UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

An interface approach to topic/focus structure Uechi, Akihiko 1998

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata

Download

Media
831-ubc_1998-346374.pdf [ 7.94MB ]
Metadata
JSON: 831-1.0099375.json
JSON-LD: 831-1.0099375-ld.json
RDF/XML (Pretty): 831-1.0099375-rdf.xml
RDF/JSON: 831-1.0099375-rdf.json
Turtle: 831-1.0099375-turtle.txt
N-Triples: 831-1.0099375-rdf-ntriples.txt
Original Record: 831-1.0099375-source.json
Full Text
831-1.0099375-fulltext.txt
Citation
831-1.0099375.ris

Full Text

A n Interface Approach to Topic/Focus Structure  by A K I H I K O UECHI B.A., Kansai University of Foreign Studies M . A . , Kansai University of Foreign Studies M . A . , The University of Northern Iowa  A THESIS SUBMITTED IN P A R T I A L F U L F I L L M E N T OF T H E REQUIREMENTS FOR T H E D E G R E E OF DOCTOR OF PHILOSOPHY in T H E F A C U L T Y OF G R A D U A T E STUDIES (Department of Linguistics) We accept this thesis as conforming to the required standard  T H E UNIVERSITY OF BRITISH C O L U M B I A August 1998 © Akihiko Uechi, 1998  In presenting this thesis in partial fulfilment of the requirements for an advanced degree at the University of British Columbia, 1 agree that the Library shall make it freely available for reference and study. I further agree that permission for extensive copying of this thesis for scholarly purposes may be granted by the head of my department or by his or her representatives.  It is understood  that copying or  publication of this thesis for financial gain shall not be allowed without my written permission.  Department of  Z A ^ / O T W  The University of British Columbia Vancouver, Canada  Date  DE-6 (2/88)  s&^j*  A  *  ii ABSTRACT An Interface Approach to Topic/Focus Structure in Japanese  This dissertation examines how information structure is represented in the different components of Japanese grammar. Each chapter investigates how topic/comment structure and focus/background structure are manifested in a given component of the grammar; that is, in phonology, syntax, and discourse/semantics. In chapter 2,1 investigate the interaction of syntax and prosody. I introduce the End-based analysis proposed by Selkirk &Tateishi (1988, 91) and propose certain revisions in order to accommodate a broader range of empirical facts than has previously been discussed in the literature. I show that presentational focus and contrastive focus are phonologically distinct and that contrastive focus overrides phonological phrasing. I then demonstrate that the system of focus projection proposed for stress languages such as English and German applies to Japanese. I show that focus projection takes place in the syntactic component, prior to mapping into prosodic structure, from the head of the phrase to its sisters (cf. Rochemont 1996) In chapter 3, I establish the discourse function of wa-marking in Japanese, extending Buring's analysis of sentence topics. I demonstrate that the discourse function of wa-marking parallels that of the L H * contour (B-accent) in English, and claim that wa-marking is equivalent to T-marking in the model of Biiring (1998). As such, wa-marking can be viewed as one of the discourse strategies available in Japanese for ensuring that a given assertion is congruent - that is, appropriate - to the question under discussion. In chapter 4,1 investigate the representation of information structure in syntax. I propose a phrase structure for Japanese based on a universal hierarchy of functional categories. I then divide the syntactic structure of Japanese into two major syntactic domains, which I call the topic domain and the comment domain. I show that both sentence topics and contrastively focused constituents must not be inside IP, which is identified as the comment domain. I further argue that subjects outside IP must be wa-marked unless contrastively focused. I conclude that syntactic structure is  HI  discourse-configurationally based. To conclude, I discuss the architecture of the grammar that emerges from the proposal defended in each chapter of this dissertation.  Table of Contents  Abstract  ii  Table  iv  of Contents  Chapter 1: Introduction  1.1. Goals : 1.2. Overview of this dissertation 1.3. What is information structure? 1.4. Theoretical background Chapter 2: Prosody-Syntax/Focus  1 2 3 6  Interface in Japanese  2.1. Introduction ; 2.2. A n outline of the syntax-prosody mapping in Japanese 2.2.1. Basic Observations 2.2.2. The End-based Analysis (Selkirk & Tateishi 1988, 91) 2.3. Extension of the End-based Analysis (Revised End-based Analysis) 2.3.1. M a P as a Metrical Boost (MB) initiator 2.3.2. Minimization of structure 2.3.2.1. VP-subjects vs. IP subjects 2.3.2.2. Relative clauses 2.3.2.3. Coordination 2.3.3. Summary 2.4. Interaction of phonological phrasing with focus 2.4.1. Projection Phrases 2.4.2. Focus projection in English 2.4.3. Focus projection and phonological phrasing in Japanese 2.4.4. ProP as the domain of total downstep and the domain of focus projection... 2.4.5. Semantic and phonological restrictions on ProPs 2.4.6. Summary 2.5. Prosody of contrastive focus 2.5.1. Phonological phrasing of contrastive focus 2.5.2. Phonological phrasing of presentational focus vs. contrastive focus 2.5.3. Phonological phrasing and heavy stress WA-marking 2.5.4. Summary 2.6. Phonological phrasing of wa-marking without heavy stress 2.7. Summary  11 12 12 16 21 26 30 30 36 40 41 43 43 46 49 52 56 59 60 60 65 69 73 73 78  Chapter 3: Semantics-Discourse Interface: The Semantics and the Discourse Function of wa-marking in Japanese  3.1. Introduction 81 3.2. What can be wa-marked? 81 3.2.1. Wa-marked elements referring to the antecedent in the previous discourse... 82 3.2.2. Wa-marked elements not referring to an antecedent in the previous discpurse.85 3.3. A semantic based analysis of wa-marking using layered focus 98 3.3.1. Biiring (1997, 98) 98 3.3.1.1. Focus and alternative semantics 98 3.3.1.2. T-marking 100  3.3.2 Layered focus and Birring (1998) 3.4. Parallelism between wa-marking in Japanese and B-accents in English 3.4.1. Application of the proposal 3.4.2. Major properties of wa-marked elements 3.4.3. T w o types of B-accents? 3.5. Extension of the analysis 3.5.1. Wa-marking and universal quantifiers , 3.5.2. Scope inversion and stressed WA-marking 3.5.3. Multiple wa-marking and contrastiveness 3.6. Alternative analyses 3.6.1. Old/new information and wa-marking 3.6.2. Definiteness/specificity and wa-marking 3.7. o-marking vs. wa-marking 3.8. Sentence topics without wa-marking 3.9. Summary Appendix: Kuroda's mini-topics  104 109 110 115 117 119 119 121 124 126 126 128 132 133 135 135  Chapter 4: Toward Syntax-Information Structure Mapping in Japanese  4.1. Introduction 4.2. Japanese phrase structure and the positions of subjects 4.2.1. Japanese phrase structure 4.2.2. Subjects of individual-level predicates 4.3. Toward a syntax-information structure mapping 4.3.1.Syntax-semantics mapping 4.3.1.1. Diesing (1992) 4.3.1.2. Problems with syntax-semantics mapping 4.3.2. Syntax-information structure mapping 4.3.2.1. Possible sites for topics 4.3.2.2. Topic vs. comment domain and wa-marking 4.3.2.3. wa-marking as the semantic subject of predication 4.4. The syntactic position of contrastively focused subjects 4.5. A comparison with Kuroda's thetic/categorical distinctions 4.6. Summary and implications Appendix: Ga-marking as the default case marker  138 138 139 147 152 152 153 155 156 157 160 163 164 166 169 171  Chapter 5 : Conclusion: O n Interface Issues  5.1. What does the analysis contribute to our understanding of Japanese grammar?  174  References  177  Chapter 1: Introduction  1.1.  Goals  This dissertation discusses how information structure is represented in the different components of Japanese grammar. I examine how topic/comment structure and focus/background structure (FBS) are manifested in each component of the grammar. The study of information structure leads me to investigate the interactions between different components of the grammar, in particular phonology, syntax, and discourse/semantics. I examine how discourse contexts restrict the distribution of topicalized and focused phrases and how they are represented in phonology and syntax. The study of Japanese grammar has been conducted along roughly two different lines: the "nihongogaku" approach (traditional Japanese linguistics) and the generative approach. The former has attempted to establish a descriptive grammar of the Japanese language whereas the latter, which was introduced about thirty years ago, has attempted to establish the status of Japanese within a theory of universal grammar (UG). Both approaches have their own strengths and weaknesses. For instance, proponents of the nihongogaku approach base their arguments on a detailed description of Japanese data. Nevertheless, such studies have not attracted the attention of generative linguists; the fact that most of the nihongogaku literature has been only available in Japanese has limited its access to non-Japanese linguists. On the other hand, the generative approach, since its rise three decades ago, has made significant contributions in revealing the structure of Japanese. However, there has been a tendency to limit empirical inquiry to those phenomena which are explicable within existing theories of UG, as established primarily on the basis of data from a limited number of languages, most notably English. Whichever approach one pursues, there is a topic which has never failed to attract the attention of native speakers of Japanese who study linguistics: the distribution of wa-marking (often known as a topic marker).  2  In Japanese, the information status of constituents is often overtly (that is, prosodically, morphologically, and syntactically) marked. The large interest in wa-marking is reflected by a robust literature in the nihongogaku tradition. In this approach, it is commonly assumed that an utterance consists of a modality and a core proposition as schematized in (1). (1)  Modality may be interpreted as a psychological attitude on the part of the speaker toward an event, an action, or a state (cf. Nakau 1976). As has been pointed out by traditional Japanese linguists, various modal phenomena in Japanese are syntactically manifested. Within this tradition, wa-marked elements belong to the domain of modality, rather than to the domain of the core proposition. Japanese generative linguists, on the other hand, have been concerned with establishing grammar mainly at the core proposition level. This dissertation aims at providing a bridge between the two approaches. Until recendy, most Japanese grammarians have investigated grammatical phenomena (e.g. the distribution of wa-marking) within the limits of a single grammatical component/module (e.g., either the phonology or the syntax of wa-marking). However, recently developed theories of information structure within the generative tradition have now made it possible to investigate phenomena such as wa-marking in Japanese from a cross modular perspective (cf. Biiring 1997, Erteschik-Shir 1996). I demonstrate that the interface approach to grammar provides new insights into the phenomena, which have previously been analyzed within the limits of a single component, and sheds new light on the architecture of grammar, and, in particular, of Japanese grammar.  3  1.2.  Overview of this dissertation  This dissertation consists of three substantive chapters, dealing respectively with the different components of Japanese grammar and their interaction with information structure. Thus, the dissertation is unified under one theme: the representation of topic/focus structure in each component of the grammar. In Chapter 2, the interaction of syntax and prosody is investigated. First, I introduce the End-based analysis proposed by Selkirk &Tateishi (1988, 91) and propose certain revisions in order to accommodate a broader range of empirical facts than has previously been discussed in the literature. I show that presentational focus and contrastive focus are phonologically distinct and that contrastive focus overrides phonological phrasing. I then demonstrate that the system of focus projection proposed for stress languages such as English and German applies to Japanese. In Chapter 3, I establish the discourse function of wa-marking in Japanese, extending Buring's analysis of sentence topics. The major proposal defended in chapter 3 is that the discourse function of wa-marking parallels that of the L H * contour (B-accent) in English. In particular, wa-marking is equivalent to T-marking in the model of Biiring (1998). A s such, wa-marking can be viewed as one of the discourse strategies available in Japanese for ensuring that a given assertion is congruent - that is, appropriate - to the question under discussion. In Chapter 4, the representation of information structure in syntax is investigated. I propose a hierarchy of functional categories. I then divide the syntactic structure of Japanese into two major domains, which I call the TOPIC domain and the comment domain. I further argue that both sentence topics and contrastively focused constituents must be outside IP, and that subjects outside IP (which corresponds to the comment domain) must be wa-marked unless contrastively focused. In Chapter 5 , I summarize the major proposals of this dissertation: how topic-focus structure manifests itself in each component of grammar. Finally, I discuss the architecture of the grammar in lights of the results of each chapter.  4 1.3. What is information structure?  A l l languages have mechanisms for introducing new information into a discourse and for keeping track of or updating old/familiar information. It is generally assumed that focused constituents represent new information, whereas sentence topics represent old/familiar information. In order to facilitate the ensuing discussion, I will begin by providing operational definitions for topic and focus; these will enable us to identify the relevant constituents.  Topic For concreteness, I tentatively identify topic as a constituent which satisfies the topic test proposed by Reinhart (1981), as shown in (2).1 (2)  Topic test (Reinhart 1981)2 Speaker A : Tell me about x Speaker B: x (x is topic)  In Japanese, wa-marked elements are often identified as sentence topics, as the topic test in (3) illustrates. (3)  . Speaker A : Tell me about Taroo. Speaker B: Taroo-w a nihon-kara ki-ta kookanryuugakusee des-u. Taroo-wa Japan-from come-past exchange student be-pres "Taro is an exchange student from Japan."  Focus I identify focus as a constituent which answers a wh-question or an element which constitutes a negative contrastive adjunct (cf. Selkirk 1995, Rochemont 1996).3 'The definition of topic will be revised in 3.3. Reinhart (1981) formally defines "aboutness" in the following manner.  2  "To say that a sentence S uttered in a context C is about a;, i.e., that the pair <a;, <|» of P P A ^ (possible pragmatic assertion) is selected in C, is to say, first, that, if possible, the proposition <|) expressed in S will be assessed by the hearer in C with respect to the subset of propostions already listed in the context set under a ; and second, that if <p is not rejected it will be added to the context set under the exntry of a ; ." ^ote the distribution between contrastive vs. presentational focus is not equivalent to narrow vs. wide focus, as pointed out by Manfred Krifka (p.c).  5  (4)  Context: What did John study yesterday? John-wa [gengogaku-o]F benkyoosi-ta. John-wa linguistics-ace study-past "John studied [LINGUISTICS] F ."  (5)  John-wa kinoo [sinrigaku]F de naku, [gengogaku-o]F benkyoosi-ta. John-wa yesterday psychology pro-verb not, psychology-acc study-past "John studied [LINGUISTICS] F yesterday, not [PSYCHOLOGY] F ." I further distinguish two types of focus: presentational focus and contrastive focus. Let us  consider the following two sentences which contain no topics.  (6)  Context: What is happening over there? Inu-ga oyoide-i-ru. (presentational focus) dog-nom swimming-be-pres A DOG is SWIMMING.  (7)  Context: What is swimming over there? Inu-ga oyoide-i-ru. (contrastive focus) dog-nom swirnming-be-pres A DOG is swimming.  The sentences in (6) and (7) are topicless sentences. If all the elements in a topicless sentence are focused as in (6), the constituents are presentationally focused. In contrast, if a topicless sentence contains focused elements as well as non-focused elements, the focused constituents are contrastively focused. Note that Presentational focus, which is typically introduced as an answer to the questions such as "What has happened?" and "What is the matter?", can introduce a new referent into discourse while contrastive focus cannot. Focus projection Following Jackendoff (1972), I assume that focus on a syntactic constituent is the syntactic, semantic and phonological reflex of an abstract syntactic feature [F] . 4 In (4) above, the focus is "linguistics", which corresponds to the wh-element in the question, and in (5), the focus is ''That is, the overt reflex of a focus feature within a sentence is visible as the phonological, morphological or syntactic marking of prominence on a constituent.  6 "linguistics", which is contrasted with the constituent "psychology" in the negative contrastive adjunct.5 (8) a.  Context: What did John study? a' John studied [LINGUISTICS] F O C .  b. Context: What did John do? b* John [studied LINGUISTICS] F O C . The answer to the wh-question in (8) determines the focus domain (Foe); as shown in (8b), every constituent in the focus domain need not be pitch accented.6 In other words, without the previous context in (8a, b), the answers in (8a', b') are structurally ambiguous. To account for this phenomenon, Selkirk (1984, 1995) and Rochemont (1986, 1996) propose a theory of focus projection.71 assume the following definition of focus projection based on that of Rochemont. (9)  The syntactic feature []F assigned on the head of an argument may optionally project to the head of an element which is in an internal argument relation (as far as the internal argument is focused).  Given (9)* (8b) may be represented as in (10). FJ^ indicates projected F-marking. (10)  1.4.  Context: What did John do? John [[studied]^ [LINGUISTICS] F ] F 0 C Theoretical  background  This dissertation investigates how information structure interacts with the different components of grammar. Adopting the principles and parameters approach (cf. Chomsky & Lasnik 1995), I assume that the grammar consists of the following three components: a transformational component, a PF (phonetic)-component, and L F (logical form)-component.  ^ h e c a p i t a l l e t t e r s i n d i c a t e the e l e m e n t s are p i t c h a c c e n t e d i n E n g l i s h . 6  J a c k e n d o f f ( 1 9 7 2 ) m a k e s stress r u l e s d e p e n d e n t o n f o c u s w h e r e a s S e l k i r k ( 1 9 8 4 ) m a d e f o c u s a s s i g n m e n t d e p e n d e n t  o n t h e o u t p u t o f s t r e s s r u l e . I w i l l p u r s u e the f o r m e r a p p r o a c h . 7  N o t e t h a t w i t h i n V P , a p i t c h a c c e n t f a l l s o n the d i r e c t o b j e c t .  7 D-structure  In this model, syntactic structure, which is derived from the transformational component by mapping the level of D-structure onto S-structure is, on the one hand, interpreted in the L F component, (that is, the level of representation in the grammar which feeds the semantic component) and on the other hand, is phonetically or phonologically interpreted in the PF-component. Syntactic assumptions about Japanese phrase structure I adopt an X-bar model of phrase structure which exploits the system of Extended Projection in Grimshaw (1991). Case assignment  in Japanese  With regard to case assignment, I assume that there are two different mechanisms of case-marking: structural case and inherent case.8 There are two structural cases in Japanese. The subject of transitive verbs and unergative verbs receives nominative case (-ga) at Spec IP (the positions of different types of subjects will be explored in detail in chapter 4) while non-oblique objects receive accusative case (-o), internally to VP.  In chapter 4,1 show that there is another mechanism for case-marking in Japanese, i.e., default case-marking. This is used when structural case-marking is not available, and inherent case is not an option, either.  8  8 (11) Structural case assignment 910  chocolate John-ga chocolate-o tabe-ta/ru John-nom chocolate-acc eat-past/pres "John ate/eat a chocolate." In contrast, constituents base-generated in adjunct positions receive inherent case, as illustrated in (12).  (12) Inherent case  DP chocolate  K -o  John-ga Megumi-ni Tyokoreeto-o morat-ta/mora-u. John-nom Megumi-from Chocolate-acc receive-past/receive-pres11 "John received/receives a chocolate from Megumi." We see that the DP "Megumi" in (12), which is base-generated in a position where it cannot receive structural case, is case-marked by the inherent case marker "-ni". Inherent case-marked^constituents 'Japanese is a headfinallanguage. 10  I make the standard assumption that INFL hosts tense.  It has been argued that the morphemes "ta" and "(r)u" in Japanese are aspectual markings (as opposed to tense markings) to describe completion vs. incompletion of a situation respectively. For ease of exposition, I will merely identify "ta" and "(r)u" as "past" and "present" tense markers. I returns to this topic in chapter 4. n  9  may either be indirect objects or adjuncts. Turning to case assignment inside DPs, I assume that genitive case "-no" is structurally assigned by D. Notice that "-no" does not only act as a genitive marker but rather case marks any  toti  -no  baisyuu  John-no toti-no baisyuu John-pit land-no purchase "purchase of a land by John" Argument  structure  I make the standard assumption that external theta-roles (e.g., agent) are compositionally assigned by the V P whereas internal theta-roles (e.g., theme) are assigned VP-internally.  Phonological assumptions I assume that phonological domains are organized according to the following prosodic hierarchy (cf. Selkirkl986)  12  "no" may also case-mark a PP inside a DP.  10 (15)  utterance international phrase phonological phrase prosodic word foot syllable  Selkirk & Tateishi (1988, 91) further divide the phonological phrase into two levels of phrasing: a Minor phonological phrase (MiP), which dominates PWds and a Major phonological phrase (MaP) which dominate MiPs. I assume that the connection between syntactic structure and prosodic structure is established by a mapping algorithm. In particular, I will assume the End-based theory based of Selkirk & Tateishi (1988, 91), who propose that the left edge of an X P identifies a MaP, and a lexical X° corresponds to a PWd. Semantic assumptions I will assume that there is a structurally mediated mapping between syntax and information structure. In other words, Japanese syntax is organized in a discourse configurational manner, and syntactic structure is mapped onto the semantics representation using a syntax-information structure mapping operation along the lines of those suggested in the work of Diesing (1988) and Jager (1997). The analysis of wa-marking presented here is based on the theoretical framework of Biiring (1997, 98) and Rooth (1985, 92).  11  Chapter 2: Prosody-Syntax/Focus Interface in Japanese  2.1.  Introduction.  The goal of this chapter is to investigate how focus is represented phonologically in Japanese. I claim that phonological phrasing is determined not only by phonological and semantic constraints but also by its focus status (whether an element is focused or not, and if focused, contrastively focused or presentationally focused) and syntax , i.e., phrase structure and argument structure. This chapter is divided into two parts. In the first half, I will establish the syntax-prosody mapping in Japanese in terms of the revised End-based Analysis (REB). In particular, I modify the End-based Analysis (EB) proposed by Selkirk and Tateishi (1988, 1991) by proposing two addtional constraints: the Branching Constraint and the Minimization of Structures. In the second half, I investigate interactions of phonological phrasing and focus. I motivate two new prosodic levels called Projection Phrases (ProPs) and contrastive focus phrases (CFPs). I argue that ProPs are sensitive to argument structure and that CFPs are induced by contrastive focus, which overrides MaP and ProP boundaries. Adopting the claim by Truckenbrodt (1995) that the representation of Metrical Structure and the representation of prosodic structure are the same, I further propose the following prosodic hierarchy, adding the level of Projection Phrases (ProPs) between Major Phonological Phrases (MaPs) and Minor Phonological Phrases (MiPs) and the level of Contrastive Focus Phrases (CFPs) above MaPs. Utterances Intonational Phrases (IntoPs) Contrastive Focus Phrases (CFPs) Major Phonological Phrases (MaPs) (Focus) Projection Phrases (ProPs) Minor Phonological Phrases (MiPs) Prosodic Words (PWds) Differences and similarities between English and Japanese prosody will be discussed in this  12 chapter as they become relevant.  PARTI  2.2. A n o u t l i n e o f the s y n t a x - p r o s o d y m a p p i n g i n J a p a n e s e  In this section, I introduce the End-based Analysis (EB) proposed by Selkirk and Tateishi (1988, 1991). Before doing so, however, I will first provide some background assumptions about Japanese prosody, including a distinction of downstep and initial lowering.  2.2.1. B a s i c  observations  McCawley (1968), Selkirk & Tateishi (1988, 91), and others claim that words in Japanese are either Accented (HL) or Unaccented ( H H ) . 1 2 (1) a. Accented ha'si-ga "chopsticks-nom" HL L hasi'-ga "bridges-nom"3 LH L  'Selkirk (1996) claims that in Japanese, the location of accents is lexically specified in nouns whereas the location of accents is on the penultimate syllables in verbs and adjectives. In Tokyo Japanese, the number of accent patterns equals the number of syllables which the word contains plus one (cf. Haraguchi 1977). As shown in (1), for instance, if the word contains two syllables, three variations are possible (i.e. two accented ones and one unaccented one). The first two are accented where a pitch fall is involved, whereas the last one is unaccented where no pitch fall occurs when a particle follows, (accents or pitch falls, are marked by the diacritic ['] throughout this chapter.)  2  Haraguchi (1977) proposes that the word "hasi" (bridges) has a floating L (marked by L) tone at the end, which is  3  associated to the following particle. hasi "bridge" --> hasi-ga "bridge-nom" LHL  LHL  13  b. Unaccented hasi-ga "edges-nom"4 LH H As we can see, an accented word has a sharp pitch fall from H to L while an unaccented word involves no pitch fall. The following data from Selkirk & Tateishi (1991) shows that f0 contours in sentences (sequences of accented/unaccented) seem to reflect the structural relations between words. Consider the following examples adopted from Selkirk & Tateishi, which are uttered in an out-of-blue context (that is, with presentational focus). (2) a  Aoyama-gen Mr. Yamaguchi-gen sister-in-law-nom absent "We cannot find the sister-in-law of Mr. Yamaguchi from Aoyama." (i.e., Mr. Yamaguchi from Aoyama's sister-in-law)  Aoyama-noY amaguchi-noaniyome-ga rnai Aoyama-gen Mr. Yamaguchi-gen sister-in-law-nom absent "We cannot find the sister-in-law of Mr. Yamaguchi from Aoyama." (i.e., Mr. Yamaguchi from Aoyama's sister-in-law) Among (spread) H toned mora at the level of PWd, the rightmost one in unaccented words or the one just before the accent is most prominent, (cf. Pierrehumbert & Beckman 1988)  4  Unaccented  Accented H  HL  I  I  V V V V -ga  V V V V -ga  14  NI  Aoyama-no Yamaguchi-no aniydme-ga inai (right-branching) Aoyama-gen Mr. Yamaguchi-gen sister-in-law-nom absent "We cannot find the sister-in-law of Mr. Yamaguchi from Aoyama."  Aoyama-no Yarfiaguchi-no aniyome-ga inai Aoyama-gen Mr. Yamaguchi-gen sister-in-law-nom absent "We cannot find Mr. Yamaguchi's sister-in-law from Aoyama. (3)  a  Oomiya-ho Inayama-no yuujin-ga inai (left-branching) Oomiya-gen Inayama-gen ffiend-nom absent "We cannot find the friend of Mr. Inayama from Oomiya." (i.e., Mr. Inayama from Oomiya's friend)  Oomiya-no Inayama-no yuujin-ga ina Oomiya-gen Inayama-gen ffiend-nom absent "We cannot find thefriendof Mr. Inayama from Oomiya." (i.e., Mr. Inayama from Oomiya's friend)  15 b  Oomiya-no Inayama-no yuujin-ga inai Oomiya-gen Inayama-gen friend-nom absent "We cannot find Mr. Inayama's friend from Oomiya." (2) contains a sequence of accented words whereas (3) contains a sequence of unaccented words. Although the sentence pairs contain the same words, the height of the f0 contour on the second word is different: (2b) is higher than (2a) ("Yamaguchi-no"): and (3b) is higher than (3a) ("Inayamano"). Why do the sentences in (2a, b) and in (3a, b) show different pitch contours? Consider their differences in syntactic structures: (2a, 3a) contain left-branching structures while (2b, 3b) contain right-branching structures. Let us examine more closely how phonological phrasings interact with syntactic parsing. (2) contains three accented words. The height of the accent peak gradually gets lower in the left-branching syntactic structure (2a).5 For Poser (1984), lowering is triggered by a preceding accent, which suppresses the following accent peak. Selkirk & Tateishi call this phenomenon downstep.6 Note that there are two downsteps in (2a). In contrast, in (2b), which contains a right-branching structure, downstep is only observed between the second and the third word. (The accent peak of the second word is as high as that of the first word when the second word starts the right-branching structure.)7 This suggests that downstep is blocked in right-branching structures ^he examples in (2) and (3) contain an existential verb, and thus, the subjects are located VP-internally. Some phonologists such as Poser (1984) and Williams (1989) utilize "catathesis" instead of downstep.  6  Kubozono (1993) identifies this phenomenon as metrical boost, which, he claims, is triggered by right-branching  7  16  and only occurs in left-branching structures. In the sequence of unaccented words in (3 a, b), no downstep is triggered since no accented words are contained in the utterance. We can also observe in (2a, b) that every accented word starts with a low (L) tone. Pierrehumbert & Beckman (1988) claim that there is a L boundary tone at the end of each phrase, which is itself associated with the beginning of the next phrase's delimitative rise. This is called initial lowering. 8 In a sequence of unaccented words, initial lowering is only observed utterance initially as shown in (3a, b) and at the beginning of a right-branching structure, as shown in (3b). McCawley (1968) proposes that two types of phonological phrasing are necessary above Prosodic Words (PWd) in order to apply rules of pitch assignment:  Minor Phrases (MiP) and  Major Phrases (MaP). He claims that a M i P has the property of containing at most one accent and a MaP consists of one or more MiPs. Selkirk & Tateishi (1988, 91), following McCawley's observation, identify MiPs as the domain of initial lowering, at which a delimitive rise of f0 is observed; MaPs as the domain of downstep, where the lowering of one accented word with respect to another preceding accented word takes place. (4) Major Phrase=the domain of downstep Downstep: A change in pitch register which is manifested as a marked lowering in the stretch of an utterance following an accented syllable. Minor Phrase= the domain of initial lowering Initial Lowering: A rise to a high tone from a low tone at the word edge. Selkirk & Tateishi argue that the sentences in (2a, b) and (3a, b) show different pitch contours because they differ in prosodic structures. In the next section, I will review the End-based analysis (EB) proposed by Selkirk & Tateishi (1988, 91) and examine their mapping algorithm.  2.2.2. The End-based Analysis (Selkirk & Tateishi 1988, 91) structures. Initial L is obligatory when the word starts an utterance or a new (Major) phonological phrase (MaP) - even with the word which starts with a H tone, e.g. ha'si-ga (chopsticks-nom). Pierrehumbert & Beckman (1988) claim that this is because the Utterance Phrase supplies an initial low tone at the beginning of the utterance, and the phrasal L tone of the preceding (Major) phonological phrase (MaP) provides a low tone at the beginning of the following MaP. 8  17  Assuming that the domain for downstep is the Major Phrase (MaP) and that the domain for Initial Lowering is the Minor Phrase (MiP), Selkirk & Tateishi (1988, 91) claim that the mapping between syntax and phonology can be accomplished by setting the following parameters. (5) Syntax-to Phonology Mapping Parameters Designated Category Parameter: Each p ; (prosodic) level has a designated category (DC;) in the syntax with which p ; is defined. End Parameter: Only one end of the D Q is relevant for defining p ; They demonstrate that the limits of MaPs in Japanese correspond to the left edge of syntactic maximal projections. MaPs are composed of MiPs which are composed of one or more Prosodic Words. PWds are delineated by the left edge of the minimal projection of lexical categories.9 (6) Japanese Parameter Setting Major Phrase; {left, X™ x}  Prosodic Word: {left, X l e x } 1 0  Selkirk & Tateishi assume the following two central properties of prosodic structure. (7) Exhaustive Parsing: A string is exhaustively parsed with respect to each prosodic category. Strict Layer hypothesis: A Prosodic constituent of type X° dominates constituents of type X n l ( t h e immediately There is another major approach which assumes that a prosodic category mediates between syntax and phonology: the Relation-based Approach (see Nesper & Vogel 1986). In contrast, the Direct Syntax Approach by Kaisse (1985) and Rochemont (1997) assumes that syntax is directly mapped onto phonology.  9  Selkirk claims that functional categories behave differently from lexical categories in that they play no major role in the syntax-prosody mapping. She reports that in Tokyo Japanese, a functional word loses a H tone if it is preceded by another accented word in the same phrase, whereas in the same circumstances, a lexical word does not lose its accent. However, as far as the following minimal pairs are concerned, the distinctions are not so clear. 10  Lexical-Functional Yonda-keredo read but "(Dread, but..."  atui-node hot because "because (it is) hot."  atui-keredo hot but "(It is hot,) but..  tukanda-noni caught but "Although (I) grabbed (it),"  atui konro hot cooking stove "hot cooking stove"  tukanda nomi caught flea "grabbed flea"  Lexical-Lexical Yonda aatikuru atui yoru read article hot evening "articles which were read" "hot evening"  In every case, the initial LH rise of the second word seems to be suppressed.  18 lower category in the prosodic hierarchy) and every prosodic constituent of type X n is exhaustively contained in a constituent of type X n + 1 . (no embedding of prosodic constituents.) Under these assumptions, MiPs are located between MaPs and PWds in the prosodic hierarchy. Hence, the following relation holds: both edges of MaPs dominate the edge of a M i P and the edge of a PWd, and both edges of a M i P dominate edges of PWds.  M a P formation Let us now reconsider (2), a sequence of accented NPs, and (3), a sequence of unaccented NP, in terms of the EB. (2) a  Aoyama-no Yamaguchi-no aniyome-ga inai (left-branching) Aoyama-gen Mr. Yamaguchi-gen sister-in-law-nom absent "We cannot find the sister-in-law of Mr. Yamaguchi from Aoyama." (i.e., Mr. Yamaguchi from Aoyama's sister-in-law)  Aoyama-no^Yaniaguchi-noahiyome-ga inai [ ] IMaP [ ][ .][ ]3MiPu Aoyama-gen Mr. Yamaguchi-gen sister-in-law-nom absent "We cannot find the sister-in-law of Mr. Yamaguchi from Aoyama." (i.e., Mr. Yamaguchi from Aoyama's sister-in-law)  "Selkirk & Tateishi (1988) observe that the subject NP "ani'yome-ga" and the verb "inai" form one MiP because a MiP can contain at most one lexical accent. However, if the verb is accented, for instance, we encounter a situation in which no initial L H rise occurs but the lexical accent is still perceived. I will later propose that initial L H rise occurs at a prosodic level between MiPs and MaPs.  19  S  b  Aoyama-no Yamaguchi-no aniybme-ga inai (right-kanching) Aoyama-gen Mr. Yamaguchi-gen sister-in-law-nom absent "We cannot find the sister-in-law of Mr. Yamaguchi from Aoyama. II  Aoyama-no Yamaguchi-no aniyome-ga inai [ ][ ] 2MaP [ ][ ][ ]3MiP Aoyama-gen Mr. Yamaguchi-gen sister-in-law-nom absent "We cannot find Mr. Yamaguchi's sister-in-law from Aoyama. In (2a), we see that the height of the second noun phrase "Yamaguchi-no" with respect to the first "Aoyama-no" and the height of the third NP "aniyome-ga" with respect to the second are lowered. That is, the second and the third NPs are downstepped with respect to the immediately preceding N P s . 1 2 This is predicted once we adopt Selkirk & Tateishi's assumptions in (6) and (7) that in Japanese the left edge of XPs coincide with the locus of the limits of a MaP, i.e., the left edge of a MaP. In (2a), the subject NPs contained in V P form a single MaP since they are left-branching and all the NPs and V P coincide at the left edge. In (2b), however, the second N P is not downstepped with respect to the first one. This is because (2b) contains a right-branching structure, where the second N P "Yamaguchi-no" creates another left-edge of X P . Assuming Selkirk & 12 Selkirk & Tateishi (1991) and Selkirk (1993) report that some speakers put a MaP boundary between the second NP and the third NP. I assume that these speakers observe strong binarity, which limits the number of MiPs in a MaP into two or that their speech is obeying the rhythmic effect (cf. Kubozono 1991, 1993), which basically rephrases two MaPs as shown below.  The rhythmic effect (cf. Kubozono 1991, 1993) MaP  MaP  MiP MiP MiP  MiP  MaP  ===>  MaP  MiP MiP MiP MiP  Notice that this in fact suggests that the number of MaPs may also be prosodically constrained. Selkirk (1995) also proposes in the OT framework that the formulation of a MaP is optimal when it consists of two MiPs.  20  Tateishi's mapping parameter, we can correctly predict that (2b) contains two MaPs. Next, consider the case of unaccented NPs in (3).  (3) a  Oomiya-ho Inayama-no yuujin-ga inai (left-branching) Oomiya-gen Inayama-gen ffiend-nom absent "We cannot find the friend of Mr. Inayama from Oomiya." (i.e., Mr. Inayama from Oomiya's friend)  Oomiya-no Inayama-no yuujin-ga rna [ ] IMaP [ ] IMiP Oomiya-gen Inayama-gen ffiend-nom absent "We cannot find the friend of Mr. Inayama from Oomiya." (i.e., Mr. Inayama from Oomiya's friend)  Oomiya-ho Inayama-no yuujih-ga iriki (right-branching) Oomiya-gen Inayama-gen ffiend-nom absent "We cannot find Mr. Inayama's friend from Oomiya."  Oomiya-no  [ [  ][ ][  Inayama-no  yuujin-ga  inai  ] 2MaP ] 2MiP  Oomiya-gen Inayama-gen ffiend-nom absent "We cannot find Mr. Inayama's friend from Oomiya." In a sequence of unaccented words, the downdrift of the f 0 contour is gradual and downsteps are  21  not observed; However, comparison of (3a) with (3b) clearly shows that the left edge of XPs starts a MaP. (3a), a left-branching structure, contains one MaP whereas (3b), a right-branching structure, two MaPs.  M i P formation  As for MiPs, we see that, in a sequence of unaccented words, the domain of initial lowering coincides with the domain of downstep in a sequence of accented words. 13 It seems, however, that additional factors are involved for M i P formation in a sequence of unaccented phrases. As shown in (3), an unaccented word followed by an unaccented word constitute one M i P within the domain of a MaP. Selkirk & Tateishi (1988) propose the following constraint to account for the distribution of MiPs. 1 4 (8)  Accent Condition on MiPs A M i P may contain at most one accented prosodic word.  With this condition in mind, let us reconsider (3). As shown in the diagram above, the only difference between (3a) and (3b) is that the former consists of left-branching NPs whereas the latter contains a right-branching structure. If we assume straight bottom-up phrasing, sentences consisting of unaccented words should always form a single M i P as in (3a); However, (3b) is also predicted to have one M i P . Then, what creates the second M i P in (3b)?  It seems that the MaP  introduced by the right-branching structure forces the existence of an extra M i P according to the Strict Layer Hypothesis by the top-down phrasing, i.e., the optimal parse is governed by constraint satisfaction.  2.3.  The  Revised End-based Analysis  13 The existence of MiPs seen in the Initial Lowering is forced by the existence of MaP boundaries under the Strict Layer hypothesis.  Poser (1984), following Kohno (1980) describes this phenomenon as MiP incorporation.  14  22  In the preceding sections, I have provided a basic description of Japanese prosodic structure. I have also discussed concatenations of word accents, and explained the necessity of positing two levels of phonological phrasing above the Prosodic Word: Major Phrases (MaP) and Minor Phrases (MiP), which define the domain of downstep (later, reanalyzed as the domain which initiates metrical boost) and initial lowering respectively. Next, I have reviewed the End-based analysis (Selkirk & Tateishi (1988, 91)) and examined how it predicts phonological phrasings. In this section, I revise the End-based analysis in order to extend its empirical coverage. I first reformulate the phrase structures used in Selkirk & Tateishi (1988, 91) in terms of a more recent syntactic framework i.e., the Principles & Parameters Approach of Chomsky (1992). To be more precise, I adopt the notion of Extended Projection (cf. Grimshaw 1991): more specifically, the DP/KP hypothesis (cf. Abney 1987, Lobbel 1994). Although adotping functional categories such as Ds and Ks does not affect the empirical coverage of the End-based theory, this step is conceptually preferable since the syntax and semantics of noun phrases and sentences are assumed to be parallel. Note that under the D P hypothesis, there are two ways of dealing with proper nouns and pronouns: 1) proper nouns and pronouns are Ds 2) proper nouns and pronouns are NPs with empty Ds. Whichever I adopt, the result will be same; I will, however, I adopt the latter for concreteness option since whether Japanese has Ds or not requires further investigation. I further adopt the K P (Case Phrase) hypothesis since it becomes obvious which nodes case particles are attached to. Henceforth, the DP and K P hypotheses are assumed. Note finally that since Japanese is a head final language, functional heads as well as case markers and conjunctions are always left-branching: [DP[D'[NP D]]] [PP[P'[DP P]]] [IP[i'[VP I]]] [CP[C'[IP C]]] [KP[K'[DP/PP K]]] [DP[?'[DP conj]]DP] This suggests that, under the Japanese Parameter setting in E B , they do not play major roles in building prosodic structures. Let us now reformulate the examples in (2) and (3). 1 5  15 Since the End-based analysis does not make any references to single-bar projections, I will simply omit them for ease of exposition.  23 (2)  a  IP  Aoyama-gen Mr. Yamaguchi-gen sister-in-law-nom absent "We cannot find the sister-in-law of Mr. Yamaguchi from Aoyama." (i.e., Mr. Yamaguchi from Aoyama's sister-in-law)  Aoyama-no Yarnaguchi-noahiyome-ga inai [ ] IMaP [ ][ ][ ]3MiP Aoyama-gen Mr. Yamaguchi-gen sister-in-law-nom absent "We cannot find the sister-in-law of Mr. Yamaguchi from Aoyama." (i.e., Mr. Yamaguchi from Aoyama's sister-in-law)  IP  Aoyama-gen Mr. Yamaguchi-gen sister-in-law-nom absent "We cannot find Mr. Yamaguchi's sister-in-law from Aoyama."  25 (3)  a  Oomiya-no Inayama-no yuujin-ga inai (left-branching) Oomiya-gen Inayama-gen friend-nom absent "We cannot find the friend of Mr. Inayama from Oomiya." (i.e., Mr. Inayama from Oomiya's friend)  Oomiya-no Inayama-no yuujin-ga ma [ ] IMaP [ ] IMiP Oomiya-gen Inayama-gen friend-nom absent "We cannot find the friend of Mr. Inayama from Oomiya." (i.e., Mr. Inayama from Oomiya's friend)  26  Oomiya-no Inayama-no yuujin-ga inai (right-branching) Oomiya-gen Inayama-gen ffiend-nom absent "We cannot find Mr. Inayama's friend from Oomiya."  Oomiya-no Inayama-no yuujin-ga inai [ ][ ] 2MaP [ ][ ] 2MiP Oomiya-gen Inayama-gen ffiend-nom absent "We cannot find Mr. Inayama's friend from Oomiya." Even after adopting these functional projections, we see that the EB can make the right predictions: one MaP is induced in (2a) and (3a) because they both contain a left-branching VP-internal subject, and thus, all the left edges of XPs are aligned at one point. In (2b) and (3b), in contrast, two MaPs are induced because the VP-internal subject contains a right-branching D P (which consists of the second and the third nouns), whose left edge creates a new MaP.  2.3.1.  M a P as a Metrical Boost (MB) initiator  Thus far, it has been recognized that a MaP is the domain of phonological downstep under the E B . Based on the number of MaPs in (9), Selkirk & Tateishi assume that downstep occurs only in left-branching structures. However, Kubozono (1993) shows that downstep occurs irrespective of branching direction. Consider the following examples adopted from Kubozono (1991).  27  na'oko-no a'ni-no ao'i eri'maki Naoko's brother's blue muffler "Naoko's brother's blue muffler"  Examples (9a, b) have identical right branching structures except that the second word of (9b) 16  n indicates lexical accents.  28  "ane-no" is unaccented. Kubozono, reports that "aoi" is realized much lower in (9a) where it follows an accented phrase than in (9b) where it follows one accented and one unaccented phrase. Kubozono thus concludes that downstep takes place irrespective of the branching direction. What then is the MaP domain? Consider the following structures. (10)a  r  ookii aoi-no e-ga aru. (Left Branching) here-loc-top big hollyhock-gen picture-nom exists "Here is a picture of a big hollyhock." ookii aoi-no e-ga arti. 1 7 (Left Branching) here-loc-top big hollyhock-gen picture-nom exists "Here is a picture of a big hollyhock." With respect to the existence of an initial L H rise before "aoi-no', I do not agree with Kubozono' judgement; I could not detect any initial L Hrisein my recording. (An initial L H rise, supposedly by the rhythmic effect or the MiP constraint, may be introduced before "e-ga', however.)  ookii aoi-no e-ga am. (Left Branching) big hollyhock-gen picture-nom exists "Here is a picture of a big hollyhock."  29  ookii aoi-no e-ga aru. (Right Branching) here-loc-top big Aoi-gen picture-nom exists "Here is a big picture of Aoi's." "Aoi"=Female name dokii aoi-no e-ga arn. (Right Branching) here-loc-top big Aoi-gen picture-nom exists "Here is a big picture of Aoi's." "Aoi"=Female name Kubozono proposes that a metrical boost (MB) occurs for disambiguation purposes. He then claims that a right-branching structure has an extra boost compared to a left-branching structure, in order to be disambiguated from a left-branching structure.18  Kubozono (1993, 94) defines  metrical boost (MB) as a phonetic realization rule which modifies the pitch range defined by the phonological process of downstep by reshifting it upwards. In particular, a MaP is defined as the domain between one Metrical Boost and the next boost. Note that Selkirk and Tateishi's system requires reference to two kinds of mechanisms, i.e., downstep (in a sequence of accented words) and right branching structures (in a sequence of unaccented words) in order to predict the domain of MaPs. On the other hand, the M B proposed by Kubozono (1993, 94), only requires right branching structures.  18 In Japanese, which is a left-branching language (cf. Kuno 1973), right-branching structures seem to be marked not only in syntax but also in phonology (cf. Kubozono 1991).  30 2.3.2. Minimization of structure  In this section, I show that the syntax-prosody mapping in Japanese obeys a principle which I call minimization of structures. This principle ensures that only branching XPs are visible (in other words, non-branching XPs are invisible) to prosody and that branching XPs create a new MaP. Along with the notion of M B , I call this modified version of the E B the Revised End-based Analysis (REB). 1 9 In order to motivate this revision of the EB, I discuss three types of constructions (unergative subjects, relative clauses, and coordination), for which the E B fails to predict the correct outputs, but for which the REB can predict the right outputs.  2.3.2.1. VP-subjects vs. IP subjects 2 0  Thus far, we have only been considering sentences with an existential predicate and a VP-internal subject as in (2) and (3). The example in (2a), for instance, contains a subject consisting of three VP-internal left-branching DPs and an existential verb "inai". Thus, the sentence constitutes a single MaP since the left edges of all the XPs coincide at one point.  For Kubozono, right-branching structures mean any maximal projections whose left-edge is not aligned at one point. Thus, in the following configurations, the XPs are considered branching whether the XPs themselves are branching or not. 19  (a)  Later, I will use the notion of (right) branchingness only in the sense of (a). I assume that the subjects of transitive verbs and unergative verbs are generated in Spec IP whereas the subjects of unaccusative verbs are located in Spec VP. For details, readers are referred to chapter 4.  20  31  (2) a  Aoyama-gen Mr. Yamaguchi-gen sister-in-law-nom absent "We cannot find the sister-in-law of Mr. Yamaguchi from Aoyama." (i.e., Mr. Yamaguchi from Aoyama's sister-in-law)  Aoyama-noYamaguchi^noaTiiyome-ga inai [ ] IMaP [ ][ ][ ]3MiP Aoyama-gen Mr. Yamaguchi-gen sister-in-law-nom absent "We cannot find the sister-in-law of Mr. Yamaguchi from Aoyama." (i.e., Mr. Yamaguchi from Aoyama's sister-in-law) Let us now consider an example, adopted from Selkirk and Tateishi (1988, 91), which contains a VP-external subject (i.e., in Spec IP). The subject DP in (11) is located in Spec IP since the verb is transitive. We see that the EB can also correctly predict that the sentence contains two MaPs because the left-edge of the KP/DP/NP dominated by the V P induces a new MaP. The example in (11) contains a transitive verb and thus, the subject is predicted to be in Spec IP.  32 (11)  Aoyama-gen Mr. Yamaguchi-nom sister-in-law-acc called "Mr. Yamaguchi from Aoyama called his sister-in-law." However, consider now my own example below, which contains three left-branching DPs inside the subject of an unergative verb. Recall that the subject of an unergative verb is located in Spec IP.  33  (12) a  Aoyama-gen Mr. Yamaguchi-gen sister-in-law-nom danced "The sister-in-law of Mr. Yamaguchi from Aoyama danced." (i.e., Mr. Yamaguchi from Aoyama's sister-in-law)  Aoyama-no YaTnaguchi-no~aniyome-ga odotta [ ] IMaP [  ][  ][  ][  ]4MiP  Aoyama-gen Mr. Yamaguchi-gen sister-in-law-nom danced "The sister-in-law of Mr. Yamaguchi from Aoyama danced."  Even though the number of MiPs differs among the examples with an unaccusative verb "exist", a transitive verb "read" and an unergative verb "dance", they all contain only one MaP. We see that (12) contains only one MaP. However, the E B would incorrectly predict that the sentence contains two MaPs since the subject of an unergative verb is in Spec IP (and, hence, the left edges of the IP and the V P are not aligned at one point), and the left edge of the V P induces a new M a P . 2 1 The E B proposed by Selkirk & Tateishi requires an awkward stipulation about syntactic structures to make their phrasing algorithm work. That is, Selkirk & Tateishi must either locate the An important thing to note is that by saying that a sentence contains three MaPs, for instance, it simply indicates that the sentence can be uttered only with three MaPs as the minimal option; Thus, it by no means excludes the option of uttering the same sentence with more MaP boundaries, in particular, in slower speech. 21  34 subject under the V P (which is syntactically not motivated) or ignore the V P node in order not to create a new MaP. However, there seems to be an alternative way of interpreting the structure in the E B . In a structure such as (12), the V P is non-branching. As long as nothing modifies the head V , the V P has no effect on prosody. The number of phonological phrasings in (12) is correctly predicted if we assume that non-branching XPs are invisible in phonology. In this vein, it seems reasonable to assume the following constraint. (13) Minimal Structure Constraint (Revised EB) Non-branching XPs, represented as [ m ], are invisible (ignored by) prosodic structures.22 Under this assumption, (12a), for instance, is reformulated as the input to the prosodic structure as shown in (12a)". We can see that the non-branching V P dominated by IP is invisible. 2 3  ^Selkirk & Tateishi (1988,91) utilize the notion of minimization of structures in a different sense. By minimization of structures, they mean that the number of phonological phrases should be as minimal as possible as long as other phonological and syntactic constraints are not violated. It is the equivalent of *Struc, proposed by Zoll (1993) in an Optimality theoreticframework(cf. Smolensky & McCarthy (1993)). In Japanese, this notion becomes relevant in a sequence of unaccented words (and an accented word) in a MaP. ^Rochemont (1997) utilizes the system of Extended Projection in the syntax-prosody mapping. For Rochemont, the VP does not create a new MaP since the IP is the Extended Projection of the verb.  35  (i.e., Mr. Yamaguchi from Aoyama's sister-in-law)  Aoyajiia-noYamagucW^noaniyome-ga oaotta [ ] IMaP [ ][ ][ ][ ]4MiP Aoyama-gen Mr. Yamaguchi-gen sister-in-law-nom danced "The sister-in-law of Mr. Yamaguchi from Aoyama danced." As observed below, the Minimal Structure Constraint makes the right prediction without any unreasonable syntactic stipulations. In the preceding section, I have shown that an MaP is defined as the domain from one M B to the next M B , which signals the existence of right-branching structure. At this stage, we need to define the notion of branching more precisely since one might say that the V P in (12) is branching in that it is a daughter (or is branching out) of the IP. Thus, I propose the following definition of branchingness. (14)  Maximal Projections are branching iff it has more than a single daughter. By (14), we can see that the V P in (13) is non-branching. Further, in the light of the  syntax-prosody mapping, I will restate the Minimal Structure Constraint in the following way.  36 (15)  Branching Constraint: Branching XPs induce a new MaP.  This constraint will ensure that the left edge of a branching X P is alligned at the left of edge of a MaP. Note further that this constraint is consistant with the Strict Layer Hypothesis assumed by the End-based Analysis since the left edge of X l e x is still aligned at the left edge of MaPs and M i P formation is unaffected. Moreover, on the assumption that left-edges of syntactic phrases in left-branching structures all coincide at one point, it does not have to be stated as the right-branching constraint. Now let us return to the examples in (2) and (11). (2a), the left-branching structure, contains four visible maximal projections: K P / K P / K P / V P whose left edges all coincide at one point. Hence, (2a) consists of one MaP. In (11), on the other hand, the left-edge of the branching V P induces a new MaP. We see that the REB, with the use of the Branching Constraint, correctly predicts the number of MaPs as two.  2.3.2.2. Relative  clauses  Let us nowconsider some examples (sequences of unaccented words) which contain a relative clause. 25  ^Inkelas & Zee (1995), in their Arboreal Syntax-Prosody Mapping, which basically claims that the domains of MaPs are defined by sisterhood relations, also utilize the notion of branching. "it is also possible to have a CP projection below the head of the relative clause, although the presence of a CP projection does not affect phonological phrasing.  37 (16) a. Adverb modifying the main clause  [ej Sensyuu nakusita nuigurumi-o mituketa [ ][ ] EB [ ][ ] 2 MaP (correct output) Last week pro [[ej ej lost] dollj-acc found "Last week, I found a doll which I had lost." (6) b. Adverb modifying the embedded relative clause  KP4  [ej  nakusita  Sensyuu nakusita nuigurumi-o mituketa *[ ][ ]EB [ ] 1 MaP (correct) Last week pro [[e^ ej lost] doll]-acc found "I found a doll which I had lost last week." The sentences contain empty categories, such as the pro subject of the main clause. However, the  38 phonological facts indicate that, whether we adopt the EB or the R E B , in order to predict the right output, we have to assume that empty categories are invisible in the phonology because they have no phonological content. Thus, in (16a), the IP2, IP3, KP1, K P 3 , and K P 4 constituents, marked by strikeouts " — " , are invisible since their heads are empty. Likewise, in (16b), IP1, IP2, IP3, K P 1 , K P 3 , and KP4 are invisible. (16) a. Adv modifying the main clause IP1  mituketa  Sensyuu nakusita nuigurumi-o mituketa ] 2 MaP (correct output) "Last week, I found a doll which I had lost.  39 (16) b. Adv modifying the embedded relative clause  [ej  Sensyuu nakusita nuigurumi-o mituketa  *[  ][  ]EB  [ ] 1 MaP (correct) Last week pro [[e, e; lost] doll]-acc found "I found a doll which I had lost last week."  The sentences in (16a) and (16b) consist of exacdy the same words; in (16a), the adverb modifies the main clause whereas in (16b), the adverb modifies the relative clause. Prosody reflects the different phrase structures: in (16a), the sentence contains 2 MaPs whereas in (16b), the sentence contains one M a P . 2 6 Once we assume that empty categories are invisible to phonology, we can see that the EB can predict the correct output for (16a). However, in (16b), the EB fails to make the correct prediction because VP2, whose head is overt, creates a new MaP. Now, let us reconsider (16a) and (16b) in terms of the REB. In (16a), IP2 is considered non-branching and thus invisible because KP1 is empty. In addition, V P 2 is invisible since KP4 [ej is empty, and thus VP2 is not considered as non-branching. Similarly, the IP3 is also invisible since KP3 [ej] is empty, and thus IP3 is considered as non-branching; This in turn makes DPI also non-branching and invisible. Hence, the left-edges of the branching (and thus visible) maximal projections, i.e., IP1 and VP1/KP2 (the left-edges of V P l / K P s coincide) respectively create a ^Similar observations are made in Uyeno et al (1971) and Venditti, Jun, & Beckman (1996).  40 MaP, and, as a result, the sentence contains two MaPs, which in fact is the correct output.  In  (16b), in contrast, IP1 and KP1 as well as IP2, IP3, KP3. KP4, and VP2 are non-branching and thus invisible; This leaves the branching (and thus visible) maximal projections VP1 and KP2, whose left edges coincide at one point. Thus, the R E B predicts that the sentence in (16b) only contains only one MaP, which is the correct output.28  2.3.2.3. C o o r d i n a t i o n  Consider the following example containing two full DPs. (17)  NP Akira  NP Eezi  [Akira-to Eezi]-o mita Akira-and Eezi-acc saw "(pro) saw Akira and Eezi." " i have assumed that empty XPs are totally invisible in the prosodic structure and that the maximal projections which directly dominate them are non-branching. Note that this operation presupposes that phonological phrases are built in a bottom-up manner. That is, MiPs must be formed prior to MaPs. Since "pro" and [e] are not lexical items, they cannot form a PWd and cannot form a MiP and of course cannot form a MaP. We have seen, however, we have seen in the discussion of the EB that MaPs must be determined prior to MiP formation in a sequence of unaccented words (top-down phrasing). The existence of both bottom-up and top-down phrasing is problematic in that it requires a structure changing operation. Readers are refened to Uechi (1995) for a possible solution to this problem within the OT theoretic framework. ^Some speakers may put a MaP boundary before the main verb, especially in (13b), which neither the EB nor the REB can predict. However, I assume that the MaP boundary is introduced for either of the following two reasons: the first possibility is that the speakers are strictly following the MiP constraint (Selkirk & Tateishi (1988) claim that a MaP can contain no more than three MiPs). The second explanation is that a MaP boundary is necessary due to another kind of disambiguation mechanism, i.e., the contour pattern of the example with an embedded adverb is very similar to the contour pattern of the example with an adverb introducing the main clause, where the adverb is contrastively focused; Therefore, by adding a MaP boundary, one may disambiguate the former from the latter. Whichever solution we adopt, this fact suggests that the syntax-prosody mapping requires the mediation of prosodic structures.  41  [A'kira-to E'ezi]-o mita [ ][ ]MaP (EB) [ ]Correct Output Akira-and Eezi-acc saw "(pro) saw Akira and Eezi." In (17), the E B wrongly predicts that there are two MaPs since "Eezi" is a full DP/NP, which creates a new MaP. However, we see that the DP/NP "Eezi" is non-branching and thus invisible. Hence, with the use of the Branching Constraint, the R E B can correctly predict that the sentence contains one MaP. 2 9 (17) Coordination  Akira  Eezi  [A'kira-to E'ezi]-o mita [ ]MaP (REB) [ ] Correct Output [Akira-and Eezi]-acc saw "(pro) saw Akira and Eezi."  2.3.3.  Summary  Thus far, I have shown that the R E B , with the aid of the Branching Constraint and an M B as the domain of a MaP, can straightforwardly accomplish the syntax-prosody mapping in Japanese.30 31  ^Manfred Krifka (p.c.) suggests that "the proposal goes against Jacobs' observation for German that coordinated phrases tend not intergrate into prosodic phrases". ^I am not claiming that non-branching XPs cannot induce a new MaP. Rather, they do not have to create MaPs. There seems to be something going on in the phrasing of DPs containing a demonstrative. Consider the following  3I  42  examples. Sono kireena huku kudasai. (context: pointing out the clothing) that beautiful clothing give "I would like to get that beautiful dress."  [ [  kireena ][  ] expected numbers of MaP = 2 ] actual numbers of MaP = 1  In the above example, even though the NP is branching, it does not seem to create a new MaP. This is the case especially when the speaker is pointing out some item or attracting the hearer's attention. Selkirk may propose that in the EB analysis, functional categories, which have no semantic content, i.e., I, P, D, C, are inserted later in the derivation. In the direct syntax approach by Rochemont (1997), we can say that the DP is the Extended Projection of the N; However, it cannot explain why the AP does not create a new MaP. In current models of syntax (e.g., the Principles and Parameters approach), both the location and the syntactic status of demonstratives have not been thoroughly studied, especially in Japanese. If demonstratives are not determiners, then Japanese may not have overt D's. For instance, it is not clear how to generate demonstratives nor is it clear whether they should be treated as adjectives. (In Japanese, a head-final language, they must occur after NPs if they are Ds.  43 PARTE  2.4. Interaction of phonological phrasing with focus  Thus far, we have learned the basics of the syntax-prosody mapping in Japanese. In this section, we will examine how focus interacts with prosody. I make the following three points: 1) in an out-of-the-blue context, (when items are presentationally focused), the domain of a MaP, (which is initiated by the left edge of XPs,) is further regulated by argument structure and government relations, which is where Focus Projection takes place. 2) contrastive focus, signaled by an extra M B starts a new phonological phrasing, overriding the boundaries of lower phonological phrases. 3) Focus, following non-focus, (a Topic/Focus partition) initiates a new MaP. In order to explain these facts, I will propose two new prosodic levels, (i.e., Projection Phrases (ProPs)) between MaPs and MiPs and Contrastive Focus Phrases (CFPs) between Intonational Phrases and MaPs. Utterances Intonational Phrases (IntoPs)  Contrastive Focus Phrases (CFPs)  Major Phonological Phrases (MaPs) (Focus) Projection Phrases  (ProPs)  Minor Phonological Phrases (MiPs) Prosodic Words (PWds)  2.4.1. Projection  Phrases  First compare the following sets of data, which consist of accented words. (Every element is presentationally focused.)  44  (18) a Context: What is happening?  Robot'to-ga zido'osya-o syu unsite-i-ru [ ][ ]2MaP29 [ ][ ][ ]3MiP robot-nom car-acc repairing-be-pres " A robot is repairing a car."  (Slight Downstep)  b Context: What is happening to the car?  (pro)  Robot'to-ga syu'urisite-i-ru [ ] IMaP [ ][ ]2MiP robot-nom repairing-be-pres "(The car), Robot is repairing."  (Partial Downstep)  c. Context: What is John doing?  (pro)  Robot'to-o syu unsite-i-ru [  ] IMaP  [ ][ ]2MiP??? robot-acc repairing-be-pres "(He) is repairing a robot."  (Total Downstep)  We see that the R E B correctly predicts that (18a), which has a branching V P , contains two MaPs whereas (18b) and (18c), in which the V P is non-branching, contain one MaP. However, we see that in (18b) the predicate "syuurisite-i-ru" starts a new MiP, which is indicated by an initial L H rise and the existence of an accent. Notice also that the second M i P is partially downstepped with respect to the first one. In (18c), in contrast, we are in a dilemma about whether it contains two MiPs or one M i P because both the subject and the predicate are accented and yet no initial L H rise is observed. That is, the predicate is totally downstepped. 29 Between two MaPs that consist of accented words, the second MaP, initiated by MB, still experiences a slight downstep with respect to the first MaP.  45 The End-based approach has not paid much attention to the degree of downstep, in particular between partial downstep (with an initial L H rise before the predicate but not as high as M B ) in (18b) and total downstep (no initial L H rise before the predicate) in (18c). 30  31  In the following section, I claim that MiPs are in fact not the domain of an initial L H rise, but merely a domain which may contain at most one accent. I then propose a new prosodic level between MaPs and MiPs, which represents the domain of an initial L H rise. This level of phonological phrasing is syntactically identified: Any constituent containing a head and its internal argument defines a prosodic phrase which I call projection phrase because it is the domain in which focus projection applies. I demonstrate that total downstep in Japanese, which occurs between two MiPs in a ProP, is the equivalent of focus projection in English (cf. Selkirk 1995, Rochemont 1996), and Focus Projection is constrained by argument structure and government relations. By assuming a level of ProP, we can thus not only distinguish three types of downstep (slight downstep, partial downstep, and total downstep) but also establish a parallelism between total downstep in Japanese and focus projection in English. 3 2 ^ h e following examples show that the VP-internal subject in (b) is also subject to total downstep. By using "tyuu", we can nominative mark "robot": in (b), which makes (a) and (b) phonologically parallel. a  (tadaima kono puranto-de-wa, kuruma-o)TJobot'to-ga syu'uri-tyuu desTt (Partial Downstep) (now this factory-in-wa, cars-acc) robot-nom repairing-in the middle be "Now, in this factory, cars, robots are repairing."  b (tadaima kono puranto-de-wa,) Robot'to-ga syu un-tyuu aesu (Total Downstep) (now this factory-in-wa,) robot-nom repairing-in the middle be "Now, in this factory, robots are being repaired." In the Japanese phonetics and phonology literature, total downstep has been treated as a phenomenon which occurs at the level of morphology, e.g., in verb-auxiliary relations, DP-case marker relations, etc. (cf. Poser 1984, Kubozono 1993) 31  An intermediate prosodic level between MaPs and MiPs has also been previously proposed by Williams (1990). She claims that "partial catathesis"(=partial downstep) occurs in the domain of MaPs and that "total catathesis" (=total downstep) occurs in the domain of intermediate phrases. Although her proposal is similar to mine in that we both claim that the intermediate phonological phrase initiates an initial LHrise,my analysis differs from hers in the following respects. First, Williams has no systematic syntactic motivations to predict the domain of the intermediate phonological phrasing. Second, Williams relates MiP formations with focus, i.e., she claims that "a MiP is assigned to every focus-bearing word." But I claim that a MiP is assigned irrespective of focus status as long as a word is accented. 32  46  2.4.2. Focus projection in English  Let us now review the system of focus projection in English. Consider the following question-answer pairs. We see that focus, identified as the constituent which answers a wh-question, contains a pitch-accented element.33 (19) Object-Verb  a. Context: What did John read? "John read [pa BOOK]." b. Context: What did John do? "John [ F readaBOOK]." 3 4 Following Selkirk (1984, 95) and Rochemont (1986, 96), let us assume that if an element is pitch accented, then the element is [F]ocus-marked. In (19a) "a B O O K " is focused. As shown in (19b), however, as long as the object is pitch-accented, the verb can remain unaccented even though it is focused. This suggests that if a pitch accent is assigned to an object, the focus projects to an unaccented predicate. This kind of phenomenon has often been discussed in English and has led to the developement of the theory of focus projection, in particular, by Selkirk (1984,1995) and by Rochemont (1986, 1996). That is, some item with a pitch accent receives a focus interpretation, while some item without a pitch accent may be interpreted as focused as long as it is in a certain syntactic relation to the pitch-accented item. 3 5 Thus, the theory of F(ocus)-projection may be summarized as a theory of the projection of F-marking from pitch accented items to non-pitch accented ones. Capital letters indicate that items are pitch accented in English.  33  ^(^a) contains contrastive focus whereas (19b) contains presentational focus. Gussenhoven (1984,92) takes a similar approach except that he uses a Sentence Accent Assignment Rule (SAAR) instead of focus projection. The definition of SAAR is given below. 35  "SAAR: If focused, every predicate, argument, and modifier must be accented, with the exception of a predicate that, discounting unfocused constituents, is adjacent to an argument." In this dissertation, I will not pursue Gussenhoven's approach although the empirical coverage of SAAR and focus projection is almost the same.  47  Selkirk (1984, 95) and Rochemont (1986, 96) argue that Focus projection is sensitive to argument structure and to government relations. In other words, simple adjacency relations are not sufficient to license F-Projection. Consider the examples in (20). We see that, in (20a), the agentive subject must bear its own accent to be F-marked since the agentive subject, whether it is in Spec V P (external) or in Spec IP, is not a governed argument of the verb. Therefore, (20b) is uninterpretable. By the same token, (20c) is uninterpretable since projection from the F-marked subject to the predicate is impossible, since the subject is not governed by the predicate. (20) VP-external Subjects  Context: What has happened? a.  [jJOHN] [ F readaBOOK].  b.  #[pJohnreadaBOOK]."  c.  #[pJOHNreadabook]."  In contrast, as shown in (21), pitch accents are not required on the predicates if the subjects are located VP-internally. (21) VP-internal Subjects (Derived Subjects)  Context: What's happened? [ F Your EYES are red] (today). [ F RACCOONS can be seen] (now). [ F TRESPASSERS are prosecuted]. [ F A TIE must be worn]. Diesing (1992) argues that these subjects are raised from its deep structure position within V P . On this assumption, the traces left by NP movement can license F-marking of the sentence and, hence, an accent on the predicate is not required. 36 Accordingly, in English, the number of pitch accents differentiates the two distinct syntactic positions that a subject can occupy: internal to V P vs. external to V P . 3 7 ^Manfred Krifka (p.c.) have suggested to me that this idea was first proposed by von Stechow & Unman (1987). Even though there is a systematic relation between syntax and F-Projection, the following examples from  37  48 To summarize, I provide below the definition of focus projection proposed by Selkirk and Rochemont. (22)  Basic Focus Rule: A n accented word is F-marked Focus Projection (a) F-marking of the head of a phrase licenses the F-marking of the phrase. (b) F-marking of a internal argument of a head licenses the F-marking of the head. (c) F-marking of the antecedent of a trace left by NP or Wh-movement licenses the F-marking of the trace3 8  The system of focus projection can also account for the argument/adjunct distinction nicely. We see that in (23a) a pitch-accented locative argument can project F-marking to the verb whereas in (23b) a pitch-accented locative adjunct cannot project F-marking to the verb. Thus, in (23b), the verb must be independently pitch accented in order to be F-marked. (23) Adjuncts (Examples are from Gussenhoven 1992) Context: What did John do? a.  John remained in the TENT  (verb + argument)  b.  John S M O K E D in the TENT, (verb + adjunct)  Schmerling (1974) demonstrate that pragmatics and, in particular, notions such as presupposition accommodation must also be taken into consideration, (cf. Lambrecht (1994), Rochemont (1996)) (i) a. b.  TRUMAN died, Truman DIED.  (i a) is used when the death of Truman was not expected whereas (i b) is used when Truman has been ill for some time. (ii) a. b.  The BABY's crying The WOMAN'S CRYING.  Regarding the pitch accent difference in (ii), Rochemont argues that the subject in (ii a) is less agentive than the one in (ii b). I agree with Rochemont in that the pragmatic notion of presupposition accommodation also plays a role on pitch-accent assignments: i.e., "crying" is an expected behavior of ordinary babies (thus easily accommodated), but it is not for women. Interested readers are also referred to Faber (1987) and Okazaki (1993), which discuss effects of the semantic/pragmatic relations between subjects and predicates on phonological phrasing. Selkirk (1995) claims that data which contain traces left by quantified NPs are problematic to the system of focus projection. 38  The PRISOners escaped. EVERybody escaped  (cf. Gussenhoven 1984,1992)  She also reports that scrambled items cannot project focus.  49  In the same vein, we see that F-marking on the leftmost argument is sufficient to license the whole DP in (24a) and the whole V P in (24b) as long as they are in a government relation or an argument-predicate relation. (24) a Context: Who came into your office? [F A strange M A N ] came into my office. b  Context: What did Tom do at the concert? He [Freceived a picture of a famous ACTOR]  2.4.3. Focus projection and  phonological phrasing in Japanese  We have seen that the system of focus projection is sensitive to - and, as such, determined by argument relations. In this section, I demonstrate that the domain of total downstep in Japanese is also bound by argument relations. Having demonstrated that total downstep in Japanese occurs exactly in the same environment of focus projection in English, I conclude that focus projection is also applicable to Japanese. Consider the following question-answer pairs in Japanese in (25). Focus may be identified as the constituent which answers a wh-question.39 (25) Object-Verb  Context: What did John do?  Zyo^wa~[jwn-o oto'sita-ncTa] [ ][ ]2MaP John-wa book-acc dropped-mood "John [F lost his BOOK]." In (25), we see that the whole predicate is F-marked, where the predicate undergoes total downstep. It suggests that, in Japanese, F may project from some item with more prominence to some item Recall that in Japanese, pitch accents are the property of lexical elements. Thus, I utilize reduced "extra-narrow font" to indicate that an item is under total downstep. 39  50  with less prominence as long as they are in an internal argument relation. Thus, F-marking of the direct object "hon-o" licenses F-marking of the verb and the whole verb phrase. Consider some more examples. (26)VP(ext) Subject Context: What has happened?  a  [pZyo'n-ga][Fho'n-o otositanda]. [  ][  ]2MaP*°  John-nom book-acc dropped-mood [pJOHN][ F lost his Book]. b.  [pjon-ga ho'n-o otositanda]. #[pFohn lost his BOOK]."  c.  [pJon-ga ho'n-o otositanda]. #[jJOHN lost his book]."  In (26b), we see that the VP-external subject, which is not an internal argument of the verb, cannot undergo total downstep. For this reason (26b) is uninterpretable. By the same token, (26c), is uninterpretable since projection from the F-marked subject to the predicate is impossible since the VP-external subject is not governed by the predicate and thus cannot license the total downstep of the whole predicate. Now, consider (27), the Japanese equivalents of (21). (27)  VP-internal  subjects  Context: What's happened?41 [FTa'nuki-ga mie'-ru.] [ ] IMaP Raccoon-nom seen-pres "RACCOONS can be seen (now)." '"Compared with "h'on-o" in (25), which follows the wa-marked subject, "hon'-o" in (26a), which follows the nominative-marked subject (in a presentational focus context) sounds slightly downstepped. 41  For some reason, an initial LHriseis observed in the following example.  Context: what has happened? [ypMe-ga akai yo]. (stage-level) eyes-nom red prt. "Your EYES are red (today)."  51  [pSinnyu'usya-ga ki'so-s-are-ru] [ ] IMaP Trespasser-nomprosecute-do-makepres "TRESPASSERS are prosecuted" [pNe'kutai-ga tyakuyo'os-are-nakerebanaranai] [ ] IMaP tie-nom wear-make-must " A TIE must be worn." We have seen that in English, focus may project in so far as elements are in internal argument relations. In Japanese, the same effect is accomplished through total downstep. Regarding the argument/adjunct distinction with respect to F-projection, we see that in (28a) the locative argument can project F-marking to the verb whereas in (28b) the locative adjunct cannot project F-marking to the verb. This contrast parallels the contrast in English. Thus, in (28b), the verb cannot undergo total downstep even though it is subject to partial downstep because the V P is non-branching. (28) Adjuncts Context: What did John do?  a. Zyo'n-wa [Fha'nmaa-o komsTTa]42 [ ][ ]2MaP John-wa hammer-acc break-past John[Fbroke A HAMMER]  b. Zyo'n-wa (nanika-o) [pha'nmaa-deJTpko'wasi-fa]. [  ][  ]2MaP  John-wa (something-acc) hammer-with break-past  Johnt^tpBROKE (something)][ with a H A M M E R ] ] . 4 3 F  Focus projection should not be understood as an obligatory phenomenon. For example, in (28a), the predicate may be uttered with an initial LH rise without affecting its interpretation, i.e., without inducing a contrastive interpretation. What is important is that in (28b), for instance, there is no option of making the predicate totally downstepped.  42  43  We can observe the same difference in a sequence of unaccented words. Zyo'n-wa [Ftanebin-ni tenkasi-i [ ][ ] 2 MaP John-wa pilot light-to ignite-past  52 We see that both (28a) and (28b) create two MaPs; however, only the verb in (28a) undergoes total downstep. (The verb in (28b) is still subject to partial downstep.) Astute readers may have already noticed that the directions of F-Projection in Japanese (right to left) and English (left to right) are mirror images. This is indeed the case. We see that in Japanese, F-marking of the rightmost argument is sufficient to license F-marking of the whole DP/VP in (29a) and (29b) whereas, in English, F-marking of the leftmost argument is sufficient.  (29) a Context: Who came into the room? [Fhenna otoko-ga] heya-ni haitteki-ta. strange man-nom room-to came in-past "[ F A strange MAN] came into the room." 4 4 (English)  (Japanese)  2.4.4. ProP as the domain of total downstep and  the domain of focus projection.  Now, recall that the existing End-based approach is not fine-grained enough to distinguish partial John[pignited a pilot light] b.  Zyo'n-wa [ptanebi-deHptenkasi-ta] [ ][ ] 2 MaP John-wa pilot light-with ignited. Johnfj^tpignited (something)][Fwith a pilot light]].  What is interesting is that despite the fact that an intial L H rise is observed before "tenkasi-ta" in (b), the range of the rise is not as high as MB initiating a new MaP, which is induced by a branching XP. ""Later, I will show that the there is a phonological constraint on the number of MiPs which may be contained in a higher phonological phrasing.  53  downstep (with an initial L H rise before the predicate but not as high as in M B ) in (18b) from total downstep (no initial L H rise before the predicate) in (18c). However, by assuming 1) that MiPs are in fact not initiating an initial L H rise, but merely define a phonological domain which contains at most one accent, and 2) that a new prosodic level called Projection Phrases (ProPs) between MaPs and MiPs is responsible for an initial L H rise, we can explain the distribution of partial downstep as well as the distribution of total downstep.45 Let us reconsider (18a, b, c). (18) a Context: What is happening?  ]2MaP ] 2ProP ] 3 MiP  (Slight Downstep)  " A robot is repairing a car." b Context: What is happening to the car?  (pro)  Robot'to-ga syu'urisite-i-ru ] IMaP ] 2ProP ]2MiP  (Partial Downstep)  "(The car), Robot is repairing."  Recall that this intermediate prosodic level between MaPs and MiPs "Projection Phrases (ProPs)" since the domains of ProPs are determined by a head-internal argument relationship, in which the system of focus projection applies.  45  54 c. Context: What is John doing?  (pro) R^b^to-o^yu^unsite-i-ru [ ] 1 MaP [ ] 1 ProP [ ][ ]2MiP robot-acc repairing-be-pres "(He) is repairing a robot."  (Total Downstep)  As we can see, (18a) contains two ProPs (since the agentive subject is not an internal argument of the verb) and two MaPs (since the branching V P induces a new MaP.) We can also see that (18b, c) contain only one MaP since the VPs in (18b, c) are non-branching. However, (18b) shows an initial L H rise between the subject and the predicate because the agentive subject, which is in Spec IP, is not an internal argument of the V P head, and thus induces a new ProP. Hence, (18b) contains two ProPs. (18c), in contrast, shows no initial L H rise since the thematic subject, which is in Spec V P , is an internal argument of the V P head. Thus, (18c) contains only one ProP. Accordingly, by assuming a new extra level of phonological phrasing (ProPs), we can not only distinguish three types of downstep (slight downstep, partial downstep, and total downstep) but also treat total downstep in Japanese and focus projection in English in a parallel manner. Truckenbrodt (1995) and Selkirk (1996) claim that "the representation of prominence (Metrical Structure) is the same as the representation of prosodic constituents (Prosodic Structure). In this representation, grid-marks that represent prominence are related to constituents in a way that is familiar from metrical structure, and that has been proposed for prosodic structure as well." To be more precise, let us put (18) into a metrical format.  55 Context: What is happening? MaP ProP MiP PWd  —1  xj  X X X X  Xl X X  Robot'to-ga zido'osya-o syu'urisite-i-ru [ ][ ,]2MaP [ ][ ] 2 ProP [ ][ ][ ]3MiP robot-nom car-acc repairing-be-pres " A robot is repairing a car."  (Slight Downstep)  b Context: What is happening to the car? MaP ProP MiP PWd  (pro)  Robot'to-ga syu'urisite-i-ru [ ] 1 MaP [ ][ ] 2 ProP [ ][ ]2MiP robot-nom repairing-be-pres "(The car), Robot is repairing."  (Partial Downstep)  c. Context: What is John doing? MaP ProP MiP PWd  x x x  x[ x (pro) Robot'to-o syu'urisite-i-ru [ ] 1 MaP [ ] 1 ProP [ ][ ]2MiP robot-acc repairing-be-pres "(He) is repairing a robot."  —Ix  (Total Downstep)  We see that each grid-mark represents the head of a prosodic constituent (i.e.,the most prominent syllable in the phrase). Truckenbrodt (1995) concludes: "Prominence is assigned rightmost or left most with regular to prosodic constituents." I conclude that, in contrast, in Japanese, prominence is assigned at the leftmost prosodic constituents. What is interesting here is that extra prominence is accumulated at the left edge of every prosodic level, with respect to lexical accents. This successfully captures the difference in the  56  ranges of initial L H rises: a slight downstep is observed between two MaPs dominated by a higher phonological phrase, a partial downstep is observed between two ProPs directly dominated by a MaP, and a total downstep is observed between two MiPs directly dorninated by a ProP. 4 6  2.4.5. Semantic and phonological restrictions on ProPs 4 7  Kori (1996) proposes that new phonological phrases (I assume that he means an equivalent of MaPs) are introduced by a semantic relation rather than by a syntactic relation. (30)  Kyooto-no tooki-o katta (restrictive) [ ] IMaP [ ] 1 ProP Kyoto-gen ceramics-acc bought "(I) bought ceramics of (made in) Kyoto." kinzyo-no/tonari-no okusan-o mita (restrictive) [ ] 1 MaP [ ] 1 ProP neighborhood-gen wife-acc met "(I) met a wife who Uves in the neighborhood." Kyooto-no toozi-o otozureta (non-restrictive) [ ] IMaP [ ][ ] 2 ProP Kyoto-gen Toozi-temple-acc visited "(I) visited Temple Toji of (which is in) Kyoto."  "'Later, I claim that MaPs are directly dominated by a Contrastive Focus Phrase (CFP). "nobita raamen" sounds different from "nobita raamen-o tabeta": i.e., the former contains two ProPs whereas the latter only contains one ProP. I assume that this is because the act of pointing out an object in this context yields a non-restrictive interpretation of the modifier (the modifier does not serve to restrict the reference of the head noun), and thus, a new ProP will be introduced before the head. 47  no'bita not raamen" gone-soft noodle "noodles which have gone soft" no'bita ke'ICflO gone-soft but "(They) have gone soft, but..."  no'bita ra amen-o la be-ia gone-soft noodle eat-past "(I) ate the noodles which have gone soft."  57 Kinzyo-no/tonari-no okasan-o mita  [  (non-restrictive)  ] 1 MaP  [ ][ ] 2 ProP neighborhood-gen Mr. Oka-acc met "(I) met Mr. Oka, who lives in the neighborhood." We see that a new phonological phrase is created between the modifier and the head when they are in a non-restrictive relation. I claim that this new phrase is in fact not a MaP but a ProP since no M B is observed at the beginning of the phrase. This also makes sense semantically because restrictive vs. non-restrictiveness is a relation between a modifier and the head, i.e., a relation between a head and its internal argument. Hence, when a non-restrictive relation holds between the modifier and the head, a new ProP is introduced, hi this circumstance, if we utter the phrase with one ProP, it would only mean that the modifier is contrastively focused, since proper nouns cannot be modified in the general case (Takeru Suzuki p.c.). 48 This explains why initial L H rises are observed in Selkirk & Tateishi's data, when the modifiers and the heads are not in a restrictive relation because in this case, there is only one sister-in-law of Mr. Yamaguchi from Aoyama. 4 9 Selkirk (1995) also points out that notions of topic/focus must play role in a theory of phonological phrasing, noting that syntactic constructions such as vocatives, appositive parentheticals, preposed clauses, and non-restrictive relative clauses must be set off in separate phonological phrases. 49  The same explanation applies to the following data, which were discussed in (9). na'oko-no a'ni-no ao'i eri'maki [ ][ ] 2 MaP [ ][ ][ ][ ] 4 ProP Naoko's brother's blue muffler "Naoko's brother's blue muffler"  Note that this phrasing is applicable only when Naoko has only one brother, who has only one muffler. Otherwise, the phrasing should be as follows: na'oko-no a'ni-no ao'i eri'maki [ ][ 12 MaP [ ][ ] 4 ProP Naoko's brother's blue muffler "Naoko's brother's blue muffler"  58 (2) a  Aoyama-no Yamaguchi-nb aniyome-ga inai (left-branching) Aoyama-gen Mr. Yamaguchi-gen sister-in-law-nom absent "We cannot find the sister-in-law of Mr. Yamaguchi from Aoyama." (i.e., Mr. Yamaguchi from Aoyama's sister-in-law)  Aoyama-no^Yl&Tiaguchi-noahiyome-ga inai [ ] IMaP [ ][ ][ ]3ProP [ ][ ][ ]3MiP Aoyama-gen Mr. Yamaguchi-gen sister-in-law-nom absent "We cannot find the sister-in-law of Mr. Yamaguchi from Aoyama." (i.e., Mr. Yamaguchi from Aoyama's sister-in-law) This is why we observe the initial L H rises before "Yama'guchi-no" and "ani'yome-ga." Note finally that the following data suggest that the distribution of ProPs are also phonologically constrained.  59 (31)  ] 1 MaP ] 2 ProP ]4MiP "(I) took a class from a handsome teacher.  it  In the data above, which contain only a left-branching structure (hence, one MaP is created), we see an initial L H rise introduced between the second DP and the third D P despite the fact that they are in a restrictive relation. We have seen that a MaP may contain at most two MiPs in 2.2. 5 0 Hence, a ProP is introduced so that strict binarity is preserved: a MaP contains two ProPs, and ProPs contains two MiPs. 5 1  2.4.6.  Summary  In this section, we have seen that by positing a level of Projection Phrases (in which total downstep takes place), whose domain is restricted by government relations, a more systematic account of prosody and argument structure (e.g., the argument-adjunct distinction and the unergativeunaccusative distinction) has become possible. We have also seen that the system of focus ^In footnote 12, we have assumed that rhythmic effects take place at the level of MaPs; however, as far as the data (31) is concerned, these effects are taking place at the level of ProPs because "zyu'gyoo-o" is partially downstepped with respect to "sense'e-no". ''However, the data in the footnote 49 suggests that this is not an obligatory phenomenon. In 49, we have seen that three accented words are contained in one ProP.  60 projection, which has been proposed only for stress languages such as English and German, is also manifested in Japanese, a tone language, through the phenomenon of "total downstep" - a fact which has never been addressed in the literature.  2.5.  Prosody of contrastive focus  In the next section, I extend the coverage of the REB to sentences containing contrastive (Heavy Stress) focus. I claim that contrastive focus elements bear an extra M B (EMB), and are not subject to downstep at a l l . 5 2 1 thus propose a prosodic level of Contrastive Focus Phrase (CFPs) above MaPs where an extra prominence is added onto M B (EMB). I then show that a CFP, indicated by an E M B overrides MaP and ProP boundaries, but not M i P boundaries.53  2.5.1. Phonological phrasing of contrastive focus  In this subsection, we will investigate how the presence of Contrastive Focus affects phonological phrasing. A l l the sentences in (32) (ditransitive constructions) have the syntactic structure in (32a) with three (right-) branching XPs, i.e, IP V P , and VP. (Data originally from Kori (1990) (32) consists of only unaccented phrases. (32a) is the utterance in out-of-the-blue discourse contexts (presentational focus). (32b-e) are uttered as answers to wh-questions inducing contrastive focus, which is represented in CAPITAL BOLD type face.  Recall that if accented words are presentationally focused in two separate MaPs, the latter MaP is subject to a slight downstep with respect to the former, (cf. Kubozono 1989,1993) "Some may say that contrastive focus also overrides MiP boundaries, which indicates that word accents are lost after contrastive focus. However, word accents are still preserved even though they are sometimes subtle (because of fo suppression of the subsequent items by contrastive focus) toward the end of an utterance.  61 (32) Unaccented  NF Mayumi-ga Mamoru-rii nuigurumi-d moratta Mayumi-nom Mamoru-from doll-acc received "Mayumi received a doll from Mamoru." a  MaP ProP MiP  x x  X  X  X  X  X  X  X  [pMayumi-ga [ vp Mamoru-ni [ypNuigurumi-o moratta]]]54 [ ][ ][ ]MaP [ ][ ][ ]ProP [ ][ ][ ]MiP Mayumi-nom Mamoru-from doll-acc received Mayumi received a doll from Mamoru."  b  CFP MaP ProP MiP  54  x x X  „55  MAYUMI-GA Mamoru-ni Nuigurumi-o moratta ]CFP ] MaP ] ProP [ ] MiP [ Q: Who received a doll from Mamoru?  The suppression of "moratta" in the out-of-the-blue context is due to focus projection.  As we can see, grids are not capable of distinguishing accented words from unaccented words. Thus, in a sequence of unaccented words, I simply use grids to indicate a H tone at the level of MiP. 55  62 c CFP MaP ProP MiP  x x x  x x x x  x x x  x x x  x x x x  x  x x x  x x x  x x x  x x x x  Mayumi-wa Mamoru-ni N U I G U R U M I - 0 moratta [ ][ ]CFP [ ][ ][ ]MaP [ ][ ][ JProP [ ][ ][ ]MiP Q: What did Mayumi receive from Mamoru?  e  CFP MaP ProP MiP  x  Mayumi-wa M A M O R U - N I Nuigurumio moratta [ ][ ] CFP [ ][ ] MaP [ ][ ] ProP [ ][ ] MiP Q: From whom did Mayumi receive a doll?  d  CFP MaP ProP MiP  x  Mayumi-wa Mamoru-ni Nuigurumi-o M O R A T T A [ ][ ] CFP [ ][ ][ ][ IMaP [ ][ ][ ][ JProP [ ][ ][ ][ ]MiP Q: What did Mayumi do with a doll from Mamoru? 5 6  We see that the presence of contrastive focus on a given accented phrase boosts its pitch and creates a new phonological phrase. I call this extra level of phonological phrase Contrastive Focus Phrases (CFPs) We see in (32c, d, e) that "Mamoru-ni", "Nuigurumi-o", "moratta" respectively start a CFP and therefore, a MaP, a ProP, and a MiP. The presence of contrastive focus also leads to the eh'mination of the initial L H rise of the following ProPs. As for MaPs, MaPs following contrastive focus are incorporated into one MaP. For instance, (32b), in which the first word is focused, forms one M i P and one MaP. Thus, contrastive focus has the effect of overriding MaP and ProP "Syntactic positions of w«-marked elements will be explained in Chapter 4.  63 boundaries in the domain of the same CFP, irrespective of whether the phrase is right-branching or left-branching. In other words, contrastive focus has the effect of making CFPs, MaPs, and ProPs in a given phonological domain form a one-to-one relationship.57 (33)  Contrastive-Focus Phrasing: The left edge of a contrastive focused constituent starts a Focus Phrase, overriding MaP and ProP boundaries in the domain of a phonological phrase. (The domain usually coincides with the end of utterance)  However, the examples in (34), a sequence of accented phrases, suggest that the situation is more complex. (34) Accented  Yuuzi-nom beer-in wine-acc "Yuuzi mixed wine with beer."  a  MaP ProP MiP  mixed  x x  x x x x x x x x [jpYuuzi-ga [ypbiiru-ni [ypwain-o mazeta] [ '][ ][ ]MaP [ ][ ][ ]ProP [ ][ ][ ][ ] M i P Yuuzi-nom beer-in wine-acc mixed "Yuuzi mixed wine with beer."  Selkirk & Shen (1990) show that in Chinese, contrastive focus forces the following elements to lose their tones, which they describe as post-focus tone deletion.  57  64  b  CFP MaP ProP MiP  x x x x  x  d  CFP MaP ProP MiP  e  CFP MaP ProP MiP  x  Y U U Z I - G A biiru-ni waino mazeta [ ]CFP [ ]MaP [ ] ProP [ ][ ][ ][ ]MiP Q:Who mixed wine with beer?  c  CFP MaP ProP MiP  x  x x x  x x x x  x  x  x  x  Yuuzi-wa B I I R U - N I waino mazeta [ ][ ]CFP [ ][ ] MaP [ ][ ] ProP [ ][ [ ][ ]MiP Q:With what did Yuuzi mix wine?  x  x  x  x  x x x  x  x  Yuuzi-wa biiru-ni WAIN-0 mazeta [ ][ ] CFP [ ][ ][ ]MaP [ ][ ][ ]ProP [ ][ ][ ][ ]MiP Q:What did Yuuzi mix with beer? x x x  x  x x  x  x x  x x x x  Yuuzi-wa biiru-ni wain-o M A Z E T A [ ][ ] CFP [ ][ ][ ][ IMaP [ ][ It ][ JProP [ ][ ][ ][ ]MiP Q:What did Yuuzi do to wine with beer?  We see that contrastive focus does indeed suppress the f0 height of following phrases; yet, the accent of each word is nonetheless, still, perceived.58 That is, contrastive focus does not override ^This is quite problematic for the derivational EB analysis in that it requires an additional condition, the Accent  65 M i P boundaries in a sequence of accented words.  2.5.2. Phonological phrasing of presentational focus vs. contrastive focus  The following data from Kubozono (1996) and Kori (1996) show that presentational focus (F) and contrastive focus (CF) can coexist, and contrastive focus has the effect of overriding the subsequent phonological phrase boundaries even if those subsequent elements are presentationally focused. The difference is that only contrastive focus overrides both MaP and ProP boundaries. (35) Data from Kubozono (1996) (Contexts are provided by the author.) a. Context: Did you know that John was in fact forty years old?  P fP  Conj -yori  John Watasi-wa[[[ZUTTO Qohn-yoriwakai]]-to] omotteita]59 [ .1(1 MaP) I-wa by far John-than young come thinking-was "I thought that John was by far younger than I."  Preservation Condition on MiP formation, for which there is no principled explanation. Thus, the derivational approach requires a stipulation to account for the difference in behavior between accented and unaccented sequences in focus. Readers are referred to Uechi (1996) for a possible solution to this problem through an Optimality Theoretic Approach (cf. Prince and Smolensky 1993). In (18a), it is in fact hard to tell whether "zutto" is contrastive focus or presentational focus. This is because we would not have an option of putting "zutto" in a separate MaP, which would end up being same as (35b). 59  66 b. Context: For how long did you think that John is younger than you?  omotteita  Watasi-wa [zutto [[[John-yori wakai]-to] omotteita]] [ ][ ] (2 MaPs) I-wa for a long time John-that young comp thinking-was "I thought for a long time that John was younger than I." Both (35a) and (35b), which consist of identical words, contain right-branching structures, in which all elements are focused except the subjects. However, we see that only "zutto" in (35a) is overriding the subsequent MaP boundaries. We see that in (35a), the adverb "zutto" only modifies the subordinate clause whereas in (35b) the adverb modifies the main clause (and both sentences are uttered in an out-of-the-blue context.). However, only contrastively focused " Z U T T O " in (35a) overrides the MaP boundaries whereas presentationally focused adverb "zutto" in (35b) does not. Thus, contrastive focus overrides MaP boundaries whereas presentational focus does not. This suggests that MaP boundaries are not overriden because the subsequent elements are not focused. Let us consider another example. (Contexts are provided by the author.)  67  Anohito-wa [KATUTE [gaikokuseki-de atta]]-tame, nenkin-ga moraenakatta. [ ](lMaP) That person-wa before foreign-register was because pension-nom receive-may-not-past "That person could not receive a pension because he used to have a foreign nationality before." b. Context: When couldn't that person receive a pension?  'aenakatta  gaikokuseki-de  nenkin -ga  Anohito-wa [katute [[gaikokusekide attatame], [Fnenkin-ga moraenakatta]]]. [ ][ ] (2 MaPs) That person-wa before foreign-register-was because pension-nom receive-may-not-past "That person could not receive a pension before because he had a foreign nationality (at that time)." A similar contrast is observed in (36). In (36a) the adverb "katute" modifies the subordinate clause whereas in (36b) the adverb modifies the subordinate clause as well as the matrix clause. Thus, both (36a) and (36b) contain right-branching structures. However, only the adverb in (36a), which is contrastively focused, is overriding the subsequent MaP boundaries. Also consider the following  68 examples. We see that (37a) and (37b) have an identical phrase structure; Thus, the R E B should predict that both contain one MaP; however, the E M B on"biiru", which starts a new CFP, creates a new MaP. (37) Data by Kori (1996) a.  Context: Anything new today?  Kyoo-wa [doitu-no biiru-o nonda] [ ] (1 MaP) Today-wa Germany-gen beer-acc drink-past "Today, I drank German beer." b.  Context: Anything new today?  Kyoo-wa [doitu-no [BIIRU-O] nonda] [ ][ ] (2 MaPs) Today-wa Germany-gen beer-acc drink-past "Today, I drank German BEER." or "Today, I drank BEER from Germany."  69  In (37a), what the speaker drank was a German beer. On the other hand, (37b) implies that Germany has many nice things to drink, and that the speaker drank a BEER. (37), not incuding the subject, consists of one MaP since the sentence has a left-branching structure, whereas in (37b) contrastive focus creates a new MaP. Notice that there is a difference between presentational focus and contrastive focus. Only contrastive focus creates a new MaP without the aid of right-branching XPs and overrides the subsequent MaP boundaries. On the other hand, phonological phrasing of presentational focus is bound by the syntax, i.e, by phrase structure and argument structure.60  2.5.3. Phonological phrasing and heavy-stress WA-marking  The following sets of data show that the WA-marked elements, when they are accompanied by stress, also induce contrastive-focus phrasing. (38) Unaccented  Mayumi-ga Mamoru-ni nuigurumi-o moratta Mayumi-nom Mamoru-from doll-acc received "Mayumi received a doll from Mamoru."  ^In the next section, I argue that presentational focus also initiates a new MaP when it follows non-focused information.  70  a  MaP ProP MiP  b  CFP MaP ProP MiP  x x x  x x x  x x  x x x x  x  x  morata]]  x  M A Y U M I - W A Mamoru-ni Nuigurumi-o moratta [ ] CFP [ ]MaP [ ] ProP [ ]MiP Q: Who received the doll from Mamoru?  x x x  x x x  x  x  x  Mayumi-wa M A M O R U - N I - W A Nuigurumi-o moratta [ ][ ] CFP [ ][ ] MaP ' [ ][ ] ProP [ ][ ]MiP Q: From whom did Mayumi receive the doll?  d  CFP MaP ProP MiP  x  [pMayumi-ga [ vp Mamoru-ni [ypNuigurumi-o [ ][ ][ ]MaP [ ][ ][ ]ProP [ ][ ][ ]MiP Mayumi-nom Mamoru-from doll-acc received Mayumi received the doll from Mamoru."  c  CFP MaP ProP MiP  x  x  x x  x  x x  x  x x x  x  Mayumi-wa Mamoru-ni N U I G U R U M I - W A moratta [ ][ ] CFP [ ][ ][' ]MaP [ ][ ][ ]ProP [ ][ ][ ]MiP Q: What did Mayumi receive from Mamoru?  71  e  CFP MaP ProP MiP  x x x  x x x  x x x x  x x x  Mayumi-wa Mamoru-ni Nuigurumi-o MORAI-WA sita61 [ ][ ]CFP [ ][ ][ ][ ]MaP [ ][ ][ ][ ]ProP [ ][ ][ ][ ]MiP Q: What did Mayumi do with the doll from Mamoru?  (39) Accented  Yuuzi-nom beer-in wine-acc "Yuuzi mixed wine with beer." a  MaP ProP MiP  x x x  x x x  x x x  mixed  x  [jpYuuzi-ga [ypbiiru-ni [ypwain-o mazeta] [ ][ ][ ]MaP [ ][ ][ JProP [ ][ ][ ][ IMiP Yuuzi-nom beer-in wine-acc mixed "Yuuzi mixed wine with beer."  "Notice that in (38e), "MORAI-" is wa-marked. Susumu Kuno (p.c.) pointed out to me that the verb followed by wa-marking is a gerund.  72 b  CFP MaP ProP MiP  x x x x  x  c  CFP MaP ProP MiP  x  x x x  x x x  x  x  x x  x x  x  x x  x x x  x  Yuuzi-wa biiru-ni WAIN-WA mazeta [ ][ ] CFP [ ][ ][ ]MaP [ ][ ] ProP [ ][ ][ ][ ]MiP Q:What did Yuuzi mix with beer?  e  CFP MaP ProP MiP  x  Yuuzi-wa BIIRU-NI-WA waino mazeta [ ][ ] CFP [ ][ ]MaP [ ][ ]ProP [ ][ ][ ][ ]MiP Q:With what did Yuuzi mix wine?  d  CFP MaP ProP MiP  x  YUUZI-WA biiru-ni waino mazeta [ ] CFP [ ]MaP [ ] ProP [ K ][ ][ ]MiP Q:Who mixed wine with beer?  x x x  x x x  x x x  x x x x  Yuuzi-wa biiru-ni wain-o MAZE-WA sita [ ][ ] CFP [ ][ ][ ][ ]MaP [ ][ ] ProP [ ][ ][ ][ ]MiP Q:What did Yuuzi do to wine with beer?  Notice that in (38b) and (39b), where "Mayumi-wa" and "Yuuzi-wa" are focused, they induce Contrastive Focus Phrasing, which overrides the MaP boundaries whereas in (39a,c,d,e) and (39a,c,d,e), where "Mayumi-wa" and "Yuuzi-wa" are not focused, they do not induce Contrastive  73 Focus Phrasing. 6 2  2.5.4.  Summary  In this section, I have demonstrated that presentational focus and contrastive focus are phonologically distinct, and only the latter creates a CFP, which overrides MaP and M i P boundaries. I have also shown that heavy-stress WM-marking behaves as contrastive focus phonologically.  2.6.  Phonological phrasing of wa-marking without heavy stress.  In this section, I show that wa-marked constituents without heavy stress also has an effect on phonological phrasing. I demonstrate that focus which follows wa-marked constituents without heavy stress starts a new MaP and that default phonological phrasing applies for wa-marked constituents without stress.63 Let us reconsider the following examples. (18) a Context: What is happening? MaP ProP MiP PWd ]2MaP ] 2 ProP ] 3 MiP  (Slight Downstep)  If the sentence contains more than one contrastive focus, each contrastive focus induces a new Contrastive Focus Phrase. !  63-  If an item is contrastively focused, then it starts a CFP.  b Context: What is happening to the car? MaP ProP MiP PWd  Ix x Robot'to-ga syu'urisite-i-ru [ ] IMaP [ ][ ]2ProP [ ][ ]2MiP robot-nom repairing-be-pres "(The car), a robot is repairing."  (pro)  (Partial Downstep)  c. Context: What is John doing? MaP  fx~|  ProP  x I  MiP PWd  ~~*L  x (pro)  —Ix  x—•—^  Robot'to-o syu'urisite-i-ru [ ] 1 MaP [ ] 1 ProP [ ][ ]2MiP robot-acc repairing-be-pres "(He) is repairing a robot."  (Total Downstep)  a Context: What is happening to the robot? MaP ProP MiP PWd  X  X  X X  X X X  —J X  Robot'to-wa zido'osya-o syu'urisite-i-ru [ ][ ]2MaP [ ][ ] 2 ProP [ ][ '][ ]3MiP robot-nom car-acc repairing-be-pres "The robot is repairing a car."  (NO Downstep)  75 b Context: What is the robot doing to the car? MaP ProP MiP PWd  (pro)  II  Robot'to-wa syu'urisite-i-ru ] IMaP ] 2 ProP ] 2 MiP  (NO Downstep)  (The car), the robot is repairing."  c. Context: What is John doing to the robot? MaP ProP MiP PWd  Robot'to-wa syu'urisite-i-ru ] 1 MaP ] 1 ProP ]] 2 M i P [ ][ robot-wa repairing-be-pres ii* The robot, (He) is repairing."  (pro)  (NO Downstep)  We have seen that in the former sets, (18a) contains two MaPs because the V P is branching. In contrast, (18b) and (18c) contain only one MaP because the V P is non-branching in (18b) and the D P is located VP-internally in (18c). However, once the subject DPs and the object DP are wa-marked, we see that the subsequent element, in this case, a verb, creates a new MaP as shown in (40).*4 6 5 Note also that bothwa-marked generics and the subject of an individual level predicate Consider the following question-answer pairs. (i)  a. Context: What did you read? pro [phon-o] yonda pro book-acc read Tread a BOOK." b. Context: What has happened? pro [phon-o yonda]. pro book-acc read. Tread a BOOK."  (ii)  a. Context: What did you read? #pro fhon-wa] yonda pro book-acc read  76 block focus projection Compare the following two sets of examples. (41) MaP ProP MiP  Stage-level Theme (Out-of-the-blue utterance is possible) Context: What's happened? x1 X  x [FTanuki-ga mieru] [ ] IMaP [ ] 1 ProP Raccoon-nom seen RACCOONS can be seen (now). [FSinnyuusya-ga kiso-s-areru] [ ] 1 MaP [ ] 1 ProP Trespasser-nom prosecute-do-make TRESPASSERS are prosecuted. [pNekutai-ga tyakuyoos-aiie-nakerebanaranai] [ ] 1 MaP [ ] 1 ProP tie-nom wear-make-must. A TIE must be worn.  (42) MaP ProP MiP  Individual-level Theme (Commonly used as generics)  Tanuki-wa [Fmieru] [ ][ ]2MaP [ ] 1 ProP Raccoon-wa seen "Raccoons are VISIBLE." Sinnyuusya-wa [Fkiso-s-areru] [ ][ ]2MaP [ ][ ] 2 ProP Trespasser-nom prosecute-do-make Trespassers are PROSECUTED. Nekutai-wa [ptyakuyoos-are-nakerebanaranai] b. Context: What has happened? #pro [hon-wa yonda]. pro book-acc read.  Notice that wa-marked elements which are not contrastively stressed cannot answer wh-questions. This indicates that non-stressed wa-marking cannot be F-marked. Selkirk claims that the separations occur at the level of Intonational Phrases.  65  77  [ ][ [ ][ tie-nom wear-make-must. A tie must be W O R N . .  ]2MaP ] 2 ProP  We see that in the latter sets of data, the predicates start a new MaP, which also forces a ProP boundary. This makes sense if we think that the subject of the individual-level predicate is located outside the domain of focus projection. This also suggests that every element in information/semantic structure must have a corresponding syntactic representation and prosodic representation. Fin (1984) also reports that utterances with a (non-stressed) wa-marked element require a longer pause before the predicate than those with an ordinary case marked element.66 In (41) and (42), we can also see that for the elements that precede the focused element, e.g., for topics, default phonological phrasing applies. That is, phonological phrasing of topics is not affected by any subsequent focus in Japanese.  Especially when sentences are short, the difference is less clear. Context: What is John doing? Jon-wa saitenkensite-orimas-u John-wa reexamining-be-pres "John is reexamining (it)." Context: What is happening? John-ga saitenkensite-orimasu John-nom reexamining-be-pres "John is reexamining (it)." In the following utterance, we see explicitly that a new MaP is created after a wa-marked subject, but not after a subject with nominative case marking. Context: What is Sunny doing with that car? Ano-kuruma-ni-kansimasite-wa Sunny-wa saitenkensite-orimasu [ ][ ][ 13 MaP [ 11 If 13 ProP that-car-prt-about-wa Sunny-wa reexamining-be-polite "As for that car and as for Sunny, he is reexamining it." Context: What is happening to my Sunny? Ano-kuruma-ni-kansimasite-wa Sunny-ga saitenkensite-or-imasu. [ ][ ] 2 MaP [ ][ ][ 13 ProP that-car-prt-about-wa Sunny-nom reexamining-be-polite "As for that car, Sunny is reexamining it."  78 2.7. Summary  In this chapter, I have proposed two new prosodic levels, ProPs and CFPs. With the use of these levels, we can now correlate degree of downstep with the prosodic hierarchy. (This holds only in a sequence of accented words.) IntoP/Utt CFP MaP ProP MiP  [ [ [ [ [  ][ ][ ][ ][ ][  ][ ][ ][ ][ ][ ][ ][ ][  ][ ][ ][  ] ] (no downstep) ][ ] (slight downstep) ][ ] (partial downstep) ][ ][ ](total downstep)*7  That is, an Intonational Phase direcdy dominates CFPs, between which no downstep is induced. A CFP directly dominates MaPs, between which slight downstep is observed. 68 A MaP in turn directly dominates ProPs, of which the ProP is subject to partial downstep. A ProP also directly dominates MiPs, between which total downstep occurs. I have shown that at the level of ProP, the system of focus projection, which has been argued for in Stress languages such as English and German, is also apphcable to Japanese, a tone language Finally, I will discuss some similarities and differences between English and Japanese prosody. Japanese is often described as a syllable- (mora) timed rhythm language compared to English, which is described as a stress-timed rhythm language (cf. Kubozono 1995). 69  What  makes these two language different is the following two phenomena: existence of stress feet and association of pitch accent. Venditti, Jun, and Beckman (1996) claim " In English, (and other similar stress-accent languages), pitch accent plays an essential role in the prominence system. The heads of stress feet can be associated to pitch accents, and this association to a pitch accent defines an extra level of I suppose that H tone suppression occurs in a sequence of unaccented words although the degree of suppression is subtle unlike in a sequence of accented words, wheretotaldownstep takes place.  CT  ^Note, however, that if wa-marked constituents are followed by focused constituents, no downstep takes place, as shown in 2.6. ^Furthermore, Japanese is often described as a mora-counting language whereas English is described as a syllablecounting language (cf. Kubozono 1995).  79  prominence. In other words, a pitch-accented syllable is more stressed than a merely heavy one. In Japanese, the long syllables are not inherendy more stressed than the short ones, and they do not define a prosodic constituent such as a stress foot in English." In Japanese, pitch accents do not function as markers for semantically prominent syllables. Recall that in Japanese, contrasts between "unaccented" phrases (which are marked only by a delimitative L H pattern at the beginning edge) and "accented" phrases (which also have a distinctive H L fall on a designated mora somewhere within the constituent) are lexical. That is, pitch accents in Japanese are lexically linked tones and make no further contribution to meaning. Beckman (1986) claims: "In English, the pitch shape for the accent is not specific to the accented lexical item. The choice of pitch shape can never contrast different lexical items but instead contrasts different intonational meanings." This variety accounts for the comparative richness in intonational shapes that characterizes English. However, the choice of an alternate shape for the accent is not a possibility in standard Japanese. Despite these differences, the following minimal pairs suggest that the two languages are very similar above the ProP level. (I tentatively assume that a nuclear accent is assigned on the rightmost MaP in Japanese.)70 7 1  Truckenbrodt (1995) claims that Japanese in fact has nuclear stress, pace Pierrehumbert & Beckman (1988). Kuno 6p.c.) also points out to me that an item in the preverbal position sounds most prominent. This is very interesting from a typological point of view in that assignment of nuclear stress is universal; nevertheless, it would also mean that in Japanese, in which an extra prominence is assigned to the leftmost items (such as at the level of MaPs and ProPs), only at the level of the utterance is an extra prominence (nuclear stress) assigned on the rightmost item. 70  Some people may say that a nuclear accent occurs not on the subject but rather on the predicate when the VP is non-branching. However, I assume that this is due to a psychological reflex that tries to reflect subject-predicate relations. 71  (a)  Taroo-ga waratta Taroo-nom laughed "Taroo laughed."  (b)  Taroo-ni Megumi-ga atta. Taroo-with Megumi-nom met "Megumi met Taroo." (empathy on Taroo)  In (b), we can see that a nuclear stress is assigned on the subject.  Context: What happened? Utt IntoP CFP MaP ProP MiP  nuclear x x  X X  X  X  X  Mary bought a book  x X X  X X X  Mary-ga hon-o katta  Context: Who bought a book? Utt IntoP CFP MaP ProP MiP  x x x x x x x Mary bought a book  x x x x x x Mary-ga hon-o katta  Context: What did Mary do? Utt IntoP CFP MaP ProP MiP  (x) (x) X  x X X  X  Mary bought a book  x X X  X X X  Mary-wa hon-o katta  Context: What did Mary buy? Utt IntoP CFP MaP ProP MiP  x x  (x) (x) X  X  X X  X  X  X  X  X  X  X  Mary bought a book  Mary-wa hon-o katta  contrastive focus phrasal H , L initial L at most one Accent  81 Chapter 3: Semantics-Discourse Interface: The Semantics and the Discourse Function of wa-marking in Japanese  3.1.  Introduction  The goal of this chapter is to explain the discourse function of wa-marking in Japanese. The major proposal defended in chapter 3 is that the discourse function of wa-marking parallels that of the L H * contour (B-accent) in English. In particular, wa-marking is equivalent to T-marking in the model of Buring (1998). As such, wa-marking can be viewed as one of the discourse strategies available in Japanese for ensuring that a given assertion is congruent - that is, appropriate - to the question under discussion. In 3.2., I discuss the distribution of wa-marked elements. I establish that wa-marking may be used in negative and generic statements and in sentences which contain non-restrictive relative clauses, indexical elements, and non-thematic subjects. I further establish that wa-marked elements do not require antecedents from the previous discourse. In 3.3., I review a recent theory of sentence topics by Buring who claims that a sentence topic, signalled by a L H * contour, creates a set of questions, i.e., a set of sets of propositions. In 3.4., I demonstrate that the function of wa in Japanese parallels that of the L H * contour (B-accent) in English. In 3.5., I extend the theory introduced in this chapter to account for the ungrammaticality of wa-marking with universal quantifiers, the scope inversion possibility that occurs with wa-marking, and multiple wa-marking. Finally, in 3.6., I briefly discuss some alternative analyses in the light of the approach I have taken in this chapter.  82  3.2. What can be wa-marked?  This section addresses the following two correlated questions: 1) What can be wa-marked? 2) Under what discourse conditions is wa-marking felicitous? I first divide the functions of wa-marking into two major categories: wa-marked elements which refer to antecedents evoked in the previous discourse and wa-marked elements which do not require an antecedent in the previous discourse. As a descriptive tool, I will adopt in this section the operational definition of topic proposed by Reinhart (1981). Reinhart's definition is based on the following test.  1  Topic test (Reinhart 1981) Speaker A: Tell me about x Speaker B: x (x is topic) Under this definition, a topic is defined as the subject of predication, i.e., what the sentence is about. Now, let us start with the first category: wa-marked elements which refer to antecedents evoked in the previous discourse.  3.2.1. Wa-marked elements referring to an antecedent in the previous discourse  The examples below illustrate the most commonly described uses of wa-marking, which have been informally described as the "anaphoric use of thematic wa" (cf. Kuno 1973 and others).  2  'The definition of a topic will be revised in 3.3. Kuno (1973) argues that there are two types of wa, i.e., thematic wa and contrastive wa., as illustrated in the following examples. 2  a wa as the theme of the sentence: Speaking of..., talking about... Gengogaku-wa Tom-ga yoku benkyoositeiru linguistics-wa Tom-nom well studying "Linguistics, Tom is studying hard." b wa for contrasted elements: "X..., but..., as for X..." Gengogaku-wa Tom-ga yoku benkyoositeiru kedo, linguistics-wa Tom-nom well studying but, "Linguistics, Tom is studying hard, but  83  (1) A : Yuubin-ga ki-masi-ta yo. mail-nom come-polite-past prt "Mail has come (for you)." B: Yuubin-vw dare-kara des-u ka? mail-wa whom-from be-polite-pres Q "The mail, who is that from?" (2) A : Dare-ga Tookyoo-e it-ta no ka ne? who-nom Tokyo-for go-past comp Q pit "Who went to Tokyo?" B: Tookyoo-e-wa Yamamoto-ga iki-masi-ta.3 4 Tokyo-for-wa Yamamoto-nom go-polite-past "To Tokyo, Mr. Yamamoto went (there)." (3) A : Kesa mata neboo-si-te, kaisya-ni tikoku-si-te-simat-ta. this morning again oversleep-do-participle, company-to late-be-participle-end up-past "I overslept this morning again, and was late for work." B: Soo iu koto de-wa komari-mas-u ne. such case being-wa embarassing-polite-pres part. "Such being the case, it is embarassing." (4) A : Aoyama-ga noohin-o matigae-ta tte senpoo-kara denwa-ga ari-masi-ta. Aoyama-nom supply-acc wrong-past quote customer-from phone-nom exist-polite-past "There was a phone call from the customer, saying that Mr. Aoyama supplied the wrong items." B: Noohin-o matigae-ta ka dooka-wa honnin-ga modor-u made kakunin-deki-na-i supply-acc wrong-past whether wa himself-nom return-pres until making sure-can-not-pres "We cannot detennine whether he supplied wrong items until he comes back." In (1) the wa-marked DP "yuubin" (mail) in (IB) refers to "yuubin" (mail) evoked in the previous Suugaku-HW Mary-ga yoku benkyoositeiru. Math-wa Mary-nom well studying Math, Mary is studying hard." Furthermore, Kuno (1973) claims that the theme must be either anaphoric or generic while there is no such constraint on the contrasted element. ^he following may also be used as an answer to the question in (2A). Tokyo-e it-ta-no-wz Yamamoto des-u. Tokyo-to go-past-comp-wa Yamamoto be-polite-pres "The person who went to Tokyo is Mr. Yamamoto." Yamamoto-ga Tookyoo-e(-wa) iki-masi-ta. Yamamoto-nom Tokyo-to(-wa) go-polite-past "It is Mr. Yamamoto who went to Tokyo." *The wa-marked phrases may be elided in (1) and (2).  84  discourse in (1A). As shown in (2) though (4), wa-marked elements need not be DPs: a PP is wa-marked in (2), an IP is wa-marked in (3), and a CP is wa-marked in (4). Anaphora between a wa-marked element and its antecedent does not require referential identity. A l l that is required is that they stand in a partially ordered salient set relationship in the sense of Ward & Prince (1991). In particular, the anaphoric relation can be a "part o f relation, a "subtype o f relation, a "member o f relation, or an "equal to" relation. Ward & Prince propose to rank referents in the discourse according to the notion of partially ordered sets (Posets). (5)  "Partially Ordered Sets (Posets) = a partial ordering R on some set of referents, b, such that, for all b l , b2, and b3 that are elements of b, R is either reflexive, transitive, and antisymmetric or alternatively, irreflexive, transitive, and asymmetric." (Ward & Prince 1991: 173) " A n element in a poset may be associated with an entity, attribute, event, activity, time or place, or, with a set of such items. Thus, we may rank a property with respect to some entity which exhibits it via the relation IS-A-PART-OF, IS-A-SUBTYPE-OF, IS-AM E M B E R - O F , IS-EQUAL-TO, and so on; an event, with other events." (Ward & Prince 1991: 173)  Consider the following examples of English topicalization, from Ward & Prince, and their Japanese equivalents. Topicalization in English and wa-marking in Japanese (6) a. When I was on surveillance, during this hijacking case, we're working for a newspaper. The guys delivering were selling papers on the side. The newspaper was earning a fortune. These guys knew they were being tailed and they still continued the same shit. People like that you have no sympathy for, they're stupid. (W & P 1991: (16a)) (J)  Soo iu yatura-ni-wa doozyoo-no yoti-ga nai that like people-to-wa sympathy-gen room-nom not-exist "People like that you have no sympathy for"  b. A : Have you finished the article yet? B: Almost, The conclusion I still have to do. (W & P (16b)) (J)  Keturon-wa mada oenakute-wa ikenai conclusion-wa still finishing-wa must "The conclusion I still have to do."  c. A : Did you get any more clues to the crossword puzzle? B: No. The cryptogram I can do like that. The crossword puzzle is hard. (W & P (16c)) (J)  CRYPTOGRAM-wa soo iu fuu-ni dekiru kedo.  85 cryptogram-wa that like can but. "The cryptogram I can do like that" We see that in (6a), "people like that" (soo iu yatura-wa) is in a superset-subset relation to "these guys". In (6b), "the conclusion" (keturon-wa) is part of "the article". In (6c), moreover, "the cryptogram" (cryptogram-wa) and "the crossword puzzle" are members of the same superset. Thus, we see that topicalized elements in English and wa-marked elements in Japanese may stand in a salient partially ordered set relation to an element which has been established in the previous discourse context. We will see, however, that this discourse function of wa-marking, i.e, referring back to an antecedent evoked in the previous discourse, despite its frequency, covers only part of the discourse function of wa-marking. In particular, I will establish in the following section that wa-marked elements do not require any kind of discourse antecedent.  3.2.2. Wa-marked elements not referring to an antecedent in the previous discourse.  Consider the following cases, in which wa-marking serves to identify the topic of a sentence. This section establishes the following important generalization: though the referent of a wa-marked element must be familiar in the discourse, it need not have been explicitly introduced. Thus, consider the cases shown in (7-10). (7)  Context: A customer is looking at some smoked salmon at a souvenir shop in downtown Vancouver. Then, he asks a clerk, Kuukoo-ni mise-wa nai des-u ka? airport-at store-wa not-exist be-pres Q "Don't you have a branch store at the airport?"  (8)  Context: Old high school classmates came across and found out that one is still single. Kekkon-w<2 mada si-na-i no?5 marriage-wa yet do-not-pres comp "Speaking of marriage, aren't you thinking of it?'  'This utterance is not possible if the speaker does not know anything about the hearer.  86  (9)  Context: A man whose love has never been returned says, Moo nayamu-no-wa yame-ta. any more be worried-nominalizer-wa stop-past "As for being worried, I stopped it."  (10)  Context: In an application form for graduate studies in Linguistics, there is a question, Gengogaku-o benkyoosi-ta koto-wa ari-mas-u ka? hnguistics-acc study-past nominalizer-wa exist-polite-pres Q "Have you ever studied linguistics?"  The following set of data show that wa-marking may serve to restrict the domain of predication in the sense of Arita (1996).' (11)  Context: A master is observing his student, and saying, Itininmae-ni nar-u-ni-wa madamada zikan-ga kakar-u. full-fledged-to become-to-wa still time-nom required-pres "In order to be independent, he still needs to practice a lot."  (12)  Context: In the morning, a husband says A , and a wife says B. A : Itte-ku-ru. going come-pres "I am leaving (for school/office..)." B: Okaeri-wz osoi no? return-wa late comp "Are you coming home late (tonight)?"  (13)  Context: A company received a complaint from a customer. A : Aoyama-ga noohin-o matigae-ta soo des-u. Aoyamna-nom delivered goods-acc make a mistake-past seem be-pres "I have heard that Mr. Aoyama delivered the wrong merchandise." B: Senpoo-ni ayamar-u ka dooka-vra honnin-ni makase-yoo. the other party-to apologize-pres whether-wa himself-to leave-let's "Let's let him decide whether or not to apologize to the customer."  6  Arita (1996) also claims that wa-marking has a domain-setting function for predication in the following sense:  "The function of wa is to separate the properties of the entity into two parts: a set of properties for identifying the entity itself and a set of predicates to be assigned to the entity, where the set of properties can be reinterpreted as a set of entities which have the properties." A similar observation is found in Iwasaki (1987).  87  In (14), the use of wa-marking implies that B actually went to school, but that "going to school" was only a part of B's actions on the day , i.e. "going to school" is in a sense contrasted to "coming home early because of illness".7 (14)  A : kyoo gakkoo-e iki-masi-ta ka? today school-to go-polite-past Q "Did you go to school today?" B: Uti-wa si-masi-ta ga kibun-ga waru-katta node gozentyuu-de sootai si-masi-ta. go-wa do-polite-past but feehng-nom bad-past since moming-by leaving early do-polite-past "I did go, but I came home in the morning because I felt sick."  In (15), two VPs are conjoined, each V P containing a wa-marked participial adjunct yielding an iterative interpretation. In (16), a wa-marked participial construction is adjoined at the IP level, and a conditional interpretation arises. (15)  iteratives Taroo-wa atti-ni itte-wa abura-o uri, kotti-ni itte-wa abura-o uri-site-i-ta. Taroo-wa there-to going-wa dawdling, here-to going-wa dawdle-doing-be-past "Taroo-was dawdling here and there."  (16)  conditionals a. Taroo-wa zibun dake-de bankyoo-site-wa hakadora-na-i koto-ga wakat-ta Taroo-wa self only-by studying-wa progress-not-pres nominalizer-nom understand-past "Taroo understood that he would not make progress if he studied by himself." b. Benkyoo-bakari site-ite-wa katayotta ningen-ni natte-simaw-u. study-only doing-being-wa unbalanced human-to becoming-end up-pres. "If all he does is studying, he will end up becoming an unbalanced person."  Temporal, locative and manner adjuncts In the following examples from (17) to (23), the wa-marked constituents restrict the time at which the event described by the main clause occurred. Thus, in (17) through (19) the temporal adjunct clauses, i.e., the "when" and "after" clauses, are wa-marked In (20) through (23), the wa-marked time adjuncts specify the boundaries and the time span of the described event. Similar examples are discussed in Arita (1996)  7  88  Again, note that there is no explicit discourse antecedent for the wa-marked element. (17)  Context: A man is recalling the time when he visited his girl friend's room for the first time. Hazimete kono heya-o tazune-ta toki-wa menkurat-ta yo otokonoko mitai de. first this room-acc visit-past when-wa surprised-past prt male like be "When I visited your room for thefirsttime, I was surprised because it looked like a boy's."  (18)  Context: A man, who people make light of, has accomplished a big project. The man says, Ore datte, yar-u toki-wa yar-u I being, do-pres when-wa do-pres "Even me, I will make it when I have to make it."  (19)  Context: A man is complaining about his unentertaining life. Yonzyuu sugite kara to i-u mono-wa ui-ta hanasi-wa hitotu-mo na-i forty past from comp say-pres property-wa happy story-wa one-also not exist-pres After I passed the age forty, I had no romantic encounters."  (20)  Context: A father is recalling his son's younger days. Itu-no koro kara ka-wa wakar-imas-en ga, when-gen time from Q-wa know-polite-net but, "Since when, I do not know, but musuko-wa pailotto-ni akogar-eru-yooni nar-mas-ita. son-wa pilot-to adore-pres seem become-polite-past my son started longing to be a pilot."  (21)  Context: A student, who got an F in his exam, is seeing his professor and saying, Kore-kara-wa kimoti-o irekaete ganbar-imas-u. now-from-wa mind-acc replacing work-polite-pres "From now on, I will work hard, refreshing my mind."  (22)  Context: The owner of a horse is selling it to a buyer, and he confesses his sad feeling. Izure-wa hanarebanare-ni nar-u koto-wa wakatte-i-ta. some day-wa separate become-pres nominalizer-wa know-be-past "I knew that, some day, we would part."  (23)  Context: A businessman is going to have a surgery for his chronic disease. Toobun-wa kaisya-o yasumi-mas-u. while-wa company-acc absent-mood-pres "I will be away from my work for a while."  The examples in (24) and (25), moreover, show that a constituent describing the location at which  89 an event ocurred may be wa-marked, once again without explicit prior mention. (24)  Context: Late in the evening at a company at the end of the month, Office-de-wa mada zyuugyooin-ga zangyoo-site-i-mas-u. office-at-wa still employee-nom overtime work- doing-be-pohte-pres "At the office, some employees are still working overtime."  (25)  Saikin-no hukeeki-no-tame, Tanaka-san-wa situgyoo-site-sima-imas-ita. recent-prt depression-comp-because, Tanaka-san-wa lose job-doing-ending up-polite-past "Due to the recent economic depression, Mr. Tanaka has lost his job." Sore-de-mo seehu-kara-vra nan-no enzyo-mo ukete-i-mas-en that-be-even government-from-wa nothing aid-polarity receiving-be-pohte-neg "Nevertheless, from the government, he has received no financial assistance."  In (24) and (25), the locatives "at the office" and "from the government" serve as the subject of predication. In all these cases, the wa-marked constituent identifies the topic, the semantic subject of predication in the sense of Reinhart (1981). Note, crucially, that the referent of the wa-marked constituent has not been previously introduced in the discourse. In the next sections, I consider several more subcategories: indexicals, non-thematic subjects (including so called "eel sentences"), non-restrictive relative clauses, negation, copula constructions, and generics.  Indexical elements Rochemont (1986) argues that spatio-temporal expressions e.g., "here, there, now, then, today, tomorrow, last night, etc.," which represent common-ground knowledge between the speaker and the hearer, can function as scene setters.8 Erteschik-Shir (1996) also claims that spatio-temporal arguments can function as topics when they specify the temporal or spatial background of the sentence.9 In Japanese, indexical elements are often wa-marked.10 "When indexical subject pronouns such as "I, you, or we" are used in the main clause, they must also be wa-marked, unless they carry contrastive focus. This is because "I, you, we" cannot be presentationally focused. They always form part of the common ground. Erteschik-Shir (1996) calls such spatio-temporal arguments stage-topics since they refer to the slice or zone of time and space where an event takes place or the time/place at which the event expressed by the sentence takes place. She also claims that only stage-level predicates can have stage topics, i.e., scenes or situations where events can take place. 9  90 (26) a. Kyoo-wa kodomo-tati-ga Disneyland-ni it-ta today-wa child-plural-nom Disneyland-to go-past "Today, my children went to Disneyland." b. Sakuya-wa yuki-ga ippai hut-ta last night-wa snow-nom much fall-past "Last night, it snowed a lot." 1 1 In the above examples, the whole sentence is taken as a presentationally focused event predicated of a temporal expression acting as a topic.  Non-thematic subjects Since the referents of indexical elements are familiar to the common ground of the discourse, wa-marking of indexical elements can be subsumed under the anaphoric function of wa-marking. These items may also be used in presentational focus sentences.  10  Context: What happened? Your grandfather looks lonely. Mago-ga kinoo Tookyco-ni kaet-ta nda grandchild-nom yesterday Tokyo-to return-past mood "His grandson went back to Tokyo yesterday." "Erteschik-Shir (1996) also claims that stage-topics specifying the here-and-now of the utterance may be implicit. We can thus assess the following sentences with respect to an implicit stage topic. Consider the following examples. (i)  Ame-ga hutteiru rain-nom falling "It is raining."  (ii)  Tegami-ga tyoodo kita. letter-nom just came "A letter has just arrived."  The following examples also suggest that wa-marked elements may be elided insofar as they are contextually reconstructable. (iii)  (koko-wal atatakai de-su ne (deletable) here-wa warm-be-pres prt "Here, it is warm."  (iv)  (kyoo-wa) tenki-ga ii de-su ne. (deletable) today-wa weather-nom good be-pres prt "Today, it is fine."  (v)  kotosi-wa musume-ga zvuken nan de-su. (non-deletable) This year-wa daughter-nom entrance exam mood be "This year, my daughter has an entrance exam."  In (v), the wa-marked element is not elidable unless the speaker has been talking about incidents of this year.  91  We now turn to a case in which there is no previously evoked discourse referent for a wa-marked constituent: wa-marking of non-thematic subjects. In the following examples, the sentence initial constituents, which are assigned no thematic roles and thus are adjuncts, are commonly wa-marked. These are often known as "hanging topics" in the literature. (27) a. Sinsyuu-wa soba-ga honba da. 1 2 Sinsyuu-wa noodles-nom best place be "Sinsyuu, it is the best place for noodles." "Sinsyuu is the home of noodles." b. Soba-wa Sinsyuu-ga honba da. noodles-wa Sinsyuu-nom best place be "Noodles, Sinsyuu is the home of them." Although this construction is more often used to make a generic statement, it is also possible with a non-generic interpretation. (28)  Aoyama-san-wa musuko-ga amerika-ni ryuugaku-si-te-i-ru Aoyama-polite-wa son-nom america-to studying abroad be-pres "Mr. Aoyama, his son is currently studying in the U.S."  The constituents following a hanging topic (verbs in the following examples) may be deleted as far as they are contextually recoverable. In the following examples, the verb "to order" is dropped and replaced by the copula "da". These sentences are called "eel sentences."13 (29)  Context: A group of people are placing orders at a restaurant. Boku-wa razania da I-wa lasagnia be "For me, (it is) lasagnia."  12 For Kuno (1973) and others, non-thematic subjects are derived from genitive constructions by subjectivization. The following data, however, suggest that genitive relations are allowed iff the wa-marked elements and the following elements are in a location-object, possessor-possessee, or whole-part relationship.  Tookyoo-wa/no bukka-ga takai. Tokyo-wa/prt prices-nom high "Tokyo, prices are high/Prices in Tokyo are high."  Bukka-wa/*no Tookyoo-ga takai. Prices-wa/prt Tokyo-nom high "Prices, Tokyo is high."  Kirin-wa/no kubi-ga nagai. giraffe-wa/prt neck-nom long "Giraffes have a long neck."  Kubi-wa/*no kirin-ga nagai. neck-wa/prt giraffes are long "Necks, giraffes are long."  "Presumably, the name of the construction originates from somebody's order of an eel dish at a restaurant.  92  Tomu-wa piza da Tom-wa pizza be "For Tom, (it is) pizza." Mearii-wa hanbaagaa da. Mary-wa hamburger be "For Mary, (it is) hamburger  "  Non-restrictive relative clauses Inoue (1982) reports that approximately 50% of Japanese sentences in newspapers start with non-restrictive relative clauses.14 The heads of non-restrictive relatives may be wa-marked without any antecedent in the previous discourse. (30) a. I£Apec-de Vancouver-o otozurete-i-ta] Hasimoto syusyoo-wa ] Apec-for Vancouver-acc visiting-be-past Hasimoto prime minister-wa "Prime minister Hasimoto, who was visiting Vancouver for Apec, kyoo UBC-de okonaw-are-ta konsinkai-ni sankasi-ta.15 today UBC-at held-past social gathering-to atttend-past attended a social gathering held at U B C today." b. Context: There was an anti-Apec demonstration at U B C , However, [[Sooitta zyookyoo-ni nankai-mo tyokumen-site-i-ru] keesatukan-tati-wa ] otituete taisyosi-ta. such situation-to many time facing-be-pres policemen-wa calmly cope-past Inoue (1982) also claims that in Japanese, there is no formal distinction between restrictive relative clauses and non-restrictive relative clauses. 14  15 In most cases, noun phrases with non-restrictive relative clauses may be either wa-marked or ordinarily case-marked, However, wa-marking is possible only if the hearer can identify the referent of the noun phrase.  (i)  Apec-o mi-ni Vancouver-o otozurete-i-ta gakusee-ga/*wa sairuigasu-o abita. Apec-acc see-in order to Vancouver-acc visiting-be past student-nom pepper spray-acc received "A student, who visited Vancouver to see Apec, was struck with pepper spray."  (ii)  Apec-o mi-ni Vancouver-o otozurete-i-tatomodati-waminna sairuigasu-o abita. Apec-acc see-in order to Vancouver-acc visiting-be past friend-wa all pepper spray-acc receive-past "All of my friends, who visited Vancouver to see Apec, were struck with pepper spray."  The example (ii) is called "characterizing sentence generic" by Krifka (1995) and "individual generic" by Kuroda (1992), and are respectively distinguished from the following type of generic statement. Neko-wa doobutu da cat-wa animal be "Cats are animals." This type of generic statement is identified as a "kind-referring NP generic" by Krifka (1995) and a "universal generic" by Kuroda (1992).  93 "The Police, who have faced such a situation many times, coped calmly." Non-restrictive relative clauses, like indexical expressions and non-thematic subjects, are topicestablishing; as we have seen, they all may be wa-marked without requiring any explicit antecedent in the previous discourse. In contrast, a sentence containing a wa-marked topic cannot be used to answer questions such as "What has happened?" or "What is happening", where the whole sentence is presentationally focused.  Negation In most cases, negative statements involve wa-marking. (31)  Nihon-wa Canada-o hinansitei-na-i. Japan-wa Canada-o criticizing-not-pres "Japan is not criticizing Canada."  Sentence negation with ga-marking, as (32), is pragmatically odd since a negated sentence usually presupposes a certain event taking place as in (31), where "criticism" is presupposed. (32)  ?#Nihon-ga Canada-o hinansitei-na-i. Japan-nom Canada-o criticizing-not-pres "Japan is not criticizing Canada." (i.e. Japan is doing some non-criticism of Canada.)  (26) sounds odd if it is uttered without any previous discourse context. As such, it would have to be assessed with respect to the "here and now" of the discourse situation (or a stage topic a la Erteschik-Shir 1996, which refers to the slice/zone of time/location in which an event takes place). The sentence in (32) would only be felicitous if a certain event, e.g., a G7 summit is taking place at the time of utterance, and a T V reporter, for instance, is reporting the situation (though such situations are rare). Nevertheless, there are a few contexts which allow a negated sentence to be assessed with respect to time and location. When the context permits negation of the whole event, negative statements without wa-marking sound natural. The following is one such example.  94  (33)  Context: The speaker has an upset stomach. Hara-no tyoosi-ga yoku-na-i. stomach-prt condition-nom well-neg-pres "My stomach is not well."  Erteschik-Shir (1996), moreover, claims that negation of the whole event is also possible with weather verbs. Interestingly, the following Japanese examples indicate that negation with wa-marking and negation without wa-marking are not semantically equivalent.  (34) Context: What is the matter? Yuki-ga hutte-i-na-i. (even if it is very cold) snow-nom falling-be-not-pres "Snow is not falling." (35) Context: What has happened to the snow since then? Yuki-wa hutte-i-na-i (for several hours) snow-wa falling-be-not-pres "As for the snow, it is not falling." We see that the example in (34), where the whole sentence is presentationally focused and the subject is ga-marked, can only be assessed as a predication of the spatio-temporal argument/the stage-topic, and thus, the whole event is negated (snow falling cannot be observed now) whereas the example in (35) can be assessed as a denial of the predication (snow may be observed, but it is not falling) (cf. Erteschik-Shir 1996).16  17 18  This indicates that a wa-marked element is outside  A similar contrast is observed in the following examples. (i) Ame-ga hutte-i-na-i (Rainfall cannot be observed now) rain-nom falling-be-not-pres "Rain is not falling." (ii)  Ame-wa hutte-i-na-i (Rain may be observed in the form of a flood, for example.) rain-wa falling-be-not-pres "As for the rain, it is not falling."  The difference between (i) where the whole event is cancelled in the sentences with ga-marking and in (ii) where only the predicate is denied comes from syntactic partition (cf. de Swart 1994). More details will be discussed in chapter 4.  95 the scope of negation, whereas an ordinary case-marked element is inside the scope of negation. In the following section, I will introduce cases of wa-marking which do not follow the same generalization as the cases introduced above. I will defer the related discussion of these cases until chapter 4.  Copula constructions As shown in (36) and (37), the subject of a copula construction must be wa-marked in written speech (unless it is contrastively focused). (36)  Asoko-ni a-ru no-wa ani-no gakusee zidai-no syasin des-u. there-at be-pres thing-wa brother-gen student time-gen picture be-pres "The thing over there is a picture of my brother's school days."  (37)  Zinsee-to-wa nan da? life-quote-wa what be "What is life?"  "McGloin (1987: p. 171-172) describes the difference between non-wa-marked negation and wa-marked negation in the following manner. "The use of non-wa-marked negation is highly evaluative and reflects the speaker/writer's subjective evaluation of the non-occurrence of an action or non-existence of a state as strange, unusual, or odd. This type of negative sentence strongly implies that the corresponding affirmative is the norm or what should be the case. The use of vw-negation, on the other hand, is objective. Here, negation is directed without the speaker's subjective evaluation toward a proposition which has been introduced in the discourse or a proposition which can be inferred." It has been often noted that when two clauses are in contrast, the clause-initial elements are wa-marked. (cf. Kuno 1973, Shibatani 1990 and others) However, I claim that these are simply cases in which two clauses exploiting wa-marking are put in a parallel manner. The following examples are from Shibatani (1990: p265). 18  (i)  Taroo-wa hisasiburi-ni tosyokan-ni it-ta ga hon-wa yom-azu-ni nete-i-ta Taro-wa after long time library-to go-past but book-wa read-not-being sleeping-be-past "After a long lapse, Taro went to the library, but books, he did not read and he was sleeping."  (ii)  Ame-wa hutte-i-ta yoo des-u ga mada yuki-ni-wa natte-i-mas-en rain-wa falling-be-past seem be-pres but yet snow-to-wa becoming-be-polite-neg "It seems that it has been raining, but it is not snowing yet."  Note that wa-marking is by no means obligatory when two clauses are contrasted as shown below. We see that the first subject may be nominative-marked without changing the contexts. (i')  Taroo-ga hisasiburi-ni tosyokan-ni it-ta ga hon-wa yom-azu-ni nete-i-ta Taro-nom after long time library-to go-past but book-wa read-not-being sleeping-bepast Taro went to the library after a long time, but books, he did not read and was sleeping."  (ii')  Ame-ga hutte-i-ta yoo des-u ga mada yuki-ni-wa natte-i-mas-en rain-nom falling-be-past seem be-pres but yet snow-to-wa becoming-be-polite-neg "It seems that it has been raining, but it is not snowing yet."  96 Likewise, the subjects of pseudo-cleft constructions, which are a subtype of the copula construction must be wa-marked. (38)  Context: The teacher was always trying to draw out each student's potential. Kodomotati-ga motomete-i-ta no-wa masani sonna kyoosi dat-ta. children-nom seeking-be-past what-wa truly such teacher be-past "The person that students were really seeking for was a teacher just like him."  (39)  Context: The daughters who ran away home were blaming their father for not paying enough attention to them after they left home. But their father is defending his position. Ie-o katte-ni tobidasi-ta no-wa omaetati-no hoo da. home-acc without consulting leave-past nominalizer-wa you-gen side be "The persons who left home without consulting me are you." "It is you who left home without consulting me."  Generics (including subjects of individual-level predicates) The subject of predication in a generic statement, and more generally, the topic of a generic statement is always wa-marked (unless it is contrastively focused) as in (40). In (41) the topic is a conditional clause, and in (42) the topic is a free relative clause. (40)  Nihon-no daigakusee-wa taiman-da Japan-gen university student-wa lazy-be University students in Japan are lazy.  (41)  Hon-o kaku-ni-wa pen-ga hituyooda. book-acc write-to-wa pen-nom needed. "In order to write, a pen is necessary."  (42)  Bankuubaa-ni syuttyoosi-ta toki-wa sumookusaamon-o katte-kita mono da. Vancouver-to business trip-past when-wa smoked salmon-acc bought-came used to be "Whenever I went to Vancouver for a business trip, I used to buy some smoked salmon."  In (43), we have a generic statement containing a non-referential subject CP "dare-no sidoo-o ukeru ka" (whose supervision one undertakes), which is wa-marked. (43)  Dare-no sidoo-o ukeru ka wa sono go-no gukusyazinsee-o kime-ru. whose supervision-acc take Q wa later-no scholarly life-o determine-pres "Whose supervision one undertakes determines one's life as a scholar."  Interestingly, it is possible to make a generic statement about more than one element. (44) is a  97  generic statement about both John and Japan; therefore, the syntactic subject (John) as well as the object (Japan) are wa-marked. We thus see that multiple wa-marking is possible.  (44)  Context: John is leaving for Japan, and his wife is worried. John's friend says to his wife,  John-wa nihon-no koto-wa yoku sitte-i-mas-u kara daizyoobu des-u yo. John-wa Japan-about things-wa well knowing-be-polite-pres since fine be-pres prt "As for John, as for Japan, he knows it well. So, he will be fine."  Multiple wa-marking The following examples show that multiple wa-marking is also possible in non-generic statements. (45)  Context: Two wives are chatting each other. SaiMn-w« gosyuzin-H>a doo site-irassyai-mas-u ka? these days-wa husband-wa how doing-be-polite-pres Q "How is your husband doing these days?"  (46)  Context: A guy with two kids visited his friend's house who has no kids. His friend says, Uti-H> a yoru-wa hutari dakara, nigiyaka-de yokat-ta ndes-u yo. our family-wa evening-wa two so, lively-being be welcomed-past mood pres prt "Because we spend evenings with just two of us, it was nice to have you since was so lively."  (47)  Context: Where did Mr. Tanaka work previously? Tanaka-san-wa motomoto-w a I B M syussin dat-ta. Tanaka-san-wa originally-wa I B M originated be-past "Mr. Tanaka was orginally from I B M . "  (48)  Context: Taro's professor recommended that he read some books. Taroo-w a sensee-ga susumete-i-ta hon-w a sudeni tyuumon-si-ta.19 Taroo-wa teacher-nom recommending-be past book-wa already order-past "Taroo already ordered the book that his professor recommended."  The following variation is also possible. Taroo-wa sensee-ga susumete-i-ta hon-o sudeni tyuumonsi-ta Taroo-wa teacher-nom recommending-be-past book-acc order-past  98  (49)  Context: A student is asking his professor, Sensee-w a Ph.D-w a dotira-de otori-ni nat-ta no-des-u ka? 2 0 teacher-wa Ph.D-wa where-at receieve-past mood Q "Where did you get your Ph.D.?"  Summary In conclusion, we have seen many cases in which wa-marked elements do not require antecedents in the previous discourse context. Thus, I conclude that the generalization that "wa-marked elements refer to an antecedent evoked in the previous discourse" covers only parity the discourse function of wa-marking. What is common to all the cases illustrated above is that wa-marking restricts the domain of elements which the speaker comments on, i.e, wa-marking has a domain-restricting function (cf. Arita 1996, Iwasaki 1987).  3.3. A semantic based analysis of wa-marking using layered focus  In section 3.2., we saw many cases in which wa-marked elements do not require discourse antecedents. In this section, I show that the role of wa-marking in Japanese parallels the role of a L H * contour (B-accent) in English. I then propose an analysis of wa-marking adopting the framework proposed by Buring (1997, 98).  3.3.1. Buring  (1997,  98)  3.3.1.1. Focus and alternative semantics  Buring (1997) proposes an analysis of sentence-internal topics in the framework of alternative ^"Ph.D." may also be accusative-marked.  99 semantics, which was originally proposed by Rooth (1985, 92) for an analysis of focus. Let us first review the Rooth's proposals. Consider the following examples. (50)  A : Who ate the pudding? B: [JOHN] F ate the pudding.  The focus part of the sentence is the information which is asked for by A's question. As shown in (50) , the focused constituent is F-marked, and the capital letter indicates that the constituent is pitch accented. Focus is marked by H * L contour, which is known as an A-accent (cf. Jackendoff 1972). In order to explain the semantic effect of focus, Rooth introduces an additional semantic value [a] f , which denotes the focus semantic value for the phrase a , in contrast to its ordinary semantic value [a] 0 . Informally, the focus semantic value is defined as follows. 2 1 "The focus semantics value [oc]f for a phrase of category S is the set of propositions obtainable from the ordinary semantic value [a]° by making a substitution in the position corresponding to the focused phrase." (Rooth 1992) If we assume that the domain of individuals is restricted to three members, e.g., {John, Mary, Tom}, the focus semantic value of (50B) will be represented as in (51). (51)  focus semantics value [JohnF ate the pudding]f = {John ate the pudding, Mary ate the pudding, Tom ate the pudding}  Buring (1998) reinterprets Rooth's account in terms of (possibly implicit) questions rather than assertions. Buring then redefines Rooth's focus value in terms of A f = the question which an assertion answers. Buring's informal definitions for A f and congruence are given below. (51)  A f : the focus question (F-question for short) associated with an assertive sentence A Form A f by replacing the F-marked constituent with a wh-word and fronting the latter. If F marks verum focus (i.e. accent on negation or finite verb), form a yes/no question. (Buring 1998: p4)  Emmon Bach (p.c.) suggests that presentational focus, out of the blue (whole sentence focus) utterance, can be seen as involving a contrast between the propositional content of the utterance and all other possible propositions in the universe of discourse. This type of presentational focus differs from contrastive focus in that contrastive focus presupposes a contextually specified set of propositions, whereas presentational focus does not. 21  100  (52)  Congruence: An assertion A must be congruent to the Question Under Discussion (QUD), i.e., A f must equal QUD. (Buring 1998: p4)  Thus, Buring argues that a question-answer pair such as (50) is congruent because the focus question A f associated with the assertion in (50B) (that is, "who ate the pudding") is equal to the question under discussion. (QUD) (that is, to the question in (50A). In contrast, the question-answer pair in (53) is not congruent. (53)  A : Who ate the pudding? B: #[JOHN] F ate the cake  (53B) is not a felicitous answer to (53A) because the focus question corresponding to (53B) (who ate the cake) is not equal to the QUD (who ate the pudding).  3.3.1.2. T-marking  Buring (1997, 98) extends the framework of alternative semantics to account for not only focus but also topic-marking. Following Jackendoff (1972), he identifies a distinctive intonation contour (the so-called B-accent, realized as L H * as T(opic) accent. Note that when a T-accent is placed over "cake" in (54) , the answer suddenly becomes appropriate - i.e., congruent, in the sense of (52). 2 2 (54)  A : Who ate the pudding? B: [JOHN] F ate the /[CAKE] T \  In order to derive the effect of B-accents on interpretation, Buring introduces another type of semantic value, parallel to the focus value: the topic semantic value. Topic value is a set of focus values, i.e., a set of sets of propositions. It may also be understood as a set of questions, A'. In the 1 indicates the LH pitch accent; \ indicates the HL pitch accent.  72  101 same sense as A f is derived from the focus value [ a ] f by replacing the set of propositions by the implicit questions which they answer, A ' is derived from [af  by replacing the set of sets of  propositions by the set of questions which they answer. Buring argues that for an answer to be felicitous to a question, the QUD must match one element in the set of questions A 1 . The revised definition of congruence is now as in (55). (55)  Congruence: For A to be congruent to QUD, a. QUD must be an element of A ' b. if A contains T-marking, some question Q in A 1 must be open to serve as the new QUD.  We see that (53B) remains incongruent with respect to (53A) because the QUD in (53A) does not match with any element in the singleton set of questions (53B). Note that in T-less sentences such as the question-answer pair in (53), A ' equals A f ; (see Buring 1998: p8) which represents the A ' of (53B). (53)'  QUDof(53A) {who ate the pudding} A f o f (53B) = A l of (53B) {who ate the cake}  However, Buring argues that in a sentence with a T-marking such as (54B), the T-marked constituent, signified by a B-accent, induces a set of alternative focus values for the sentence, one of which matches the QUD. We now see that in the question-answer pair in (54), the answer becomes felicitous with respect to the question because the QUD in (54) matches one element in the topic value of the answer, i.e, the set of questions derived from (54B). This satisfies (55a). (54)'  QUDof(54A) {who ate the pudding} A of(54B) {{who ate the pudding}, {who ate the cake}, {who ate the pie}}  Now, since a T-marking creates a set of questions A', it also creates at least one question which remains open, as stated in (55b). This gives rise to the 'implicature' effect associated with T-marking  102 (topic implicature). Notice, in contrast, that without T-marking, there is no such an effect, as shown in (50) repeated below, where B's answer has the effect of removing all the propositions other than "John ate the pudding" 2 3 (50)  A : Who ate the pudding? B: [JOHN] F ate the pudding. Buring demonstrates several other uses of the B-accent. Consider the following examples,  from Buring (1997), and their Japanese equivalents.  Contrastive topics 2 4 (56)  A : Which book would Fritz buy? (context) B:  Eeto, watasi-w a Hotel New Hampshire-o kaw-u na. well, I-wa Hotel New Hampshire-acc buy-pres prt Well, [I]T would buy ['The Hotel New Hampshire*]F  We see in (56B) that a B-accent is used to move the conversation away from an entity given in the previous discourse, by answering a question A 1 rather than the question given by A f . (56)'  QUD(=A f ) {which book would Fritz buy} A' {{which book would I would buy}, {which book would Fritz buy},...}  We see that the person who answers the question does not answer the original question but another related one contained in A'.  ^Szabolcsi's (1981) claim below supports Biking's analysis of contrastive topic. "The characteristic difference is that by using a sentence with a contrastive topic, one suggests (or, implicates) that the claims he is making need not be true of something else, whereas by using a contrastive focus one asserts that the claim he is making is in fact not true of anything else. An additional difference is that whether the topic of a sentence is contrastive or not usually depends on whether it receives an extra intonation contour while most foci (i.e. most expressions in F position) are necessarily contrastive in the above sense." ^ote that later in this chapter, I will use the terminology "contrastive topic" in a different sense.  103 Partial topics The example in (57) illustrates that a B-accent might be used to 'narrow down' a given QUD." (57) a. A:What did the pop stars wear? (context) B:  (57) '  Onna-no pop star-M> a caftan-o kite-i-ta. female-prt pop star-wa caftan-acc wearing-be-past The [female]T pop stars wore [caftans]F  QUD (=Af) {what did the pop stars wear} A1 {{what did the female pop stars wear}, {what did the male pop stars wear},...}  Without a B-accent, (57) would be incongruent because B would convey insufficient information with respect to A's question. More precisely, B's answer is not felicitous because it is not exhaustive with respect to the QUD - B answers as if A asked only about a subset of the pop stars. The B-accent on 7 [ F E M A L E ] T \ " , however, creates a set of questions {{what did the female pop star wear}, {what did male pop star wear}} from {what did the pop stars wear}. Once the A' has been created, B's answer becomes felicitous.  Implicational topics (58)  A : Did your wife kiss other men? (context) B:  (58)'  Watasi-no tuma-H' a hoka-no otoko-ni kisusi-na-katta. my-prt wife-wa other-prt men-to kiss-not-past [My] T wife [didn't]F kiss other men. (but Fritz's did...)  QUD(=A f ) {Did my wife kiss other men, Didn't my wife kiss other men}  A  {{Did my wife kiss other men, Didn't my wife kiss other men}, {Did Fritz's wife kiss other men, Didn't Fritz's wife kiss other men}...}  Implicational topics are a little different from contrastive topics and partial topics in that the question-answer sequence would be well-formed without the implicational topics. In other words,  104  A f is congruent to the QUD unlike in the previous two cases. However, T-marking is still possible, giving rise to the implicature effect described in (58B)  (ordinary') topics (=Topicalization') (59)  A : What did you buy on 59th street? (context) B:  (59)'  59bangai-de-H' a (watasi-wa) kutu-o kaw-ta. 59th street-on-wa (I-wa) shoes-acc buy-past [On 59th street]T, I bought [the shoes]F  QUD (=A*) {what did you buy on 59th street} A  {{what did I buy on 59th street}, {what did I buy on 60th street}, {what did I buy on 61th street}}  Note once again that the B-accent on "59th street" creates a set of questions A', one of which matches the answer in (59B).  3.3.2 Layered focus and B u r i n g (1998)  Buring (1998) notes that in order to deal with answers to multiple wh-questions in English, his account of T-marking must be further enriched. To see why, first of all, note that answers to multiple wh-questions contain an A-accented and a B-accented constituent. Consider the following data from Liberman & Pierrehumbert (1984).  105  (60)  Question: What about Manny? Who came with him?  Answer: Anna came with Manny. A-accent B-accent Japanese equivalent Anna-ga Mannyy-to-wa ki-ta (or, Manny-to-wa Anna-ga ki-ta) 2 5 Anna-ga Mannyy-with-wa come-past "Anna came with Manny." A-accent B-accent (61)  Question: What about Anna? Who did she come with?  Answer: Anna came with Manny. B-accent A-accent Japanese equivalent Anna-wa Manny-to ki-ta Anna-wa Manny-with come-past "Anna came with Manny." B-accent A-accent We see in (60) that the A-accent (H*L) is followed by the B-accent ( L H * ) whereas in (61) the B-accent is followed by the A-accent. 27 The presence of a B-accent implies that T-marking must be contained in the answer to multiple wh-questions. Note also that wa-marking is present in Japanese exactly where the B-accent is present in the English cases. Now, consider how the model introduced so far will deal with the following cases. (60)"  QUD: Who came with who? A 1 : Who came with Manny? Who came with Mary?  "The question may also be answered without using a sentence topic if the answerer provides a single answer repeatedly. ^ h e following answer is also possible. Keeki-wa John-ga tabe-ta. cake-wa John-nom eat-past ^Liberman & Pierrehumbert also claim that B-accented elements may optionally be deaccented.  106  (61)"  QUD: Who came with who? A 1 : Who did Anna come with? Who did John come with?  We see that the definition of congruence adopted in (55) does not make the right prediction for (61)', for instance, because the QUD is not an element of A'. Thus, (61)" is problematic for (55). Buring (1998) approaches this problem by adopting from Roberts (1996) the idea that A ' can be reformulated in terms of a hierarchically organized set of subquestions. Buring organizes these questions by employing what he refers to as D(iscourse)-trees. Some central notions of D-trees are given in (62) and an example of a D-tree is given in (63). (62)  Buring (1998: p 18 (15)) a. any node in a D-tree is called Move b. for any Move M , the question minimally dominating it is called the immediate question under discussion, IQUD (M). c. for any Move M , those questions dominating M form the question under discussion stack for M , QUD(M); elements in QUD (M) are totally ordered according to the following conventions: If Q l dominates Q2 (and both dominate M), Q l precedes Q2 in QUD (M) d. for any Move M , if M is a question, the sub-tree rooted in M is called the strategy to answer M e. for any Moves M l and M2, M l precedes M 2 if M l dominates M2, or there are Moves M3 and M4 which are sisters, M 3 is to the left of M4, and M 3 dominates or equals M l and M 4 dominates or equals M 2 .  (63) a. A Question-answer without B-accents (without T-marking) Q <~M(ove)  I  A <~M(ove) b. A Question-answer with B-accents (with T-marking) Q <~M(ove) SQ1 I A  SQ2 I A  SQ3...<-M(ove)  I  A . . . <~M(ove)  Now, we can replace A f by M f and A' by M 1 and redefine congruence in the following fashion. (64) Congruence For M to be congruent to IQUD (M) the following conditions must be met:  107  a. F-condition: IQUD (M) =M f b. T-condition: M* indicates a strategy S to answer IQUD (IQUD (M)) (provided M contains T-marking) Thus, (64) entails that the answer to a multiple wh-question must both answer a subquestion (the IQUD (M) and a super question (the strategy IQUD (IQUD(M)). 2 8  2 9  Now, returning to (61)",  we see that M f equals the IQUD (SQ1), and thus the F-condition is satisfied. Buring claims that "if we assume that a B-accent hints at a main question of the form i.e., 'who came with who?', it becomes obvious that each element in (61) is a subquestion to that main question. In other words, the element in (61) indicates a strategy to answer the main question." We see that the IQUD (M) is an element of M \ and thus indicates a strategy to answer the main question. Therefore, the T-condition is also satisfied. (61)"  Q: Who came with who? SQ1: Who did Anna come with? SQ2: Who did John come with?... A : Anna came with Manny. [ A N N A T came with M A N N Y F ] f = {who did Anna come with} [ A N N A X came with M A N N Y F ] 1 = {{who did Anna come with},  Assuming that Focus-Background Structures (FBS) may be recursive (layered focus), Kanerva & Gabriele (1995) propose that "Difference in the relative prominence of accents can indicate alternative orders for packing/unpacking information." They relate the phonological structuring of accent prominence to the information structure in the focus layers in the following way. 28  (i) Kanerva & Gabriele (1995: p340) a. Within a focus layer, focus is more prominent than non-focus. b. Prominence accumulates across focus layers. Kanerva & Gabriele map the relative prominence of A-accents (H*L contour) and B-accents (LH* contour) to the information structures with two levels of focus layer. Let us suppose that (60) and (61) are uttered at a party, where many people come as couples. We see that in the outer layers, both ANNA and MANNY are focused in (60) and (61); However, in the inner layers, ANNA is focused in (60) whereas MANNY is focused in (61). That is, B-accented elements play roles as embedded focus: the item itself is not only focused but also is the part of background information for the other (A-accented) focus. Williams (1997) also considers B-accented items to be subordinated foci. He claims that they are subordinated in that A-accented items are clearly the focus of the entire sentence. For instance, in (61), x came with ANNA is the presupposition; However, the presupposition itself also contains a focus-presupposition structure: x came with y is the presupposition, in which ANNA is the focus. ANNA then is a focus embedded in the presupposition of another (higher) focus, i.e. MANNY. Accordingly, he identifies the subordinate focus as topic. Westmoreland (1994) and Roberts (1996) also claim that in multiple wh-questions, a B-accent establishes a series of subquestions that fix the value of one wh-element. ^For focus only questions, only the F-condition is necessary.  108  {who did John come with}...} We can now revise the analysis of the various kinds of topic we have examined in the light of Buring. In (56-59)", the question-answer sequences containing a T-marking are congruent because equals SQ1 and SQ1 is an element of M 1 and thus induces a strategy to answer the Q. (Implicit questions are italicized).  Contrastive topics (56)"  A : Which book would Fritz buy? B:  Eeto, watasi-w a Hotel New Hampshire-o kaw-u na. well, I-wa Hetel New Hampshire-acc buy-pres prt Well, [fJT would buy ['The Hotel New Hampshire']F  Q: Which book would Fritz buy? SQ1: Which book would I buy? SQ2: Which book would Fritz buy? A : I would buy The Hotel New Hampshire. Nf*  .  '  {which book would Fritz buy}  M1 {{which book would I would buy}, {which book would Fritz buy},...}  Partial topics (57)"  A:What did the pop stars wear? B:  Onna-no pop star-w a caftan-o kite-i-ta. female-pit pop star-wa caftan-acc wearing-be-past The [female]T pop stars wore [caftans]F  The [female]x pop stars wore [caftans]F Q: what did the pop stars wear? SQ1: what did the female pop stars wear? SQ2: what did the male pop stars wear? A : The female pop stars wore caftans.  rvf {what did the pop stars wear}  M {{what did the female pop stars wear}, {what did the male pop stars wear},...}  109  Implicational topics (58)"  A : Did your wife kiss other men? Watasi-no tuma-w a hoka-no otoko-ni kisusi-na-katta. my-prt wife-wa other-prt men-to kiss-not-past [My] T wife [didn't]F kiss other men. (but Fritz' did...)  B:  Q: Did your wife kiss other men? SQ1: Did my wife kiss other men? SQ2: Did Fritz's wife kiss other men? SQ3... A : M y wife didn't kiss other men. M* {Did my wife kiss other men, Didn't my wife kiss other men} M {{Did my wife kiss other men, Didn't my wife kiss other men}, {Did Fritz's wife kiss other men, Didn't Fritz's wife kiss other men}...} (Ordinary) topics ftopicalization) (59)"  A : What did you buy on 59th street? B: 59bangai-de-M> a (watasi-wa) kutu-o kaw-ta. 59th street-on-wa (I-wa) shoes-acc buy-past [On 59th street]T, I bought [the shoes]F Q: what did you buy on 59th street? SQL what did I buy on 59th street? SQ2: what did I buy on 60th street? A : On 59th street, I bought the shoes. M  f  {what did you buy on 59th street} M {{what did I buy on 59th street}, {what did I buy on 60th street}, {what did I buy on 61th street}} 3.4. Parallelism between wa-marking in Japanese and B-accents in English  In Japanese, the use of B-accents is not an option because accents are properties of lexical items as demonstrated in chapter 2; However, I propose that T-marking is introduced by wa-marking in Japanese. In the last section, I have provided the Japanese equivalents of the English examples  110  discussed by Buring and others. We can see that the distribution of wa-marking in Japanese exacdy parallels the distribution of B-accents in English. This suggests that in Japanese, wa-marking introduces T-marking, which creates M l in a D-tree. I will show that the theory introduced in the previous section accounts for the Japanese data provided in 3.2.  3.4.1. Application of the proposal  First, let me emphasize that none of the examples given in 3.2. can answer questions such as "What has happened?" or What is the matter?", which force the whole answer to be presentationally focused. The fact that these utterances cannot be used in a context which forces the whole sentence to be presentationally focused suggests that these utterances must be partitioned into focus-background structure, i.e., implicit question-answer sequences. For instance, if we consider the examples in (65), a question-answer sequence without a topic would be represented as follows. (65) A : Dare-ga Tokyo-e it-ta no ka ne? who-nom Tokyo-for go-past comp Q prt "Who went to Tokyo?" B: Yamamoto-ga Tokyo-e iki-masi-ta. Yamamoto-nom Tokyo-to go-polite-past "Mr. Yamamoto went to Tokyo." (65B) is congruent to (65A) because M f matches the (main) question under discussion, as shown in (65)". (65)"  Q: Who went to Tokyo? A : Mr. Yamamoto went to Tokyo. {who went to Tokyo}  In contrast, the existence of wa-marking in (66) indicates that it is not a simple question-answer sequence but rather one which contains a topic and thus a sequence of subquestions.  Ill  (66)A: Dare-ga Tokyo-e it-ta no ka ne? =(2) who-nom Tokyo-for go-past comp Q prt "Who went to Tokyo?" B: Tokyo-e-wa Yamamoto-ga iki-masi-ta. Tokyo-for-wa Yamamoto-nom go-polite-past "To Tokyo, Mr. Yamamoto went." (66B) has an explicit main question and a set of implicit subquestions. (66) "  Q: Who went to Tokyo? SQ1: Who went to Tokyo? (SQ2: Who went to Osaka?, SQ3: Who went to Nagano?...) A : To Tokyo, Mr. Yamamoto went. {Who went to Tokyo} M {{Who went to Tokyo}, {Who went to Osaka}, {Who went to Nagano}...}  We see that (66) is congruent because M f equals the SQ1 (F-condition), and the SQ1 is an element ofM1.30 As shown in (67), moreover, if a question is not provided by the questioner, the main question may also be implicit. (67) =(24)  Context: Late in the evening at a company at the end of the month Office-de-wa mada zyuugyooin-ga zangyoo-site-i-mas-u. office-at-wa still employee-nom overtime work- doing-be-pohte-pres "At the office, some employees are still working overtime."  In the following context, only (66B), but not (65B) can be used. Context: A company send its employees to its branches in Tokyo, Osaka, and Nagano. A: Who went to Tokyo branch, but not to Osaka branch? B: Tokyo-e-wa Yamamoto-ga iti-masi-ta. Tokyo-to-wa Yamamoto-nom go-polite-past "To Tokyo, Mr. Yamamoto went." #B: Yamamoto-ga Tokyo-e iki-masi-ta. Yamamoto-nom Tokyo-to go-polite-past "Mr. Yamamoto went to Tokyo."  112  (67) "  Q: What is happening at the company? What is happening at the office? (SQ2: What is happening at the factory?, SQ3:...) A : At the office, some employees are still working overtime.  SQL  M {what is happening at the office} M {{what is happening at the office}, {what is happening at the factory}...} However, in order to maintain discourse coherence, the main question must contain information provided by the discourse context. We see in (67) and (67)" that "(at) the company" in the context is accommodated into the implicit main question". This maintains discourse coherence. (67) is congruent because it meets the F-condition ( M f equals SQ1) and induces a strategy to answer the main question (thus, satisfies the T-condition). In sentences containing a non-thematic subject such as (68), or negated sentences such as (69), the use of wa-marking is predominant, because there are few contexts that allow the whole utterance to be presentationally focused. 31 Thus, (68) and (69) are uttered in answer to the following main questions (68Q and 6 9 0 and subquestions (6SSQ1 and 69SQ1).  Non-thematic subjects (68) =(28)  Aoyama-san-wa musuko-ga amerika-ni ryuugaku-si-te-i-ru Aoyama-polite-wa son-nom america-to studying abroad be-pres "As for Mr. Aoyama, his son is currently studying in the U.S."  (68) "  Q: Who is doing what? (or How are your friends doing?) SQ1: How is Mr. Aoyama doing? SQ2: How is Ms. Tanaka doing?... A : As for Mr. Aoyama, his son is currently studying in the U.S. M5 {How is Mr. Aoyama doing} M {{How is Mr. Aoyama doing}, {How is Ms. Tanaka doing},,,}  Negations (69) Context: What has happened to the snow since then? 31 In Japanese, if the nominative-case-marked element is not focused in the main clause, it has to be wa-marked. With the other cases, such as accusative case, the original case marking may be retained.  113 =(35)  (69)"  Yuki-wa hutte-i-na-i (for several hours) snow-wa falling-be-not-pres "As for the snow, it is not falling." Q: What has happened to the snow since then? (explicit) SQ1: What has happened to the snow since then? (SQ2: what has happened to the storm, SQ3:,„) A : As for the snow, it is not falling. Ivf  {What has happened to the snow since then} M {{What has happened to the snow since then}, {what has happened to the storm},,,} In the same vein, we see that both (68) and (69) are congruent because they satisfy the F-condition as well as the T-condition.  Non-restrictive relative clauses Non-restrictive relative clauses are also commonly wa-marked because the clause which modifies the head of the relative clause can establish common-ground information, which is accommodated into the (implicit) main question.32 (70) =(30)  UApec-de Vancouver-o otozurete-i-ta] Hasimoto syusyoo-wa ] Apec-for Vancouver-acc visiting-be-past Hasimoto prime minister-wa "The prime minister Hasimoto, who was visiting Vancouver for Apec, kyoo UBC-de okonaw-are-ta konsinkai-ni sankasi-ta. today UBC-at held-past social gathering-to atttend-past attended a social gathering held at U B C today."  That is, the subject relative clause in (70) establishes the context: "The prime minister Hasimoto was visiting Vancouver for Apec", a part of which is adopted as background information of the implicit question. (70)"  Q: Who did what? (or What did Mr. Hasimoto do?) SQ1: What did Mr. Hasimoto do? (SQ2: What did the vice prime minister do?, SQ3:,„)  Yoko Collier-Sanuki (p.c.) suggests to me that restrictive relative clauses can also establish common-ground information. 32  114  A : The prime minister Hasimoto, (who was visiting Vancouver for Apec,) attended a social gathering held at U B C today.  {What did Mr. Hasimoto do} M {{What did Mr. Hasimoto do}, {What did the vice prime minister do},,,} (70) " is well formed because M f equals the SQ1 and M l induces a strategy to answer the main question.  Indexical elements In utterances containing indexical elements such as (71), the indexical elements are necessarily a part of common-ground knowledge. Thus, implicit questions are formed in the following manner. (71) =(26)  Kyoo-wa kodomo-tati-ga Disneyland-ni it-ta today-wa child-plural-nom Disneyland-to go-past "Today, my children went to Disneyland."  (71)"  Q: What happened when? (What happened today?) SQL What happened today? (SQ2: What happened yesterday?, A : Today, my children went to Disneyland  SQ3:,„)  {What happened today} M {{What happened today}, {What happened yesterday},,,} Again, we see that the question-answer sequence is congruent because M * equals SQ1, which satisfies the F-condition and the answer is not only directly congruent to the SQ1 but also mdirectly congruent to the main question, which satisfies the T-condition.  Cases not accounted for by this model Generic statements and copula constructions that induce property interpretations, can never be presentationally focused. Therefore, elements whose properties are talked about are also wa-marked (unless they are contrastively focused). In these cases, although our analysis may well explain why  115  wa-marking may be used, it does not explain why certain elements must be wa-marked (unless they are focused). I will claim in chapter 4, that wa assignment in these cases is syntactically enforced.  3.4.2. Major properties of wa-marked elements  In this section, I will show that the analysis of wa-marking proposed here can derive some common properties of wa, which have been observed in the previous literature. These are contrastiveness, aboutness, and partially ordered set relations.  Contrastiveness Kuroda (1965) claims that wa inherently marks contrastiveness. This proposal captures native speakers' intuitions that they perceive some kind of implicit contrastive implicature in a proposition containing a wa-marked element. Nevertheless, as we have seen, being contrastive does not distinguish topic from focus because focus also involves contrast. I have shown that wa induces a set of questions. This suggests that contrastiveness, i.e., topic implicature, is induced by the contrast among sets of focus values which are the elements of the topic value.  Aboutness & domain restriction Aboutness defined by Reinhart (1981) is derivable from the relation between a subquestion and main question. Since a subquestion is a "proper" subpart of a main question, [i.e., does not start a new sentence] the main question entails the subquestion. In that case, the main question must be about the topic element specified by the subquestion (Henry Davis p.c).  Poset relations Our analysis of topics also explains why wa-marked elements must be in a Poset relation to the QUD of the main question. However, it does not necessarily mean that wa-marking is used to  116 mark an element which is in a Poset relation with an item which has been evoked in the previous discourse. (72) =(6a)  When I was on surveillance, during this hijacking case, we're working for a newspaper. The guys delivering were selling papers on the side. The newspaper was earning a fortune. These guys knew they were being tailed and they still continued the same shit. People like that you have no sympathy for, they're stupid. (W & P 1991: (16a)) The Japanese translation of the italicized sentences Soo iu yatura-ni-wa doozyoo-no yoti-ga nai that like people-to-wa sympathy-gen room-nom not-exist "People like that you have no sympathy for"  (72)'  ...these guys... accomodation Q: What feeling do I have towards whom (people like that)? SQ1: What feeling do I have towards people who are like that? SQ2: What feeling do I have towards people who are not like that?... A : People like that you have no sympathy for.  As shown in (72)', accommodation is made between the contextually given information and the main question. Thus, cases like so-called thematic use of wa-marking (cf. Kuno 1973) are captured as a subsidiary effect of the semantics of wa-marking.33 Further, it is now evident that this analysis can also explain why referential elements as well as non-referential elements such as preposition-wa, verb-wa adverb-wa, etc. may be wa-marked. Prince (1995) claims that Topicalization in English has the following two discourse functions (Left-Dislocation has the first function only.). 33  Double Discourse Function of Topicalization: 1. "Topicalization triggers an inference on the part of the hearer that the entity represented by the initial NP stands in a salient partially-ordered set (Poset) relation to some entity or entities already evoked in the discourse-model. Posets are defined by a partial ordering R on some set of entities, {e}, such that, for all el, e2, and e3 that are elements of {e}, R is either reflexive, transitive, and antisymmetric, or, alternatively, irreflexive, transitive, and asymmetric. Poset relations include, along with the usual set relations and the identity relation, relations like "is-part-of" and "is-a-subtype-of' but they do not include functional dependency relations, e.g., the "has-a" relation between, say, a house and a door. 2. "Topicalization triggers an inference on the part of the hearer that the proposition is to be structured into a focus and a focus-frame as follows. First, if the entity evoked by the leftmost NP represents an element of some salient set, make that set-relation explicit. Then, in all cases, the open proposition resulting from the replacement of the tonically stressed constituent in the clause with a variable is taken to represent information salienfly and appropriately on the hearer's mind at that point in the discourse, the tonically stressed constituent representing the instantiation of the variable and the new information in the discourse." Our analysis suggests that the First discourse function of topicalization is reduceable to the second function.  117  (73)  Kono mise-ni-wa yoku irassar-u ndes-u ka? this shop-to-wa often come-pres mood-pres Q "Do you often come to this restaurant ?" Iya, sonnani yoku-wa ko-na-i. (#Iya, sonnani yoku ko-na-i.) No, so often-wa come-not-pres "No, I do not come so often." To conclude, the claim that wa creates a set of questions, i.e, a set of sets propositions  explains why wa-marking is used in sentences containing indexical expressions, non-thematic subjects, non-restrictive relative clauses, and negation.  3.4.3. T w o types of B-accents?  In this subsection, I will demonstrate that through wa-marking in Japanese, which is equivalent to B-accenting in English, an extra layer of prominence may be associated with wa-marked elements. This extra prominence signals that the stressed, wa-marked item is a topic in explicit contrast (TEC) as opposed to (simple) topic. However, before I start to show how the theory introduced in the previous section accounts for the Japanese data provided in 3.2., I would like to demonstrate that there are two different types of B-accent. We have seen that Buring identifies B-accents as L H * contours; however, in the following example, a rise from L-tone in the B-accent is less significant. If we pretend that English also allows a mid (M) tone, the B-accent in (74)' may be described as a L M * contour. (The element in (74) " may be (LH*) B-accented but it does not have to be.) (74)"  What did you buy on 59th street? On [fifty ninth street]T, I bought the [SHOES] F  118  (74J)  (The Japanese equivalent of 59) a. Gozyuukyuu-bangai-de-wa, (watasi-wa) kutu-o kai-masi-ta. fifty ninth-street-on-wa, (I-wa) shoes-acc buy-mood-past b. GOZYUUKYUU-BANGAl-de-wa, (watasi-wa) kutu-o kai-masi-ta. fifty ninth-street-on-wa, (I-wa) shoes-acc buy-mood-past  If we compare (59) with (56, 58), we see that the purpose of using a ( L H * ) B-accent is to make the contrast between each set of propositions explicit, as in Buring's implicational topics in (58), contrastive topics in (56), and multiple wh-questions in (60-61). Note also that partial topics may be either (LH*) B-accented or (LM*) B-accented depending on the speaker's intention. That is, a (LH*) B-accent is not essential (if a topic element is identifiable as a member of the superset) unless the speaker's intention is to make an explicit contrast. Thus, I call T-marking with ( L M * ) B-accents a (simple) "topic" and a topic with (LH*) B-accents a "topic in explicit contrast "(TEC). As shown in (74J), the distinction between a topic and a T E C is made in terms of the prominence of wa-marked elements. We see that only the T E C requires extra prominence on wa-marked elements. The same principle applies to generics. Generic subjects do not require extra prominence unless the speaker's intention is to make the contrast explicit. (75) a. John-wa titeki da John-wa intelligent be "John is intelligent." b. JOHN-wa titeki da ga MARY-wa kinben da John-wa intelligent be but Mary-wa diligent be "John is intelligent, but Mary is diligent." The same principle even applies to non-referential elements as shown in (76). (76) a. Context: What can you do in what way? What can you do well? A painter is saying, Mosi zyoozuni siagero to iw-are-reba [zyoozuni-wa] kak-e-ru.  119  if nicely finish quote say-passive-if nicely-wa draw-can-pres "If I am asked to finish nicely, I can draw [nicely]." (Topic) b. A painter is saying, Mosi zyoozuni siagero to iw-are-reba [ZYOOZUNI-wa] kak-e-ru. if nicely finish quote say-passive-if nicely-wa draw-can-pres "If I am asked to finish nicely, I can draw [NICELY]." (TEC) But, it may take more time and require more money... B y (LH*) B-accenting or by adding extra prominence on wa-marked elements, the speaker explicitly reminds the hearer that the elements are in contrast. In actual discourse, non-referential elements are more likely to be used as TECs; however, the use of a (simple) topic is also possible if an appropriate context is provided.  3.5.  E x t e n s i o n o f the a n a l y s i s  In this section, I extend the analysis of wa-marking, explaining why wa-marking is incompatible with universal quantifiers, why wa-marking induces non-exhaustiveness when it is used to answer wh-questions, and why multiple wa-marking induces further contrast.  3.5.1.  W a - m a r k i n g a n d universal  quantifiers  Let us first compare the following two sets of examples. (77) Context: What did your colleagues already do? a. Hotondo-no tomodati-wa sudeni sotugyoosite-simat-ta most-gen colleague-wa already graduating-end up-past "Most colleagues of mine have already graduated." b. Tomodati-no hotondo-wa sudeni sotugyoosite-simat-ta colleague-gen most-wa already graduating-end up-past "Most of my colleagues have already graduated."  120  (78) Context: What did your colleagues already do? a. *Zenin-no tomodati-wa sudeni sotugyoosite-simat-ta all-gen colleague-wa already graduating-end up-past " A l l colleagues of mine have already graduated." " A l l of my colleagues have already graduated." b. ?Tomodati-no zenin-wa sudeni sotugyoosite-simat-ta colleague-gen all-wa already graduating-end up-past " A l l of my colleagues have already graduated." We see in (78a, b) that wa-marking is incompatible with universal quantifiers.34 This incompatibility is captured by our analysis. Let us suppose that the answerer has six colleagues, i.e., Myles, Ping, Nike, Kimary, Tom and John. Recall that wa-marking induces a set of questions. Notice that choosing " A l l of my colleagues have graduated" forces us to eliminate all other options in the topic value, as shown in (78)'. This, however, violates the congruence of a sentence containing a T-marking, which states that some question Q in A 1 must be open to serve as the new QUD. 3 5 (78)'  36  {{Myles has already graduated}, {Ping has already graduated}, {Nike has already graduated}, {Kimary has already graduated}, {Tom has already graduated}, {John has already graduated}}  ^ote that wa-marking is compatible with afloatedquantifier, as shown below. Tomodati-wa zenin sudeni sotugyoosite-simat-ta colleage-wa all already graduating-end up-past "My colleagues have all graduated already." I assume that this is due to the fact that the quantifier in this case is in the comment domain, but not in the topic domain. See chapter 4 for details. See Krifka (1998:p83-84) for a similar observation in English and German.  35  Wa-marking is also incompatible with the Japanese equivalents of sentences containing negatives such as no, no one, and nothing. This is parallel to the use of B-accents on negative quantifiers in English. 36  Dare-mo ko-na-katta someone-pol come-not-past "No one came." *Dare-mo-wa ko-na-katta someone-pol-wa come-not-past "*[No one]T came."  121 In (78)', following Biiring's notation, I have crossed out all the values of A ' which have been eliminated by the universal quantifier. It can be seen that there is no open question available to serve as the new QUD. On the other hand, with the use of "most", for instance, other elements, i.e., Tom and John are still disputable. (77)'  {{Myles has already graduated}, {Ping has already graduated}, {Nike has already graduated}, (Kimary hao already graduated}, {Tom has already graduated}, {John has already graduated}}  3.5.2. Scope inversion and stressed WA-marking  As shown in (79), L H * B-accents, (not L M * B-accents), are only used for T E C , not (simple) topics: (79)  Context: What did you do today? a. In KITS, I bought a shirt, and in DOWNTOWN, I saw a movie. b. #In Kits, I bought a shirt, and in downtown, I saw a movie.  Jacobs (1997) claims, however, that true contrastive topics (which are signalled by a H L H * contour) are even more restricted in their distribution. 3 7 (80)  Krifka (1998: p85) A : What about Maria and Hans? What did they read? (Who read which novel?) B: Ma/RIa hat den Schatz imNSILbersee gelesen und /HANs deWTNnetou. Maria has the Treasure in-the Silver-Lake read and Hans the Winnetou "Maria read The Treasure in the Silver-Lake and Hans read Winnetou." B':  ??MaVRIa hat den Schatz imNSILbersee gelesen undV H A N S denWINnetou.  Jacobs claims that contrastive topics can occur only in clauses that express illocutionary acts, and not in relative clauses. Furthermore, they are restricted to assertive and directive sentences. 37  122  Jacobs (1997) claims that contrastive topic constructions involve a focus within the topic constituent that is realized by a rise accent, or rather by a slight fall followed by a strong rise. (The second accent is a regular A-accent). He also claims that this contour (which he calls "a root contour" ( H L H * or V)), but not a B-accent (LH*), induces scope inversions. 3 8 I will call Jacobs' root contour C-accent for convenience. In (80), we can observe that the T E C (LH*) is acceptable as in (80B) but not the contrastive topics (HLH*) as in (80B'). 39 Jacobs (1997) and Krifka (1998), moreover, claim that only contrastive topics (which contain focus), but not TECs, induce scope inversion. (81)  Krifka (1998: p.80 (16b)) Mindestens ein Student hat jeden Roman gelesen 3 V at least one-nom student has every-acc read "There is at least one student who read every novel." Mindestens /EIN Student hat\ JEden Roman gelesen V 3 at least one-nom student has every-acc read "Every novel is read by at least one student."  In Japanese, the equivalent of a L H * B-accent (TEC) is signalled by stressed XP-wa. In contrast, the equivalent of a H L H * contour is indicated by stressed WA-marking (or sometimes by stressed XP-wa). 4 0 (80J)  A : What about Maria and Hans? What did they read? (Who read which novel?) B: MARIA-wa The Treasure in the Silver-Lake-o yon-da ga Maria-wa The Treasure in the Silver-lake-acc read-past but HANSU-wa Winnetou-o yon-da Hans-wa Winnetou-acc read-past "Maria read The Treasure in the Silver-Lake and Hans read Winnetou."  ^Prior to Jacobs (1997), it was assumed that B-accents (LH*) were responsible for scope inversions. Jacobs (1997) calls it I- (Intonational) topicalization.  39  "°Note, however, that a TEC cannot be marked by XP-WA.  123  B': #Maria-WA The Treasure in the Silver-Lake-o yon-da ga Maria-wa The Treasure in the Silver-lake-acc read-past but Hans-WA Winnetou-o yon-da Hans-wa Winnetou-acc read-past "Maria read The Treasure in the Silver-Lake and Hans read Winnetou." ??MaVRIa hat den Schatz imNSILbersee gelesen undV HANS denXWINnetou. Moreover, parallel to the German use of the C-accent, the use of stressed WA-marking is necessary in order to induce scope inversion. (81 J) Sukunakutomo hitori-wa/ga dono syoosetu-mo yonde-i-ru at least one-wa/nom every novel-also reading-be-pres "There is at least one student who read every novel. Sukunakutomo hitori-WA dono syoosetu-mo yonde-i-ru at least one-wa every novel-also reading-be-pres "Every novel is read by at least one student. Thus, stressed WA-marking corresponds to a C-accent, just as wa-marking corresponds to a B-accent.  41  4 1  A r e l a t e d f u n c t i o n o f s t r e s s e d W A - m a r k i n g i s to i n d u c e a n o n - e x h a u s t i v e f o c u s . T h e d i f f e r e n c e b e t w e e n e x h a u s t i v e  f o c u s a n d n o n - e x h a u s t i v e f o c u s is illustrated i n (i) a n d ( i i ) . N o t e that i n (i) e x h a u s t i v e f o c u s is e x p r e s s e d b y g a - m a r k i n g w h e r e a s i n ( i i ) , n o n - e x h a u s t i v e f o c u s is e x p r e s s e d b y s t r e s s e d W A - m a r k i n g . (i)  S p e a k e r A : W h o is s t u d y i n g l i n g u i s t i c s ? Speaker B:  [ T o m to Mary-ga] gengogaku-o benkyoosite-iru. F  [Exhaustive Focus (EF)] T o m and M a r y - n o m linguistics-ace study-ing " ( O n l y ) T o m a n d M a r y are studying linguistics."  (ii)  S p e a k e r A : W h o is s t u d y i n g l i n g u i s t i c s ? Speaker B:  [ T o m - W A } gengogaku-o benkyoosite-iru... p  [ N o n - E x h a u s t i v e F o c u s (NEF)1 T o m - w a linguistics-acc study-ing " ( A t least) T o m is s t u d y i n g l i n g u i s t i c s . " O b s e r v e t h e d i f f e r e n c e b e t w e e n g a - m a r k e d a n d w a - m a r k e d c o n t r a s t i v e f o c u s , (i) i s f a l s e i f J o h n o r J o e i s a l s o s t u d y i n g l i n g u i s t i c s ' ; T o m a n d M a r y h a v e t o b e t h e o n l y m e m b e r s w h o a r e s t u d y i n g l i n g u i s t i c s i n t h e d i s c o u r s e - s p e c i f i e d set. O n t h e c o n t r a r y , w a - m a r k e d c o n t r a s t i v e f o c u s , a s s h o w n i n ( i i ) , d o e s n o t r e q u i r e the f o c u s e d e l e m e n t t o b e t h e o n l y m e m b e r o f that set. I n o t h e r w o r d s , a l l m e m b e r s w h i c h s a t i s f y t h e c o n d i t i o n h a v e t o b e l i s t e d e x h a u s t i v e l y i n ( i ) b u t n o n - e x h a u s t i v e l y i n ( i i ) . I n o t h e r w o r d s , ( i ) m a y b e i d e n t i f i e d as E x h a u s t i v e F o c u s ( E F m a r k e d b y a c a s e m a r k e r ) a n d ( i i ) as N o n - E x h a u s t i v e F o c u s ( N E F , m a r k e d b y  wa).  It i s p o s s i b l e that s t r e s s e d W A ( a n d its C - a c c e n t c o u n t e r p a r t i n G e r m a n a n d E n g l i s h ) c a n b e t h o u g h t o f as f o c u s i n g a T - m a r k e d c o n s t i t u e n t . I n that c a s e , u n d e r B i i r i n g ' s ( 1 9 9 8 ) m o d e l , s t r e s s e d W A w i l l c o n v e r t a s u b q u e s t i o n  124  3.5.3. Multiple wa-marking and contrastiveness  In this section, I will show that the position of wa-marking delimits the scope of focus and thus has truth conditional effects. First let us consider the following example. (82)  Taroo-wa musuko-wa tamani-wa suppotu-wa si-wa su-ru Taroo-wa son-wa sometimes-wa sport-wa do-wa do-pres "As for Taro, as for his son, sometimes, as for sports, he DOES them."  (82) shows that a sentence may contain more than one wa-marking. Moreover, the position of wa-marking delimits the scope of focus. (83) a Bankuubaa-wa saamon-wa oisi-i Vancouver-wa salmon-wa tasty-pres "As for Vancouver, as for its salmon, they are tasty." b Saamon-wa Bankuubaa-wa oisi-i Salmon-wa Vancouver-wa tasty-pres "As for salmon, as far as Vancouver is concerned, they are tasty." (84) a Taroo-wa hon-wa yoku yom-u Taroo-wa book-wa often read-pres "As for Taro, as for books, he often reads them." b Hon-wa Taroo-wa yoku yom-u book-wa Taroo-wa often read-pres "As for books, as for Taro, he often reads them." We also see that the examples in (83) and (84) (generic statements) do not have the same interpretations. For instance, (83a) is a generalization about Vancouver and salmon in Vancouver; (83b), in contrast, is a generalization about salmon and salmon in Vancouver. That is, (83a) would be paraphrased as "Regarding Vancouver, there are many good things, but speaking of salmon, they are good." whereas (83b) would be paraphrased as "Regarding salmon, there are many places where they are good, but as far as Vancouver's salmon) is concerned, they are good." Hence, there is a containment relation between the first wa-marked element and the second wa-marked into a main question, thus forming the answer to accommodate a new QUD (Henry Davis p.c).  125  element. In fact, some speakers may feel that the second wa-marked element is more contrastive.42 This is because the set of questions induced by the second wa-marking is derived from the set of questions induced by the first wa-marking. Thus, as shown in (82), the more deeply the wa-marking is embedded, the more contrastiveness is implied. In other words, the narrower the domain of predication is, the more explicit the contrast becomes. Nevertheless,time/locationelements, which set up a stage where an event takes place do not induce strong contrastiveness, as shown in (85) and (86). (85)  Gosyuzin-wa saikin-wa doo site-irassyai-mas-u ka? husband-wa these days-wa how doing-be-mood-pres Q "As for your husband, as for these days, how is he doing?"  (86)  Taroo-wa gakko-de-wa ninki-ga ar-u Taroo-wa school-at-wa popularity-nom exist-pres "As for Taro, at his school, he is popular there."  Moreover, some speakers perceive more contrastiveness when wa-marked elements are non-referential, (e.g. PP + wa, adverb-wa, verb +wa..) Tookyoc-wa Taroo-ga iki-masi-ta Tokyo-wa Taro-nom go-mood-past Tookyoo-e-wa Taroo-ga iki-masi-ta. Tokyo-to-wa Taro-nom go-mood-past. It has been also claimed that wa-marking in an embedded clause requires a contrastive interpretation. However, this is only true in some types of embedded clauses, i.e., in a subordinate clause adjoined lower than an IP. For instance, in the following example, the use of wa-marking is perfectly fine and does not force a contrastive interpretation. [Hasimoto-syusyoo-wa ima Kanada-o otozurete-i-ru to] Taroo-ga itte-i-ta. Hasimoto-prime minister-wa now Canada-acc visiting-be-pres quote Taroo-nom saying-be-past "Taroo was saying that the Prime Minister Hashimoto is now visiting Vancouver." When wa-marked elements occur in DPs, such as in relative clauses, however, a contrastive interpretation becomes obligatory. [Kanada-wa otozurete-i-ta Hasimoto-syusyoo]-wa kyoo buzi-ni Nihon-ni kikokusi-ta Canada-wa visiting-be-past Hasimoto-prime miniser-wa today safely Japan-to return-past "The Prime Minister Hashimoto, who was visiting at least Canada, returned to Japan safely." I can think of two reasons why a contrastive interpretation is forced. First, as observed in Kuno (1973), a relative pronoun, which is an empty operator in Japanese, is a topic; however, the clause containing the topic is embedded in the main clause. Thus, the situation is analogous to (83) and (84). Secondly, in order to indicate that an element is in a relative clause, the clause has to be uttered in one MaP; otherwise, the wa-marked element will be perceived as a main clause constituent. Thus, a contrastive interpretation is also forced prosodically. However, in the appositive construction, which does not involve an empty operator, contrastiveness is reduced. Note that in the following example, the subordinate clause is a generic statement. [Hasimoto-syusyoo-wa maiasa kami-ni pomaado-o tuke-ru koto]-wa yoku sir-arete-i-ru. Hasimoto-prime minister-wa every morning hair-to pomade put-pres-nominalizer-wa well know-passive-be-pres "It is well-known that the Prime Minister Hashimoto applies pomade to his hair every morning."  126  Similarly, the example in (87), in spite of containing several wa-marked elements, does not induce strong contrastiveness compared to (82). This is due to the fact that this is a generic statement about Taroo (thus, wa-marking is required), the graduate school is a location, and linguistics is the subject of negation and thus, is likely to be background information when the sentence is uttered. (87)  3.6.  Taroo-wa daigakuin-de-wa gengogaku-wa zyukkosite-i-na-i Taroo-wa graduate school-at -wa linguistics-wa taking a class-be-neg-pres43 "As for Taro, at his graduate school, as for linguistics, he is not taking classes. Alternative analyses  In this section, I will compare the approach proposed here with two others: the old/new distinction proposed by Prince (1992) and Inoue (1982) and an approach based on definiteness or specificity.  3.6.1. Old/new information and  wa-marking  In this subsection, I will show that the distribution of wa-marking cannot be predicted in terms of the information-based approaches proposed by Prince (1992) and Inoue (1982). Prince (1992) proposes that the information represented by an utterance is classified into two distinct divisions: H E A R E R - N E W / O L D and DISCOURSE-NEW/OLD. This system allows Ono (1990) claims that a generic judgment like "Whales are mammals" sets the class of whales against the whole set of living creatures, minus whales.  43  Examples from Ono (1990) a.  Kuzira-wa honyuurui da "A whale is a mammal."  b. Taroo-wa gakusee da "Taro is a student." c.  Gonin-wa kita. "Five (of them) came."  Ono claims that in terms of figure-ground opposition, the vaguer and broader the ground is, the less salient the figure becomes. In (a), where "of X" is conventionally definable by our knowledge of the world, the contrastive effect almost fades out. (cf. Inoue 1982 for a similar proposal)  127  us to establish a category in which the information is N E W in discourse but N O T N E W to the hearer. (88)  Hearer Discourse  I I  old old  old new new new 44  wa/ga wa/ga *wa/ga This system, nevertheless, cannot predict the distribution of wa-marking, because the only situation where the use of wa-marking is not possible is in the Hearer NEW-Discourse N E W combination. Thus, in the Hearer OLD-Discourse O L D combination and the Hearer OLD-Discourse N E W combination, either wa-marking or ordinary case-marking may be used, depending on the context. Inoue (1982), furthermore, utilizes the notion of "KNOWN", which indicates that something is known to the speaker and to people other than the hearer (cf. Prince 1981). However, this system cannot help distinguish wa-marking from ordinary case marking either because, again, the use of wa-marking is impossible only in the combination of Speaker NEW-Hearer NEW-Discourse NEW, but, in the other combinations, either wa-marking or ordinary case-marking may be used.  (89)  speaker hearer discourse  I I I  old old old  old old new  old new new new new new45  4 6  wa/ga wa/ga wa/ga *wa/ga  ^The combination of Hearer new and Discourse old is not possible. 4S  The following combinations are not possible.  speaker hearer discourse  old new old  new old old  new old new  new new old  ^Condoravdi (1994) has developed Strong/Weak Novelty and Strong/Weak Familiarity distinctions based on Heim's Novelty/Familiarity distinctions; however, this system, too, cannot predict the distribution of wa-marking, because it is based on definiteness. In 3.6.2., we have seen that the use of wa-marking is possible irrespective of (in)definiteness and (non-)specificity.  128 3.6.2. Deflniteness/specificity  and wa-marking  Ward & Prince (1991) show that topicalization is possible irrespective of the definiteness and specificity of the topicalized DP. They demonstrate that, in English, for DPs to be topicalized, they do not need to be definite. I will show that the same principle also holds for wa-marking. That is, wa-marking is possible irrespective of definiteness and specificity. In (92B) and (93A2), we see that a non-specific indefinite DP and a specific definite DP are topicalized, respectively. (90) Non-specific Definite  A : Which part of the procedures do you dislike most? B: The job talk, I like. But the interview makes me nervous. BJ: Zyobutooku-wa sukida. 4 7 Demo mensetu-wa kintyoosuru Job talk-wa like but interview-wa nervous-be "In general, the Job talk, I like. But the interview makes me nervous."  (91) Specific Definite  A l : I'll have to introduce one new principle in today's class. A2: But, this principle, I will introduce after the break. A2J: Kono genii-wa, bureeku-no atode syookai-simasu. This principle-wa, break-gen after introduce-do "This principle, I will introduce after the break."  (92) Non-specific Indefinite A : Do you think you'd be more nervous in a job talk or a job interview? (W&P 1991: pl68) B: A job talk, I think you'd have somewhat more control over. BJ: Zyobutooku-wa mootyotto nantokanaru to omou Job talk-wa somewhat manageable comp think prt " A job talk, I think you'd have somewhat more control over." (93) Specific Indefinite  A l : I'll have to introduce two principles. (W&P 1991: p 168) A2: One I'm going to introduce now and one I'm going to introduce later A2J: Hitotu-wa ima syookai-site, moo hitotu-wa atode syookaisimasu. One-wa now introduce-do, another one-wa later introduce-do "One I'm going to introduce now and one I'm going to introduce later"  (J) represents the Japanese equivalent of an English example. For instance, (91BJ) is the Japanese equivalent of (91B).  47  129  This is also true for wa-marking in Japanese. In (92BJ) and (93A2J), we see that elements need not be definite or specific to be marked by wa in the sentence-initial position. Thus, wa-marked elements need not be definite, contrary to what is often claimed in the literature. See, for instance, Kuroda (1992) and Kuno (1973). (94)  According to a new report in Japan, Syoonen-no san-nin-ni hitori-wa naihu-o motte-ir-u koto-ga hanmeesi-ta youth-prt three-cl-per one-wa knife-acc owning-be-pres nominalizer-nom reveal-past "It turns out that one in three teenagers owns a knife."  Thus, it is not the case that wa-marking itself cannot mark indefinites. Rather, the discourse contexts, in which indefinites are used are less likely to support the contextual restrictions necessary for the use of wa-marking. 48 (95)  Context: There is heavy snow. Teachers are talking about whether they should close the school or not. One teacher says, Suunin-wa gakkoo-ni kite-i-ru yoodes-u several-wa school-to coming-be-pres seem-pres "It seems that several students have already come to school. But, another teacher reports, Sudeni, hitori-wa kaette-simat-ta yoodes-u. already, one-wa returning-end up-past seem-pres "It seems that one already went home."  Thus, given an appropriate context, the use of wa-marking with indefinites (whether specific or non specific) is possible, as in (94) and (95). However, as shown in (96), wa-marking of indefinites is impossible if the indefinites cannot be in a partially ordered set relationship with any ^Kuroda (1992) uses specificity (specific events) only in contrast to genericity (generic events). Thus, his definition of specificity is different from the one used by Kennedy (1993: p 93), for instance, who analyzes specificity as an effect of speaker use. "A speaker uses a specific NP to talk about an individual whose existence he presupposes, and he uses a non-specific NP when the existence of a constant value for that NP is not presupposed." Moreover, applicability of wa-marking to the NPs in (90-93) is the same even if we adopt different definitions of specificity such as those of Enc (1991) or Ludlow and Neal (1991).  130  previous contextual knowledge, i.e, the hearer's beliefs. 49 Ward & Prince (1991: p 176) also claim that constituents introducing brand new information cannot be topicalized in English. The same holds true for wa-marked constituents in Japanese. (96)  #Ippiki-no inu-wa hasitte-i-ru.50 one-gen dog-nom running-be-pres  Without any prior knowledge, it is not possible to utter (96) as an answer to "What is happening?". However, the fact that the wa-marked elements cannot be used to start presentational focus sentences does not mean that wa-marking cannot be used with (non-specific) indefinites. That is, it is not the case that indefinites cannot be wa-marked, as shown in (97B), but rather, wa-marked elements cannot occur in the sentence-intitial position of presentational focus sentences, as shown in (98B). (97) a. Context: The speaker (A) has realized that he forgot his textbook in the department office. He is asking (B) if there is still someone in the department so that he can ask that person to keep the textbook for him. A : Gakubu-ni mada dareka i-mas-u ka ne? deparment-in still somebody be-mood-pre Q pit "Do you think if there is still somebody at the departmemt?" B: Mada dareka-wa i-ru desyoo. still somebody -wa be-pres will "Somebody should be still there." or Mada dareka i-ru desyoo. still somebody -wa be-pres will "There should be somebody there still." "'Referring to Prince's (1981: p237) taxonomy, shown below, Hinds (1987: p87) also claims that unanchored (brand-new) NP subjects are only marked by ga-marking, but not by wa-marking. Prince's assumed familiarity itaimng inferrable brand-new (unanchored)  containing inferrable  textually evoked  ionally evoked  brand-new anchored  Extending the strong/weak distinction as proposed by elements must be strong.  Milsark (1974, 77), Ogihara (1984) claims that wa-marked  131  b. A : Kanada-de nanika konsaato-ni iki-masi-ta ka? Canada-in something concert-to go-mood-past Q "Did you go to any concerts in Canada?" B: Ee, nankoka-wa iki-masi-ta. yes, some-wa go-mood-past "Yes, I went to some." or Ee, nankoka iki-masi-ta. yes, some-wa go-mood-past "Yes, I went to some." (98)  What happened? a. #Sir-ana-i hito-wa totuzen hanasikakete-ki-ta nda. know-not-pres person-wa suddenly speaking-come-past mood b. Sir-ana-i hito-ga totuzen hanasikakete-ki-ta nda. know-not-pres person-wa suddenly speaking-come-past mood "I was suddenly spoken to by a stranger."  Thus, I conclude that the use of non-specific indefinite wa is simply hmited contextually because (non-specific) indefinites are most often used to introduce new discourse entities/referents, and therefore, it is hard to find common-ground knowledge from which a topic can be created. As is often observed in the literature, moreover, the use of stress on wa-marked indefinites also makes the sentence interpretable. This is because the stress guarantees the existence of a (super) set with which the wa-marked element may be contrasted. (99)  Hakasegoo-o toru-ni-wa saitee sannen-wa kakar-u. Ph.D-acc receive-to-wa at leat three years-wa required-pres "In order to get a Ph.D., a student must study at least three years."  This situation is analogous to the cases of partitives and answers to wh-questions, in which the use of definites is most frequent but the use of indefinites is rare but is not impossible. (100^a.San-nin-no kodomo-wa gengogaku-o benkyoositeiru. three-cl-prt children-wa linguistics-acc studying-be-pres "(At least) THREE children are studying linguistics." (partitive reading)  132  b. Context: Who is studying linguistics? Kodomo-tati-ga gengogaku-o benkyoositeiru. three children-nom linguistics-acc studying-be-pres " A G R O U P OF C H I L D R E N are studying linguistics."  3.7. o-marking vs. wa-marking  We have seen that in English, T-marked elements induce a set of questions. We have also seen that the Japanese equivalent of T-marking is wa-marking. I now turn briefly to the distinction between wa-marked and o-marked (accusative case marked) elements. Compare the following examples. (101)  A : Who ate the cake? B: John-ga keeki-o tabe-ta. John-nom cake-acc eat-past "John ate the cake."  (102)  A : Who ate the cake? B: Keeki-wa John-ga tabe-ta. cake-wa John-nom eat-past "The cake, John ate."  (103)  A : Who ate the cake? B: #Keeki-o John-ga tabe-ta. cake-acc John-nom eat-past  Comparison of (103) with (101) and (102) indicates that fronted o-marking in (103) has a discourse function distinct from either in-situ o-marking as in (101), which contains a simple (focus only) question-answer pair, or wa-marking (in (102)), which induces a set of questions. Interestingly, if the question is not explicitly provided in the previous discourse, i.e., if the speaker can create an implicit question, fronted o-marking as well as wa-marking becomes possible. (104)a. Amai mono-wa John-ga takusan tabe-ru sweet thing-wa John-nom much eat-pres  133  b. Amai mono-o John-ga takusan tabe-ru. sweet thing-acc John-nom much eat-pres I leave the distinction between wa-marking and o-marking in these cases for further investigation.  3.8.  Sentence topics without wa-marking  Thus far, we have discussed the semantics and the discourse function of wa-marking. However, there are some cases in which wa-marking is absent. Consider the following examples. (105)a.Context: A Japanese student has finally decided to study abroad. Ore, Amerika-ni ik-u koto-ni si-ta. I, America-to go-pres nominalizer decide-past "I decided to go the U.S." b. Context: After corning back from one year of studying abroad, a female student is saying, Watasi, kono itinen-de hontooni kawat-ta tte omo-u no. I, this one year-in really change-past quote think-pres prt "I think that I have really changed in the past year." c. Context: Grandfather looks lonely since yeserday. Something happened yesterday. Kinoo, Mago-ga Tookyoo-ni kaet-ta nda. yesterday, grandchild-nom Tokyo-to return-past mood "Yesterday, his grandson went back to Tokyo." d. Context: After the speaker saw a commercial for a new Nissan model, Kondo-no Nissan, watasi-wa sukosi kiniitte i-mas-u. this time-prt Nissan, I-wa a little like-be-polite-pres "This year's Nissan, I somewhat like." e. Context: A host found out that his guest came by car after he had arranged for a taxi for his guest. The host said, Takusii, kaette moratte. taxi, return having "Have the taxi leave." Hasegawa (1993) and Maruyama (1996) claim that 0-marking is not just a result of wa-drop but has an independent discourse function. Maruyama also shows that 0-marked elements are still sentence topics because the 0-marked element in (106c) as well as the wa-marked element in (106b) do not induce negation of the whole event. Compare the following examples.  134  (106)a. Context: What happened at the birthday party? Keeki-o Mary-ga tabe-na-katta. (The whole event is negated) cake-acc Mary-nom eat-neg-past "It is not the case that Mary ate the cake." b. Context: What happened at the birthday party? How was the cake? Keeki-wa Mary-ga tabe-na-katta. cake-wa Mary-nom eat-neg-past "As for the cake, (as opposed to the pudding, etc.) Mary did not eat it." "The cake, Mary did not eat." c. Context: What happened at the birthday party? How was the cake? Keeki, Mary-ga tabe-na-katta. cake-0 Mary-nom eat-neg-past "As for the cake, Mary did not eat it." I claim that a topic without wa-marking is exploited to avoid topic implicature. Thus, 0-marking appears to be possible in cases which do not induce any implicit contrast, as opposed to wa-marking, which always induces a set of alternatives. Note that 0-marking is particularly frequent with indexicals and generics, neither of which involve contrastive nor presentational focus. 5 1 Thus, 0-marking does not involve merely the deletion of wa-marking^ but has a different discourse function. The following data by Shibatani support this observation. Shibatani (1990) claims: "The clearest instance of this involves utterances appealing to the hearer expressing the speaker's internal feeling." (p 368) (107)  Watasi samisii no. I lonely pit "I feel lonely." (female speech)  Takubo (1990) observes that "wa" cannot mark unidentifiable objects.  51  A:  Kinoo Tanaka-ga ne, yesterday Tanaka-nom pit, "Yesterday, Mr. Tanaka..."  B: Tanaka-tte/*wa dare? Tanaka-quote/wa who "Who is Tanaka?" A: Tanaka-to iu-no-wa gengogakukooza-no zyosyu da. Tanaka-quote-say-nominalizer-wa linguistics division-prt assistant be "(The person called) Mr. Tanaka is an assistant in the linguistics division."  135  Ore uresii nda. I happy mood "I feel happy." (male speech) Shibatani also claims that, " A neutral descriptive reading does not obtain, since what is expressed are internal feelings that can be felt only by the speaker who cannot describe them as an outside observer. Thus, utterances of this type, which are a direct expression of the speaker's internal feeling cannot mark their subject with any particle." However, Shibatani's analysis does not explain the obligatory use of 0-marking in (105). We can, however, extend our explanation from (105) to (107), since the subjects in (107) are indexical elements. As such, they could only receive a contrastive focus reading with ga-marking and a (contrastive) topic reading with wa-marking; neither interpretation is associated with (107).  3.9.  Summary  In this chapter, I have reviewed the distribution and interpretation of wa-marked elements in Japanese. I have shown that their use corresponds closely to that of the so-called B-accent in English. Adopting Biiring's (1998) model of topic-focus structure, I have claimed that wa induces a set of subquestions one of which must be congruent to the main question under discussion of a discourse.  Appendix: Kuroda's mini-topics  Kuroda (1992) claims that DPs and PPs may also have their own topics, which he calls "mini topics." (1)  Kuroda (1992: p37). Tanaka-ga ano kaigi-ni Huransu-zin-wa gengogakusya-o yon-da. Tanaka-nom that meeting-to French-person-wa linguist-acc invite-past "Mr. Tanaka invited linguists, so far as the French are concerned, to that conference,..."  136 Tanaka-ga ano kaigi-de Huransu-zin-ni-wa gengogakusya-ni at-ta. Tanaka-nom that meeting-at Freanch-person-to-wa linguist-with meet-past "Mr. Tanaka met linguists, as far as the French are concerned, at that conference,..." Kuroda considers "Huransu-zin-wa gengogakusya" and "Huransu-zin-ni-wa gengogakusya" DPs, which are embedded in the main clause. He states that: "In the above examples, the mini-topicalized DPs associate French with linguists and contrast it with Americans, Koreans, and so forth. What is associated with Americans or Koreans is not expressed." (p.39) French psychologists linguists philosophers.... American psychologists linguists philosophers Korean psychologists linguists philosophers...  Kuroda (p. 40) further claims that: "The function of a mini-topicalized DP is to indicate that a particular pairing, say, (the French, linguists) is selected as a reality, from among other possible pairings, and then the mini-topic is contrasted with other possible mini-topics given or implied in the pragmatic context." 52 This suggests that wa-marking may be used in presentational focus if it is embedded. Of course, some knowledge of the characters in discourse is essential for using mini topics, i.e., their use is different from a sentence like " A man came into the room", which does not require any information from the previous discourse. Nevertheless, it is also true that presentational focus may reintroduce entities evoked in the previous discourse. Thus, we assume that presentational focus may contain wa-marked elements if they are embedded as mini-topics, as shown in (2) and (3). (2)  Context: A family was going for a ski trip, and their child insisted that he wanted to help his parents. Everything was set, and they left. But, after they had arrived, guess what happened? Kodomo-ga kuruma-ni zibun-no omotya-wa tun-da child-nom car-to self-gen toy-wa load-past "The child loaded his toys to the car,  If the value of the predicate is fixed, the use of wa-marking becomes awkward.  !  137 motiron bentoo-mo tun-da of course lunch-also load-past and of course loaded the lunch, too demo kanzin-no sukii-wa wasurete i-ta but the most important ski-wa forgeting-be-past but, forgot skiis, which are the most important." (3)  Context: M y family subscribes to two newspapers, but we heard that one or both of the newspaper companies would be on strike. Then, next morning, I saw somebody delivering newspapers. I said, Arubaito-no gakusee-ga Nyuuyooku Taimuzu-wa haitatusite-i-ru yoo da. part time-gen student-nom New York Times-wa delivering-be-pres seem be "It seems that student part-timers are delivering as far as the New York Times is concerned."  138  Chapter 4: Toward syntax-Information structure mapping in Japanese  4.1.  Introduction  The goal of this chapter is to show that syntactic structure in Japanese reflects topic-comment structure. In 4.2., I first provide some syntactic evidence that topic elements, including wa-marked elements, must be located outside the domain of IP. In 4.3., I then show that there is a direct correspondence between S-structure positions and semantic interpretation in Japanese. I review Diesing's (1992) mapping hypothesis, a pioneering work which proposes an explicit mapping between syntax and semantics. I then argue against a mapping between syntax and L F and for a mapping between syntax and information structure. In particular, I propose that the mapping takes place between syntax and topic-comment structure (cf. Erteschik-Shir 1996, Jager 1997 for similar proposals).1  4.2. Japanese phrase structure and the positions of subjects  In this section, I establish that different types of subjects occupy different syntactic positions. In 4.2.1., I start with a review of Kuroda's (1988) VP-internal subject hypothesis and show that (stage-level) non-thematic subjects (=non-theta-marked subjects or major subjects) are located in Spec IP whereas (stage-level) theta-marked subjects occupy Spec V P . I then review Diesing (1988) and show that the subjects of individual-level predicates occupy a syntactically higher position than the subjects of stage-level predicates in Japanese. Finally, I review Nakayama & Koizumi (1991), who demonstrate that the subjects of (stage-level) transitive verbs and unergative verbs occupy higher syntactic positions than those of (stage-level) unaccusative subjects. I then unify these claims and propose that non-thematic subjects of stage-level predicates occupy Spec IP, 'Erteschik-Shir (1996: p4) also argues that "f-structure and not LF is the input to a semantic rule of predication."  139 subjects of stage-level transitive and unergative verbs occupy Spec V P (ext), and subjects of stage-level unergative verbs occupy Spec V P (int) as summarized in (l). 2 (1) Japanese (Head-final)3  stage-level non-thematic stage-level transitive unergative stage-level unaccusative  V  Then, in 4.2.2., I demonstrate that the subjects of individual-level predicates must be located outside the domain of IP.  4.2.1. Japanese phrase structure  Non thematic subjects vs. theta-marked subjects Kuroda (1988) proposes the VP-internal subject hypothesis, observing that in English (which has subject agreement in number and gender), subjects must move into Spec IP to receive nominative Case to fulfill agreement.  I a s s u m e r e c u r s i v e V P - s t r u c t u r e s . V P ( e x t ) i d e n t i f i e s the h i g h e s t s p e c i f i e r a n d V P ( i n t ) i d e n t i f i e s t h e l o w e s t s p e c i f i e r  2  i n a r e c u r s i v e V P structure.  ' i n t h i s c h a p t e r , I o m i t K P n o d e s f o r ease o f e x p o s i t i o n .  140  (2) English  In contrast, in Japanese (which lacks subject agreement) there is no legitimate reason for subjects to move into Spec IP to be case-marked, and subjects, thus, remain in situ. (3)Japanese  nanika-o kaw-ta something bought In particular, since the case marker ga is not assigned by INFL, we lack an initial direct motivation to assume that subjects raise to Spec IP, as in English. Note also that Spec V P is a theta-marked position whereas Spec IP is a non-theta-marked position. According to Kuroda's (1988) VP-internal subject hypothesis, Japanese allows double subject constructions precisely because it is a "non agreement forcing language". Following Kuroda (1988), I assume that non-thematic subjects (adjuncts) as in (4) occupy Spec IP. 4 5 (4) a. Context: While the speaker and hearer are driving toward a mountain, the speaker says, [jpYama-ga [ypki-ga kiree des-u]] ne (cf. Kuno 1973) mountain-nom tree-nom beautiful be-pres prt "Trees look beautiful in the mountain."  "Takahashi (1994) also claims that major subjects (=non-thematic subjects/adjuncts) are located in Spec D?. Kuroda (1992) also claims that scrambled objects can occupy Spec IP in Japanese.  5  141  b. [jpTanaka-san-ga [ kodomo-ga infuruenza-ni kakatte-i-ru]]. Tanaka-Mr/Ms.-nom child-nom flu-by having-be-pres "Tanaka's child has the flu." vp  c. Context: John fainted thirty minutes ago. But, now... [jpJohn-ga [vpki-ga tui-ta]]. John-nom consciousness-nom become-past "John has become conscious." d. Context: My car broke down. A mechanic explains to me, [jpEnzin-ga moo [ypgata-ga kite-i-mas-u]] ne. engine-nom already damage-nom becoming-be-mood-pres pit "The engine has already started to become damaged." Notice that sentences containing this type of subject can be used in out-of-the blue contexts, i.e., they can be used to answer questions such as "What happened?, What is the matter?" etc. This suggests that the whole sentence is presentationally focused. Let us now turn to examples which contain thematic subjects.  6  (5) a. John-ni Huransugo-ga wakari-mas-u ne John-for French-nom understand-mood-pres prt "John understands French." (For John, French is understandable/understood.) b. John-ga Huransugo-ga wakari-mas-u ne John-nom French-nom understand-mood-pres prt "John understands French." (6) a. John-ni ryoori-ga deki-mas-u ne John-for cooking-nom capable-mood-pres prt "John can cook." (For John, cooking is doable.) b. John-ga ryoori-ga deki-mas-u ne John-nom cooking-nom capable-mood-pres-prt "John can cook." As shown in (5b) and in (6b), the sentences containing ga-marked subjects of stative verbs (the Saito (1982) claims that the first ga-marked elements such as (5b) and (6b) are not theta marked. He, thus, assumes that verbs such as "to understand" and "to be able" are not capable of assigning nominative case, and proposes that the subjects of these predicates receive default case, which is also marked by ga. A similar assumption will be adopted later in this chapter. Notice, moreover, that these verbs are not capable of assigning accusative case to the objects "French" and "cooking", which thus also receive default case-marking. 6  142  first ga-marked elements) do not have to be contrastively focused. The whole sentence may be presentationally focused. I thus conclude that regardless of whether or not the subject is theta-marked, the whole sentence containing it may be presentationally focused as far as the sentence can describe a temporary state, yielding a state description interpretation (cf. Ladusaw 1994).7 In contrast, individual-level predicates denoting properties cannot be presentationally focused, as shown below. (7)  Context: What is happened?/ What is the matter? #John-wa otoko da John-wa male be "#John is male."  Thus, it is important to distinguish predicates describing temporary state descriptions of objects from predicates describing properties.8 In particular, the subjects of stative verbs differ from the subjects of individual-level predicates (predicates describing inherent properties) in that the subjects of individual-level predicates are wa-marked unless they are focused.9 We have also seen in chapter 3 that wa-marked NPs cannot initiate an answer to questions such as "What happened?" or "What is the matter?", which force the whole sentence in the answer to be presentationally focused.  In the literature, distinctions between stative verbs and individual-level predicates are not always made explicit. This is because some authors assume that the subject of an individual-level predicate, which usually denotes a property, may induce a temporary state description interpretation. 7  Individual-level predicate - Property -State Stage-level predicate  - Event  In contrast, I will strictly associate individual-level predicates with property interpretations. Thus, I do not consider any predicates that allow temporary state description interpretations to be individual-level predicates. This means that predicates such as "altruistic" may be used as individual-level predicates (properties) or as stage-level predicates (temporary states). Individual-level predicate - Property Stage-level predicate  - State - Event  de Hoop & de Swart (1990) propose that the stage- vs. individual-level distinction should be abandoned since some stage-level predicates such as "to die", "to destroy", etc., whose action cannot be repeated (once-only predicates), behave like individual-level predicates. 8  'Note that they can only be contrastively focused, but not presentationally focused.  143  (8)  Context: What is happening?/ What is the matter? #John-wa gengogaku-o benkyoosite-i-ru. John-wa linguistics-acc studying-be-pres "#John is studying linguistics."  I now move to the question of where wa-marked subjects are generated.  Subjects of stage-level predicates vs. subjects of individual-level predicates Diesing (1988) investigates the relationship between the stage/individual-level distinction and focus projection. She claims that the subject of a stage-level predicate is base-generated in Spec V P whereas the subject of an individual-level predicate is base-generated in Spec IP as in the following examples. (9) a. [jpBLOWFISHp [^are poisonous]], b. [jptvp BLOWFISH are available]F]. Claiming that focus can only project from the inner V P position but not from the outer V P subject position as shown in (9), she maps non-focus-projectable materials (and contrastive focus elements) into the restrictor of a tripartite quantificational structure and focus-projectable materials like presentationally focused elements into the nuclear scope, (cf. Selkirk 1984)1 Diesing extends her analysis to Japanese, where such differences are clearly distinguished by the use of wa-marking and ga-marking. (10) a. Kuzira-wa mie-ru. whale-wa see-pres " W H A L E S are VISIBLE." b. Kuzira-ga mie-ru. whale-nom see-pres " W H A L E S can be seen."  (individual-level: property)  (stage-level unaccusative: state description)  As observed in chapter 2, in a sentence containing a derived subject such as the subject of an unaccusative or passive verb, focus projection from a subject to a (stage-level) predicate is possible. This is allowed because the trace of the subject in Spec V P licenses the focus projection. Thus, in  144  (10b), a pitch accent on the subject in English and one on a MaP in Japanese is sufficient to license focus projection on the whole sentence because the subject of the stage-level predicate (unaccusative/passive) is base-generated inside VP. In (10a), which contains an individual-level predicate, however, a pitch accent is required on both the subject and the predicate in English and two MaPs are required in Japanese because focus projection is confined to V P . Observing that the S-structure position of a subject (in particular its syntactic order with respect to the verb) correlates with its interpretation in Japanese, Diesing proposes that non-focus-projectable material such as NP-wa and contrastively focused elements are mapped into the restrictor whereas focus-projectable material (such as presentationally focused ga-marked subjects) are mapped into the nuclear scope, as in (11). (11)  IP  IP  IP=restrictor=presupposed=non-focus projection VP=nuclear scope=not presupposed=focus projection Diesing motivates this mapping hypothesis on the following basis. Contrastively focused elements trigger existential presuppositions and as such must be mapped onto the restrictor of a tripartite quantificational structure. In contrast, presentationally focused elements do not trigger existential presuppositions and as such must be mapped onto the nuclear scope. She claims that this establishes a distributional distinction between subjects of individual level predicates and subjects of unaccusative predicates. This leaves open the question of whether the subject of event-denoting verbs (transitive/unergative) occupy a distinct subject position.  145  Subjects of transitive/unergative verbs vs. subjects of unaccusative verbs (stage-level) Nakayama & Koizumi (1991) claim that the subjects of unergative and transitive verbs are base-generated outside the domain of V P , whereas the subjects of unaccusatives and passives are base-generated in Spec V P . The core evidence for this proposal comes from numeric floated quantifiers (NQ). (12) a Gakusee-ga 3-nin kono naihu-de niku-o kit-ta student-nom 3-CL this knife-with meat-acc cut-past "Three students cut meat with this knife." b  *Gakusee-ga kono naihu-de 3-nin niku-o kit-ta student-nom this knife-with 3-CL meat-acc cut-past "Three students cut meat with this knife."  c  *Gakusee-ga niku-o 3-nin kit-ta student-nom meat-acc 3-CL cut-past "Three students cut meat."  d  *Gakusee-ga John-ni 3-nin tegami-o kai-ta. students-nom John-dat 3-CL letter-acc write-past "Three students wrote letters to John."  Following Miyagawa (1989), they assume that an NP can be construed with a numeric floating quantifier (NQ), only if mutual c-command holds between the NP and the NQ. Observing that VP-internal elements such as either an instrumental adjunct or the indirect and direct object of the verb cannot appear between the subject and the N Q with which the subject is construed in (12), they claim that the subject of a transitive verb must be base-generated outside Spec V P . They also compare subjects of unergatives with the subjects of unaccusatives and passives. They argue that (13a), which contains an unergative verb, is ungrammatical because mutual c-command does not hold between the subject and the NQ. In contrast, (13b), which contains an unaccusative verb, and (13c), with a passive, are grammatical since mutual c-command holds between the subject trace in Spec V P and the NQ. (13) a *Gakusee-ga geragera 3-nin warat-ta. student-nom loudly 3-CL laugh-past "Three students laughed loudly."  ;  146  b Gakusee-ga attoiumani 3-nin sin-da. student-nom momendy 3-CLdie-past "Three students died suddenly." c Gakusee-ga Mary-no kawari-ni 3-nin sikar-are-ta student-nom Mary-gen behalf 3-CL scold-passive-past "Three students were scolded in place of Mary." Hence, they conclude that the subjects of transitive and unergative verbs cannot be base-generated in Spec V P , in contrast to the subjects of unaccusatives and passives.  Unifying the three analyses Let us recapitulate the generalizations that we have established so far. 1) Non-thematic subjects (hanging topics) are located higher than thematic subjects. 2) Subjects of individual-level predicates are located higher than subjects of unaccusative predicates. 3) Subjects of transitive and unergative verbs are located higher than subjects of unaccusative predicates. To explain the above distribution of subjects, we need a model of phrase structure that allows three distinct subject positions. 1 0 1 1  "Tateishi (1991) proposes a lexical AGR projection between IP and VP, to host theta-marked subjects in Japanese. "The following examples indicate that the subject in Spec VP may move into Spec IP. In both (i) and (ii), the sentences are presentationally focused. (i)  Context: At the sea world, Dainamikku-ni kuzira-ga mie-ru dynamically whale-nom seen-pres "The whale can be seen dynamically (with 3D glasses)."  (ii)  Context: At the sea world, Kuzira-ga dainamikku-ni mie-ru dynamically whale-nom seen-pres "The whale looks dynamic." "The whale can be seen dynamically (with 3D glasses)."  Interestingly, in (ii), the sentence becomes ambiguous. This suggests that the subject is moved from Spec VP to Spec IP.  147  (14) Subjects of stage-level predicates IP stage-level non-thematic subject stage-level (agent) transitive/unergatr subject stage-level (theme) unaccusative subject The question that remains to be answered is whether subjects of individual level predicates occupy a position distinct from non-thematic subjects and subjects of transitive/unergative verbs. If so, what position do they occupy?  4.2.2. Subjects of individual-level  predicates  In this section, I provide three pieces of evidence to demonstrate that the subjects of individual-level predicates and wa-marked elements are located in a position which is structurally higher than any of the three subject positions I identified in ( 1 4 ) . 1 2 1 3 1 temporarily label the position occupying the subject of an individual-level predicate a topic phrase.14 12  I will exclude the cases of mini-topics.  In contrast, Tateishi (1991) claims that theta-marked subjects, whether wa-marked or ga-marked, are located in Spec AGRP, which is between IP and VP. 13  Gussenhoven (1992) claims that a topicalized argument cannot form a focus domain with its predicate.  14  (i) Context: What happened? a.  John met Mary.  b.  #Mary, John met.  As shown in (ib), a sentence containing a topicalized element cannot answer a question such as "What happened?", which requires the whole sentence to be presentationally focused. As we have already seen in Chapter 3, this claim also holds in Japanese. Consider the following examples. (ii) Context: What happened? a. John-ga Mary-ni at-ta John-nom Mary-to meet-past  148  (15) individu subject stage-level non-thematic subject stage-level (agent) transitive/unergati subject stage-level (theme) unaccusative subject  Argument 1: Conditionals and IP adjuncts I establish that an element is located outside IP by arguing that it is higher than an IP adjunct (conditional clause). 15 The contrast between (16) and (17) establishes that topicalized elements "John met Mary" b.  Mary-ni-wa John-ga at-ta John-wa Mary-to meet-past "#Mary, John met."  (iib) cannot be used as an answer to "What happened?"; However, it becomes acceptable in the following context. (iii) Context: What happened to Mary? Mary-ni-wa John-ga at-ta John-wa Mary-to meet-past "#Mary, John met." This suggests that topicalized constituents/wa-marked elements fall outside the domain of presentational focus, i.e. the domain of IP. Note also that scrambling within the domain of presentational focus is acceptable. Context: What happened? Mary-ni John-ga at-ta Mary-with John-nom meet-past "John met Mary." I take this to indicate that there are two types of preposing (topicalization and scrambling), which have different landing sites. "However, wa-marked elements may occur in clausal adjuncts such as "kara" (because) clauses, which are adjoined higher than IP. John-wa eega-ni ik-u-kamorena-i kara, watasi-mo issyoni ik-u. John-wa movie-to go-pres-may-pres because, I-also together go-pres  149  cannot occur within IP adjuncts in English, (cf. Sawada 1993) (16)  When John goes to see the movie, I will go with him. If John goes to the university, I will pay the tuition for him.  (17)  *When, the movie, John goes to see, I will go with him. *If, the university, John goes to, I will pay the tuition for him.  Likewise, we see that wa-marked NPs in Japanese are not licensed within conditional clauses, which are adjoined at the IP level. We thus conclude that a wa-marked N P must occur outside the domain of IP. (18) a. John-ga eega-ni ik-u toki, watasi-mo issyoni ik-u. John-nom movie-to go-pres when, I-also together go-pres "When John goes to the movie, I will go with him." b. *John-wa eega-ni ik-u toki, watasi-mo issyoni ik-u. John-nom movie-to go-pres when, I-also together go-pres *"When as for John, he goes to the movie, I will go with him." (19) a. John-ga daigaku-ni ik-u-nara, watasi-ga kare-no zyugooryoo-o hara-u. John-nom university-to go-pres-if, I-nom he-gen tuition-acc pay-pres "If John goes to the university, I will pay his tuition." b. *John-wa daigaku-ni ik-u-nara, watasi-ga kare-no zyugooryoo-o hara-u. John-nom university-to-if, I-nom he-gen tuition-acc pay-pres *"If as for John, he goes to the university, I will pay his tuition." Argument 2: Pro IP Pro-form "soo (da)" functions as a pro-IP. Koizumi (1993) shows that "soo (da)" replaces the propositional content of IP, (excluding epistemic modals). In contrast, "soo suru" functions as a pro V P and "da" functions as a pro V . Given this, consider the following examples.  The following example, however, shows that CP elements cannot be either in conditional clauses or in clausal adjuncts such as "kara" clauses. *John-wa eega-ni ik-u no toki, watasi-mo issyoni ik-u. John-wa movie-to go-pres comp when, I-also together go-pres *John-wa eega-ni ik-u no kara, watasi-mo issyoni ik-u. John-wa movie-to go-pres comp because, I-also together go-pres This indicates that wa-marked elements are located below CP elements.  150  (20)  A : Yuuree-wa [^mie-ru]. (for psychics) ghost-wa visible "Ghosts are visible." (generic statement) B: a.  Syogoree-mo soo da guide-also soo be "So are spiritual guides."  b.  #Syugoree-mo soo sur-u guide-also soo do-pres  c. ?Syugoree-mo da. guide-also be (20) indicates that "yuuree-wa" in (20a) is outside the domain of IP because if the wa-marked element is in the domain of IP, (20Ba) should not be possible. The following examples in (21) suggests that presentationally focused ga-marked subjects occupy a different position, i.e., they occur IP/VP internally. 16 (21)  A : What happened? B: [ IP/vp Yuuree-ga mie-ru] ghost-nom see-pres "Ghosts can be seen." B: a  ?Syogoree-mo soo da guide-also soo be "So are spiritual guides."  b. #Syugoree-mo soo sur-u guide-also soo do-pres c. Syugoree-mo da. guide-also be  "pro IP" (soo da) replaces all IP elements. However, (21A) cannot be replaced by "pro IP". This indicates that a presentationally focused non-wa-marked element occurs inside IP.  Evidence 3: Negation of events Recall from chapter 3 that negation with wa-marking and negation without wa-marking are Uttering (21Bb) in one MaP makes the contrast between (20Ba) and (21Ba) more explicit.  16  151  semantically different. (22)  (Moo gozen 3-zi da kara,) hito-ga hitori-mo aruite-i-na-i. (already A . M . three o'clock be because,) people-nom one-polarity walking-be-not-pres "Since it is already three A . M . , people are not walking."  (23)  (Koko-wa zitensya senyoo da kara,) hito-wa hitori-mo aruite-i-na-i. (here-wa bicycle proper be because,) people-wa one-polarity walking-be-not-pres "Since this (road) is for bicycles only, people are not walking."  We see that in (22), the whole event is negated (no people can be observed now) whereas in (23) only the predication is denied, (people may be observed, but they are not walking). Thus, a wa-marked element falls outside the scope of negation whereas an ordinary case-marked element falls inside the scope of negation. To capture this fact, I assume that negation takes scope over IP and that wa-marked elements occur outside IP. Finally, before closing this section, I will show that not only subject-wa but also other wa-marked elements, such as object-wa must be located outside IP. Thus, consider the following examples. (24) a. *Eega-ni-wa John-ga ik-u toki, watasi-mo issyoni ik-u. movie-to-wa John-nom go-pres when, I-also together go-pres *"When, the movie, John goes to see, I will go with him."  b. Eega-ni John-ga ik-u toki, watasi-mo issyoni ik-u. movie-to John-nom go-pres when, I-also together go-pres "When John goes to see the movie, I will go with him." (25) a. *Daigaku-ni-wa John-ga ik-u-nara, watasi-ga kare-no zyugooryoo-o hara-u. university-to-wa John-nom go-pres-if, I-nom he-gen tuition-acc pay-pres *"If, the university, John goes to, I will pay the tuition for him." b. Daigaku-ni John-ga ik-u-nara, watasi-ga kare-no zyugooryoo-o hara-u. university-to John-nom go-pres-if, I-nom he-gen tuition-acc pay-pres "If John goes to the university, I will pay the tuition for him." The examples in (24) and (25) contain an IP adjunct. Note that the object of the main clause cannot be wa-marked. This indicates that a wa-marked object cannot be inside IP, either. Let us now compare (26) and (27).  152 (26) pro IP A : Tom-wa tokidoki zassi-o yomi-mas-u Tom-wa often magazine-acc read-mood-pres "Tom often reads magazines." B: ?Manga-mo soo da comic-also so be (27) pro IP A : Tom-wa zassi-wa tokidoki yomi-mas-u Tom-wa magazine-wa often read-mood-pres "Tom read magazines through many times. B: Manga-mo soo da. comic-also so be The comparison of (26) and (27) indicates that "zassi-wa" in (27), but not "zassi-o" in (26), is located outside I P . 1 7 1 8  4.3. Toward a syntax-information structure  4.3.1. Syntax-semantics  mapping  mapping  Diesing's pioneering work (1992) proposes an explicit mapping between syntactic structure and 17 Jager (1995) argues that the different syntactic positions of the object in German correlate with an atelic vs telic distinction. Consider the following examples from Jager (p.3).  (i)  Peter hat oft die Biebel gelesen, (aber selten gebetet) Peter has often the Bible read, but seldom prayed) "Peter often read the Bible (but seldom prayed)."  (ii)  Peter hat die Biebel oft gelesen Peter has the Bible often read "Peter read the Bible many times."  Jager argues that when the object is internal to IP, an atelic reading is obtained whereas when the object is external to IP, a telic reading is obtained. Even VP may be topicalized with the aid of do-support, (cf. Mihara 1994 and Tateishi 1991).  18  Tokidoki zassi-o yomi-wa Tom-wa su-ru. often magazine-acc read-wa Tom-wa do-pres However, this construction does do not allow pro-IP replacements such as "soo da" because the topicalized elements themselves contain verbal elements.  153 semantics. She correlates presupposed information and new information with the availability of two syntactic positions i.e., IP subjects and V P subjects, thus deriving differences in interpretation from syntactic structure. In the following section, I review the mapping hypothesis proposed by Diesing (1992).  4.3.1.1.  Diesing (1992)  To explain the possible interpretation of subjects (that is, generic, specific, non-specific), Diesing assumes two subject positions (VP-external and VP-internal) and then proposes that material in V P is mapped onto the nuclear scope whereas material outside V P is mapped onto the restrictor of a tripartite quantificational structure at L F . (28)  This explains why the subject of a stage-level predicate allows at least two interpretations, as shown in (29). (29)  Firemen are available. a. Genx[firemen (x)][available (x)] b. 3x[firemen (x) & available (x)]  (generic) (existential)  (29) is derived as in (30): presupposed information is mapped onto the restrictor domain at L F while non-presupposed information is mapped onto the nuclear scope.  154  (30) Diesing's S-structure to L F mapping of stage-level predicates a. Firemen are available (generic reading) S-S LF  [pFiremenfyp t are available]] S  Cvperator Restrictor Nuclear-Scope Genx[firemen (x)] [available(x)] Spec IP VP b. Firemen are available (existential reading) S-S LF  [jpFiremen[VF/ are available]] S  3x  [firemen (x) & available(x)] Spec V P VP  In (30a), the generic interpretation is derived by generating the subject in Spec IP where it falls under the scope of a generic operator. In contrast, in (30b), the VP-internal subject is under the scope of an existential operator. In contrast, an individual-level predicate only allows a generic reading. (31) Firemen are altruistic a. Genx [firemen (x)][altruistic (x)] b. *3x[firemen (x) & altruistic (x)]  (generic) (*existential)  To block the existential interpretation of the individual-level predicate in (31b), Diesing must prevent the subject N P from being lowered into VP. She assumes that the lexical N P in Spec IP is assigned a special theta-role "has the property x" by INFL and further controls a P R O subject base-generated in Spec VP, which is also assigned a theta-role by the predicate. (32) Diesing's S-structure for individual-level predicates S-str LF  [jpFirementvpPflO are available]] [ g Firpmen[ypryO are available]]  Diesing further claims that the syntax-semantics mapping takes place at L F in English whereas in a  155  language like German, S-structure can be mapped onto L F because there is a direct correspondence between S-structure and semantic interpretation, as demonstrated in (33) and (34). (33) a....weil Linguisten ja doch Kammermusik spielen (Diesing (1992): p36, 38). since linguists PRT PRT chamber music play "... since (in general) linguists play chamber music." b....weil ja doch Linguisten Kammermusik spielen since PRT PRT linguists chamber music play "... since there are linguists playing chamber music." (34) a....weil Wildschweine ja doch intelligent sind since wild boars 'indeed' intelligent are since (in general) wild boars are intelligent." b*...weil ja doch Wildschweine intelligent sind since 'indeed' wild boars intelligent are We see that the bare plural subject of the stage-level predicate which occupies the VP-external positions at S-structure induces a generic interpretation, as in (33a), whereas the subject which occupies the VP-internal position at S-structure induces an existential interpretation, as in (33b). In contrast, we see in (34) that the bare plural subject of the individual-level predicate can only appear in the VP-external position.  4.3.1.2. Problems with syntax-semantics mapping  Diesing's claim that a direct correspondence holds between the S-structure position of an N P and its semantic interpretation is challenged by Jager (1997). He claims that "the stage/individual contrast is in fact purely semantic in nature, while the apparent syntactic differences can be explained as consequences of general principles governing the interaction of information structure with both syntax and semantics." Consider the following German examples in Jager (1997: p7) (35) a. weil ja die meisten Studenten Franzosisch konnen since PRT most student French know "Most students know French." b. weil die meisten Student ja Franzosisch konnen since most students PRT French know  (thetic)  (categorical)  156  Jager shows that the subject of an individual-level predicate can be generated inside V P even when it is introduced by a strong quantifier, as shown in (35a). Jager then concludes that subjects of individual-level predicates may be base-generated in Spec VP, contra Diesing. A similar generalization can be made in Japanese. Consider the following example: (36)  A : What is the matter? B: Odoroi-ta koto-ni, hotondo-no hitobito-ga huransugo-o sitte-i-ru. surprisingly, most-prt people-nom French-acc knowing-be-pres "Surprisingly, most people/most of the people know French (in Montreal)."  In (36), we see that the presupposed elements, the strong quantifier "most" or the partitive N P "most of the people" is used in the presentationally focused sentence. I thus conclude that the mapping between syntax and interpretation cannot be enforced by appealing to the notion of presupposition. In support of this conclusion, recall the examples in (4) where the whole sentence can be presentationally focused as long as the predicate describes a state as opposed to a property.  4.3.2. Syntax-information structure mapping  I propose that a mapping is enforced between syntax and information structure (topic-comment structure).19  2 0  2 1  In the next section, I propose a hierarchy of functional projections, and  correlate the distribution of subjects with this hierarchy. I then define topic-comment in terms of We have to exclude cases of mini-topics.  19  20  A partition between background and focus alone induces a wrong prediction. Context: A: Kono ziki sake-wa oisii ne this season salmon-wa tasty prt "Salmon is tasty this season." B: Soo i-eba, watasi-wa kinoo sake-o tabe-ta say, I-wa yesterday salmon-acc eat-past "By the way, I ate salmon yesterday."  B's answer is also appropriate to a question like "What did you do yesterday?", where the VP is presentationally focused. If a partition is enforced in terms of background vs. focus, the object cannot be mapped onto VP (IP in our terms) because the object "salmon" has already been uttered by A. Jager (1997) correlates the topic-comment vs. comment only distinction with a categorical vs. thetic judgment distinction, proposed by Kuroda (1972). He then proposes that the subject of individual-level predicates, but not the subject of stage-level predicates, must obligatorily move out of the thetic (comment) domain. 21  this hierarchical structure.  4.3.2.1. Possible sites for topics  In this section, I will argue for the following hierarchy of post-verbal heads.  benkyoosi "study"  158  John-ga eego-o benkyoosi-nakerebanarana-i kamosirena-i. John-nom English-acc study-must-pres may-pres "John may have to study English." I assume that the morphological order of the heads directly feeds into the hierarchical structure of functional projections of clauses, as shown in (37). Note that a root sentence, excluding embedded clauses, may contain two modals besides tense and aspect. The higher modal is an epistemic modal whereas the lower modal is a root (deontic/dynamic) modal. 2 3 Brennan (1993: p i ) claims that "epistemic modal sentences are those expressing possibility or necessity relative to some state or knowledge." In contrast, root modal sentences describe either what is required or allowed by some normative system (deontic modals) or the abilities, susceptibilities and dispositions of the subject (dynamic modals). ^ h e following example indicates that the epistemic modal and tense projections may also be dominated by certain modals, such as "yooda/sooda/rasii". (i)  John-ga eego-o benkyoosi-nakerebanarana-i kamosirena-i yooda/sooda/rasii. John-nom English-acc study-must-pres may-pres seem "It seems that John may have to study English." However, their distribution is the same as that of "noda", which is literally a complementizer/nominalizer plus be. (ii)  John-ga eego-o benkyoosi-nakerebanarana-i kamosirena-i noda John-nom English-acc study-must-pres may-pres mood be "It is that John may have to study English."  Also, it is not possible to switch the order of "yooda/sooda/rasii" with the epistemic modal. (i)'  *John-ga eego-o benkyoosi-nakerebanarana-i yooda/sooda/rasii kamosirena-i. John-nom English-acc study-must-pres seem may-pres (ii) ' *John-ga eego-o benkyoosi-nakerebanarana-i noda kamosirena-i. John-nom English-acc study-must-pres mood be may-pres Thus, I conclude that when "yooda/sooda/rasii/noda" are used, the whole sentence is embedded as in (iv). (iii)  John-ga eego-o benkyoosi-nakerebanarana-i kamosirena-i. John-nom English-acc study-must-pres may-pres "John may have to study linguistics."  (iv)  John-ga eego-o benkyoosi-nakerebanarana-i kamosirena-i yooda. John-nom English-acc study-must-pres may-pres "It seems that John may have to study linguistics."  This is another class of epistemic modal elements that occupy a seperate higher position . I assume that these higher modals are seperate predicates which take a subordinate clause as an argument. ^Japanese has no special morphology specific to modals; Most of them are morphologically complex and have bound adjective endings. However, I will treat them as simple modals, considering that they are morphologically established as one word.  159  (38)  Root  Modal  Epistemic  Modal |24  may  (te)moyo(i) (permission)  kamosirena(i) (possibility)  must  nakerebanarana(i) (necessity)  (ni)tigaina(i) (certainty)  will  (y)oo (volition)  daroo (probability)  should  beki(da)25 (obligation)  hazu(da) (supposition)  Note that root modals cannot dominate aspect whereas epistemic modals must dominate aspect and must be dominated by tense.26 (39) a. John-ga eego-o benkyoosi-nakerabanarana-i. (root modal) John-nom English-acc study-must-pres "John must study English." (John has to study English.)  The following examples indicate that "soo da" (seem) may also be used as a root modal.  24  John-ga nemuri-sooda John-nom sleep-seem "John is falling asleep." John-ga (yoku)nemur-u-sooda. John-nom (often) sleep-pres seem "It seems that John (often) sleeps Moreover, the following examples show that, in Japanese, "e" (can) is located lower than root modals. John-wa eego-ga hanas-e-temomoyo-i (noni, hanas-e-na-i.) John-wa English-nom speak-can-may-pres but, speak-can-not-pres "John should be abletospeak English (because he was born in the U.S., but he cannot." There is another equivalent of (can) "koto-ga deki", which requires the embedding of a clause. John-wa [eego-o hanas-u koto]-ga deki-ru. John-wa English-acc speak-pre nominalizer-nom can-pres Thus, I do not consider "e" (can) as a modal. Rather, I assume that it forms a complex predicate. "The status of "beki" as a root modal is questionable since it may be followed by tense in many cases. ^Jackendoff (1972) claims that there is a structural difference between epistemic modals and root modals, i.e., epistemic modals are analogous to rasing verbs and root modals are analogous to control verbs. He also observes that the distributions of epistemic modals and root modals are parallel to those of speaker-oriented adverbials and subject-oriented adverbials.  160 b. *John-ga eego-o benkyoosu-ru-nakerabanarana-i. (root modal) John-nom English-acc study-pres must-pres (40) a. John-ga eego-o benkyoosu-m-nitigaina-i. (epistemic modal) John-nom Enghsh-acc study-pres-must-pres "John must study English." (John is definitely going to study English.) b. *John-ga eego-o benkyoosi-nitigaina-i. (epistemic modal) John-nom English-acc study-must-pres Thus, the order of root modals and epistemic modals cannot be reversed. (41)  *John-ga eego-o benkyoosi-kamosirena-i-nakerebanarana-i John-nom English-acc study-may-pres-must-pres  4.3.2.2. Topic vs. comment domain and wa-marking  I now address the question of how to correlate the articulated phrase structure in (37) with information structure - that is, topic-comment structure, as shown in (42).  (42)  TOPIC  Comment  161  Distributional correlation of wa-marked elements and E M In section, 4.2., we argued that wa-marked elements cannot occur in a conditional adverbial clause. I first show that an E M also cannot occur in a conditional adverbial clause. Consider the following examples from Sawada (1993). (43) a. John may go to see the movie. (RM/EM) b. When John may go to see the movie, I will go with him. (RM/*EM) (44) a. John must go to the university. (RM/EM) b. If John must go to the university, I will pay the tuition for him. (RM/*EM) In (43a) and (44a), the modal may be either interpreted as a root modal or an epistemic modal. However, in conditional clauses as in (43b) and (44b), only a root modal interpretation is possible. Let us now consider the Japanese equivalents. (43J) aJohn-ga/wa eega-ni it-temoyo-i/ik-u-kamosirena-i. John-nom/wa movie-to going-may/go-pres-may-pres "John may go to see the movie." (RM/EM) b.John-ga/*wa eega-ni it-temoyo-i/*ik-u-kamosirena-i toki, watasi-mo issyoni ik-u. 2 7 John-nom/wa movie-to going-may-pres/go-pres-may-pres when, I-also together go-pres "When John may go to see the movie, I will go with him." (RM/*EM) (44J) a.John-ga/wa daigaku-ni ika-nakerebanarana-i/ik-u-nitigaina-i. John-nom/wa university-to go-must-pres/go-pres-must-pres "John must go to the university." (RM/EM) b.John-ga/*wa daigaku-ni ika-nakarebanaranak-ereba/?ik-u-nitigainak-ereba John-nom/wa university-to go-must-if/go-must-if watasi-ga kare-no zyugyooryoo-o hara-u. I-nom he-gen tuition-acc pay-pres "If John must go to the university, I will pay the tuition for him." (RM/*EM)  o  "The following example indicates that an " i f (ereba) clause is adjoined lower than an aspect phrase since an aspect suffix such as " - i " or "-katta" is not allowed in the clause. John-ga daigaku-ni ika-nakarebanaranak-ereba/*ika-nakerebanaranak-u-ereba John-nom/wa university-to go-must-if/*go-must-pres-if "If John must go to the university, watasi-ga kare-no zyugyooryoo-o hara-u. I-nom he-gen tuition-acc pay-pres I will pay his tuition."  162 We see that in (43) and (44), modal cannot be interpreted as an E M . This indicates that the E M projection as well as wa-marking must be located higher than conditional adverbial clauses. In 4.2., we saw that sentences which contain a wa-marked subject allows a pro-IP replacement, but that sentences which contain a presentationally focused ga-marked subject do not. We concluded that pro IP replacement targets a constituent which excludes wa-marked subjects but includes ga-marked subjects. The following examples also indicate that only epistemic modals, but not root modals, permit a pro-D? replacement. (45) a. Tom-wa zibun-de ie-o tate-nakerebanarana-i. *Mary-mo soo nakerebanarana-i. Tom-wa self-by house-acc build-must-pres. Mary-also so must "Tom must build the house for himself." "So must Mary." (RM) b. Tom-wa zibun-de ie-o tete-ru-nitigaina-i. Mary-mo soo nitigaina-i. Tom-wa self-by house-acc build-must-pres. Mary-also so must-pres "Tom must build the house for himself." "So must Mary." (EM) Assuming the structure in (37), what "soo" does is to serve as a pro-ASP-phrase. Having observed these facts, I propose that the domain above the E M phrase is the topic domain whereas the domain below the Aspect Phrase is the domain of comment.  Temporal adverbials and wa-marking I now present further evidence for the claim that wa-marked NP's occur in the topic domain. This evidence is based on the interpretation of wa-marked temporal adverbs. I first briefly review the interpretation of temporal adverbials in English. de Swart (1996) demonstrates the following correlation between the position and interpretation of time adverbials in English, as illustrated by the paradigm in (46). (46) a. The bomb didn't explode at three o'clock, b. At three o'clock, the bomb didn't explode.  163  According to de Swart, (46a) is ambiguous between the following two interpretations: i) the adverb specifies the reference time and takes wide scope over negation. What happened at three o'clock was that the bomb did not explode. ii) the adverb does not specify the reference time. The bomb exploded, but this accident did not happen at three. Note, however, that (46b), in which the adverb is preposed, is unambiguous. It only allows the first interpretation. She concludes that the pragmatic scope of negation is over what is focused (i.e., over the main clause) in (i) but only over the time adverbial in (ii). The part of the sentence that is not focus (topic) is outside the scope of negation. As shown in (47), in Japanese, these two readings are distinguished in terms of wa-marking. In (47b) where the time adverb is wa-marked, the wa-marked time adverb must be outside the scope of negation. (47) a. Bakudan-ga sanzi-ni bakuhatusi-na-katta. bomb-nom three-at explode-not-past The bomb didn't explode at three o'clock. b. Sanzi-ni-wa bakudan-wa bakuhatusi-na-katta three-at-wa bomb-wa explode-not-past At three o'clock, the bomb didn't explode. This further supports the claim that wa-marked elements occur in the topic domain, which is above ASP.  4.3.2.3. Wa-marking as the semantic subject of predication 2 8  I now address the question of why the subject of a generic statement is obligatorily wa-marked as illustrated in (48-49). As noted in Carlson (1989), the subject of a generic sentence need not be the ^For inherently assigned cases, such as by prepositions, wa-marking is simply attached to the PP without overriding the original case-marking, e.g. kaisya-de-wa company-loc-wa  John-ni-wa John-to-wa  164  syntactic subject. Consider the following examples from Brockett (1990: p.61). (48) a. Taihuu-wa Taiheeyoo-no kono hen-de hasseesu-ru. Typhoon-wa Pacific Ocean-gen this part-loc occur-pres "Typhoons (in general) occur in this part of the Pacific." b. Taiheeyoo-no kono hen-de-wa taihuu-ga t hasseesu-ru. Pacific Ocean-gen this part-loc-wa typhoon-nom occur-pres "In this part of the Pacific, (in general) typhoons occur." (49) a. Konpyuutaa-wa mainiti-no tenki-o utidas-u. computer-wa daily-gen weather-acc print-out-pres "Computers (in general) print out the daily weather." b. Mainiti-ni tenki-wa, konpyuutaa-ga t uti-das-u. daily-gen weather-wa computer-nom print-out-pres "The daily weather, (in general) a computer prints it out." In (48b), the locative "in this part of the Pacific", and in (49), the object "the daily weather" are located outside the domain of focus (comment domain). Chierchia (1995) claims that "individual-level predicates express properties of individuals that are permanent or tendentially stable." I propose, following Krifka (1988) that constituents move outside the domain of focus (=comment in our term) to be bound by the generic operator. Thus, an element that induces a property interpretation is wa-marked.  4.4. The syntactic position of contrastively focused subjects  In the previous section, we also saw that wa-marked subjects must be outside the domain of comment, identified as the domain of A S P . 2 9 In this section, I will demonstrate that wa-marked ^Mihara (1994) and Whitman (1991) claim that a subject that follows a VP adverb as in (ia) can only be interpreted as presentational focus. In contrast, where the subject precedes the VP adverb as in (ib), it may be interpreted as either contrastive focus or presentational focus. (i) VP adverb (the examples from Mihara) a. Matigatte, Taroo-ga Chomsky-no hon-o kat-ta by mistake, Taroo-nom Chomsky-gen book-acc buy-past By mistake, Taro bought a book by Chomsky." b. Taroo-ga matigatte Chomsky-no hon-o kat-ta Taroo-nom by mistake Chomsky-gen book-acc buy-past "Taro by mistake bought a book by Chomsky."  165  elements as well as contrastively focused (ga-marked) subjects are outside the domain of comment, i.e., outside the domain of ASP. Consider the following examples. (50) A : Tyuukosya-wa yoku koware-ru. (generic statement) used car-wa often broken-pres "Used cars break down often B:(Tyuuko) konpyutaa-mo soo-da. (used) computer-also so-be "So do used computers." (51) Context: What is often broken? A : T Y U U K O S Y A - g a yoku koware-ru. (generic statement) used car-nom often broken-pres "USED C A R S break down often B:(Tyuuko) konpyutaa-mo soo-da. (used) computer-also so-be "So do used computers." Notice that in (51), the subject is ga-marked instead of being wa-marked; however, the sentence in (51) tolerates a pro-ASP replacement, which indicates that the ga-marked subject in (51) is outside the domain of comment.30 3 1 Let us consider another set of examples. As seen in 4.2., the whole proposition is under the scope of negation when the sentence is presentationally focused. (52)  (Moo gozen 3-zi da kara,) hito-ga hasitte-i-na-i. (already A . M . three o'clock be because,) people-nom running-be-not-pres "Because it is already three in the morning, no people are running."  We see that in (52), the whole event is negated. Thus, the sentence in (52) means that there is no 30  As shown below, if the subject is presentationally focused, a pro-ASP replacement is not possible.  Context: What happened? Tyuukosya-ga koware-ta. used car-nom broken-past "My used car broke down. ??(Tyuuko) konpyutaa-mo soo-da. (used) computer-also so-be "So do used computers." Herburger (1996) also demonstrates that contrastively focused elements quantify over events.  3,  166  one who is running. However, if the subject is contrastively focused, it cannot be inside the scope of negation. We see that the sentence in (53aB) can only mean that among those which are not running, it is human being who is not running. Also recall that the wa-marked subject cannot be in the scope of negation, as shown in (53b). (53) a. A : (Koko-wa zitensya senyoo da kara,) hito-ga aruite-i-na-i. (here-wa bicycle proper be because,) people-nom walking-be-not-pres "Because this (road) is only for bicycles, people are not walking." Context: A friend of A could not hear what she said, and asks B: Nani-ga hasitte-i-na-i te? What-nom running-be-not-pres quote "What is not running?" And A answers, A : Hito-ga hasitte-i-na-i. people-nom running-be-not-pres "People are not running." b. (Koko-wa zitensya senyoo da kara,) hito-wa aruite-i-na-i. (here-wa bicycle proper be because,) people-wa walking-be-not-pres "Because this (road) is only for bicycles, people are not walking."  4 . 5 . A comparison with Kuroda's thetic/categorical distinctions  In order to differentiate wa-marked elements from other case-marked elements, especially from nominative case marking, Kuroda (1965, 72, 92) uses the notions of categorical judgement and thetic judgement. He claims that "a categorical judgement consists of two distinct cognitive acts, one the recognition of a subject and another the act of acknowledging or disavowing a predicate of a subject."32 In this sense, a categorical judgement is also called a double judgement. In contrast, "a thetic judgment is a unitary cognitive act, which simply expresses recognition of the existence of an entity or a situation." In this sense, a thetic judgement is called a simple judgement. Thus, the subject of a categorical judgment is a logical subject (semantic subject=topicalized), and the subject of a thetic judgment is a grammatical subject.33  34  He assumes that the subject of a  Kuroda uses "Subject" to refer to the semantic/logical subject and "subject" to refer to the syntactic subject.  32  33  As we have seen, a wa-element need not be a syntactic subject.  167  thetic judgment is mapped onto the position inside IP and the subject of categorical judgment is mapped onto the position outside IP. 3 5 1 will show, however, that the thetic/categorical distinction does not correctly predict the distribution of wa-marking and ga-marking in Japanese. In particular, I will show that categorical judgment cannot be correlated with wa-marking. Consider the following examples from Kuroda (1992). (54) Categorical judgement =wa-marked Neko-wa nete-i-ru. cat-wa sleeping-be-pres "The cat is sleeping." (But, "#A cat is sleeping.")36 (55) Thetic judgement=non-wa-marked Neko-ga nete-i-ru. cat-nom sleeping-be-pres "The cat is sleeping." (or " A cat is sleeping.") The sentence in (54) involves a categorical judgement, and thus the subject is wa-marked in Japanese. According to Kuroda, the categorical judgement consists of two separate cognitive acts: the act of recognition of the cat and the act of affirming the cat is sleeping; thus making a assertion about the cat. In contrast, (55), a thetic judgement, only constitutes an act of affirming a specific situation: " A cat is sleeping." Thus, in Japanese, a categorical judgement is morphologically distinguished from a thetic judgement in terms of wa-marking. 3 7  38  Consider the following  ^Drubig (1994) and Winkler (1996) make similar distinctions. They call the semantic subject (the subject of a categorical judgement) "the subject of predication" and the syntactic subject (the subject of a thetic judgement) "the subject of theta saturation". However, I will not utilize this distinction because Japanese has non-thematic subjects, which may still be presentationally focused. Ogihara (1984) claims that if an element is wa-marked, it must be strong. Ladusaw (1994), however, proposes that the strong/weak distinction is an epiphenomenon of the categorical/thetic judgement distinction.  3S  ^Kuroda (1992) claims that an indefinite NP cannot be used in a categorical judgment because the cognitive act of recognizing the Subject cannot be simply be a thetic judgment affirming the existence of an entity. However, as we have seen in chapter 3, indefinite NPs may also be wa-marked in Japanese. This means, in Kuroda's terms, that indefinite NPs may be used in a categorical judgement. The distinction between a categorical and a thetic judgement is not overtly expressed in English.  37  ^For Kuroda, the example below would require both categorical and thetic judgements on two different objects simultaneously, although he does not discuss this kind of example. UBC-de-wa gakkai-ga okonawarete-i-ru. UBC-at-wa conference-nom held-be-pres  168 examples, which involve generic statements. (56)  Neko-wa nezumi-o oikake-ru. cat-wa mice-acc chase-pres "Cats chase mice."  (57) A : Nani-ga nezumi-o oikake-ru-no? what-nom mice-acc chase-pres-comp "What chases mice?" B: Neko-ga nezumi-o oikake-ru. cat-nom mice-acc chase-pres. "CATS chase mice." He claims that a topicalized sentence (wa-marked), whether specific or generic, expresses a categorical judgement and a non-topicalized sentence (non-wa-marked) expresses a thetic judgement. Thus, for Kuroda, the contrast between sentences with and without topicalization reduces to the distinction between categorical and thetic judgements. However, this correlation is not valid for generic sentences because he assumes that a thetic judgment is obligatorily specific. 3 9 Kuroda's correlation incorrectly predicts that a generic statement requires wa-marked subjects since it is a categorical statement. As shown in (57), this is empirically incorrect since the subject of the generic statement in (57) is ga-marked. Further evidence is provided in (58-59).40 4 1 "At UBC, a conference is being held." As shown in chapter 3, indefinites may also be wa-marked. This suggests that pace Kuroda, indefinites may be used as subjects of categorical judgements.  39  ""Kuroda (1992) claims that a generic statement must be a categorical judgement since it must involve "substance". In contrast, a specific statement can be either categorical or thetic. Kuroda (1992) claims that the subject of a categorical judgement may be maintained by switching the word order.  41  (i)  Nani-ga nezumi-o okikakeru no? what-nom mouse-acc chase comp "What chases mice?"  (ii)  Nezumi-wa nani-ga okikakeru no? mouse-wa what-nom chase-comp "What chases mice?"  He explains that the sentence in (i) does not involve a thetic judgement since no elements are wa-marked in (i); However, preposing the object "Nezumi" (mouse) and wa-marking it as in (ii) yields a sentence involving a categorical judgement. Nevertheless, Kuroda himself admits that this strategy cannot be used if the predicate is intransitive. Nani-ga hasiru-no? what-nom run-comp "What runs?"  169  (58) A : Nani-ga nezumi-o oikakeru-no? what-nom mice-acc chase-comp "What chases mice?" B: There are many animals which chase mice, but... Neko-WA nezumi-o oikakeru. cat-wa mice-acc chase. "CATS at least chase mice." In (58), the subject is WA-marked with stress (yielding a Contrastive Topic, as argued in chapter 3) and as such is topicalized in Kuroda's sense. However, Kuroda would predict this sentence to constitute a thetic judgment since it is an answer to a wh-question. Finally, Kuroda correlates wa-marking with definites. This correlation does not hold, as shown in (59). (59) A : Nani-ga nete-i-ru-no? what-nom sleeping-be-pres "What is sleeping (among your pets/cats)?" B: Neko-WA nete-i-ru. cat-nom sleeping-be-pres " A C A T is at least sleeping." In a context where the answerer has several pets or cats, and the answerer can check what his cats/pets are doing, "a cat" is WA-marked. However, Kuroda would predict that (59) involves a thetic judgment despite the fact that "a cat" is WA-marked. To conclude, under Kuroda's hypothesis, the syntax-information structure mapping is not straightforward because he correlates the distribution of wa-marking with the subject of a categorical judgment.  4.6. Summary and implications  In this chapter, I have shown that i) there are four distinct subject positions, and ii) there is a hierarchical structure of functional projections. For me, this is not a problem since contrastively focused subjects are also outside the comment domain. That is, in Kuroda's terms, they are also the subjects of categorical judgements.  170  I have correlated the distinct subject positionswith the hierarchy of functional projections and have demonstrated that topic elements (wa-marked elements) are located in the highest subject position. We have also seen that epistemic modals and topicalization are not allowed in conditional clauses in English as well as in Japanese. The examples are repeated in (60-61). This suggests that even in English, topics and epistemic modals are outside the domain of comment, (cf. Kiss 1996, Nakamura 1994) 4 2 (60) a. *When the movie, John goes to, I will go with him. (ASP adjunct) b. Because movies, John likes, I will go out with him. (TP adjunct) (61) a. When John may/must go to the movie, I will go with him (R/*E) b. Because John may like movies, I will go out with him. (R/E) This structure further distinguishes preposing to the TOPIC domain (topicalization/contrastive focus movement) from preposing inside the comment domain (scrambling), e.g. adjunction to ASP. The theoretical, empirical or structural difference, however, has not received much attention in the literature.43 4 4  4 5  Before I close this chapter, I would like to discuss what the effect of discourse on syntactic structure in Japanese is. In this chapter, I have proposed a syntactic structure which reflects the ^Japanese linguists such as Kuno (1983), Takubo (1987), Masuoka (1991), and Mihara (1994) notice that a wh-question requires "no da" (complementizer plus be). (i)  Tom-wa naze kinoo gakkoo-o yasunda no des-u ka? Tom-wa why yesterday school-acc absent comp be-pres Q "Why was Tom absent from school yesterday?"  (ii)  ?Tom-wa naze kinoo gakkoo-o yasumi-masi-ta ka? (sounds like an echo question) Tom-wa why yesterday school-acc absent-mood-past Q  As shown above, the sentence in (ii) can only be acceptable as an echo-question-like utterance. This suggests that Wh-questions may in fact contain an embedded structure. If this is true, the final destination of Wh-elements should be the superordinate Spec CP, but not the subordinate Spec CP. Gundel (1974) also claims that there are two types of topicalization: topic topicalization and focus topicalization.  43  I would also like to address the issue of base-generation vs. movement of wa-marked elements. My answer is both are possible; nevertheless, I propose that the base-generation of wa-marked elements in Spec TP is the equivalent of left-dislocation in English whereas movement of a Case-marked element (whether or not wa-marked) is the equivalent of topicalization in English. (Thus, Japanese and English are quite parallel.)  M  In Japanese, subjects (ga-marking is obligatorily replaced with wa-marking when a subject is a topic) and objects (o-marking may be replaced with wa-marking) are wa-marked when they are topics. Thus, in Japanese, left-dislocation of NPs is almost indistinguishable from topicalization of NPs.  45  171  discourse functions of topic and focus. However, in chapter 3, I have proposed a model of semantic interpretation which was insensitive to the syntactic presentation of F-marked and T-marked elements. How do I reconcile these two apparently contradictory apparatuses? I argue that focus and topic are uttered with respect to particular positions and/or domains in syntax in Japanese. For instance, wa-marking is no longer simply inducing a sentence topic, but is an instruction to syntax to locate a particular constituent in a position where it can be interpreted as a sentence topic (Henry Davis p.c). I have shown that the domain of comment is identified as the domain defined by presentational focus. This is the domain where a proposition resides. I thus propose that movement (and base-generation of left-dislocated elements) is necessary to derive a set of sets of propositions (a sentence with a topic) and a set of propositions (a sentence with contrastive focus).  Appendix: Ga-marking as the default case marker  I propose that ga not only marks the structural nominative case of subjects in Spec V P (ext) as in (1) but also functions as the default case-marker in Japanese. (1)  John-ga eego-o hanasite-i-ru. John-nom English-acc speaking-be-pres "John is speaking English."  As shown in (2) and in (3), if the verb is not capable of assigning accusative case to an element, ga is assigned to the element as a default case marker in Spec V P (int) (cf. Inoue 1989). Recall that the subject of an unaccusative verb stays in Spec V P (int) in Japanese.46  """Note that ga-marking cannot be identified as a focus marker since it marks non-focused elements in embedded clauses, as shown in the following example. A: John-ga sensyuu Canada-ni it-ta. John-nom last week Canada-to go-past "John went to Canada last week." B: John-ga Canada-ni it-ta koto-wa sira-na-katta. John-nom Canada-to go-past fact-wa know-not-past "I did not know that John went to Canada." Thus, ga should be simply regarded as a case-marker, (cf. Heycock 1993)  172 (2) Unaccusative verb Kuruma-ga koware-ta car-nom broken-past "The car broke." (3) a. Tough-construction Tom-ni/ga eego-ga hanasi-yasu-i. Tom-for/nom English-nom speak-easy-pres "English is easy for Tom to speak." b. Ability-verb Tom-ni/ga eego-ga hanas-e-ru. Tom-for/nom English-nom speak-can-pres "It is possible for Tom to speak English." c. Necessity-verb/construction Tom-ni/ga eego-ga hituyooda. Tom-for/nom English-nom necessary "It is necessary for Tom to speak English." d. Possession relation Tom-ni/ga kyoodai-ga oo-i (Mihara 1994: p 131) (as a temporary description) Tom-for/nom siblings-nom many-pres "Tom has many siblings." e. Volition predicate (watasi-ga) eego-ga hanasi-ta-i. (I-nom) English-nom speak-want-pres "I want to speak English." f. Non-accomplished states Tom-ga eego-ga tokuida. (as a temporary description) Tom-nom English-nom good "Tom is good at English." We have seen in 4.4. that a contrastively focused subject must be outside the domain of comment. (4) Context: What is often broken? a. YASU-I K U R U M A - g a yoku koware-ru. (exhaustive contrastive focus) cheap-pres car-ga often broken-pres " C H E A P C A R S break down often." b. Yasu-i kuruma-WA yoku koware-ru. (non-exhaustive contrastive focus) cheap-pres car-wa often broken-pres " C H E A P CARS at least break down often." Recall from our discussion in chapter 3 that wa-marking is used to mark non-exhaustive contrastive focus, but not exhaustive contrastive focus. The latter will thus be marked by the default case  173 marker, that is, by ga-marking as illustrated in (4).  174  Chapter 5: Conclusion: On Interface Issues  5.1. What does the analysis contribute to our understanding of Japanese grammar?  In this dissertation, I have presented a unified model of information structure in Japanese. In chapter 2, I have presented a Revised End-based Analysis of the syntax-prosody mapping in Japanese. I have claimed that presentational focus and contrastive focus are phonologically distinct and that the system of focus projection proposed for stress language like English and German is applicable to Japanese, a tone language. In chapter 3,1 have analyzed the discourse function of wa-marking in terms of Buring (1997, 98). I have proposed the distribution of wa-marking in Japanese parallels that of B-accents in English. In chapter 4,1 have argued that syntactic structure in Japanese is discourse-configurationally based. Let us briefly summarize the major finding of each chapter.  Syntactic component I have proposed the following hierarchy of functional categories. I then divided the syntactic structure of Japanese into two major domains, which I have called the TOPIC domain and the comment domain, as shown in (1).  175  (1)  TOPIC domain  Epistemic Modal) stage-level :vel norj- thematic sub] ect matic ^ J*MP comment domain  J^l AS (A;  V P (ext)  pect)  R M (Moot Modal)  staj e-level (agent) tran sitive/unergaj sub ect jiiP(int) stade-level (theme) unabcusative subjeet  V  I have further argued for the existence of three distinct subject positions within the comment domain.  Phonological component I have assumed that syntactic structure is mapped onto prosodic structure via an End-based theory of the type advanced in Selkirk & Tateishi (1988,91). I have proposed two major revisions of the End-based theory. First, I have argued that the End-based theory must be modified in order to incorporate sensitivity to syntactic (non) branching. Secondly, I have argued that phonological phrasing is constrained by the revised prosodic hierarchy which contains two new levels of prosodic constituency: projection phrases and contrastive focus phrases. I have assumed that focus projection takes place in the syntactic component, prior to mapping into prosodic structure, from  176  the head of the phrase to its sisters (cf. Rochemont 1996). Finally, I have claimed that contrastive focus overrides normal focus projection mechanisms.  Semantic component The major proposal defended in chapter 3 is that the discourse function of wa-marking parallels that of the L H * contour (B-accent) in English. In particular, wa-marking is equivalent to T-marking in the model of Buring (1998). As such, wa-marking can be viewed as one of the discourse strategies available in Japanese for ensuring that a given assertion is congruent - that is, appropriate - to the question under discussion.  Architecture of the grammar I have assumed that information structure is represented in the syntax via the assignment of Topic/Focus features. Under this view, the picture of the grammar which emerges is schematized in (2). (2) syntax [+F, +T] Since syntax is the level which mediates between PF on the one hand, and L F on the other, these features will be interpreted at both interface components. The model in (2), thus, claims that topic/focus structure is represented at all levels of the grammar. It determines phonological phrasing in the PF component. It triggers movement (be it overt or covert) to discourse configurational positions in the syntax (be it at S-structure or LF). Finally, it triggers interpretative mechanisms of the type proposed in Buring (1997,98) in the semantic component.  177  References Abney, S. (1987) The English noun phrase in its sentential aspect. Ph.D. dissertation. MIT. Amano, M . (1990) "Hukusyukakubunkoo - hiikusyukakubun-no imi to seeritu-ni kakawaru imitekiseeyaku -" Nihongogaku, 9. Arita, S. (1996) "On the function of the Japanese particle wa." Japanese/Korean  6.  Linguistics,  Beckman, M . (1986) Stress and non-stress accent. Dordrecht: Foris. Beckman, M . , S. Hertz, & O. Fujimura (1983) "SRS prosody rules for Japanese." Journal of acoustic society of America,  73.  Beckman, M . & J. Pierrehumbert (1986) "Intonational structure in Japanese and English." Yearbook 3: 255-309.  Phonology  Beckman, M & J. Pierrehumbert (1986) "Japanese prosodic phrasing and intonation synthesis." Birner, B. (1994) "Information status and word order: an analysis of English inversion." Language, 70. Blok, P. (1993) The Interpretation of Focus: An Epistemic Approach Dissertation. University of Groningen. Brennan, V . (1993) Root and Epistemic modal auxiliary  verbs.  to Pragmatics.  Ph.D.  Ph.D. Dissertation. UMass.  Brockett, C. (1991) Wa-marking in Japanese and the syntax and semantics of generic Ph.D Dissertation. Cornell University.  sentences.  Brockett, C. (1992) "The syntax of generics: Japanese evidence for the quantificational model." WCCFL,  11.  Buring, D. (1994) "The interaction of Focus, phrase structure, and quantification." MLT working papers in linguistics,  23.  Buring, D. (1997) The meaning of topic and focus: The 59th street bridge accent. London: Roudedge. Buring, D. (1998) "Focus and topic in a complex model of discourse." ms. Cologne University. Carlson, G. (1977) Reference to kinds in English.  Ph.D.. Dissertation. UMass.  Carlson, G. (1989) "On the semantic composition of English Generic sentences." In G. Chierchia et al (eds.) Properties, types and meaning. Vol. II: Semantic Issues. Dordrecht: Kluwer. Chierchia, G. (1995) "Individual-level predicate as inherent generics." In G. Carlson & F. Pelletier (eds.) The Generic Book. Chicago: University of Chicago Press. Chomsky, N . & H . Lasnik (1995) "The theory of principles and parameters." In N . Chomsky Minimalist program. Cambridge: MIT Press. Cinque, G . (1993) " A null theory of phrase and compound stress." Li 24  178  Condoravdi, C. (1994) Descriptions Cresti, D. (1996) Indefinite  in context. Ph.D. Dissertation. Yale University.  topics. Ph.D. Dissertation. MIT.  Culicover, P. (1991) "Topicalization, Inversion and complimantizers in English." Currah, S. & G. Prideaux (1991) "The pragmatic function of wa in Japanese." Canadian of  Linguistics.  Dechaine, R - M . (1994) "Ellipsis and the position of subjects." NELS,  Journal  24.  de Hoop, H . (1998) "Optional scrambling and interpretation." ms. U I L OTS. de Hoop, H . & H. de Swart (1989) "Indefinite objects." Linguistics  in the  Netherlands.  de Swart, H . (1994) "Position and meaning: Time adverbials in context." To appear in P. Bosch and R. van der Sandt (eds.) The Focus book. Cambridge: Cambridge University Press. Diesing, M . (1988) "Bare plural subjects and the stage/individual contrast" In M . Krifka (ed.) Genericity  in natural language: Proceedings  of the 1988 Tubingen  conference.  Diesing, M . (1992) Indefinites. Cambridge: MIT Press. Drubig, B. (1993) "Zur frage der grammatischen representation thetischer und kategorischer satze." In J Jacobs (ed.) Information  sruktur und grammtik. Opladen: Wetdeutscher Verlag.  Drubig, B. (1997) "Some cross - categorial generalization on Focus structure." Arbeitspapiere des SFB 340, Nr. 109. Eckardt, R. (1996) "Intonation and predication: an investigation in the nature judgement structure." Arbeitspapiere des SFB340. Bericht Nr. 77. En?, M . (1991) The semantics of specificity. LI, 22. Erteschik-Shir, N . (1992) "Topic-chaining and Dominance-chaining." Erteschik-Shir, N . (1996) The Dynamics Negev.  of Focus Structure,  ms. Ben Gurion University of the  Faber, D. (1987) "The accentuation of intransitive sentences in English." Journal Finn, A . (1984) "Intonational accompaniments of Japanese morphemes wa and ga." and Speech,  27  Fujisaki, H . (1989) ""Nihongo-no ontyoo-no bunseki to moderuka." Kooza nihongokyooiku, 2. Tokyo: Meijishoin.  ofLinguistcs Language  nihongo-to  Fukuda, M . (1991) " A movement approach to multiple subject constructions in Japanese." Journal of Japanese  Linguistics  13.  Gbbbel, E. (1995) "On the syntax of topic and focus in Romanian." In I. Kohlhof et al (eds.), Proceedings  of Goettingen  focus  workshop.  179  Grimshaw, J. (1991) "On Extended Projections." ms. Rutgers University. Gussenhoven, C. (1984) On the grammar and semantics of sentence accents. Dordrecht: Foris. Gussenhoven, C. (1992) "Sentence accents and argument structure." In I. Roca (ed.) structure: Its role in grammar. Dordrecht: Foris.  Thematic  Hajicova, E. (1993) Issues of sentence structure and discourse patterns. Institute of theoretical and computational linguistics. Charles University. Haraguchi, S. (1977) The tone pattern of Japanese: An autosegmental Kaitakusha.  Tokyo:  theory oftonology.  Haraguchi, S. (1988) "Pitch-accent and intonation in Japanese." In van der Hulst and Smith (eds.) Haraguchi, S. (1991) A theory of stress and accents. Dordrecht: Foris. Hasegawa, Y . (1993) "Hanasikotoba-ni okeru 'Muzyosi'-no kinoo." Nihongokyooiku,  80.  Herburger, E. (1995) "Presupposition vs. assertion: a Davisonian account." Proceedings workshop  on Focus.  Heycock, C. (1993) "Syntactic predication in Japanese." Journal Heycock, C. (1994) "Focus Projections in Japanese." NELS,  ofUMass  of East Asian Linguistics,  2.  24.  Hinds, J. (1987) "Thematization, assumed familiarity, staging, and syntactic binding in Japanese." In J. Hinds et al (eds) Perspective on Topicalization: The case of Japanese Wa'. Amsterdam: John Benjamins Hirose, K., H . Fujisaki, & H . Kawai (1985) " A system for synthesis of connected speech special emphasis on the synthesis of prosodic features." Onsei kenkyuukai kaiho (The Acoustical society of Japan), S85-43. Hirschberg, J. (1985) A Theory of Scalar Implicature. Pennsylvania.  Ph.D. dissertation, University of  Hitzeman, J. (1997) "Semantic partition and the ambiguity of sentences containing temporal adverbials." Natural Language Semantics, 5. Hobbs, J. (1990) "The Pierrehumbert-Hirschberg theory of intonational meaning made simple: comments on Pierrehumbert and Hirschberg." In P. Cohen, J. Morgan & M . Pollack (eds.) Intentions in Communication. Cambridge: MIT Press. Home, M . (1987) Towards a discourse-based dissertation. Lund University.  model of English  Inkelas, S. & D. Zee (1990) The phonology-syntax Press.  connection.  sentence intonation.  Ph.D.  Chicago: University of Chicago  Inkelas, S. & D. Zee (1996) "Syntax-phonology interface." In J. Goldsmith (Ed) The handbook of phonological theory. Cambridge: Blackwell.  180  Inoue, K. (1982) "An interface of syntax, semantics, and discourse structures." Lingua, 57. Inoue, K. (1984) "Bun-bunpoo to Danwabunpoo-no setten." Gengo Kenkyuu.  84.  Inoue, K. (1989) "Syugo-no imiyokuwari to kakuhairetu." In S. Kuno & M . Shibatani (eds.) Nihongogaku-no sintenkai. Tokyo: Kurosio. Inoue, K. (1991) "Three types of elliptical sentences in Japanese." In O. Kamada & W. Jacobsen (eds.) On Japanese and how to teach it. Tokyo: The Japan times. Iwasaki, S. (1987) "Identifiability, scope-setting, and the particle wa: A study of Japanese spoken expository discourse." In J. Hinds et al (eds) Perspective on Topicalization: The case of Japanese 'Wa'. Amsterdam: John Benjamins Jackendoff. R. (1972) Semantic interpretation  in generative grammar. Cambridge: MIT Press.  Jacobs, J. (1997) "I-Topikalizerung." Linguistische  Berichte,  168.  Jager, G. (1993) Topic, scrambling and aktionsart. ConSOLE,  3.  Jager, G. (1995) Topics in dynamic semantics. Ph.D. Dissertation. Humboldt University. Jager, G. (1996) "Topic, focus and weak quantifiers". To appear in: Bosch, P & R.v.d. Sandt (eds.) The Focus Book. Cambridge: Cambridge University Press. Jager, G. (1997) "Thetic categorical distinctions revisited." WCCFL Jenkins, L. (1972) Modality  in English  15.  Syntax. Ph.D. Dissertation. MIT.  Kadmon, N . & C. Roberts (1986) "Prosody and scope" The role of discourse structure." CLS, 22.  Kaisse, E. (1985) Connected speech. San Diego: Academic Press. Kanerva, J. (1989) Focus and phrasing University.  in Chichewa  phonology.  Ph.D. dissertation. Stanford  Kanerva, J. (1990) "Focusing on phonolgocia phrases' in Chichewa." In S, Inkelas & D. Zee (eds.) Kanerva, J. & L. Gabriele (1995) "Intonation and focus layers." NELS,  25  Kato, Y . (1985) Negative sentences in Japanese. Sophia Linguistica. Kawashima, R. (1994) The Structure of Noun Phrases and the Interpretation NPs in Japanese. Ph.D. Dissertation. Cornell University.  on  Quantificational  Kennedy, R. (1993) Reference and focus in English. Ph.D. dissertation, Harvard University. Kenstowicz, M . & H., Sohn (1996) "Phrasing and focus in Northern Kyungsang Korean." P. Bertinetto (ed.) Certamen phonoloigcum, 3 Kiss, K. (1994) Generic and Existential Bare Plurals and the Classification of Predicates. Working Papers in the Theory of Grammar, Vol. 1, No. 2. Research Institute for Linguistics, Hungarian Academy of Sciences.  181  Proceedings  Kiss, K . (1995) "Focus as a non-uniform phenomenon." In Kohlhof, I. et al (eds.) of the Goettingen Focus workshop, Arbeitsbericht (SFB 340) Nr. 69. Kiss, K. (1995) "Definiteness effect revisited." In I. Kenesei (ed.) Levels and structures to Hungarian,  (Approaches  5.  Kiss, K. (1996) "Two subject positions in English." Linguistic  Review.  Koganemaru, H . (1990) "Muudo-no noda to sukoopu-no noda."  Nihongogaku.  Kohno, T. (1980) "On Japanese phonological phrases." Descriptive  and applied  Koizumi, M . (1993) "Modal and adjuncts". Japanese/Korean  Linguistics,  Kori, S. (1989) "Kyootyoo-to intoneesyon." In Kooza nihongo-to Meijishoin.  linguistics,  2. 2. Tokyo:  nihongokyooiku,  Kori, S. (1996) "Nihongo-no intoneesyon: Kata-to kinoo." In Nihongo-no  13.  onsei. Tokyo: Sanseido.  Kratzer, A . (1991) "The representation of focus." In: Stechow and Wunderlish (eds.) Kratzer, A . (1995) "Stage-level and individual-level predicates." In G. Carlson & F. Pelletier (eds.) Krifka, M . (1988) "The relational theory of genericity." In M . Krifka (ed.) Genericity in natural language.  Krifka, M . (1992) " A framework for focus-sensitive quantification." In C. Baker & D. Dowty (eds.) SALT,  2.  Krifka, M . (1992) " A compositional semantics for multiple focus constructions." In J. Jacobs (ed.) Informationsstruktur und grammatik. Linguistische  Berichte,  4.  Krifka, M . (1993) "Focus and presupposition in dynamic interpretation." Journal 10.  of  semantics,  Krifka, M . (1998) "Scope inversion under the Rise-Fall Contour in German." LI, 29. Kubozono, H. (1991) "Syntactic and rhythmic effects on downstep in Japanese." Phonology,  6.  Kubozono, H . (1992) On the metrical structure of Japanese downstep. In W. Dressier et al. (eds.) Phonologica 1988. Cambridge: Cambridge University Press. Kubozono, H . (1993) "Temporal regulation of phonological prominence." ms. Kubozono, H . (1993) The organization  of Japanese prosody.  Tokyo: Kurosio.  Kubozono, H. (1996) "Akusento/Intoneesyonkoozoo to bunpoo." In M . Sugito (ed.) no sinkenkyuu. Tokyo: Sanseido.  Nihongoonsee-  Kuhn, J. (1996) "Context effects on interpretation and intonation." ms. University of Stuttgart. Kuno, S. (1973) The Structure of the Japanese Language.  Cambridge. MIT Press.  182  Kuno, S. (1976) "Subject, theme, and the speaker's empathy - a reexamination of relativization phenomena. In C. L i (ed.) Subject and Topic. New York: Academic Press. Kuno, S. (1978) Danwa-no  bunpoo. Tokyo: Taishukan.  Kuno, S. (1982) "The focus of the question and the focus of the answer." CLS, 18. Kuno, S. (1991) "Passivization and thematization." In O. Kamada & W . Jacobsen (eds.) On Japanese and how to teach it. Tokyo: The Japan times. Kuroda, S-Y. (1965) Generative Grammatical MIT.  Studies in the Japanese Language.  Ph.D. Dissertation.  Kuroda, S-Y (1992) Japanese Syntax and Semantics: Collected Papers. Dordrecht: Kluwer Academic Publishers Group. Ladd, R. (1996) Intonational  phonology.  New York: Cambridge University Press.  Ladsaw, W . (1994) "Thetic and categorical, stage and individual, weak and strong." SALT, Lambrecht, K. (1994) Information representations  of discourse  IV.  structure and sentence form: Topic, focus and the mental referents. Cambridge: Cambridge University Press.  Larson, R. (1988) "On the nature of grammatics relations." LI, 19. Liberman, M & J. Pierrehumbert (1984) "Intonational invariants under changes in pitch range and length." In M . Aronoff & R. Oehrle (eds.) Language sound structure. Cambridge: MTT Press. Link, G. (1983) "The logical analysis of plurals and mass terms: A lattice-theoretical approach." In R. Bauerle, et al (eds.) Meaning, use and interpretation of language. Berlin: de Gruyter. Maekawa, K. (1991) "Perception of intonational characteristics of wh and non-wh questions in Tokyo Japanese." In Proceedings of the 12th international congress of phonetic sciences. Maekawa, K. (1994) "Is there 'dephrasing' of the accentual phrase in Japanese?" OSU papers in linguistics,  44.  wroking  Maruyama, N . (1996) "Zyosi-no daturakugansyoo." Masunaga, K . (1982) "Bridging." Proceeding Masuoka, T. (1987) Meedai-no Masuoka, T. (1992) Modaritii-no  of the Xlllth  International  Congress  of  linguistics.  bunpoo. Tokyo: Kurosio. bunpoo. Tokyo: Kurosio.  Masuoka, T. & Y . Takubo (1987) Kiso Nihongobunpoo. McCawley, J. (1968) The phonological  Tokyo: Kurosio  component of a grammar  of Japanese.  McCawley, J. (1977) "Accent in Japanese." In L. Hyman (ed.) McCawley, J. (1978) "What is a tone language?" In V . Fromkin (ed.)  Hague: Mouton.  183  McGloin, N . (1987) "The role of wa in negation." In J. Hinds et al (eds) Perspective Topicalization: The case of Japanese Wa'. Amsterdam: John Benjamins  on  McGloin, N . (1991) "The pragmatics of object topicalization in Japanese." In O. Kamada & W . Jacobsen (eds.) On Japanese and how to teach it. Tokyo: The Japan times. McNally, L. (1995) "On recent formal analyses of "topic"." ms. The Ohio State University. Meinunger, A . (1995) Discourse Potsdam.  dependent DP (de-) placement.  Ph.D Dissertation. University of  Mihara, K. (1990) "Takakusykaku koobun-o megutte." Nihongogaku, Mihara, K. (1994) Nihongo-no Milsark, G . (1974) Existential  toogokoozoo.  Tokyo: Shohakusha  sentences in English.  Minami, F. (1974) Gendainihongo-no  9.  Ph.D. Dissertation. MIT  koozoo. Tokyo: Taishukan  Minami, F. (1993) Gendainihongobunpoo-no  rinkaku.  Tokyo: Taishukan.  Miyagawa, S. (1987) "Wa and W H phrases." In J. Hinds et al (eds) Perspective The case of Japanese Wa'. Amsterdam: John Benjamins Nagahara, H . (1994) Phonological  phrasing  on  Topicalization:  in Japanese. Ph.D. Dissertation. U C L A .  Nakamura, M . (1996) "Topicalization, Neg-preposing, and locative preposing." Nakayama, M . & M . Koizumi (1991) "Remarks on Subjects in Japanese." Lingua Nesper, M . & I. Vogel (1986) Prosodic  phonology.  Dordrecht: Foris.  Nesper, M & I. Vogel (1989) "On clashes and lapses." Phonology,  6.  Noda, H . (1996) "wa" to "ga". Tokyo: Kurosio Numata, Y . (1992) "Toritatesi to siten" Nihongogaku,  11-8  Ogihara, T. (1984) "Strong and weak determiners and wa and ga in Japanese." tame-no gengogaku-ni  kansuru  kenkyuu.  Ogihara, T. (1989) "Topics, definiteness and predication." Texas Linguistics  Forum,  Okazaki, M . (1995) " A semantic analysis of sentence accent assignment in English." 107. Ono, M . (1990) From Sentence to Discourse: Phenomena  in Japanese.  Integrated Explanations Ph.D. Dissertation. U B C .  Partee, B. (1991) "Topic, focus and quantification." SALT,  Sizengengosyori-no  for Certain  31.  Gengokenkyuu, Linguistic  1.  Partee, B. (1992) "Syntactic categories and semantic type." In M . Rosner & R. Johnson (eds.) Computational linguistics and formal semantics. Cambridge: Cambridge University Press.  184  Partee, B. (1996) "Allegation and accommodation." In B. Partee & P. Sgall (eds.) Discourse and meaning. Amsterdam: John Benjamins. Peregrin, J. (1996) "Topic and focus in a formal framework." In B. Partee & P. Sgall (eds.) Discourse and meaning. Amsterdam: John Benjamins. Pierrehumbert, J. & M . Beckman. (1988) Japanese tone structure. Cambridge. M I T Press. Pierrehumbert, J. & J. Hirschberg. (1990) "The meaning of intonational contours in the interpretation of discourse." In P. Cohen, J. Morgan & M . Pollack (eds.) Intentions in communication. Cambridge: MIT Press Portner, P. & K. Yabushita (1996) "The semantics and pragmatics of topic phrases." ms. Georgetown University & Naruto University of Education. Poser, W. (1984) The phonetics and phonology oftone and intonation in Japanese. Ph.D. Dissertation. MIT. Prince, A & P. Smolensky (1993) Optimality theory: constraint interaction  in generative  grammar.  Prince, E. (1981) "Topicalization, Focus-movement, and Yiddish-movement: a pragmatic differentiation." BLS. Prince, E. (1981) "Toward a taxonomy of given-new information. In P. Cole (ed.) pragmatics. New York: Academic Press.  Radical  Prince, E. (1988) "Discourse Analysis: A Part of the Study of Linguistics Competence." In Newmeyer (ed.) Prince, E. (1996) "On the Emits of syntax, with reference to Left-Dislocation and Topicalization." To appear in P. Culicover & L. McNally (eds) The limits of syntax. Reichenbach, H. (1947) Elements of symbolic logic. New York: Macmillan. Reinhart, T. (1981) "Pragmatics and linguistics: an analysis of sentence topics." Philosophica,  27.  Reinhart, T. (1996) Topics and the conceptual interface." In H. Kamp & B.Partee (eds.) Context-dependence  in the analysis of linguistics  meaning.  Rizzi, L. (1995) "The fine structure of the left periphery." ms. University of Genenve. Roberts, C. (1996) Information structure in discourse: Towards an integrated formal theory of pragmatics." OSU working papers in linguistics, 49. Rochemont, M . (1986) Focus in generative grammar. Philadelphia. John Benjamins Publishing Company. Rochemont, M . (1996) "Phonological focus and structural focus." To appear in P. Culicover & L. McNally (eds.) Limits of syntax. Orlando: Academic Press. Rochemont, M . (1997) "Phrase structure relation and the syntax-phonology mapping" ms. U B C . Rooth, M . (1985) Association  with Focus. Ph.D. Dissertation. UMass.  185  Rooth, M . (1992) " A theory of focus interpretation." Natural Language  Semantics,  1.  Saito, M . (1982) "Case marking in Japanese: A preliminary study, ms. MIT. Saito, M . (1985) Some asymmetries in Japanese and their theoretical consequences. Dissertation. MIT.  Ph.D..  Sawada, H . (1993) Siten to syukansee. Tokyo: Hituzi Schmerling, S. (1976) Aspects of English  stress. Austin: University of Texas Press.  Schwardzchild, R. (1996) "Givenness and optimal focus." ms. Rutgers University. Selkirk, E. (1984) Phonology  and syntax. Cambridge: MIT Press.  Selkirk, E. (1986) "On derived domains in sentence phonology." Phonology  Yearbook,  3.  Selkirk, E. (1990) "Comments on Beckman and Edwards Paper." In J. Kingston & M . Beckman (eds.) Papers in laboratory phonology I: Between grammar and the physics of speech. Cambridge: Cambridge University Press. Selkirk, E. (1995) "Sentence prosody: Intonation, stress and phrasing." In J. Goldsmith (ed.) The handbook of phonological theory. Cambridge: Blackwell. Selkirk, E. (1996) "The prosodic structure of function words." In J. Morgan and K. Demuth (eds.) Signal to syntax: Boothstrapping form speech of grammar in early acquisition. New Jersey: Lawrence Erlbaum Associates. Selkirk, E. (1996) "The interaction of constraints on prosodic phrasing." ms. Selkirk, E. & T. Shen (1990) "Prosodic domains in Shanghai Chinese." In S. Inkelas & D. Zee (eds.) p313-337. Selkirk, E. & K. Tateishi (1988) "Constraints on minor phrase formation in Japanese." CLS, 25. Selkirk, E. & K. Tateishi (1991) "Syntax and phonological phrasing in Japanese. In Georgopoulos & Ishihara (eds.) Shibatani, M . (1990) The languages of Japan. Cambridge: Cambridge University Press. Shirai, K . (1987) Japanese Noun-phrases and particles wa and ga." In. J. Groenedndijk et al (eds.) Studies in Discourse presentation theory and the theory of generalized quantifiers. Dordrecht: Foris. Suzuki, S. (1995) "The functions of topic-encoding zero-marked phrases: A study of the interaction among topic-encoding expressions in Japanese." Journal of Pragmatics, 23. Szabolcsi, A . (1981) "The semantics of topic-focus articulation." In J. Groenendijk et al. (eds.) Formal methods in study of language, vol 2. Amsterdam: Mathematisch Centrum. Szwedek, A . (1990) "What is topic? A conrrastivist's view." In J. Fisiak (ed.) Further into contrastive analysis. Amsterdam: John Benjamins.  insights  186  Szwedek, A . (1991) "Sentence stress on proper names." Acta universitatis 1061.  No.  wratislaviensis.  Takami, K. & Kamio, A (1996) "Topicalization and subjectivization in Japanese: Characterizational and identificational information." Lingua, 99. Takubo, Y . (1985) "On the scope of negation and question in Japanese." Papers in Japanese Linguistics,  10.  Tancredi, C. (1997) "Focus types and associatative operators." ms. Yokohama National University. Tateishi, K (1989) "Subjects, SPEC, and DP in Japanese." NELS,  19.  Tateishi, K. (1993) The syntax of subjects. Stanford: CSLI. Terakura, H . (1984) "Discourse function of sentence-initial NP in conversation." Descriptive applied linguistics, XVII. Tokyo: ICU.  and  Tonoike, S. (1989) "wa, mo, ga-no ronrikeesiki: Bunbunpoo to danwabunpoo-no intaafeesu." ms. Meijgakuin University. Truckenbrodt, H . (1995) Phonological Ph.D. Dissertation. MIT.  phrases:  Their relation  to syntax focus, and  prominence.  Tsubomoto, A . (1989) "Null subject phenomena in Japanese: Incorporation, null expletives, and topic-agreement." English Linguistics, 6. Tsumaki, J. (1994) "Intonational properties of adverbs in Tokyo Japanese." Proceedings international  conference  on spokedn languaage  processing.  of the  Uechi, A . (1994) "Revised Mapping Hypothesis." In P. Bosch & R. van der Sandt (eds.) Focus and natural language processing,  vol. 1.  Uechi, A . (1995) "Focus phrasing in Japanese: An optimality approach" UMass focus.  Uechi, A . (1996) "Toward Syntax-Information Mapping." Japanese/Korean  workshop  Linguistics,  Uechi, A . (1996) "The discourse function of wa-marking in Japanese." NWLC,  on  5.  13.  Uechi, A . (1997) "Focus projection and stage/individual distinctions." In K. Shahin (et al.) Proceedings  of Northwest  Linguistics  Conference.  UBC.  Ueyama, A . (1989) "Focus-no ga to Nihongo-no kukoozoo." Kansai Linguistic  Society,  9.  Uyeno, T., H . Hayashibe, K. Imai, H . Imagawa, & S, Kiritani (1981) "Syntactic structure and prosody in Japanese: A case study on pitch contours and the pauses at phrase boundaries." Annual Bulletin (Research Institute of Logopaedics and Phoniatrics, University of Tokyo) 15. p91-108. Vallduvi, E. (1992) The informational  component.  New York: Garland  Vallduvi, E. (1994) "Information packaging:A survey." ms. University of Edinburgh.  187  Vallduvi, E. & R. Zachaski (1994) "Accenting phenomena, association with focus, and the recursiveness of focus-ground." In P. Dekker & M . Stokfhof (eds) Proceedings of the ninth Amsterdam  colloquium.  van Hoof, H . (1992) "Focus projection and the double subject hypothesis." ms. Venditti, J., S-A Jun, & M . Beckman (1996) "Prosodic cues to syntactic and other linguistics structures in Japanese, Korean, and English." In J. Morgan & K. Demuth (eds.) Signal to syntax: Boothstrapping form speech of grammar in early acquisition. New Jersey: Lawrence Erlbaum Associates. Vogel, I. & I. Kenesei (1990) "Focus and phonological structure." ms. von Fintel, K. (1994) Restrictions  on Quantifier  Domains.  Ph.D. Dissertation. UMass.  von Stechow, A . & S. Uhmann (1986) "Some remarks on focus projection." In W. Abraham and S. de Meij (eds.) Topic, focus and configurationality. Amsterdam: Benjamins. Ward, G. (1985) "Discourse function of preposing." Ph.D. Dissertation. University of Pennsylvania. Ward, G. & B. Birner (1994) " A unified account of English fronting constructions." University Pennsylvania  working papers  in Linguistics,  1.  Ward, G. & E. Prince (1991) "On the Topicalization of indefinite NPs." Journal of 16.  of  Pragmatics,  Westmoreland, R. (1994) "The relation of topic to focus-frame in topic-prominent languages." ms. Indiana University. Williams, E. (1997) "Blocking and anaphora." LI, 28. Williams, N . (1990) Japanese intonation: Meaning and intonational Georgetown University.  phrases.  Ph.D. dissertation.  Winkler, S. (1994) Secondary predication in English: A syntactic and focus-theoretical Ph.D. dissertation. University of Tubingen. Yamazaki, M . (1990) "Hitee-no syooten." Nihongogaku,  approach.  9-12  Yamada, S. (1993) "Modaritii to hitee." Kokubungakukaisyaku  to kansyoo,  58-1.  Yoshimoto, K. (1982) "Wa to ga - sorezore-no kinoosuru reberu-no tigai-ni tyuumokusite -." Gengokenkyuu,  81.  Yoshimoto, K . (1993) "Nihongo-no bunkaisookoozoo to syudai, syooten, zisee." 103.  Gengokenkyuu,  Zwart, J-W. (1994) "Word order, intonation, and noun phrase interpretation in Dutch.  WECOL.  

Cite

Citation Scheme:

        

Citations by CSL (citeproc-js)

Usage Statistics

Share

Embed

Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                        
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            src="{[{embed.src}]}"
                            data-item="{[{embed.item}]}"
                            data-collection="{[{embed.collection}]}"
                            data-metadata="{[{embed.showMetadata}]}"
                            data-width="{[{embed.width}]}"
                            async >
                            </script>
                            </div>
                        
                    
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:
http://iiif.library.ubc.ca/presentation/dsp.831.1-0099375/manifest

Comment

Related Items