WH-CONSTRUCTIONS IN JAPANESE
A Dissertation Presented by JUNKO SHIMOYAMA
Submitted to the Graduate School of the University of Massachusetts Amherst in partial fulfillment of the requirements for the degree of DOCTOR OF PHILOSOPHY February 2001 Department of Linguistics
© Copyright by Junko Shimoyama 2001 All Rights Reserved
WH-CONSTRUCTIONS IN JAPANESE
A Dissertation Presented by JUNKO SHIMOYAMA
Approved as to style and content by:
Kyle B. Johnson, Co-Chair
Angelika Kratzer, Co-Chair
F. Roger Higgins, Member
Chisato Kitagawa, Member
Elisabeth Selkirk, Department Head Linguistics
ACKNOWLEDGMENTS
I am most grateful to the members of my dissertation committee: Kyle Johnson, Angelika Kratzer, Roger Higgins and Chisato Kitagawa. I have been very fortunate to have Kyle and Angelika as the co-chairs of the committee. I thank them for their patient guidance and encouragement during all these years. Sitting in Kyle's office and talking to him always let me see what I was doing more clearly and what to look for next. Angelika's enthusiasm has been a great source of encouragement for me, especially at those times when I had Japanese data that were so different from English that I didn't know what to do with them in semantic terms. Roger has served in all the committees I had in Amherst, including the one for my phonology generals paper, and provided me with generous guidance. I have learned a lot from discussions with Chisato. I have shared with him the great fun of thinking about our own language. A few more people who are not on the committee deserve special thanks. Barbara Partee has given me many helpful comments and encouragement along the way. What is reported in chapter 3 started out in weekly meetings with Hagit Borer, who is now at USC. I thank Hagit for her generous advising. Sigrid Beck got me interested in the syntax and semantics of questions when she taught a proseminar on the topic during her visiting semester at UMass. Ever since then, Sigrid's comments and interest have helped me greatly. I am very grateful to the other faculty members at the department, from whom I have learned how to become a linguist: Lyn Frazier, John Kingston, John McCarthy, Tom Roeper, Lisa Selkirk, Peggy Speas and Ellen Woolford. I would also like to thank our visiting faculty, Orin Percus, Norvin Richards and Yael Sharvit for discussions. Many thanks to the department staff, Kathy Adamczyk and Lynne Ballard, who are responsible for making everything work so smoothly. Thanks to all the people who have contributed to making the department what it is, and those who recommended that I come to UMass.
iv
Many thanks go to my fellow students. I thank my classmates for having gone through the first year and beyond with me. They are, Isadora Cohen, Kiyomi Kusumoto, Elliott Moreton, Bernhard Schwarz, and Wayne Worley. I would in particular like to thank Bernhard for many conversations, linguistic and non-linguistic, over many meals we shared on Linden Street. I also wish to thank, among others, John Alderete, Ana Arregui, Elena Benedicto, Katy Carlson, Maria Nella Carminati, Mako Hirotani, Bart Hollebrandse, André Isaak, Caroline Jones, Ji-yung Kim, Cecilia Kirk, Winnie Lechner, Jo-wang Lin, Maribel Romero, Jen Smith, Mariko Sugahara, Pius Tamanji, Mike Terry, Satoshi Tomioka, and Elisabeth Villalta. Thanks to Tohru Noguchi for his hospitality when I first arrived in Amherst. I am very happy that my years in the Amherst area overlapped with the presence of Shin Watanabe and of Hisao Tokizaki in the area. I thank Shin and Hisao for friendship and for the fun we had together. Thanks also go to Masumi Matsumoto for her company on my first New Year's Day away from Japan. Many other people have contributed to the research reported in this dissertation. Comments from Jun Abe, Veneeta Dayal, Alexander Grosu, Irene Heim, Richard Larson and three anonymous reviewers for Journal of East Asian Linguistics were particularly helpful. I thank Takashi Toyoshima for sending me his 1996 manuscript on questions. It turned out that some of the ideas that I was pursuing at the time were very similar to what he had discussed in his manuscript. The e-mail discussions with him that followed were very encouraging and fun. I also had fruitful conversations with Paul Hagstrom and Masao Ochi on various issues in Japanese syntax and semantics, as well as with Masaaki Fuji, Hideki Maki, Toshi Ogihara, Uli Sauerland, Akira Watanabe and Tomo Yoshida. I also benefited from helpful comments from audiences at NELS 29 at the University of Delaware, MIT/UConn/UMass Semantics Workshop I, Cornell University, University of Connecticut, UMass Semantics reading group, the LSA annual meeting in Los Angeles, and Kanda University of International Studies. I am very grateful to those who provided judgments: Yoshi Dobashi, Kiyomi Kusumoto, Toshi Ogihara, Hisao Tokizaki, Satoshi
v
Tomioka, Takashi Toyoshima, Satoshi Shimoyama, Mariko Sugahara, Akira Watanabe, Shin Watanabe, and Kyoko Yamakoshi. I feel very fortunate to have spent the academic year 1999-2000 at Cornell University. Many thanks go to my colleagues and students who made my stay an enjoyable and interesting one. I would particularly like to thank Molly Diesing and Sally McConnell-Ginet for always being there for help, and Elisabeth Rieken and Ayako Tsuchida for good company. I also thank Lorey Pro and the volleyball crowd, as well as the hills, waterfalls and gorges in Ithaca for having kept me in shape. I thank Noriko Terazu Imanishi for introducing me to the field, as well as the many people back home, teachers and colleagues, who nurtured my interest in Linguistics. The first two years of my study at UMass was supported by a Fulbright Graduate Study Grant. I wish to thank the Japan-United States Educational Commission in Tokyo and the Institute of International Education in New York. My first year at UMass was also supported by a Rotary International Graduate Scholarship, for which I am grateful. I would also like to thank the University of Massachusetts for Teaching Assistantships. For her professional assistance, I am grateful to Pat Vokbus at the International Programs Office at UMass. Many people have helped me go through these years directly or indirectly. Thanks to, among others, Maryam Bekrani, Masahiko Fujiwara, Shizuko Fukuma, Buzz and Sue Kandler, Yuko Kumano, Yukiko Morita, Nagisa Moritoki, Karyn Saemann, Gertrud and Hermann Schwarz, Atsuko Sugio, Yukie Tamura, and Haruko Tanaka. Finally, my greatest gratitude goes to my parents Shimoyama Teruko and Shimoyama Tsukasa for their love and support. Many thanks also go to my sister Yoko, my brother Satoshi, Yoko's husband Takenori, my nephews Yuuya and Takumi, my grandmothers Ishii Yuiko and Shimoyama Setsuko, my aunts Gamô Ikuko, Gotô Muneko, Kaji Takuko and Suwa Kakuko. My special thanks go to my late great-aunt Shimoyama Yoshiko, who sent me letters transcribed by her nurse from her bed.
vi
ABSTRACT WH-CONSTRUCTIONS IN JAPANESE FEBRUARY 2001 JUNKO SHIMOYAMA, B.A., OCHANOMIZU UNIVERSITY M.A., OCHANOMIZU UNIVERSITY Ph.D., UNIVERSITY OF MASSACHUSETTS AMHERST Directed by: Professor Kyle B. Johnson and Professor Angelika Kratzer
This dissertation investigates the syntax and semantics of two types of whconstructions in Japanese. One is the construction that involves so-called "indeterminate phrases", such as dare 'who' and nani 'what'. The other is what is known as the "internally headed relative clause construction". In both of these wh-constructions, association of two non-local positions in a sentence takes place. The association has been most commonly analyzed in terms of invisible movement. This dissertation reexamines this assumption. Indeterminate phrases differ from English wh-phrases in that they occur in the universal construction as well. Previous studies have put primary focus on indeterminate phrases in the interrogative construction, and tried to extend their analysis to the universal construction. Chapter 2 shows that examining the universal construction provides a new window on the nature of association between indeterminate phrases and the question particle ka and the universal particle mo. As a consequence of the proposed semantics of the universal construction, a uniform analysis of the two constructions emerges that does not involve English-type wh-movement. The fact that the association between indeterminate phrases and ka/mo can take place across islands for movement, except for wh-islands, is a long-standing puzzle . This presents a challenge to any movement theory. The analysis presented in this chapter allows a switch of perspective on this issue, and the puzzle is shown to follow from the interpretative process. vii
Chapter 3 examines the internally headed relative clause construction, with particular focus on new data that involve quantificational NPs and indeterminate phrases. The data provide arguments for representations in which the internal head remains internal at LF. Furthermore, it is shown that the kind of "argument sharing" observed in this construction is better analyzed in terms of anaphora rather than in terms of movement. Evidence is presented that the interpretation of this construction involves E-type anaphora in particular. An explicit mechanism for compositional interpretation is proposed, which also derives a restriction on possible internal heads. This study has the cross-linguistic implication that the constructions called "internally headed relative clauses" in various languages do not form a homogeneous class.
viii
TABLE OF CONTENTS
Page ACKNOWLEDGMENTS .........................................................................iv ABSTRACT ........................................................................................vii Chapter 1.
INTRODUCTION ......................................................................... 1
2.
THE SYNTAX AND SEMANTICS OF INDETERMINATE PHRASE CONSTRUCTIONS ....................................................................... 9 2.1 Introduction............................................................................. 9 2.1.1 Indeterminate Phrases in Japanese ............................................ 9 2.1.2 A Need for a Uniform Account .............................................. 13 2.1.3 A Uniform Account Based on a Particular Interpretation of the Locality Effects ............................................................ 17 2.1.4 The Island Puzzle and the Pied-Piping Analysis........................... 19 2.1.5 The Syntax, Semantics, and Their Mapping................................ 23 2.2 The Interpretation of the Mo-Construction ........................................ 25 2.2.1 Universal Quantification over Pairs of Individuals ........................ 25 2.2.2 The Partitive and Weak Readings............................................ 27 2.2.3 The Embedded Restrictor View.............................................. 29 2.2.4 Apparent Support for the Embedded Restrictor View .................... 31 2.3 A New Way of Looking at the Semantics of the Mo-construction............. 36 2.3.1 The Direct Restrictor View ................................................... 36 2.3.2 Focus and the Domains of Quantification .................................. 39 2.3.3 Indeterminate Phrases as Introducing Free Variables: The Ka-Construction .......................................................... 41 2.3.4 Extension to the Direct Restrictor View of the Mo-Construction ....... 43 2.3.5 Mo as a Cross-Categorial Universal Quantifier ............................ 49 2.4 The Direct Restrictor View and the Island Puzzle................................ 53 2.4.1 The Island Puzzle .............................................................. 53 2.4.2 The Apparent Wh-Island Effect as a Property of Unselective Binding .......................................................................... 58 2.4.3 Unselective Binding and Locality ........................................... 59 2.5 An Alternative: Indeterminate Phrases as Denoting Sets of Individuals ...... 62
ix
2.6 A Further Look at the Embedded Restrictor View ............................... 65 2.6.1 Anaphora Facts ................................................................. 65 2.6.2 The Mapping from Syntax to Semantics in the Embedded Restrictor View................................................................. 67 2.6.2.1 The LF of the Ka-Construction Extended to the Mo-Construction ...................................................... 67 2.6.2.2 Other Possible LF Representations of the Mo-Construction .... 74 2.6.2.3 Summary ............................................................... 75 2.6.3 The Nature of the Restriction on the Domain of Universal Quantification................................................................... 76 2.6.3.1 Different Predictions in the Two Views ........................... 76 2.6.3.2 Burdens on Presupposition Accommodation...................... 78 2.6.3.3 Further Questions ..................................................... 83 2.7 Remaining Questions: Kadooka 'whether'-Questions ............................ 85 2.7.1 Kadooka 'whether'-Island Effect ............................................. 85 2.7.2 Ka, Doo, and Ka................................................................ 87 2.8 Summary ............................................................................... 92 3.
INTERNALLY HEADED RELATIVE CLAUSES IN JAPANESE AND E-TYPE ANAPHORA ................................................................... 93 3.1 Introduction............................................................................ 93 3.2 Does the Internal Head Move at LF or Stay In-Situ? ............................ 95 3.2.1 The LF Head Raising Analysis............................................... 95 3.2.2 The Truth-Conditional Difference between IHRs and EHRs ............ 96 3.2.3 The Position of the Internal Head at LF and Scope Phenomena......... 98 3.2.3.1 Background: The Scope of Quantifiers and Scrambling......... 98 3.2.3.2 The Relative Position of the Internal Head with Respect to Material in the Matrix Clause ..................................... 101 3.2.3.3 The Relative Position of the Internal Head with Respect to Other Internal Material ............................................. 103 3.3 E-type Anaphora in the IHR....................................................... 105 3.3.1 Bound Variable Anaphora and Referential Anaphora .................. 105 3.3.2 The Need for E-type Anaphora............................................. 107 3.3.3 Further Arguments for the E-type Analysis: The IHR and Wh-Questions................................................................. 108 3.3.3.1 The E-type Interpretation of the Anaphoric Element........... 109 3.3.3.2 A Contrast in EHR-IHR Pairs that Contain Wh-phrases ...... 110
x
3.4 The LF and Compositional Interpretation of IHRs............................. 116 3.4.1 Hoshi's (1995) Proposal ..................................................... 116 3.4.2 Amended E-type Analysis .................................................. 120 3.5 Consequence: The So-Called "Indefiniteness" Restriction.................... 128 3.5.1 The Pro Analysis............................................................. 128 3.5.2 Recovering a Property ....................................................... 129 3.5.3 IHR-DPs in the Direct Object Position.................................... 132 3.6 The Apparent Lack of Pragmatic Variability in the IHR ...................... 134 3.6.1 A Predicate Recovered from VP ........................................... 135 3.6.2 Predicative Internal Heads and Predicate Modification................. 137 3.6.3 The Relevancy Condition ................................................... 139 3.7 Concluding Remarks ............................................................... 143 BIBLIOGRAPHY................................................................................ 146
xi
CHAPTER 1 INTRODUCTION
The sentences in (1) illustrate two types of wh-constructions in English. (1a) is an example of a content question (wh-question), and (1b) is an example of a relative clause construction. These constructions are most commonly assumed to involve overt movement of wh-phrases, which book and which in (1), as shown in (2a) and (2b).
(1) a. Which book did Yoko read? b. The book which Yoko read was interesting. (2) a. [which book]1 did you read t1 b. the book which1 Yoko read t1
The surface syntactic representations of these constructions are suitable for deriving their standard semantic representations compositionally. In wh-questions, we can assume, for instance, that wh-phrases introduce an existential quantifier that binds variables created by their movement, and that there is an interrogative operator that forms a set of propositions, along the lines of Hamblin (1973) and Karttunen (1977). In the relative clause construction in (2b), we can assume that the relative pronoun which abstracts over the free variable created by its movement. The resulting predicate modifies the noun book (see Quine 1960). This dissertation is a study of the syntax and semantics of two types of whconstructions in Japanese that are related to the English wh-constructions in (1). One is the construction that involves so-called "indeterminate phrases", such as dare 'who' and nani 'what'. The other is what is known as the "internally headed relative clause construction". What these two constructions in Japanese have in common is that no overt movement relation holds between the two positions that correspond to those in English
1
related via movement as shown in (2). These two constructions in Japanese have been most commonly analyzed in terms of invisible movement. Chapter 2 examines the syntax and semantics of constructions that involve indeterminate phrases. An example of a wh-question in Japanese is given in (3). Dono hon 'which book' occurs in the position in which kono hon 'this book', for example, would have occurred in the corresponding affirmative sentence.
(3) Yoko-wa dono hon-o
yomimasita ka?
Yoko-Top which book-Acc read
Q
'Which book did Yoko read?'
As I mentioned above, a common assumption has been that the association of dono hon 'which book' and the question particle ka is mediated through covert movement at LF (see Huang 1982, Nishigauchi 1986, 1990, Lasnik and Saito 1992, Watanabe 1992a, b and Richards 1997, among many others). Indeterminate phrases like dono hon 'which book' are different from English whphrases in that the former can show up in the universal context as well, as shown in (4).
(4) a. Yoko-wa dono hon-mo
yonda.
Yoko-Top which book-MO read 'Yoko read every book.' b. [[Dono hon-o
yonda] kodomo]-mo yoku nemutta.
which book-Acc read
child
-MO well slept
'For every book x, the child who read x slept well.'
Previous research has put primary focus on indeterminate phrases in the interrogative context, and tried to extend their analysis to the universal context. I will show in chapter
2
2 that examining the universal context, which does not have a direct counterpart in English, provides a new window on the nature of association between indeterminate phrases and the particles ka and mo. It has been well-known that in English, wh-phrases cannot be overtly extracted out of certain islands (see Ross 1967, Chomsky 1973, 1977). The examples in (5) illustrate the effect of Wh-island, Complex NP island and Adjunct island, respectively.
(5) a. *What did Yoko wonder where Satoshi bought? b. *Who did Julia eat carrot pasta that Mike made for? c. *Who did Karyn get upset after Eric called?
The Japanese example in (6) shows that indeterminate phrases are also subject to whisland effect. Neither of the two indeterminate phrases in situ can take matrix scope, and the matrix question is interpreted as a yes-no question, rather than as a wh-question.
(6) Taro-wa [Yamada-ga
dare-ni nani-o
okutta ka] tazunemasita ka?
Taro-Top Yamada-Nom who-Dat what-Acc sent a.
Q asked
Q
'Did Taro ask what Yamada sent to whom?'
b. ?* 'Whox did Taro ask what Yamada sent to x?' c. *
'Whatx did Taro ask to whom Yamada sent x?'
d. ?* 'Whox did Taro ask whether Yamada sent what to x?'
This kind of data naturally led many authors to believe that the unavailability of the interpretations in (6b, c, d) has the same source as the ungrammaticality of the English example in (5a). More specifically, the presence of wh-island effect is taken to indicate that movement is involved in the association of indeterminate phrases and the question
3
particle ka (see, for example, Nishigauchi 1986, 1990, Pesetsky 1987, Watanabe 1992a, b, von Stechow 1996 and Richards 1997, among many others). There is, however, a long-standing puzzle about the locality effects observed in Japanese, which I will call "the island puzzle". As illustrated in (7), Japanese whquestions do not show any effect of Complex NP island or Adjunct island.
(7) a. Taro-wa [[dare-ga
katta] mochi]-o
tabemasita ka?
Taro-Top who-Nom bought rice cake -Acc ate
Q
'Whox did Taro eat rice cakes that x bought?' b. Taro-wa [dare-ga
kita-kara]
kaerimasita ka?
Taro-Top who-Nom came-because left
Q
'Whox did Taro leave because x came?'
Any movement analysis would have to posit a special mechanism so that the unexpected behavior of indeterminate phrases when they occur within Complex NP and Adjunct islands is circumvented. Nishigauchi (1986, 1990) made an influential proposal in this area, which subsequent movement analyses build upon (see also Choe 1987). Nishigauchi proposed that the apparent lack of Complex NP and Adjunct island effects in (7) is due to the fact that these islands somehow count as [+wh]. This makes the whole islands pied-pipe and undergo movement at some invisible level where island constraints are at work (his LF). Thus the indeterminate phrases embedded inside these islands do not get extracted out of them. Von Stechow (1996) pointed out that the LF representations that result from Nishigauchi's pied-piping mechanism are not suitable for deriving the desired semantics (see also Chomsky 1977). However, von Stechow (1996) still adopts Nishigauchi's piedpiping mechanism, only at a different level of representation. Other authors who argue
4
for or assume the movement analysis also rely on Nishigauchi's pied-piping analysis or variants of it. The challenge for the pied-piping analysis is not just to come up with a mechanism that makes pied-piping of certain islands possible, but also, and perhaps more importantly, to let the correct distinction between which island is pied-pipable and which island is not follow from the analysis. We would want to look for an analysis in which the exceptionality of the Complex NP islands and Adjunct islands follow from something inherent in the analysis. This study approaches this familiar topic from a slightly different direction from what is found in the literature, namely, by first looking at the universal context. This is worthwhile for the following reasons. As long as the analysis of wh-questions in Japanese is modeled after English, as in the movement analysis, questions about their semantics and syntax-semantics mapping that are specific to Japanese do not arise. This is because one would expect that any plausible analysis of English wh-questions would carry over to the Japanese cases. Paying attention to the universal context, which does not have a direct counterpart in English, will provide an insight in the nature of the association between indeterminate phrases and the particles ka and mo that we might overlook if we only compared the wh-questions in English and Japanese. I will propose a new way of looking at the semantics of the universal construction in (4). This leads us to a uniform analysis of the indeterminate phrases in interrogative and universal contexts in which the association between indeterminate phrases and the particles does not involve movement. The new analysis also allows a switch of perspective on the long-standing island puzzle. The fact that only wh-island effect shows up is expected in the analysis. Chapter 3 examines the syntax and semantics of the internally headed relative clause construction in Japanese. In the familiar type of relative clause construction in Japanese, the head noun follows a relative clause, as shown in (8). Unlike English, there
5
is no overt relative pronoun. What corresponds to the trace in the English example in (2b) is indicated by a gap (ø).
(8) Yoko-wa [[Taro-ga
sara-no
ue-ni ø oita] keeki]-o
Yoko-Top Taro-Nom plate-Gen on-Loc put
tabeta.
cake -Acc ate
'Yoko ate a piece of cake which Taro put on a plate.'
Despite these surface differences from English, it is not difficult to imagine how appropriate semantic interpretations can be derived. We may, for example, assume that a relative pronoun that is phonologically empty creates a predicate by abstracting over the variable introduced in the position of the gap. The example in (9) presents a more interesting and challenging case. (9) is an example of what is often called the internally headed relative clause construction.
(9) Yoko-wa [[Taro-ga
sara-no
ue-ni
keeki-o
oita] -no] -o
tabeta.
Yoko-Top Taro-Nom plate-Gen on-Loc cake-Acc put -NM -Acc ate 'Taro put a piece of cake on a plate and Yoko ate it.'
In (9), keeki 'cake' occupies the position that a gap occupies in (8). The embedded clause in (9) looks just like a simple sentence, rather than a relative clause. It is then followed by the nominalizer (NM) -no, and the whole constituent receives the accusative marker -o. Keeki 'cake', the object of the embedded verb oita 'put', seems to serve as the object of the matrix verb tabeta 'ate' as well. A question arises as to how the underlined internal head in (9) is understood to play the dual role of being an argument of the embedded verb and an argument of the matrix verb at the same time. Another question is how the internally headed relative
6
clause construction in (9) relates to the more familiar, externally headed relative clause construction in (8) or the English relative clause construction in (1b). One might hypothesize that the semantic head keeki 'cake' in (9) raises at an invisible level, and that the internally headed relative clause is more or less assimilated to its external counterpart (see Ito 1986 and Watanabe 1992a, b). A theory that assumes movement of the internal head should make a certain set of predictions. First, the theory would predict that the two types of relative clause constructions share truth conditions. I will show that this is not the case. Second, one might expect for internal heads not to occur within islands for movement. The relevant data, unfortunately, are very subtle, with authors reporting conflicting judgments (see discussions in Ito 1986, Watanabe 1992a,b, Hoshi 1995 and Kuroda 1999). Therefore I will not focus on island effects in this construction. Finally, the movement theory would make predictions about the behavior of the raised head with respect to scope. In particular, if a quantificational internal head occupies the external head position at LF, or the position in which the relative pronoun which occurs in English example (1b), one should expect that the internal head takes wider scope than other scope bearing elements in the internally headed relative clause. I will examine the behavior of quantificational internal heads and provide evidence for representations in which they remain internal at LF. This leads me to conclude that the interpretation of this construction should be mediated by some element in the matrix clause that is anaphorically related to the internal head. The next question to ask is what kind of anaphoric relation is involved in the construction. One view assumes that there is a phonologically null pronoun, pro, in the matrix clause (Mihara 1994a,b, Murasugi 1994, Kitagawa 1996). It will be shown that this theory overgenerates. Evidence will be presented that the interpretation of this construction involves E-type anaphora. This provides further support for Hoshi (1995). An explicit mechanism of how the semantic interpretation of the construction is composed of its parts will be spelled out. It assumes proforms that recover salient
7
properties from the context, rather than all-purpose pros. This has a nice consequence in deriving a certain restriction on possible internal heads.
8
CHAPTER 2 THE SYNTAX AND SEMANTICS OF INDETERMINATE PHRASE CONSTRUCTIONS
2.1 Introduction 2.1.1 Indeterminate Phrases in Japanese The Japanese words in (1) are polarity items in the sense that they should always co-occur with the particle ka or mo.
(1) dare
'who'
doko
'where'
nani
'what'
itu
'when'
dore
'which (one)'
naze
'why'
dono
'which' (Det)
doo
'how'
When these words occur in the scope of the question particle ka, they take on interrogative meanings as shown in (2).1
(2) a. Dare-ga
denwasi-masi-ta ka?
who-Nom call-Pol-Past
Q
'Who called?' b. Taro-wa [dare-ga
denwasi-ta ka] siritagatteiru.
Taro-Top who-Nom call-Past
Q want_to_know
'Taro wonders who called.'
1Direct
questions that contain the particle no, which is often described as another question marker, can be viewed as a reduced form of no da/desu ka 'NO Cop Q'. There are some semantic differences between questions with or without no. See Sugahara (1998) for related discussions.
9
Example (3a) shows that (2b) becomes ungrammatical without ka. Example (3b) shows that when no word in (1) occurs in the scope of ka, the clause is interpreted as a yes-no question.
(3) a. *Taro-wa [dare-ga
denwasita] siritagatteiru.
Taro-Top who-Nom called
want_to_know
'Taro wonders who called.' b. Taro-wa [Yuka-ga
denwasi-ta ka] siritagatteiru.
Taro-Top Yuka-Nom call-Past
Q want_to_know
'Taro wonders whether Yuka called.'
In other environments, the words in (1) take on non-interrogative meanings. As (4) indicates, when the particle mo attaches to these words, they take on the universal meaning. As (5) indicates, when the particle ka attaches to these words, they take on the existential meaning.2 Those in parentheses do not seem to have productive uses.
(4) dare-mo
'everyone'
doko-mo 'everywhere'
(nani-mo)
'everything'
itu-mo
dore-mo
'everything'
* naze-mo
dono N-mo
'every N'
'always' 'for every reason'
(doo-mo)
2There
are at least two more contexts in which a subset of the words in (1) can be found. First, non-accented dare 'who' followed by mo behaves as a negative polarity item (like anybody in English). Second, if dare (either accented da're or non-accented) is followed by de-mo 'Cop-MO', it is interpreted like free choice anybody in English. Even though these are obviously related to (4) and (5), I will leave a discussion of these contexts for future. See, for example, Haspelmath (1995), Kawashima (1994), Kuroda (1965) and Lin (1996).
10
(5) dare-ka
'someone'
doko-ka
'somewhere'
nani-ka
'something'
itu-ka
'sometime'
dore-ka
'something'
naze-ka
'for some reason'
dono N-ka
'some N'
(doo-ka)
The words in (1) are most often referred to as wh-words because they are most easily translated to English by the use of wh-words such as who, when, etc. when they co-occur with the question particle ka as in (2). Kuroda (1965), following a view held in the traditional Japanese grammar, refers to the words in (1) as "indeterminate pronouns", which reflects their varying semantics. These words do not have interrogative meanings inherently, as shown in (3), (4) and (5). I will follow Kuroda's terminology in what follows. The universal particle mo is like the question particle ka in that it can occur in a position not local to indeterminate phrases. In (6), mo is attached to NPs or DPs that contain indeterminate phrases.3 In (7), mo is attached to PPs that contain indeterminate 3Non-local
association of indeterminate phrases and the existential particle ka seems to be difficult for some reason, as shown in (i) and (ii). (i) *
[Dono gakusei-no okaasan]-ka-ga kita. which student-Gen mother -KA-Nom came 'Some student's mother came.' (ii) * [[Dono gakusei-ga ø syootaisita] sensei] -ka-ga kita. which student-Nom invited teacher]-KA-Nom came 'A teacher that some student had invited came.' The following example from Nishigauchi (1990, pp. 121-122) may seem like a non-local case at a first glance. (iii)
Dare-kara-ka henna tegami-ga todoita. who-from-KA strange letter-Nom arrived 'A strange letter came from God knows who.'
I suspect, however, that the particle ka in this example is the question marker ka, as hinted at in Nishigauchi's paraphrase 'a letter came from someone, but I don't know who it is from' and discussion in the text surrounding (iii). Dare-kara-ka 'who-from-KA' in (iii) is interchangeable with dare-kara-da-ka 'who-from-Cop-KA', which suggests that it might very well be an elliptical form of embedded question.
11
phrases. The English translations given may not necessarily convey the precise interpretations of these sentences. They will be discussed shortly.
(6) a. [Dono gakusei-no okaasan]-mo odotta. which student-Gen mother -MO danced 'Every student's mother danced.' b. [[Dono gakusei-ga
ø syootaisita] sensei]-mo odotta.
which student-Nom invited
teacher -MO danced
'Every teacher that any student had invited danced.' (7) a. [Doko-kara]-mo syootaijoo-ga
todoita.
where-from- MO invitation card-Nom arrived 'From every place, an invitation card arrived.' b. Yoko-wa [[dare-no tukue-no ue]-ni]
-mo ame-o
oita.
Yoko-Top who-Gen desk-Gen top -LOC -MO candy-Acc put 'Yoko put candies on everyone's desk.' c. Yoko-wa [[dare-ga
kita
hi] -ni]-mo byookidatta.
Yoko-Top who-Nom came day -on -MO sick_was 'Yoko was sick on any day someone visited her.'
Further, (8) shows that mo can attach to a clausal element, and (9) shows that mo can attach to a than-phrase or clause in comparatives. 4
(8) a. Taro-wa [dare-ga
denwasi-te]-mo deru.
Taro-Top who-Nom call
-TE -MO answer
'No matter who calls, Taro answers.' 4Mo
in the comparative context should be examined further. Mo is optional in the examples in (9), whereas it is not in the examples in (8).
12
b. Taro-wa [Yoko-ga itu
denwasi-te]-mo rusudatta.
Taro-Top Yoko-Non when call
- TE -MO was_not_home
'No matter when Yoko called, Taro was not home.' (9) a. Yuuya-wa mikan-o
[dono kodomo-yori]-mo takusan tabeta.
Yuuya-Top orange-Acc which child-than
-MO many
ate
'Yuuya ate more oranges than any child.' b. Yuuya-wa mikan-o
[[dono kodomo-ga tabeta] yori]-mo takusan tabeta.
Yuuya-Top orange-Acc which child-Nom ate
than -MO many
ate
'Yuuya ate more oranges than any child did.'
What these non-local cases in (6) through (9) tell us is that the Japanese universal particle mo seems to behave slightly differently from the universal quantificational determiners in well-studied languages, e.g., every in English, in the sense that it seems to be able to occur away from its restrictor in surface syntax.
2.1.2 A Need for a Uniform Account The nature of association of indeterminate phrases with the particle ka in the interrogative context (the ka-construction) and with the particle mo in the universal context (the mo-construction) has attracted attention in the literature.5 Even though there has generally been much more focus on the syntax of the ka-construction in the literature than on the syntax and semantics of the mo-construction, if we were to understand the nature of indeterminate phrases in general, we should also pay attention to the latter. An additional reason to look for parallel analyses for the ka-construction and the mo-construction comes from the fact that both constructions show exactly the same 5See,
for example, Kuroda (1965), Hoji (1985), Nishigauchi (1986, 1990), Ohno (1989, 1991), Watanabe (1992a), Brockett (1994), Maki (1995), von Stechow (1996), Tanaka (1998, 1999), Hagstrom (1998), Takahashi (1999), and Yamashina (in progress).
13
locality effects. As we saw in chapter 1, the association of indeterminate phrases and ka shows wh-island effect. In (10), it is difficult to construe the indeterminate phrases dare 'who' and/or nani 'what' within the embedded interrogative clause as taking matrix scope. The sentence is most readily interpreted as a yes-no question as indicated in (10a).
(10)
Taro-wa [Yamada-ga
dare-ni nani-o
okutta ka] tazunemasita ka?
Taro-Top Yamada-Nom who-Dat what-Acc sent a.
Q asked
Q
'Did Taro ask what Yamada sent to whom?'
b. ?* 'Whox did Taro ask what Yamada sent to x?' c. *
'Whatx did Taro ask to whom Yamada sent x?'
d. ?* 'Whox did Taro ask whether Yamada sent what to x?'
Sentence (11) shows that when no indeterminate phrases are in the scope of mo, mo is interpreted as 'also' or 'even'. In the mo-construction in (12), dare 'who' or nani 'what' in the embedded interrogative clause cannot be associated with the universal particle mo, just as in the case of association with ka we saw in (10). The only available reading is expressed in (12a), where mo is interpreted as 'also' or 'even', and it is expected from what (11) shows.6
6Nishigauchi's
(1990, p. 164) example that shows the same point is given in (i).
(i) [[[[Dare-ga ø kaita ka] Mary-ga siritagatteiru] tegami]-ni]-mo who-Nom wrote Q Mary-Nom want_to_know letter -to -MO John-ga henzi-o kaita. John-Nom answer-Acc wrote a. 'Also to the letter such that Mary wants to know who wrote it, John wrote a reply.' b. *'For all x, x a letter, Mary wants to know for y, y a person, y wrote x, John wrote a reply to x.' b'. *'For every person y, and for every letter x, such that Mary wants to know whether y wrote x, John wrote a reply to x.' The unavailable interpretation expressed in (ib) is Nishigauchi's paraphrase. Judging from his remarks in the text, his intention may be better expressed as (ib').
14
(11)
Sono syoonin-mo damatteita. that witness-MO was_silent 'That witness was also silent./Even that witness was silent.'
(12)
[[ ø [Yamada-ga [[
dare-ni nani-o
okutta ka] sitteiru] syoonin]-mo
[Yamada-Nom who-Dat what-Acc sent
Q] know] witness]-MO
damatteita. was_silent a.
'The witness who knew what Yamada sent to whom was also silent.' 'Even the witness who knew what Yamada sent to whom was silent.'
b. *
'For every person x, the witness who knew what Yamada sent to x was silent.
c. ** 'For every thing x, the witness who knew to whom Yamada sent x was silent.' d. *
'For every person x, for every thing y, the witness who knew whether Yamada sent y to x was silent.'
On the other hand, when indeterminate phrases are embedded inside complex NPs or adjunct clauses, the sentences are fine in the ka-construction as shown in (13) and (14), as well as in the mo-construction as shown in (15) and (16).7
7Naze
'why' is sensitive to islands, while doo 'how' is not, as shown in (i) and (ii). See, for example, Huang (1982), Lasnik and Saito (1992) and Tsai (1994).
(i) ?* John-wa [Taro-ga naze kita-kara] kaerimsita ka? John-Top Taro-Nom why came-because left Q 'Whyx did John leave because Taro came for reason x?' (ii) John-wa [Taro-ga doo hannoosita-kara] okorimasita ka? John-Top Taro-Nom how reacted-because got_upset Q 'Howx did John get upset because Taro reacted in manner x?'
15
(13)
Taro-wa [[dare-ga ø katta] mochi]-o
tabemasita ka?
Taro-Top who-Nom bought rice cake -Acc ate
Q
'Whox did Taro eat rice cakes that x bought?' (14)
Taro-wa [dare-ga T.-Top
kita-kara]
kaerimasita ka?
who-Nom came-because left
Q
'Whox did Taro leave because x came?' (15)
[[[[Dono T.A.-ga
ø osieta] gakusei]-ga ø syootaisita] sensei]-mo kita.
which T.A.-Nom taught student -Nom invited
teacher -MO came
'For every T.A. x, the teacher(s) that the student(s) that x had taught invited came.' (16)
[[ ø [Taro-ga
nani-o
katta-kara]
okotta]
hito]-mo
Taro-Nom what-Acc bought-because got_angry person -MO heya-o deteitta. room-Acc left 'For every thing x, the people who got angry because Taro had bought x left the room.'
The situation is summarized as follows.
(17)
*
[..... [..... ind ..... ka]WH .....]IP-ka z--------m
(18)
a. [..... [..... ind .....]CNP .....]IP-ka z-------m
b. [..... [..... ind .....]Adjunct .....]IP-ka z--------m
16
(19)
*
[..... [..... ind ..... ka]WH .....]CNP -mo z---------m
(20)
a. [..... ind .....]CNP -mo z----m
b. [..... [..... ind .....]CNP .....]CNP -mo z--------m
c. [..... [..... ind .....]Adjunct .....]CNP -mo z---------m Not only is the association of indeterminate phrases and the particles ka and mo blocked by the presence of a wh-island, which comes with ka, but the same association is also blocked by the presence of particle mo. This is schematically shown in (21), and the relevant examples will be discussed in section 2.4.1.
(21)
a. * b. *
[..... [..... ind ..... mo] .....]-ka z-------m
[..... [..... ind ..... mo] .....]-mo z-------m
The fact that the ka- and mo- constructions exhibit exactly the same pattern with respect to locality indicates that these two constructions should receive a uniform account. Further, we would want the uniform account to derive the island puzzle.
2.1.3 A Uniform Account Based on a Particular Interpretation of the Locality Effects A well-accepted interpretation of the locality pattern we just saw is that it indicates that the association of indeterminate phrases and the particles ka and mo exhibits a property reminiscent of movement relations at least partially (e.g., Nishigauchi 1990, Watanabe 1992a, b, von Stechow 1996, Richards 1997, Hagstrom 1998, among
17
many others.).8 The view that movement is involved in the ka-construction developed along with a more general theory of LF movement of wh-in-situ (Huang 1982, Lasnik and Saito 1992, etc.). According to the LF wh-movement theory, the English whquestion in (22a) and its Japanese counterpart in (23a) share fundamental structural properties at LF, as in (22b) and (23b).
(22)
a. What did Yoko read? b. [whati [C0 [IP Yoko read ti ]]]
(23)
a. Yoko-ga Yoko-Nom
nani-o
yomimasita ka?
what-Acc read
Q
'What did Yoko read?' b. [nani-oi [[IP Yoko-ga ti what-Acc Yoko-Nom
yomimasita] ka]] read
Q
At least as far as syntax is concerned, a uniform account of the ka- and moconstructions based on this familiar movement analysis of the former is most explicitly pursued in Nishigauchi (1990). We have then nice parallel structures for the ka- and moconstructions as in (24). I will refer to the projection that mo heads as MoP, although the exact categorial status is not clear (see, for example, Nishigauchi 1990, pp. 159-161). This issue is not crucial to the discussion to follow.
8There
are some variations among these authors as to what moves at what level. For simplicity, I will present below the traditional picture in which indeterminate (or wh-) phrases move at the level of LF.
18
(24)
a.
CP 2 indeterminate 2 IP C ka @ Q ....t....
b.
MoP 2 indeterminate 2 XP mo @ ∀ ....t....
This syntactic parallelism is pursued further, for example, in Watanabe (1992a) and von Stechow (1996) with some revisions.
2.1.4 The Island Puzzle and the Pied-Piping Analysis A challenge for the movement analysis is to account for the island puzzle, that is, the exceptional behavior of indeterminate phrases when they occur within Complex NP and Adjunct islands. Authors who argue for or assume the movement analysis rely on Nishigauchi's (1986, 1990) pied-piping analysis or variants of it. Nishigauchi (1986, 1990) proposed that the apparent lack of Complex NP/Adjunct island effects in (13) through (16) is due to the fact that these islands somehow count as [+WH]. This makes the whole islands pied-pipe and undergo wh-movement at some invisible level where island constraints are at work (his LF), and therefore the indeterminate phrases embedded inside these islands are not extracted out of them.9 In sentence (25) for example, repeated from (13), (i) the indeterminate phrase dare 'who', which is marked as [+WH], moves to the specifier position of the relative clause CP at a covert level, and (ii) the relative clause CP occupies the specifier position of the NP. He further assumes with Selkirk (1982) that when the head of some projection is unmarked for some feature, the relevant feature on the specifier percolates up. These assumptions together allow the whole complex NP to receive the feature [+WH], as indicated in (26). 9There
is perhaps extraction out of islands and syntactic reconstruction of the rest of the islands at a level where no island constraints are at work, in order to obtain appropriate interpretations. See Chomsky (1977), von Stechow (1996) and Sharvit (1998).
19
(25)
Taro-wa [[dare-ga
katta] mochi]-o
tabemasita ka?
Taro-Top who-Nom bought rice cake -Acc ate
Q
'Whox did Taro eat rice cakes that x bought?' (26)
NP 2 [+WH] CP N' 2 indeterminate C' 2 IP C # .......t...... [+WH]
[+WH]
Assumptions (i) and (ii), which are made crucial use of for the feature percolation mechanism to work, raise a number of questions. First, assumption (i) says that indeterminate phrases move to SpecCP.10 According to standard assumptions, however, the specifier of a relative clause CP is occupied by an empty relative pronoun, so that the interpretation of the relative clause can proceed successfully. 11 In English, for example, this is what is believed to make a complex NP an island for movement. The specifier of the relative clause CP is not available for a wh-phrase to stop over, hence, it has to cross IP and NP nodes in one movement. If the specifier of the relative clause CP is available in Japanese, as Nishigauchi's pied-piping mechanism forces it to be, then there is no need to discuss pied-piping at all in the first place. Japanese indeterminate phrases can move out of complex NPs without violating Subjacency.
10Subsequent
movement analyses such as Watanabe (1992a, b), Hagstrom (1998) and Richards (2000) make use of this kind of "internal wh-movement" (Riemsdijk 1984) as well. 11Von
Stechow (1996, fn. 6) and Toyoshima (1996, p. 5) also questions this assumption. Von Stechow (1996), however, still relies on Nishigauchi's pied-piping mechanism.
20
Second, assumption (ii) says that the relative clause CP occurs in the specifier position of NP. This assumption is necessary so that the [+WH] feature on CP can be percolated up to NP. As Toyoshima (1996) points out, however, relative clauses are modifiers, and are typically assumed to be adjoined to N' (or NP in the DP system). Thus the second assumption that is crucial for the pied-piping mechanism to work also requires justification.12 Let us assume for now that the above questions are taken care of in some way or another, and that complex NPs and adjunct clauses successfully count as [+WH] and hence pied-pipe. A more important question now is whether the pied-piping mechanism draws the correct line between which islands are pied-pipable and which islands are not. More specifically, the pied-piping mechanism should not allow wh-islands to pied-pipe, since otherwise, one would not expect the wh-island effects in (10) and (12). Further, the example in (27) shows that when a complex NP contains a wh-island in it, the complex NP cannot undergo pied-piping. Similar examples are discussed in Nishigauchi (1990), Watanabe (1992a), Maki (1995), Toyoshima (1996) and Hagstrom (1998).
12Toyoshima
(1996) discusses more questions about the pied-piping mechanism. See also Watanabe (1992a) for a summary of general questions about pied-piping, such as parametric differences between Japanese and English with respect to the availability of pied-piping and LF Subjacency constraint.
21
(27)
[[[Dare-ga
dare-no
ronbun-o yonda ka] siritagatteiru]
[[[who-Nom who-Gen paper-Acc read
hito]-ga
Q] want_to_know] person]-Nom
kimasita ka? came a.
Q 'Did the person who wonders who read whose paper come?'
b. ?* 'Whox did the person who wonders whose paper x read come?' c. *
'Whose papery did the person who wonders who read y come?'
d. ?* 'Whox did the person who wonders whether x read whose paper come?'
Thus the pied-piping mechanism should not allow the complex NP of the kind we find in (27) to pied-pipe. It seems that what allows complex NPs and adjunct clauses in the schematic pictures in (28a) and (28b) to pied-pipe should also allow wh-islands in (28c) to pied-pipe. That is, in (28c), the specifier of CP is marked [+WH], and it should percolate up to the CP node as indicated.13
(28) a. Complex NP island [+WH]
NP
2 [+WH] CP 2 2
[+WH] indeterminate1 2
IP C # .......t1.......
N
b. Adjunct island [+WH]
XP
2 [+WH] indeterminate12
IP X # .......t1........
13The
c. Wh-island [+WH]
CP
2 [+WH] indeterminate12
IP C # ka ...indeterminate...t1...
head C in (28) is probably marked as [+WH] as well. In that case, it is that feature that percolates up to CP, according to Nishigauchi’s pied-piping mechanism.
22
Since it is this syntactic or morphological feature [+WH] that determines what undergoes movement at a covert level, wh-islands should undergo movement as well. 14 Thus the exceptionality of the complex NP islands and adjunct islands does not seem to follow from anything inherent in the movement analysis.
2.1.5 The Syntax, Semantics, and Their Mapping As we have just seen, the movement analysis is forced to say something special about the lack of island effects other than the wh-island. This is a challenge to the syntactic side of the movement analysis. As far as the semantics and the syntaxsemantics mapping of the ka-construction is concerned, not much discussion specific to Japanese is found. One would expect that a general theory of the semantics of interrogatives would extend to Japanese. Further, once we adopt the movement analysis, in which the LF of the ka-construction is modeled after the LF of English wh-questions, questions of the syntax-semantics mapping specific to the Japanese case do not arise since one should expect that any plausible story of mapping in English, for instance, as described in Karttunen (1977), will take care of Japanese as well. The semantics of the mo-construction has been discussed in Nishigauchi (1986, 1990), Ohno (1989) and von Stechow (1996) among others, which will be discussed shortly. Regarding the question of how the appropriate semantics of the mo-construction is derived from its structure, von Stechow (1996) is the only work that provides an explicit analysis.15 As we will see in section 2.6.2, an attempt to provide a uniform account by extending the well-accepted LF wh-movement analysis of the ka-construction 14Richards
(2000) addresses the question of distinguishing what is pied-pipable and what is not, and proposes a mechanism which also involves internal wh-movement. Due to time limitations, I leave a detailed examination of the proposed mechanism for future work. 15Yamashina
(in progress) will be the only other attempt. At the time of writing, I have not seen a complete version of it.
23
to the mo-construction faces difficulty since ad-hoc assumptions will be needed for the mo-construction. My strategy in this chapter is a slightly different one from what is found in the literature: I will first examine the mo-construction and look for a good analysis. It will turn out that the new analysis of the semantics of the mo-construction can be spelled out using existing ideas that have been proposed or at least made use of in the context of interrogative pronouns in English. This has consequences in how we want to analyze the ka-construction. The uniform analysis of the mo- and ka-constructions that emerges from the new way of looking at the semantics of the mo-construction will provide a switch of perspective on the long-standing island puzzle. In section 2.2, I will discuss the standard analysis of the semantics of the moconstruction, which I will refer to as "the embedded restrictor view". What is commonly believed to be support for this standard view will be presented there. It will be shown in section 2.3 that by departing from the standard way of looking at the semantics of the moconstruction, we can provide a straightforward analysis of the syntax-semantics mapping of the construction. Furthermore, I will show in sections 2.4 and 2.5 that the alternative view provides not only a uniform analysis of the mo- and ka-constructions, but also a new way of understanding the island puzzle. In section 2.6, I will go back to the embedded restrictor view of the mo-construction and discuss challenges that it meets. Remaining questions regarding kadooka 'whether'-clauses will be discussed in section 2.7. Section 2.8 summarizes the chapter. In the main discussions to follow, I will focus on the nominal mo-construction as introdued in (6). I will briefly come back to the postpositional mo-construction as in (7) and the clausal mo-construction as in (8) in section 2.3.5.
24
2.2 The Interpretation of the Mo-Construction 2.2.1 Universal Quantification over Pairs of Individuals The sentence in (25) is an example of the local mo-construction, in which the indeterminate phrase dono gakusei 'which student' occurs next to the universal particle mo. The sentence in (26), on the other hand, is an example of the non-local moconstruction, where dono gakusei 'which student' is embedded in a complex NP.
(25)
Dono gakusei-mo kita. which student-MO came 'Every student came.'
(26)
[[Dono gakusei-ga
ø syootaisita] pianisuto]-mo
which student-Nom invited
pianist
kita.
-MO came
'Every pianist that a student had invited came.'
According to Nishigauchi (1990), (26) has the truth condition in (27), in which the universal quantification is over student-pianist pairs, such that the former invited the latter.16
(27)
Åx,y[[student(x) & pianist(y) & invite(y)(x)] ∞ come(y)]
16Nishigauchi's
(1990, p. 126) example is given in (i).
(i) [[Dare-ga ø kaita] tegami]-ni-mo onazi koto-ga kaite-atta. who-Nom wrote letter -in-MO same thing -Nom written-was 'For all x, y, x a person, y a letter x wrote, the same thing was written in y.' Note that (27) is logically equivalent to (ii). I will come back to this in section 2.3. (ii) Åy[˛x[student(x) & pianist(y) & invite(y)(x)] ∞ came(y)]
25
Let's consider a situation where three students and three teachers each invited a pianist (six pianists all together), and three other students each invited a drummer to a party. The three pianists invited by the students came to the party. In this situation, the sentence is judged to be true, in accordance with (27). 17 Regarding the question of how to derive the semantics in (27) from the surface syntax of the sentence in (26), Nishigauchi (1990) does not present a complete analysis. It is assumed that both the indeterminate phrase dono gakusei 'which student' and the head noun of the relative clause pianisuto 'pianist' introduce free variables, in the way English indefinites are analyzed in Kamp (1981) and Heim (1982). The pied-piping mechanism introduced in section 2.1.4 allows the [+WH] feature of dono gakusei 'which student' to percolate up the tree and makes the whole NP count as [+WH]. This process is responsible for allowing the bare noun pianisuto 'pianist' to be treated like an indeterminate phrase, and the variable introduced by it to receive universal force from mo. Further, the [+WH] feature of the whole NP makes it move to the specifier position of the MoP. The LF of (26) thus looks like (28).18
17More
than one indeterminate phrase can occur in the scope of mo, as shown in (i).
(i) [[Dono gakusei-ga dono ie-ni ø syootaisita] pianisuto]-mo kita. which student-Nom which house-to invited pianist -MO came 'Every pianist that a student had invited to some house came.' So more generally, Nishigauchi's (1990) paraphrase involves universal quantification over (n+1)-tuples of individuals for n many indeterminate phrases in the scope of mo. In (i), it is a triple of a student, a house, and a pianist. 18The
constituent I labeled as NP in (28) may be better analyzed as DP.
26
(28)
IP 5 MoP VP 5 # NP 3 kita %t mo came dono gakusei-ga syootaisita pianisuto every which student-Nom invited pianist
No explicit analysis of how exactly the LF in (28) is mapped to (27) is presented in Nishigauchi (1990).19
2.2.2 The Partitive and Weak Readings Ohno (1989) and von Stechow (1996) correctly point out that the interpretation that Nishigauchi's (1990) proposal predicts is not the only available reading of the moconstruction in general. The examples in (29) show that bare NPs in Japanese do not have any markings for definiteness/indefiniteness or singularity/plurality. Depending on the context, definite, indefinite, singular or plural interpretations are possible for the subject NPs in (29).
(29)
a. Pianisuto-ga kita. pianist-Nom came 'A pianist/pianists/the pianist(s) came.' b. [[Yoko-ga
ø syootaisita] pianisuto]-ga
Yoko-Nom invited
pianist
kita.
-Nom came
'A pianist/pianists/the pianist(s) that Yoko had invited came.' 19Von
Stechow (1996) is the only previous work that attempts to provide an explicit analysis of the syntax-semantics mapping of the mo-construction. His analysis incorporates revisions of Nishigauchi's (1990) semantics of the construction, as well as its syntax. The semantic revision will be discussed shortly. The syntactic revision is that syntactic reconstruction is added to Nishigauchi's LF representation in, for example, (28), in which the whole complex NP is moved out of the scope of mo (see section 2.6.2).
27
In sentence (26) above, too, pianisuto 'pianist' could be interpreted as plural given an appropriate context. Let's modify the scenario above slightly: three students each invited three pianists to a party (nine pianists all together), and three other students each invited three drummers (nine drummers all together). Sentence (26) has a reading in which it is true if and only if all the nine pianists invited by the students came to the party. This reading is expected in the translation in (27). I will refer to this reading as the strong reading. Given an appropriate context, sentence (26) seems to allow another reading which I will refer to as the partitive reading. Let's start out with the same scenario as described above. We now want to play a game which requires us to form pairs of a student and a pianist or a drummer who he or she had invited. We look around the house and find out that for every student who had invited pianists, at least one pianist invited by him or her has arrived. We can now utter sentence (26) truthfully. Obtaining this interpretation in (26) requires some help from the context, compared to the preferred strong reading described in (27).20 This reading of sentence (26) can be expressed more clearly by (30). An example similer to (30) is provided in Ohno (1989, p. 241).
(30)
[[Dono gakusei-ga which
ø syootaisita] pianisuto]-mo hito-ri-wa
student-Nom invited
pianist
kita.
-MO one-Cl-at least came
'For each student who had invited pianists, at least one of the pianists he or she had invited came.'
The partitive reading is not expected in Nishigauchi's (1990) semantics.
20More needs to be said if one wants to prove that they are independent readings. It may turn out, as suggested by Angelika Kratzer (p.c.), that (27) is a special case of the weak reading.
28
The example in (31) from Ohno (1989) also demonstrates that there are interpretations of the mo-construction that Nishigauchi's (1990) proposal fails to capture. (32) is the truth condition of the sentence that Nishigauchi (1990) would provide, where mo universally quantifies over country-book pairs. Ohno (1989) points out that sentence (31) can be truthfully uttered if "each country is represented with at least one publication (p. 241)". I will refer to this type of reading as the weak reading.
(31)
Kono mise-de-wa [[dono kuni-de
syuppansareta] hon]-mo
this store-in-Top which country-in was_published book- MO
utteiru. sell
'Books published in any country are sold in this store.' (32)
Åx,y[[country(x) & book(y) & published_in(x)(y)] → sell(y)(this store)]
With this example, Ohno shows that Nishigauchi’s (1990) idea that both indeterminate phrases and the head noun of a relative clause always receive universal force from mo cannot be maintained.
2.2.3 The Embedded Restrictor View Ohno (1989), as well as von Stechow (1996), conclude from observations of the above kind that what provides the restriction for the universal quantifier mo is only the denotation of indeterminate phrases that are embedded inside relative clauses. The head of the relative clause is interpreted independently, either as definite or indefinite, through whatever mechanism that allows the bare NPs in the examples in (29) to receive various interpretations. In particular, they claim that the meaning of sentence (31) should be represented as (33), rather than as (32). According to this proposal, mo only quantifies over countries, and the weak reading arises from the indefinite interpretation of the head
29
of the relative clause, hon 'book(s)'. The sentence is predicted to assert that for every country x, there are books published in x that this store sells.
(33)
Åx[country(x) → ˛y[book(y) & published_in(x)(y) & sell(y)(this store)]]
The mo-construction thus involves a universal quantifier that takes sentential scope, even though indeterminate phrases that provide a domain for quantification could be embedded in a larger constituent. I will refer to their idea as "the embedded restrictor view". The strong reading, according to Ohno (1989) and von Stechow (1996), arises in the following manner. The representation of the strong reading of sentence (26) is as in (34). The *-operator and the Í-operator are defined in (35), in which y ≤ x means "y is a mereological part of x" (Link 1983).
(34)
Åx[student(x) ∞ *come(Íy[pianist(y) & invite(y)(x)])]
(35)
a. *P(x) = Åy[y ≤ x & atomic(y) ∞ P(y)] b. ÍxP(x) = ιx[*P(x) & Åy[*P(y) ∞ y ≤ x]]
The apparent universal force associated with the head of the relative clause pianisuto 'pianist' (see the translation in (27)) comes about due to the bare NP's semantics that is assumed to be analogous to that of plural definite NPs in English. The partitive reading of (26) also makes use of the plural definite interpretation of pianisuto 'pianist', and can be represented as (36).
(36)
Åx[student(x) ∞ ˛z[z ≤ Íy[pianist(y) & invite(y)(x)] & come(z)]]
30
To summarize, Nishigauchi (1990) only discusses one of the possible readings available in the nominal mo-construction, and presents a (partial) analysis for that reading. Ohno (1989) and von Stechow (1996) recognize the partitive and weak readings and propose that only indeterminate phrase denotations provide the domain of quantification for mo, and the head NP of the relative clause is interpreted independently of mo (the embedded restrictor view). The strong reading is achieved by assuming the semantics of plural definites for the head NP of the relative clause. It appears, then, that the embedded restrictor view represents the meaning of the mo-construction better than Nishigauchi's analysis. The embedded restrictor view, indeed, seems to be accepted as the standard semantics of the mo-construction (see, for example, Watanabe 1992a and Takahashi 1999; see also Yamashina (in progress) for a slightly revised version).
2.2.4 Apparent Support for the Embedded Restrictor View According to the embedded restrictor view, the universal quantifier mo takes sentential scope and it binds a variable or variables associated with indeterminate phrases, as in (33), (34) and (36). This view then predicts that a pronoun in the matrix clause can be bound by the universal quantifier. The examples in (37) and (38) show that this kind of variable binding is indeed possible. (For simplicity, I assume singularity of pianisuto 'pianist' and hito 'person'.)
(37)
a. [[Dono gakusei1 -ga ø syootaisita] pianisuto]-mo sono1 sensei datta. which student-Nom
invited
pianist
-MO his
teacher was
'For every student1 , the pianist he1 had invited was his1 teacher.' b. Åx[student(x) ∞ teacher_of(x)(ιy[pianist(y) & invite(y)(x)])]
31
(38)
a. [[ø Dono kuni1 -e
ryokoosita] hito] -mo sono1 syuto-o
which country-to traveled
person -MO its
tazuneta.
capital-Acc visited
'For every country1 , the person who traveled there1 visited its1 capital.' b. Åx[country(x) ∞ visit(ιz[capital(x)(z)])(ιy[person(y) & travel_to(x)(y)])]
The examples in (39) from Nishigauchi (1990, p. 205 & p. 219) show the same point.21 (39b) is an example of the clausal mo-construction.
(39)
a. [[Dono ronbun1 -o kaita] hito] -mo sore1 -o LI-ni okutta. which paper-Acc wrote person- MO it-Acc LI-to sent 'For every paper1 , the person who worte it1 sent it1 to LI.' b. [Dono hon1 -o
kaw-te]-mo Mary-wa kanarazu {e1 /sore1 -o} yonda.
which book-Acc buy-TE -MO Mary-Top without_fail
it-Acc
read
'For every book1 that Mary bought, she read it1 without fail.'
Although the anaphora facts may appear to provide initial support for the embedded restrictor view, more data will be provided in section 2.6.1 that support a different interpretation of the facts. The embedded restrictor view also seems to fit with the movement analysis mentioned earlier. The simplified structures are repeated below.
21Nishigauchi's
own paraphrases for (39a) and (39b) are given in (ia) and (ib), with correction of typographical errors. (i) a. 'For all x, y, x a paper, y a person who wrote x, y sent x to LI.' b. 'For all x, x a book, when Mary bought x, Mary read x without fail.'
32
(24)
a.
CP 2 indeterminate 2 IP C ka @ Q ....t....
b.
MoP 2 indeterminate 2 XP mo @ ∀ ....t....
Establishing a syntactic relation between an indeterminate phrase and mo, by moving the former to the specifier position of MoP as shown in (24b), appears to be motivated by the semantic relation between the indeterminate phrase and mo: namely, the relation of the quantifier and what provides its domain (see Nishigauchi 1990). It will be shown in section 2.6.2, however, that on the contrary to what is believed in some syntactic literature, it is not a straightforward matter to derive the semantics based on the embedded restrictor view from the structure in (24b). Furthermore, as we saw earlier, the movement analysis has to rely on some or other form of the pied-piping mechanism, which has problems in providing a natural account of the island puzzle. Further, as we saw above, the embedded restrictor view construes the moconstruction as involving a universal quantifier mo which takes sentential scope while indeterminate phrases it associates with can be embedded in its sister constituent. This state of affairs is somewhat reminiscent of what is known as "inverse linking" in English (May 1985), exemplified in (40). The sentence is known to have a reading in which every basket takes the widest scope, as shown in (41), even though it is embedded in a larger DP.
(40)
One apple in every basket is rotten.
(41)
Åx[basket(x) → ˛y[apple(y) & in(x)(y) & rotten(y)]]
33
There are various analyses as to how to derive this interpretation from the syntax in (40). For example, this interpretation can be derived from an LF like (42), in which every basket is extracted out of the larger DP and adjoined to IP.
(42)
[IP [every basket]1 [IP [DP one apple in t1 ] is rotten]]
The embedded restrictor view of the mo-construction, along with the movement analysis, allows us to see the similarity between the mo-construction and the Englsih inverse linking construction. In fact, von Stechow (1996) and Takahashi (1999) view the moconstruction as a kind of inverse linking construction analogous to (40). Let me point out, however, a couple of differences between the mo-construction and inverse linking in English. The first obvious difference is that in the English inverse linking sentence in (40), the determiner every is embedded, while in its Japanese counterpart, the non-local mo-construction in (43), the universal particle mo is a matrix element in surface syntax. Further, the sentences in (43) are unambiguous. They only allow the inversely linked reading.
(43)
a. [Dono kago-no
naka-no
ringo]-mo hito-tu kusatteiru.
which basket-Gen inside-Gen apple - MO one-Cl rotten 'One (of the) apple(s) in every basket is rotten./For every basket x (that contains apples), there is one apple in x that is rotten.' b. [Dono gakka-no
gakusei daihyoo]-mo
toohyoosita.
which department-Gen student representative -MO voted 'A student rep of every department voted.'
When mo occupies the position that every occupies in (40), as in (44), the relevant inverse linking interpretation is not available. If (44a) is grammatical at all, it only has a
34
reading that corresponds to the nonsensical reading of (40), that some apple is in every baseket and it is rotten. (44b) only has a reading in which we are talking about a student representative that represents all the departments.
(44)
a. ?? [Dono kago-mo-no
naka-no
ringo]-ga
hito-tu kusatteiru.
which basket-MO-Gen inside-Gen apple]-Nom one-Cl rotten 'There is one apple that is in every basket and is rotten.' b. ?
[Dono gakka-mo-no
gakusei daihyoo]-ga
toohyoosita.
which department-MO-Gen student representative -Nom voted 'A student rep who is from every department voted.'
This suggests that if movement is responsible for the inversely linked readings, it can only apply to the indeterminate phrases in (43), but not to [indeterminate phrase-mo] in (44), unlike the English case in (42). Secondly, in English, every student in (45), which is embedded in a relative clause, cannot take matrix scope. On the other hand, for its Japanese counterpart in (46a), such a reading is the only reading available, just like the examples we saw in (43), which did not involve complex NPs. The interpretation of the example in (46b) corresponds to the sole reading in (45).
(45)
The teachers that every student invited came.
(46)
a. [[Dono gakusei-ga which
ø syootaisita] sensei]-mo kita.
student-Nom invited
teacher -MO came
'For every student x, the teacher(s) x invited came.' b. [[Dono gakusei-mo ø syootaisita] sensei]-ga which
student-MO
invited
kita.
teacher -Nom came
'The teacher(s) that every student invited came.'
35
These differences, then, are obstacles to assimilating the mo-construction to the inverse linking construction.
2.3 A New Way of Looking at the Semantics of the Mo-construction This section presents an alternative view of the semantics of the mo-construction. It will be shown that the alternative view can be spelled out using an existing analysis of interrogative pronouns. This will have consequences in how we want to analyze the kaconstruction. The island puzzle will be shown to follow from the interpretation process, and the new view also makes the mapping from syntax and semantics of the moconstruction simple.
2.3.1 The Direct Restrictor View I propose an alternative to the embedded restrictor view of the mo-construction that assumes that the domain of quantification for mo is provided directly by its sister constituent, rather than by indeterminate phrases embedded in its sister. According to this alternative way of looking at the semantics of the mo-construction, which I refer to as "the direct restrictor view", the universal quantification contributed by mo in (47a), for example, is over a set of mothers rather than over a set of students. If Yoko, Satoshi and Takumi were all the students, the interpretation of (47a) according to this view can be expressed as in (47b).
(47)
a. [Dono gakusei-no okaasan]-mo odotta. which student-Gen mother -MO danced 'Every mother of some student danced.'
36
b. ∀x[x ∈ {ιy[mother(Yoko)(y)], ιy[mother(Satoshi)(y)], ιy[mother(Takumi)(y)]} → dance(x)]
Likewise in sentence (48a), the universal quantification is over a set of plural individuals that are pianists invited by some student or other, rather than over a set of students, as shown in (48b). (48b) represents the strong reading of the sentence.
(48)
a. [[Dono gakusei-ga
ø syootaisita] pianisuto]-mo kita.
which student-Nom invited
pianist
-MO came
'Every pianist that a student had invited came.' b. ∀x[x ∈ {Íy[pianist(y) & invite(y)(Yoko)], Íy[pianist(y) & invite(y)(Satoshi)], Íy[pianist(y) & invite(y)(Takumi)]} → ∗come(x)]
More generally, the interpretations of (47a) and (48a) according to the direct restrictor view can be expressed as in (49).
(49)
a. Åx[˛z[x = ιy[mother(z)(y) & student(z)]] → dance(x)] b. Åx[˛z[x = Íy[pianist(y) & student(z) & invite(y)(z)]] → ∗come(x)]
Note that the strong reading of (48a) by itself does not motivate the plural quantification in (49b), because (50) would express the strong reading equally well.22
22Note
that (50) is equivalent to (i), Nishigauchi's (1990) translation of sentence (48a).
(i) Åx,y[[pianist(x) & student(y) & invite(x)(y)] ∞ come(x)]
37
(50)
Åx[˛y[pianist(x) & student(y) & invite(x)(y)] → come(x)]
To see that (49b) is more appropriate than (50), we must wait for a discussion of the partitive reading of (48a) in section 2.3.4. The embedded restrictor view, as well as its predecessor Nishigauchi (1990), relates the non-local mo-construction as in (47a) and (48a) to the local mo-construction as in (51) by assuming that mo always quantifies over what is denoted by indeterminate phrases.
(51)
Dono gakusei-mo kita. which student-MO came 'Every student came.'
The non-local mo-construction, thus, can be viewed as a case of "quantification at a distance",23 since one of the two arguments of mo is distanced from the quantifier at Sstructure. The direct restrictor view, on the other hand, relates the local and non-local moconstructions by assuming that mo always quantifies over what is denoted by its sister constituent. According to this view, the non-local mo-construction is no longer a case of quantification at a distance. This is summarized schematically in (52), where double underlining indicates which syntactic forms provide the domains for quantification.
23The
term is used here as a descriptive term. The term was originally introduced by Obenauer (1983) for a certain French construction in which beaucoup/peu 'a lot/little', for example, occurs in distance from its associate de livres 'of books'.
38
(52)
a. The embedded restrictor view local:
non-local:
ind-mo
[......ind.......]-mo
zm
z---m
b. The direct restrictor view local:
non-local:
ind-mo
[......ind.......]-mo
zm
z---m
2.3.2 Focus and the Domains of Quantification The contrast between the two views of the non-local mo-construction has a close similarity to the contrast in two of the major alternative views of a phnomenon known as association with focus. (53) shows examples of association of only with focus from Rooth (1985). Capitalized constituents are marked with a focus feature F in syntax, and receive phonological prominence.
(53)
a. John only introduced [BILL]F to Sue. b. John only introduced Bill to [SUE]F.
The different placement of focus affects the truth-conditions of the sentence. If John introduced Bill and Tom to Sue and performed no other introduction, (53a) is false while (53b) is true. One view sees (53) as involving quantification over individuals (Chomsky 1976). The truth-conditions of the two sentences are assumed to be (54). (54a) says that the unique individual that John introduced to Sue is Bill, and (54b) says that the unique individual that John introduced Bill to is Sue. The clauses in parentheses are presuppositions.
39
(54)
a. ∀y[introduce(y)(Sue)(John) → y = Bill] (& introduce(Bill)(Sue)(John)) b. ∀y[introduce(Bill)(y)(John) → y = Sue] (& introduce(Bill)(Sue)(John))
The quantifier only and the focused phrase, one of its arguments, do not occur locally in (53). In a similar manner to the movement of indeterminate phrases in the embedded restrictor view of the mo-construction, this apparent "quantification at a distance" is assumed to be fixed at LF by moving the focused phrases closer to only. An alternative view due to Rooth (1985) assumes that the sentences in (53) involve quantification over properties. In this view, it is not the focused phrase at a distance but the sister constituent of only (VP in (53)) that serves as an argument of only. The sentences in (53) are then translated as (55), which gives similar truth-conditions to (54).
(55)
a. ∀P[P ∈ {introduce(y)(Sue) | y ∈ D} & P(John) → P = introduce(Bill)(Sue)] b. ∀P[P ∈ {introduce(Bill)(y) | y ∈ D} & P(John) → P = introduce(Bill)(Sue)]
(55a) says that if John has a property of introducing someone to Sue, then it is the property of introducing Bill to Sue. (55b) says that if John has a property of introducing Bill to someone, then it is the property of introducing Bill to Sue. In this alternative view, the first arguments of only in (53) are not the focused phrases. Thus the S-structure of the sentences does not need to be rearranged and the focused phrases can be interpreted in-situ. A piece of argument in favor of this view comes from the lack of locality effects expected for movement of focused phrases.
40
2.3.3 Indeterminate Phrases as Introducing Free Variables: The Ka-Construction If sets like those in (47b) and (48b) can be obtained from the sister constituent of mo which contains one or more indeterminate phrases, the syntactic structure of the nonlocal mo-construction can now be transparently mapped into the tripartite structure just like the local mo construction or sentences with every in English. One way of making such sets available can be spelled out by making use of an idea that indeterminate phrases introduce free variables. It has been often assumed that wh-phrases in English introduce free variables that are bound by some operator. See, for example, Baker (1970), Pesetsky (1987), Berman (1991), and Reinhart (1998). Work by Kamp (1981) and Heim (1982) on indefinites in English also had much influence in this area. As for indeterminate phrases in Japanese, Kuroda (1965, p. 101) noted that "the role of the indeterminate pronouns are very much like that of yet unbounded logical variables". Nishigauchi (1986, 1990) pursued this insight and made an important connection to the analysis of English indefinites by Kamp (1981) and Heim (1982).24 The direct restrictor view of the mo-construction proposed above can be spelled out by making use of this existing idea about wh- and indeterminate phrases. Let us first see how this idea derives the desirable semantics of the ka-construction. I will then show that adopting this idea also makes it possible to spell out the direct restrictor view of the mo-construction, hence, providing a unified analysis of the ka- and mo-constructions. In the ka-construction in (56a), the indeterminate phrase dare 'who' receives index 1 and is interpreted as person(x1 ). Let us assume that the question particle ka selects as its argument a constituent that is of a suitable type for questions, namely, a set of propositions, along the lines of Hamblin (1973) and Karttunen (1977). The sister of ka in (56a), however, is an open sentence, rather than a set of propositions. Suppose that 24For
related discussions on Chinese wh-phrases, see, for example, Aoun and Li (1993), Tsai (1994) and Lin (1996). See also Cole and Hermon (1994, 1998).
41
interpretability forces an introduction of an operator Op, at LF as in (56b). Op combines with the clause and forms the set of propositions {that x is a person and x dances: x ΠD} by abstracting over the free variable that it is co-indexed with.
(56)
a. [Dare-ga
odorimasu] ka?
who-Nom dance
Q
'Who dances?' b. [[dare1 -ga odorimasu] Op1 ] ka who-Nom dance
Q
The Op and indeterminate phrases are generally interpreted as in (57), along the lines of Heim's (1982) analysis of indefinites (see also Berman 1991). In (57b), g'≈ 1,...,n g indicates that assignment g' is like assignment g, except possibly for values assigned to the variables 1, ..., n.
(57)
a. Op-indexing: Copy the index of each indeterminate phrase onto the c-commanding Op. b. ª IP Op1,...,n º g = {ª IP º g': g'≈ 1,...,ng} = {p: ∃g'[g'≈ 1,...,ng & p = ª IP º g']}
Applying (57), we obtain (58) for sentence (56a). The third line in (58) shows that since dare 'who' is interpreted as an open sentence, it adjoins to a position where it is interpretable, for example, VP or IP. It is then interpreted conjunctively with the rest of the sentence (see Heim 1982).
42
(58)
ª[[dare1 -ga odorimasu] Op1 ]º g = {p: ∃g'[g'≈ 1 g & p = ªdare1 -ga odorimasuº g']} = {p: ∃g'[g'≈ 1 g & p = ªdare1 [t1 -ga odorimasu]º g']} = {p: ∃x[p = that x is a person and x dances]}
If Yoko, Satoshi and Takumi are all the people, then the question denotes the set {that Yoko dances, that Satoshi dances, that Takumi dances}. This is a suitable basic denotation for the interrogative sentence in (56a). Multiple questions are interpreted in the same fashion, as shown in (59).
(59)
a. [Dare-ga
dare-to
odorimasu] ka?
who-Nom who-with dance
Q
'Who dances with whom?' b. [[dare1 -ga dare2 -to
odorimasu] Op1,2] ka
who-Nom who-with dance
Q
c. {p: ˛g'[g'≈1,2g & p = ªdare1 -ga dare2 -to odorimasuº g']} = {p: ˛x˛y[p = that x is a person and y is a person and x dances with y]}
2.3.4 Extension to the Direct Restrictor View of the Mo-Construction Japanese NPs do not carry any specification of definiteness, indefiniteness, singularity or plurality.25 Linguistic and non-linguistic context plays a role in narrowing 25One
could indicate plurality of animate objects by the use of –tati, and definiteness by the use of the demonstrative sono, as in (i). (i) Kodomo-tati-ga kyoositu-ni haittekita. Sono kodomo-tati-ga uta-o utatta. child-Pl-Nom classroom-to entered the child-Pl-Nom song-Acc sang 'Children came into the classroom. The children sang.'
43
down the possible interpretations of bare NPs. Due to the occurrence of bare NPs as part of mo-phrases, the mo-construction as a whole receives different readings. I assume that indeterminate phrases in the mo-construction introduce free variables, just like indeterminate phrases in the ka-construcion. If pianisuto 'pianist' in sentence (60) receives definite interpretation, the complex NP that mo is attached to denotes something like Íy[pianist(y) & student(x) & invite(y)(x)] (the pianist(s) that student x invited).
(60)
[[Dono gakusei-ga which
ø syootaisita] pianisuto]-mo kita.
student-Nom invited
pianist
-MO came
'Every pianist that a student had invited came.'
Since mo as a universal quantifier requires a set as its argument, the structure is not interpretable as it is. Here, the same operator Op as introduced in the interpretation of the ka-construction is inserted, for reasons of interpretability. It combines with the complex NP and forms the relevant set, {Íy[pianist(y) & student(x) & invite(y)(x)]: x ∈ D}, by abstracting over the free variable x.26 In (61), the index 1 on dono gakusei 'which student' is copied onto Op, following the rule of Op-indexing in (57a). And by applying (62), a generalized version of (57b), we obtain the denotation in (63).
(61)
[[[dono gakusei1 -ga
ø syootaisita] pianisuto] Op1 ]-mo kita.
which student-Nom invited (62)
pianist
-MO came
ª XP Op1,...,n º g = {ª XP º g': g'≈ 1,...,ng} = {z: ∃g'[g'≈ 1,...,ng & z = ª XP º g']}
26Free
variables introduced by indeterminate phrases should be distinguished from free variables associated with pronouns. I leave it open whether bare NPs in Japanese should receive an analysis along the lines of Heim's (1982) and Kamp's (1981) analyses of
44
(63)
ª[[dono gakusei1 -ga ø syootaisita] pianisuto] Op1 º g = {ª[dono gakusei1 -ga ø syootaisita] pianisutoº g ': g≈ 1 g'} = {ª[dono gakusei1 [t1 -ga ø syootaisita]] pianisutoº g ': g≈ 1 g'} = {Íy[pianist(y) & student(x1 ) & invite(y)(x1 )]: x1 Œ D} = Òz[˛x1 [z = Íy[pianist(y) & student(x1 ) & invite(y)(x1 )]]]
Once this denotation of the NP/DP is available, the universal quantifier mo, defined in (64), directly quantifies over it, as shown in (65).
(64)
ª mo º = ÒPÒQÅx[P(x) ∞ Q(x)]
(65)
ª[[dono gakusei1 -ga ø syootaisita pianisuto Op1 ] mo] kitaº g = ªmoº g (ªdono gakusei1 -ga ø syootaisita pianisuto Op1 º g )(ªkitaº g ) = Åz[˛x1 [z = Íy[pianist(y) & student(x1 ) & invite(y)(x1 )]] ∞ *come(z)]
(65) shows that mo allows quantification over plural individuals.27 As pointed out in an earlier section, however, the strong reading of sentence (61) by itself does not motivate this treatment of mo. The strong reading of (61) can be paraphrased as (66), or "every pianist that some student had invited came", which does not involve plural quantification.
(66)
Åy[˛x[pianist(y) & student(x) & invite(y)(x)] → come(y)]
indefinites in English. If it turns out that they should, and hence, they introduce free variables, then they also should be distinguished from indeterminate phrase free variables (see Kratzer 1991 for a related discussion). 27Every
in English can also quantify over pluralities, as shown in Schwarzschild's (1996, p. 78) example in (i). (i) One out of every three handguns in America is made by Smith and Wesson.
45
In order to see that mo can quantify over pluralities, we should look at partitive readings. The example in (67) is repeated from section 2.2.2. It can be paraphrased as: at least one from the set of pianists invited by student a came, at least one from the set of pianists invited by student b came, at least one from the set of pianists invited by student c came, ..., etc.
(67)
[[Dono gakusei-ga which
ø syootaisita] pianisuto]-mo hito-ri-wa
student-Nom invited
pianist
kita.
-MO one-Cl-at least came
'At least one pianist invited by each student came.'
We can derive the partitive reading by first making use of the denotation of the sister of mo which was responsible for the strong reading, repeated in (68) from above.
(68)
ª[[dono gakusei1 -ga ø syootaisita] pianisuto] Op1 º g = Òz[˛x1 [z = Íy[pianist(y) & student(x1 ) & invite(y)(x1 )]]]
The numeral quantifier hito-ri-wa 'at least one' (or hito-ri 'one') that occurs to the left of the matrix verb can be treated as a kind of VP modifier (Dowty and Brodie 1984, Fukushima 1991, 1993), as shown in (69).
(69)
a. ªhito-ri-waº = ÒPÒx˛z[z ≤ x & P(z)] b. ªhito-ri-wa kitaº = ªhito-ri-waº(ªkitaº) = ÒPÒx˛z[z ≤ x & P(z)](Òz[come(z)]) = Òx˛z[z ≤ x & come(z)]
46
Putting (68) and (69b) together, we now interpret sentence (67) as (70).
(70)
ªmoº (Òz[˛x1 [z = Íy[pianist(y) & student(x1 ) & invite(y)(x1 )]]])(Òx˛z[z ≤ x & come(z)]) = Åu[˛x[u = Íy[pianist(y) & student(x) & invite(y)(x)]] ∞ ˛z[z ≤ u & come(z)]]
Unlike (67), the example in (71), repeated also from section 2.2.2, does not have an overt numeral quantifier in the matrix clause. Still, as pointed out in that section, it allows a partitive reading like in (67) in an appropriate context.
(71)
[[Dono gakusei-ga which
ø syootaisita] pianisuto]-mo kita.
student-Nom invited
pianist
-MO came
'At least one pianist invited by each student came.'
We could assume something like an invisible counterpart of hito-ri-wa 'at least one' in (71). Alternatively, we could assume that the existential interpretation comes from the verb kita 'came' (see Carlson 1977 and Chierchia 1998). This question is independent of the mo-construction, since the simple sentence in (72) is ambiguous with respect to the interpretation of the bare NP pianisuto 'pianist'.
(72)
[[Yoko-ga ø syootaisita] pianisuto]-ga Yoko-Nom invited
pianist
kita.
-Nom came
'A pianist/pianists that Yoko had invited came.' 'The pianist(s) that Yoko had invited came.' 'One/some of the pianists that Yoko had invited came.'
47
Future investigation on how bare NPs in Japanese are interpreted in general would help us understand how exactly the interpretation of a sentence like (71) is derived. As pointed out in footnote 17, more than one indeterminate phrase can occur in the scope of mo, as in (73a). The translation is intended to indicate the strong reading. After Op-indexing in (57a) applies, its LF representation looks like (73b).
(73)
a. [[Dono gakusei-ga dono ie-ni
ø syootaisita] sensei]-mo kita.
which student-Nom which house-to invited
teacher - MO came
'Every teacher that some student invited to some house came.' b. [[[dono gakusei1 -ga dono
ie2 -ni ø syootaisita] sensei] Op1,2 ]-mo kita.
which student-Nom which house-to invited
teacher
- MO came
The semantics is derived as in (74), using (62), just like the case in which there was only one indeterminate phrase.
(74)
ª(73b)º = ªmoº(ª[[dono gakusei1 -ga dono ie2 -ni ø syootaisita] sensei] Op1,2º)(ªkitaº) = ªmoº(Òz[˛x1 ˛x2 [z = Íy[teacher(y) & student(x 1 ) & house(x2 ) & invite_to(y)(x2 )(x1 )]]])(Òx[come(x)]) = Åz[˛x1 ˛x2 [z = Íy[teacher(y) & student(x 1 ) & house(x2 ) & invite_to(y)(x2 )(x1 )]] ∞ come(z)]
We can assume that Op is freely available at LF. If there is no indeterminate phrase in its scope, it simply returns the denotation of its sister. If Op is freely introduced at LF, it may create sets in positions where they are not selected as arguments by suitable lexical items. In those cases, the computation will simply crash and the sentences will be ruled out as uninterpretable. Recall that in the local mo-construction as in (75), mo takes
48
the indeterminate phrase dono gakusei 'which student' directly. In this case, insertion of a lambda operator will make the set (λx[student(x)]) available from the indeterminate phrase for mo to combine with.
(75)
Dono gakusei-mo kita. which student-MO came 'Every student came.'
2.3.5 Mo as a Cross-Categorial Universal Quantifier We have focused on the non-local mo-construction in which mo takes an NP or DP that contains indeterminate phrases. As introduced at the beginning of this chapter, mo can also be attached to phrases other than NP/DP that contain one or more indeterminate phrases. For instance, in the examples in (76), mo takes PPs as its sister.
(76)
a. [Doko-kara]-mo syootaijoo-ga
todoita.
where-from- MO invitation card-Nom arrived 'An invitation card arrived from everywhere.' b. Yoko-wa [[dare-no tukue-no ue]-ni]
-mo
Yoko-Top who-Gen desk-Gen top -LOC- MO
ame-o
oita.
candy-Acc put
'Yoko put candies on everyone's desk.' c. Yoko-wa [[dare-ga
kita
hi] -ni]-mo byookidatta.
Yoko-Top who-Nom came day -on -MO sick_was 'Yoko was sick on any day someone visited her./For every person x, Yoko was sick on the day when x visited her.'
According to the direct restrictor view, mo universally quantifies over what its sister denotes. In (76), it quantifies over PP denotations. In this sense, mo is a cross-categorial 49
universal quantifier. Of course, we want to know what exactly the denotations of PPs are, and this is not a trivial question. Whatever the exact meanings of the PPs in (76) turn out to be, we can give a general definition of mo as in (77).
(77)
ª mo º = ÒPÒQÅx[P(x) ∞ Q(x)], where x Œ D† and P, Q Œ D<†,t>.
For instance, let us take the predicative PP sono kooen-ni 'in that park' in (78a), and turn it into the mo-construction as in (78b).
(78)
a. Yoko-ga
sono kooen-ni ita.
Yoko-Nom that
park-in
was
'Yoko was in that park.' b. Yoko-ga
[dono kooen-ni]-mo ita.
Yoko-Nom which park-in - MO was 'Yoko was in every park.'
With the help of the set forming operator Op, the PP in (78b) denotes a set of properties, namely, ÒP[˛y[P = Òx[x is in park y]]]. Since mo quantifies over this set, we obtain ÅP[˛y[P = Òx[x is in park y]] ∞ P(Yoko)], which says that every property of being in some park or other is true of Yoko. As mentioned at the beginning of this chapter, mo can also attach to a clausal constituent, as in (79).
(79)
a. Taro-wa [dare-ga
denwasi-te]-mo deru.
Taro-Top who-Nom call-TE
-MO answer
'No matter who calls, Taro answers.'
50
b. Taro-wa [Yoko-ga
itu
denwasi-te]-mo rusudatta.
Taro-Top Yoko-Non when call
-TE -MO was_not_home
'No matter when Yoko called, Taro was not home.'
Sentence (79a), for instance, would be paraphrased by Nishigauchi (1990) as "For every person x, if x calls, Taro answers", judging from his paraphrases of similar sentences. In the direct restrictor view, on the other hand, mo should quantify over what its sister denotes, namely, the set of situations in which someone calls. This seems to give rise to an appropriate semantics for the whole sentence: Every situation in which someone calls is extendable to a situation in which Taro answers. I leave more detailed discussion of the clausal mo-construction for future. Finally, how does the weak reading of the mo-construction, pointed out in section 2.2.2, arise? Let me first illustrate the weak reading using the following scenario. Mika, a new student from Japan, wants to buy a car. In particular, she is considering buying one of the Geo models (Geo Metro, Geo Prizm, Geo Storm, Geo Tracker). She asks me whether I know any Japanese in the area who own a Geo model, so that she could get some information from them. Then I can answer with (80).
(80)
[[ø Geo-no
dono kuruma-ni notteiru] nihonjin]-mo sitteiru.
Geo-Gen which car-on
drive
Japanese -MO know
'For every Geo model, I know a Japanese who drives it.'
This context makes it easier to see that sentence (80) has a reading in which it is judged to be false if, say, there is a model, say Geo Storm, that no Japanese in the area drives. Its semantic representation in (81), which is based on the embedded restrictor view, captures this intuition.
51
(81)
Åz[Geo_model(z) ∞ ∃y[Japanese(y) & drive(z)(y) & know(y)(I)]]
If the direct restrictor view is on the right track, then mo should be a rather flexible, cross-categorial universal quantifier that quantifies over whatever its sister constituent denotes. If we assume that the bare noun nihonjin 'Japanese' in (80) in this particular scenario is of generalized quantifier type (ÒP˛y[Japanese(y) & P(y)]), this would mean that the set forming operator Op creates a set of generalized quantifiers, as in (82), which mo quantifies over. Accordingly, the type of the basic mo needs to be shifted to (83) in this case.
(82)
ª [[ø Geo-no dono kuruma1 -ni notteiru] nihonjin] Op1 º
g
g'
= {X: ∃g'[g'≈ 1g & X = ª[[ø Geo-no dono kuruma1 -ni notteiru] nihonjin]º ]} = {X: ∃z[X = ÒP˛y[Japanese(y) & Geo_model(z) & drive(z)(y) & P(y)]]} (83)
ª mo º = Ò℘<<,t>,t>ÒQÅX<,t>[℘(X) ∞ X(Q)]
As a result of applying (83) to (82), and then to the matrix predicate 'I know', (84) is obtained. This is equivalent to (81).
(84)
ÅX<,t>[∃z[X = ÒP˛y[Japanese(y) & Geo_model(z) & drive(z)(y) & P(y)]] ∞ X(Òy[know(y)(I)]) ]
Quantification over generalized quantifiers in (84) may be avoided if it turns out that indefinite readings of bare NPs in Japanese are best described as involving an empty
52
determiner position that is interpreted as a choice function.28 Then mo can quantify over a set of individuals rather than a set of generalized quantifiers. Many questions remain, but I hope to have set a starting point for further investigation. I have assumed that the sister constituent of mo denotes (possibly plural) individuals (definite) or generalized quantifiers (indefinite) in the nominal moconstruction, but these are by no means final words on the matter. One cannot talk about definiteness or indefiniteness in Japanese easily, unless a much larger question of how bare NPs in Japanese are interpreted in general is addressed seriously.
2.4 The Direct Restrictor View and the Island Puzzle 2.4.1 The Island Puzzle As introduced in section 2.1.2, the locality effects observed in the association of indeterminate phrases and ka are exactly the same as the locality effects observed in the association of indeterminate phrases and mo. This was one of the reasons why we decided to look for a parallel analysis for the two constructions. Furthermore, the locality effects in the two constructions show a puzzling pattern, which I called the island puzzle. The relevant examples are repeated below from section 2.1.2. Example (85) is most readily interpreted as a yes-no question as indicated in (85a). The other interpretations indicated in (85b) through (85d), in which the indeterminate phrases dare 'who' and/or nani 'what' within the embedded interrogative clause take matrix interrogative scope, are at least much more difficult to obtain than (85a).
28This
was suggested to me by Angelika Kratzer (p.c.). See, for example, Reinhart (1997), Kratzer (1998) and Matthewson (1999).
53
(85)
Taro-wa [Yamada-ga
dare-ni nani-o
okutta ka] tazunemasita ka?
Taro-Top Yamada-Nom who-Dat what-Acc sent a.
Q asked
Q
'Did Taro ask what Yamada sent to whom?'
b. ?* 'Whox did Taro ask what Yamada sent to x?' c. *
'Whatx did Taro ask to whom Yamada sent x?'
d. ?* 'Whox did Taro ask whether Yamada sent what to x?'
The mo-construction in (86) shows the same pattern. Dare 'who' and/or nani 'what' in the embedded interrogative clause cannot be associated with the universal particle mo in the matrix clause. The only available reading is the one shown in (86a), in which mo is interpreted as 'also'. This is like mo in (87), in which no indeterminate phrase occurs in the scope of mo.
(86)
[[ ø [Yamada-ga [[
okutta ka] sitteiru] syoonin]-mo
dare-ni nani-o
[Yamada-Nom who-Dat what-Acc sent
Q] know] witness]-MO
damatteita. was_silent a.
'The witness who knew what Yamada sent to whom was also silent.' 'Even the witness who knew what Yamada sent to whom was silent.'
b. *
'For every person x, the witness who knew what Yamada sent to x was silent.
c. ** 'For every thing x, the witness who knew to whom Yamada sent x was silent.' d. *
'For every person x, for every thing y, the witness who knew whether Yamada sent y to x was silent.'
54
(87)
Sono syoonin-mo damatteita. that witness-MO was_silent 'That witness was also silent./Even that witness was silent.'
Just as the question particle ka blocks the association of indeterminate phrases and ka/mo as we have just seen, the universal particle mo has the same blocking effect. That the association with ka is blocked by the universal mo is exemplified in (88).
(88)
Yoko-wa [[[dono gakusei-ga
dono ie-ni
ø
Yoko-Top which student-Nom which house-to kita
syootaisita]
sensei]-mo
invited
teacher -MO
ka] siritagatteiru.
came Q want_to_know a.
'Yoko wonders whether for every student x, every house y, the teacher that x had invited to y came.'
b. ?* 'Yoko wonders for which student x, for every house y, the teacher that x had invited to y came.' c. *
'Yoko wonders for which house y, for every student x, the teacher that x had invited to y came.'
d. (?) 'Yoko wonders for which student x and for which house y, the teacher that x had invited to y also came.'
In the reading in (88a), dono gakusei 'which student' and dono ie 'which house' are both associated with mo, and the embedded question is interpreted as yes-no question. The ungrammatical (88b) and (88c) are the readings in which one of the indeterminate
55
phrases is associated with ka, across the universal mo. In the reading in (88d), both indeterminate phrases are associated with ka, and mo is interpreted as 'also'. 29 Similarly, in (89), the association with the higher universal mo is blocked by the lower universal mo.
(89)
[[ ø [[Dono gakusei-ga
nani-nituite ø kaita] ronbun]-mo yonda] sensei]-mo
which student-Nom what-about totemo
wrote paper -MO read
teacher -MO
tukareta.
very_much got_tired a.
'The teacher who read, for every student x, every topic y, the paper that x wrote on y also got very tired.'
b. ?* 'For every student x, the teacher who read, for every topic y, the paper x wrote on y got very tired.' c. *
'For every topic y, the teacher who read, for every student x, the paper x wrote on y got very tired.'
d. (?) 'For every student x, every topic y, the teacher who also read the paper that x wrote on y got very tired.'
In the reading in (89a), both dono gakusei 'which student' and nani 'what' are associated with the lower mo, and the higher mo is interpreted as 'also'. In the ungrammatical (89b) and (89c), one of the indeterminate phrases is associated with the higher mo, skipping the
29In
(88) and (89), associating both of the indeterminate phrases with the higher particle across the intervening mo 'also', as in the d-readings, does not seem to make the sentences syntactically bad (see Nishigauchi (1990, p. 213) for a different judgment; cf. (85d) and (86d)). To obtain these readings, appropriate contexts need to be set up that satisfy the presuppositions that come with mo 'also'. (Reducing the number of indeterminate phrases to one in these examples makes it easier to obtain the relevant type of readings, and still serves the purposes here.) This suggests that the universal mo and mo 'also' should be treated as two distinct lexical items.
56
lower universal mo. In (89d), both of the indeterminate phrases are associated with the higher mo, and the lower mo is interpreted as 'also'. When indeterminate phrases are embedded inside complex NPs or adjunct clauses, the sentences are fine in the ka-construction as shown in (90) and (91), as well as in the mo-construction as shown in (92) and (93).30
(90)
Taro-wa [[dare-ga ø katta] mochi]-o
tabemasita ka?
Taro-Top who-Nom bought rice cake -Acc ate
Q
'Whox did Taro eat rice cakes that x bought?' (91)
Taro-wa [dare-ga
kita-kara]
kaerimasita ka?
Taro-Top who-Nom came-because left
Q
'Whox did Taro leave because x came?' (92)
[[[[Dono T.A.-ga
ø osieta] gakusei]-ga ø syootaisita] sensei]-mo kita.
which T.A.-Nom taught student -Nom invited
teacher -MO came
'For every T.A. x, the teacher(s) that the student(s) that x had taught invited came.' (93)
[[ ø [Taro-ga
nani-o
katta-kara]
okotta]
hito]-mo
Taro-Nom what-Acc bought-because got_angry person -MO heya-o deteitta. room-Acc left 'For every thing x, the people who got angry because Taro had bought x left the room.'
The following table summarizes the pattern.
30See
footnote 7.
57
(94) (95)
*
[..... [..... ind ..... ]-ka/mo .....]-ka/mo z----------m
[..... [..... ind .....]CNP/Adjunct .....]-ka/mo z----------m The most prevalent view on the island puzzle, as we saw earlier, has been to take
the apparent wh-island effect in (94) to be suggestive of movement of indeterminate phrases (or some other items, as proposed, for example, in Watanabe 1992a, b and Hagstrom 1998) across islands in general. This is in analogy to English, in which whphrases move overtly and the movement shows a wh-island effect. In this view, the apparent lack of other island effects in (95) is taken to be a result of some special mechanism, which does not follow from anything inherent in the analysis. The analysis of the ka- and mo-constructions that I presented in section 2.3 now allows us to switch our perspective and read something else from the patterns in (94) and (95). In my analysis, indeterminate phrases do not undergo movement analogous to whmovement, even though they do adjoin to a position where they can be interpreted. The lack of complex NP and Adjunct island effects in (95) can be seen as suggestive of the general absence of movement across islands. The question now is where the apparent wh-island effect comes from if not from movement. In what follows, I will show that the island puzzle in fact follows from the interpretational process involved in the non-whmovement analysis.
2.4.2 The Apparent Wh-Island Effect as a Property of Unselective Binding The lack of Complex NP/Adjunct island effects schematized in (95) is expected in this analysis since there is no movement across islands. The task for the non-whmovement anlysis then is to account for the apparent wh-island effect schematized in (94). Notice that only in the case of (94), but not in (95), there is a particle ka or mo
58
intervening between indeterminate phrases and the matrix particle ka or mo with which they want to be associated. What seems to be disallowed is the configuration in (96a), where an index is copied onto the higher Op across the intervening Op. On the other hand, sentences are fine when the lower Op catches all the free variables in its scope, as in (96b).
(96)
a. *
[[..... [[[..... ind1 ..... ind2 .....] Op1 ]-ka/mo] .....] Op2 ]-ka/mo
b.
[..... [[[..... ind1 ..... ind2 .....] Op1,2]-ka/mo] .....]-ka/mo
The next section shows that there is some evidence that the type of co-indexing shown in (96a) should be disallowed as a general property of unselective binding.
2.4.3 Unselective Binding and Locality In section 2.3 I presented one particular way of spelling out the direct restrictor view of the mo-construction, in which free variables introduced by indeterminate phrases are bound by a c-commanding Op. I mentioned there that how these variables are bound goes along the lines of Heim's (1982) analysis of indefinites, according to which indefinites introduce free variables. Operators such as adverbs of quantification unselectively bind them, in the sense that they may bind more than one variable (see also Lewis 1975 and Kamp 1981). In particular, the Op-indexing introduced earlier is based on (97) from Heim (1982).
(97)
Quantifier Indexing: Copy the referential index of every indefinite NP as a selection index onto the lowest c-commanding quantifier.
59
When we apply the rule of quantifier indexing to the example in (98), the rough LF representation in (99a) is obtained. This gives rise to the truth condition that is equivalent to (99b), which can be paraphrased as (99c).
(98)
If a cat likes a friend of mine, I always give it to him.
(99)
a.
S always1,2
S S % # a cat1 likes [a friend of mine] 2 I give it1 to him2
b. ∀x1 ,x2 [[cat(x1 ) & friend_of_mine(x2 ) & like(x1 , x2 )] →[give(I, x1 , x2 )]] c. For every x, y, if x is a cat and y is a friend of mine and x likes y, I give x to y.
It is shown in Heim (1982) that the locality condition in the rule of Quantifier Indexing in (97) ("the lowest c-commanding quantifier") is necessary because otherwise we would predict a reading of sentence (98) that is not available. In (100a), the index 2 of the indefinite a friend of mine is copied onto the existential quantifier, skipping always. This representation, if possible, would give rise to the truth condition that is equivalent to the formula in (100b), which is paraphrased in (100c).
(100) a. ∃2 S always1
S S % # a cat1 likes [a friend of mine] 2 I give it1 to him2
b. ∃x2 ∀x1 [[cat(x1 ) & friend_of_mine(x2 ) & like(x1 , x2 )] →[give(I, x1 , x2 )]] c. There is some y, such that for every x, if x is a cat and y is a friend of mine and x likes y, then I give x to y.
60
As Heim points out, the sentence is incorrectly predicted to be true whenever there is some individual that is not a friend of mine. So the locality condition in (97) is necessary in order to avoid a representation like (100a).31 If this loclality requirement is a general property of this kind of index copying, then the island puzzle in fact follows from the analysis presented in section 2.3. Specifically, the rule of Op-indexing introduced earlier is now revised as (101), in which the locality requirement is built in.
(101) Op-indexing: Copy the index of each indeterminate phrase onto the lowest c-commanding Op.
The set-forming Op binds all variables in its scope introduced by indeterminate phrases unselectively. Thus, the patterns in (96) are expected in the analyisis — the lower Op should catch both indeterminate phrases. 32,33 To summarize, the direct restrictor view of the mo-construction made an analysis that involves only a very simple syntax-semantics mapping possible, which does not involve wh-movement. This then let us switch our perspective on the puzzling pattern of island effects observed in both ka- and mo-constructions. The lack of Complex NP and Adjunct island effects is no longer puzzling.34 The only island effect observed turns out 31See
Reinhart (1997) for a different view.
32See
Toyoshima (1996), who independently reached a very similar conclusion.
33This
analysis derives the effect of Minimal Scope Principle proposed in Tanaka (1998, 1999), which states that a wh-phrase must take the scope of the closest possible Q-marker at LF. 34This
suggests that, even though overt pied-piping is observed, for example, in Basque (Ortiz de Urbina 1989) and Imbabura Quechua (Hermon 1984), motivation for invisible large-scale pied-piping in Japanese of the kind proposed in Nishigauchi (1990) is lost (see also Webelhuth 1992).
61
to result not from movement, but from part of the process of interpretation. Thus the island puzzle in fact provides further support for the analysis pursued here.
2.5 An Alternative: Indeterminate Phrases as Denoting Sets of Individuals There is an alternative way of spelling out the semantics of the ka- and moconstructions, which also builds upon an existing analysis in the area of interrogatives just as in the analysis presented so far. That is to extend Hamblin's (1973) proposal about English wh-pronouns to Japanese indeterminate phrases.35 The alternative analysis of indeterminate phrases may turn out to be favored over the analysis I presented above, in which indeterminates introduce free variables, with respect to the question of how the island puzzle is derived. I leave, however, a detailed spelling out of the analysis and discussion of empirical consequences for the future, for reasons of time. According to Hamblin (1973), wh-pronouns like who and what denote sets of individuals. For instance, the denotation of who is the set of all human individuals. If we apply this analysis to the Japanese example in (102), dare 'who' denotes the set of human individuals as in (103a). The denotations of lexical items that are not indeterminate phrases are singleton sets whose sole members are their normal denotations. For example, odorimasu 'dance' in (102) denotes a singleton set whose only member is its ordinary denotation, as shown in (103b).
(102) [Dare-ga odorimasu] ka? who-Nom dance
Q
'Who dances?'
35Ramchand
(1997) extends Hamblin's proposal to indeterminate phrases in Bengali. Hagstrom (1998) extends the idea to indeterminate phrases in the Japanese kaconstruction. See also Lin (1996), who assumes a free variable analysis for Chinese whphrases in some contexts, and a set analysis for those in other contexts.
62
(103) a. ªdareº w = {x Œ De: human(x)(w)} b. ªodorimasuº w = {ÒxÒw[dance(x)(w)]}
We compose the NP dare 'who' and the VP odorimasu 'dance' by applying the function ÒxÒw[dance(x)(w)] to each member of the set of humans, and thus forming a set of propositions. A general rule for semantic composition can be stated as follows, which is based on Rooth (1985, 1996).
(104) Let å be a complex phrase with component phrases å1 , ..., åk , and let Ï be the semantic rule for å, e.g., functional application. The semantic value of å is the set of objects that can be obtained as Ï(x1 , ..., xk ), where x1 Œ ªå1 º ¡ ... ¡ xk Œ ªåk º.
For sentence (102), we obtain the set of propositions in (105).
(105) ªdare-ga odorimasuº w = ªodorimasuº w(ªdareº w) = {f(x): f Œ ªodorimasuº w, x Œ ªdareº w} = {p: ˛x[human(x)(w) & p = Òw[dance(x)(w)]]}
This is a suitable basic denotation for the interrogative sentence (102), along the lines of Hamblin (1973) and Karttunen (1977). If indeterminate phrases denote sets of individuals, they trigger the generation of non-singleton Hamblin denotations compositionally, up to a point where an operator is encountered that selects a non-singleton Hamblin denotation. Suppose that ka and mo are such operators. In example (85), repeated below as (106), the denotation of the embedded IP is as in (107).
63
(106) Taro-wa [[ IPYamada-ga Taro-Top a.
dare-ni nani-o
okutta] ka] tazunemasita ka?
Yamada-Nom who-Dat what-Acc sent
Q
asked
Q
'Did Taro ask what Yamada sent to whom?'
b. ?* 'Whox did Taro ask what Yamada sent to x?' c. *
'Whatx did Taro ask to whom Yamada sent x?'
d. ?* 'Whox did Taro ask whether Yamada sent what to x?' (107) {p: ˛a,b[p = Òw[Yamada sent a to b in w]]}
Though I cannot provide a definite answer to the question of what the exact semantic contribution of ka is at this point, it seems clear that both of the indeterminate phrases, dare 'who' and nani 'what', are caught by the first ka they encounter.36,37 In this alternative, then, the island puzzle follows from the analysis more naturally than in the free variable analysis of indeterminate phrases. There are no variables involved in the interpretation of indeterminate phrases. Therefore there is no need to assume the setcreating operator, and no need for the Op-indexing rule that has the locality condition built in. Furthermore, the alternative analysis may be favored over the free variable analysis in that the source of the polarity sensitivity of indeterminate phrases may be stated in simpler terms. The alternative analysis could say that the indeterminate phrases are the only items that have the Hamblin denotations that are non-singleton sets, and that the particles ka and mo are the only items that select such denotations. If the final denotations of sentences should end up as singleton sets, indeterminate phrases always
36One
idea, due to Angelika Kratzer (p.c.), is that ka is an operator that takes a nonsingleton Hamblin set and returns a singleton set whose sole member is, for example, a question denotation as proposed in Groenendijk and Stokhof (1982). See Heim (1994).
37See
Wold (1996) for a related discussion in the area of association of focus.
64
find themselves in the scope of ka or mo, and ka and mo always find indeterminates in their scope.
2.6 A Further Look at the Embedded Restrictor View In section 2.2.4, I briefly mentioned that initial support for the embedded restrictor view was only apparent. This section gives a further look at some of the issues discussed there.
2.6.1 Anaphora Facts Certain anaphora facts observed in the mo-construction, which we saw in section 2.2.4, appeared to support the embedded restrictor view at first sight. For example, in sentence (108a), repeated from that section, both the variable associated with the indeterminate phrase dono gakusei 'which student' and the variable associated with the pronoun in the matrix clause sono 'his' seem to be bound by the universal quantifier. That this kind of variable binding is possible is expected in the embedded restrictor view, which assumes (108b) as the translation of the sentence. (For simplicity, I assume that pianisuto 'pianist' here receives a definite singular interpretation.)
(108) a. [[Dono gakusei1 -ga ø syootaisita] pianisuto]-mo sono1 sensei datta. which student-Nom
invited
pianist
-MO his
teacher was
'For every student1 , the pianist he1 had invited was his1 teacher.' b. Åx[student(x) ∞ teacher_of(x)(ιy[pianist(y) & invite(y)(x)])]
In the direct restrictor view, on the other hand, we can assume that sono 'his' in (108a) is an E-type pronoun, which is interpreted as something like "the student who had invited him or her", as shown in (109).
65
(109) Åz[˛x[z = ιy[pianist(y) & student(x) & invite(y)(x)]]] ∞ teacher_of(ιx[student(x) & invite(z)(x)])(z)]
Thus, either the bound variable analysis or the E-type analysis seems to work for sono 'his' in example (108a). The following pair of examples, however, show that there are cases in which only the E-type strategy can account for the interpretation of pronouns that occur in the moconstruction. Let us first consider sentence (110a). Like in (108a), sono 'its' in sentence (110a) can be interpreted either as a bound variable pronoun as in (110b), or as an E-type pronoun, as in (110c).
(110) a. [[Dono hune-o
eranda] hito] -mo sono nedan-ni odoroida.
which boat-Acc chose person- MO its
price-at was_surprised
'Every person who chose some boat was surprised at its price.' b. 'For every boat x, the person who chose x was surprised at the price of x.' c. 'For every person y who chose some boat, y was surprised at the price of the boat y chose.'
Now, sentence (111a) is based on (110a), but is slightly more complex. For this sentence, the paraphrase according to the bound variable pronoun analysis in (111b) does not provide an appropriate interpretation. In (111b), the variables associated with dono kaisya 'which company' and sono 'its' are co-bound by the universal quantifier. Instead, sono 'its' should be interpreted as an E-type pronoun, as in (111c).
66
(111) a. [[[[Dono kaisya-ga
tukutta] hune]-o
which company-Nom made
eranda] hito] -mo sono nedan-ni
boat -Acc chose person -MO its
price-at
odoroita. was_surprised 'Every person who chose a boat that some company made was surprised at its price.' b. *'For every company x, the person who chose a boat that x made was surprised at the price of x.' c. 'For every person y who chose a boat that some company made, y was surprised at the price of the boat y chose.'
Thus, even though the embedded restrictor view might appear to provide a simpler semantic representation for anaphora facts such as (108a) and (110a), cases like (111a) in which only the E-type strategy works indicate that examples like (108a) and (110a) by themselves do not necessarily support the embedded restrictor view.
2.6.2 The Mapping from Syntax to Semantics in the Embedded Restrictor View 2.6.2.1 The LF of the Ka-Construction Extended to the Mo-Construction Recall from section 2.2.4 that one apparent advantage of the embedded restrictor view was that it seemed to fit with the idea of LF movement of indeterminate phrases in the ka-construction. This provides nice parallel structures for the ka- and moconstructions depicted in (112), repeated from that section.
(112) a.
CP 2 indeterminate 2 IP C ka @ Q ....t....
b.
67
MoP 2 indeterminate 2 XP mo @ ∀ ....t....
A closer look at the syntax-semantics mapping, however, will show that despite what is believed in some syntactic literature, it is not a straightforward matter to derive the semantics based on the embedded restrictor view from the structure in which indeterminate phases are moved to the specifier of MoP. The question is whether we can derive (113b), the translation of sentence (113a) according to the embedded restrictor view, from the LF in (113c).38
(113) a. [[Dono gakusei-ga which
ø syootaisita] sensei]-mo kita.
student-Nom invited
teacher -MO came
'For every student x, the teacher(s) x invited came.' b. Åx[student(x) ∞ *come(Íy[teacher(y) & invite(y)(x)])]
38In
Nishigauchi's (1990) original proposal, the whole complex NP occupies the specifier position of MoP at LF. For the mo-phrase in his own example in (i), the LF in (ii) is assigned. I use Mo' for his C', and MoP for his CP, to stay away from the question of the categorial status of mo. (ii) also shows that dare 'who' moves to the specifier position of the relative clause CP (see section 2.1.4). (i) [[Dare-ga ø kaita] tegami]-ni-mo onazi koto-ga kaite-atta. who-Nom wrote letter -in -MO same thing -Nom written-was (ii) [MoP [NP[CP dare1 [IP t1 -ga ø kaita]] tegami]2 [Mo'[PP t2 -ni] mo]] … It is then assumed that the LF in (ii) "can be 'interpreted' as (p. 163)" (iii). However, it is not shown how this can be achieved. (iii) 'For all x, y, x a person, y a letter, x wrote y, the same thing was written in y.' Von Stechow (1996) provides a revised analysis, in which it is assumed (i) that the entire complex NP moves (as in Nishigauchi's analysis) for syntactic reasons at the first invisible level, and (ii) that at the second invisible level, the material other than the indeterminate phrase is syntactically reconstructed. The result looks like (113c).
68
c.
IP 5 MoP VP 3 # dono gakusei-ga1 3 kita which student-Nom NP mo came %every t1 syootaisita sensei invited teacher
We are now faced with at least two difficulties. First, mo in (113) has to be a different lexical item than the one that occurs in the simple local case in (114).
(114) a. Dono gakusei-mo kita. which student-MO came 'Every student came.' b.
IP 4 MoP VP 3 # dono gakusei mo kita which student every came
Mo in (114b) is just like English every, in the sense that it first combines with its restrictor, and then with its scope. On the other hand, mo in (113c) first combines with part of its scope, and then with its restrictor. Thus we need to assume two radically different mos as in (115): (a) for (114) and (b) for (113).39 This is in fact what von Stechow (1996) assumes. We should then make sure, for instance, that mo2 in (115b) is not used in (114). 39Further,
(115b) should be generalized so that it accommodates the occurrence of multiple indeterminate phrases in mo's scope, as in (ia). (i) a. [[Dare-ga doko-de ø kaita] hon]-mo omosiroi. [[who-Nom where-at wrote] book]-MO interesting 'The books that whoever wrote wherever are interesting.' b. For everyone x and every place y, the books that x wrote at y are interesting.
69
(115) a. mo1 :
λPλQ∀x[P(x) → Q(x)]
for (114)
b. mo2 :
λPλQ∀x[Q(x) → P(x)]
for (113)
Second, assuming the special kind of mo in (115b) is of course not enough to derive the semantics of the entire sentence, due to the discontinuous scope of mo in (113c). The first argument of mo in (113c) is only a part of its scope, and the VP denotation is the other part of the scope. In order for the denotation of the VP to end up as part of mo's scope, a special mechanism is necessary. Such a mechanism is not impossible to come up with, as we will see shortly. However, the mechanism is a technical one, with ad hoc assumptions. Von Stechow (1996) attempts to provide an anlysis that does exactly what we are trying to do here: to derive the standard translation (113b) from the LF in (113c). Although providing a thorough analysis of the mo-construction is not one of the main concerns of his article, it is worth looking at his analysis since it is the only explicit analysis of the syntax-semantics mapping of the mo-construction we can find in the literature. First of all, von Stechow (1996) assumes the special mo in (115b) above for the non-local mo-construction. With respect to the second difficulty that I mentioned above, namely, the difficulty of discontinuous scope, he proposes what he calls the "inverse linking mechanism". (116) is the informal description of the mechanism with minor modification so that it fits our example (113a).
(116) λQ [MoP every student x, the teachers that x invited Q] (came) = For every student x, the teachers that x invited came (von Stechow 1996, pp. 92-94) 70
It is not clear to me what exactly the connection is between (116) and the inverse linking construction in English, but what (116) is intended to do is clear: the predicate 'came' is lambda-ed in to the position of the silent predicate Q of type . In (117), I present the LF for sentence (113a), based on von Stechow's partial tree (his (62)) for a structually similar sentence.40
(117)
∀x[student(x) → *came(Íx3 [teacher(x 3 ) & invited(x3 )(x)])]
5
λQ∀x[student(x) → *Q(Íx3 [teacher(x 3 ) & invited(x3 )(x)])] VP g # ∀x[student(x) → *Q(Íx3 [teacher(x 3 ) & invited(x3 )(x)])] kita MoP came 5 Mo' λR∀x[R(x) → λx1 [student(x1 )] dono gakusei-ga which student-Nom 4 *Q(Íx3 [teacher(x 3 ) λx2 [*Q(Íx3 [teacher(x 3 ) & invited(x3 )(x2 )])] mo & invited(x3 )(x)])] g λPλR∀x[R(x) → P(x)] *Q(Íx3 [teacher(x 3 ) & invited(x3 )(x2 )]) 4 Íx3 [teacher(x 3 ) & invited(x3 )(x2 )] NP ? % *Q [t2 ø syootaisita] sensei invited teacher
Notice that the silent predicate Q is generated before NP combines with mo, and it is abstracted over at the point where MoP combines with VP. The effect of this is that the VP denotation, which is the other half of mo's scope, is successfully put into the position where Q is generated.
40Recall
that von Stechow's (1996) system requires the whole complex NP to move to the Spec of MoP for syntactic reasons at a level where island constraints apply, as in Nishigauchi's (1990) proposal (see footnote 38). (117) is the structure in which the complex NP minus dono gakusei 'which student' has been reconstructed.
71
Let us suppose for a moment that a silent predicate of this kind is independently available in Japanese. Then we would expect it to appear in positions where predicates appear as in (118a). Further, we would expect it to be anaphoric to some other linguistic material, like soo sita 'did so' in (118b), which is anaphoric to kaetta 'left'. Contrary to these expectations, (118c) not only lacks the anaphoric interpretation, but is ungrammatical. (P in (118c) is a silent predicate.)
(118) a. Yoko-ga
kaetta ato de Taro-ga
Yoko-Nom left
after
kaetta.
Taro-Nom left
'Taro left after Yoko left.' b. Yoko-ga
kaetta ato de Taro-ga
Yoko-Nom left
soo sita.
after Taro-Nom so did
'Taro left after Yoko left.' c. *Yoko-ga kaetta ato de Taro-ga P Yoko-Nom left
after Taro-Nom
'Taro left after Yoko left.'
This shows that silent predicates are not freely available in Japanese, and it must be stipulated that they are only generated as a sister of NP that combines with mo. Even if we stipulate that silent predicates are only allowed to sprout up right before mo combines with NP, it still holds that they cannot be anaphoric to other predicates, as (119) shows.
(119) *Yoko-ga
kaetta ato de [[dare-ga
Yoko-Nom left
after
syootaisita sensei] P ]-mo
who-Nom invited
teacher
-MO
'After Yoko left, for every person x, the teacher x had invited left.'
72
What needs to be added as stipulation is a requirement that the property variable be lambda-abstracted at the right point, namely, immediately after mo combines with its sister. If we didn't have such a requirement, the sentence in (120) would be allowed.
(120) *Yoko-wa [[[dare-ga syootaisita sensei] P ]-mo]-to sakenda omotta. Yoko-Top who-Nom invited
teacher
- MO-that shouted thought
'Yoko thought that for every person x, the teacher x had invited shouted.' 'Yoko shouted that for every person x, the teacher x had invited thought.'
Since the underlined constituent in (120) denotes a truth-value, it should in principle be able to occur as a complement clause of some appropriate verb. Depending on whether the lambda-abstraction takes place before or after the complement clause combines with sakenda 'shouted', either sakenda 'shouted' or omotta 'thought' ends up in the position of the silent predicate, and the sentence should be fine. The ungrammaticality of (120) shows that where the lambda-abstraction must take place needs to be stipulated. Extending the standard LF movement analysis of the ka-construction to the moconstruction as shown in (112) seemed promising at first sight, given the standard embedded restrictor view of the construction. I have shown that the movement analysis of the mo-construction based on the analysis of the ka-construction does not give us a straightforward way of mapping from syntax to semantics. In particular, in order to derive the standard translation from the structure in (112b), we need to (i) postulate a special mo for the non-local mo-construction, (ii) postulate silent predicates which are only allowed to occur in the mo-construction, and (iii) stipulate that the silent predicates are abstracted over immediately after mo combines with its sister.41
41Furthermore,
the inverse linking mechanism does not directly apply to the clausal moconstruction as in (i).
73
2.6.2.2 Other Possible LF Representations of the Mo-Construction One advantage of von Stechow's (1996) LF for the mo-construction, a revised version of Nishigauchi's (1986, 1990) LF, was that it paralleled the standard LF for the ka-construction. I have shown, however, that there are difficulties in deriving the standard semantics based on the embedded restrictor view from the proposed LF. Other LF representations for the mo-construction are conceivable, once we consider alternative LFs for the ka-construction, or depart from the idea of pursuing a parallelism between the mo- and ka-constructions. For example, the S-structure of (121a) may be passed on to LF, with no invisible movement. This is an extension of Hagstrom's (1998) idea about the ka-construction, presented in Takahashi (1999). A simplified LF should look like (121c).
(121) a. [[Dono gakusei-ga which
ø syootaisita] sensei]-mo kita.
student-Nom invited
teacher -MO came
'For every student x, the teacher(s) x invited came.' b. Åx[student(x) ∞ *come(Íy[teacher(y) & invite(y)(x)])] c.
IP 2 MoP VP 2 NP mo # … indeterminate …
Takahashi (1999), however, does not show how the standard semantics based on the embedded restrictor view, which he adopts, can be derived from this LF.
(i) Taro-wa [dare-ga denwasi-te]-mo deru. Taro-Top who-Nom call -TE -MO answer 'Taro answers no matter who calls.'
74
Other than the structure I reviewed in section 2.6.2.1, there are ways of rearranging the pieces of the sentence in order to arrive at LF representations that are better suited for the purpose of deriving the standard semantics based on the embedded restrictor view. For instance, (122a) and (122b) are possible LFs obtained by moving both indeterminate phrases and mo.
(122) a.
4 2 IP indeterminate1 mo2 2 MoP VP 2 NP t2 ! …t1 …
b.
wgo mo2 indeterminate1 IP 2 MoP VP 2 NP t2 ! …t1 …
It is not difficult to see how we can derive the semantics based on the embedded restrictor view in (121b) from either of these structures. However, these movement analyses have problems of accounting for the island puzzle, just as in the case of the movement analysis discussed above. Furthermore, we may have to drop the idea of providing a uniform analysis of the ka- and mo-constructions.
2.6.2.3 Summary I have examined an attempt to achieve parallel analyses for the ka- and moconstructions by extending the LF wh-movement analysis of the former to the latter, a general idea pursued in Nishigauchi (1986, 1990) and von Stechow (1996). I have shown that once we pay attention to the syntax-semantics mapping of the mo-construction, the apparent parallelism fails. The question of syntax-semantics mapping in the kaconstruction did not arise since its LF was modeled after wh-questions in English, whose analysis was expected to extend to the Japanese ka-construction.
75
This result, along with the brief review of other possibilities in section 2.6.2.2, indicates that deriving the standard semantics of the mo-construction based on the embedded restrictor view by rearranging the parts of the sentences at LF is not a straightforward matter, contrary to first appearances. Even though one cannot dismiss the embedded restrictor view simply because the mapping is difficult, here we see a clear contrast to the straightforward syntax-semantics mapping that the direct restrictor view offers.
2.6.3 The Nature of the Restriction on the Domain of Universal Quantification 2.6.3.1 Different Predictions in the Two Views Let us now turn to some semantic predictions that the embedded restrictor view makes. Recall that the strong reading of sentence (123a), repeated from earlier sections, is represented as (123b) in the embedded restrictor view proposed by Ohno (1989) and von Stechow (1996). It is represented as (123c) in the direct restrictor view proposed in section 2.3. (123d) is Nishigauchi's (1990) representation, which amounts to the same as (123c) in this case.
(123) a. [[Dono gakusei-ga which
ø syootaisita] pianisuto]-mo kita.
student-Nom invited
pianist
-MO came
'Every pianist that a student had invited came.' b. Åx[student(x) ∞ *come(Íy[pianist(y) & invite(y)(x)])] c. Åz[˛x[z = Íy[pianist(y) & student(x) & invite(y)(x)]] ∞ *come(z)] d. Åx,y[[student(x) & pianist(y) & invite(y)(x)] ∞ come(y)]
Notice that the embedded restrictor view and the direct restrictor view make different claims with respect to the restriction on the domain of students. Let us consider the scenario from section 2.2.1 again: Three students and three teachers each invited a
76
pianist (six pianists all together), and three other students each invited a drummer to a party. The three pianists invited by the students came to the party. In (123c), if an individual assigned to x was not a student, or was a student who did not invite any pianist, then the equation fails for any individual z. The sentence is predicted to be true if and only if every pianist invited by some student came. This conforms to our intuitions, and the sentence is judged to be true in this situation. According to (123b), on the other hand, the sentence is true if and only if for every student x, the pianist(s) invited by x came. Recall that in the above scenario, not every student invited a pianist. In this analysis, the fact that the sentence is judged to be true in the scenario is accounted for by appealing to (local) accommodation of the existence presupposition that there are pianists invited by a given student x. The existence presupposition supposedly comes from whatever is responsible for the definitelike interpretation of pianisuto 'pianist'. As a result, we get something like (124).
(124) Åx[student(x) & ˛z[pianist(z) & invite(z)(x)] ∞ *come(Íy[pianist(y) & invite(y)(x)])]
This way, if a given student invited no pianist, he or she is not in the domain of the universal quantification. (See, for example, Heim 1983 and Beaver 1995 for general discussions of presupposition accommodation from the nuclear scope into the restrictive clause.) Thus, the ways in which the two analyses get the appropriate restriction on the domain of indeterminate phrase denotation are quite different. In (123c), that we only look at the set of students who invited some pianist(s) is part of the semantics. In (123b), on the other hand, it is a result of a pragmatic process of presupposition accommodation.
77
The example in (125a) also shows a similar point. Suppose that Yoko put apples in basket 1 and basket 2, and Satoshi put apples in basket 3; Taro tried an apple from basket 1 and one from basket 2, as depicted in (125b).
(125) a. Taro-wa [[Yoko-ga
dono kago-ni ø irete oita] ringo]-mo azimisita.
Taro-Top Yoko-Nom which basket-in put Aux apple -MO tried 'Taro tried some of the apples that Yoko had put in any basket.'
b.
z---m z---m z---m
basket1 Yoko
basket2 Yoko
( = apple that Taro tried)
basket3 Satoshi
Sentence (125a), in its partitive reading, is judged to be true in this scenario.42 This indicates that the domain of baskets is restricted so that one only looks at those baskets that Yoko put some apples in. Following the embedded restrictor view, one would paraphrase the sentence as "For every basket x, Taro tried some of the apples that Yoko had put in x". Then the domain of quantification should be narrowed, by appealing to accommodation of an existence presupposition that there are apples that Yoko had put in basket x.
2.6.3.2 Burdens on Presupposition Accommodation The following data may suggest that the embedded restrictor view puts too much burden on the pragmatic process of presupposition accommodation. Suppose that we are talking about housing situations in the Pioneer Valley, and their impact on the graduate 42It
seems to me that if one puts focus on dono kago 'which basket', a reading of the sentence emerges that is felt to involve presupposition failure in this context. However, this is not always the case with other structurely similar sentences. More detailed examination is necessary.
78
students at our department. Ten of them live in Amherst, and the other ten live in Northampton. Recently the landlord association of the Pioneer Valley announced a significant increase in apartment rents for the next academic year. The sentence in (126) is judged to be true in a situation where all ten of the grad students who live in Northampton complained.
(126) [[ ø Northampton-no dono apaato-ni
sundeiru] daigakuinsei]-mo
Northampton-Gen which apartment-in live monku-o
grad student - MO
itta.
complaints-Acc said 'Every grad student who lives in an apartment in Northampton complained.'
In the embedded restrictor view, the sentence is translated as (127).
(127) Åx[apartment(x) & in_Northampton(x) ∞ *complain(Íy[grad_student(y) & live_in(x)(y)])]
In order to capture the intuition that in sentence (126), what is relevant is not just any apartment in Northampton, but only those apartments where graduate students live, the embedded restrictor view needs to assume (i) that there is an existence presupposition that there is/are (a) graduate student(s) who live(s) in a given apartment x, and (ii) that it is accommodated locally as in (128). (128) says: Every apartment in Northampton in which some graduate student(s) live(s) is such that the graduate student(s) who live(s) in it complained.
(128) Åx[apartment(x) & in_Northampton(x) & ˛z[grad_student(z) & live_in(x)(z))] ∞ *complain(Íy[grad_student(y) & live_in(x)(y)])]
79
Thus, as a result of accommodation, the relevant set of apartments becomes smaller. Coupled with the definite-like semantics of the bare NP daigakuinsei 'graduate student', (128) predicts the same truth-conditions as (129), which is how the direct restrictor view translates sentence (126).
(129) Åz[∃x[z = Íy[grad_student(y) & apartment(x) & in_Northampton(x) & live_in(x)(y)]] ∞ *complain(z)]
In the embedded restrictor view, however, it is expected that the hearer can always accommodate the presupposition with ease so that he or she revises (127) to (128). How to get from the syntax of sentence (126) to translation (127) is not a straightforward matter to begin with, as we saw in section 2.6.2. A further claim in the embedded restrictor view is that once the sentence receives the semantics in (127), accommodation takes place easily and yields (128). Let us now consider the sentence in (130), which is a "back-translation" of (127). Sentence (130) is designed in such a way that its surface syntax can be straightforwardly mapped to translation (127). In this sentence, Northampton-no dono apaato-mo 'which apartment in Northampton-MO' occurs sentence-initially, and a pronoun soko 'there' fills the position occupied by Northampton-no dono apaato 'which apartment in Northampton' in the earlier sentence in (126).
80
(130) Northampton-no
dono apaato-mo,
[[ø soko-ni sundeiru] daigakuinsei]-ga
Northampton-Gen which apartment-MO monku-o
itta.
complaints-Acc
said
there-in live
grad student-Nom
'For every apartment in Northampton, the student(s) who live there/in it complained.'
When this sentence is uttered in the same context as (126), the presupposition that there is/are (a) graduate student(s) who live(s) in a given apartment x cannot get accommodated into the restrictive clause in the way shown in (128). Rather, the presupposition is (globally) accommodated into the common ground, so that the sentence suggests that every apartment in Northampton is inhabited by graduate students. The embedded restrictor view has to rely on accommodation for restricting the domain of universal quantification in (126), while for (130), the same kind of accommodation should not apply. To show that this is in fact what is going on, independent evidence is needed that the different surface syntax in (126) and (130) plays a crucial role in the possibility of accommodation. I leave this issue for future studies.43 43In
the analysis presented in section 2.5 based on Hamblin (1973), in which indeterminate phrases denote sets of individuals, dono gakusei 'which student' in example (123a) denotes, for example, {Yoko, Satoshi, Takumi}, if these are all the students. The sister constituent of mo then denotes the set in (i). (i) {Íy[pianist(y) & invite(y)(Yoko)], Íy[pianist(y) & invited(y)(Satoshi)], Íy[pianist(y) & invite(y)(Takumi)]} Suppose that Takumi didn't invite any pianist. Assuming that maximal individuals that are pianists invited by some student are only defined on non-empty sets, the last member of the set in (i) is not defined. Thus the whole set in (i) ends up undefined. The set mo quantifies over in this example should be (ii). (ii) {Íy[pianist(y) & invite(y)(Yoko)], Íy[pianist(y) & invited(y)(Satoshi)]} This is not a problem specific to the mo-construction. As pointed out to me by Angelika Kratzer (p.c.), in Hamblin's (1973) semantics of wh-phrases, there must be special provisions for undefinedness for cases like: Whose dog is barking?.
81
The following examples may pose a similar difficulty for the embedded restrictor view. Each of the underlined phrases in (131) consists of an indeterminate phrase and an appropriate classifier. The indeterminate phrase nan 'what' in these examples ranges over (natural) numbers.
(131) a. [[ ø Konpyuutaa-o nan-dai motteiru] sensei]-mo atarasii no-o computer-Acc what-Cl have
teacher – MO new
tyuumonsita.
one-Acc ordered
'Every teacher who owns any number of computers ordered a new one.' b. [[ ø Sono wain-o that
wine-Acc
nan-bai
nonda] hito]
-mo
what-Cl
drank] person –MO
yopparatta. got_drunk
'Everyone who had any number of glasses of that wine got drunk.' c. [[ ø Tokyo-ni nan-do Tokyo-to what-times Yanaka-o
susumeta.
Yanak-Acc
recommended
itta koto-ga aru]
amerikajin]-mo
have_been_to
American –MO
'Every American who has been to Tokyo any number of times recommended Yanaka.'
According to the embedded restrictor view, the sentences in (131) all involve universal quantification over numbers. For example, (131a) is translated as (132a).
(132) a. Ån[number(n) ∞ ˛z[order(z)(Íx[teacher(x) & ˛y[own(y)(x) & computer(y) & | y | = n]]) & new computer(z)]] b. Ån[[number(n) & ˛x[teacher(x) & ˛y[own(y)(x) & computer(y) & | y | = n]]] ∞ ˛z[order(z)(Íx[teacher(x) & ˛y[own(y)(x) & computer(y) & | y | = n]]) & new computer(z)]]
82
The fact that speakers have no trouble understanding what these sentences mean forces the analysis to claim that the domain of numbers is always narrowed with ease via local presupposition accommodation, yielding (132b). As we have seen above, this claim requires further empirical evidence.44
2.6.3.3 Further Questions The mo-construction allows the strong reading, the partitive reading and the weak (pure indefinite) reading. Our expectation now is that whenever the head NP of the relative clause receives a definite-like interpretation, either the strong or the partitive reading arises, and therefore no apparent universal quantification over indeterminate phrase denotations. I point out here that there are certain cases that are not as clear as we want them to be. The sentence in (133) is interpreted as expected, namely, every female governor elected in some prefecture or other recently attended the meeting in Tokyo.
44If
we placed the mo-phrase in (131a) in a context that induces a pure indefinite reading, as in (ib), it sounds pragmatically odd. This is expected not only in the direct restrictor view, but also in the embedded restrictor view, since the domain of natural numbers is infinite. (i) a. Kono daigaku-ni-wa [[ ø konpyuutaa-o roku-dai motteiru] sensei]-ga this university-at-WA computer-Acc six-Cl have teacher-Nom 'There is a teacher who owns six computers at this university.' b. #Kono daigaku-ni-wa [[ ø konpyuutaa-o nan-dai motteiru] sensei]-mo this university-at-WA computer-Acc what-Cl have teacher–MO 'For every number n, there is a teacher who owns n many computers at this university.'
83
iru. is iru. is
(133) [[ø Konkai-no
senkyo-de dono ken-kara
sensyutusareta]
this_time-Gen election-in which prefecture-from was_elected jyosei chiji]
-mo Tokyo-de-no kaigi-ni
syussekisita.
female governor –MO Tokyo-in-Gen meeting-Dat attended 'Every female governor who was elected in any prefecture in the recent elections attended the meeting in Tokyo.'
When the restriction on jyosei chiji 'female governor' is not in the form of a relative clause, but something much smaller, for example, the genitive phrase in (134), intuitions become less clear. With an appropriate context, the sentence does have an expected reading, in which it is true in a situation where every female governor of some prefecture or other attended the meeting, even when, say, only five of the forty-seven prefectures had female governors.45
(134) [Dono ken-no
jyosei chiji]
-mo Tokyo-de-no
kaigi-ni
syussekisita.
which prefecture-Gen female governor –MO Tokyo-in-Gen meeting-Dat attended 'Every female governor of some prefecture or other attended the meeting in Tokyo.'
It seems that this sentence allows another reading that comes with a certain presupposition. If you hear (134) out of the blue, you may get the interpretation in (135),
45The
sentence in (i) clearly says something distinct from "there is a female governor of Okayama who is very competent". If there is no female governor of Okayama, (i) is felt to lack a value, even though jyosei chiji 'female governor' does not have any marking of definiteness like English the. (i) [Okayama-ken-no jyosei chiji]-wa hijyooni Okayama-prefecture-Gen female governor- WA very 'The female governor of Okayama is very competent.'
84
yuunoo da. competent is
where there is a presupposition that there is a female governor of a given prefecture x. Notice that the translation in (135) fits with the embedded restrictor view.
(135) Åx[prefecture(x) ∞ attend(ιz[meeting_in_Tokyo(z)])(ιy[female_governor(x)(y)])]
If you are someone who does not know anything about Japan, you accommodate into the common ground that all prefectures have a female governor. If you are someone who knows that most of the prefectures in Japan have male governors, the sentence is felt to involve presupposition failure. This is not expected on the direct restrictor view proposed here. The question of why there seems to be a correlation between the size of the restriction and this kind of reading should be investigated in future studies.
2.7 Remaining Questions: Kadooka 'whether'-Questions 2.7.1 Kadooka 'whether'-Island Effect The examples that I have focused on in this chapter in demonstrating a wh-island effect in Japanese involve islands created by content questions, as in (136b,c).
(136) Taro-wa [Yamada-ga
dare-ni nani-o
okutta ka] tazunemasita ka?
Taro-Top Yamada-Nom who-Dat what-Acc sent a.
Q asked
'Did Taro ask what Yamada sent to whom?'
b. ?* 'Whox did Taro ask what Yamada sent to x?' c. *
'Whatx did Taro ask to whom Yamada sent x?'
d. ?* 'Whox did Taro ask whether Yamada sent what to x?'
85
Q
But as is well-known, sentences which contain embedded yes-no questions also create a similar island effect, as shown in (136d), in which the lower ka is interpreted as 'whether', as well as in (137).
(137) ?* Taro-wa [Yamada-ga
dare-ni wairo-o
okutta kadooka]
Taro-Top Yamada-Nom who-Dat bribe-Acc sent
whether
tazunemasita ka? asked
Q
'Whox did Taro ask whether Yamada sent a bribe to x?'
In a movement account, the ungrammaticality of reading (136d) and (137) is viewed as stemming from the same source as the ungrammaticality of readings (136b) and (136c): The specifier position of CP is occupied by some wh-element, namely, kadooka 'whehter' or a part of it (see Watanabe 1992a, for example), thereby creating an island for movement of another wh-element. In the non-wh-movement analysis proposed in this chapter, it is not clear at this point whether the account for (136b,c) extends to (136d) and (137) straightforwardly. More specifically, in the proposed analysis, we expect interveners to be unselective binders. If it turns out that the general syntactic and semantic analysis of kadookaquestions involves an unselective binder, the account for (136b,c) may extend to (136d) and (137). If not, it suggests that the ungrammaticality of (136b,c) and (136d)/(137) may have different sources, contrary to standard assumptions. In what follows, I present some questions that need to be addressed in looking for a general analysis of kadookaquestions.
86
2.7.2 Ka, Doo, and Ka Even though kadooka is often translated as 'whether', the examples in (138) show that kadooka is not to be equated with whether in English. In examples (138b) through (138d), in which the alternative clauses are spelled out, the second of the alternative clauses takes place of doo of kadooka in (138a). In their respective English translations, on the other hand, whether remains intact.
(138) a. Satoshi-wa [Yoko-ga
kuruma-o katta
kadooka]
tazuneta.
Satoshi-Top Yoko-Nom car-Acc bought KADOOKA asked 'Satoshi asked whether Yoko bought a car.' b. Satoshi-wa [Yoko-ga
kuruma-o katta
Satoshi-Top Yoko-Nom car-Acc
ka soo-de-nai ka] tazuneta.
bought KA so-is-not
KA
asked
'Satoshi asked whether Yoko bought a car or it is not so.' c. Satoshi-wa [Yoko-ga
kuruma-o katta
Satoshi-Top Yoko-Nom car-Acc
ka kawanakatta ka] tazuneta.
bought KA didn’t_buy KA asked
'Satoshi asked whether Yoko bought or didn’t buy a car.' d. Satoshi-wa [Yoko-ga
kuruma-o katta
Satoshi-Top Yoko-Nom car-Acc
ka Takenori-ga
kuruma-o katta
bought KA Tekenori-Nom car-Acc bought
ka] tazuneta. KA
asked
'Satoshi asked whether Yoko bought a car or Takenori bought a car.'
Thus, kadooka should be decomposed into three pieces, ka, doo, and ka, which is assumed in Kuroda (1965) as well.
87
It has been suggested that (at least) the first ka in kadooka is in fact an occurrence of the disjunction marker ka, as used in nominal and clausal disjunctions in (139) (see Kuroda 1965, Tonoike 1992, Whitman 1998 and Hagstrom 1998, among others).46
(139) a. Yukie-ka Shizuko-ga Yukie-or Shizuko -Nom
denwasita. called
'Yukie or Shizuko called.' b. Yukie-ga Yukie-Nom
denwasita ka Shizuko-ga called
tegami-o kaita.
or Shizuko-Nom letter-Acc wrote
'Yukie called or Shizuko wrote a letter.'
A piece of support for the idea that kadooka-questions always involve overt clausal disjunction comes from the following contrast in kadooka-questions in Japanese and whether-questions in English. The whether-clause in (140a) has the two readings in (141). (141a) is the yes-no question reading, while (141b) is the alternative question
46Ka
can optionally follow the second disjunct, as in (i). This might suggest that the second occurrence of ka in each example in (138) is also an instance of disjunction marker ka. (i) a. Yukie-ka Shizuko-(ka)-ga denwasita. Yukie-or Shizuko-or -Nom called 'Yukie or Shizuko called.' b. Yukie-ga denwasita ka Shizuko-ga tegami-o kaita (ka da). Yukie-Nom called or Shizuko-Nom letter-Acc wrote or is '(It is that) Yukie called or Shizuko wrote a letter.
The second occurrence of ka in (138), however, does seem to be different from the disjunction marker ka and seem to mark interrogativity. First, the second occurrence of ka in interrogative clauses is obligatory, unlike in (i). Second, sentence (ii) only allows the question reading of the embedded clause. (ii) Taro-wa [Yukie-ga denwasita ka Shizuko-ga tegami-o kaita ka] tutaeta. Taro-Top Yukie-Nom called or Shizuko-Nom letter-Acc wrote KA reported a. 'Taro reported whether Yukie called or Shizuko wrote a letter.' b. * 'Taro reported that either Yukie called or Shizuko wrote a letter.'
88
reading. If we add the overt or not to the sentence as in (140b), the alternative question reading in (141b) disappears (Larson 1985).
(140) a. Yuriko asked whether Hitoshi goes to Sapporo or Okinawa. b. Yuriko asked whether Hitoshi goes to Sapporo or Okinawa or not. (141) a. {p: p = Hitoshi goes to Sapporo or Okinawa ⁄ p = ¬Hitoshi goes to Sapporo or Okinawa} b. {p: p = Hitoshi goes to Sapporo ⁄ p = Hitoshi goes to Okinawa}
The example in (142) is the Japanese counterpart of (140).
(142) Yuriko-ga
[Hitoshi-ga
Sapporo-ka Okinawa-ni iku ka doo ka] tazuneta.
Yuriko-Nom Hitoshi-Nom Sapporo-or Okinawa-to go KA DOO KA asked 'Yuriko asked whether Hitoshi goes to Sapporo or Okinawa or not.'
Example (142) only allows the reading in (141a), but not (141b). This serves as support for the idea that kadooka-questions always involve overt clausal disjunction: with respect to possible readings, the Japanese sentence in (142) behaves like the English sentence in (140b), which involves overt clausal disjunction, rather than like (140a). Thus, one possible analysis of the examples in (138) is that they all involve the structure in (143).
(143)
CP 4 egi ka IP1 ka IP2 Q or
89
(138a) is a special case of (143), and its LF is as in (144). Doo of kadooka in (144) should be interpreted as ¬ªIP1 º.
(144)
CP 4 egi ka IP1 ka IP2 Q or doo 'how' Even though this might look like a plausible structure for a kadooka-clause, it is
not a suitable LF for its desired interpretation. Let us assume for now that the second ka in (143) is the same question marker as in the content questions that we focused on in section 2.3, and that it forces the introduction of the set forming Op, which gets coindexed with free variables introduced by indeterminate phrases. For the embedded interrogative clause in (138d), repeated here as (145), we would obtain the unsuitable singleton set in (146a) as its denotation, rather than the desired denotation in (146b).
(145) Satoshi-wa [Yoko-ga
kuruma-o katta
Satoshi-Top Yoko-Nom car-Acc
ka Takenori-ga
kuruma-o katta
bought or Tekenori-Nom car-Acc bought
ka] tazuneta. KA asked
'Satoshi asked whether Yoko bought a car or Takenori bought a car.' (146) a. {p: p = Yoko bought a car or Takenori bought a car} b. {p: p = Yoko bought a car ⁄ p = Takenori bought a car}
One way of obtaining the proper denotation in (146b) is to assume that the clausal disjunction in (145) introduces a free propositional variable (see Rooth and Partee 1982, Larson 1985; see also von Stechow 1993), and that it is abstracted over by Op. However, once we allow the set forming Op to be freely co-indexed with variables introduced by
90
non-indeterminate phrases, many consequences follow, which I am not ready to discuss in detail here. Though I have presented the idea that the first ka in kadooka is a disjunction marker, a possibility that this ka is a question marker cannot be excluded. The example in (147) shows that soretomo, which means something like or, can optionally occur between the two clauses.47
(147) [Yuki-ga snow-Nom
hutta
ka (soretomo)
fell
KA
or
ame-ga
hutta
rain-Nom fell
ka] siritai. KA
want_to_know
'I wonder whether it snowed or it rained.'
This reminds us of English examples like (148), in which two whether/if clauses are connected with or. (148b) is taken from Whitman (1998, p. 126).
(148) a. I wonder whether Yoko bought a car or whether Takenori bought a stereo. b. I don’t know if Eri bought that hat or if she didn’t (buy that hat).
Thus it will be worth exploring the structure in (149), as an alternative to the structure in (143).
47Soretomo
'or' seems to have a more limited distribution than ka 'or', as shown in (i).
(i) a.
Ayako-ka Elisabeth(-ka)-ga Ithaca-ni Ayako-or Elisabeth(-or)-Nom Ithaca-to 'Ayako or Elisabeth came to Ithaca.' b. ?? Ayako-ka soretomo Elisabeth(-ka)-ga Ayako-or or Elisabeth(-or)-Nom 'Ayako or Elisabeth came to Ithaca.'
91
kita. came Ithaca-ni kita. Ithaca-to came
(149)
CP egi CP (soretomo) CP 2 2 IP ka IP ka Q Q
Needless to say, the details of the syntax-semantics mapping of the kadooka-questions remain to be worked out. This is for future work.
2.8 Summary In this chapter I proposed a new way of looking at the semantics of the moconstruction, namely, the direct restrictor view, as opposed to the standard embedded restrictor view. It was shown that one way of spelling out this idea is to assume that indeterminate phrases are like Heimian indefinites (Heim 1982, Kamp 1981; Kuroda 1965, Nishigauchi 1990), in that they introduce free variables that will be bound by a higher operator, in this case, a set-forming Op in particular. The new view of the semantics of the mo-construction made possible the straightforward analysis of the syntax-semantics mapping, as well as the unified analysis of the ka- and mo-constructions in which indeterminate phrases do not undergo LF whmovement. This in turn allowed a switch of perspective on the long-standing island puzzle, as discussed in section 2.4. In particular, the lack of Complex NP and Adjunct island effects is no longer a puzzle, since indeterminate phrases do not undergo LF whmovement. The wh-island effect, on the other hand, follows from the way in which the indeterminate phrases are interpreted. Thus the island puzzle in fact serves as empirical support for the non-wh-movement analysis. There are recent claims that wh-in-situ does not undergo covert movement for theoretical reasons (Chomsky 1995, Reinhart 1998; see also Ochi 1998). This chapter reached a similar conclusion for Japanese, but on empirical, new grounds.
92
CHAPTER 3 INTERNALLY HEADED RELATIVE CLAUSES IN JAPANESE AND E-TYPE ANAPHORA1
3.1 Introduction In the familiar type of relative clause construction in Japanese, the head noun follows a relative clause, which contains a gap (ø) that corresponds to the head. This is shown in example (1).2
(1) Yoko-wa [[Taro-ga
sara-no
ue-ni ø oita] keeki]-o tabeta.
Yoko-Top Taro-Nom plate-Gen on-Loc put
cake -Acc ate
'Yoko ate a piece of cake which Taro put on a plate.'
Unlike English, for example, no overt relative pronoun is used, and the relative clause in (1) can be either restrictive or non-restrictive. Since the head occurs "external" to the relative clause, I refer to this type as the "externally headed relative clause construction (EHR)". The sentence in (2) is an example of what is often called the "internally headed relative clause construction (IHR)".
(2) Yoko-wa [[Taro-ga
sara-no
ue-ni keeki-o
oita] -no] -o
Yoko-Top Taro-Nom plate-Gen on-Loc cake-Acc put
1This
tabeta.
-NM -Acc ate
chapter is based on Shimoyama (1999).
2In
keeki 'cake' in (1), the singular/plural as well as definite/indefinite distinctions are unspecified, and it could mean 'a/the cake' or '(the) cakes'. This ambiguity will not be reflected in English translations throughout the chapter.
93
In (2), keeki 'cake' occupies the position that a gap occupies in (1). The italicized embedded clause in (2) (which I call an "IHR-clause") is followed by the morpheme -no, whose exact syntactic and semantic status is controversial. 3 For now, I simply gloss it as a 'nominalizer (NM)', since it is this morpheme that seems to be responsible for turning the sentence embedded under it into a constituent that has a nominal-like external distribution. This constituent will be referred to as an "IHR-DP". In (2), the IHR-DP as a whole receives the accusative marker -o, and seems to function as the object of the matrix verb tabe 'eat'. This verb, however, cannot take a propositional complement. Rather, keeki 'cake' is understood to be the "semantic head" of the IHR-clause, which serves as the object of the matrix verb. The constituent that serves as a semantic head of an IHRclause will be referred to as an "internal head", and will be put in boldface. This chapter addresses the question of how the internal head is understood to play the dual role of being an argument of the embedded verb and an argument of the matrix verb. There are two major approaches to this question in the literature. The first assumes that the internal head is raised to the external head position at the level of LF. The second assumes no head raising, but posits a phonologically null anaphoric element in the matrix argument position. The first goal of the chapter is to present evidence for representations in which internal heads remain internal to IHR-clauses at LF, and to provide further arguments for the idea (first presented in Hoshi 1995) that the interpretation of IHRs involves E-type anaphora. The second goal is to embed the E-type analysis of IHRs within a general theory of E-type anaphora that overcomes problems that arise with Hoshi's (1995) particular implementation. The chapter is organized as follows. Section 3.2 provides evidence for a head internal LF representation of IHRs. The evidence comes from the behavior of 3See,
for example, Kuroda (1976-77), Kitagawa and Ross (1982), and Murasugi (1991) for discussion.
94
quantificational NPs that occur in IHRs. Section 3.3 demonstrates that the relation between an IHR-clause and its matrix clause is mediated by E-type anaphora. I discuss new data that involve IHRs in the wh-question context. In section 3.4, an LF representation that is suitable for deriving the appropriate semantics of the IHR compositionally is proposed. A consequence of the proposal is discussed in section 3.5. Section 3.6 examines a property of IHRs that is not shared by E-type pronouns in English. Section 3.7 concludes the chapter.
3.2 Does the Internal Head Move at LF or Stay In-Situ? 3.2.1 The LF Head Raising Analysis One approach to the IHR in Japanese as well as in other languages in the literature assumes that the internal head undergoes raising at the level of LF, and as a result the LF structure for the IHR is reduced to that of the EHR. I call this approach "the LF head raising analysis". The LF structure of the IHR in this type of analysis looks like (3).
(3)
NP
2
CP
2
NPi
C'
2
IP
! ... ti ...
C
This kind of LF representation for the IHR is proposed by Ito (1986) for Japanese,4 Broadwell (1986) for Choctaw, Cole (1987) for Ancash and Imbabura Quechua, Lefebvre and Muysken (1988) for Quechua, among others. 5,6 4Ito
(1986) in fact assumes that the internal head raises at a level later than LF, which she calls LF'. Thus at her level of LF, the structure of the IHR is still distinct from the structure of its external counterpart. She observes that the IHR and the EHR exhibit different properties with respect to certain LF phenomena (e.g., Weak Crossover), and concludes that raising of the head applies after LF. The main motivation for the raising is that it ensures that the matrix verb assigns the appropriate thematic role to the head noun. 95
A crucial point here is that in this analysis, the LF representation of the IHR is basically identical to the S-structure and LF representations of the EHR.7 In other words, with the internal head being raised, the IHR is assimilated to its external counterpart. An apparent advantage of this kind of uniform analysis is that a unique interpretational mechanism can take care of both IHR and EHR. This is advantageous, of course, to the extent that it is empirically accurate. We will see below that this is not the case.
3.2.2 The Truth-Conditional Difference between IHRs and EHRs The IHR-clause in (4) contains a quantificational NP as its internal head.
(4)
Taro-wa [[Yoko-ga
reezooko-ni
kukkii-o hotondo
Taro-Top Yoko-Nom refrigerator-Loc cookie-Acc most irete-oita]-no]-o
paatii-ni motte itta.
put-Aux -NM -Acc party-to brought 'Yoko put most cookies in the refrigerator and Taro brought them to the party.'
While her observations are compatible with the point I will make in this section (i.e., the internal head is located internal to the clause at LF), my general assumption is that semantic interpretation (including thematic properties) is read off from a single level of representation. 5This
type of movement was first proposed by Brame (1968) and Vergnaud (1974) for externally headed relative clauses in English and French. See also Kayne (1994).
6There
is a slightly different subtype in the LF head raising analysis. It is assumed by, for example, Williamson (1987, Lakhota), Barss et al. (1991, Navajo) Watanabe (1992a, b, Japanese) and Bonneau (1992), that the internal head undergoes movement and lands in the specifier position of CP, rather than in the external head position.
7Since
the S-structure representation of the EHR is suitable for compositional derivation of the meaning, I assume that its LF representation is essentially the same as its Sstructure representation (see Quine 1960).
96
If the internal head raises at LF, the LF representation of (4) should look like the Sstructure and LF representations of its external counterpart in (5).
(5)
Taro-wa [[Yoko-ga
reezooko-ni
ø irete-oita] kukkii-o
Taro-Top Yoko-Nom refrigerator-Loc
put-Aux cookie -Acc
hotondo] paatii-ni motte itta. most
party-to brought
'Taro brought most cookies that Yoko had put in the refrigerator to the party.'
The two sentences, however, do not share truth-conditions, as informally indicated in the translations. First, the translation for (5) shows that the relative clause 'Yoko put x in the refrigerator' constitutes the restriction for the domain of 'most' along with 'cookies', whereas the translation for (4) shows that the domain of 'most' is restricted only by 'cookies', but not by the rest of the IHR-clause 'Yoko put x in the refrigerator'. Second, the translation for (5) shows that the matrix clause 'Taro brought x to the party' is the nuclear scope of 'most', whereas the translation for (4) shows that 'Taro brought x to the party' is not the nuclear scope of 'most'. Rather, (4) implies that Taro brought all the cookies Yoko put in the refrigerator to the party.8 Suppose that Yoko put 10 cookies in the refrigerator and Taro brought 7 of them to the party. In this situation, (5) is true while (4) is false. This kind of contrast in IHR and EHR is also noted by Hoshi (1995) with respect to similar examples.9 8This
so-called "maximality" effect (Sells 1986) can be cancelled by the use of a floating numeral, as shown in (i), where we obtain a partitive reading. (i) Taro-wa [[Yoko-ga reezooko-ni kukkii-o hotondo irete-oita]-no] -o Taro-Top Yoko-Nom refrigerator-Loc cookie-Acc most put-Aux -NM-Acc 3-tu dake paatii-ni motte itta. 3-CL only party-to brought 'Yoko put most cookies in the refrigerator and Taro brought only three of them to the party.'
9Srivastav
(1991, p. 683) points out a similar fact in the following internally headed relative in Ancash Quechua: 97
The LF head raising analysis has no account of the truth-conditional difference between (4) and (5). In this analysis, (4) is reduced to (5) at LF, and it is predicted that the IHR-DP in (4) is interpreted as 'most cookies that Yoko put in the refrigerator'. The restriction difference and the scope difference between (4) and (5) suggest that in (4), the internal head remains inside the IHR-clause (with the assumption that the scope of quantifiers is clause-bounded) and the IHR-clause is interpreted as a simple sentence. This was reflected in the informal translation I gave to (4), and in fact, the translation anticipates the LF and semantic representations of IHRs that I propose in section 3.4.
3.2.3 The Position of the Internal Head at LF and Scope Phenomena This section provides further arguments for head internal LF representations of IHRs, i.e., representations in which the internal head remains internal to the IHR-clause at LF. If the internal head is raised and located in the external head position at LF, it is predicted (i) that a quantificational internal head interacts with other scope bearing elements in the matrix clause, and (ii) that it takes the widest scope with respect to other scope bearing elements within the IHR-clause. It will be shown below that neither of the predictions is correct.
3.2.3.1 Background: The Scope of Quantifiers and Scrambling Let me briefly introduce the relevant background about the interpretation of the relative scope of quantifiers in Japanese. In general, the S-structure c-command relation determines the relative scope of quantifiers. I illustrate this with sentence (6a) and its (i) nuna ishkay bestya-ta ranti-shqa-n alli bestya-m ka-rqo-n. man two horse-Acc buy-Perf-3 good horse-Validator be-Past-3 'The two horses that the man bought were good horses.' She notes that "the relative clause [in (i)] contains the information that the total number of horses bought by the man is two."
98
variant (6b), in which the surface order of the subject and the object is reversed by scrambling.
(6) a. Hotondo-no gakusei-ga most-Gen
dono syukudai-mo teisyutusita.
student-Nom which homework-MO turned_in
'Most students turned in every homework.' (i) Most > ∀
(ii) *∀ > Most
b. Dono syukudai-moi
hotondo-no gakusei-ga ti teisyutusita.
which homework-MO most-Gen
student-Nom
turned_in
Lit. 'Every homework, most students turned in.' (i) *Most > ∀
(ii) ∀ > Most
Reading (i) (For most student x, for every homework y, x turned in y.) and Reading (ii) (For every homework y, for most student x, x turned in y.) can be described as follows. Suppose that there are four students, A, B, C and D, and three homework assignments, #1, #2 and #3, under consideration; consider the scenarios in (7a) and (7b). (For example, the first line of (7a) is read as: Student A turned in homework #1, #2 and #3.)
(7) a. A. B. C. D.
#1
#2 #3
b.
#1 #2 #3 A. B. C. D.
Reading (i), in which 'most' takes wide scope, is true under scenario (7a). Reading (ii), in which 'every' takes wide scope, also comes out as true under (7a) because reading (i) is a
99
special case of reading (ii), so whenever reading (i) is true, reading (ii) is also true.10 Reading (i) Most > ∀ is false but reading (ii) ∀ > Most is true under scenario (7b). Sentence (6a) only has the reading that is true in (7a) and false in (7b), namely, reading (i), which is represented by the surface order. Sentence (6b), on the other hand, only has the reading that is true both in (7a) and in (7b). That is reading (ii), which is, again, represented by the surface order.11 This state of affairs also holds, naturally, for quantificational NPs which have a relative clause as an additional restriction. The only difference between the sentences in (6) and those in (8) is that in the latter, the direct object contains a relative clause ('that Taro assigned before the exam'). In (8a), as in (6a), 'every homework' takes narrow scope, but not wide scope, with respect to 'most students'. In (8b), in which the whole direct object is scrambled and precedes the subject, the wide scope reading of the universal quantifier becomes available as was the case with (6b).
(8) a. Hotondo-no gakusei-ga most-Gen
[[Taro-ga
ø sikenmae-ni
student-Nom Taro-Nom
dasita]
before_exam-at assigned
dono syukudai-mo] teisyutusita. which homework-MO turned_in 'Most students turned in every homework that Taro assigned before the exam.' (i) Most > ∀
(ii) *∀ > Most
10Thus,
more needs to be said if we want to show that these are independent readings. For ease of reference, I call them readings here. 11With
different choice of quantifiers and their relative orders, the reading represented by the basic word order may remain available after scrambling, unlike (6b). See, for example, Kuroda (1970) and Hoji (1985), among others.
100
b. [[Taro-ga ø sikenmae-ni Taro-Nom
dono syukudai-mo]i
before_exam-at assigned which homework-MO
Hotondo-no gakusei-ga most-Gen
dasita]
student-Nom
ti teisyutusita. turned_in
Lit. 'Every homework that Taro assigned before the exam, most students turned in.' (i) *Most > ∀
(ii) ∀ > Most
With this much in mind, let us move on to IHRs.
3.2.3.2 The Relative Position of the Internal Head with Respect to Material in the Matrix Clause We have already seen an example in section 3.2.2 that illustrated that the matrix clause was not in the scope of a quantificational internal head. The sentences in (9), which are the internal versions of the sentences in (8), confirm this descriptive generalization. Our focus is on how the internal head dono syukudai-mo 'every homework' interacts with the matrix subject hotondo-no gakusei 'most students'.
(9) a. Hotondo-no gakusei-ga most-Gen sikenmae-ni
[[Taro-ga dono syukudai-mo
student-Nom Taro-Nom which homework-MO dasita]
-no] -o
teisyutusita.
before_exam-at assigned]-NM]-Acc turned_in 'Taro assigned every homework before the exam and most students turned them in.' (i) Most > ∀
(ii) *∀ > Most
101
b. [[Taro-ga dono syukudai-mo sikenmae-ni
dasita]
-no] -o i
Taro-Nom which homework-MO before_exam-at assigned]-NM]-Acc hotondo-no gakusei-ga most-Gen
student-Nom
ti teisyutusita. turned_in
'Taro assigned every homework before the exam and most students turned them in.' (i) Most > ∀
(ii) *∀ > Most
The only available reading of (9a) is the one in which 'most students' takes wider scope. The reading in which 'every homework' takes wider scope is not available. This pattern seems to be the same as the one we observed in (8a), the external version of (9a). When object scrambling is applied to (9a), resulting in (9b), the sentence exhibits a scope pattern that is not observed in its external counterpart (8b). In (9b), reading (ii), in which the universal quantifier takes wider scope (a set of students varying with each homework) is not available. Recall that in (8b), the external counterpart of (9b), this scope pattern is available. This contrast is rather surprising under the LF head raising analysis, which assumes that the internal head 'every homework' in (9b) occupies the same configurational position as 'every homework' in its external counterpart (8b) at LF. On the other hand, the unavailability of reading (ii) in (9b) falls out if we assume that the internal head 'every homework' remains inside the IHR-clause, and the IHR-clause is interpreted as a simple sentence as indicated by the English translation. It is expected that this quantificational NP does not take wider scope with respect to the matrix subject 'most students', whether scrambling is involved or not, under the assumption that the scope of quantifiers is clause-bounded.12 12Recall
that there is a subtype of the LF head raising analysis which assumes that the internal head occupies the specifier position of the embedded CP (see footnote 6). This analysis might be able to predict the scope pattern in (9b). However, it also predicts that 102
3.2.3.3 The Relative Position of the Internal Head with Respect to Other Internal Material Let us turn to the question of how a quantificational internal head interacts with another quantificational NP within the same clause. When sentence (6a) is embedded as an IHR-clause as in (10a), the scope pattern remains the same. When object scrambling is applied within the IHR-clause in (10a), we obtain (10b). The scope pattern we find in (10b) is again parallel to the one we saw in the simple sentence in (6b).
(10)
a. Taro-wa [[hotondo-no gakusei-ga
dono syukudai-mo
Taro-Top most-Gen student-Nom which homework- MO teisyutusita]-no] -o turned_in
yatto
saitensioeta.
-NM -Acc finally finished_grading
'Most students turned in every homework and Taro finally finished grading them.' (i) Most > ∀
(ii) *∀ > Most hotondo-no gakusei-ga ti
b. Taro-wa [[dono syukudai-moi
Taro-Top which homework- MO most-Gen student-Nom teisyutusita] -no]-o turned_in
yatto
saitensioeta.
-NM -Acc finally finished_grading
Lit. 'Every homework, most students turned in and Taro finally finished grading them.' (i) *Most > ∀
(ii) ∀ > Most
The fact that the IHR-clauses in (10a) and (10b) exhibit scope patterns identical to those found in the simple sentences in (6a) and (6b), respectively, indicates that there is the internal head, since it is in the spec of CP, always has the widest scope with respect to other material within that CP (unless adjunction to CP is assumed). The data in section 3.2.3.3 will show that this prediction is not borne out.
103
no LF head raising and that the IHR-clause is interpreted as a simple sentence. If the internal head dono syukudai-mo 'every homework' were located in the external head position at LF, both (10a) and (10b) should be reduced to their external counterpart (11) in their meaning and scope possibilities. However, this is not the case (as informally indicated in the translations).
(11)
Taro-wa [[hotondo-no gakusei-ga
ø teisyutusita] dono syukudai-mo]
Taro-Top most-Gen student-Nom turned_in yatto
which homework-MO
saitensioeta.
finally finished_grading 'Taro finally finished grading every homework that most students turned in' (i) *Most > ∀ (ii) ∀ > Most
This sentence can be truthfully uttered, for example, in the following situation: There are four students and three homework assignments, #1, #2, and #3. A (possibly different) group of three students turned in homework #1 and homework #2, but only one student turned in homework #3. Taro finished grading homework #1 and #2. As expected from the clause-boundedness of the quantifier scope, 'most' cannot take wide scope in this sentence.13, 14 13The
relative clause in (11) may also be interpreted as non-restrictive. The scenario provided under (11) is intended only for the restrictive interpretation. 14Hoshi
(1995, pp. 63-65) briefly mentions that the following kind of contrast is a problem for the LF head raising analysis. He states that "when the HERC [EHR, J.S.] is headed by a universally quantified expression, it yields two readings corresponding to the two relative scope orders," observing that sentence (i) has a reading in which dareka 'someone' takes wide scope and a reading in which it takes narrow scope. (i) John-wa [[dareka-ga ø muitekureta] dono ringo-mo] hitori-de John-Top someone-Nom peeled which apple-MO alone tabetesimatta rasii. ate up seem 'I heard that John ate up alone every apple that someone peeled for him.' He notes that sentence (ii) poses a problem for the LF head raising analysis because it 104
I have presented evidence for a head internal LF representation of IHRs, and against the LF head raising analysis. This leads me to conclude that the interpretation of the IHR construction must be mediated by some element in the matrix clause that is somehow anaphorically related to the internal head. The question of what kind of anaphoric relation is involved is the topic of the next section.
3.3 E-type Anaphora in the IHR 3.3.1 Bound Variable Anaphora and Referential Anaphora Hoshi (1995) proposes that an IHR-clause is externally headed by an empty argument [e], as in (12), which remains empty at LF.
predicts that the IHR in (ii) ends up having the same configuration as the EHR in (i) and that it should also be ambiguous, which is not the case. Sentence (ii) only has the reading in which 'someone' takes wider scope. (ii) John-wa [[dareka-ga dono ringo-mo muitekureta]-no]-o peeled ]-NM]-Acc John-Top [[someone-Nom which apple-MO hitori-de tabetesimatta rasii. alone ate up seem 'I heard that John ate up alone every apple that someone peeled for him.' (Hoshi's translation) It should be noted that the availability of the two readings in (i) is not a general property of relative clauses headed by a universal quantifier (see, for example, (11)), but is presumably due to a general property of an indefinite to show a wider scope interpretation than its syntactic scope, or due to the fact that the reading ∃ > ∀ is a special case of the other reading, ∀ > ∃. The important point here is that while the reading in which dono ringo-mo 'every apple' takes wider scope is available in (i), it is not in (ii), which is not explained by the LF head raising analysis. While Hoshi does not make a connection between these examples and his E-type proposal, the difference in scope properties between (i) and (ii) in fact follows from a general E-type analysis, as we will see.
105
(12)
John-wa [NP [CP [IP Mary-ga John-Top -no] -o
san-ko-no
ringo-o
muitekureta]
Mary-Nom three-CL-Gen apple-Acc peeled [e] ] tabeta.
-NM -Acc
ate
'Mary peeled three apples and John ate them all.'
(Hoshi 1995, p. 131)
As I noted earlier, he makes the same kind of observation as the one I made in section 3.2.2 concerning the scope of the quantificational internal head. He observes that the scope of the internal head san-ko-no ringo 'three apples' in (12) does not extend to the matrix clause and bind the empty argument [e]. If it did, the sentence would mean that three apples are such that Mary peeled them and John ate them, and would be compatible with a scenario where Mary peeled five apples and John ate three of them. This, however, is not the correct meaning of the sentence. Rather, it means that Mary peeled only three apples and John ate them all. Hoshi (1995) further observes that the property of the empty argument [e] noted above is shared by, for instance, the pronoun they in sentence (13) from Evans (1980, p. 339).
(13)
Few congressmen admire Kennedy, and they are very junior.
As Evans (1980) discussed, although the pronoun they in (13) seems to be somehow "anaphorically related" to the quantificational NP few congressmen, it is not interpreted as a bound variable pronoun, as we can tell from the fact that the sentence does not mean that few congressmen are such that they admire Kennedy and are very junior. What the sentence means is rather that few congressmen admire Kennedy and that all the congressmen who admire Kennedy are very junior.
106
What Hoshi's example in (12) (and all the examples of IHRs we have seen up to this point) as well as Evans' example in (13) show is that the relation between the anaphoric element ([e] in (12) or they in (13)) and the quantificational antecedent is neither bound variable anaphora nor coreferential anaphora. In (12), for instance, the matrix object [e] cannot corefer with its quantificational antecedent, which does not refer in the first place. The object [e], however, could still be considered to be referential, namely, it refers to the apples that Mary peeled, a plural individual referent made salient by the utterance of the IHR-clause. Thus, Hoshi's observation is compatible with assuming referential anaphora in the interpretation of IHRs, and it has not yet been shown that the third kind of anaphora that is neither bound variable anaphora nor referential anaphora is truly involved in the interpretation of IHRs.
3.3.2 The Need for E-type Anaphora The example in (14) crucially shows that it is not referential anaphora that is involved in the interpretation of the sentence. The closest paraphrase of (14) in English is given in (14').
(14)
Dono gakusei-moi [soitui -ga /proi kongakki
peepaa-o
3-bon
which student-MO (s)he-Nom/pro this semester term paper-Acc 3-CL kaita] -no] -o
kesa
teisyutusita.
wrote -NM -Acc this morning turned_in (14')
Every student wrote three term papers this semester and turned in the term papers he or she wrote this semester this morning.
The matrix object in (14) does not refer to any particular set of term papers. The interpretation we want is best paraphrased by the use of the definite description 'the term papers that he or she wrote this semester', which contains a variable 'he/she' bound by the
107
matrix subject 'every student'. This allows the semantic value of the matrix object to vary with each choice of a student. If referential anaphora were involved in (14), the reading paraphrased in (14') should not be possible. Thus, example (14) crucially demonstrates that we cannot assume a referential pronoun, a simple free variable that receives its value from contextual assignment, in the matrix object position. It is this kind of example that motivates the need for a third kind of anaphoric relation ("E-type anaphora") in the interpretation of IHRs that is neither bound variable anaphora nor referential anaphora.15, 16
3.3.3 Further Arguments for the E-type Analysis: The IHR and Wh-Questions In this section I provide further evidence that it is E-type anaphora that is involved in the interpretation of IHRs, focusing on IHR-clauses that contain wh-phrases. 17 I also show how the interpretation of an IHR with a wh-phrase contrasts with the interpretation of its external counterpart. The observed contrast serves as further support for the earlier conclusion that the internal head remains internal to the IHR-clause, hence the IHR does not have the same LF representation as either a restrictive EHR or a non-restrictive EHR.
15For
a general introduction to issues related to E-type anaphora, readers are referred to Heim and Kratzer (1998, Ch. 11). 16Despite
his use of the term "E-type pronoun" (referring to his empty argument [e]), Hoshi (1995) seems to assume that [e] is a referring expression. This assumption cannot be maintained (i) for the reason explained in the text, and (ii) in light of data I will observe in sections 3.3.3.1. and 3.5.1. below. 17I
will use the standard term "wh-phrases" rather than "indeterminate phrases" in this chapter, since our concern will be only on indeterminate phrases that receive interrogative interpretation.
108
3.3.3.1 The E-type Interpretation of the Anaphoric Element In (15a), the head of the EHR-clause is a wh-phrase dono neko 'which cat', while the same wh-phrase appears internal to the IHR-clause in (15b). 18
(15)
a. Taro-wa [[[Yoko-ga
ø turete kita]
dono neko]-ga
Taro-Top Yoko-Nom brought_along which cat -Nom nigedasita ka] siritagatte iru. ran_away Q want_to_know 'Taro wonders which cat that Yoko brought along ran away.' b. Taro-wa [[[Yoko-ga
dono neko-o
turete kita]
-no]-ga
Taro-Top Yoko-Nom which cat-Acc brought_along -NM -Nom nigedasita ka] siritagatte iru. ran_away Q want_to_know Lit. 'Taro wonders which cat Yoko brought along and that the cat that Yoko brought along ran away.'
As we saw in chapter 2, it is possible in Japanese for a wh-phrase to occur inside a normal relative clause. This is shown in (16a). (16b) is the internal version of (16a).
(16)
a. Taro-wa [[[dare-ga ø turete kita] Taro-Top
neko]-ga
who-Nom brought_along cat
nigedasita ka ]
-Nom ran_away Q
sitte iru. know Lit. 'Taro knows a cat that who brought along ran away.'
18Recall
that the noun neko 'cat' is unspecified with respect to number and definiteness. For simplicity, I treat it uniformly as a singular indefinite noun in the text. The main point would not be affected with another choice.
109
b. Taro-wa [[[dare-ga
neko-o turete kita]
-no] -ga
Taro-Top who-Nom cat-Acc brought_along -NM -Nom nigedasita ka] sitte iru. ran_away Q know Lit. 'Taro knows whox brought along a cat and that the cat that x brought along ran away.'
The translations for (15b) and (16b) informally indicate that the IHR-clauses are interpreted as simple sentences, rather than as relative clauses, as argued in section 3.2. Sentence (16b) crucially shows that what is involved is E-type anaphora. As indicated informally in the translation, the interpretation of the subject of the verb nigedasita 'ran away' is paraphrased by the definite description 'the cat that x brought along'. It contains a variable x that ranges over people. This shows that the anaphoric relation between the internal head neko 'a cat' and the subject of the verb nigedasita 'ran away' is that of E-type anaphora.
3.3.3.2 A Contrast in EHR-IHR Pairs that Contain Wh-phrases My informants share the intuition that in EHR-IHR pairs of the kind exemplified in (15) and (16), there is some subtle meaning difference between EHR and IHR. I am going to sharpen this intuition and show what the difference is and how it fits with the Etype analysis. Before examining an EHR-IHR pair that contains wh-phrases, I introduce a phenomenon called "(weak) exhaustivity" in the literature, which shows up in certain constructions with embedded interrogatives, as in (17).19
19See,
for example, Karttunen (1977), Groenendijk and Stokhof (1982), Bäuerle and Zimmermann (1991), Berman (1991) and Heim (1994).
110
(17)
Taro-wa [dare-ga
neko-o turete kita
ka] sitte iru.
Taro-Top [who-Nom cat-Acc brought_along Q] know 'Taro knows who brought along a cat.'
This sentence implies that, of every person who brought along a cat, Taro knows that he/she brought along a cat. Suppose that Yoko brought along her cat to a party and that Isadora also brought along her cat to a party and nobody else did. In this situation, this sentence implies that Taro knows two propositions: that Yoko brought along a cat and that Isadora brought along a cat. If Taro knows only that Yoko brought along a cat and he does not know that Isadora brought along a cat, we would not want to say that Taro knows who brought along a cat. This property is called (weak) exhaustivity.20 It licenses the inferences of the following form:
(18)
Taro knows who brought along a cat. x brought along a cat.
⇒ Taro knows that x brought along a cat.
We are now ready to examine the meaning contrast in EHR-IHR pairs. The EHRIHR pair in (16) is repeated below.
(16)
a. Taro-wa [[[dare-ga ø turete kita] Taro-Top
neko]-ga
who-Nom brought_along cat
nigedasita ka ]
-Nom ran_away Q
sitte iru. know Lit. 'Taro knows a cat that who brought along ran away.' 20Since
what is known as "strong exhaustivity" is not directly relevant to my concerns, I use the term "exhaustivity" in the sense of "weak exhaustivity" in this chapter.
111
b. Taro-wa [[[dare-ga
neko-o turete kita]
-no] -ga
Taro-Top who-Nom cat-Acc brought_along -NM -Nom nigedasita ka] sitte iru. ran_away Q know Lit. 'Taro knows whox brought along a cat and that the cat that x brought along ran away.'
For sentence (16a) to be true, Taro has to know, in accordance with exhaustivity, of every person x such that a cat that x brought along ran away, that a cat that x brought along ran away. In other words, Taro has to know all the true propositions of the following form.
(19)
that a cat that x brought along ran away
Let us now assume the scenario in (20).
(20)
Carol brought along Lucky, Yoko brought along Nyaaon, and Isadora brought along Spencer. Lucky and Nyaaon ran away.
For sentence (16a) to be true under scenario (20), Taro has to know the set of propositions in (21).
(21)
{that a cat that Carol brought along ran away, that a cat that Yoko brought along ran away}
Let us turn to (16b), the internal version of (16a). Most of my informants agree that there is a contrast between (16a) and (16b) in that (16b) does not sound perfectly
112
felicitous under the scenario in (20). When can we use sentence (16b) then? It is compatible with, for example, the following types of scenarios.
(22)
a. Carol brought along Lucky. Lucky ran away. b. Carol brought along Lucky and Yoko brought along Nyaaon. Lucky and Nyaaon ran away. c. Carol brought along Lucky, Yoko brought along Nyaaon, and Isadora brought along Spencer. Lucky, Nyaaon and Spencer ran away.
The incompatibility of sentence (16b) with scenario (20) and its compatibility with those scenarios in (22) fit with the E-type analysis, in which the IHR-clause is interpreted as a simple sentence and the relation between the IHR-clause and the clause embedding it is mediated by E-type anaphora. In this analysis, for (16b) to be true, Taro has to know two sets of true propositions of the form:
(23)
[that x brought along a cat] and [that the cat that x brought along ran away]
The first proposition in (23) corresponds to the IHR-clause in (16b). The second proposition corresponds to the matrix clause in (16b). The anaphoric element (the subject) of this clause is expressed by the definite description 'the cat that x brought along'. Focusing on the first proposition in (23), we see that for sentence (16b) to be true, Taro has to know, in accordance with exhaustivity, of every person x that brought along a cat, that x brought along a cat. In other words, he has to know the complete answer to the
113
question "Who brought along a cat?". For instance, in both scenarios (20) and (22c), this means that Taro knows the following set of propositions:
(24)
{that Carol brought along a cat, that Yoko brought along a cat, that Isadora brought along a cat}
Since x ranges over Carol, Yoko and Isadora in the first proposition in (23), the same should hold for the occurrence of x in the second proposition. This means that sentence (16b) also implies that Taro knows the following set of propositions:
(25)
{that the cat that Carol brought along ran away, that the cat that Yoko brought along ran away that the cat that Isadora brought along ran away}
In other words, (16b) presupposes that all the people that can truthfully be named in an answer to the question "Who brought along a cat?" let his/her cat run away. This can be compatible only with scenario (22c), and causes incompatibility with scenario (20). The sentence is true under scenario (22c) if Taro actually knows these propositions.21 Recall
21In order to capture the intuition that there is a contrast between (16a) and (16b), it is essential to assume, as I do in the text, that for (16b) to be true, Taro has to know the two sets of true propositions of the form in (23). This should be distinguished from the kind of interpretation available in sentences that involve Across-the-Board (ATB) movement as in (ia). For (ia) to be true, Taro needs to know the set of true propositions of the form in (ib).
(i) a. Taro knows [whox Karyn likes tx and Eric dislikes tx ]. b. that Karyn likes x and Eric dislikes x If (16b) were to be analyzed analogously to the hypothetical English sentence in (iia), the obtained ATB-style interpretation would be indistinguishable from the interpretation of its external counterpart (16a) (with the definite interpretation of the head noun).
114
that the truth conditions for sentence (16a), the external counterpart of (16b), do not require Taro to know the complete answer to the question "Who brought along a cat?". This explains the contrast between (16a) and (16b). 22, 23 I have shown how the interpretation of an IHR that contains a wh-phrase contrasts with the interpretation of its external counterpart. The contrast fits with the E-type analysis, whereas it is unexpected if it is assumed that an IHR is reduced to its external counterpart at LF. The earlier observation that the content of the IHR-clause does not restrict the domain of the quantificational internal head (section 3.2.2) may have suggested a possibility of assimilating IHRs to non-restrictive EHRs.24 The grammaticality of a sentence like (16b), however, shows that IHRs cannot be literally equated with non(ii) a. Taro knows [whox tx brought along a cat and the cat that t x brought along ran away.] b. that x brought along a cat and the cat that x brought along ran away Note also that the two occurrences of the variable x in (ib) are both due to overt movement, while in (23), neither of the occurrences of the variable x is created by overt movement. 22Some
of the speakers who find that (16b) does not sound felicitous in scenario (20) still feel that it is not impossible to somehow force the use of (16b) in that scenario. This is in a way expected in my analysis, because, as is well-known, the effect of exhaustivity is not always observed, depending on the choice of matrix verbs or pragmatic content of the sentence. (For instance, in Taro told us where we can buy miso.) Thus it is possible that the subtlety of the judgment has something to do with this general nature of exhaustivity. More specifically, if one can get a non-exhaustive reading in sentence (17), it is not surprising if he/she feels that (16b) can be uttered felicitously in scenario (20). I leave further examination for future studies. 23One
can also find a scenario in which (15a) can be uttered felicitously but (15b) cannot be. For instance, Yoko brought along Lucky, Nyaaon and Spencer; Lucky and Nyaaon ran away. This contrast also supports the E-type analysis, because the infelicitousness of (15b) in this scenario fits with the "maximality" associated with the definite description 'the cat(s) that Yoko brought along'. I use the EHR-IHR pair (16a) and (16b) rather than (15a) and (15b) in the discussion in the text because the effect of exhaustivity of the kind observed in (16b) is not as directly observable in (15b), due to maximality. 24Kuroda
(1975-76; 1992, p. 174) notes: "I believe it is appropriate to assume that p.-i. relative clauses [= IHR-clauses, J.S.] are 'nonrestrictive' in a certain semantic sense, ..." Kitagawa (1996) in fact proposes that the IHR is a non-restrictive relative clause with a phonologically null pronominal head (pro).
115
restrictive EHRs. (26a) is an example of a non-restrictive EHR. (26b) shows that a nonrestrictive EHR cannot contain a wh-phrase, while a restrictive EHR can, as we saw in example (16a) above.
(26)
a. [[Becky-ga
ø sukina] Lucky]-ga
Becky-Nom
like
nigedasita.
Lucky -Nom ran_away
'Lucky, which Becky likes, ran away.' b. *Taro-wa [[[dare-ga Taro-Top
ø sukina] Lucky]-ga
who-Nom
like
nigedasita ka]
Lucky -Nom ran_away Q
sitteiru. know Lit. 'Taro knows Lucky, which who likes, ran away.'
A sentence like (16b), which contains an IHR-clause with a wh-phrase in it, cannot find a corresponding non-restrictive EHR to be assimilated to.
3.4 The LF and Compositional Interpretation of IHRs I have presented new arguments for (i) a head internal LF representation of IHRs and (ii) the E-type interpretation. In this section I spell out this general idea in full detail. Section 3.4.1 is a brief review of Hoshi's (1995) proposal, where problems associated with his particular implementation are discussed. Section 3.4.2 provides a revised analysis that overcomes the shortcomings of Hoshi (1995).
3.4.1 Hoshi's (1995) Proposal Hoshi (1995) proposes that the IHR-DP has the structure in (27), headed by an empty argument [e]. Modifiying Cooper's (1979) analysis of E-type pronouns in English, Hoshi proposes the lexical meaning of the empty argument [e] in (28).
116
(27)
(28)
NP 2 CP N 2 [e] IP C ! no ... NP... NM λK∃y[∀z[[∨T(p)](z)↔z=y] & K(y)] <,t>
According to (28), [e] denotes "the set of characteristics of the unique entity which stands in the relation T to p" (p. 142). T is a variable of type <, >>, a function from propositions to properties, which does not appear in Cooper's (1979) original proposal. It remains as a free variable and its value is determined by "the relevant semanticopragmatic information coming from the embedded clause and the matrix clause" (p. 142). It is assumed that λ-abstraction over a propositional variable p is applied to (28) when it combines with the denotation of CP, which results in (29).
(29)
λpλK∃y[∀z[[∨T(p)](z)↔z=y] & K(y)]
<,<,t>>
Let us look at a concrete example. The sentence in (30) receives the interpretation in (31a), which Hoshi paraphrases as (31b). I provide its compositional derivation in (32).25
25
The representation in (32) departs from Hoshi's original in that in (32), the object NP is moved (by Quantifier Raising) to a position where it is interpretable. The index of the moved NP is interpreted as a lambda abstractor over a coindexed variable, as is standardly assumed.
117
(30)
John-ga [[Mary-ga
teeburu-no ue-ni orenzi-o
John-Nom Mary-Nom table-Gen on
oitekureta]-no]-o
orange-Acc put
-NM -Acc
tenitotta. picked_up 'Mary put an orange on the table and John picked it up.' (31)
a. ∃y[∀z[[∨T(^∃x[orange(x) & put_on_table(x)(mary)])](z)↔z=y] & pick_up(y)(john)] b. There is some unique entity that bears the relation T to the proposition 'there is some orange which Mary put on the table', and John picked it up.
(32)
∃y[∀z[[∨T(^∃x[orange(x) & put_on_table(x)(mary)])](z)↔z=y] & pick_up(y)(john)]
λK∃y[∀z[[∨T(^∃x[orange(x) & put_on_table(x)(mary)])](z)↔z=y] & K(y)] λy[pick_up(y)(john)] NP1 5 VP ∃x[orange(x) & λpλK∃y[∀z[[∨T(p)](z)↔z=y] & K(y)] 2 put_on_table(x)(mary)] 1 John-ga 2 CP N t1 V 3 [e] y tenitotta ∨ picked up IP C λK∃y[∀z[[ T(p)](z)↔z=y] & K(y)] no % Mary-ga teeburu-no λp[p] ue-ni orenzi-o oitekureta Mary put an orange on the table
The crucial part of Hoshi's compositional interpretation involves the function assigned to the variable T in the denotation of the empty argument [e]. T takes propositions and returns properties. For example, in (31a), T(^∃x[orange(x) & put_on_table(x)(mary)]) is claimed to represent "the property of being the 'salient entity' with respect to the proposition 'Mary put an orange on the table'" (p. 147). Elsewhere (Hoshi 1996, p. 87), T
118
is described as "a kind of 'saliency' operator which takes a proposition and gives a unique salient entity with respect to the proposition." There are two problems associated with this proposal. First, the semantic operation Hoshi seems to have in mind with respect to the interpretation of T requires more explicit specification. In particular, it is not obvious from his proposal how the semantics can 'look into' unstructured propositions and return properties. On standard assumptions, which he seems to adopt, too, the semantic values of sentences are sets of possible worlds/situations (or characteristic functions thereof). Thus, in (32), the embedded CP denotes a set of possible worlds/situations. The crucial function that is assumed for the variable T, then, is to recover a property of being the salient entity with respect to a particular set of worlds/situations. How the relevant properties are actually recovered is not discussed in his proposal.26 Second, sentences like (14) from section 3.3.2 pose another problem for Hoshi's implementation.
(14)
Dono gakusei-moi [soitui -ga /proi kongakki
peepaa-o
3-bon
which student-MO (s)he-Nom/pro this semester term paper-Acc 3-CL kaita] -no] -o
kesa
teisyutusita.
wrote -NM -Acc this morning turned_in 'Every student wrote three term papers this semester and turned in the term papers he or she wrote this semester this morning.'
It was shown in section 3.3.2 that the kind of anaphora involved in (14) is neither bound variable anaphora nor referential anaphora, hence motivating a need for an independent
26As
pointed out by Angelika Kratzer (p.c.), the relevant process may be made explicit if we define some notion of "aboutness" as in, for instance, 'a proposition p is about an individual a, iff when w ∈ p, a exists in w'.
119
mechanism, E-type anaphora. The IHR-DP in (14) does not refer to a particular group of term papers but its denotation varies with each choice of a student. However, Hoshi's analysis of the empty argument [e], unlike Cooper's (1979) analysis of English E-type pronouns, does not seem to allow an occurrence of an individual variable in it that is bound from outside, even though [e] is assumed to be a definite description. It does not seem straightforward to allow a bound variable in [e] in the LF representation, given his analysis in which T is a free variable. The process involved in the interpretation of the empty argument [e] that he seems to have in mind essentially boils down to a process for a referential pronoun to pick up its referent, some contextually salient individual. Since a sentence like (14) is the clearest example that illustrates the need to assume E-type anaphora in the interpretation of IHRs, it is essential that a proper E-type analysis accommodate it.
3.4.2 Amended E-type Analysis I now spell out an LF representation and compositional interpretation of the IHR, a more explicit version of the general E-type analysis that overcomes the weakness of Hoshi's particular implementation. My formulation is based on the analysis of English Etype pronouns spelled out in Heim and Kratzer (1998), which is in turn based also on an analysis due to Cooper (1979). In Heim and Kratzer (1998), the sentence in (33a) is assigned the representation in (33b). (The indices of moved phrases are interpreted as variable binders adjoined to the sisters of the moved phrases.)
(33)
a. Every host bought just one bottle of wine and served it with the dessert.
120
b.
S 4 DP 4 2 1 S every host 5 VP 3 4 and VP DP 2 2 2 2 VP t1 2 just one bottle of 2 served_with_ DP => it wine t1 2 dessert 2 the NP bought t2 2 N DP R<7,>> pro<1,e>
The DP in the boldface in (33b) is assumed to be spelled out as pronoun it. It contains a definite article and a predicate that consists of two variables, R and pro. The R-variable is a free variable of type >, and it receives its denotation from the context of utterance c. The context of utterance c specifies a salient two-place relation, through the variable assignment gc in (34).
(34)
gc := [7 →λx ∈ De. λy ∈ De. y is a bottle that x bought]
The second variable pro in (33b) is bound from the subject position, and it saturates the first argument of the two-place relation denoted by the R-variable. The whole sentence receives the interpretation in (35). Pronoun it in (33a) is interpreted as 'the bottle that x bought', a definite description that contains a bound variable, allowing its value to vary with each choice of a host.
(35)
For every host x, there is just one bottle y such that x bought y, and x served the bottle that x bought with the dessert .
121
Let us go back to IHRs. I propose that an IHR-DP contains an unpronounced proform, which is a predicate that denotes an n-place property. As we saw in the previous paragraph, a similar assumption is independently necessary in a general theory of E-type anaphora. I also assume that the morpheme -no takes this predicate, and, like the English definite article the, returns a maximal individual that satisfies the property denoted by the predicate. How is the whole IHR construction interpreted? I spell that out using example (36a). (36b) is its (simplified) LF representation.27
(36)
a. Taro-wa [[Hanako-ga
dono sinbun-mo
katte kita] -no] -o
Taro-Top Hanako-Nom which newspaper-MO buy_came -NM -Acc tana-ni narabeta. shelf-on placed 'Hanako bought (and brought) every newspaper and Taro shelved them.'
27The
LF structure in (36b) is greatly simplified, and is intended only to demonstrate how the basic syntax-semantics mapping works. (The same remark applies to the structure in (39) below.) Various other internal structures of the IHR-DP are of course conceivable. For example, the morpheme -no could be decomposed into a definite determiner and a proform that corresponds to N in (36b), or the base positions of CP (or IP) and NP could be different from those depicted in (36b). These issues do not affect the main point of my proposal.
122
b.
IP 4 CPi IP 4 VP I 2 ta Taro 2 Past Hanako-ga dono sinbun-mo DP V katte kita 2 narabe Hanako bought every newspaper ti D' shelve 2 NP D 1 no N' NM 1 N P<3, >
It was demonstrated in sections 3.2 and 3.3 that an IHR-clause is interpreted as a simple sentence that is separate from the matrix sentence, even though it seems to form a constituent with the rest of the IHR-DP. We find closely analogous properties in nonrestrictive relative clauses. It has been observed that non-restrictive relative clauses are interpreted as separate sentences, and do not compose semantically with phrases they appear to modify (Ross 1967; Emonds 1979; Demirdache 1991). Applying Demirdache's (1991) implementation to the case of IHRs, I assume an LF representation in which an IHR-clause is adjoined to the matrix IP, as shown in (36b). 28,29 28Given
the way I define the meaning of -no shortly in (38), it follows that the IHRclause cannot be semantically composed with the rest of the IHR-DP because their types do not match. We could say, then, that what drives the "IHR-clause lifting" is a type mismatch, as in instances of Quantifier Raising. Since what is crucial here is for an IHRclause to be adjoined to any position where it is interpretable, the adjoined position could be a VP, rather than an IP (as assumed in the text), as pointed out by Satoshi Tomioka (p.c.). I assume that the trace of the IHR-clause is not interpreted. The IHR-clause lifting by itself is not a crucial part of my analysis, because it is conceivable that fixing the type of -no will make the IHR-clause interpretable in-situ. I would like to thank a reviewer for bringing up relevant questions. 29I
simply assume here that the IHR-clause and the matrix clause are connected with logical connective '&' in the semantic representation, leaving the exact semantic relation between the two clauses indeterminate. This leaves space for pragmatics to play a role in determining the semantic relations (e.g., causation, temporal sequence, concessivity, accompanying circumstances, etc.). In this respect, I am following Kuroda (1975-76). 123
In the matrix object DP in (36b), the N-position is occupied by an unpronounced proform, a variable P of type . The P-variable remains as a free variable and receives its denotation via the assignment gc in (37) in the utterance context c in which the sentence is uttered. The content of the IHR-clause provides a salient property, in this case, the property of being 'newspapers that Hanako bought'.
(37)
gc := [3 →λx ∈ De. x is newspapers that Hanako bought]
The anaphoric NP then combines with the morpheme -no, which I take to be semantically non-vacuous. While there has been much discussion on the categorial status of this morpheme (for example, whether it is a C or an N—see the references cited in footnote 3), virtually no serious attention has been paid to the question of what the semantic contribution of the morpheme -no is. I assume that the function of -no is similar to that of the definite article the in English, as defined in (38).30
(38)
ªnoº ∈ D<,e> ªnoº(f) denotes the maximal individual a such that f(a) = 1.
Given the variable assignment in (37) and the lexical meaning of -no as defined in (38), the IHR-DP (the boldfaced DP in (36b)) denotes the maximal individual that satisfies the property of being newspapers that Hanako bought, which is equivalent to what the English definite description 'the newspapers that Hanako bought' denotes.
The semantic relations that hold between the two clauses overlap with those observed between two clauses in asymmetric conjunctions (e.g., Taro went to the store and bought a bottle of sake; Lakoff 1986) and in absolute constructions (e.g., With the children asleep, Mary watched TV; Stump 1985). 30I
leave open the question of whether this characterization of -no in IHRs carries over to other occurrences of a homophonous morpheme.
124
In section 3.4.1, example (14), repeated below, was pointed out to be problematic for Hoshi's (1995) analysis.
(14)
Dono gakusei-mo4 [soitu4 -ga /pro4 kongakki
peepaa-o
3-bon
which student-MO (s)he-Nom/pro this semester term paper-Acc 3-CL kaita] -no] -o
kesa
teisyutusita.
wrote -NM -Acc this morning turned_in 'Every student wrote three term papers this semester and turned in the term papers he or she wrote this semester this morning.'
The sentence is represented as (39) at LF in the amended E-type analysis.
(39)
IP 4 DP 4 @ 4 IP dono gakusei-mo 4 every student CPi IP 4 VP I 2 ta t4 2 Past V pro4 kongakki peepaa-o 3-bon kaita DP 2 teisyutusi he4 wrote 3 papers this semester D'turn_in ti 2 NP D 1 no N' NM 2
DP N pro<4,e> R<9,>>
In (39), there is an N-position which is occupied by an unpronounced proform, this time a free variable R of type >. This proform is a relational noun, like sister, and it
125
denotes a salient two-place relation in the utterance context. In this example, it receives its value from the variable assignment in (40).
(40)
gc := [9 →λx ∈ De. λy ∈ De. y is term papers that x wrote this semester]
The bound variable pro in (39) saturates the first argument of the relational noun. After combining the denotation of NP with the denotation of -no, we end up with the desired interpretation of the matrix object, which can be expressed as the definite description 'the term papers x wrote this semester'. To summarize, an IHR-DP contains a proform that is a predicate that denotes an n-place property. The proform remains a free variable, and receives its value from the context of utterance. It then combines with the morpheme -no, and the IHR-DP as a whole is interpreted as a definite description. In Hoshi's (1995) analysis, too, an IHR-DP is interpreted as a definite description, with the minor difference that his analysis is stated in terms of Russell's (1905) analysis of definite descriptions, while the amended E-type analysis is stated in terms of Link's (1983) analysis of definite descriptions. The amended analysis differs from Hoshi's in that it assumes a property variable in a definite description that remains free (as opposed to the variable T of type <, >> in his analysis), and that the definite description could contain an individual variable bound from outside which serves as an argument of the unpronounced predicate. In this respect, the amended analysis is more "Cooperian" than Hoshi's analysis. Earlier, I made a connection between IHRs and non-restrictive relative clauses (NRCs), in that they share the property that an embedded clause is interpreted as if it were an independent sentence and is not composed with the rest of the DP it occurs in. The two constructions share another property with respect to the type of anaphoric relation involved. Sells (1986) shows that NRCs involve what he calls "cospecificational anaphora," which is basically the same as E-type anaphora. He notes that an example
126
like (41) must involve cospecificational/E-type anaphora between which and its antecedent two doors, because the anaphoric relation here is neither a referential nor a bound variable relation.
(41)
Each car has (exactly) two doors, which open only from the outside.
I illustrate anaphoric relations involved in NRC and IHR schematically in (42). For ease of comparison, (i) the NRC in (42a) is adjoined to the matrix clause following Demirdache (1991), and (ii) it is adjoined to IP although Demirdache leaves open the question whether the adjoined node is IP or CP.
(42)
a. NRC
b. IHR IP 2 IP CPNRC
IP 2 CPIHR IP
# #
# #
..........DP wh.........
.....DP..... .....DP.....
2
DP
2
tIHR 2 NP D -no
tNRC
In (42a), wh in the NRC has a cospecificational/E-type interpretation, and its antecedent DP is in the matrix clause. On the other hand, in the case of IHR, shown in (42b), the antecedent or head DP is in the IHR-clause, while the DP in the matrix clause receives a cospecificational/E-type interpretation. Despite these similarities, IHR-clauses cannot be equated with NRCs, contrary to Kitagawa's (1996) proposal. The content of an IHR-clause (indirectly) supplies a relevant restriction for a definite description (=IHR-DP) in the matrix clause, while an NRC only adds additional information to its antecedent.31 Moreover, it was pointed out 31Hoshi
(1995, p. 127, fn. 7) notes a similar point.
127
in section 3.3.3.2 that the presence of IHRs that contain wh-phrases clearly shows that they cannot be NRCs.
3.5 Consequence: The So-Called "Indefiniteness" Restriction 3.5.1 The Pro Analysis The amended E-type analysis not only predicts the correct semantic interpretations for the data I have observed so far, but also has another welcome consequence with respect to deriving a restriction on possible internal heads. Notice that in my analysis, an anaphoric element in an IHR-DP is a phonologically null N-proform, rather than a phonologically null pronoun (pro), which is known to occur frequently in Japanese. Suppose instead that we took this latter option and assumed a pro in a matrix clause, as, for instance, Mihara (1994a,b), Murasugi (1994) and Kitagawa (1996) in fact do. Since pros in Japanese are known to have varying semantic functions including an Etype pronoun use, it may appear that the pro analysis can accommodate the data we have observed so far. The rather sharp contrast between the (a) and (b) examples in (43) and (44), however, is unexpected under the pro analysis.
(43)
a. [[Daidokoro-no mado-kara
siroi neko-ga haitte kita]-no]-ga
kitchen-Gen window-from white cat-Nom came_in -NM -Nom sakana-o totte nigeta. fish-Acc steal_ran_away 'A white cat came in from the kitchen window and it stole a fish and ran away.'
128
b.?*[[Daidokoro-no mado-kara kitchen-Gen
Lucky-ga
haitte kita]-no]-ga
window-from Lucky-Nom came_in -NM -Nom
sakana-o totte nigeta. fish-Acc steal_ran_away 'Lucky came in from the kitchen window and it stole a fish and ran away.' (44)
a. John-wa [[[Kathy-ga John-Top -no
ofisu-ni yagi-o
2-too turete kita]-no]
Kathy-Nom office-to goat-Acc 2-Cl brought
ke] -o
-NM
katta.
-Gen hair -Acc cut 'Kathy brought two goats to the office and John cut their hair (sheared them).' b. ?*John-wa [[[Kathy-ga John-Top -no
ofisu-ni Sebastian-o
turete kita]-no]
Kathy-Nom office-to Sebastian-Acc brought
ke] -o
-NM
katta.
-Gen hair -Acc cut 'Kathy brought Sebastian to the office and John cut his hair (sheared him).'
Since a pro also has a referential use, taking an entity of type e as an antecedent, it is unexpected in the pro analysis that the proper names are not acceptable as internal heads in (43b) and (44b). The same problem arises with Hoshi's (1995) analysis as well, in which the matrix empty argument [e] is assumed to be a referring expression.
3.5.2 Recovering a Property In the proposal I made above, what is recovered is a property, through Nproforms—see (36) and (39). Furthermore, the generalization holds that a recovered property, which forms a restriction for the definite description in the matrix clause, should obligatorily include the predicative part of the internal head. (This is in fact what causes a particular DP in the IHR-clause to be construed as its 'internal head'.) Suppose
129
that in (45), a variation of (36a), a recovered property consisted only of a property of being 'bought by Hanako', excluding a property of being 'newspapers'. Then the IHR-DP would denote the maximal individual a such that Hanako bought a, which can be paraphrased by the English free relative 'what Hanako bought'. This predicts that for the second sentence in (45) to be true, Taro has to shelve the newspapers she bought and the books she bought. These are not the correct truth conditions for the sentence. They are too strong.
(45)
Hanako-ga
hon-o
3-satu katte kita.
Hanako-Nom book-acc 3-CL buy_came Taro-wa [[Hanako-ga
mata
Taro-Top Hanako-Nom also
dono sinbun-mo
katte kita]-no]-o
which newspaper-MO buy_came-NM-Acc
tana-ni narabeta. shelf-on placed 'Hanako bought (and brought) three books. Hanako also bought (and brought) every newspaper and Taro shelved them.'
A similar generalization holds for an English E-type pronoun and its antecedent. The question of how this fact should be captured in a principled way is left open for future research (see a related discussion in Heim 1990). Whatever the explanation for this generalization is, it follows that possible internal heads are those that contain predicates. Williamson (1987) observes that only indefinite NPs can appear as the internal head in Lakhota internally headed relative construction. Watanabe (1992a) claims that the same restriction applies to the IHR in Japanese, but this claim has aroused some controversy (see e.g., Hoshi 1995). The preceding discussion on this issue is based on a dichotomy between indefinite NPs on the one hand, and definite NPs and proper names on the other hand. Notice, however, that the above prediction about a possible internal
130
head draws a line between predicative and non-predicative NPs, rather than between indefinite and definite NPs.32 The contrast in (43) and (44) now falls out. In (43a), for instance, the internal head siroi neko 'white cat' is or at least contains a predicate of type , and can be included in the denotation of the proform as in (43a'). On the other hand, the internal head Lucky in (43b) is of type e and cannot form part of the denotation of the proform. Thus (43b') cannot be a recovered property because it fails to include a predicative part of the internal head (for there is none).
(43a') λx ∈ De. x is a white cat and x came in from the kitchen window (43b') λx ∈ De. x came in from the kitchen window
The example in (46) shows that a definite NP as an internal head has a degraded status, but is not as bad as proper names.
(46) ?/?? [[Daidokoro-no mado-kara kitchen-Gen
Isadora-no neko-ga haitte kita]-no] -ga
window-from Isadora-Gen cat-Nom came_in -NM -Nom
sakana-o totte nigeta. fish-Acc steal_ran_away 'Isadora's cat came in from the kitchen window and it stole a fish and ran away.'
The in-between status of definite NPs as internal heads is expected if we assume that they can be either referential (type e) or predicative (type ).33 In example (46), if the 32I
am grateful to Angelika Kratzer, who first suggested to me that the relevant data should be re-examined from a different perspective. 33See,
for example, Heim (1982) and Diesing and Jelinek (1995).
131
head 'Isadora's cat' is construed as denoting a property, it can form part of a restriction for the matrix subject as in (46'), whereas if it is construed as referential, this is not possible.
(46')
λx ∈ De. x is Isadora's cat and x came in from the kitchen window
3.5.3 IHR-DPs in the Direct Object Position Proper names and definite NPs often sound better as internal heads when the IHRDP occurs in the direct object position of the matrix clause, as in (47). (Notice that in the earlier examples in (43), (44) and (46), an IHR-DP occurs as the subject of the matrix clause or as a genitive-marked phrase.)
(47)
Taro-wa [[daidokoro-no mado-kara Taro-Top kitchen-Gen haitte kita]-no] -o
{siroi neko/Lucky/Isadora-no neko} -ga
window-from white cat/ Lucky/Isadora-Gen cat
-Nom
tukamaeta.
came_in -NM -Acc caught 'Taro caught a white cat/Lucky/Isadora's cat as she came in from the kitchen window.'
Examples like this may present a challenge to the generalization above on possible internal heads based on the contrast between (43a) and (43b), and between (44a) and (44b). I only speculate here that the apparent IHR construction in (47) may be more accurately characterized as something analogous to the so-called tokoro circumstantial adverbial construction as shown in (48), rather than the genuine IHRs that this chapter is concerned with.
132
(48)
Taro-wa [daidokoro-no mado-kara
Lucky-ga
haitte kita tokoro] -o
Taro-Top kitchen-Gen window-from Lucky-Nom came_in TOKORO -Acc tukamaeta. caught 'Taro caught Lucky as she came in from the kitchen window.'
I cannot review analyses of this construction here, and readers are referred to works such as Harada (1973), Hale and Kitagawa (1976-77), and Kuroda (1978). If it should turn out that (47) shares general properties with the tokoro construction, which allows a proper name as in (48), then the fact that a proper name sounds good as an internal head in (47) would not be surprising.34 The examples in (49) and (50) show that a tokoro-clause cannot occur in subject position or in a genitive-marked position (The first point is also noted in Kitagawa 1996, pp. 11-13).
(49) *[Daidokoro-no mado-kara kitchen-Gen
siroi neko-ga
haitte kita tokoro] -ga
window-from white cat-Nom came_in TOKORO-Nom
Taro-ni tukamatta Taro-by was_caught 'A white cat was caught by Taro as it came in from the kitchen window.'
34The
data of the tokoro construction seem to be compatible with the idea that there is a pro in the matrix clause. If examples like (47) were to be assimilated to this construction, as suggested by Mihara (1994a, b) and Murasugi (1994) (for all IHRs), the pro analysis would remain as a candidate for these cases.
133
(50) *John-wa [[Kathy-ga
ofisu-ni yagi-o
2-too turete kita tokoro] -no
John-Top Kathy-Nom office-to goat-Acc 2-Cl brought TOKORO -Gen atama]-o
nadeta.
head -Acc patted 'John patted the heads of the two goats as Kathy brought them to the office.'
Thus, when an IHR-DP occurs in a subject or genitive-marked position as in (43) and (44), it cannot be related to the tokoro construction. Hence I take the sharp contrast in acceptability between indefinite NPs and proper names as internal heads in (43) and (44) to represent a true generalization about possible internal heads in genuine IHRs. In summary, the fact that proper names do not sound good as internal heads falls out automatically in my analysis, in which what is recovered is a property, by means of N-proforms, but crucially not by means of pros, given the descriptive generalization that a restriction for the definite description in the matrix clause should obligatorily include a predicative part of the internal head. The internal head must contain a predicative part which can form part of the restriction for the definite description, hence, no proper names can serve as internal heads.
3.6 The Apparent Lack of Pragmatic Variability in the IHR In my current proposal, what property is recovered is determined by a variable assignment gc under the utterance context c, leaving the work to pragmatics. In this section I discuss a particular property that is observed in typical E-type anaphora in English, but apparently not shared by IHRs, which poses a problem for any E-type analysis of the IHR. The data may suggest that the relevant information should be recovered not by a pragmatic mechanism, but by a syntactic one. I will show, however, that this does not provide a solution to the problem. This section ends with a suggestion
134
that the so-called Relevancy Condition (Kuroda 1975-76), when formally spelled out, may be a key factor in solving the problem.
3.6.1 A Predicate Recovered from VP One advantage of assuming the pragmatic way of recovering salient properties for E-type pronouns in English is that it accounts for the pragmatic variability which is observed with respect to what property (other than the property recovered from the antecedent) enters the restriction for the definite description. Consider the example in (51).35
(51)
{Almost no/only few} graduate student(s) came to the party. They were busy.
In (51), the pronoun they should be interpreted as 'the graduate students', rather than 'the graduate students who came to the party'. The Japanese IHR shows seemingly different behavior from English E-type pronouns in this respect. The sentence in (52) is analogous to (51), except that the part that corresponds to the first sentence of (51) is expressed by an IHR in (52). NP+SIKA, in combination with negation, means 'only NP'.
(52)
#[[Honno suunin-no insee-sika only -no]-ga
doyoobi-no
party-ni ikanakatta]
a_few-gen grad_student-SIKA Saturday-Gen party-to go-Neg-Past jitsuwa uchi-de term paper-o
kaite ita.
-NM-Nom in_fact home-at term paper-Acc was_writing 'Only a few graduate students went to the party on Saturday. In fact they were writing term papers at home.'
35I
thank Angelika Kratzer for drawing my attention to this kind of examples.
135
Compare (52) with (53), which is a more direct paraphrase of the English example (51). Here, the pronoun karera 'they' is used.
(53)
Honno suunin-no insee-sika only
doyoobi-no
party-ni ikanakatta.
a_few-gen grad_student-SIKA Saturday-Gen party-to go-Neg-Past
Karera-wa jitsuwa uchi-de term paper-o
kaite ita.
they-Top in_fact home-at term paper-Acc was_writing 'Only a few graduate students went to the party on Saturday. In fact they were writing term papers at home.'
The fact that (52) does not have the intended interpretation shows that the kind of pragmatic variability observed in the interpretation of pronouns in (51) and (53) is not observed in (52). That is, it appears that recovering only the property of being a graduate student out of the IHR and forming the definite description 'the graduate students' is not available in the case of Japanese IHR. It seems, then, that the way in which the P/R-variable in the IHR gets its denotation from the context is not as free as the way in which the P/R-variable in English E-type pronouns gets its denotation from the context. In the IHR, it appears that the content of the VP in the embedded clause plays a crucial role in the restriction of the definite description in the main clause. This may suggest that some structural mechanism is at work. In the next sub-section, I consider a structural analysis that ensures that the content of VP in the embedded clause is obligatorily included in the restriction of a definite description in the matrix argument position. I point out advantages and disadvantages of the structural analysis.
136
3.6.2 Predicative Internal Heads and Predicate Modification One conceivable attempt to make sense of the different behavior in the IHR and English E-type pronouns may be to assume that in the case of IHR, the contribution of the VP content is syntactically accommodated into the restriction of a matrix definite description. More specifically, suppose that the denotation of the internal head combines with the denotation of the rest of the clause via the operation of predicate modification. The resulting property then combines with -no, whose denotation is the same as that of English definite determiner the. The compositional derivation of the meaning of sentence (54) is roughly depicted in (55).
(54) [[Haiiro-no neko-ga mado-kara
haitte kita]-no]-ga
catnip-o
tabeta.
gray-Gen cat-Nom window-from came_in -NM-Nom catnip-Acc ate 'A gray cat came in from the window and she ate catnip.' (55)
VP 5 'the gray cat that came in from the window' DP 3 4 DP V λx[gray_cat(x) & came_in_from_window(x)] CP D catnip ate 2 no IP C NM 2 λx[gray_cat(x)] DP 2λx[came_in_from_window(x)] gray cat VP I 2 PP V from the window came in
This method ensures that the content of the embedded VP is included in the restriction of the definite determiner, thereby ensuring the contribution of the VP content to the semantics of the entire sentence. Moreover, this analysis could potentially account for the set of data observed in section 3.5, namely, the fact that proper names are not very good internal heads,
137
indefinite NPs are perfect heads, and definite NPs have an in-between status. The analysis predicts that only those heads that denote an (n-place) property are allowed. Proper names are correctly predicted to be not possible as a head, because they are referential expressions, and indefinites are correctly predicted to be possible heads, because they are predicative. Definites are of marginal status since they can be either referential or predicative. This analysis, however, faces serious problems. First, it makes a strong prediction that only those NPs that denote a property (of type , >, etc.) can be internal heads, and that truly quantificational NPs cannot occur as internal heads. We have seen, however, a range of examples of IHRs that involve quantificational NPs. For instance, the sentence in (36a) contains the internal head dono simbun-mo, which means 'every newspaper'. The second problem with this analysis is that it cannot accommodate the following kind of data.
(56)
a. Wasaburo-wa [[dono gakusei-mo Wasaburo-Top which student-MO -no] -o
itiniti-de
peepaa-o
3-bon dasita]
term paper-Acc 3-CL
turned_in
yonda.
-NM -Acc one_day-in read 'Every student turned in three term papers and Wasaburo read them in one day.' b. Wasaburo-wa [[3-nin-no kodomo-ga sorezore ringo-o Wasaburo-Top 3-CL-Gen child-Nom each 2-tu-zutu katte kita] -no] -o
apple-Acc
tana-ni oita.
2-CL-each buy_came -NM -Acc shelf-on put 'Three children bought two apples each and Wasaburo put them on the shelf.'
138
In (56a), what Wasaburo read is all the term papers that were turned in to him, rather than 'three term papers.' If n is the number of students, Wasaburo read (3xn)-number of papers. Similarly, in (56b), what Wasaburo put on the shelf are (3x2) apples, rather than 'two apples'. This means that we cannot simply assume the internal heads 'three term papers' or 'two apples' to be predicative and to form part of the restriction for the definite description. Thus, the structural analysis is not a suitable account for the lack of pragmatic variability.
3.6.3 The Relevancy Condition In this section I suggest a possible line of approach to the apparent lack of pragmatic variability, as an alternative to adopting some structural mechanism. The alternative keeps the main proposal based on the pragmatic recovery of properties, and appeals to an already existing well-formedness condition on the IHR to account for the difference between IHRs and English E-type pronouns. It has been noted in the literature that the IHR has much more limited distribution than its external counterpart. If we take any externally headed relative clause and turn it into an internally headed one, it is very likely that we end up having an unacceptable sentence. (57a) contains an EHR-DP 'the gray cat that came in from the window yesterday', while (57b) contains the internal version of it and the sentence is unacceptable.
(57)
a. [ø Kinoo
mado-kara
haittekita [haiiro-no neko]]-ga
yesterday window-from came_in kesa
gray-Gen cat
-Nom
mata yattekita.
this morning again came 'The gray cat that came in from the window yesterday came back this morning.'
139
b. ?*[[Haiiro-no neko-ga kinoo
mado-kara
haittekita]-no]-ga
gray-Gen cat-Nom yesterday window-from came_in -NM-Nom kesa
mata yattekita.
this morning again came 'A gray cat came in from the window yesterday, and it came back this morning.'
Kuroda (1975-76; 1992, p. 147) gave the following generalization which is supposed to describe this peculiar property of the IHR.36
(58)
The Relevancy Condition: For a pivot-independent relative clause [= IHR, J.S.] to be acceptable, it is necessary that it be interpreted pragmatically in such a way as to be directly relevant to the pragmatic content of its matrix clause.
The condition, in essence, requires that some intrinsic connection should exist between the two events represented by the IHR and the matrix clause. Kuroda (1975-76; 1992, p. 151) paraphrases it in the following way. (59)
"... the two events represented by the relative and the matrix clause constitute, so to speak, a superordinate event either in the physical world, thanks to simultaneity or co-locationality, or in the consciousness of a protagonist in the sentence, thanks to purposiveness, motivation, etc."
The unacceptable status of (57b) can thus be considered to be due to violation of this condition. The two events described by the IHR and the matrix cluase are temporally distinct, and they are not easily perceived as "two subevents of a continuum of events". 36The
Relevancy Condition consists of several sub-conditions (simultaneity, colocationality, purposive or motivational connections), and the IHR should satisfy at least one of them to be acceptable.
140
Let us now go back to our concern: the apparent lack of pragmatic variability in the IHR. We observed that in (51), repeated below as (60), the E-type pronoun they can mean 'the graduate students', rather than 'the graduate students who came to the party.' On the other hand, in (52), repeated below as (61), it appears as if this strategy is not available.
(60)
{Almost no/only few} graduate student(s) came to the party. They were busy.
(61)
#[[Honno suunin-no insee-sika only -no]-ga
doyoobi-no
party-ni ikanakatta]
a_few-gen grad_student-SIKA Saturday-Gen party-to go-Neg-Past jitsuwa uchi-de term paper-o
kaite ita.
-NM-Nom in_fact home-at term paper-Acc was_writing 'Only a few graduate students went to the party on Saturday. In fact they were writing term papers at home.'
This contrast led us to consider the syntactic mechanism that obligatorily included the embedded VP part in the restriction for the matrix argument. However, there is an alternative way to look at this phenomenon: we could see the lack of the intended reading in (61) as stemming from its violation of the Relevancy Condition. In (61), it is not immediately clear how the event described in the IHR constitutes one super event with the event described in the matrix clause. In this approach, a salient property is still recovered via a variable assignment, just as in English E-type pronouns, rather than through some syntactic mechanism. The approach sees the contribution of the content of the embedded VP to the semantics of the whole sentence as an effect of the Relevancy Condition, which must be satisfied in Japanese IHRs, but not in English E-type pronouns. I suggested that the contribution of the content of an embedded VP to a sentence as a whole can be captured without assuming a syntactic mechanism. Notice that simply
141
guaranteeing the VP part to be included in the restriction structurally, by itself, does not give us a way to explain the effect of the Relevancy Condition. The structural solution would make no prediction about the acceptability contrast observed between sentence (57a), which contains an externally headed relative clause, and sentence (57b), which contains its internal version. Needless to say, a formal characterization of the Relevancy Condition in terms of the event structure/temporal structure of the matrix and embedded clauses needs to be done, so that we can pinpoint what exactly is violated in sentences such as (57b) and (61) (see Fuji 1996 for discussion on the temporal structure of IHRs and matrix clauses). We have examined a property of Japanese IHRs that English E-type pronouns do not seem to share. Here, I touch upon another area where the interpretation of IHRs shows some peculiar property that is not observed in the interpretation of English E-type pronouns. It involves genitive phrases. In (63) and (64), her or pro can have NP his mother or soitu-no hahaoya 'his mother' that occurs inside another NP as an antecedent. However, the sentence in (65), which involves an IHR, is ungrammatical under the reading in which every host praises his mother.
(63)
Every host served [his mother's sushi] and praised her immediately.
(64)
Dono hosuto1 -mo [[soitu1 -no hahaoya-no] sushi]-o which host-MO suguni
his
dasite,
mother-Gen sushi -Acc served
pro hometa
immediately
praised
'Every host served his mother's sushi and praised her immediately'
142
(65) *Dono hosuto1 -mo [[pro1 [[soitu1 -no hahaoya-no] sushi]-o which host-MO suguni
his
dasita]-no]-o
mother-Gen sushi -Acc served-NM-Acc
hometa.
immediately praised 'Every host served his mother's sushi and praised her immediately'
This contrast is rather puzzling because we cannot appeal to a lesser degree of saliency of soitu-no hahaoya 'his mother' in (65). Moreover, the two events described in the main and embedded clauses seem to be rather closely related. It seems to be the case that only thematic role bearers of the event in the lower clause can be the internal head. This may be due to part of a constraint on event structure that makes sure that the two events have a tight connection. Questions about constraints on event structure are left for future research. The event structure of sentences that contain IHRs should also be compared with asymmetric conjunctions (e.g., Taro went to the store and bought a bottle of sake; Lakoff 1986) and absolute constructions (e.g., With the children asleep, Mary watched TV; Stump 1985).
3.7 Concluding Remarks I have provided evidence that the apparent dual role played by the internal head should be characterized in terms of anaphora, rather than in terms of LF head raising. Furthermore, I have shown that the relevant anaphoric relation is that of E-type anaphora. An explicit mechanism of how the meaning of the construction is composed of its parts was provided, which was couched in a general theory of E-type anaphora. This study of Japanese IHRs supports the general idea expressed in, for instance, Basilico (1996) and Grosu and Landman (1998) that IHRs in various languages do not form a homogeneous class. Basilico (1996) tries to derive certain word order facts in IHR-clauses in some languages, e.g., the Yuman languages (Diegueño, Cocopa and
143
Mojave), Northern Athabaskan (Koyukon and Tanaina) and Mooré, from the assumption that IHR-DPs in these languages are associated with a definite operator. In my analysis of Japanese IHRs, too, the morpheme -no functions as a definite operator. Basilico's analysis and mine crucially differ in the following respect: In his analysis, the IHR-clause itself, which is interpreted as an open sentence, functions as a restriction on the definite operator, whereas in my analysis, the IHR-clause is a closed sentence and what functions as a restriction is a salient property recovered contextually.37 Basilico's analysis makes a prediction that truly quantificational NPs cannot occur as internal heads because they do not provide a variable to be bound by the definite operator (as it is already bound). We have seen, however, a range of examples of Japanese IHRs that involve quantificational NPs in sections 3.2.2 and 3.2.3. 38 If Basilico's analysis is on the right track with respect to the IHRs in the languages he examines, and if what I have called quantificational NPs in the examples of Japanese IHRs are indeed instances of truly quantificational NPs, it must be the case that Japanese IHRs belong to a different category than the IHRs in the languages he examines. Even if it turned out that what I have been calling quantificational NPs in Japanese are not truly quantificational and can be considered to be predicative, leaving Basilico's analysis still as an alternative, the examples in (56), repeated below as (66), show that in Japanese IHRs, forming of a definite description must be mediated through a pragmatically recovered property, rather than the more direct way assumed in his analysis.
37Grosu
and Landman's (1998) analysis of IHRs in Quechua and Japanese is similar to Basilico's (1996) in this respect. Their "maximalization" also operates directly on embedded CPs, which is problematic for Japanese for the same reason that I will point out below with respect to Basilico's analysis. 38Basilico
(1996, pp. 524-525; pp. 528-529) also discusses apparent counter examples to his claim in Mooré and Navajo.
144
(66)
a. Wasaburo-wa [[dono gakusei-mo peepaa-o
3-bon dasita]
Wasaburo-Top which student-MO term paper-Acc 3-CL turned_in -no] -o
itiniti-de
yonda.
-NM -Acc one_day-in read 'Every student turned in three term papers and Wasaburo read them in one day.' b. Wasaburo-wa [[3-nin-no kodomo-ga sorezore ringo-o Wasaburo-Top 3-CL-Gen child-Nom each 2-tu-zutu katte kita] -no] -o
apple-Acc
tana-ni oita.
2-CL-each buy_came -NM -Acc shelf-on put 'Three children bought two apples each and Wasaburo put them on the shelf.'
As we saw in section 3.6.2, in (66a), what Wasaburo read is all the term papers that were turned in to him, rather than 'three term papers'. Likewise in (66b), what Wasaburo put on the shelf were (3x2) apples, rather than 'two apples'. This means that the internal head 'three term papers' or 'two apples' cannot directly form part of the restriction for the definite description. What this suggests is that even those IHRs that seem to make use of a definite operator may not belong to one single category.
145
BIBLIOGRAPHY
Aoun, Joseph, and Yen-hui Audrey Li (1993) "Wh-Elements in Situ: Syntax or LF?," Linguistic Inquiry 24, 199-238. Baker, C. L. (1970) "Notes on the Description of English Questions: The Role of an Abstract Question Morpheme," Foundations of Language 6, 197-219. Barss, Andrew, Ken Hale, Ellavina T. Perkins and Margaret Speas (1991) "Logical Form and Barriers in Navajo," in C.-T. James Huang and Robert May (eds.), Logical Structure and Linguistic Structure, Kluwer, Dordrecht, pp. 25-47. Basilico, David (1996) "Head Position and Internally Headed Relative Clauses," Language 72, 498-532. Bäuerle, Rainer and Thomas Ede Zimmermann (1991) "Fragesätze," in Arnim von Stechow and Dieter Wunderlich (eds.), Semantics: An International Handbook of Contemporary Research, Walter de Gruyter, Berlin, pp. 333-348. Beaver, David (1995) Presupposition and Assertion in Dynamic Semantics, Ph.D. dissertation, University of Edinburgh. Berman, Stephen (1991) On the Semantics and Logical Form of WH-Clauses, Ph.D. dissertation, University of Massachusetts, Amherst. Bonneau, José (1992) The Structure of Internally Headed Relative Clauses: Implications for Configurationality, Ph.D. dissertation, McGill University. Brame, Michael (1968) "A New Analysis of the Relative Clause: Evidence for an Interpretive Theory," ms., MIT. Broadwell, George A. (1986) "A-Bar Anaphora and Relative Clauses," in Stephen Berman, Jae-Woong Choe and Joyce McDonough (eds.), Proceedings of the North East Linguistic Society 16, GLSA, University of Massachusetts, Amherst, pp. 47-56. Brockett, Chris (1994) "Mo: Quantificational Evidence for a Non-Quantificational Analysis," in Hiroyuki Ura and Masatoshi Koizumi (eds.), Formal Approaches to Japanese Linguistics 1, Proceedings of the First Conference on Formal Approaches to Japanese Linguistics, MIT Working Papers in Linguistics 24, pp. 45-59. Carlson, Greg (1977) Reference to Kinds in English, Ph.D. dissertation, University of Massachusetts, Amherst. Chierchia, Gennaro (1998) "Reference to Kinds Across Languages," Natural Language Semantics 6, 339-405. Choe, Jae-Woong (1987) "LF Movement and Pied-Piping," Linguistic Inquiry 18, 348353.
146
Chomsky, Noam (1973) "Conditions on Transformations," in Stephen R. Anderson and Paul Kiparsky (eds.), A Festschrift for Morris Halle, Holt, Rinehart and Winston, New York. Chomsky, Noam (1976) "Conditions on Rules of Grammar," Linguistic Analysis 2, 303351. Chomsky, Noam (1977) "On Wh-Movement," in Peter Culicover, Thomas Wasow and Adrian Akmajian (eds.), Formal Syntax, Academic Press, New York, pp. 71-132. Chomsky, Noam (1995) The Minimalist Program, MIT Press, Cambridge, Mass. Cole, Peter (1987) "The Structure of Internally Headed Relative Clauses," Natural Language and Linguistic Theory 5, 277-302. Cole, Peter, and Gabriella Hermon (1994) "Is There LF Wh-Movement?," Linguistic Inquiry 25, 239-262. Cole, Peter, and Gabriella Hermon (1998) "The Typology of WH-Movement: WH Questions in Malay," Syntax 1, 221-258. Cooper, Robin (1979) "The Interpretation of Pronouns," in Frank Heny and Helmut S. Schnelle (eds.), Syntax and Semantics 10, Selections from the Third Groningen Round Table, Academic Press, New York, pp. 61-92. Demirdache, Hamida (1991) Resumptive Chains in Restrictive Relatives, Appositives and Dislocation Structures, Ph.D. dissertation, MIT. Diesing, Molly and Eloise Jelinek (1995) "Distributing Arguments," Natural Language Semantics 3, 123-176. Dowty, David and Belinda Brodie (1984) "The Semantics of 'Floated' Quantifiers in a Tranformationless Grammar," in M. Cobler, S. MacKaye and M. T. Wescoat (eds.), Proceedings of the Third West Coast Conference on Formal Linguistics, CSLI, Stanford University, pp. 75-90. Emonds, Joseph (1979) "Appositive Relatives Have No Properties," Linguistic Inquiry 10, 211-243. Evans, Gareth (1980) "Pronouns," Linguistic Inquiry 11, 337-362. Fuji, Masaaki (1996) "Temporal Interpretation of Internally Headed Relative Clauses in Japanese," paper presented at Formal Linguistics Society of Midamerica VII, Ohio State University, Columbus. Fukushima, Kazuhiko (1991) "Phrase Structure Grammar, Montague Semantics, and Floating Quantifiers in Japanese," Linguistics and Philosophy 14, 581-628. Fukushima, Kazuhiko (1993) "Model Theoretic Semantics for Japanese Floating Quantifiers and Their Scope Properties," Journal of East Asian Linguistics 2, 213228. Groenendijk, Jeroen and Martin Stokhof (1982) "Semantic Analysis of WhComplements," Linguistics and Philosophy 5, 175-233. 147
Grosu, Alexander and Fred Landman (1998) "Strange Relatives of the Third Kind," Natural Language Semantics 6, 125-170. Hagstrom, Paul (1998) Decomposing Questions, Ph.D. dissertation, MIT. Hale, Ken and Chisato Kitagawa (1976-77) "A Counter to Counter Equi," Papers in Japanese Linguistics 5, 41-61. Hamblin, C. L. (1973) "Questions in Montague English," Foundations of Language 10, 41-53. Harada, Shin-Ichi (1973) "Counter Equi NP Deletion," Annual Bulletin, Research Institute of Logopedics and Phoniatrics, University of Tokyo, 7, 113-147. Haspelmath, Martin (1995) "Diachronic Sources of 'All' and 'Every'," in Emmon Bach, Eloise Jelinek, Angelika Kratzer and Barbara H. Partee (eds.), Quantification in Natural Languages, Kluwer, Dordrecht, pp. 363-382. Heim, Irene (1982) The Semantics of Definite and Indefinite Noun Phrases, Ph.D. dissertation, University of Massachusetts, Amherst. Heim, Irene (1983) "On the Projection Problem for Presuppositions," in M. Barlow, D. Flickinger and M. T. Wescoat (eds.), Proceedings of the Second West Coast Conference on Formal Linguistics, Stanford University, pp. 114-125. Heim, Irene (1990) "E-type Pronouns and Donkey Anaphora," Linguistics and Philosophy 13, 137-177. Heim, Irene (1994) "Interrogative Semantics and Karttunen's Semantics for Know," in Rhonna Buchalla and Anita Mittwoch (eds.), IATL 1, Akademon, Jerusalem, pp. 128-144. Heim, Irene and Angelika Kratzer (1998) Semantics in Generative Grammar, Blackwell, Oxford. Hermon, Gabriella (1984) Syntactic Modularity, Foris, Dordrecht. Hoji, Hajime (1985) Logical Form Constraints and Configurational Structures in Japanese, Ph.D. dissertation, University of Washington. Hoshi, Koji (1995) Structural and Interpretive Aspects of Head-Internal and HeadExternal Relative Clauses, Ph.D. dissertation, University of Rochester. Hoshi, Koji (1996) "On the Necessity of a "Cooperian" Treatment of E-type Pronouns: Evidence from the Head-Internal Relative Clause in Japanese," in Marek Przezdziecki and Lindsay Whaley (eds.), ESCOL '95, Proceedings of the 12th Eastern States Conference on Linguistics, CLC Publications, Cornell University, pp. 77-88. Huang, C.-T. James (1982) Logical Relations in Chinese and the Theory of Grammar, Ph.D. dissertation, MIT.
148
Ito, Junko (1986) "Head-Movement at LF and PF," in Nobuko Hasegawa and Yoshihisa Kitagawa (eds.), Oriental Linguistics, University of Massachusetts Occasional Papers in Linguistics 11, pp. 109-138. Kamp, Hans (1981) "A Theory of Truth and Semantic Representation," in J. Groenendijk, T. Janssen, and M. Stokhof (eds.) Formal Methods in the Study of Language. Mathematical Centre Tracts 135, Mathematisch Centrum, Amsterdam, pp. 277322. Reprinted in J. Groenendijk, T. Janssen, and M. Stokhof (eds.) (1984) Truth, Interpretation and Information, Foris, Dordrecht, pp. 1-41. Karttunen, Lauri (1977) "Syntax and Semantics of Questions," Linguistics and Philosophy 1, 3-44. Kawashima, Ruriko (1994) The Structure of Noun Phrases and the Interpretation of Quantificational NPs in Japanese, Ph.D. dissertation, Cornell University. Kayne, Richard (1994) The Antisymmetry of Syntax, MIT Press, Cambridge, Mass. Kitagawa, Chisato (1996) "Pivot Independent Relatives in Japanese and the Pronominal Head," ms., University of Massachusetts, Amherst. Kitagawa, Chisato and Claudia N. G. Ross (1982) "Prenominal Modification in Chinese and Japanese," Linguistic Analysis 9, 19-53. Kratzer, Angelika (1991) "The Representation of Focus," in Arnim von Stechow and Dieter Wunderlich (eds.), Semantics: An International Handbook of Contemporary Research, Walter de Gruyter, Berlin, pp. 825-834. Kratzer, Angelika (1998) "Scope or Pseudo-Scope? Are There Wide-Scope Indefinites?," in Susan Rothstein (ed.) Events in Grammar, Kluwer, Dordrecht, pp. 163-196. Kuroda, S.-Y. (1965) Generative Grammatical Studies in the Japanese Language, Ph.D. dissertation, MIT. Kuroda, S.-Y. (1970) "Remarks on the Notion of Subject with Reference to Words Like Also, Even or Only, Part II," Annual Bulletin, Research Institute of Logopedics and Phoniatrics, University of Tokyo, 4, 127-152. Reprinted in Papers in Japanese Linguistics 11, 121-157. Reprinted in Kuroda (1992). Kuroda, S.-Y. (1974) "Pivot-Independent Relativization in Japanese I," Papers in Japanese Linguistics 3, 59-93. Reprinted in Kuroda (1992). Kuroda, S.-Y. (1975-76) "Pivot-Independent Relativization in Japanese II," Papers in Japanese Linguistics 4, 85-96. Reprinted in Kuroda (1992). Kuroda, S.-Y. (1976-77) "Pivot-Independent Relativization in Japanese III: Types of Japanese Relatives," Papers in Japanese Linguistics 5, 157-179. Reprinted in Kuroda (1992). Kuroda, S.-Y. (1978) "Case Marking, Canonical Sentence Patterns and Counter Equi in Japanese," in John Hinds and Irwin Howard (eds.), Problems in Japanese Syntax and Semantics, Kaitakusha, Tokyo, pp. 30-51. Reprinted in Kuroda (1992).
149
Kuroda, S.-Y. (1992) Japanese Syntax and Semantics: Collected Papers, Kluwer, Dordrecht. Kuroda, S.-Y. (1999) "Syubu naizai kankeisetu [Internally Headed Relative Clauses]," in S.-Y. Kuroda and Masaru Nakamura (eds.), Kotoba-no Kaku-to Syuuen: Nihongoto Eigo-no Aida [Language, Its Core and Periphery: Explorations in English and Japanese], Kuroshio Publishers, Tokyo, pp. 27-103. Lakoff, George (1986) "Frame Semantic Control of the Coordinate Structure Constraint," in CLS 22, Part 2, Papers from the Parasession on Pragmatics and Grammatical Theory at the 22nd Regional Meeting, Chicago Linguistic Society, pp. 152-167. Larson, Richard (1985) "On the Syntax of Disjunction Scope," Natural Language and Linguistic Theory 3, 217-264. Lasnik, Howard and Mamoru Saito (1992) Move α : Conditions on its Application and Output, MIT Press, Cambridge, Mass. Lefebvre, Claire and Pieter Muysken (1988) Mixed Categories: Nominalizations in Quechua, Kluwer, Dordrecht. Lewis, David (1975) "Adverbs of Quantification," in Edward Keenan (ed.), Formal Semantics of Natural Languages, Cambridge University Press, Cambridge, pp. 315. Lin, Jo-wang (1996) Polarity Licensing and Wh-phrase Quantification in Chinese, Ph.D. dissertation, University of Massachusetts, Amherst. Link, Godehard (1983) "The Logical Analysis of Plurals and Mass Terms: A LatticeTheoretical Approach," in Rainer Bäuerle, Christoph Schwarze and Arnim von Stechow (eds.), Meaning, Use, and Interpretation of Language, Walter de Gruyter, Berlin, pp. 302-323. Maki, Hideki (1995) The Syntax of Particles, Ph.D. dissertation, University of Connecticut. Matthewson, Lisa (1999) "On the Interpretation of Wide-Scope Indefinites," Natural Language Semantics 7, 79-134. May, Robert (1985) Logical Form: Its Structure and Derivation, MIT Press, Cambridge, Mass. Mihara, Ken-ichi (1994a) "Iwayuru syuyoobu naizaigata kankeisetu-ni tuite [On the SoCalled Head-Internal Relative Clauses]," Nihongogaku 13, 80-92. Mihara, Ken-ichi (1994b) "On the Proper Treatment of Postpositions in Japanese," in Masaru Nakamura (ed.), Current Topics in English and Japanese, Hituzi Syobo, Tokyo, pp. 131-150. Murasugi, Keiko (1991) Noun Phrases in Japanese and English: A Study in Syntax, Learnability and Acquisition, Ph.D. dissertation, University of Connecticut.
150
Murasugi, Keiko (1994) "Head-Internal Relative Clauses as Adjunct Pure Complex NPs," in S. Chiba et al. (eds.), Synchronic and Diachronic Approaches to Language: A Festschrift for Toshio Nakao on the Occasion of His Sixtieth Birthday, Liber Press, Tokyo, pp. 425-437. Nishigauchi, Taisuke (1986) Quantification in Syntax, Ph.D. dissertation, University of Massachusetts, Amherst. Nishigauchi, Taisuke (1990) Quantification in the Theory of Grammar, Kluwer, Dordrecht. Obenauer, Hans-Georg (1983) "Une Quantification Non Canonique: La 'Quantification à Distance'," Langue Française 58, 66-88. Ochi, Masao (1998) "Move or Attract?," in Emily Curtis, James Lyle, and Gabriel Webster (eds.), Proceedings of the Sixteenth West Coast Conference on Formal Linguistics, CSLI, Stanford University, pp. 319-333. Ohno, Yutaka (1989) "Mo," in Emmon Bach, Angelika Kratzer and Barbara H. Partee (eds.), Papers on Quantification, NSF Grant Report, Department of Linguistics, University of Massachusetts, Amherst, GLSA, University of Massachusetts, Amherst, pp. 224-250. Ohno, Yutaka (1991) "Arguments against Unselective Binding in Korean," in Susumu Kuno et al. (eds.), Harvard Studies in Korean Linguistics IV, Hanshin Publishing, Seoul, pp. 553-562. Ortiz de Urbina, Jon (1989) Parameters in the Grammar of Basque, Foris, Dordrecht. Pesetsky, David (1987) "Wh-in-Situ: Movement and Unselective Binding," in Eric Reuland and Alice ter Meulen (eds.), The Representation of (In)definiteness, MIT Press, Cambridge, Mass., pp. 98-129. Quine, Willard Van Orman (1960) Word and Object, MIT Press, Cambridge, Mass. Ramchand, Gillian C. (1997) "Questions, Polarity and Alternative Semantics," in Kiyomi Kusumoto (ed.), Proceedings of the North East Linguistic Society 27, GLSA, University of Massachusetts, Amherst, pp. 383-396. Reinhart, Tanya (1997) "Quantifier Scope: How Labor is Divided between QR and Choice Functions," Linguistics and Philosophy 20, 335-397. Reinhart, Tanya (1998) "Wh-in-Situ in the Framework of the Minimalist Program," Natural Language Semantics 6, 29-56. Richards, Norvin (1997) What Moves Where When in Which Language? Ph.D. dissertation, MIT. Richards, Norvin (2000) "An Island Effect in Japanese," Journal of East Asian Linguistics 9, 187-205. Riemsdijk, Henk van (1984) "On Pied-Piped Infinitives in German Relative Clauses," in Jindrich Toman (ed.), Studies in German Grammar, Foris, Dordrecht, pp. 165-192.
151
Rooth, Mats (1985) Association with Focus, Ph.D. dissertation, University of Massachusetts, Amherst. Rooth, Mats (1996) "Focus," in Shalom Lappin (ed.), The Handbook of Contemporary Semantic Theory, Blackwell, Oxford, pp. 271-297. Rooth, Mats and Barbara Partee (1982) "Conjunction, Type Ambiguity and Wide Scope Or," in D. Flickinger, M. Macken and N. Wiegand (eds.), Proceedings of the First West Coast Conference on Formal Linguistics, Stanford University, pp. 353-362. Ross, John Robert (1967) Constraints on Variables in Syntax, Ph.D. dissertation, MIT. Russell, Bertrand (1905) "On Denoting," Mind 14, 479-493. Schwarzschild, Roger (1996) Pluralities, Kluwer, Dordrecht. Selkirk, Elisabeth (1982) The Syntax of Words, MIT Press, Cambridge, Mass. Sells, Peter (1986) "Coreference and Bound Anaphora: A Restatement of the Facts," in Stephen Berman, Jae-Woong Choe and Joyce McDonough (eds.), Proceedings of the North East Linguistic Society 16, GLSA, University of Massachusetts, Amherst, pp. 434-446. Sharvit, Yael (1998) "Possessive Wh-Expressions and Reconstruction," in Pius N. Tamanji and Kiyomi Kusumoto (eds.), Proceedings of the North East Linguistic Society 28, GLSA, University of Massachusetts, Amherst, pp. 409-423. Shimoyama, Junko (1999) "Internally Headed Relative Clauses in Japanese and E-Type Anaphora," Journal of East Asian Linguistics 8, 147-182. Srivastav, Veneeta (1991) "The Syntax and Semantics of Correlatives," Natural Language and Linguistic Theory 9, 637-686. von Stechow, Arnim (1993) "Die Aufgaben der Syntax," in Joachim Jacobs, Arnim von Stechow, Wolfgang Sternefeld and Theo Vennemann (eds.), Syntax: An International Handbook of Contemporary Research, Walter de Gruyter, Berlin, pp. 1-88. von Stechow, Arnim (1996) "Against LF Pied-Piping," Natural Language Semantics 4, 57-110. Stump, Gregory T. (1985) The Semantic Variability of Absolute Constructions, Reidel, Dordrecht. Sugahara, Mariko (1998) "The Syntax of the Japanese Noda-Sentence," ms. University of Massachusetts, Amherst. Takahashi, Daiko (1999) "Determiner Raising and Scope Shift," ms., Tohoku University. Tanaka, Hidekazu (1998) Conditions on Logical Form Derivations and Representations, Ph.D. dissertation, McGill University. Tanaka, Hidekazu (1999) "LF Wh-islands and the Minimal Scope Principle," Natural Language and Linguistic Theory 17, 371-402. 152
Tonoike, Shigeo (1992) "Operator Movements in Japanese," Meiji Gakuin Ronsoo 84, 79-142. Toyoshima, Takashi (1996) "LF Subjacency and Stationary Wh-in-Situ," ms., Cornell University. Tsai, Wei-tien (1994) On Economizing the Theory of A-Bar Dependencies, Ph.D. dissertation, MIT. Vergnaud, Jean Roger (1974) French Relative Clauses, Ph.D. dissertation, MIT. Watanabe, Akira (1992a) "Wh-in-Situ, Subjacency, and Chain Formation," MIT Occasional Papers in Linguistics 2. Watanabe, Akira (1992b) "Subjacency and S-structure Movement of Wh-in-Situ," Journal of East Asian Linguistics 1, 255-291. Webelhuth, Gert (1992) Principles and Parameters of Syntactic Saturation, Oxford University Press, Oxford. Whitman, John (1998) "Gojun-to kukoozoo [Word Order and Phrase Structure]," Part II in Takezawa Koichi and John Whitman, Kaku-to Gojun-to Toogokoozoo [Case, Word Order and Syntactic Structure], Kenkyuusya, Tokyo. Williamson, Janis S. (1987) "An Indefiniteness Restriction for Relative Clauses in Lakhota," in Eric Reuland and Alice ter Meulen (eds.), The Representation of (In)definiteness, MIT Press, Cambridge, Mass., pp. 168-190. Wold, Dag E. (1996) "Long Distance Selective Binding: The Case of Focus," in Teresa Galloway and Justin Spence (eds.), Proceedings from Semantics and Linguistic Theory VI, CLC Publications, Cornell University, pp. 311-328. Yamashina, Miyuki (in progress) Ph.D. dissertation, University of Connecticut.
153
![free [download] midnight in broad daylight: a japanese ...](https://p.pdfkul.com/img/300x300/free-download-midnight-in-broad-daylight-a-japanes_5a2817701723dd073e6dbef5.jpg)
