this structural difference of the focus position between the base-generation analysis and the movement analysis, the two analyses will make different predictions on the acceptability of the relevant CM-Cleft examples. NSE "licensing" is one of such phenomena. I have summarized in Chapter 4.3 Kataoka’s (2006, 2007) discussion about the NSEs in Japanese, especially about rokuna-N 'good/decent N'. One of the main claims in Kataoka 2006, 2007 is that rokuna-N 'good/decent N' does, but XP-sika 'all but XP' and dare-mo (indefinite person+suffix -mo)/nani-mo (indefinite thing+suffix -mo) 'nobody/nothing' do not obey the condition in (6), which has been generally agreed upon in the literature as the condition on 206 NSEs in English. (6) (Kataoka 2007: 2, (4)) An NSE must be c-commanded by Neg at LF. (cf. Klima 1964) In regard to the position of the sentential negative element Neg (-nai), Kataoka (2007) assumes (7). (7) (Kataoka 2007: 6, (11)) Sentential negative element Neg (-nai) is in the sister position of the VP at LF. (e.g., Masuoka 1989, Pollock 1989) As one of the arguments for her claim that rokuna-N 'good/decent N' obeys the condition in (6), Kataoka (2006) presents the following asymmetry in acceptability of rokuna-N between (8) and (9). (8) (Kataoka 2006: 105, (27)-(30)) a. * Rokuna hito-ga [Taroo-ga manga-o yoma-nai]-to it-ta. good person-NOM Taroo-NOM comic-ACC read-Neg-that say-Past '(Lit.) Good persons said that Taroo does not read comics.' b.* Rokuna gakusee-ga [Yamada-sensee-ga syukudai-o dasa-nai]-riyuu-o tazune-ta. good student-NOM Yamada-teacher-NOM homework-ACC assign-Neg-reason-ACC ask-Past '(Lit.) Good students ask the reason why Mr.Yamada does not assign homework.' 207 c.* [Hon-o yoma-nai]-gakusei-ga, rokuna tooan-o kaku. book-ACC read-Neg student-NOM good answer-ACC write '(Lit.) Students that do not read books write good answers.' d.* Konkaino goodoo setumeekai-ni syatyoo-ga syussekisi-na-kereba, this:time joint recruiting:fair-to president-NOM attend-Neg-if rokuna gakusee-ga kuru. good student-NOM come '(Lit.) If the president does not attend the next joint recruiting fair, good students will come.' (9) (Kataoka 2006: 105, (31)) a. ? Boku-ni-wa [Yamada-ga rokuna ronbun-o kaku]-to omo-e-nai. me-to-TOP Yamada-NOM good paper-ACC write-that think-POT-Neg '(Lit.) To me, it cannot be thought that Yamada writes good paper.' b. ? Kono-kurasu-ni-wa [sensee-ni rokuna koto-o iu]-yatu-ga i-nai. this-class-in-TOP teacher-to good thing-ACC tell person-NOM be-Neg '(Lit.) There is no one in this class that tells good things to his/her teacher.' In (8) rokuna-N is in the main clause and Neg is in the embedded clause. The sentences are all unacceptable. On the other hand, in (9) rokuna-N is in the embedded clause and Neg is in the main clause. The sentences are acceptable though they are a little degraded. Thus, the generalization is that when rokuna-N is in the main clause and Neg is in the embedded clause, sentences are never acceptable, while when Neg is in the main clause and rokuna-N is in the embedded clause, sentences can be acceptable. In the latter case, Neg in the main clause c-commands rokuna-N in the embedded clause. Based on these facts Kataoka 208 (2006) argues that rokuna-N obeys the condition in (6). In CM-Cleft, both the main copula verb da/desu and the verb in the presuppositional CP can take Neg as shown in (10a) and (10b), respectively. (10) a. Taroo-ga kinoo atta no-wa Hanako-ni zya-nakat-ta. 2 Taroo-Nom yesterday met that-Top Hanako-Dat be-Neg-Past 'It was not Hanako that Taroo met yesterday.' b. Taroo-ga kinoo awa-nakat-ta no-wa Hanako-ni da/desu. Taroo-Nom yesterday meet-Neg-Past that-Top Hanako-Dat is 'It was Hanako that Taroo did not meet yesterday.' Under the proposed LF-structure of CM-Cleft in (1b), together with the assumption on the position of Neg in (7), the focus phrase in CM-Cleft is c-commanded by Neg in the main clause at LF, but not by Neg in the presuppositional CP. On the other hand, under the LF-structure in (5b) proposed by Takahashi (2006), the focus phrase in CM-Cleft is c-commanded by Neg in the presuppositional CP at LF, but not by Neg in the main clause as a result of the reconstruction. This difference of the position of the focus phrase at LF makes different predictions on the possibility of NSEs between the base-generation analysis and the movement analysis. 2 In (10a) the main copula verb is expressed as zya. This is in fact the contracted form of the copula verb de + a restrictive particle wa, and has a colloquial usage. Although the de + Neg nai sequence is possible, it sound less natural than the dewa + nai sequence, especially when it comes at the end of the sentence. Moreover, such a context as in (10a), colloquial expressions sound more natural. Though there is some difference in meaning with or without the restrictive particle wa, I use zya as a main copula verb coming before Neg nai here as well as in the EPSA experiment to be presented, leaving the existence of the restrictive particle wa in zya aside. 209 6.3 NSEs in CM-Cleft under the movement analysis 6.3.1 Predictions in the movement analysis: Takahashi 2006 I first discuss what predictions the movement analysis proposed by Takahashi (2006) would make in regard to the possibility of NSEs in CM-Cleft. Given the position of Neg in (7), the LF-structure of CM-Cleft in Takahashi 2006 in (5b) comes as in (11) with the positions of Neg. 3,4 (11) LF-structure of CM-Cleft (cf. (5b)) TP CP T' t 1 C' FP T TP C t 1 F' | NegP T no-wa NegP Foc | VP Neg ta VP Neg t CP V Taroo-ga Hanako-ni 1 aw | da reconstruction When the focus phrase is the NSE rokuna-N 'good/decent N', it is c-commanded by Neg in the 3 I omit the specifier position of NegP in (11) and the subsequent discussion just for the ease of exposition. It is not the case that NegP cannot have the specifier position. 4 I assume here without discussion that the Focus Phrase (i.e., FP) projects above NegP. Takahashi (2006) does not discuss the examples with negation, so it is not clear how Takahashi (2006) analyzes the position of NegP. It would be possible that Takahashi (2006) assumes that NegP projects above FP. However, there is no difference in the c-command relation between Neg and the focus phrase depending on this positional difference as long as the focus phase is asssumed to be reconstructed inside the presuppositional CP. I will discuss below the possibility that the focus movement is not reconstructed in the movement analysis. 210 presuppositional CP, but not by Neg in the main clause at LF in (11). Given the condition on rokuna-N proposed in Kataoka 2007 presented in Chapter 4.3, which is repeated here as (12), we can obtain a *Schema in (13) under Takahashi’s (2006) movement analysis. 5 (12) (Kataoka 2007: 16, (43)) Rokuna-N must be c-commanded by Neg at LF. (13) *Schema: [… … V-T]-no-wa rokuna-N-CM zya-nai-T Since the case-marker on rokuna-N in the focus position is not specified in (13), it can be the subject, accusative object, dative object or postpositional phrase. Thus, one of the *Examples conforming to the *Schema in (13) is shown in (14), in which rokuna-N in the focus position is the dative object with the dative case-marker ni. (14) *Example: Sono paatii-de Hanako-ga koeokaketa-no-wa rokuna otoko-ni zya-nak-katta. that party-at Hanako-Nom spoke:to-that-Top decent man-Dat be-Neg-Past '(Lit.) It wasn’t decent men that Hanako spoke to at that party.' The example in (14) and three other examples conforming to the schema in (13) are included in 5 As we will see shortly, the *Schema in (13) is the ok Schema in EPSA[25]-3, which I have been conducting. Thus, the *Example in (14) is one of the ok Examples in EPSA[25]-3. 211 EPSA[25]-3. 6 The average score of the judgments of eight informants is 50 on the scale of 0-100, with 0 being completely unacceptable and 100 being fully acceptable. As we have seen in Chapter 3.3.3, *Examples conforming to a *Schema should be completely unacceptable. Therefore, this result disconfirms the*Schema-based prediction in (13), which was made on the basis of the structure of CM-Cleft in (11) under Takahashi’s (2005) movement analysis. As noted in Chapter 3.3.3, what is fatal to the validity of Takahashi’s (2005) movement analysis is the disconfirmation of the *Schema-based prediction and not the failure of the corresponding ok Schema-based prediction to be confirmed. It is perhaps worth noting that the ok Schema-based prediction in question is not confirmed. Since rokuna-N 'good/decent N' in the focus position is reconstructed to the position inside the presuppositional CP at LF, it is c-commanded by embedded Neg at LF in the structure in (11). Therefore, given the condition on rokuna-N in (12), we can obtain the ok Schema in (15). 7 (15) ok Schema: [… … V-nai-T]-no-wa rokuna-N-CM da-T Again, since the case-marker on rokuna-N in the focus position is not specified in (15), it can be the subject, accusative object, dative object or postpositional phrase. Thus, one of the ok Examples conforming to the ok Schema in (15) is shown in (16), in which rokuna-N in the focus position is the dative object with the dative case-marker ni. 6 The details of the result of EPSA[25]-3 will be shown in Chapter 6.3.2. 7 As we will see shortly, the ok Schema in (15) is the *Schema in EPSA[25]-3, which I have been conducting. Thus, the ok Example in (16) is one of the *Examples in EPSA[25]-3. 212 (16) ok Example: Sono paatii-de Hanako-ga koeokake-nak-katta-no-wa rokuna otoko-ni da-tta. that party-at Hanako-Nom spoke:to-Neg-Past-that-Top decent man-Dat be-Past '(Lit.) It was decent men that Hanako didn’t spoke to at that party.' The example in (16) and three other examples conforming to the schema in (15) are included in EPSA[25]-3. 8 The average score of the judgments of eight informants is 3. The example (16) is lexically as well as structurally similar to (14). Therefore, it is implausible to reduce such low acceptability solely to parsing difficulty or the unnaturalness of the interpretation of (16). The fact that the ok Schema-based prediction with regard to (15) fails to be confirmed thus also goes against Takahashi’s (2005) movement analysis. 9 6.3.2 Some possible modifications We have seen that the *Schema-based prediction that Takahashi’s (2006) movement analysis would make has been disconfirmed, while the corresponding ok Schema-based prediction has not been confirmed. The disconfirmation of the *Schema-based prediction, in particular, is fatal to Takahashi’s (2006) movement analysis. Thus, he might have to modify his analysis in order to account for the NSE facts. I will consider next what modifications might be made under Takahashi’s (2006) movement analysis. What has disconfirmed the *Schema-based prediction is the high acceptability of the *Examples such as (14). Thus, Takahashi (2006) would want to predict that these examples are acceptable, in other words, want to turn the *Schema in (13) to an ok Schema. Under the LF-structure in (11), the focus 8 The details of the result of EPSA[25]-3 will be shown in Chapter 6.3.2. 9 Recall again the discussion in Chapter 3.3.3 that ok Schema-based predictions cannot be disconfirmed since it is not possible to exhaustively check all the possible ok Examples of an ok Schema. 213 phrase is not c-commanded by the matrix Neg, and that has made the schema in (13) a *Schema. Thus, in order to make the *Schema in (13) an ok Schema, he would need the structure in which the focus phrase is c-commanded by the matrix Neg. One way to achieve this is to assume the V-to-I movement of the copula verb da by way of other heads intervening between V and I. 10 Given this V-to-I movement, the LF-structure in (11) would be as in (17). (17) LF-structure of CM-Cleft after V-to-I movement (cf. (11)): 11 TP CP T' t 1 C' FP V-Neg-Foc-T TP C t 1 F' | NegP V-Neg-T no-wa NegP t V-Neg-Foc VP t V-Neg VP t V-Neg t CP t V Taroo-ga Hanako-ni 1 t V reconstruction As a result of the V-to-I movement, the matrix Neg is situated at the position of the matrix T, c-commanding the presuppositional CP and thus the elements inside it. Thus, the matrix Neg now c-commands the focus phrase that is inside the presuppositional CP after the reconstruction. The schema in (13) is therefore predicted to be acceptable under the revised structure in (17). 10 For the arguments for the existence of V-to-I movement in Japanese, see Otani & Whitman 1991, Koizumi 1995, 2000 and Hatakeyama, Honda & Tanaka 2008. 11 In the matrix clause in (17), the head V is first incorporated into the head Neg, forming the amalgamated head [V-Neg]. This head is then incorporated into the head Foc, forming the amalgamated head [V-Neg-Foc]. This head is finally incorporated into the head T, forming the amalgamated head [V-Neg-Foc-T]. The head Foc does not have any phonological realization in Japanese. 214 However, the other problem still remains. Recall that the ok Schema-based prediction has not been confirmed. Since the focus phrase is still inside the presuppositional CP after the reconstruction in the structure in (17), it is c-commanded by the embedded Neg. Therefore, the ok Schema in (15) would remains as an ok Schema, even under the revised structure in (17). In order to solve this problem, the focus phrase needs to be situated in the position that is not c-commanded by the embedded Neg at LF. One way to achieve this is to assume that the focus movement is not reconstructed. This is in fact what is assumed in Hiraiwa & Ishihara 2002. 12 If the focus phrase remains in the focus position at LF, the LF-structure in (17(17) comes as in (18). (18) LF-structure of CM-Cleft after V-to-I movement and without focus reconstruction TP CP T' t 1 C' FP V-Neg-Foc-T TP C Hanako-ni 1 F' | NegP V-Neg-T no-wa NegP t V-Neg-Foc VP t V-Neg VP t V-Neg t CP t V Taroo-ga t 1 t V In (18) the focus phrase is c-commanded by the matrix Neg but not by the embedded Neg. Thus, the schema in (13) is now an ok Schema, while the schema in (15) is a *Schema, so that the example in (14) is predicted to be acceptable, while the example in (16) unacceptable. The results of ESPA[25]-3 presented above are in harmony with the predictions. However, these modifications give rise to at least three problems. First, as I have pointed 12 See Chapter 2.4.1.1.2 for the discussion in regard to this assumption by Hiraiwa & Ishihara (2002). 215 out in Chapter 2.4.1.1.2 in regard to Hiraiwa & Ishihara’s (2002) movement analysis, the assumption of the A′-movement that does not 'reconstruct' introduces a new type of A'-movement, which is different from the commonly assumed one. Reconstruction effects are one of the representative properties of A′-movement and it seems costly to introduce this type of A'-movement into the current theory because such an A'-movement virtually has no difference from the A-movement in its effects. Second, the existence of V-to-I movement in Japanese is controvercial. 13 V-to-I movement should be triggered by grammatical agreement features. However, Fukui (1986) and Kuroda (1988) suggest that grammatical agreement features are totally absent in the Lexicon of Japanese. In the generative tradition, movements that are triggered by grammatical agreement features are assumed to take place obligatory. However, Japanese lucks such obligatory movements, which casts doubt on the existence of grammatical agreement features in the Lexicon of Japanese. Unless the existance of obligatory movements, and thus the existence of grammatical agreement features, were confirmed in Japanese, and presumably on the basis of some *Schema-based predictions made under the relevant hypotheses surviving a rigorous disconfirmation attempt (together with the confirmation of the corresponding ok Schema-based predictions, arguments for the existence of the V-to-I movement also would remain to be rather weak, to say the least. Moreover, I (or T) in Japanese is a bound morpheme and nothing can intervene between V and I except for other functional heads such as the modal and Neg, which are also bound morphemes. Therefore, it is difficult to test the existance of the V-to-I movement by making recourse to the ways taken in English and some other language, for example, by checking the position in relation to VP-adverbs, negation and floating quantifiers. Third and most crucially, as a result of the assumption that the focus movement is not 13 For example, Hoji (1998) argues against Otani & Whitman 1991 and Takano (2002) argues against Koizumi 1995, 2000. 216 reconstructed, the BVA and DR facts observed in the preceding chapter cannot be accounted for under the movement analysis. The facts show that the focus phrase is interpreted for BVA and DR at the trace position of the empty operator movement inside the presuppositional CP under the base-generation analysis. In order to account for these facts, I have introduced the mechanism of chain binding assumed in Hoji & Ueyama 2003. Under the movement analysis, on the other hand, this is a natural consequence of the reconstruction of the focus phrase to the originated position inside the presuppositional CP. No additional assumption is necessary. However, introducing the assumption that the focus movement is not reconstructed eliminates this advantage of the movement analysis. Some additional assumption(s) is/are thus necessary under the movement analysis, too, in order to account for the BVA and DR facts. 14 In conclusion, unless the introduction of the new type of A'-movement that does not get reconstructed is justified and the existence of the V-to-I movement in Japanese is attested, the above modifications that would be made by Takahashi (2006) could not be validated. 6.4 NSEs in CM-Cleft under the base-genaration analysis 6.4.1 Predictions in the base-generation analysis Let us now turn to the base-generation analysis proposed in this work. Under the assumption in (7) that Neg (-nai) is in the sister position of the VP at LF, the LF-structure of CM-Cleft proposed here in (1b) comes as in (19) with the positions of Neg in the matrix clause 14 Another possible modification one might make under Takahashi’s (2006) analysis would be to assume that the NSE licensing is computed in the overt syntax (i.e., before Spell Out), while BVA and DR are computed at LF, together with the V-to-I movement in the overt syntax. Given this, the focus phrase is interpreted at the focus position for the NSE licensing, while it is interpreted at the originated position inside the presuppositional CP for BVA and DR. This modification would correctly account for both the NSE facts and the BVA and DR facts. The validity of this modification depends on whether they would be able to show that (i) the NSE licensing is not an LF-phenomenon, and (ii) the V-to-I movement exists (and is obligatory) in Japanese. We should also ask whether this modification would lead to any new *Schema-based predictions, to ensure the testability of the proposal. 217 and in the presuppositional CP. (19) LF-structure of CM-Cleft with Neg positions (cf., (1b)): TP NegP T VP Neg SC V | CP NP 1 -CM da TP C | Op 1 TP no-wa NegP T VP Neg NP-CM…t 1 …V In (19), the focus phrase NP 1 -CM is c-commanded by the matrix Neg but not by the Neg in the presuppositional CP. Given the condition on rokuna-N in (12), then, we can obtain the schematic asymmetry in (20). (12) (Kataoka 2007: 16, (43)) Rokuna-N must be c-commanded by Neg at LF. (20) Schema group: a. ok Schema A (= (13)) [… … V-T]-no-wa rokuna-N-CM zya-nai-T 218 b. *Schema B (= (15)) [… … V-nai-T]-no-wa rokuna-N-CM da-T c. ok Schema C [… … V-nai-T]-no-wa NP-CM da-T It should be noted that the nature of the ok Schema C in (20) is different from the one in BVA and DR. In the cases of BVA and DR, the ok Schema C is phonetically identical to the *Schema. The only difference is that the *Schema does but the ok Schema C does not involve interpretation γ(a, b). In (20), however, the ok Schema C is not phonetically identical to the *Schema. Kataoka (pc in February, 2011) assumes that in the case of the NSE licensing, γ(a, b) is an interpretation between Neg and an NSE. She then assumes that under γ(Neg, rokuna-N), rokuna-N 'good/decent N' is interpreted as 'bad/worthless N' in combination with Neg. However, if we adopt Kataoka’s view, it is not clear what interpretation the sentence has without γ(a, b). Therefore, instead of using the phonetically identical one to the *Schema, the ok Schema C in the Neg licensing is constructed by removing rokuna from rokuna-N in the *Schema, so that the resulting sentence is a simple negation sentence. The purpose of having an ok Schema C is to ensure that the unacceptability of the *Examples cannot be due to parsing difficulty. Since the ok Schema C here differs from the *Schema B only in the absence of rokuna and all the other parts are the same, it can serve for this purpose of ok Schema C . 6.4.2 Experiment and its results The predicted schematic asymmetry in (20) is what is tested in EPSA[25]-3. 15 One 15 The experiment on rokuna-N does not involve the kind of γ(a, b) that we see in the case of BVA and DR, and we might consider the experiment in question as a preliminary to experiments on the availability of 219 ok Example conforming to the ok Schema A is (14), while one *Example conforming to the *Schema B is (16). Rokuna-N in (14) and (16) is an NP expressing a person with the dative case-marker ni. In each schema, there are three more examples. In one example, rokuna-N is the same as the one in (14) and (16), which expresses a human, but with the nominative case-marker ga. In the other two examples, rokuna-N expresses a non-human and one has the dative case-marker ni and the other has the nominative case-marker ga. Thus, a total of 12 examples are tested, which are shown in (21)-(24). (21) (Rokuna-N-Dat, N = person) a. ok Examples A-1 (= (14)): Sono paatii-de Hanako-ga koeokaketa-no-wa rokuna otoko-ni zya-nak-katta. that party-at Hanako-Nom spoke:to-that-Top decent man-Dat be-Neg-Past '(Lit.) It wasn’t decent men that Hanako spoke to at that party.' b. *Examples B-1 (= (16)): Sono paatii-de Hanako-ga koeokake-nak-katta-no-wa rokuna otoko-ni da-tta. that party-at Hanako-Nom spoke:to-Neg-Past-that-Top decent man-Dat be-Past '(Lit.) It was decent men that Hanako didn’t spoke to at that party.' c. ok Examples C-1 : Sono paatii-de Hanako-ga koeokake-nak-katta-no-wa seenen zitugyooka-no that party-at Hanako-Nom spoke:to-Neg-Past-that-Top young businessman-Gen Yamada-san-ni da-tta. Yamada-Mr.-Dat be-Past '(Lit.) It was Mr. Yamada, a young businessman, that Hanako didn’t spoke to at that party.' BVA with a rokuna-N as A of BVA(A, B), as Kataoka has been conducting. 220 (22) (Rokuna-N-Nom, N = person) a. ok Examples A-2 : Sono paatii-de Hanako-ni koeokaketa-no-wa rokuna otoko-ga zya-nak-katta. that party-at Hanako-Dat spoke:to-that-Top decent man-Nom be-Neg-Past '(Lit.) It wasn’t decent men that spoke to Hanako at that party.' b. *Examples B-2 : Sono paatii-de Hanako-ni koeokake-nak-katta-no-wa rokuna otoko-ga da-tta. that party-at Hanako-Dat spoke:to-Neg-Past-that-Top decent man-Nom be-Past '(Lit.) It was decent men that didn’t spoke to Hanako at that party.' c. ok Examples C-2 : Sono paatii-de Hanako-ni koeokake-nak-katta-no-wa seenen zitugyooka-no that party-at Hanako-Dat spoke:to-Neg-Past-that-Top young businessman-Gen Yamada-san-ga da-tta. Yamada-Mr.-Nom be-Past '(Lit.) It was Mr. Yamada, a young businessman, that didn’t spoke to Hanako at that party.' (23) (Rokuna-N-Dat, N = non-person) a. ok Examples A-3 : Dentuu-ga zyobu feaa-e-no sanka-o yobikaketeiru-no-wa rokuna kaisya-ni zya-nai. Dentsu-Nom job fair-to-Gen participation-Acc calling:on-that-Top decent company-Dat is-Neg '(Lit.) It isn’t decent companies that Dentsu is calling on to participate in the job fair.' b. *Examples B-3 : Dentuu-ga zyobu feaa-e-no sanka-o yobikaketei-nai-no-wa rokuna kaisya-ni da. Dentsu-Nom job fair-to-Gen participation-Acc calling:on-Neg-that-Top decent company-Dat is 221 '(Lit.) It is decent companies that Dentsu is not calling on to participate in the job fair.' c. ok Examples C-3 : Dentuu-ga zyobu feaa-e-no sanka-o yobikaketei-nai-no-wa Toyota-ni da. Dentsu-Nom job fair-to-Gen participation-Acc calling:on-Neg-that-Top Toyota-Dat is '(Lit.) It is Toyota that Dentsu is not calling on to participate in the job fair.' (24) (Rokuna-N-Nom, N = non-person) a. ok Examples A-4 : Kotosi zyobu feaa-ni sankasiteiru-no-wa rokuna kaisya-ga zya-nai. this:year job fair-to participating-that-Top decent company-Nom is-Neg '(Lit.) It isn’t decent companies that are participating in the job fair this year.' b. *Examples B-4 : Kotosi zyobu feaa-ni sankasitei-nai-no-wa rokuna kaisya-ga da. this:year job fair-to participating-Neg-that-Top decent company-Nom is '(Lit.) It is decent companies that are not participating in the job fair this year.' c. ok Examples C-4 : Kotosi zyobu feaa-ni sankasitei-nai-no-wa Toyota-ga da. this:year job fair-to participating-Neg-that-Top Toyota-Nom is '(Lit.) It is Toyota that is not participating in the job fair this year.' Table 6.1 summarizes the result of EPSA[25]-3 on the predicted schematic asymmetry in (20), which has now 11 participants. 16 16 I have reported the average scores of the examples conforming to ok Schema A and *Schema B above, though the former was a *Schema and the latter was an ok Schema then. 222 Table 6.1: A summary of the results of EPSA [25]-3 EPSA [25]-3 (Total 11 participants; 190 answers) ... as of Mar/1/2011 OK Schema A *Schema B OK Schema C (21) Rokuna-N-Dat, N = person 10 values 72 10 values 2 11 values 85 (22) Rokuna-N-Nom, N = person 10 values 35 10 values 0 11 values 68 (23) Rokuna-N-Dat, N = non-person 10 values 75 10 values 0 11 values 82 (24) Rokuna-N-Nom, N = non-person 8 values 25 10 values 10 10 values 63 Total 38 values 53 40 values 3 36 values 75 The average score on the four *Examples conforming to the *Schema is very low, while the average scores of the corresponding ok Examples are significantly higher than the *Examples. Therefore, although the total number of informants is still small, the predicted schematic asymmetry in (20) has so far been confirmed. Under the formulation of the ok Schema-based prediction presented in Chapter 3.3.3, an ok Schema-based prediction gets confirmed when the informants' judgments on the ok Examples conforming to the ok Schema are not completely unacceptable. Thus, the score of 53 for the ok Schema A in Table 6.1 confirms the ok Schema based prediction. However, one may feel that the score of 53 for the ok Schema A is not as high as one might expect, though this is not crucial to the confirmation of ok Schema-based predictions as it is understood here. The (un)acceptability of ok Schemata can be affected by factors outside the Computational System such as parsing difficulty and the unnaturalness of the interpretation of the entire sentence. As I have pointed out in Chapter 5.3.3.2, fn.14, when the focus phrase has the nominative case-marker ga, some informants find the resulting sequence NP-ga da a little unnatural as compared to the examples where the nominative case-marker ga is replaced by the other case-markers (e.g., o, ni) or 223 postpositions (e.g., kara 'from', to 'with'). This seems to be applicable here, too, though what follows ga is not da but zya-nai/nakatta. If we recalculate the results in Table 6.1 based on the choice of the case-marker on the focus phrase, we obtain Table 6.2. 17 Table 6.2: A summary of the results of EPSA [25]-3 (dative focus vs. nominative focus) EPSA [25]-3 (Total 11 participants; 190 answers) ... as of Mar/1/2011 OK Schema A *Schema B OK Schema C (21) & (23) Rokuna-N-Dat 20 values 73 20 values 1 22 values 83 (22) & (24) Rokuna-N-Nom 18 values 30 20 values 5 21 values 66 Total 38 values 53 40 values 3 43 values 75 The results in Table 6.2 show that by using the dative case-marker ni (and possibly other case-markers than the nominative ga), we can obtain much clearer asymmetry between the *Schema and ok Schemata. Recall also that in Chapter 5.3.3.2 we have classified the informants into "linguistically informed" and "linguistically-naïve" informants based on whether they state they understand (i) "bound variable anaphora" or "bound readings" and (ii) "A takes wide scope over B." It is reasonable to assume that "Linguistically informed" informants, in general though perhaps not always, have more structural/contextual resourcefulness than "linguistically naïve" informants. If we classify the 11 informants in EPSA[25]-3 into "linguistically-informed" informants and "linguistically-naïve" informants and recalculate the average score of each group, the result of EPSA[25]-3 in Table 6.2 comes as shown in Table 6.3. 17 Person/non-person distinction of the N of rokuna-N, on the other hand, does not make much difference in acceptability. 224 Table 6.3: A summary of the results of EPSA [25]-3 (dative vs. nominative focus by group) EPSA [25]-3 (Total 11 participants; 190 answers) ... as of Mar/1/2011 OK Schema A *Schema B OK Schema C (21) & (23) Rokuna-N- Dat linguistically-informed (7 informants) 12 values 91 12 values 0 14 values 100 linguistically-naïve (4 informants) 8 values 46 8 values 3 8 values 55 Total 20 values 73 20 values 1 22 values 83 (22) & (24) Rokuna-N- Nom linguistically-informed (7 informants) 12 values 50 12 values 0 13 values 89 linguistically-naïve (4 informants) 8 values 0 8 values 12 8 values 4 Total 18 values 30 20 values 5 21 values 66 When rokuna-N in the focus position is the subject with the case-marker ga, the average score on the *Examples conforming to the *Schema by the three "linguistically-naïve" informants is very low, but the average scores of the ok Examples conforming to the ok Schemata are also very low. Therefore, we cannot attribute the unacceptability of the *Examples conforming to the *Schema by the three informants solely to the hypothesized property/ies of the CS. On the other hand, the average score on the *Examples conforming to the *Schema by the six "linguistically-informed" informants is very low, while the average score of the ok Examples conforming to the ok Schema C is very high. This suggests that it is highly unlikely that the low score on the *Examples conforming to the *Schema is due to parsing difficulty or the unnaturalness of the interpretation of the entire sentence, thereby suggesting that it is due to (12). However, the average score of the ok Examples conforming to the ok Schema A is not very high even by these "linguistically-informed" informants. This suggests that some factors other than the hypothesized property/ies of the CS might be involved in the unacceptability of the *Examples conforming to the *Schema. We should thus avoid using nouns with the 225 nominative case-marker ga as the focus phrase in constructing examples for the NSE-related experiments. When rokuna-N in the focus position is the dative object with the case-marker ni, on the other hand, the average score on the *Examples conforming to the *Schema by the three "linguistically-naïve" informants is very low, but the average scores of the ok Examples conforming to the ok Schemata are not very high. This suggests that these three informants' judgments might have been affected by parsing difficulty or the unnaturalness of the interpretation of the entire sentence and the same factor(s) might have affected their judgments of the *Examples conforming to the *Schema. Therefore, we cannot attribute the unacceptability of the *Examples conforming to the *Schema by these three informants solely to the hypothesized property/ies of the CS. On the other hand, the average score on the *Examples conforming to the *Schema by the six "linguistically-informed" informants is very low, while the average scores of the ok Examples conforming to the ok Schemata are significantly higher than that of the *Examples. Therefore, we can provisionally assume that the predicted schematic asymmetry in (20) gets confirmed within this informant group. Thus, the result of EPSA[25]-3 is in favor of the proposed base-generation analysis of CM-Cleft. 6.5 Resumption in CM-Cleft 6.5.1 Preliminary It is commonly assumed in English that resumptions appear at the trace position of A'-movement to remedies the subjacency violation as shown in (25), while it is not allowed in a 'local' context as shown in (26). 226 (25) (Hoji & Ueyama 2003: fn.1, (i), referring to Ross 1967: 6, (154a)) I just saw that girl who 1 [Long John's claim that {she/*ec} 1 was a Venusian] made all the headlines. (26) (Hoji & Ueyama 2003: fn.1, (i), referring to Sells 1984:6 (4)): * This is the man who 1 I saw him 1 yesterday. In regard to resumption in Japanese, Saito (1985) claims that 'Scrambling' construction (in Ueyama’s (1998) term, the OS-type construction) never allows resumption. Saito (1995) presents the examples as in (27) to demonstrate this point. (27) (Saito 1985:115 (118), 164 (9)) a. * John i -o, Mary-ga kare i -o mita (koto) John-ACC Mary-NOM he-ACC saw fact '(the fact that) John i , Mary saw him i ' b. * So-no hon i -o John-ga [Mary-ga so-re i -o katta to] omotteiru] (koto) that-GEN book-ACC John-NOM Mary-NOM that-thing-ACC bought COMP think fact '(the fact that) [that book] i , John thinks [that Mary bought it i ]' Under Saito’s (1985) analysis, 'Scrambling' is an adjunction operation, which is usually classified as an A'-movement. Therefore, Saito’s analysis of resumption is basically in line with the analysis of resumption in English in that A'-movement does not allow resumption in both analyses. Hayashishita (1997), on the other hand, observes that the Deep OS-type allows resumption in principle, while the Surface OS-type does not. Examples like (28a) allow the overt form 227 soko-ni in the position of the ec in (28b) (28) (Ueyama 1998: 69, (103a) referring to Hayashishita 1997) a. Toyota-ni-sae Nissan-ga so-ko-ni syatyoo-to-no mendan-o moosiiretekita. Toyota-DAT-even Nissan-NOM that-place-DAT president-with-GEN appointment-ACC requested 'Even to Toyota, Nissan applied to it for an appointment with the president.' b. Toyota-ni-sae Nissan-ga ec syatyoo-to-no mendan-o moosiiretekita. Toyota-DAT-even Nissan-NOM president-with-GEN appointment-ACC requested Resumption is thus possible in the OS construction in Japanese. Now consider (29). (29) (Ueyama 1998: 70, (104a) referring to Hayashishita 1997) a. * [55%-no konpyuutaa] i -o Toyota-ni [so-re] i -no sekkeisya-ga so-ko-ni okurituketekita. 55%-GEN computer-ACC Toyota-DAT that-thing-GEN planner-NOM that-place-DAT sent:directly '[55% of the computers] i , to Toyota, its i planner sent e i directly to it.' b. [55%-no konpyuutaa] i -o Toyota-ni [so-re] i -no sekkeisya-ga ec okurituketekita. 55%-GEN computer-ACC Toyota-DAT that-thing-GEN planner-NOM sent:directly (29a) and (29b) are instances of the multiple OS-type construction. Ueyama (1998: Chapter 2.4.2) demonstrates that at most one DL (a dislocated NP) can be a Deep DL in the multiple OS-type construction. 18 Since the first DL 55%-no konpyuutaa '55% of the computers' must be a Deep DL in order to obtain BVA(55%-no konpyuutaa, sore), the second DL Toyota-ni has to be a Surface-DL. Thus, the impossibility of resumption in (29a) shows that the Surface 18 Deep DL is a dislocated NP that shows the properties of A-movement, for example, absence of the WCO effects, while Surface DL is a dislocated NP that shows the properties of A'-movement, for example, reconstruction effects. 228 OS-type does not allow resumption. On the other hand, given that the Surface OS-type does not allow resumption, the possibility of the resumption in (28a) shows that the DL Toyota-ni-sae 'Even to Toyota' has to be a Deep-DL, and thus the Deep OS-type allows resumption. Based on this observation as well as the possibility of resumption in other constructions, Hoji & Ueyama (2003) propose the analysis of resumption in Japanese in (30). (30) (Hoji & Ueyama 2003: 10-11) "Resumption" in Japanese (that has a unique grammatical property) is an overt category that appears in the launching site of a covert IP-adjunction. PF: [ . . . {pro/so-ko} 1 . . . ] LF: [ IP {pro/so-ko} 1 [ IP . . . t 1 . . . ]] Semantic Representation: λx [ . . . x . . . ] Recall the analyses of the Deep OS-type and Surface OS-type constructions in Ueyama 1998, which are repeated here as (32) and (32), respectively. (31) (Ueyama 1998: 63, (91)) Deep OS-type: PF: NP 1 -DAT/ACC NP-NOM ec 1 ... LF: NP 1 -DAT/ACC Op 1 [ NP-NOM t 1 ... (32) (Ueyama 1998: 63, (92)) Surface OS-type: PF: NP 1 -DAT/ACC NP-NOM ... t 1 ... LF: NP-NOM NP-DAT/ACC ... 229 Hoji (to appear: 5) assumes that the empty nominal that is represented as pro in (30) acts as an empty operator. Given this, the covert IP-adjunction of pro in (30) is equivalent to the empty operator movement assumed in the Deep-OS type in (31). 19 Therefore, resumption can appear at the trace position of Op in (31), which is the launching site of the covert IP-adjunction in (30). On the other hand, under Ueyama’s (1998) analysis of the Surface OS-type in (32), the position where the resumption would appear is where the DL (NP-DAT/ACC in (32)) is at LF. The DL appears in its surface position as the result of the PF-movement. 20 There is thus no position for resumption to appear in Surface OS-type in (32). Therefore, under the analyses of the Deep OS-type and Surface OS-type constructions in Ueyama 1998, resumption is predicted to be possible in the Deep OS-type but not in the Surface OS-type. 21 6.5.2 Possibility of resumption in CM-Cleft Let us now turn to the resumption in CM-Cleft. In Chapter 1, we have seen Hoji & Ueyama’s (2003) observation that CM-Cleft allows resumption (sore 'that thing') in a local context. The example presented there is repeated in (33). (33) (Hoji & Ueyama 2003: 2, (7a)) [Kokuren-ga kibisiku [so-re-o] 1 hihansita no]-wa [amerika-no booei-seisaku-o] 1 da. the:United:Nations-NOM harshly that-thing-ACC criticized COMP-TOP USA-GEN defense-policy-ACC COPULA 19 See Chapter 1, fn.11 for the details of this assumption. 20 Ueyama (1998:63) assumes that "a PF movement freely occurs without requiring any syntactic motivation." 21 The experiments on resumption have been conducted by Hoji (EPSA[23]-1,2 and 5), which test what can be used as resumption in Japanese. The results of EPSA[23]-1 and 2 show that so-words such as soko 'that place' and soitu 'that person' are much more easily used as resumption than a-words such as asoko 'that place' and aitu 'that person', which are impossible for most of the informants. Furthermore, the results of EPSA[23]-5 shows that among so-words, "small" so-NPs such as soko and soitu are more easily used as resumption than "large" so-words of the form so-no NP 'that NP'. 230 'It was [the USA's defense policy-ACC] 1 that the United Nations harshly criticized it 1 .' In the base-generation analysis proposed in this work, the same empty operator movement as assumed in the Deep-OS type is assumed as shown in the LF-structure in (1b). Therefore, given the analysis of the resumption in (30), it is predicted under the base-generation analysis that resumption is possible at the launching site of the empty operator movement. The example in (33) is one such example and Hoji & Ueyama (2003) report that it is acceptable. I have been testing more examples in (34) and (35) to see if resumption is in fact possible in CM-Cleft. (34) Resumption = soko 'that place' a. Bee-gikai-ga san-syuu renzoku so-ko-o shookansita-no-wa U.S.-Congress-Nom 3-week consecutively that-place-Acc summoned-that-Top kanarinokazu-no kaisya-o da. a:considerable:number-Gen company-Acc is '(Lit.) It was a considerable number of companies that the U.S. Congress summoned it for three consecutive weeks.' b. Iroirona syoohisya-dantai-ga so-ko-ni kookaisitumonzyoo-o okurituketa-no-wa various consumer-organization-NOM that-place-Dat open:letter-ACC sent:out-that-Top kanarinokazu-no kaisya-ni da. a:considerable:number-Gen company-Dat is '(Lit.) It was to a considerable number of companies that many consumer organizations sent out open letters to it.' 231 (35) Resumption = soitu 'that person' a. Bee-gikai-ga san-syuu renzoku so-itu-o shookansita-no-wa U.S.-Congress-Nom 3-week consecutively that-person-Acc summoned-that-Top kanarinokazu-no syuutizi-o da. a:considerable:number-Gen governor-Acc is '(Lit.) It was a considerable number of governors that the U.S. Congress summoned him/her for three consecutive weeks.' b. Iroirona syoohisya-dantai-ga so-itu-ni kookaisitumonzyoo-o okurituketa-no-wa various consumer-organization-NOM that-person-Dat open:letter-ACC sent:out-that-Top kanarinokazu-no syuutizi-ni da. a:considerable:number-Gen governor-Dat is '(Lit.) It was to a considerable number of governors that many consumer organizations sent out open letters to him/her.' The average score of the above four examples by 11 informants is 45. The score is not particularly high, but the result at least shows that resumption is in fact possible for not a few informants. Resumption is not a common phenomenon and some people, especially those who have little or no linguistic background, may not tolerate resumption. Thus, let us provide some more details of the informants' judgments by classifying them into "linguistically informed" and "linguistically naïve" informants. Among the 11 informants, seven informants are classified into "linguistically informed" informants, while four are into "linguistically naïve" informants. The average score of the seven "linguistically informed" informants is 57, while the one of the four "linguistically naïve" informants is 25. In fact, only one "linguistically informed" informant judged all of the above examples totally unacceptable and one judged all the examples 25 or lower. Thus, quite a few informants, especially "linguistically informed" 232 informants, accept resumption in CM-Cleft. This result is as expected under the base-generation analysis of CM-Cleft proposed in this work. What would this result tell about the movement analysis of CM-Cleft? As mentioned at the beginning of this section, resumptions are not allowed in English at the trace position of A'-movement if it is in a 'local' context. If it is indeed the case in Japanese, too, the movement analysis would predict that the examples in (34), (35) and (36) were unacceptable with resumption since the focus movement is assumed to be an A'-movement and the resumptions of all these examples are in a 'local' context. Thus, the result of the experiment just presented would disconfirm the prediction under the movement analysis. However, since neither Takahashi (2006) nor Hiraiwa & Ishihara (2002) address the resumption in their analyses of CM-Cleft, it is unclear what they might say about resumption in CM-cleft, making it difficult to construct a predicted schematic asymmetry involving resumption under their analyses. 6.5.3 Resumption in the Surface OS-type In the above discussion, I crucially depend on the hypothesis about resumption in (30). Thus, the validity of (30) would strengthen the validity of the claim I have made above, namely, the resumption is possible in CM-Cleft as the result of covert IP-adjunction of the empty operator inside the presuppositional CP. One way to demonstrate the validity of (30) is to show that the Deep OS-type does but the Surface OS-type does not allow resumption. Based on the hypothesis about resumption in (30) and the analyses of the Deep OS-type and Surface OS-type constructions in (31)) and (32), we obtain the following predicted schematic asymmetry. 233 (36) Schema group: a. ok Schema A QP-CM [… so-NP … ]-ga resumption-CM V BVA (QP, so-NP) b. *Schema B [… so-NP … ]-CM QP-ga resumption-CM V BVA (QP, so-NP) c. ok Schema C [… so-NP … ]-CM QP-ga soko/soitu-CM V (With soko/soitu "referring to" a particular individual/object) As presented in Chapter 4.2.1, when we use what Ueyama (1998) calls A-type QPs that do not refer to a specific group of individuals as A of BVA(A, B), we can be reasonably confident that the BVA, if possible, is FD-based. In the FD-based BVA, (the QR trace of) the QP must c-command the dependent term at LF. Thus, in (36a) in order to get BVA (QP, so-NP), (the QR trace of) the QP must c-command so-NP at LF. Therefore, the QP must be a Deep DL, which is base-generated at an IP-adjoined position. Since the structure in (36a) is the Deep OS type, resumption is predicted to be possible, so that the schema in (36a) is an ok Schema. On the other hand, in (36b) if the DL containing so-NP is a Deep DL, so-NP cannot be c-commanded by (the QR trace of) the QP at LF, so that BVA (QP, so-NP) is impossible. In order to get BVA (QP, so-NP), then, the DL must be a Surface DL, i.e., the DL must occupy the position of the resumption in (36b) at LF. Since resumption cannot occupy the same position as the Surface DL, under the hypothesis in (30), resumption is predicted to be impossible in (36b), making it a *Schema. Kiyoko Kataoka conducted an experiment several years ago to check the acceptability of the examples corresponding to the schemata in (36a) and (36b) although the experiment in 234 question does not have the examples conforming to (36c) by using a general experimental design that is not as systematic as EPSA. Given in (37) and (38) are ok Example conforming to the ok Schema A in (36a) and a *Example conforming to the *Schema B in (36b), respectively. (37) ok Example Itutuizyoo-no ginkoo-ni so-ko-no torihikisaki-ga soko-ni Nissan-o suisensita. 5:or:more-Gen bank-Dat that-place-Gen customer-Nom that-place-Dat Nissan-Acc recommended '(Lit.) To more than five companies, its customer recommended Nissan to it.' (38) *Example So-ko-no torihikisaki-ni itutuizyoo-no ginkoo-ga soko-ni Nissan-o suisensita. that-place-Gen customer-Dat 5:or:more-Gen bank-Nom that-place-Dat Nissan-Acc recommended '(Lit.) To its customer, more than five companies recommended Nissan to it.' There were 16 informants who participated in that experiment. The average score on the ok Example in (37) is 86 on the scale of 0-100 with 0 being completely unacceptable and 100 being fully acceptable. On the other hand, the average score on the *Example in (38) is 14 on the scale of 0-100. These results seem to be basically in line with the predicted schematic asymmetry between (36a) and (36b). 6.6 Summary In this chapter, I have presented two arguments for base-generation analysis of CM-Cleft; NSE-based argument and resumption-based argument. 235 Under the base-generation analysis proposed in this work, I have introduced the two hypotheses; (i) the focus position of CM-Cleft is the A'-position and (ii) the focus phrase is interpreted by way of chain binding. As a result of the introduction of these two hypotheses, however, the base-generation analysis makes the same predictions as the movement analysis as far as the BVA facts and the DR facts of CM-Cleft are concerned. Thus, in order to discriminate between the two analyses, we need to have phenomena in which the movement analysis and the base-generation analysis make different predictions and to show that the predicted schematic asymmetries obtained under the base-generation analysis get confirmed, while the ones under movement analysis do not. NSEs in Japanese serve the purpose. In the NSE-based argument, I have shown that the focus phrase is c-commanded by the matrix Neg but not by the Neg inside the presuppositional CP under the proposed base-generation analysis, while it is c-commanded by the Neg inside the presuppositional CP but not by the matrix Neg under the movement analysis by Takahashi (2006). Under the hypothesis that the NSE rokuna-N 'good/decent N' must be c-commanded by Neg at LF, the movement analysis predicts that the sentence is acceptable when rokuna-N is the focus phrase and Neg is inside the presuppositional CP, while it is not when rokuna-N is the focus phrase and Neg is in the matrix clause. The former is an ok Schema-based prediction, while the latter a *Schema-based prediction. Informants' judgments on the examples conforming to these Schemata go against both of these predictions; the ok Schema-based prediction fails to get confirmed, while the *Schema-based prediction gets disconfirmed. The base-generation analysis proposed in this work, on the other hand, predicts that the sentence is acceptable when Neg is in the matrix clause, while it is not when Neg is inside the presuppositional CP. The result of the experiment on this predicted schematic asymmetry has so far complied with the predictions though the number of the informants is still small. 236 I have then discussed possible modifications that would be made under Takahashi’s (2006) movement analysis. If he assumed that (i) the V-to-I movement is obligatory in Japanese and (ii) the focus movement, which is the A'-movement, is not reconstructed, the NSE facts could be accounted for. However, both of these two assumptions are controversial, which casts doubt on the validity of these modifications. In the resumption-based argument, I have shown that under the hypothesis about resumption proposed in Hoji & Ueyama 2003 in (30), resumption is predicted to be possible and it is indeed possible for quite a few informants. Whether the result would constitute evidence against the movement analysis depends on how resumption would be analyzed under the movement analysis, making it difficult to obtain a predicted schematic asymmetry involving resumption under the movement analysis. Instead of speculating how they might do that, I have introduced the results of Kataoka’s experiment several years ago that seem to provide support for the validity of the hypothesis about resumption in (30), indicating that the Surface OS-type does not allow resumption. In short, the two arguments presented in this chapter go in favor of the base-generation analysis of CM-Cleft. The base-generation analysis of CM-Cleft proposed in this work accounts for the NSE and resumption facts without any additional assumptions or ad-hoc stipulations. 237 Chapter 7: Concluding remarks 7.1 Summary of the discussion Cleft constructions in Japanese have been a topic widely discussed in the generative literature in Japanese. The analyses proposed in the literature are divided into two types based on how the focus phrase gets to its surface position: One base-generates the focus phrase in its surface position with an empty operator movement inside the presuppositional clause. The other involves the movement of the focus phrase and the subsequent remnant movement of the presuppositional clause. I have referred to the former type of analyses as "base-generation analysis" and the latter as "movement analysis." The focus phrase in cleft constructions can arise with or without a case-marked. Hoji & Ueyama (2003) refer to the cleft sentences with the case-marker on the focus phrase as CM-Cleft and those without the case-marker as non-CM-Cleft. As observed in Hoji & Ueyama 2003, they behave differently in regard to the subjacency effects and the resumption possibility. I have confined my discussion in this work to the CM-Cleft constructions on the ground that they are more likely to reflect the properties of the Computational System. I have proposed a base-generation analysis of CM-Cleft in line with Hoji 1987, 1990. The proposed structure is illustrated in (1). 238 (1) IP VP I SC V | CP NP 1 -CM da IP C | Op 1 IP no-wa … t 1 … In (1), the focus phrase NP 1 -CM forms a small clause (SC in (1)) with the presuppositional CP. This small clause is selected as a complement by the copula verb da. This copular structure in CM-Cleft is in line with the one proposed by Heggie (1988) for English cleft constructions (both (it-)cleft and pseudocleft). A predication relation is established between the focus phrase and the presuppositional CP. I initially assumed that the focus position was an A-position. What is crucial in the structure in (1) is that the focus position c-commands elements inside the presuppositional CP. I made comparisons between the proposed analysis and other analyses based on the predictions each analysis makes. I first made a comparison within the base-generation analysis, specifically, between the proposed analysis and Kizu’s (2005) analysis. One crucial difference between the two analyses is that in the proposed analysis but not in Kizu’s (2005), the focus position c-commands elements inside the presuppositional CP. This structural difference has made different predictions between the two analyses in regard to BVA and DR in CM-Cleft. 1 More specifically, in order for FD-based BVA to obtain, the QP must c-command the dependent 1 Recall that BVA (Bound Variable Anaphora) refers to any anaphoric relation between a distributive/ quantificational expression (QP) and a singular-denoting anaphoric expression, while DR (Distributive Reading) refers to a wide scope distributive reading of a QP over the other QP. 239 term. 2 Thus, when a QP is in the focus position and a dependent term is inside the presuppositional CP, Kizu’s (2005) analysis predicts that FD-based BVA is impossible. I have demonstrated that this *Schema-based prediction on BVA is disconfirmed. 3 The *Schema under Kizu’s (2005) analysis is an ok Schema under the proposed analysis, so that the BVA fact that has disconfirmed the *Schema-based prediction under Kizu 2005 does not pose a problem to the proposed analysis. However, BVA is impossible in some ok Examples conforming to the ok Schema, more specifically, when the dependent term inside the presuppositional phrase is nominative-marked. Thus, the ok Schema-based prediction under the proposed analysis failed to be confirmed. In order to account for this fact, I have made some modifications on the analysis by introducing the following two hypotheses. (2) The focus position in CM-Cleft is an A'-position. 2 FD is an abbreviation of Formal Dependency and its structural and lexical consditions are repeated in (i) for the convenience. (i) (Ueyama 1998: 115, (65)) a. Structural condition on FD: *FD(α,β) if α does not c-command β at LF. b. Lexical condition on FD: *FD(α,β) if β is a large NP. In order for the FD-based BVA to obtain, the QP must c-command the dependent term and the dependent term must be a small NP. 3 The formulations of the*Schema-based prediction and the ok Schema-based prediction are repeated in (i) and (ii), respectively. (i) (Hoji 2010a: 31, (6)) A *Schema-based prediction: Informants judge any *Example conforming to a *Schema to be completely unacceptable under interpretation γ(a, b). (ii) (Hoji 2010a: 31, (7)) An ok Schema-based prediction: Informants judge ok Examples conforming to an ok Schema to be acceptable (to varying degrees) under interpretation γ(a, b). 240 (3) A single member A'-chain can be composed with another A'-chain, forming a single A'-chain, only if the single member A'-chain c-commands the other A'-chain. Based on these hypotheses and the proposed structure of CM-Cleft in (1), I obtained a predicted schematic asymmetry in regard to BVA and conducted an experiment. The results have so far confirmed the predicted schematic asymmetries, which is in favor of the proposed base-generation analysis. DR is another relation that is crucially based on c-command if we carefully exclude the "quirky" type of DR. Thus, if the argument constructed with regard to BVA could be duplicated with regard to DR, we could obtain further support for the proposed analysis. I conducted experiments on the relevant predicted schematic asymmetries. The results of the experiments, however, are not quite in harmony with the predicted schematic asymmetries. Crucially, the scores on the *Examples conforming to the *Schemata are not low enough for us to justifiably conclude that the *Schema-based prediction has survived rigorous disconfirmation attempts. Since the total number of informants is still small, it would be too early to conclude that the *Schema-based predictions were disconfirmed, one might say. However, at least we cannot (yet) obtain strong support for the introduction of the two hypotheses in (2) and (3) from the DR-based argument. The BVA- and DR-based arguments have demonstrated that the proposed base-generation analysis of CM-Cleft is preferable to Kizu’s (2005) base-generation analysis. However, as a result of the introduction of the two hypotheses in (2) and (3), the focus phrase in CM-Cleft is interpreted not in the focus position but in a position inside the presuppositional CP. Under the movement analysis, the focus phrase is base-generated in the position inside the presuppositional CP and gets interpreted at that position after the reconstruction applies at LF as 241 illustrated in (4), for instance. (4) (cf. H&I 2002: (25)-(26)) a. Underlying structure = 'No da' in-situ focus construction: [ TopP [ FocP [ FinP/CP [ TP … XP (Focus) …] no] da] Top] b. Focus movement: [ TopP [ FocP XP (Focus) [ FinP/CP [ TP … t XP …] no] da] Top] c. Remnant Topicalization: [ TopP [ FinP/CP [ TP … t XP …] no]-wa [ FocP XP (Focus) t FinP/CP da] Top] Therefore, as far as the BVA facts and the DR facts of CM-Cleft are concerned, the movement analysis and the proposed base-generation analysis make the same predictions. We thus cannot discriminate between the two analyses in these respects. In order to argue for the base-generation analysis pursued in this work, then, we needed to consider phenomena about which the movement analysis and the base-generation analysis make different predictions and showed that the predicted schematic asymmetries obtained under the base-generation analysis get confirmed, while the ones under movement analysis do not. NSE ('Neg(ation)-sensitive' element) lisencing and resumption in Japanese served the purpose. 4 Rokuna-N 'good/decent N', one of the NSEs, must be c-commanded by Neg at LF. Under the proposed base-generation analysis, the focus phrase is c-commanded by the matrix Neg but not by the Neg inside the presuppositional CP at LF, while under the movement analysis by Takahashi (2006), the focus phrase is c-commanded by the Neg inside the presuppositional CP but not by the matrix Neg at LF. The movement analysis thus predicts that the sentence is 4 NSE ('Neg(ation)-sensitive' element) is an elements that must occur with sentential negation (henceforth, Neg), which is in general called Negative Polarity Items (NPI) in the literature. 242 acceptable when rokuna-N is the focus phrase and Neg is inside the presuppositional CP, while it is not when rokuna-N is the focus phrase and Neg is in the matrix clause. The former is an ok Schema-based prediction, while the latter a *Schema-based prediction. Informants' judgments on the examples conforming to these Schemata go against both of these predictions. Crucially, the *Schema-based prediction has been disconfirmed. The proposed base-generation analysis, on the other hand, predicts that the sentence is acceptable when Neg is in the matrix clause, while it is not when Neg is inside the presuppositional CP. The result of the experiment on this predicted schematic asymmetry has so far complied with the predictions though the number of the informants is still small. As for the possibility of resumption in CM-Cleft, I have shown that under the hypothesis about resumption proposed in Hoji & Ueyama 2003, resumption is predicted to be possible in CM-Cleft and it is indeed possible for quite a few informants. Whether the result would constitute evidence against the movement analysis depends on how resumption would be analyzed under the movement analysis, making it difficult to obtain a predicted schematic asymmetry involving resumption under the movement analysis. Instead of speculating how they might do that, I have introduced the results of Kataoka’s experiment that seem to provide support for the validity of the hypothesis about resumption, indicating that the Deep OS-type does, but the Surface OS-type does not allow resumption. The arguments based on the NSE and resumption facts in CM-Cleft go in favor of the base-generation analysis of CM-Cleft. The proposed base-generation analysis accounts for the NSE and resumption facts without any additional assumptions or ad-hoc stipulations. 7.2 The significance of this work The goal of generative grammar is to discover the properties of the language faculty and a 243 fundamental hypothesis in generative grammar is the existence of the Computational System (CS) at the core of the language faculty. The main goal of generative grammar can therefore be understood as demonstrating the existence of the CS by discovering its properties. Then, the question is how we can discover properties that are reflections of the CS. The interpretation between the two elements (referred to as γ(a, b)) that is crucially based on the c-command relation between an LF object corresponding to a and one corresponding to b is a good candidate for a linguistic intuition that is a reflection of properties of the CS under the hypothesis that Merge is the only structure-building operation in the CS. Therefore, I have focused on the phenomena that are crucially based on the c-command relation at LF in designing the experiments, for example, FD-based BVA, DR and NSE licensing. In this work, I have adopted the EPSA (Evaluation of Predicted Schematic Asymmetries) method advocated by Hoji (2009 and subsequent works). The main general thesis of the EPSA method is that it is possible to study the language faculty as an exact science. The EPSA method is a method of hypothesis testing in line with the hypothetico-deductive method and it aspires to obtain testability, reproducibility and quantitative analyzability in language faculty science. In this work, I have confined my discussions to CM-Cleft on the ground that it is more likely to reflect the properties of the CS and investigated the CM-Cleft construction utilizing the EPSA method. What is tested under the EPSA method is the predicted schematic symmetry, which consists of a *Schema and the two corresponding ok Schemata. When the *Schema-based prediction survives a rigorous disconfirmation attempt, while at the same time the corresponding ok Schema-based predictions are confirmed, we say that we obtain a confirmed schematic asymmetry. Confirmed schematic asymmetries so obtained can be regarded as "basic units of facts" for research concerned with the properties of the CS. The *Schema-based prediction makes our hypotheses testable in the most crucial way. 244 *Examples conforming to *Schema must be judged as completely unacceptable by informants. Thus, when an informant accepts an example conforming to *Schema, the *Schema-based prediction is disconfirmed (at least in principle). When the *Schema-based prediction is disconfirmed, the relevant hypotheses need to be abandoned or modified in a manner that the modification introduces a new *Schema-based prediction. Otherwise, the modification is a content-reducing (i.e., degenerating) problemshift in the terms of Lakatos (1970/1978). I have conducted the experiments on CM-Cleft following the EPSA method, so that the tested predictions are disconfirmable, being prepared to face the challenge of disconfirmation attempts. In fact, one *Schema-based prediction had gotten disconfirmed, which led to the introduction of the two hypotheses in (2) and (3). It is important to point out that the introduction of these hypotheses did not just eliminate the *Schema-based prediction, but it introduced a new *Schema-based prediction. When the predicted schematic asymmetry gets confirmed, we could gain strong support for the proposed analysis. Although the number of informants in each experiment is still small, the results of the experiments are so far in favor of the base-generation analysis proposed in this work. Given the small number of informants, it may be too early to say that the predicted schematic asymmetries in question have been confirmed. When they indeed get further confirmation by the judgments by more informants and by the replication of the same results in more experiments of the same basic design, we will be in a position to say, with some confidence, that the properties of CM-Cleft discussed in this work are the reflections of the CS, and are revealing about some aspects of the language faculty, thereby making a step toward achieving the goal of generative grammar. 245 Bibliography Akmajian, Adrian (1970) Aspects of the grammar of focus in English, Doctoral dissertation, MIT, Cambridge, Mass. Baker, Mark (1996) The Polysynthesis Parameter, New York: Oxford University Press. Barss, Andrew (1984) "Chain Binding," ms. MIT, Cambridge. Barss, Andrew (1986) Chains and Anaphoric Dependence, On Reconstruction and Its Implications, Doctoral dissertation, MIT, Cambridge. Browning, Marguerite (1987) Null operator constructions, Doctoral dissertation, MIT, Cambridge, Mass. Chomsky, Noam (1970) "Deep structure, surface structure, and semantic interpretation," in R. Jakobson & S. Kawamoto, eds., Studies in General and Oriental Linguistics presented to Shiro Hattori, TEC, Tokyo 52-91. Chomsky, Noam (1977) "On Wh-movement," in P. W. Culicover, T. Wasow and A. Akmajian, eds., Formal Syntax, New York: Academic Press, 71-132. Chomsky, Noam (1986) Knowledge of Language, its Nature, Origin, and Use, Praeger, New York. Chomsky, Noam (1993) "A Minimalist Program for Linguistic Theory," in K. Hale and S.J. Keyser, eds., The View from Building 20, The MIT Press, Cambridge; also in (1995) The Minimalist Program, The MIT Press, Cambridge, pp.167-218 Chomsky, Noam (1995) The Minimalist Program. Cambridge, Massachusetts: The MIT Press. Chomsky, Noam (2007) "Biolinguistic Explorations: Design, Development, Evolution," International Journal of Philosophical Studies, 15:1, 1-21. Chung, Daeho (2006) "Toward a Unified Approach to Negative Polarity Triggered by Inherently Negative Predicates", Japanese/Korean Linguistics 14, 165-175. 246 Chung, Daeho and Hong-Keun Park (1998) "NPIs outside of negation scope," Japanese/Korean Linguistics 6: 415-435, Stanford: CSLI Publications Delahunty, Gerald (1982) Topics in the syntax and semantics of English cleft sentences, distributed by the Indiana University Linguistics Club, Bloomington, Indiana. Duhem, Pierre (1906/1954) The aim and structure of physical theory. Princeton: Princeton University Press. (The original publication in French in 1906, its original English translation in 1954, and its renewed edition in 1982. Emonds, Joseph (1976) A Transformational Approach to English Syntax: Root, Structure- Preserving, and Local Transformations, New York: Academic Press. Fauconnier, Gilles (1975) "Pragmatic scales and logical structure," Linguistic Inquiry 6-3: 353- 375 Feynman, Richard (1965/1994) The character of physical law. New York: The Modern Library. Fiengo, Robert (1974) Semantic Conditions on Surface Structure, Doctoral dissertation, MIT. Fiengo, Robert (1977) "On Trace Theory," Linguistic Inquiry 8: 35-62. Fukaya, Teruhiko and Hajime Hoji (1999) "Stripping and Sluicing in Japanese and Some Implications." A paper presented at the 18th West Coast Conference on Formal Linguistics, University of Arizona, April 9, 1999. Fukui, Naoki (1986) A Theory of Category Projection and its Applications, Doctoral Dissertation, MIT. Haraguchi, Shosuke (1973) Remarks on Right Dislocation in Japanese, ms., MIT. Hasegawa, Nobuko (1997) "A copula-based analysis of Japanese clefts: Wa-cleft and ga-Cleft." Report (1): Researching and Verifying an Advanced Theory of Human Language (COE Research), Kanda University of International Studies, Japan, 15-38. Hatakeyama, Yuji, Kensuke Honda and Kosuke Tanaka (2008) "Verb movement in Japanese revisited," Journal of Japanese Linguistics 24, 89-103. 247 Hayashishita, J.-R. (1997) "On the Scope Ambiguity in the Scrambling Construction in Japanese," ms., University of Southern California Hayashishita, J.-R. (1999) "Two Ways of Deriving Distributive Readings," in Proceedings of the 23 rd Annual Penn Linguistics Colloquium, pp. 201-216, University of Pennsylvania Working Papers in Linguistics Volume 6.1. Hayashishita, J.-R. (2000a) "Scope Ambiguity and 'Scrambling'," WCCFL 19, 204-217, Somerville, Massachusetts: Cascadilla Press. Hayashishita, J.-R. (2000b) "More on Two Ways of Deriving Distributive Readings," Japanese/Korean Linguistics vol. 9, pp.283-296, CSLI Publications. Hayashishita, J.-R. (2004) Syntactic and Non-Syntactic Scope, Doctoral dissertation, University of Southern California. Heycock, Caroline and Anthony Kroch (1996) "Identity, predication and connectivity in pseudoclefts." Paper presented at the 19th GLOW Colloquium, Athens, Greece. Heggie, Lorie (1988) The syntax of copula sentences, Doctoral dissertation, Los Angeles: University of Southern California. Higgins, Francis Roger (1979) The Pseudo-cleft Construction in English, New York: Garland Publishing. Hiraiwa, Ken (2003) "Eliminating PBC: Multiple Spell-Out, Scrambling and the Edge Operation," in The Proceedings of the 26th Penn Linguistics Colloquium. UPenn Working Papers in Linguistics 9:1. 89-103. Philadelphia, PA.: University of Pennsylvania. Hiraiwa, Ken (2005) "Indeterminate-Agreement: Some Consequences for the Case System," in Ken Hiraiwa and Joey Sabbagh, eds., MITWPL #50, 93-128. Cambridge, MA.: MITWPL. Hiraiwa, Ken and Shinichiro Ishihara (2002) "Missing Links: Cleft, Sluicing, and "No da" Construction in Japanese," MIT Working Paper in Linguistics 43, 35-54. Hoji, Hajime (1985) Logical Form Constraints and Configurational Structures in Japanese, Doctral Dissertation, University of Washington. 248 Hoji, Hajime (1987) "Japanese Clefts and Reconstruction/Chain Binding Effects." A talk presented at WCCFL VI held at University of Arizona, 3/21/1987. Hoji, Hajime (1990) "Theories of Anaphora and Aspects of Japanese Syntax," ms., University of Southern California. Hoji, Hajime (1991) "Raising-to-object, ECM and the major object in Japanese." A talk given at Japanese Syntax Workshop at University of Rochester. Hoji, Hajime (2003a) "Falsifiability and Repeatability in Generative Grammar: A Case Study of Anaphora and Scope Dependency in Japanese," Lingua 113/4-6, 377-446. Hoji, Hajime (2003b) Surface and Deep Anaphora, Sloppy Identity, and Experiments in Syntax, in A. Barss, and T. Langendoen, eds., Anaphora: A Reference Guide, Blackwell. Hoji, Hajime (2009) "A foundation of generative grammar as an empirical science," ms., University of Southern California. Hoji, Hajime (2010a) "Hypothesis testing in generative grammar: Evaluation of predicted schematic asymmetries," Journal of Japanese Linguistics 26 Hoji, Hajime (2010b) "Language Faculty Science as an Exact Science." A talk presented at Sophia University, 12/10/2010. Hoji, Hajime (2010c) "Language Faculty Science as an Exact Science." A lecture note at University of Southern California, 12/2/2010. Hoji, Hajime (2011) "Language Faculty Science as an Exact Science: Empirical Illustration." A talk presented at Kyoto University, 1/16/2011. Hoji, Hajime (to appear) "Reconstruction Effects in Passive and Scrambling in Japanese," Papers from JK conference 13. Hoji, Hajime and Ayumi Ueyama (1998) "Resumption in Japanese," ms., University of Southern California. 249 Hoji, Hajime and Ayumi Ueyama (2003) "Resumption in Japanese: A Preliminary Study." A talk presented at WCCFL 2003 held at University of Arizona, 3/26/2003. Huang, C.-T. James (1982) Logical Relations in Chinese and the Theory of Grammar, Doctoral dissertation, MIT, CambridgeInoue, Kazuko (1976) Henkeibunpoo to Nihongo (Transformational Grammar and Japanese), Japan: Taishukan. Jaeggli, Osvaldo (1986) Topics in Romance Syntax , Foris, Dordrecht. Kataoka, Kiyoko (2004) Nihongo Hiteibun-no Koozoo: Kakimaze-bun to Hiteekoohyoogen (Syntactic Structure of Japanese Negative Sentences: Scrambling Construction and Negation-sensitive Elements), Doctoral dissertation, Fukuoka, Japan: Kyushu University. Kataoka, Kiyoko (2006) Nihongo Hiteibun-no Koozoo: Kakimaze-bun to Hiteekoohyoogen (Syntactic Structure of Japanese Negative Sentences: Scrambling Construction and Negation-sensitive Elements), Kurosio Publishers, Tokyo. Kataoka, Kiyoko (2007) "Two Types of Neg-Sensitive Elements," ms., The National Institute for Japanese Language. Kikuchi, Akira and Daiko Takahashi (1991) "Argument and small clauses," in H. Nakajima and S. Tonoike, eds., Topics in Small Clauses, Japan: Kuroshio Publishers, 75-105. Kitagawa, Yoshihisa (1990) "Anti-Scrambling," ms., University of Rochester. (A paper presented at the Tilburg University Workshop on Scrambling, 10/19/90.) Kizu, Mika (2005) Cleft Constructions in Japanese Syntax, New York: Palgrave Macmillan. Koizumi, Masatoshi (1995) Phrase Structure in Minimalist Syntax, Doctoral dissertation, MIT. Koizumi, Masatoshi (2000) "String vacuous over verb raising," Journal of East Asian Linguistics 9: 227-285. Kuno, Susumu (1973) The structure of the Japanese Language, Cambridge, Massachusetts: The MIT Press. Kuno, Susumu (1978) Danwa-no Bunpoo [Grammar of Discourse], Tokyo: Taishuukan. 250 Kuno, Susumu, Ken-ichi Takami, & Yuru Wu (1999) "Quantifier Scope in English, Chinese, and Japanese," Language 75, pp.63-111. Kuroda, S.-Y. (1969/1970) "Remarks on the Notion of subject with Reference to Words like Also, Even or Only," Annual Bulletin, Research Institute of Logopedics and Phoniatrics, University of Tokyo, vol.3, pp.111-129, and vol.4, pp.127-152; Also in Kuroda (1992) Japanese Syntax and Semantics, Kluwer Academic Publishers, Dordrecht, pp.78-113. Kuroda, S.-Y. (1988) "Whether We Agree or Not: A Comparative Syntax of English and Japanese," Linguisticae Investigationes 12, 1-47. (Reproduced as Kuroda, 1992: chap. 10. (315-357).) Kuroda, S.-Y. (1992) "Judgment Forms and Sentence Forms," in Japanese Syntax and Semantics, Kluwer Academic Publishers, Dordrecht, pp.13-77 Kuroda, S.-Y (1994) "Quantifier Scope in Japanese," ms., University of California at San Diego. Kuwabara, Kazuki (1995) "Some problems of sluiced clause in Japanese," ms., Kanda University of International Studies, Japan. Kuwabara, Kazuki (1996) "Multiple Wh-phrases in elliptical clauses and some aspects of clefts with multiple foci," in M. Koizumi, M. Oishi and U. Sauerland, eds., Formal Approaches to Japanese Linguistics 2: The MIT Working Papers in Linguistics 29, MIT, Cambridge, Massachusetts, 97-116. Ladusaw, William A. (1979) Polarity sensitivity as inherent scope relation, New York and London: Garland Publishing, Inc. Lakatos, Imre (1970) "Falsification and methodology of scientific research programmes," in I. Lakatos and A. Musgrave, eds., Criticism and the growth of knowledge, 91-195. Cambridge: Cambridge University Press. Reprinted as Lakatos (1978: chapter 1). Lakatos, Imre (1978) "The methodology of scientific research programmes," in J. Worrall and G. Currie, eds., Philosophical papers volume 1, Cambridge: Cambridge University Press. Lasnik, Howard, and Mamoru Saito (1984) "On the Nature of Proper Government," Linguistic Inquiry 15, 235-289. 251 Masuoka, Takashi (1989) "Modaliti-no-koozoo to Gimon/Hitei-no-Sukoopu (The Structure of Modality and the Scope of Question and Negation)," in Y. Nitta and T. Masuoka, eds., Nihongo-no Modalitii (Modality in Japanese), Kurosio Publishers, Tokyo, 193-210. Müller, Gereon (1993) On Deriving Movement Type Asymmetries, Doctoral dissertation, Universität Tübingen. Müller, Gereon (1996) "A Constraint on Remnant Movement," Natural Language and Linguistic Theory 14-2, 355-407. Nishiyama, Kunio (1998) The morphosyntax and morphophonology of Japanese predicates, Doctoral dissertation, Cornell University. Noda, Hisashi (1996) Wa to Ga (Wa and Ga), Kuroshio Publishers, Japan. Otani, Kazuyo and Whitman, John (1991) "V-raising and VP-ellipsis," Linguistic Inquiry 22: 345-358. Pinkham, Jessie and Jorge Hankamer (1975) "Deep and shallow clefts," Chicago Linguistc Society, Vol. 11: 429-450. Pollock, Jean-Yves (1989) "Verb Movement, Universal Grammar, and the Structure of IP," Linguistic Inquiry, Volume 20, Number 3, 365-424 Reinhart, Tanya (1983) Anaphora and Semantic Interpretation, The University of Chicago, Chicago. Rizzi, Luigi (1997) The fine structure of the left periphery, in L.Haegeman, ed., Elements of Grammar: Handbook in Generative Syntax, Dordrecht: Kluwer, 281–337. Rochemont, Michael (1986) Focus in Generative Grammar, Amsterdam: John Benjamins. Saito, Mamoru (1985) Some Asymmetries in Japanese and Their Theoretical Implications, Doctoral dissertation, MIT, Cambridge. Sells, Peter (2006) "Interactions of negative polarity items in Korean," Harvard Studies in Korean Linguistics 11: 724-737, Harvard University 252 Sells, Peter and Kim, Shin-Sook (2006) "Korean NPIs Scope Over Negation," Language Research, 42 . pp. 275-297. Takahashi, Daiko (2006) "Apparent Parasitic Gaps and Null Arguments in Japanese," Journal of East Asian Linguistics 15, 1-35. Takano, Yuji (2002) "Surprising constituents," Journal of East Asian Linguistics 11: 243-301. Tsuboi, Yukiko (2006) "Raising or ECM is not an option for RTO constructions in Japanese," ms., University of Southern California. Ueyama, Ayumi (1997) "Scrambling in Japanese and Bound Variable Construal," ms., University of Southern California. Ueyama, Ayumi (1998) Two Types of Dependency, Doctoral dissertation, Los Angeles: University of Southern California. Distributed by GSIL publications, Los Angeles: University of Southern California. Ueyama, Ayumi (2003) "Two Types of Scrambling Constructions in Japanese," in: A. Barss & T. Langendoen (eds.), Anaphora: A Reference Guide, Cambridge: Blackwell. Ueyama, Ayumi (2010) "Model of judgment making and hypotheses in generative grammar," in S. Iwasaki, H. Hoji, P. Clancy and S. Sohn, eds., Japanese/Korean linguistics 17, 27-47. Stanford, CA: CSLI (available at: http://www.gges.org/hoji/research/hp-Ayumi.cgi). Ueyama, Ayumi and Hoji Hajime (2001) "Resumption in Japanese." A talk presented to Workshop on Empty Categories and Ellipsis (UNICAMP) Aug., 2001. Williams, Edwin (1980) "Predication," Linguistic Inquiry 11, pp.203-238.

Abstract (if available)

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

Binding and scope dependencies with 'floating quantifiers' in Japanese

PDF

Sluicing and stripping in Japanese and some implications

PDF

Functional categories: the syntax of DP and DegP

PDF

Towards a correlational law of language: three factors constraining judgement variation

PDF

Agreement on the left edge: the syntax of left dislocation in Spanish

PDF

A unified syntactic account of Mandarin subject nominals

PDF

Existential constructions: a syntactic predication approach

PDF

Towards the unity of movement: implications from verb movement in Cantonese

PDF

Dominant language influence in acquisition and attrition of the Chinese reflexive ziji By Chinese-English bilinguals

PDF

Factors affecting relative clause processing in Mandarin

PDF

The theory of empty noun in Chinese: with special reference to the right node raising construction

PDF

Rethinking sources for errors with case marking related to transitive predicates in learning Korean

PDF

A study on the use of lexical collocations of Korean heritage learners: identifying the sources of errors

PDF

On the structure of English subtitles in select German films

PDF

Software security economics and threat modeling based on attack path analysis; a stakeholder value driven approach

PDF

A corpus-based discourse analysis of Korean grammatical constructions: Focus on the multifold functions and meanings of the pragmatic construction e kaciko

PDF

Detrimental effects of dental encroachment on secondary alveolar bone graft outcomes in the treatment of patients with cleft lip and palate: a cone-beam computed tomography study

PDF

A reduplicative analysis of sentence modal adverbs in Spanish

PDF

The speaker's choice of using object-marking in informal spoken Korean

PDF

The interplay between the state and civil society: a case study of honor killings in Turkey

Asset Metadata

Creator Miura, Yasuhiko (author)

Core Title The syntax of cleft constructions in Japanese: A base-generation analysis

School College of Letters, Arts and Sciences

Degree Doctor of Philosophy

Degree Program East Asian Languages and Cultures

Publication Date 05/04/2011

Defense Date 03/11/2011

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

Tag bound variable anaphora (BVA),cleft,FOCUS,NPI,OAI-PMH Harvest,presupposition,resumption

Language English

Contributor Electronically uploaded by the author (provenance)

Advisor Hoji, Hajime (committee chair), Li, Audrey Yen-Hui (committee member), Saltarelli, Mario (committee member), Simpson, Andrew (committee member)

Creator Email yasmiura@hotmail.com,ymiura@usc.edu

Permanent Link (DOI) https://doi.org/10.25549/usctheses-m3905

Unique identifier UC1218040

Identifier etd-Miura-4514 (filename),usctheses-m40 (legacy collection record id),usctheses-c127-474495 (legacy record id),usctheses-m3905 (legacy record id)

Legacy Identifier etd-Miura-4514.pdf

Dmrecord 474495

Document Type Dissertation

Rights Miura, Yasuhiko

Type texts

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

Repository Name Libraries, University of Southern California

Repository Location Los Angeles, California

Repository Email cisadmin@lib.usc.edu