Rightward Movement Phenomena in Linguistics [1 ed.] 1527511472, 9781527511477

Citation preview

Rightward Movement Phenomena in Linguistics

Rightward Movement Phenomena in Linguistics Kohji Kamada

By

Cambridge Scholars Publishing

Rightward Movement Phenomena in Linguistics ByKohjiKamada This book first published 2018 Cambridge Scholars Publishing Lady Stephenson Library, Newcastle upon Tyne, NE6 2PA, UK British Library Cataloguing in Publication Data A catalogue record for this book is available from the British Library Copyright© 2018 byKohjiKamada All rights for this book reserved. No part ofthis book may be reproduced, stored in a retrieval system, or transmitted, in any fonn or by any means,

electronic, mechanical, photocopying, recording or otherwise, without the prior pennission ofthe copyright owner. ISBN (10): 1-5275-1147-2 ISBN (13): 978-1-5275-1147-7

CONTENTS

PREFACE

....................................................................................................

ABBREVIATIONS AND SPECIAL SYMBOLS RELATING TO ACCEPTABILITY CHAPTER

Vll

....

IX

ONE ............................................................................................. 1

INTRODUCTION

1.1 Two main questions ......................................................................... 2 1.2 Outline ............................................................................................. 3 CHAPTER

Two ............................................................................................. 7

GENERAL DESCRIPTION

2.1 Introduction ..................................................................................... 7 2.2 Japanese ........................................................................................... 7 2.3 Rightward Movement Phenomena in English ............................... 26 2.4 Rightward Movement Phenomena in Other Languages ................ 41 2.5 Summary ....................................................................................... 64 CHAPTER THREE

........................................................................................ 66

A THEORETICAL FRAMEWORK 3.1 Introduction ................................................................................... 66 3.2 The Grammar ................................................................................ 66 3.3 The Parser ...................................................................................... 77 3.4 Conclusion .................................................................................... 91 CHAPTER FOUR

......................................................................................... 92

THE POSTVERBAL CONSTRUCTION IN JAPANESE

4.1 Introduction ................................................................................... 92 4.2 Previous Studies ............................................................................ 92 4.3 A Proposa!: Another Type ofNomnovement Approach .............. 141 4.4 Deriving the Properties of JPVCs ............................................... 170 4.5 Island Effects and Parsing Strategies ........................................... 181 4.6 Linear Distance Effects ............................................................... 231 4.7 Conclusion .................................................................................. 243

VI

Contents

CHAPTER FIVE 244 A CROSS-LINGUISTIC PERSPECTIVE 5.1 English Rightward Movement Constructions 244 5.2 NP Shift Constructions from a Cross-linguistic Perspective ....... 281 5.3 Conclusion .................................................................................. 298 .........................................................................................

..............................

CHAPTER

SIX ........................................................................................... 300

CONCLUSION ........................................................................................

304

......................................................................................................

325

BIBLIOGRAPHy

INDEX

PREFACE

This book is based on my Ph.D. dissertation, which I submitted to the University of Edinburgh in 2009. It deals with the properties of what is called the "rightward movement phenomena in human language," from which it took its title. Many of the theoretical assumptions here and conclusions are almost the same as in the dissertation. In Chapter 4, however, I propose the revised licensing condition and interpretive rules for adjoined phrases, and one more subsection is added for multiple post­ verbal phrases. Another difference from the dissertation is in Chapter 5, where rightward movement phenomena are discussed in terms of comparative syntax, with more linguistic data added. I composed my original dissertation while I was a PhD. candidate at the University of Edinburgh between 2005 and 2009. Many kind people helped me during that time. First and foremost, I wish to express my deepest gratitude to Peter Ackerna, who served as my first supervisor during my final three years, for his meticulous feedback and constant encouragement I am also extremely grateful to Jim Hurford, who served as my supervisor during my first two years, for his kindness and generosity. I also wish to thank Caroline Heycock, who served as my second supervisor during my final two years, for her useful comments and remarks. Furthermore, I must thank Simon Kirby for serving as my second supervisor during my first year. I am also deeply thankful to Masaru Kajita for having introduced me to the study of generative grammar. Finally, I also wish to thank Naoki Fukui and Yasuhiko Kato for affording me opportunities to give talks to their classes about the ideas advanced in the thesis. At various stages, many people have commented on parts of this study, or were otherwise of help. Among them, I would like to thank Daigo Akiba, Masahiro Akiyama, Dan Dediu, Kazuaki Ezure, Hideaki Gen-ei, Nikolas Gisbome, David Hawkey, Stefan Hofler, Yoshiyuki Igarashi, Masayuki Ike-uchi, Akira Ikeya, Takao Ito, Sachie Kajita, Yasuki Kamiyama, Anna Parker-Kinsella, Mayumi Kumada, Kaori Miura, Atsuko Miyajima, Tsuguro Nakamura, Sakae Ohwada, Ryoya Okabe, Kyoko Otsuki, Graham Ritchie, Emi Sakamoto, Nobufumi Sasaki, Julia Schultz, Tohru Seraku, Andrew Smith, Kenny Smith, Hiroshi Takahash� Koichi Takahashi, Mineko Takahashi, Tomokazu Takehisa, Monica Tamariz,

V111

Preface

Mitsuo Tani, Hajime Yamauchi, Yoko Takahashi-Yamauchi, Keiichi Yasu, Masaya Yoshida, Tomoko Yoshino, and the audiences at various conferences and seminars in which I presented my ideas. Last but not least, my special thanks go to my mother, Michiko, and my sister, Michiyo, for their support. This work was partially supported by JSPS KAKENID Grant Number 26370439 and a grant from Sophia Linguistic Institute for International Communication, Sophia University, Tokyo.

A BBREVIATIONS AND SPECIAL SYMBOLS R ELATING TO A CCEPTABILITY

Abbreviations

A Abl Ace ADJ Adv CFC CI Camp Conj CSC CRP Dat DO EA EC ECM ECP EPP EX FGD FP Gen GTA HNPS

adjective ablative accusative adjunct adverb core functional category classifier cornplernentiser conjlUlction Coordinate Structure Constraint Case Resistance Principle dative direct object external argurn ent empty category Exceptional Case Marking Empty Category Principle Extended Projection Principle Extraposition from NP filler-gap domain sentence final particle genitive Generalised Theta Attachinent Heavy NF Shift

Han

10

IR JPVC LA LBC LC Lit

ME MiD MP

N nla Neg NFl NFSC Nom NonP NS OBJ OE OLLC Op P Poss PVC

Q

honorific indirect object Interpretive Rule Japanese postverbal construction lexical array Left Branch Condition Licensing Condition literal Magnitude Estimation 11inirnise Domain minimalist program noun not applicable negative negative polarity item NF Shift construction nominative non-past narrow syntax object Old English On-Line Locality Constraint null operator postpositionl preposition posseSSIve postverbal construction question particle/quantifier

x

RA

RD RDC RNR RRC sg SUBJ

Abbreviations and Special Symbols Relating to Acceptability right association principle Right Dislocation Right Dislocation construction Right Node Raising Right Roof Constraint singular subject

Top UG UMC

topic Universal Grammar Unambiguous Modification Condition UREC unconSClQUS reinterpretation condition V verb

Special Symbols Relating to Acceptability

*, 7*, 7?, 7 Unacceptability, decreasing in degree from * to 7 ! Indicates "garden path sentences." #

&

Indicates semantic deviance. Indicates a different interpretation from what is intended.

CHAPTER ONE INTRODUCTION

The aim of this work is (I) to demonstrate that some properties of "rightward movement phenomena" (a cover term referring to sentences in which an element appears to be "displaced" to the right) may be derived from syntactic principles and interface conditions within the framework of the generative grammar/minimalist program, and (2) to argue that certain properties, which up to now have been dealt with purely in syntax, can be better accounted for in terms of language processing; accordingly, the human parser should undertake explanations of part of the competence system's output (see e.g. Ackema and Neeleman 2002; Frazier 2013a; Gibson 1998, 2000; Hofmeister et a1. 2013, 2015; Kluender 1998, 2004; cf Goodluck, Saah, and Tsiwah 2015)1 Generally, few theoretical linguists seem to take into consideration what psycholinguists do, and vice versa, although the relationship between the competence system (the grammar) and the performance systems (the human parser/processor) has been occasionally discussed (e.g. Belletti and Rizzi. 2013; Berwick, Abney, and Tenny 1991; Berwick and Weinberg 1984; Bresnan and Kaplan 1982; Crocker 1996; Culicover 2013; Gorrell 1995; Guasti 2014; Hawkins 2004, 2014; Koot 1990; Lewis and Phillips 2015; Mulders 2002; Neeleman and Koot 2010; Phillips 1996, 2003; Pritchett 1992; Prichett and Whitman 1995; Reinhart, 2006; Siloni 2014, etc.).2 As far as I know, with respect to rightward movement phenomena, a theoretical linguistic discussion that takes processing into account from the minimalist perspective has scarcely been initiated (cf Chesi 2013). My proposed analysis of these phenomena leads to the conclusion that phrasal 1 Trotzke, Bader and Frazier (20 13, 16) argue that some language universals can be relegated to the independently motivated systernaticity of the performance systems, which is compatible with the reduction ofUG to a minimum (Chomsky 2005). See also Good]uck and Zweig (20 1 3), Phillips (2013), and references therein. 2 The articles, books, and theses listed in the parentheses are limited to those published after the so-called ''principles-and-pararneters approach" period in the history of generative granunar.

2

Chapter One

rightward movement rules in syntax fail to follow specific principles. At first glance, this conclusion seems identical with Kayne's (1994) claim that no rightward movement rules exist. However, his and my work provide completely different grounds for the absence of rightward movement rules, meaning that the present work presents an original view of rightward movement phenomena. 1,1 Two Main Questions

In general, Japanese is a verb-final language. In colloquial speech, however, a phrase frequently follows a verbal element, as exemplified in (1)' (1) , ] hon-o age-masita (yo), Taro-ga, Kenji-nij book-Acc give-Past FP Taro-Nom Kenji-Dat IKenji-nij Taro-gal. IKenji-Dat Taro-Nom

Lit

", ]

gave a book, Taro! to Kenjij/to Kenjij Taro!."

C/Jj age-masita (yo), Taro-gai hon-oj FP Taro-Nom book-Acc Taro-gai. Ihon-oj lbook-Acc Taro-Nom Lit ", gave ] to Kenji, Taro! a booky a bookj Taro!." d. , } k age-masita (yo), Taro-ga, Kenji-nij han-a. give-Past FP Taro-Nom Kenji-Dat book-Acc IHon-ok Kenji-nij Taro-gal. Ibook-Acc Kenji-Dat Taro-Nom Lit ", gave ] k, Taro! to Kenjij a boo.... "

c.

tP! Kenji-ni

Kenji-Dat give-Past

There is a different type of multiple postverbal element (23) Taro-ga , itte-ta yo, [Mari-ga Said Manager Kim Kessels: "This is our best result in 20 years! " d. More important are the moral objections. (Huddleston and Pullum 2002, 1382, 1 384) e. There walked into the room a man with long blond hair. (Rochemont and Culicover 1 990, 1 1 6) f There visited us last night a large group of people who travelled all the way from India.

(Chomsky 1995, 343) The comparative construction allows for the postposing of the subject, as in (46a). In (46b, c), verbs of reporting are used. (46b) shows that subj ect shifting can occur in the parenthetical position. According to Huddleston and Pullum (2002), subject shifting is also found inj ournalistic style, as in (46c). In (46d), subject postposing is accompanied by another element that is being preposed In (46e, I), insertion of there in the subject position enables the sentences to become acceptable. It may be that the insertion of there is a kind of resumptive pronoun strategy, although the question would arise why there is used instead of a pronoun such as it (see also 2.3.3, where Right Dislocation is discussed). 2.3.1.2 Subject position in the non-tensed clause

Rightward extraction of an NP from the subject position of a small clause is possible, as shown in (47). (47) a. I consider '1> stupid anyone who would support a Socialist bid for power.

b. I saw '1> leaving the room last night a man with long bond hair. (Rochemont and Culicover 1 990, 126, 19ln38)

28

Chapter Two

c. I would consider >4 4-------- P SEM

::::

t

[

R ( s pel out)

I : !�� t

: op PHON+-44f--- p ase2 i

TRANSFER (Spell-Out) piece by piece, cyclically. There are, therefore, no LF properties and no interpretation of LF, strictly speaking, though :L: and cl> interpret units that are part of something like LF in a noncyclic conception." Hence, it seems to me that :L: and SEM could be regarded as fimctionally corresponding to LF plus semantic comp:ments and LF plus semantic representations in a GB theory, respectively (see also Ike-uchi 2003, 12). 7 EPP stands for the Extended Projection Principle, the property of which is assumed to enable each CFC to allow an extra Spec. S See footnote 1 2 for order in NS.

A Theoretical Framework

(2)

69

� CP

Sp

C

''A '" � A A � Spec

(EA)

v

VP/AP

3.2.3 Merge In this section. I lay out the assumptions that I will adopt or make to account for syntactic properties of JPVCs. First, I follow Chomsky (1995. 225; 2000. 1 1 3. 1 1 8; 2004b. 1 07) in adopting the Inclusiveness Condition as formulated in (3). assuming that the condition in (3) holds for NS alone. (3) Inclusiveness Condition: No new elements are added in the course of computation. It follows from (3) that no indices. no phrasal categories. and no intermediate nodes are admitted because such elements are not included in the lexicon. 9 Thus, principles referring to indices such as binding principles are no longer in NS. and they should be reformulated as interpretation rules in the interface (SEM) or the semantic component (see Chomsky 2008). More importantly. X-bar theory is also reformulated in terms of bare phrase structure (i.e. the elimination of X-bar theory). Ai; for structure-building, it is therefore assumed that a syntactic structure is 9 I aSSlUlle that traces should exist in the lexicon, although it is claimed (e.g. Chomsky 1995) that traces are replaced by copies in the copy theory of movement, which states that ''the trace left behind is a copy of the moved element, deleted by a principle of the PF comp:ment in the case of overt movement. But at LF [= SEM] the copy remains, providing the materials for 'reconstruction'" (Chomsky 1995, 202).

70

Chapter Three

constructed by the operation called Merge, as mentioned above. For example, this operation takes two syntactic objects a and � and then fonns from them a new object, as illustrated in (4). (4)

l\

Merge also determines the label of a new larger object by projection of a

head (Chomsky 1995, 243-245). 10 If a is the head, the new object will be (Sa), and if � is the head, the new object will be (5b). (5) a.

b.

It is worth noting that as Chomsky (2005, 16) points out, "[w]ithout further stipulations, external Merge yields n-ary constituents." l 1 In other words, it is not strictly necessary that a structure be formed in a binary way. Hence, it is assumed here that a constituent can have more than two branches (see also Chomsky 2004a, 168; Fukui 201 1 , 80n?). It is also important to note that the structures in (4) are informal notation only: it is assumed in MP that syntactic objects are fonnulated as sets like {a {a, � } } for (Sa) (Chomsky 1995 24lff; 2004b, 109; 2005, 14). However, I follow Saito and Fukui (1998) in assuming that order is introduced in NS 12 Let us consider the simple example in (6) to see how Merge forms a syntactic structure. (6) John loves Mary. Phrase structures are built in a bottom-up fashion. First, Merge puts together Mary and loves, which has an assign-accusative Case feature,

10 For a discussion oflabelling, see Chomsky (2013, 201 Sa). 1 1 There are two subcases of Merge: external and internal Merge (see Chomsky 2005, 12). Internal Merge corresponds to the so-called "Move." External Merge is Merge that is not part of internal Merge. The operation llllder discussion is external Merge. 12 The MP has assumed that there is no order in NS except for adjllllcts (see Chomsky 2004b, 1 1 7ff.). Further, Bruening (2014 384) states that "linear order must be decided early in the syntax, at Merge. Merge must create an ordered set .

A Theoretical Framework

71

detennining that loves projects, because loves selects Mary as a direct object The structure built thus far is as follows: (7)

loves

�Mary

loves

At this stage of derivation, Mary (internal argument) is assigned a theta­ role by the verb, which follows from the theta-theoretic principle given in (8) 13 (8) The theta-theoretic principle: External Merge ill theta-position is required of (and restricted to) argurn ents. Adapted from (Chomsky 2000, 103)

VP

Next, the is merged with the light verb v, to which loves adjoins. 14 Mary checks its Case feature against the Case featme of the light verb, and the Case feature is deleted (henceforth, strikeout is used to indicate deleted elements). IS , 16 (9)

�VP V /':--,. A v

loves tv

Mary

fAee1

Then, John (EA) is merged with the vP and it is assigned a theta-role by v.

1 3 The principle in (8) states that the theta-roles are regarded as "a relation between two syntactic objects, a configuration and an expression selected by its head" (Chomsky 2000, 103). 1 4 I assume that V raises to v obligatory in NS (see Chomsky 1995, 331 ; 2004b, 1 1 2, 122). 1 5 The precise mechanisms for Case-assigmnentlchecking are not discussed here. 16 For ease of exposition, I will use informal notations like X' and XP in the present work.

72

(10)

Chapter Three

A �

John

A AMary

v

loves tv

Thereafter, the vP is merged with T, which is assumed to have an EPP­ feature and an assign-nominative Case feature at least Finally, as (II) shows, John raises to the specifier position of TP, where John checks its Case feature against the Case feature of the finite T, and the EPP and the relevant Case featmes are deleted.

(II)

3,2,4 Two Internal Arguments Based on the principle in (8), two internal arguments yP and ZP in a ditransitive verb are assigned in the following manner One of them (e.g. YP) is first merged with V, as shown in (l 2a). The other is then merged with V as another sister of V in (l2b) or a sister of V' in (l 2c). Since the structure in (l2b) is possible without further stipulations (see 3.2.3), I assume in this work that a ditransitive verb has a structure like that in ( l 2b) unless there is evidence to the contrary 17

1 7 See e.g. Larson (1988) and Citko (201 1 ) for another possible structures.

A Theoretical Framework

(12) a.

)\YP

V

b.

�ZP

c.

V yp

73

�

VP

t'--...

V'

ZP

V

YP

3.2.5 Phases As mentioned earlier, derivations proceed phase by phase. Then the question will arise as to what the phases are. I assume that the phases are CP and Vp 18 Given a phase � [a [ H �]], where H is a head of a phase, � is called the domain of H, and a the edge of a phase. At a phase level, TRANSFER hands � over to and to Z, not a whole phase. Hence, the Phase Impenetrability Condition would follow. (13) The domain of H is not accessible to operations [that apply outside the phase], but only the edge ofHP. (Chomsky 2004b, 108) "Subjacency" effects are derived from the Condition in (13). As Chomsky (2000, 1 1 0) points out, syntactic objects are fonned in parallel by NS. In the course of derivation of (14), for instance, each phase in (IS) is the syntactic object built via repeated Merge based on the lexical items that a subarray provides for a new phase, with steps (I Sa) and (l 5b) unordered. (14) The demonstration that glaciers are receding showed that global warming must be taken seriously. (Chomsky 2000, 1 1 0) (IS) a. Phase! � [cp lhat global wanning must be taken seriously] b. Phase, � [cp that glaciers are receding] c. Phase, � [,p [the demonstration Phase, [show Phase ! ]]] d Phase, � (14) Adapted from (Chomsky 2000, 1 1 0) Phase, is built from the subarray {the, demonstration, show} by repeated Merge to yield [the demonstration Phase,] as well as [show Phase ! ], with 18 Chomsky (2005 17) mentions that DP may also be a phase. However, I assrune here that DP is not a phase. See also Chomsky (2008, 2012) and Citko (2014).

74

Chapter Three

theta-roles assigned. The subarray containing T and C is then submitted to the derivation to proceed to Phase4 . 3.2.6 Adjuncts The principle in (8) does not refer to adjuncts. I assume that adjunction structures are formed by External Merge. 19 If External Merge is costless. nothing will prohibit elements from adjoining to any elements unless syntactic principles are violated. There is evidence that some adjuncts must be considered as verb-phrasaL As observed by Takami (1985). some adverbial clauses are verb-phrasal. as shown in (16). 20 (16) a. What John did was [,p go to bed [before he finished his ho mework]] b *What John did before he finished his homework was [,Pgo to bed]. c. I wanted John to go to bed before he finished his homework, and [,p go to bed [before he finished his homework]] he did. d *1 wanted John to go to bed before he finished his homework. and [,p go to bed] he did before he finished his homework Adapted from (Takami 1985. 273-274) The example in (16a) shows that before he finished his homework is inside a focused vP. which is assumed to be a constituent. and (1 6b) indicates that before he finished his homework cannot be outside vP. This suggests that before he finished his homework must be inside vP. If a fronted vP is a constituent, the contrast in acceptability between (1 6c) and (1 6d) also suggests that before he finished his homewori:r uwaSQ-O / """ "rurnour-Acc"

� . } toiu A C

gakuseltatz-ga · ·

(u =)

, sonXKlszte-lru �

�espect-NonP"

"students-Nom" VF

v

/"-v

v

In (1 79a), the null subject e (= �) in the main clause c-commands the null object e (= a) in the embedded clause. They are non-distinct in terms of Case features because their Case features are uninterpretable (i.e. deleted). Thus, the matrix null subject has priority over the embedded null object for association with the postverbal NP. 126 Therefore, (1 79a) is unacceptable. 126 My proposed analysis allows (1 79a) to have the reacting thatMr. Yamada knows the rumour that the students respect someone. This interpretation, however, is

Japanese Postverbal Contructions

197

Likewise, in (I 79b-e), each of the postulated matrix null subjects prevents the postverbal NP from being associated with the embedded null object, so the examples are difficult to comprehend. It should be noted that ( l 79a-b) can be accounted for in a different manner: with respect to Case features, in each example, the postverbal element is non-distinct from the complex NP that contains the null object, and thus the postverbal NP is blocked by the complex NP from being associated with the null object (see ((1 77)) 127 Furthermore, the above argument also applies to examples where overt matrix subjects appear in the initial position of a sentence. (1 80) a.

&Taro-ga [NF[ep e, Yamada sensei-o

sonkeisite-iru toiu]

Taro-Nom

Yamada teacher-Acc respect-NonP Comp uwasa]-0 sitte-iru yo, gakuseitatir-ga. rumour-Acc know-NonP FP students-Nom Lit "Taro knows the rumour that they, respect Mr. Yamada, the studentsi. " b. &Taro-ga [NF[ep Op,[lP e, t, sonkeisite-iru]] senseio]-o sitte-iru Taro-Nom respect-NonP teacher-Acc know-NonP yo, gakuseitatir-ga. FP students-Nom Lit "Taro knows the teacher who they, respect, the students." c. &Taro-ga [cp e) Yamada sensei-o sonkeisite-iru kotoJ-o Taro-Nom Yamada teacher-Acc respect-NonP Comp-Acc sitte-iru yo, gakuseitatirga. know-NonP FP students-Nom Lit "Taro knows that they, respect Mr. Yamada, the student... " d &Taro-ga [ep e, Yamada sensei-o sonkeisite-iru kadooka]-o Taro-Nom Yamada teacher-Ace respect-NonP whether-Ace sitte-iru yo, gakuseitatirga. know-NonP FP students-Nom Lit "Taro knows whether they, respect Mr. Yamada, the studentsi. " In (1 80), on encountering the embedded verbs sonkeisite-iru "respect," the parser misanalyses Taro-ga as an external argument of the verb. In other words, Taro in each example is construed as the subject corresponding to the embedded null subject in the example in (I 78a). The matrix null impossible. I will not enter into details here. See footnotes 75 and 9 1 . 127 Two of three informants judge the examples in (1 79c--d) to be better than those in (1 79a-b) , although all the examples sound uuuatural to them

198

Chapter Four

subjects postulated by the parser hence have priority over the true null subjects for association with the postverbal NPsYs

Type III: � neither containing nor c-commanding a Let us now consider (1 76c) (i.e. the case where � neither contains nor c-commands o) Observe (18 1), where postverbal NPs have accusative Case, matrix subjects are complex NPs containing null objects, and matrix objects appear in the initial position of sentences by undergoing the operation of scrambling. ( 1 8 1 ) a.

& Minna-o [NP[cpgakuseitati-ga e, sonkeisite-im toiu] uwasa]-ga everyone-Ace students-Nom respect-NonP Comp rum our-Nom Yamada senseit-o. surprise-Past FP Lit "The rumour that the students respect him, surprised everyone,

odorokase-ta yo,

Yamada teacher-Ace Mr. Yamadat"

b.

&Minna-o [NP[cpOpo[lP t, e, sonkeisite-iruII gakuseitati-ga everyone-Ace

respect-NonP

students-Nom

nayamase-ta yo, Yamada senseiro.

Yamada teacher-Ace

annoy-Past FP Lit. "The students who respect

him! annoyed everyone,

Mr.

Yamadat." In (1 8la), when the embedded verb sonkeisite-iru "respect" is encountered, the parser incorrectly analyses minna-o "everyone-Ace" and gakuseitati­ ga "students-Nom" as arguments of the embedded clause verb. The parse tree at this point thus contains no null arglll11 ents. Minna-o "everyone­ Ace" should also be construed as a scrambled element. 129 On reaching toiu "Comp," the parser amends the main clause analysis such that the clause can be assigned a theta-role, and thereby the clause is kept in store until a theta-role assigner appears. \Vhen encountered, the theta-role assigner uwasa-ga "rurnour-Nom" is merged to the stored clause, and assigns the clause a theta-role, and the complex NP is created. However, the complex NP has no theta-role at this stage, and hence it is stored. When reaching a matrix verb, the parser postulates a null object as an argument of the 128 If Taro were reanalysed as a matrix subject tlrrough reattachment, this reanalysis would be high-cost because of a violation of the Unconscious Reanalysis Condition in (173). I will later consider some examples that seem to be llllacceptable due to syntactic reanalyses. 1 Z9 See footnote 1 2 1 .

Japanese Postverbal Contructions

199

matrix verb and subsequently integrates both the null object and the complex NP to the matrix verb so that both of them can be assigned theta­ roles and Case-checked. As soon as the postverbal NP is attached to a root CP, the LC in (1 27) attempts to apply in order to guarantee that the postverbal NP is licensed. The parse tree at this point is illustrated in ( 1 8 l a).

c� seneiro CP

( 1 8 1 ) a'

�;�'.

'�"',,�A« "

uwa n; v0' �o�a

�

urprise-past"

(� =) !\ ;(5 minnarO

u A "everyone-Ac,;;/' "

),�o�ru

.A gaku ei�a "stude ts-Nom " C. �espect-NonP" T

(u =)

v A

In (1 8 l a�, the postverbal NP Yamada sensei-o "Yamada teacher-Acc" fails to be associated with the embedded object t, (=a), which is incorrectly analysed as the trace of the scrambled object minna-o "everyone-Acc." Furthermore, the null obj ect of the matrix verb is closer to the postverbal NP than any other element non-distinct from it The matrix null object hence takes precedence over such elements for association with the

200

Chapter Four

postverbal NP. 130 The alternative analysis would reattach minna-o to the matrix TF as a scrambled element (see also footnote 128). This reanalysis, 1 30 If the nominative subject (i.e. uwasa-ga "rurnour-Noffi'') moved to the specifier ofT, the null object ofthe matrix verb would be prevented by minna-o "everyone­ Ace" from being associated with the postverbal NP, because the null object has a longer path to the IX'stverbal NP than minna-o does, as shovm below.

�

(i)

T

ada senei-o {A't'Yamada teacher-Ace"

A � (\ /\ toiu

((1=) minna;-o T "everyone-A gakuseitatis-ga "students-Nom" vP

tk

!\'surprise-Past"

VP

v' oaorokase-ta v

r---.. ts/\ sonkeisite-iru (X,, "respect-NonP"

)\

(a=) t;

v

V

At first sight, there seem to be no empirical differences between the assmnption immediately above and my proposed analysis. However, this assumption is lllltenable. Let US consider the example in Cii) where the postverbal NP is expected to be associated with the null object ofthe matrix verb. (ii) [NP[ Cp minna-o gakuseitati-ga sonkeisite-irn toiu] uwasa]-ga ei everyone-Acc students-Nom respect-NonP Comp nnnour-Nom odorokase-ta yo, Yamaada senseii-o. surprise-Past FP Yamada teacher-Ace Lit. "The nnnour that the students respect everyone surprised him!, Mr. Yamadal." In (ii), the postverbal NP is easy to associate with the matrix null object. This suggests that the complex NP subject remains in the specifier of vP, and hence, that

Japanese Postverbal Contructions

201

however, is costly (see (1 73)). The postverbal NF in the above example is hence difficult to associate with the null object within the complex NF lJ ! In ( l 8 1b), the relative clause contains a null argument that is expected to be associated with the postverbal NF . On encountering the embedded verb sonkeisite-iru "respect," the parser analyses minna-o "everyone-Acc" as an internal argument of the verb, and thereby the embedded clause is assumed not to contain a null object. When encountering the matrix verb, the parser postulates a null argument as a matrix object to maximally satisfy syntactic principles. The parse tree at this point would thus be as follows. ( 1 8 1 ) b'[NP[ cP()P,[n> t, minna-o sonkeisite-iru]] gakuseitati]-ga everyone-Acc respect-NonP students-Nom

e

nayamase-ta. annoy-Past Lit "The students who respect everyone annoyed e." When the postverbal NF Yamada sensei-o "Yamada teacher-Ace" IS encountered, it is attached to the root CP. The null object in the matrix clause is closer to the postverbal NF than the embedded object is, and the null matrix object is closer to the postverbal NP than minna-o is (see also footnote 62). 1 3 1 "When null objects appear in matrix clauses as well as in embedded clauses, if postverbal elements can be construed as arguments sharing properties with the null matrix objects, they can be associated with the null embedded objects as well. (i) a.[NP[ Cp gakuseitati-ga el sonkeisite-iru tOiu]uwasa]-ga ei odorokase-ta yo, students-Nom respect-NonP Comp rurnour-Nom surprise-Past FP Yamada senseii-o. Yamada teacher-Ace

Lit. "The rurnour that the students respect ei surprised e i, Mr. Yamadai." b.[NP[crOpo[Tp to e! sonkeisite-iru]] gakuseitati-ga el nayamase-ta yo, Yamada respect-NonP students-Nom armoy-Past FP Yamada senseiro. teacher-Ace

Lit. '''The students who respect ei armoyed ei, Mr. Yamadai." In (ia), Yamada sensei-o "Yamada teacher-Acc" can be associated with the null embedded object. The same is true of (ib). My proposed analysis, however, incorrectly predicts examples like (i) to be unacceptable because the null matrix objects would block the association of the postverbal elements with the null embedded objects in the same manner as in (181). I leave this problem open for future research.

202

Chapter Four

hence the postverbal NP is associated with the postulated null object Alternatively, if minna-o was reanalysed as a matrix object in such a way that a null object may be postulated in the embedded clause. this type of reattachment would be costly because the final attachment site is not sufficiently close to the original attachment site according to the Unconscious Reanalysis Condition in (173). Therefore, the postverbal NP may not be associated with the null object in the embedded clause without conscious efforts. I will now turn to further examples in which incorrect syntactic­ analysis leads to the wrong association. (1 82) a. ?*Taro-ga [NP[CP gakuseitati-ga e, sonkeisite-iru toiu] uwasa]-0 Taro-Nom students-Nom respect-NonP Comp rumour-Acc sitte-iru yo, Yamada senseirO.132 know-NonP FP Lit "Taro knows the rumour that the students respect him"

Yamada teacher-Ace

Mr.

Yamada." b. ?*Taro-ga [NP[cpOpo[ TP to e, sonkeisilR-iru]] gakuseitati-o sitte-iru Taro-Nom respect-NonP students-Acc know-NonP yo, Yamada senseiro. FP Lit. "Taro knows the students who respect him), c. ?*Taro-ga [cp gakuseitati-ga el sonkeisite-iru kola]-0 sitte-irn Taro-Nom students-Nom respect-NonP Comp-Acc know-NonP yo, Yamada senseiro. FP Lit "Taro knows that the students respect him"

Yamada teacher-Ace

Mr. Yamadat."

Yamada teacher-Ace

Mr. Yamada."

d.?*Taro-ga [cp gakuseitati-ga e, sonkeisite-iru kadooka]-o Taro-Nom students-Nom respect-NonP whether-Acc sitte-iru yo, Yamada senseiro. know-NonP FP Lit "Taro knows whether the students respect him"

Yamada teacher-Ace Yamada."

Mr.

1 32 The example in (i) is unacceptable probably because the complex NP containing a null argrunent has the same Case as the postverbal NP. (i) ?*Taro-wa [NP[ Cp gakuseitati-ga e{ sonkeisite-iru toiu] uwasa]-o sitte-iru yo, Taro-Top students-Nom respect-NonP Comp nnnour-Acc know-NonP FP Yamada senseij-o. Yamada teacher-Ace

Lit. "Taro knows the rurnour that the students respect himi. Mr. Yamada."

Japanese Postverbal Contructions

203

In each example in (182), whether the matrix object is a complex NP or a clause, Taro-ga "Taro-Nom" is incorrectly analysed as an element in the embedded clause. IJ3 In other words, Taro-ga is construed as an argument of sonkeisite-irn "respect." Thus, there are no appropriate elements with which the postverbal NP can be associated. An alternative analysis would be to say that Taro-ga should be reattached to the matrix TP as the subject Yet, this syntactic reanalysis would be costly according to the Unconscious Reanalysis Condition in (173). Hence, the postverbal NPs in the above examples are difficult to associate with the null objects in the embedded clauses. The above claim may be supported by the acceptability of the following example, where incorrect analyses are avoided with the use of an overt object instead of a covert one. (1 82) c'Taro-ga [oP gakuseitati-ga ano hito,-o sonkeisite-iru koto]-0 Taro-Nom students-Nom that person-Acc respect-NonP Comp-Acc sitte-iru yo, Yamada senseiro. know-NonP FP Lit "T aro knows that the students respect that personi,

Yamada teacher-Ace

Mr, Yamada,."

In (182c), when the embedded verb is encountered, gakuseitati-ga "students-Nom" and ana hito-o "that person-Acc" are correctly analysed as arguments of the embedded verb; at the same time, Taro-ga is kept in store until the matrix verb appears. 134 "When sitte-iru "know" is encountered, Taro-ga is analysed as the matrix subject without syntactic reanalyses. The phrase that contains the overt object ano hito-o "that person-Acc" is a clause (i.e. CP), and hence no elements prevent the postverbal NP from being linked to ana hito. The contrast in acceptability between (1 82c) and (182c) therefore indicates that (1 82c) involves a wrong analysis, but (182c) does not It is important to note that the example in (1 82a) remains unacceptable even if an overt object is inserted in the embedded clause like in (1 82c). (1 82) a'?*Taro-ga [NF[cP gakuseitati-ga ana hito,-o sonkeisite-iru toiu] Taro-Nom students-Nom that person-Acc respect-NonP Comp uwasa]-0 sitte-iru yo, Yamada senseiro. rumour-Acc know-NonP FP Lit "Taro knows the rumour that the students respect that

Yamada teacher-Ace personi Mr Yamadai." ,

.

1 33 Taro-ga "Taro-Nom" could be construed as the NP marked with the exhaustive­ listing ga (see Kuno 1 973a, 38). 1 14 Cf. (l86b).

204

Chapter Four

In (l 82a), Taro-ga would not be analysed as an element of the embedded clause. However, a matrix object containing the embedded object under discussion blocks the association of the postverbal NP with ano hito "that fellow", so the postverbal NP is difficult to inteIpret as coreferential with it I will next move on to a couple of examples without syntactic misanalyses, where matrix objects precede matrix subjects, as in (183) and matrix subjects precede matrix objects, as in (I 84). l3 l (1 83) a.?*[NP[cP e,

Yamada sensei-o

sonkeisite-iru toiu] uwasa]-o

Yamada teacher-Acc respect-NonP Comp rumour-Acc Taro-ga sitte-iru yo, gakuseitatirga. Taro-Nom know-NonP FP Lit "Taro knows the rumour that they, respect Mr. Yamada,

students-Nom

the

studentsi. " b.?*[NP[cPOpo[11' e, to sonkeisite-irulJ senseio]-o Taro-ga respect-NonP

sitte-irn

teacher-Ace Taro-Nom

yo, gakuseitatirga.

students-Nom

know-NonP FP Lit "Taro knows the teacher who they, respect, (1 84) a.

the students."

?*[NP[cpgakuseitati-ga e, sonkeisite-iru toiu] uwasa]-ga students-Nom

respect-NonP Comp rum our-Nom Yamada senseiro. everyone-Ace surprise-Past FP Lit "The rumour that the students respect him, sUIprised everyone,

minna-o

odorokase-ta yo,

Yamada teacher-Ace

Mr. Yamadat."

b.

?*[cp gakuseitati-ga e, sonkeisite-iru koto]-ga students-Nom

odorokase-ta yo,

minna-o

respect-NonP Comp-Nom everyone-Acc Yamada senseiro.

Yamada teacher-Ace

sUIprise-Past FP Lit "That the students respect

him,

surprised everyone,

Mr.

Yamadai." c.

?*[cpgakuseitati-ga el sonkeisite-iru kadookaJ-ga minna-o students-Nom

nayamase-ta yo,

respect-NonP whether-Nom everyone-Acc Yamada senseiro.

Yamada teacher-Ace

annoy-Past FP Lit. "Whether the students respect him! annoyed everyone,

Mr.

Yamadat." m At first sight, it seems that the examples in (183) are similar to those in (177c-d), but they are different In (183), the complex NPs and postverbal elements are distinct in terms of Case features, but in (1 77c-d) they are non-distinct.

Japanese Postverbal Contructions

205

When the matrix verb sitte-iru "know" in (I 83a) is encountered, as ( l 83a) shows, the complex NP (i.e. the matrix object) is analysed as a scrambled element at the same time as the matrix subject and the trace of the complex NP have theta-roles assigned and their Case features checked.

Taro-J;a

� �

(1 83) a'

f ')

J\

o

A��� )\\ '"

/'Jf\

(u =)

eitatil-ga

"students-Nom"

�

C (� =)Taro-ga.

toiu

T

A v'

,1\

,Sltte-iru 'know-NonP"

m

Taro-ga

The matrix subject neither c-commands nor contains the null subject e, (= a) , which is postulated within the matrix obj ect However, e, has a longer path to the postverbal NP than does. Thus, blocks the association of e, with the postverbal NP. It is hence difficult to construe the postverbal NP as the subject within the complex NP. The same is true of the example in (I 83b), where the complex NP containing the relative clause not only has an accusative Case feature, but it also contains a null subject within it. 136 The example in (1 84a) has a partial parse tree as follows.

Taro-ga

Taro-ga

1 36 The llllacceptability of (i) would be accOlmted for in the same way as in (1 83a). (i) ?*[NP[CP e! Yamada sensei-o sonkeisite-irn toiu] uwasa]-o Taro-wa Yamada teacher-Acc respect-NonP Comp rurnour-Acc Taro-Top sitte-irn yo, gakuseitatij-ga. know-NonP FP students-Nom Lit. "'Taro knows the rurnour that they! respect Mr. Yamada, the studentsi."

206

Chapter Four

(1 84) a'

CP

� !"'-. "Yamada

ada seneiro

C

�v' en" 0 /"-. toiu (�)

C

teacher-Ace"

vP

�

uwas

0

orokase-ta

urprise-past"

minna-o V

T

� ' son� � vP

T

gakuSei

"students-Nom

espect-NonP"

AV

(a =) e,

As (1 84a) shows, the relevant null object e, (=a) is farther from the postverbal NP than minna-o (= �) is. As a result, the postverbal NP is blocked by minna-o from being associated with the null object e, (=a). The same is true of the examples in (1 84b-c).

4.5.3.2 The absence of island effects In this subsection, I will discuss acceptable examples where postverbal elements can be associated with null arguments that are contained in such embedded clauses as complement clauses and relative clauses. These examples are grouped into the three types listed below. Type

I :

Type

IT :

Type

ill :

Phrases containing null arguments are different from phrases with respect to categorial features. Phrases containing null arguments are different from phrase with respect to Case features. Phrases containing null arguments are different from phrases with respect to both categorial features features.

postverbal postverbal postverbal and Case

Japanese Postverbal Contructions

207

These three types will be presented in order

Type I: Different categorialfeatures I will first consider Type I, in which phrases containing null arguments are different from postverbal phrases with respect to categorial features. Type I is subdivided into two patterns, as shown below. (1 85) a. [ep e, . . . J-Nom . . . NPi-Nom ei can be associated with a postverbal nominative Case-marked NP. b. [NP e, . . . J-Nom . . . CPi-Nom ei can be associated with a postverbal nominative Case-marked

CP.

Observe the examples in (1 86), which belong to pattern (1 85a) (cf (177a)). (1 86) a. [cp e, Yamada sensei-o sonkeisite-iru kotoJ-ga hontoo dat-ta Yamada teacher-Acc respect-NonP Camp-Nom true be-Past

yo, gakuseitath-ga.

students,-Nom

FP Lit "That they, respect Mr. Yamada was true, b. [epe, Yamada sensei-o sonkeisite-iru kadookaJ-ga zyuuyoo Yamada teacher-Acc respect-NonP whether-Nom important desu yo, gakuseitalirga. be-NonP FP Lit "Whether they, respect Mr. Yamada was important,

the students;."

students-Nom studentsi. "

the

In (1 86a), the nominative Case-marked NP gakuseitat-ga "students-Nom" appears in postverbal position. It is different in terms of categorial features from the clause [ep Yamada sensei-o sonkeisite-iru kotoJ-ga "[that e respect Yamada teacherJ-Nom," which contains a null argument That is, the clause is not similar to the null argument in the sense of (175). Thus, the clause does not prevent the postverbal NP from being associated with the null argument, and hence (1 86a) is acceptable. The same is true of (1 86b). Let us next tum to pattern (1 85b). (1 87) [NP [ep

e, Yamada sensei-o

nayamasase-ta toiuJ uwasaJ-ga

Yamada teacher-Ace armoy-Past Comp rurnour-Nom hontoo dat-ta yo, [epgakuseilali-ga benkyoo si nai koloJ-ga. true be-Past FP

students-Nom study do not Comp-Nom

208

Chapter Four

Lit. "The rumour that iti annoyed Mr. Yamada was true, ] ."

[that the

students do not study i

In (1 87), the clause [gakuseitati-ga benkyoo si nai koto]-ga "[that the students do not study]-Nom" appears postverbally. With respect to categorial features, the clause is different from the complex [NF e, Yamada sensei-o nayamasase-ta toiu uwasa]-ga "[the rurnour that it annoyed Mr. Yamada]-Nom," which contains a null argument. As in the case of (1 86), nothing blocks the postverbal element from being associated with the null argument. Therefore, (1 87) is acceptable.

NP

Type II: Different Casefeatures I will next consider Type IT, where phrases containing null arguments are different from postverbal phrases with respect to Case features. Type IT is further classified into two patterns, as shown below.

]-Nom . . . NPi-Acc ]-Nom i Ace

(1 88) a. [NF . . . e, . . . ej can be associated with a postverbal accusative Case-marked NP. b. [ep . . . e, . . . . . . CP ej can be associated with a postverbal accusative Case-marked CPo Now let us look at the examples in (1 89), which belong to pattern (1 88a). (1 89) a. [NF[ep gakuseitati-ga e, sonkeisite-iru toiu] uwasa]-ga hontoo students-Nom respect-NonP Camp rumour-Nom true dat-fa yo, Yamada senseiro. be-Past FP Lit. "The rumour that the students respect him, was true,

Yamada teacher-Ace Yamadat."

Mr,

b. [NF[ Cp Op, ['" e, to sonkeisite-iru]] gakuseitatio-ga huete-imasu yo, respect-NonP students-Nom increase-NonP FP Yamada senseiro.

Yamada teacher-Ace Yamadat."

Lit. "The number of students who respect him! is increasing,

NP

Mr.

In (1 89a), the accusative Case-marked Yamada sensei-o "Yamada teacher-Ace" appears in postverbal position. It is different in teITI1S of Case features from the complex [ep gakuseitati-ga sonkeisite-iru toiuu

NP

Japanese Postverbal Contructions

209

uwasaJ-ga "[the rumour that the students respect eJ-Nom," which contains a null argument In other words, the complex is not similar to the null argument in the sense of (1 75). Thus, the complex does not block the postverbal from being associated with the null argument, and hence (1 89a) is acceptable. The same account can be given for (1 89b). 137 The following example belongs to pattern (188b).

NP

NP

NP

(1 90) [cp gakuseitati-ga e, sinzite-iru kataJ-ga hantaa dat-ta ya, students-Nom believe-NonP Comp-Nom true be-Past FP [cp Yamada sensei-ga yameru kalaJ-a. Yamada teacher-Nom resign Comp-Acc Lit "That the students believe it, was true, [t at Mr. Yamada will resign]i."

h

In (1 90), the clause [Yamada sensei-ga yameru kataJ-a "[that Yamada teacher will resign]-Acc" appears postverbally. With respect to Case features, the clause is different from the clause [gakuseitati-ga sinzite-iru kataJ-ga "[the students believe eJ-Nom," which contains a null argument As in the case of (1 89), nothing blocks the postverbal element from being associated with the null argument Thus, (190) is acceptable.

Type III: Different categarial and Case features Now let us turn to Type patterns.

I1I.

This type is also subdivided into two

(191) a. [ep . . . e, . . . J - o m . . . NPI-Acc ej can be associated with a postverbal accusative Case-marked

N

NP.

\J7

The example in (i) is less acceptable than those in (189) even though the postverbal phrase is different from the complex NP that contains a null argument in terms of Case features. (i) ?*John-ga [NP[ cw\1ary-ga el age-fa] hon]-0 1UIsunda yo, Billrni. 101m-Nom Mary-Non gave book-Ace stole FP Bill-Dat Lit. "John stole a book that Mary gave to himi, to Bilk" The reason why (i) is llllacceptable may be that an NP marked with the dative particle ni is likely to be analysed as a locative PP, and hence, that Bill-ni "BiU­ Dat" is interpreted as a potential modifier of the matrix predicate. For a discussion of the dative particle ni in Japanese, see e.g. Miyagawa and Tsujioka (2004), Sadakane and Koizurni (1 995), and Sakamoto and Walenski (1998, 1 04-105).

210

Chapter Four

]-Nom . . . CP,-Acc

b. [NP . . . e, . . . ej can be associated with a postverbal accusative Case-marked

CP.

Observe the following examples: (1 92) a. [cp gakuseitati-ga e, sonkeisite-iru koto]-ga hontoo dat-ta yo, students-Nom respect-NonP Comp-Nom true be-Past FP Yamada senseiro.

Yamada teacher-Ace Mr, Yamada..

Lit "That the students respect him, was true, " b. [ep gakuseitati-ga e, sonkeisite-iru kadooka]-ga zyuuyoo desu students-Nom respect-NonP whether-Nom important be-NonP yo, Yamada senseiro. FP Lit "Whether the students respect him, is important,

Yamada teacher-Ace Yamadai."

Mr,

NP

In (1 92a), the accusative Case-marked Yamada sensei-o "Yamada teacher-Acc" appears in postverbal position. The postverbal is different from the clause [gakuseitati-ga e, sonkeisite-iru koto]-ga "[the students respect e Comp]-Nom," which contains a null argument with respect not only to categorial features but also to Case features. Hence, the clause is not similar to the null argument in the sense of (175), resulting in failure to block the association of the postverbal with the null argument Thus, (1 92a) is acceptable. The same is true of (1 92b). The example in (193) belongs to pattern (19Ib).

NP

NP

(1 93) [NP[Cp gakuseitati-ga e, sinzite-iru toiu] uwasa]-ga hontoo dat-ta students-Nom believe-NonP Comp rumour-Nom true be-Past yo, [cp Yamada sensei-go yameru kala]-a. FP Lit "The rumour that the students believe it, was true,

Yamada teacher-Nom resign Comp-Acc Yamada will resign],."

[that Mr,

In the above example, the clause [Yamada sensei-ga yameru koto]-0 "[Yamada teacher-Nom resign Comp]-Acc" appears in postverbal position. With respect to categorial and Case features, the clause is different from the complex [gakuseitati-ga e, sinzite-iru toiu uwasa]-ga "[the rumour that the students believe e ]-Nom," which contains a null argument As in the case of (1 92), nothing blocks the postverbal element from being associated with the null argument, resulting in the acceptability of (1 93).

NP

Japanese Postverbal Contructions

211

All the examples observed above contain nothing corresponding to �, a potential intervener, in (1 74). Postverbal elements can thus be associated with null arguments within embedded clauses. This indicates that the island effects observed in JPVCs are irrelevant to syntactic constraints on movement. Finally, I will mention a problematic case where copulative sentences involve complex predicative NPs. (1 94) a. Kore-wo [NP[cP OPI e, tl oikoke-to11 neko do yo, ana inurKU . this-Top chase-Past cat is FP dog-Nom Lit. "This is a cat which it, chased, dog!." b. Kore-wo [NP[CP OPI tl e, oikoke-to 11 neko do yo, ana inuro. this-Top chase-Past cat is FP dog-Ace Lit. "This is a cat which chased it" dog,."

that

that

that

that

In (1 94), the NPs marked with the topic particle wo (i.e. kore-wo "this­ Top") are non-distinct from the postverbal NPs in terms of categorial features and Case features, c-commanding null arguments that are contained in the complex predicative NPs [NPe oikoketa neko] "a cat which it chased" or "a cat which chased it." Therefore, kore-wa would prevent the postverbal elements from being associated with the null arguments (i.e. e). Furthermore, even if the NPs marked with the topic particle wa are not c-commanded by the postverbal elements, the complex predicative NPs containing the null arguments block the association of the postverbal NPs with the null arguments because the complex NPs are non-distinct from the postverbal NPs with respect to Case features. However, each postverbal element in (1 94) is easy to associate with the null argument e within the complex predicate NP. At this point, I have no clear idea of how this problem should be solved. 138 1 33 Kore-wa could be reattached to CP as schematically illustrated below. (i)

�

�

korej-wa "1hiS-TO "

CP

P

the postverba. NP

TP

AT'

Ii

yo

212

Chapter Four

4.5.3.3 Postverbal adjuncts In this subsection. I will discuss the case where adjuncts appear in postverbal position. Let us first consider examples like (1 95). which display island effects. (1 95) a &

[syusyoo-ga

kinoo

at-to

zyosei]-o

prime minister-Nom yesterday meet-Past with woman-Ace mi-tanda yo, Shinbashino-n0 ryootei-de. see-Past FP "(1) saw the woman whom the prime minister met with yesterday." (Soshi and Hagiwara 2004. 423)

Shinbashi-Gen Japanese-style restaurant at at a Japanese-style restaurant in Sinbashi

b.

& Watasi-wa [cp Taro-ga ana mise de koinu-o

kat-ta to]

I-Top

Taro-Nom that shop at puppy-Acc buyPast Comp arnot-fa yo, kinoo. think-Past FP Lit. "(1) thought [that Taro had bought a puppy at that shop].

yesterday. yesterday."

In (l 95a). after encountering the postverbal PP. the parser realises that there are no further elements, so it starts to associate the postverbal phrase with a modifiee. The matrix verb mi-ta "see-Past"" can be modified by the locative PP. and it also contains the complex which the verb at-to "meet-Past with" is contained, so the matrix verb is chosen as a rnodifiee over the embedded one. In other words. the postverbal locative PP is difficult to associate with the verb at-to within the relative clause. The same account can be given for (I 95b). The matrix verb omot-ta "think-Past"· not only contains the embedded verb kat-to "buy-Past:' but it can also be modified by kinoo "yesterday." Hence. the matrix predicate blocks the association between kinoo and kat-to. Next. let us turn to the case where. although one element asymmetrically c-cornrnands another one, the fOITI1er has no priority over the latter for association:

NP in

(1 96) Kyoozyu-ga kuruma-o kat-to yo, yuumei-na. professor-Nom car-Acc buy-Past FP

well-known

Iftopicalised elements occupy the specifier position ofTP, the reattaclunent to CP is low-cost according to the Reanalysis Condition. In the structure in (i), kore-wa is no longer a potential intervener because it is not c-conunanded by the postverbal NP.

Japanese Postverbal Contructions

Lit. "A professor bought a car,

213

well-known."

The example in (1 96) has two possible readings: the postverbal adjective

yuumei-na "well-known" may modify kyoozyu-ga "professor-Nom" or kuruma-o "car-Acc." This ambiguity can be derived from the UREC in (1 74). That is, although a subject c-commands an object, the subject does not block the association between the object and the postverbal phrase because the subject is contained in every phase (i.e. vP) that contains the object (see (1 76b)) (note that kyoozyu-ga occupies the specifier position of vP). Hence, yuumei-na "well-known" may be associated with both arguments without conscious effort. This account is further supported by the following unambiguous example. (196) Kuruma,-o kyoozyu-ga I, kal-Ia yo, yuumei-na. car-Acc Professor-Nom buy-Past FP Lit. "A cari, a professor bought I"

well-known well-known."

In (196), the object kuruma-o "car-Acc" is moved to the specifier position of TP by scrambling. The scrambled c-commands kyoozyu-ga "professor-Nom" and is not contained in every phase that contains kyoozyu-ga. Hence, kuruma-o has priority over kyoozyu-ga for association with the postverbal element yuumei-na "well-known," resulting in the absence of ambiguity. Finally, let us consider the example in (1 97) in which the postverbal relative clause can be associated with an element in the embedded clause.

NP

(1 97) [cp Taro-ga kuruma-o damatte unlensi-Ia (I-Top) Taro-Nom car-Acc without permission drive-Past kOlo]-o sitte-iru yo, [Hanako-ga kinoo kat-ta). Camp-Acc know-NonP FP "I know that Taro drove the car without peITI11SSlOn

Hanao-Nom yesterday buy-Past Hanako bought yesterday

There are no potential modifiees between the postverbal relative clause and kuruma "car" in the embedded clause, so the postverbal relative clause can be associated with kuruma.

4.5.3.4 Interim summary The examples discussed in the previous subsections (i.e. 4.5.3. 1 ; 4.5.3.2; 4.5.3.3) are summarised in table 4-2, where � is an intervening

214

Chapter Four

element similar to both UREC in (1 74).

a

and a postverbal phrase

in the sense

of the

Table 4-2 The presence and absence of island effects in JPVCs

examples

(dis)similarity pattern (1 77a-b) [NP e! . · .]-Nom . . . NPI-Nom

relation between a and � � contains a

island effect

(1 77c-d) [NP e! . · .]-Acc . . . NPI-Acc

� contains a

Yes

Yes

( l 77e)

[cpe! . . . ]-Nom . . . CPI-Nom

� contains a

Yes

(1 77f)

[cp ei · . ]-Acc . . . CPI-Acc

� contains a

Yes

(1 78a-b) [NP e) . . ]-Acc . . . NPI-Nom

� c-cornrnands a

Yes

(1 78c-d) [cpe! . · .]-Acc . . . NPI-Nom

� c-cornrnands a

Yes

[cpe! . · .]-Acc . . .CPI-Nom

� c-cornrnands a

Yes

(1 79a-b) [NP ej. · .]-Acc . . . NPI-Acc

� c-commands a

Yes

(1 79c-d) [cp ej . . . ]-Acc . . . NPI-Acc

� c-cornrnands a

Yes

[cp ej . . . ]-Acc . . . CPI-Acc

� c-commands a

Yes

� c-cornrnands a

Yes

(1 SOc-d) SUBJ [cp ei . · .]-Acc. . .NPI-Nom � c-cornrnands a

Yes

( l S l a-b) OBJ [NP e,. · .]-Nom . . . NPI-Acc

Neither

Yes

( l S2a-b) SUBJ [NP e, . · .]-Acc . . . NPI-Acc

Neither

Yes

( l S2c-d) SUBJ [cp e, . · .]-Acc . . . NPI-Acc

Neither

Yes

(1 83a-b) [NP e, . · .]-Acc . . . SUBJ, NPI-Nom

Neither

( l 78e)

(1 7ge)

( l SOa-b) SUBJ [NP ei . · .]-Acc NPI-Nom

Yes (Continued)

Japanese Postverbal Contructions

215

Table 4-2 Continued examples

(dis)similarity pattern [NP e!. . ]-Nom . . OBJ,

relation island effect between a and Neither Yes

�

NP.-Acc (1 84b--l = a postverbal element

Chapter Four

228

CP

(215)

�=

�CP2

a postverbal element

CPI Con]

CP.

The RA chooses (214) over (215) so that postverbal elements can adjoin to the lowest Thus, the postverbal element keiki-o "cake-Ace" in (213a) c-cornrnands the null argument in the second clause, can be associated with the null argument, and is construed as an argument Therefore, (213a) is acceptable. As with (211), one might think of the possibility of the postverbal element undergoing reanalysis. However, this reanalysis is also impermissible under the Unconscious Reanalysis Condition in (173). As for (21 3b), the postverbal element susi-o "sushi-Ace" is associated with keiki-o "cake-Ace" in the second clause. Even though susi-o is licensed, susi and keiki are semantically different This results in the unacceptability of (21 3b) 149 Based on the arguments presented above, we can conclude that the RA should be employed by the parser

4.5.5 Multiple Postverbal Phrases Revisited In this subsection, from the point of view of language processing, I re-consider multiple postverbal elements, which were discussed in 4.3.5. 1 0. Let us first consider the example in (1 33), repeated below in (216). (216) Taro-ga omotte-imasu, kawaii to, Hanako-ol-ga. Taro-Nom think-NonP pretty Comp Hanako-Acc/Nom "T aro thinks Hanako to be pretty." Suppose that the preceding clause has a structure in (21 7a), which contains a null argument ej. Then, when encountering to "Comp," the parser postulates the null argument eJ within the adjunct CP, as in (21 7b). (217) a. [ep Taro-ga [" e, omotte-imasuII b. [ep Taro-ga [" e, omotte-imasuII [cpe, kawaii tali

1 49 Susi may be used to correct keiki. In this case, it SOlUlcis to me like Taro made sushi and Hanako ate it See ( 1 1 3a) and footnote 73.

Japanese Postverbal Contructions

229

In (21 7b), [ ep ei kawaii tali "[pretty Comp]" is associated with the null argument el in the preceding clause because of non-distinction in terms of agreement-features. Thus, r cp e, kawaii tal can be licensed. The IRs in (1 05) allow it to be construed as an argument sharing properties wifh the relevant element. \\Then Hanako]-o/-f;a is encountered, the RA requires it to adjoin to r ep ei kawaii tok The simplified parse tree is thus as follows (see footnote 90) (218)

� �

P2

CI'3

koro/-ga "Hanako-AcclNom"

In (21 8), Hanako,-o/-Ka c-commands fhe null argument e, and fhey are non-distinct in terms of agreement-features, so it is licensed. The IRs in (1 05) allow Hanoka to be construed as an argument sharing properties with the null argument The example in (216) is therefore acceptable. This shows that the final parse tree is not always identical to the structure built in syntax even when the relevant sentence is acceptable (cf ( l 34b)). Next, I return to the examples in (1 30), reproduced as (219). (219) a. e, ej tabe-masita yo, Tarot-ga, keikiro. eat-Past FP Taro-Nom cake-Ace Lit "Hei ate i�, Tarol, cake]." b. e, keiki-o tabe-masita yo, Taro-gal, gatugatuto. cake-Acc eat-Past FP Taro-Nom hungrily Lit "Hei ate cake, Tarol, hungrily." In (21 9a), when keiki-o "cake-Acc" is encountered, it is required by the RA to adjoin to Taro-go "Taro-Nom" instead of the topmost CP, so the schematic parse tree might be like (220).

Chapter Four

230

(220)

�

C I

TartJ-ga "Taro-Nom" ei ej

a

kezki-o "cake-Ace"

e-maSl a yo

However, in (220), keiki "cake" fails to be licensed because of its inability to c-command the null argument ej in the preceding clause. Suppose here that keiki undergoes reanalysis. Then, two possible parse trees will be created as illustrated in (221) (footnote 90; 3.2.3; Chomsky 2004b, 168). (221) a.

� C� Ta)o-g:J;;;ki-o

b.

� CP2

Taro-ga

If CPI in (220) is a phase, under the Unconscious Reanalysis Condition in (1 73), the parse tree in (22l a) is not permissible, but that in (22lb) is. In (221 b), keiki-o can be licensed because it c-commands the null argument eJ in the preceding clause. The example in (21 9a) is thus acceptable. The same can be said of (219b). Again, this suggests that the final parse tree is not necessarily identical to the structure built in syntax even in the case of the acceptable sentences. (Note that a structure corresponding to (218) cannot be generated by the grammar because of the Extension Condition, but the one corresponding to (22lb) can.)

4.5.6 Conclusion In the present section, I have argued that the presence or absence of island effects in JPVCs follows from the interaction of syntactic principles with the parsing strategies that I have laid out I have also argued that the proposed parsing strategies can deal with cases such as the preferred reading of scopally ambiguous JPVCs. Furthermore, I have claimed that the parser should employ a parsing strategy proposed in Kimball (1 973), namely the right association principle (RA). Finally, I have discussed the

Japanese Postverbal Contructions

231

case in which the final parse tree is not identical to the structure built in syntax when the relevant sentence is acceptable. 4.6 Linear Distance Effects

In this section, I will discuss the results of an experiment to test the effect of linear distance on the acceptability of JPVCs. I will examine whether there is a parsing strategy relating to linear distance. If such a parsing strategy exists, it will predict that linear distance will affect the acceptability of JPVCs. The results of the experiment indicate that the acceptability of JPVCs decreases as the linear distance between postverbal elements and relevant gaps/modifiees becomes longer I will illustrate that the linear distance effects observed in JPVCs may be justified by the parsing strategy proposed by Hawkins (2004), who argues that the parser should minimise a processing domain.

4.6. 1 Background In the previous section, I discussed instances in which island effects in the JPVC may follow from parsing strategies such as the UREC in (1 74) and the RA in (209), rather than the violation of syntactic constraints such as the Complex Constraint However, some island effects in JPVCs cannot be explained by the parsing strategies I have adopted thus far, which are defined in terms that do not refer to linear distance. Let us compare the following sentences.

NP

no]-wo zizitu do yo, kinkor-o. someone-Nom open-Past Comp-Top fact Copula FP safe-Ace Lit "[That someone opened e,] is certain, the safe•. " b.?? [ep [dare-ka-go e, ake-ta] no]-wo zetloi-ni ugokanu someone-Nom open-Past Comp-Top absolutely undeniable zizitu da yo, kinkoro. fact Copula FP safe-Ace Lit "[That someone opened e,] is absolutely an undeniable fact, the safe.. " Adapted from (Simon 1989, 173_ 1 74)l lO

(222) a. [cp [dare-ko go ei oke-to]

1 50 Simon (1989) provides the following acceptable example as well. (i) 7 [cp [dare-ka-ga ej ake-ta] no]-wa ugokarm zizitu da yo, kinkoj-o. someone-Nom open-Past Comp-Top lUldeniable fact Copula FP safe-Ace Lit. "[That someone opened el] is an lUldeniable certain, the safei."

232

Chapter Four

The contrast in acceptability between (222a) and (222b) cannot be explained by the parsing strategies I have adopted thus far In (222b), there are no interveners corresponding to � such as DREC. Hence, it is incorrectly predicted that the postverbal phrase in (222b) is as easy to associate with the relevant null argument as in (222a). Based on the contrast between the examples in (222), Simon (1989, 173) proposes that "the shorter the syntactic distance (in terms of the number of nodes that intervene linearly) between the embedded clause and the end of the sentences, the easier it is to associate the postverbal element with that clause." In other words, there are instances where linear distance, not hierarchical distance, is related to parsing difficulty. Thus, in order to examine whether there is another parsing strategy for dealing with linear distance effects, I conducted an experiment designed to test the effect of the intervening elements' length on the acceptability of JPVCs. 4,6,2 Experiment Objective

The experiment was designed to test the linear distance effect on the acceptability of JPVCs where there is an adjunct clause intervening between a postverbal phrase and its relevant gap/modifiee (i.e. to examine whether the JPVC becomes less acceptable as the linear distance between the relevant elements is longer). If JPVCs may or may not contain adjunct clauses intervening between postverbal elements and relevant gaps/modifiees and are predicted to be acceptable by the parsing strategies I have adopted thus far, then the predictions are as follows: (223) a. U there is a parsing strategy relating to linear distance (i.e. defined in terms which refer to linear distance), JPVCs with intervening adjunct clauses will be less acceptable than those without b. If there are no parsing strategies relating to linear distance, JPVCs with intervening adjunct clauses will be as acceptable as those without Participants

Twenty-seven native speakers of Japanese participated in the experiment All were students at Sophia University in Tokyo, Japan.

Japanese Postverbal Contructions

233

Twenty-five students were undergraduates enrolled m an introductory linguistics course, while the rest were postgraduates studying linguistics. They were not informed about the purpose of the experiment The undergraduates were presented with questionnaires a classroom at the university which were collected after completion, while each postgraduate received the questionnaire in MS Word fOffilat and was asked to return it to the experimenter via e-mail. No payment was made.

in

Moteriol ond design

� elements

fewer words (� short distance)

more words (� long distance)

(224a) [SUBJ OBJ VP]

(224b) [SUBJ OBJ VP]

koto-go VP

koto-go

type

non-JPVC

[Intervening CP] VP

postverbal elements

SUBJECT

(224c) [e OBJ VP] koto-go VP, SUBJ

(224d) [e OBJ VP] koto-go [Intervening CP] VP SUBJ

OBJECT

(224e) [SUBJ e VP] koto(224g) [SUBJ OBJ VP]

(2241) [SUBJ e VP] kotogo [Intervening CP1VP OBJ (224h) [SUBJ OBJ VP]

koto-go VP,

koto-go

go VP, OBJ

ADJUNCT

ADJ

[Intervening CP] VP

ADJ

Figure 4-1: Intervening Elements and Types

The experiment employed a token set of eight conditions in 2x3x2 factorial design, as illustrated in Figure 4-1, where one of the three intersection factors-the postverbal element type--contained three levels (subject, obj ect, and adjunct), while the others had two levels each (short vs. long distance between the matrix subject and matrix predicate; non­ JPVCs vs. JPVCs).

Chapter Four

234

The eight experimental sentences that each participant read were randomly combined with 24 filler sentences. A set of experimental sentences is displayed in (224) with the intervening elements underlined. In all the conditions except non-JPVC, the element (gap/modifiee) associated with the postverbal phrase appears in the sentential subject where the predicate is transitive and the matrix clause predicate is not transitive. Short-Non-JPVC (224a) [ono koteikyoosi-go yososii sonkoosyo-o kou koto]-go that tutor-Nom easy study-aid book-Acc buy Comp-Nom hituyoo

dot-to nodesu

ka.

necessary be-Past Emontion Q "Was it necessary for that tutor to buy an easy-to-read study-aid book?" Long-Non-JPVC (224b) [ono koteikyoosi-ga yososii sonkoosyo-o kou kotol-go that tutor-Nom easy study-aid book-Acc buy Comp-Nom Ichiro-no siken gookaku notameni-wa hituyoo dat-fa nodesu

ka.

Ichiro-Poss examination pass for-Top necessary be-Past Emotion Q "Was it necessary for that tutor to buy an easy-to-read study-aid book so that Ichiro could pass the examination?" Short-SUBJ-JPVC (224c) [€i yososii sonkaosyo-o kou koto]-go hituyoo dot-to easy study-aid book-Acc buy Comp-Nom necessary be-Past nodesu ka, anD kateikyoosirga. Emotion Q that tutor-Nom

Long-SUBJ-JPVC (224d) [e,yososii sonkoosyo-o kou koto]-go Ichiro-no easy study-aid book-Acc buy Comp-Nom Ichiro-Poss siken gookaku notameni-wa hituyoo

dat-fa nodesu ka,

examination pass for-Top necessary be-Past Emotion Q

ano kateikyoosi-ga. that tutor-Nom

Japanese Postverbal Contructions

235

Short-OBJ-JPVC (224e) [ano kateikyoosi-ga €I kau kotoJ-ga hituyoo dat-ta that tutor-Nom buy Comp-Nom necessary be-Past nodesu ka, yasasii sankoosyo-o. Emotion Q easy study-aid book-Ace

Long-OBJ-JPVC (2241) [ano kateikyoosi-ga e, kau koto]-ga Ichiro-no siken that tutor-Nom buy Camp-Nom Ichiro-Poss examination gookaku notameni-wa hituyoo

pass for-Top

dat-ta

nodesu ka,

necessary be-Past Emotion Q

yasasii sankoosyo-o. easy study-aid book-Ace

Short-ADJ-JPVC (224g) [ano kateikyoosi-ga sankoosyo-o kau koto]-ga that tutor-Nom study-aid book-Ace buy Comp-Nom hituyoo

dat-ta nodesu ka, yasasii.

necessary be-Past Emotion Q easy Long-ADJ-JPVC (224h) [ano kateikyoosi-ga sankoosyo-o kau koto]-ga that tutor-Nom study-aid book-Ace buy Comp-Nom Ichiro-no

siken

gookaku notameni-wa hituyoo

lchiro-Poss examination pass for-Top

nodesu ka, yasasii. Emotion Q easy

dat-ta

necessary be-Past

Procedure:

The experiment was conducted on a paper basis. The data were elicited utilizing Magnitude Estimation (ME) (see below). The ME procedure requires participants to report their judgements based on the range of variations they prefer employing. The value must be greater than zero. otherwise, it would be impossible to log-transform the result. Three participants marked certain sentences zero. Magnitude Estimation:

ME is a method which was developed by psychophysicists to measure judgements of sensory stimuli (Keller 2003. 652). It has recently been

Chapter Four

236

in

utilised linguistics to elicit linguistic judgements just as in the field of psychophysics (e.g. Bard, Robertson, and Sorace 1996; Cowart 1997; Featherston 2005). In ME, participants intuitively assign a numeric value to each stimulus. The number they choose may be any positive real number or fraction. For instance, if you are presented with two sentences, the first is the reference sentence to which you can assign an arbitrary number. If you believe the second sentence is three times more acceptable than the reference (i.e. the first sentence), you would assign it a value three times the value of the reference. If you believe the second is half as acceptable as the reference, you would assign half the value of the reference (Sprouse 2007, 16). Thus, participants provide comparative judgements. Results:

Table 4-3 Results and paired t-tests for sentence type conditions

� non-JPVC

postverbal elements

mean for each condition short chstance

long chstance

degrees of freedom Cd!)

t - 1 . 86

p

0.010

0.087

23

0038 (>0.01)

SUBl

-0.24

-0.46

23

2.82

0.0049 « 0.01)

OBl

-0. 1 3

-0.58

23

5.39

0.0000090 « 0.01)

ADJ

-0.58

-0.78

23

3.67

0.00060 « 0.01)

Data were first divided by the reference sentence's score and then the normalised data were log-transformed. Paired t-tests were conducted on each sentence type to verify the prediction that there is a distance effect on the acceptability of JPVCs. The t-test results are summarised in table 4-

in

Japanese Postverbal Contructions

237

3. Mean acceptability is graphed in Figure 4-2 for the "intervening elements" and "sentence types."

0.2 0. 1 o

-0. 1

� �

-0.2

-0.3

-0.4

-0.5 -0.6 -0.7 -0.8 -0.9

,-------,

f---"-'--'-'--� r-----f-------�: f-------�: f------------�: f--------------'=='--f------------------j f------------------'="---j

III SHORT o LONG

L_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ -----' Type

Figure 4-2: The effect ofthe intervening elements on acceptability

The analysis revealed that when more elements intervene between a postverbal phrase and a gap/modifiee, lPVCs are significantly less acceptable; the acceptability difference between short and long distance conditions was highly significant In contrast, non-lPVCs display no significant difference in acceptability regardless of elements intervening between matrix sentential subj ects and predicates. Discussion

This experiment reveals that the lPVC displays the linear distance effect if more elements intervene between a postverbal phrase and a gap/modifiee, whereas the non-lPVC does not display such an effect even if there are elements intervening between matrix subjects and predicates. Thus, it is necessary to propose a parsing strategy to manage the contrast between lPVCs and non-lPVCs regarding the length effect Below, I will attempt to provide a tentative explanation for this contrast within Hawkins' (2004) framework of a parsing principle called Minimise Domain (MiD), as offered in (225): 151

1 5 1 See footnote 32 in Chapter 3.

Chapter Four

238

(225) Minimise Domains (MiD) The human processor prefers to minimise the connected sequences of linguistic and their conventionally associated syntactic and semantic properties in which relations of combination and/or dependency are processed. The degree of this preference is proportional to the number of relations whose domains can be minimised in competing sequences or structures, and to the extent of the minimisation difference in each domain. (Hawkins 2004, 31)

[OnTIS

"Combination" is defined as follows: (226) Combination Two categories A and B are in a relation of combination iff they occur within the same syntactic mother phrase or maximal projection (phrasal combination), or if they occur within the same lexical cooccurrence frame (lexical combination). (Hawkins 2004, 1 8) In big car, for example, big is in phrasal combination with car because both big and car are in the same NP, ate combines with the cake in the same VP, and the subject the girl combines with this VP within vP. According to Hawkins, these phrasal combinations are determined by general phrase structure rules. Subjects and objects also exist in lexical combinations with verbs and are listed alongside them in its lexical entry. The definition of "dependency" is given in (227): (227) Dependency Two categories A and B are in a relation of dependency iff the parsing of B requires access to A for the assignment of syntactic or semantic properties to B with respect to which B is zero-specified or ambiguously or polysemously specified. (Hawkins 2004, 22) In the boy} praised himseljj, the boy (A) and himself (B) are in a dependency relation because the parser needs to access the boy so that himself, which is zero-specified with respect to co-indexation, can be co­ indexed with the boy The subject the boy (B) in the boy ran depends on ran (A) in the sense that the boy is zero-specified regarding its theta role and the parser is required to access the verb to assign a theta role to the boy As Hawkins (2004) mentions, the verb run also depends on its subject

Japanese Postverbal Contructions

239

and object (if any). Hawkins (2004, 20-21) remarks that "Run is syntactically ambiguous in English between intransitive and transitive uses (the boy ran/the boy ran the water), and it is semantically ambiguous or polysemous between a whole range of interpretations depending on the choice of subject (the water ran/the stocking ran/the advertisement ran) or object (the boy ran the race/ran the water/ran the advertisement) (cf Keenan 1 979)." Hence, the parser needs to access the NPs so that run can be assigned syntactic and semantic properties. Hawkins (2004) notes that as with theta role assignment, several relationships between a given A and B are not only dependent but also combination relationships and he therefore refers to them as "combinatorial dependencies."u2 Hawkins (2004) defines the domain where combinatorial and dependency relations are processed: (228) A combinatorial or dependency domain consists of the smallest connected sequence of terminal elements and their associated syntactic and semantic properties that must be processed for the production and/or recognition of the combination or dependency relation in question. U3 (Hawkins 2004, 23) In the sentence he relied on them for support, the domain sufficient for processing the VP and its three immediate constituents (V, PP" PP,) is relied on them for I" The domain sufficient for processing the lexical meaning of the verb in the above sentence is relied on them and (in an 1 52 It seems to me that all relations between a given A and B are combination relations, by the definition in (226), where both A and B are within the same maximal projections as long as they are within the same clause. However, this problem is irrelevant to the present discussion. I hence assume combination to be a necessary condition for dependency for the moment. 1 53 With respect to the case of movement, besides (226), Hawkins (2004, 1 75) proposes a filler-gap domain (FGD): (i) An FGD consists of the smallest set of tenninal and non-tenninal nodes dominated by the mother of a filler and on a cOIlllected path that must be accessed for gap identification and processing; for subcategorised gaps the path COIlllects the filler to a co-indexed subcategorisor on which the gap depends for its processing; for non-subcategorised gaps the paths cormects the filler to the head category that constructs the mother node containing the co-indexed gap; all constituency relations and co-occurrence requirements holding between these nooes belong in the description of the FGD. 1 54 Like Hawkins I have displayed the relevant domain in bold.

Chapter Four

240

appropriate context) relied on . In the sentence the man} believed himseljj to be honest, the dependency domain for co-indexation is the manl believed himselfl, which only consists of the set of connected words and their associated properties (i.e. indices), because the parser does not have to access the other words (i.e. to be honest) which are irrelevant to co­ indexation. Likewise, in the sentence the girl ate the cake yesterday, the domain for theta role assignment to the subject NP is the girl ate the cake, which only involves the set of connected words and their associated properties (i.e. theta role assignment). As Hawkins (2004, 32) states, the MiD in (225) "defines a preference for the most minimal surface structure domains sufficient for the processing of each combinatorial and dependency relation" Essentially, the intuition behind the MiD is that the processing domain size for combinatorial or dependency relations should be as small as possible so that simultaneous processing of additional phonological, syntactic, and semantic properties can be reduced, decreasing working memory load (Just and Carpenter 1 992). Regarding the linear distance effect on JPVCs, I will consider experimental sentences that display length effects. For expository purposes, I will compare (224b) (i.e. non-JPVC) with (2241) (i.e. Long-OBJ JPVC), reproduced in (229) and (230), respectively. (229) [ano kateikyoosi-ga yasasii sankoosyo-o kau kotoJ-ga that tutor-Nom easy study-aid book-Acc buy Comp-Nom Ichiro-no siken

gookaku notameni-wa hituyoo

Ichiro-Poss examination pass

for-Top

dat-fa

necessary be-Past

nodesu ka.

Emotion Q "Was it necessary for that tutor to buy an easy-to-read study-aid book so that Ichiro could pass the examination?" (230) [ano kateikyoosi-ga e, kau kotoJ-ga Ichiro-no siken gookaku that tutor-Nom buy Comp-Nom Ichiro-Poss examination pass notameni-wa hituyoo dat-ta

nodesu

ka, yasasii sankoosyo-o.

for-Top necessary be-Past Emotion Q easy study-aid book-Ace "Was it necessary for that tutor to buy an easy-to-read study-aid book so that Ichiro could pass the examination?" In (229), when the matrix copula is encountered, the matrix sentential subject [cpano kateikyoosi-ga yasasii sankoosyo-o kau kotoJ-ga "[for that tutor to buy an easy-to-read study-aid bookJ-Nom" is analysed as a

Japanese Postverbal Contructions

241

scrambled element because an adjunct clause [cp Ichiro-no siken gookoku nOlomeni]- wo "[so that Ichiro could pass the examination]-Top" should be adjoined to the matrix vP (at lowest), as (23 1) schematically illustrates.

f"-,... CP

(231)

Y C�a \'

C

A

CP-wa

T

vP

/\'

I

-

"

' AP/\v

hituyoo do

" necessary

Assuming that a trace can be assigned a theta role, the most minimal domain sufficient for assigning it to the matrix subject is [I hiluyoo] "I necessary," which does not involve the adjunct clause CP_wa. 155 Therefore, no length effect can be found. (230) supposes a dependency relation (227) between a postverbal phrase and its relevant element (i.e. a licenser). Thus, in (230), the minimal domain for licensing the postverbal phrase is [e kou kololl-Jla fep

Ichiro-no siken gookaku notamenzl-wa ti hituyoo dat-ta nodesu ka, yasasii sankoosyo-o "was it necessary to buy an easy-to-read study-aid

book so that Ichiro could pass the examination." The domain contains an adjunct clause because the relevant null argument precedes the adjunct clause unlike (229). If an alternative like (232) exists, where the adjunct clause precedes the null argument, (232) would be chosen over (230).

1 55 I do not analyse the domain for cOImecting a subject with its trace (see footnote 153).

242

Chapter Four

(232) Ichiro-no siken gookaku notameni-wa [ano kateikyoosi-ga e, kau Ichiro-Poss examination pass for-Top that tutor-Nom buy koto ]-ga

hituyoo

dat-ta nodesu ka, yasasii sankoosyo-o.

Camp-Nom necessary be-Past Emotion Q easy study-aid book-Ace "Was it necessary for that tutor to buy an easy-to-read study-aid book so that Ichiro could pass the examination?" In (232), the minimal domain for licensing the postverbal phrase is [e kau koto]i-ga ti hituyoo dat-ta nodesu ka, yasasii sankoosyo-o "was it necessary to buy an easy-to-read study-aid book," which does not involve the adjunct clause under discussion The minimal domain in (232) is more minimal than in (230). Hence, (230) would be excluded because it is not the most minimaL If the present account were accurate, the length effect in (230) would follow from the MiD. The same argument holds true of both (224d) and (224h). Therefore, MiD seems to be a promising parsing strategy for dealing with linear distance effects. 156 Finally, it should be noted that even if there is a parsing strategy, which is defined in terms that refer to linear distance, parsing strategies still need to be defined in terms that do not refer to linear distance. This claim is supported by several examples. In both (178) and (1 79), for instance, embedded null subjects are c-commanded by matrix null subjects, and the latter prevents the former from associating with postverbal elements, but the former are closer to postverbal phrases in terms of linear distance than the latter 1 17 4.6.3. Summary The results of the experiment indicate that the JPVC displays the linear distance effect if more elements intervene between a postverbal phrase and a gap/modifiee, whereas the non-JPVC does not display this effect even if there are elements intervening between matrix subjects and predicates. I have attempted to provide a tentative explanation for the contrast between JPVCs and non-JPVCs with respect to the length effect by adopting a parsing principle proposed by Hawkins (2004), namely MiD.

1 56 Gibson (1998, 2000) proposed a similar proposal with respect to the minimisation of dependency length 1 57 See (196) and Chapter 5 for further evidence.

Japanese Postverbal Contructions

243

4.7. Conclusion

In this chapter, I discussed the JPVc. Firs� I presented a critical review of some of the previous accounts of JPVCs and claimed that movement analyses are untenable. Second, I proposed the licensing condition for postverbal elements, assuming that deriving JPVCs involves no movement. Third, I argued that several JPVC syntactic properties can be derived from the interaction of syntactic principles and parsing strategies. Finally, I attempted to provide a tentative explanation for the contrast between JPVCs and non-JPVCs with respect to the length effect based on the results of the experiment on the acceptability of the JPVC

CHAPTER FIVE

A CROSS-LINGUISTIC PERSPECTIVE In the fIrst half of this chapter, I will mainly attempt to show how the locality effecls observed in three types of rightward movement construction in English can follow from the parsing strategies that I proposed in the previous chapters. First, I will argue that right-dislocated NPs in English should be licensed by the licensing condition that holds true for postverbal NPs in JPVCs. I will further claim that the effect of the Right Roof Constraint displayed by English Right Dislocation constructions can be accounted for in terms of parsing strategies. I will then show that the licensing condition is also applicable to Extraposition from NP constructions in English, and that locality effecls shown in Extraposition from NP constructions can also follow from the parsing strategies. I will next consider Heavy NP Shift constructions in English, paying particular attention to locality from the perspective of language processmg. In the second half of the present chapter, I will claim that languages fall into three types with respect to the possibility of NP Shift constructions, as one of the consequences of the interaction of the syntactic principles with the parsing mechanisms. 5.1 English Rightward Movement Constructions

My main purpose in this section is to show how the licensing condition and parsing strategies, which hold true for JPVCs, apply to three types of English Rightward Movement construction: Right Dislocation (5. 1 . 1), Extraposition from NP (5. 1 .2) and Heavy NP Shift (5. 1 .3). 5. 1.1 English Right Dislocation Constructions' The Right Dislocation construction (henceforth, the RDC) ,s a construction in which the "dislocated" NP, a cover term referring to an NP 1 This subsection is based on Kamada (2010, 201 5).

A Cross-linguistic Perspective

245

that appears in the sentence-final position, refers to a pronoun, as observed in Chapter 2. A few examples are reproduced in (1) with the relevant pronoun in italics and the dislocated NP in boldface. (1) a. He was a great novelist, that Charles Dickens. b. Are they learned men, your priests? As (2) shows, the dislocated NP cannot occur outside the embedded clause that contains the relevant pronoun (see 2.3.3). This seems to suggest that the dislocated NP is derived by movement, because a violation of a movement constraint-i.e. the Right Roof Constraint (RRC)--appears to be reflected (Ross, 1986, 1 79) 2, J (2) ?*That they spoke to the janitor about that robbery yesterday is terrible, the cops. (Ross 1986, 25 8)

However, there is an example that violates the RRC but is still acceptable, as shown in (3). (3) [That they spoke to the janitor about that robbery yesterday] is terrible, I mean, the cops. (Whitman, 2000, 450)

The sentence in (3) differs from that in (2) only in that it has I mean inserted between the preceding clause and the dislocated element This suggests that the derivation of the RDC should at least not involve rightward movement. 4 Note that the relevant pronoun is not a "resurnptive" pronoun that repairs an island violation: it would otherwise be difficult to account for the unacceptability of the example in (2), in which the pronoun seems to play no role in repairing the violation of the RRC.' 2 Another possibility is a violation of the Sentential Subject Constraint, which is not operative in some languages such as Japanese (see Ross 1986, 1 49). 3 It should be assumed that the relevant pronollll is inserted somewhere in course of derivation. 4 An example like Ilia! in (3) is originally provided by Tsubomo!o (1 995), who argues against a movement analysis for the RDC and accollllts for some of its properties in terms of information structure. 5 If movement were involved in the derivation of the RDC and the relevant pronOlUl were a resumptive pronoun, the RRC would be a condition on a

246

Chapter Five

Further acceptable examples exist that appear to violate movement constraints such as the Coordinate Structure Constraint (CSC) and the Left Branch Condition (LBC), as witnessed by the acceptability of the examples in (4) (see 2.3.3) 6,7 (4) a. I saw Mary and him downtown yesterday, your friend from Keokuk

b. I noticed his car in the driveway last night, your friend from Keokuk

(Ross 1986, 260) In (4a), the second conjunct (i.e. him) of the coordinate NPs can refer to the dislocated NP, and in (4b), the NP his on the left branch of the NP (i.e. his car) may be linked with the dislocated NP. In other words, if each dislocated NP in (4) were extracted from the site at which the relevant pronoun appears, (4a-b) would violate movement constraints. Hence, neither dislocated NP could be linked with the relevant pronoun, contrary to fact Irrespective of whether an element moves rightward or leftward, English observes the CSC and LBC, as shown in (5) and (6) respectively, where t indicates a trace of the moved elements. (5) a. *What sofa; will he put the chair between some table and t;? (Ross 1986, 97) b. *1 saw Mary and t; downtown yesterday, your friend from Keokuk;. (Ross 1986, 260) (6) a. *Whosei did you steal ti money? (McCawley 1998, 526) b. *I noticed t; car in the driveway last night, your friend from Keokuk;. (Ross 1986, 260)

representation. 6 With respect to the esc, Ross (1986 98-99) states that ''in a coordinate structure, no conjllllct may be moved, nor may any element contained in a conjunct be moved out ofthat conjllllCt." It is assumed that the esc and LBe are regarded as conditions on movement rather than on representations.

7

A Cross-linguistic Perspective

247

Accordingly, the data in (4) also suggest that the derivation of the RDe involves no rightward movement: the dislocated NP IS adjoined to the preceding element via External Merge.8 This subsection is organised as follows. First, I argue that the derivation of the RDe involves no leftward movement, by pointing out empirical problems with Whitman (2000), who claims that the derivation of the RDe in English involves the operation of deletion after leftward movement. I then demonstrate that the interaction of the licensing condition with the parsing principles can account for the cases with which movement analyses fail to cope. Finally, I give a brief summary.

5.1.1.1 Whitman (2000): Problems with a bic1ausal plus deletion analysis

In the previous subsection, I discussed certain empirical problems with rightward movement analyses. In this subsection, I take up Whitman (2000) as an example of leftward movement analyses, and demonstrate that it fails to account for several properties of the English RDe Whitman (2000) follows Kayne (1 994) in claiming that a sentence like that in (7) is derived from the biclausal structure shown in (8a), as in (8b). (7) He is real smart, John. (8) a.

�XP � � � eiSieaSi1iiirt [

P,

John is real smart

Adapted from (Whitman 2000, 452)

S Both the NP and an auxiliary can be right-dislocated. (i) a. He's amazing, that man Is. b. He's jilted her, Jack has. (Brinton 2000, 299. The structure o/modern English: A linguistic Introduction) The examples in (i) suggest that the RDC is derived by ellipsis (Kayne 1994, 78ff.; Poutsma 1928, 169). However, the case where dislocated NPs are associated with pronouns in positions other than the subject (e.g. (lb) and (4)) would pose a problem for an ellipsis approach

248

Chapter Five

I

b [cP1He is real smart], Johnj, [CPl 11. jQURIQJ18ftJ

•

As (8b) shows, John is left-adjoined/dislocated to CP2, and the remaining elements (i.e. the underlined parts) are deleted under an identity condition, thereby generating (7) 9 According to Whitman (2000), the RRC effect displayed in (2) is explained as follows: As in (7), (2) is formed by first conjoining two clauses, and then, as shown in (9), the cops is extracted from the sentential subject in CP2 to adjoin to the left side of CP2 . This extraction, however, violates the Sentential Subject Constraint, resulting in the RRC effect (9) *[cPI That they spoke to the janitor about that robbery yesterday]

IS

terrible, [the cops ]i, [CP2 [that A spoke to the janitor about that robbery yesterday] is terrible] . (Whitman 2000, 458)

However, the analysis above is empirically problematic, because (3) would be excluded in the same way as is (2). 10 Furthermore, the analysis is not adequate to account for the case in which a relevant pronoun is in the object position, as in (10), of which the underlying structure is given in (1 1). (10) I gave him a dollar, that man back there. [� (83d) in Ch. 2] ( 1 1 ) [CPI I gave him a dollar], that man back therel [CP2 I gave A a dollar]

9 Whitman (2000) claims that his analysis is also applicable to the RDClPVC in Japanese and Korean (Chapter 4). Further similar proposals are made by, e.g. Ch\lllg (2012) for Korean and Ott and de Vries (2012) for Dutch and German \¥hat these proposals have in conunon is that the RDC!PVC has a biclausal structure and lllldergoes left-adjoirunent to the second clause before deletion llllder an identity condition. Hence, the application of these approaches to English RDCs will face similar sorts of empirical problems as does \Vhitrnan (2000). AB for Japanese PVCslRDCs, see chapter 4, where I have shown that analyses for JPVCs based on a biclausal plus deletion analysis give rise to serious empirical problems. 10 Although Whitman (2000) provides (3) in his paper, it is melear how he accOlllslt for its acceptability as he does not elaoorate on this type of example.

A Cross-linguistic Perspective

249

As CP1 contains him, the underlined part in CP2 does not appear to be identical to CP1 if an identity condition is defined in phonetic/phonological terms. 11 ,12 Even if the "revised" identity condition could cope with the case in (10), the examples in (4) would still challenge the analysis. That is, your friend from Keokukf s) would be extracted from the respective second clauses [/ saw Mary and your friend from Keokuk downtown yesterday] and [/ noticed your friend from Keokuk's car in the driveway last night]. These respective extractions, however, violate the CSC and LEC, as discussed in the previous subsection (see footnote 7). Thus, the biclausal + deletion analysis cannot account for the acceptability of the examples in (4) either Furthermore, the biclausal + deletion analysis faces another empirical problem. (12) The girl who ate it, the potato salad, was rushed to the hospital.'" 14 (Gundel 1988, 132) The example in (12) shows that the RDC within an embedded clause is possible." There are at least two possibilities for (12) to be derived under the analysis in question. The relevant possible structures corresponding to that in (8) before deletion takes place would be those in (13), with the content ofCP1 in (1 3a) ignored. (13) a.[CPl . . . ], the potato salad1 [ePl the girl who ate t1 was rushed to the hospital] b. (the girl who) [[ep1 ate it], the potato salad1 [ePl ate hll In (1 3a), the potato sai1ld moves out of the relative clause. This movement violates the Complex NP Constraint, and so this possibility should be excluded. As for (1 3b), the potato sai1ld moves leftward inside a relative

1 1 See also the examples in (9) in Chapter 4. 1 Z Ott and de Vries (2012, 3) assume that ''the deleted domain in CPz and its antecedent domain in CPl must be semantically equivalent . . 1 3 Left dislocation i s not permitted in an embedded clause. (i) *The woman who that book, wrote it is a well-known linguist. (Gundel 1988, 84) 1 4 The example has already been provided in Chapter 2. 1 5 The RDC in an embedded clause is not always possible. See (15) in this regard.

Chapter Five

250

clause. As Gundel (1988, 1 5 1) points out, however, leftward movement in a relative clause is not permitted, as illustrated by (14). (14) *The one who [topic-comment structurei doesn't understand tl ] is me. Adapted from (Gundel 1988, 151) Hence, the structure in (1 3b) would not be appropriate either The biclausal + deletion analysis might claim that the internal structure of the embedded clause in (12) is different from that of the relative clause in a sentence like (14). If so, the analysis would be unable to cope with an unacceptable example such as (15), of which the embedded clause appears to have the same structure as that in (12). (15) *Bill gave the girl who [ate it, the potato salad], a dollar 16 Thus, biclausal + deletion analyses such as that of Whitman (2000) have empirical problems (see also Chapter 4). On the basis of the discussion thus far, it seems safe to say that the derivation of the English RDC does not involve movement (i.e. that the RDC is base-generated).

5,1.1.2 The licensing of dislocated NPs Before discussing how the RRC effect in the RDC follows from the parsing strategies I set out in Chapter 3, let us consider the licensing of dislocated NPs, which are assumed to be adjoined to the preceding elements. I7 Recall that I have proposed in Chapter 4 the licensing condition for adjoined phrases, reproduced below in (16). (16) The Licensing Condition (LC) for adjoined phrases (Revised) (where � any syntactic category): A phrase a adjoined to is only licensed if a is associated with an element �, such that (i) a c-commands �, and (ii) a is non-distinct from � in terms of agreement features (i.e. '1'­ features, Case features and honorific features).

X

XP

16 The examples have already been provided in Chapter 2. 1 7 It seems that nothing prevents dislocated NPs from being adjoined to their preceding elements as long as adjlUlction structures may be formed by External Merge (see Chapters 3 and 4). I will later discuss the position to which dislocated NPs are adjoined.

A Cross-linguistic Perspective

251

To see how the LC applies to English RDCs, let us consider the following acceptable examples. (17) a. They spoke to the janitor about that robbery yesterday, the cops. b. The cops spoke to him about that robbery yesterday, the janitor. c. The cops spoke to the janitor about it yesterday, that robbery. (Ross 1986, 258)

CP

Suppose that dislocated NPs are adjoined to via External Merge (see footnote 17). In (17), then, each dislocated NP can c-command the relevant pronoun. In (1 7a), given that NPs do not always have Case features, the dislocated NP the cops is non-distinct from they in terms of agreement features. Is The cops can thus be associated with they, and it is hence licensed. The cops can be construed as the subject of spoke via the interpretive rules proposed Chapter 4, repeated in (18).

in

( 1 8) Interpretive Rules (IRs) for adjoined phrases: Suppose that a phrase a. is adjoined to (where category) and is associated with an element �; then,

XP

X

�

any syntactic

18 If the right-dislocated NPs had Case features, uninterpretable Case features would remain unchecked, yielding a violation ofthe principle of Full Interpretation This point is supported by the observation that fronted NPs can appear in nonargurnent positions without Case features being checked, as sho"M1. in (ib) and (id). (i) a. *1 assured you Jo1m to be a nice guy. b. JOM, I assure you ti to be a nice guy.

(Rizzi 1990, 60) c. *He alleged Melvin to be a pimp. d WhOi did he allege ti to be a pimp? (Postal 1974, 304-305) The above observation falls under the generalisation that overt NPs in peripheral positions do not have to have Case features. This generalisation may extend to the case where, as Endo (1996, 2) observes, postverbal NPs in Japanese lack case markers (e.g. -ga, -0) when the NPs are specific. (ii) Taro-ga eJ tabe-ta yo, ano Teino SUSii. Taro-Nom eat-PAST FP that in question sushi Lit. "Taro ate it, that sushi in question." The example in (ii) indicates that a Case feature is morphologically llllfealised because the postverbal NP does not have it (see (99) in Chapter 4).

Chapter Five

252

(i)

a

is construed as an element sharing properties with � only if is non-distinct from � in terms of categorial features (i.e. syntactic categories), semantic features and semantic types. (ii) a is construed as a potential modifier of � only if a cannot be construed as an element sharing properties with �. a

In (1 7a), the cops is non-distinct from they in terms of categorial features, semantic features and semantic types. The same holds true for (1 7b-