University of Pennsylvania ScholarlyCommons

IRCS Technical Reports Series Institute for Research in Cognitive Science

April 1993

Scrambling as Case-Driven Obligatory Movement

Young-Suk Lee University of Pennsylvania

Follow this and additional works at: https://repository.upenn.edu/ircs_reports

Lee, Young-Suk, "Scrambling as Case-Driven Obligatory Movement" (1993). IRCS Technical Reports Series. 15. https://repository.upenn.edu/ircs_reports/15

University of Pennsylvania Institute for Research in Cognitive Science Technical Report No. IRCS-93-06

This paper is posted at ScholarlyCommons. https://repository.upenn.edu/ircs_reports/15 For more information, please contact [email protected]. Scrambling as Case-Driven Obligatory Movement

Abstract In this thesis I explore the nature and properties of scrambling in Korean. Contrary to the widely accepted view that scrambling is truly optional, I propose that scrambling is a consequence of case-driven obligatory movement, a proposal consistent with the "last resort" condition on movement in [Chomsky 1991] and [Chomsky 1992]. I assume that scrambling is adjunction and defend this view in Ch. 5.

In Ch. 2 and Ch. 3 based on binding facts and scope reconstruction, I claim that scrambling is best analyzed as A-movement. Scrambling either creates a binding relation which does not obtain in the base order, or destroys a binding relation which obtains in the base order. A scrambled element undergoes optional reconstruction for scope interpretation. All these properites are consistent with those of standard A-movement.

In Ch. 4, I propose that scrambling is a consequence of case-driven movement. On the basis of case and word order possibilities in event nominal clauses, I first establish that in orK ean nominative case is licensed by INFL, and accusative case by a complex category formed by the head raising of VERB-to-INFL. Under the VP-internal Subject Hypothesis, all the arguments have to move out of VP to be assigned case. As long as the case licensing conditions are met, arguments may be arranged in any order, and therefore, scrambling is a consequence of case driven movement.

The combination of the assumption that scrambling is adjunction with the proposal that scrambling is A- movement leads to the conclusion that adjoined positions are A-positions, contrary to the view in [Chomsky 1986] that adjoined positions are A'-positions. In Ch. 5, I defend the conclusion that adjoined positions are A-positions in Korean, on the basis of facts involving case assignment to adverbials, binding by a nominative adjunct NP in multiple nominative constructions, and absence of island effects in scrambling out of a scrambled clause.

In Ch. 6, I examine island effects and discourse constraints on scrambling. I argue that islandhood of various clause types is determined by the selectional properties of the clause, as argued by [Cinque 1990] for wh-movement. I also argue that the relevant discourse notion characterizing the scramblability of an element is "presuppositionality" as defined in [Diesing 1990], atherr than specificity as ariousv authors including [Moltmann 1990], [Mahajan 1990] and [Enc 1991] advocate.

Comments University of Pennsylvania Institute for Research in Cognitive Science Technical Report No. IRCS-93-06

This thesis or dissertation is available at ScholarlyCommons: https://repository.upenn.edu/ircs_reports/15 The Institute For Research In Cognitive Science

Scrambling as Case-Driven Obligatory Movemen (Ph.D. Dissertation)

by P

Young-Suk Lee E

University of Pennsylvania Philadelphia, PA 19104-6228 N April 1993

Site of the NSF Science and Technology Center for Research in Cognitive Science N

University of Pennsylvania IRCS Report 93-06 Founded by Benjamin Franklin in 1740

Scrambling as Casedriven ObligatoryMovement

YoungSuk Lee

A Dissertation

in

Linguistics

Presented to the Faculties of the UniversityofPennsylvania in Partial Fulllmentofthe

Requirements for the Degree of Do ctor of Philosophy

AnthonySKroch Sup ervisor of Dissertation

Sabine Iatridou Committee Member

James Yo on Committee Member Donald Ringe Graduate Group Chairp erson

c

Copyright

by

YoungSuk Lee

Acknowledgements

Ihavewaited so long to have this opp ortunity toexpressmy gratitude to the p eople who

have help ed me come this far I am very grateful to my committee memb ers My advisor

AnthonyKroch has shap ed my view of linguistics and every page of this thesis reects his

inuence His emphasis on indep endent thinking and high academic standards has always

driven me to push my limits It was an immeasurable blessing to have Sabine Iatridou

on my committee Even though Sabine came to Penn at the latter stages of my thesis

research having her around was critical in strengthening my ability as a researcher My

desire to emulate her as a teacher and a linguist prompted me to serveasavolunteer TA

for her intro ductory semantics class Besides b eing an intellectual mentor she was always

willing to share her exp erience with me as a friend and has b een a great source of emotional

supp ort James Yoon read every single page of this thesis despite the fact that I asked him

to serveonmy committee at a very late stage His challenging comments got me to think

ab out problems whichIwould have never addressed bymyself Had he sp ent the time on

his research whichhespent in reading my thesis and typing all the comments whichhesent

to me via email he could have written at least two substantial research pap ers of his own

My teac hers at Penn havecontributed greatly to myintellectual development Naoki

Fukuis thesis and his seminar on GB syntax convinced me that Government and Binding

Theory is a suitable theoretical framework even for languages like Korean Aravind Joshis

seminar on Tree Adjoining Grammar and Mark Steedmans seminar on Combinatory Cate

gorial Grammar taught me that there are manyways to approach formal syntax in addition

to enabling me to gain insights into the formalisms Ellen Princes courses on pragmatics

emphasized the close interaction b etween syntax and discourse and prevented me from

drawing hastysyntactic generalizations Various so ciolinguistics classes which I to ok with

Gillian Sanko made me b e aware of the immense variation among sp eakers Besides the

eld work course Sp eech Community which I to ok in my rst semester at Penn was such

awelcome change I got to talk to p eople and learn ab out the Philadelphi a community

instead of sp ending hours and hours in tracking down a huge list of linguistics articles Even

though I never had a chance to take his course Bill Lab ov has b een a role mo del for me

My exp erience with him as a TAforIntro duction to Linguistics was enough to motivate

me to aspire to b e a go o d linguist who is full of energy and enthusiasm

Shortly after I came to Penn I wanted to work on scrambling for my thesis People

talked ab out scrambling a lot as a distinguishing characteristic of Korean and Japanese as

hwork on it aside from Mamoru Saitos thesis opp osed to English But I didnt see muc

on Japanese I felt frustrated every time I heard the word scrambling and wanted to

learn ab out it Conversations with JeeIn Kim who was working on scrambling within

Combinatory Categorial Grammar in inspired me to lo ok at the phenomenon from a iii

theoretical p ersp ective The term pap er which I wrote with Michael Niv on how to handle

scrambling in Combinatory Categorial Grammar was the starting p ointofmy research

Whenever I got stuckinmy research there were always p eople who were willing to help

me out the scrambling reading group which Beatrice Santorini organized guided me to

raise the relevant issues and gave me lots of reading material The pro cess of working

with Beatrice Santorini made writing my prop osal much easier The insightful pap er which

Shigeru Miyagawa wrote and gavemea copyofaftermy presentation at ESCOL has

b een an invaluable resource for my thesis In retrosp ect my thesis work is an attempt to

improve on his work I am so indebted to him that I cannot thank him enough

Ihave also b eneted from discussions with other p eople in the pro cess of writing my

thesis Chapter of this thesis draws heavily from myjointwork with Owen Rambow and

Rob ert Frank Discussions which I had with Owen Rambowonhow to handle scrambling in

Tree Adjoining Grammar help ed me clarify a numb er of assumptions whichhave rarely b een

made explicit on the issue Detailed comments on my thesis prop osal byGertWeb elhuth

and DongWhee Yang rep eated email corresp ondence with Ano op Maha jan conversations

t with Mamoru Saito Peter Sellss comments on my pap er on case p ossibilitie s in even

nominal clauses Caroline Heyco cks challenging questions and comments in the scrambling

seminar and conversations with Michael Hegarty at the latter stages of my researchhave all

made essential contributions to this thesis The Tilburg Workshop on scrambling organized

by Henk van Riemsdijk and Nob ert Corver to ok place when I was ab out to write my thesis

prop osal and enabled me to lo ok at the issue from a broader p ersp ective I also cannot forget

Umit Turan and Beryl Homan whose researchonscrambling in Turkish has always made

me feel that my next step in research should b e a thorough comparative study of scrambling

between Korean and Turkish Michael Niv Dan Hardt and B Srinivas graciously allowed

me to share their oce space whichwas absolutely crucial in exp editing the pro cess of

writing this thesis Alexis Dimitriadis pro ofread this thesis until the last minute of my stay

at Penn

I am grateful to the following p eople for having b een my informants whenever I needed

them Saeko Urushibara Hiroaki Tada Ken Matsuda and Naoki Fukui for Japanese Umit

Turan for Turkish Caroline Heyco ck Beatrice Santorini Michael Hegarty Rob ert Frank

Beth Ann Ho ckey Michael Niv and Dan Hardt for English ChangBong Lee So onHyun

Hong Wonchul Park Sunny Chae Hyokon Kim Jong Park Inhye Kang JeeIn Kim

JinYoung Choi DongIn Cho KiSun Hong and Mo on Jung for Korean

My stayatPenn was very pleasan t and intellectually stimulating b ecause of the great

colleagues I have had Victoria Tredinnick Alexis Dimitriadis Raaella Zanuttini Megan

Moser Julie Auger Christine Nakatani Enric Vallduv Sharon Cote Beryl Homan Libby

Levison Pamela Saunders Michael Mo ore Michael White Jamie Henderson and Je Mark

Siskind WheneverIwas tired of study and needed constructive distractions such as going

swimming going to Wawa and having a meal together Anuj Dawar Barbara Di Eugenio

PatrickParoub ek Tilman Becker and Ulf Cahn von Seelen were always there Ihave

acquired a lot of knowledge by going to the CLiFF meetings Computational Linguistics

Informal FeedbackForum Several visitors to the Linguistics Department and the Institute

for Research in Cognitive Science at Penn have enriched my linguistic exp erience Richard

Oehrle SteveFranks James Higginbotham David PesetskyPeter Sgall and Angelica

Kratzer

It was always fun to go to the Korean linguistics conferences and talk to the following iv

p eople KiSun Hong YoungMee Yu Cho DongIn Cho Saeko Urshibara HyunWoo Lee

SungHo Ahn HyunSo ok Shin So okHwan Cho JongHo Jun HeeRahk Chae EunJo o

Han John Whitman Joan Maling YoungJo o Kim YoungKey Kim Renaud SeungHo

Nam James Yo on Peter Sells JaeHyung Cho JeongSheik Lee So onJa Choi DaeHo

Chung IkHwan Lee Chungmin Lee JoungRan Kim So oYeon Kim JongHo Jun I am

esp ecially grateful to DongIn Cho for sending me several interesting articles My teachers

back in Korea constantly encouraged me to do mybesteven after I came to Penn My MA

thesis advisor Cho onHak Cho MA thesis committee NamSheik Park and DongWhee

Yang my advisors in college SukSan Kim and ByungGun Lee and my college teachers

InSo ok Kim HanKon Kim and ByungTae Cho

There are two p eople whose help I would liketoacknowledge the most Beatrice San

torini and Michael Niv Beatrice Santorini was always willing to pro ofread and discuss my

pap ers She sat down with me for hours and hours in front of a terminal to teachmehow

to write a pap er Her great sense of humor made it all the more enjoyable to b e in her

company Her emphasis on hard work also made me feel less lonelyMichael Niv has b een

my great informal teacher for English Computer Science hextris etc He always listened to

my half baked ideas and made me realize their merits and drawbacks Only his friendship

enabled me to sustain my sanity in hard times

My study at Penn was made p ossible through the nancial supp ort from the Scho ol

of Arts and Sciences of the UniversityofPennsylvania the Mellon Foundation and the

summer fundings from Aravind Joshi The grantnumb ers are

DARPA Sp onsor NJ

DARPA Sp onsor NJ

Sp onsor MC NSF

NSF Sp onsor DCR

Thanks also go to Ellen Prince who encouraged me to apply for funding in my last year at

Penn

Finally I deeply thank my family memb ers my mother my brothers ManJae InJae

WonJae and So oJae and my sister InSo ok Without their supp ort I cannot imagine

myself coming this far I dedicate this thesis to my mother who never exp erienced but

knows b etter than anyone else what it is like tobeinscho ol v

Abstract

Scrambling as Casedriven Obligatory Movement

Author YoungSuk Lee

Sup ervisor Anthony S Kro ch

In this thesis I explore the nature and prop erties of scrambling in Korean Contrary to

the widely accepted view that scrambling is truly optional I prop ose that scrambling is a

consequence of casedriven obligatory movement a prop osal consistent with the last resort

condition on movement in Chomsky and Chomsky I assume that scrambling

is adjunction and defend this view in Ch

In Ch and Ch based on binding facts and scop e reconstruction I claim that scrambling

is b est analyzed as Amovement Scrambling either creates a binding relation whichdoes

not obtain in the base order or destroys a binding relation which obtains in the base order

Ascrambled element undergo es optional reconstruction for scop e interpretation All these

prop erites are consistent with those of standard Amovement

In Ch I prop ose that scrambling is a consequence of casedriven movement On the

basis of case and word order p ossibilities in event nominal clauses I rst establish that

in Korean nominative case is licensed by infl and accusative case by a complex category

formed by the head raising of verbtoinfl Under the VPinternal Sub ject Hyp othesis

all the arguments havetomove out of VP to b e assigned case As long as the case licensing

conditions are met arguments may b e arranged in any order and therefore scrambling is

a consequence of casedriven movement

The combination of the assumption that scrambling is adjunction with the prop osal that

t leads to the conclusion that adjoined p ositions are Ap ositions scramblingisAmovemen

contrary to the view in Chomsky that adjoined p ositions are A p ositions In Ch I

defend the conclusion that adjoined p ositions are Ap ositions in Korean on the basis of facts

involving case assignmenttoadverbials binding by a nominative adjunct NP in multiple

nominative constructions and absence of island eects in scrambling out of a scrambled

clause

In Ch I examine island eects and discourse constraints on scrambling I argue that

islandho o d of various clause typ es is determined by the selectional prop erties of the clause

as argued by Cinque for whmovement I also argue that the relevant discourse

notion characterizing the scramblability of an element is presupp ositionality as dened

in Diesing rather than sp ecicity as various authors including Moltmann

Maha jan and Enc advo cate vi

List of Abbreviations

The following abbreviations are used in the glosses

acc accusative case asp asp ect morpheme

ce causative ending comp complementizer

cop copula datdative case

dec declarative marker dir directional

gen genitive case inst instrumental

loc lo cative mod mo difying sux

nmz nominalizer nom nominative case

pass passive morpheme perf p erfect asp ect

pres present tense prog progressive asp ect

pstpasttense qm question marker

rel relativizer pl plural marker

top topic marker uq universal quantier

vstemverb stem vii

Contents

Acknowledgements iii

Abstract vii

List of Abbreviations ix

Intro duction

Theoretical Concepts

The Xbar schema and phrase structure

The AA distinction

Aprop erties

Denition and characteristics of scrambling

Multiple scrambling

Unb ounded dep endency

Scrambling of clausal elements

Assumptions

Scrambling as movement

Scrambling as Adjunction

Leftward scrambling vs Rightward scrambling

Scrambling Topicalization and Leftdislo cation

The Organization of the Thesis

Scrambling and Binding

Diagnostics for AA movement

Amovement passive raising

A movement whmovement topicalization

Scrambling and Binding

Lo cal scrambling and Pronoun Binding

Lo cal scrambling and Principle C

Summary

Sp ecial status of sub jects

Long distance scrambling and binding

ApparentParasitic Gaps

Previous analyses

Maha jan

Web elhuth viii

Saito

Scrambling and Scop e

AA distinction and scop e

Scop e reconstruction

LFmovement

Scrambling of whphrases

Licensing condition on whphrases

Scop e reconstruction

LFmovementofscrambled whphrases

Scrambling of NegativePolarity Items

Licensing of NPIs

Reconstruction of a scrambled NPI

Implications

Scrambling as movement

Reconstruction for binding and scop e

Scrambling as casedriven Amovement

Scrambling as casedriven adjunction to IP

Case and word order p ossibilities in event nominal clauses

An analysis

Extension to verbal clauses

Comparison with Miyagawa

Implications

Thetaindex and case assignment

Prop erty of the predicate and hierarchy

Thetaindex

Case assignment rule

Examples

Apparent problems

Arguments without overt case morphemes

Co ordination in event nominal clauses

Case assignment in innitival clauses

Scrambling of nominative arguments

Long distance scrambling

Derivation of longdistance scrambling

Absence of lo cality constraints on scrambling

Achain

Where do es long distance scrambling diverge from lo cal scrambling

Deriving the parametric dierence b etween English and Korean

The Adjoined Argument Hyp othesis

Case assignmenttoadverbials

Sensitivity to the stative distinction

Passivization Test

Adjectival vs Verbal Passives ix

Implications

Binding by an adjunct

Islandho o d of scrambled clauses

Scrambling out of a scrambled complement clause

Islandho o d of extrap osed clausesphrases in English

Why is there a complementadjunct distinction

Constraints on Scrambling

Island eects on scrambling

Scrambling of denite expressions

Scrambling of whexpressions

Summary

Discourse constraints on scrambling

Referentiality and scrambling

Sp ecicity and scrambling

Presupp ositionality and scrambling x

Contents xi

List of Figures

Derivation of GENGEN combination

Derivation of NOMGEN combination

Derivation of NOMACC combination

Derivation of GENACC combination

Sstructure representation of a verbal clause xii

Chapter

Intro duction

Assuming that scrambling is adjunction cf Saito Web elhuth I explore the

nature and prop erties of scrambling in Korean in terms of A and A movement I pro

p ose that scrambling is a consequence of casedriven movement on a par with standard

Amovement contrary to the widely accepted view that scrambling is truly optional This

prop osal is consistent with Amovement prop erties of scrambling with resp ect to binding

but contradicts the view in Chomsky that adjunction is A movement To reconcile

these contradicting views I hyp othesize that IPadjoined p ositions can b e Ap ositions in

Korean on the basis of case and binding facts adjoinedargument hypothesis Through

out this thesis I will assume as a theoretical framework the principles and parameters

approach to syntax and will presupp ose that the reader is familiar with the basics of this

theory

In this chapter I intro duce some theoretical concepts and assumptions which are min

imally necessary for an understanding of the discussion in the following chapters In sec

tion I discuss the theoretical concepts the Xbar schema and the AA distinction

In section I summarize the characteristics of scrambling which I take as basic They

include multiple scrambling unb ounded dep endency and scrambling of clausal arguments

In section I discuss some assumptions ab out scrambling which I will not justify in other

chapters of this thesis ie scrambling as movement scrambling as adjunction leftward

vs rightward scrambling and topicalization Finally I outline the con tents of eachchapter

of this thesis in section

Theoretical Concepts

The Xbar schema and phrase structure

I adopt the Xbar schema in Chomsky and Chomsky

X

Z

Z

ZP X

c

c

X YP

An Xbar structure is comp osed of pro jections of heads selected from the lexicon Basic

relations are typically lo cal In structures of the form there are two lo cal relations

the SPEChead relation of ZP to X and the headcomplement relation of X to YP order

irrelevant The headcomplement is the core lo cal relation

Along with the Xbar schema in I assume the VPinternal sub ject hyp othesis which

is originally due to Fillmore and McCawley and later was adopted byanum

ber of syntacticians Fukui Fukui and Sp eas Ko opman and Sp ortiche

among others According to the VPinternal Sub ject Hyp othesis sub jects originate and

are assigned a role in a p osition dominated by a maximal pro jection of the verb Several

dierentvariants of this hyp othesis have b een prop osed with some authors arguing that

sub jects originate in the sp ecier p osition of VP and others that they o ccupy a p osition

max

adjoined to V cf Ko opman and Sp ortiche Here I assume that sub jects o ccupy

SPECVP to b e consistent with the generalized Xbar schema

Under the generalized Xbar schema and the VPinternal sub ject hyp othesis the D

structure representation of a clause I assume is given in pace directionality

CP

b

b

b

SPEC C

c

c

C IP

b

b

b

SPEC I

c

c

VP I

Q

Q

Sub j V

c

c

V Ob j

Regarding the functional pro jection IP in a numb er of authors including Pollo ck

Chomsky and Maha jan have argued for the existence of a multitude of func

tional heads such as TENSE AGR AGR asin

S O

AGR

S

H

H

H

SPEC AGR

S

Q

Q

AGR T

S

b

b

b

SPEC T

Q

Q

T AUX

H

H

H

SPEC AUX

H

H

H

AUX AGR

O

H

H

H

SPEC AGR

O

Q

Q

AGR VP

O

b

b

b

SUBJ V

Z

Z

V OBJ

t system under the assumption that the nature of agreements Given the lack of an agreemen

is pronominal and hence the lack of the motivation for various SPEChead relations I do

not adopt such an elab orate phrase structure for Korean

The AA distinction

The AA distinction has a numb er of consequences in various mo dules of the grammar

and plays a central role in this thesis In Chomsky an Ap osition is dened as

in any p osition which is not an Ap osition is an A p osition At the time when was

prop osed SPEC IP was considered an Ap osition since it is the p osition where a sub ject

is generated and assigned a role

An Ap osition is one in which an argumentsuch asanameoravariable

may app ear in Dstructure it is a p otential p osition

Under the VPinternal Sub ject Hyp othesis however SPEC IP is not an Ap osition

according to the denition given in a sub ject is generated and marked in SPEC

VP even though it maymove to SPEC IP p osition to b e assigned case as argued in

Ko opman and Sp ortiche Nevertheless SPEC IP has all the prop erties asso ciated

Most recently Yoon and Yo on argue for the multitude of functional heads in Korean on the basis

of co ordination facts On the other hand Sells argues against the existence of functional pro jections

in the syntax of Korean and Japanese

While Ko opman and Sp ortiche argue that a sub ject always has to move to SPEC IP to b e

assigned nominative case Iatridou claims that a sub ject can b e assigned case either in SPEC VP or

SPEC IP According to Iatridou agreement is a feature on the verb and therefore the sub ject and the verb

with an Ap osition ie nominative case assignment participation in binding and b eing the

landing site of Amovement

Chomsky intro duces three p osition typ es which replace the traditional AA

distinction narrowly Lrelated p ositions broadly Lrelated p osition and nonLrelated p o

sitions The following is from Chomsky emphases mine

The functional elements Tense and AGR therefore incorp orate features of

the function of the Vfeatures the verb Let us call these features V features

of an inectional element I is to check the morphological prop erties of the verb

selected from the lexicon More generally let us call such features of a lexical

item L L features Keeping to the Xbartheoretic notions wesay that a p o

if it is in a lo cal relation to an Lfeature ie in the internal sition is L related

domain or checking domain of a head with an Lfeature The checking domain

can furthermore b e sub divided into twocategories nonadjoined SPEC and

adjoined Let us call these p ositions narrowly and broadly Lrelated resp ec

tively A structural p osition that is narrowly related has the basic prop erties of

Ap ositions one that is not Lrelated has the basic prop erties of A p ositions

in particular the sp ecier of C not Lrelated if C do es not contain a Vfeature

The status of broad ly Lrelated adjoined positions has been debated particu

larly in the theory of scrambling For our limited purposes we may leave the

matter open

In this thesis I adopt Chomskys threeway distinction of p osition typ es ie narrowly L

related nonLrelated and broadly Lrelated p ositions while continuing to call them A

A and adjoined p ositions resp ectively

Ap ositions narrowly Lrelated p ositions

A p ositions nonLrelated p ositions

Adjoined p ositions broadly Lrelated p ositions

One of the main goals of this thesis is to explore the status of adjoined broadly Lrelated

p ositions by examining the prop erties of scrambling which I assume to b e adjunction

Aprop erties

Prop erties of an Ap osition which I use as diagnostics for identifying the status of adjoined

p ositions include structural case assignment and participation in binding

First structural case is assigned to an Ap osition Structural case typically nominative

and accusative case assignment is dened in terms of either governmentby a caseassigner

cf Chomsky or SPEChead agreement cf Chomsky Whatever mechanism

of case assignmentwe adopt structural case is assigned to an Ap osition and this is reected

in the denition of an Achain stated in

are in a sp echead relationship at Dstructure Nominative case is assigned to the sub ject in SPEC VP by

the agreement feature on the verb when the verb is governed by nite Tense When the pro jection of an

auxiliary verb blo cks governmentoftheverb byTense however a sub ject has to move to SPEC IP to b e

assigned case

A maximal Achain has exactly one Casemarked p osition namely

n

and exactly one marked p osition namely Chomsky

n

An Achain consists of either a single memb er if there is no movement or more than one

element if there is movement to an Ap osition The condition that the rst member of

an Achain in particular when the chain involves movement is the Casemarked p osition

instantiates the idea that Case is assigned to an Ap osition

Second the binding theory whichischaracterized by the three subtheories stated in

refers to a relation b etween two elements in Ap ositions cf Chomsky

A An anaphor is b ound in a lo cal domain

B A pronominal is free in a lo cal domain

C An rexpression is free in the domain of the head of its chain

Denition and characteristics of scrambling

I use the term scrambling b oth in its descriptive and technical senses DescriptivelyI

dene scrambling to b e the p ossibility that arguments of verbs may b e arranged in any

order ie free word order Technicallyscrambling refers to an op eration which either

ord order derives non base word orders or all the p ossible word orders including the base w

dep ending on the particular analysis one adopts In most parts of this thesis except for

in Ch where I prop ose my analysis of scrambling I use the technical term scrambling

to refer to an op eration deriving nonbase word orders Throughout the thesis I will not

sp ecify whether I use the term in its descriptiveoritstechnical sense unless a clarication

is required

I assume that scrambling has the following characteristics which need to b e accommo

dated byany analysis

More than one argument which b elongs to the same argument structure can b e scram

bled ie multiple scrambling

There is no limit to the numb er of clauses which a scrambled element can cross

ie unb ounded dep endency

Not only phrasal but also clausal arguments can undergo scrambling

Some remarks are due on the unb oundedness of scrambling It has b een controversial

whether long distance scrambling scrambling across clause b oundaries is the same syntac

tic phenomenon as lo cal scrambling scrambling within a single clause in various languages

On the basis of diagnostics such as anaphor binding and weak crossover Maha jan

and Saito argue that lo cal and long distance scrambling are dierent phenomena in

Hindi and Japanese resp ectively On the other hand on the basis of the same kind of

diagnostics Homan and Turan and Frank et al argue that lo cal and long

distance scrambling are the same phenomenon in Turkish Korean and German Even

See Lee and Ch of this thesis for a detailed discussion

German do es not allow long distance scrambling out of a tensed clause and all the long distance

scrambling for German involve scrambling out of innitival clauses Web elhuth and Web elhuth

argue for the view that lo cal and longdistance scrambling are the same phenomenon in general

though there are such parametric variations in longdistance scrambling I will assume that

at least in German Korean and Turkish lo cal and long distance scrambling are the same

phenomenon

Multiple scrambling

The three arguments of the ditransitiveverb senmwulhata present may b e arranged in any

of the six logically p ossible orders assuming that the p osition of the verb is xed sentence

nal p osition in this case This is illustrated in

a Sunheeka Youleeeykey chayk han kwenul senmwulhayssta

Sunheenom Youleedat b o ok one CLacc gaveapresent

Sunhee gave a b o ok to Youlee as a present

b Sunheenom chayk han kwenacc Youleedat senmwulhayssta

c Youleedat Sunheenom chayk han kwenacc senmwulhayssta

d Youleedat chayk han kwenacc Sunhee nom senmwulhayssta

e chayk han kwenacc Sunheenom Youleedat senmwulhayssta

f chayk han kwenacc Youleedat Sunheenom senmwulhayssta

Assuming that the base order for a ditransitiveverb sentence is sub jectIODOverb d

and f are instances of multiple scrambling

Multiple long distance scrambling is also p ossible as in c

a nanun Sunheeka Youleeeykey chayk han kwenul senmwulhaysstako

Itop Sunheenom Youleedat b o ok one CLacc gaveapresentcomp

sayngkakhanta

think

I think Sunhee gavea booktoYoulee as a present

ysstako b Youlee eykey nanun Sunheekat chayk han kwenul senmwulha

i i

sayngkakhanta

c Youlee eykey chayk han kwen ul nanun Sunheekat t senmwulhaysstako

i j i j

sayngkakhanta

Considering that Japanese and Korean are similar to each other in so many resp ects it is not clear

what causes such a dierence in the nature of long distance scrambling b etween the two languages It could

well b e the case that the dierence is due to the diculty in getting the grammaticality judgments for

sentences involving long distance scrambling and binding andor weak crossover Contrary to Saito

Yoshimura argues that lo cal and long distance scrambling are the same in Japanese with resp ect to

weak crossover Concerning the dierence b etween Korean and Hindi the dierence might b e related to

the fact that in Hindi whichisanSOV language a nite clause is obligatori ly extrap osed resulting in SVO

order cf Maha jan while Korean is a rather rigid SOV language with no such constraint

Unb ounded dep endency

Scrambling is p ossible not only within the same clause but also across clause tensed and

untensed b oundaries

Scrambling across a clause b oundary

Arguments can b e scrambled across a clause b oundaryasshown in b

a nanun Sunheeka Youleeeykey chayk han kwenul senmwulhaysstako

Itop Sunheenom Youleedat b o ok one clacc gaveapresentcomp

sayngkakhanta

think

I think that Sunhee gave a b o ok to Youlee as a present

b Youlee eykey nanun Sunheekat chayk han kwenul senmwulhaysstako

i i

sayngkakhanta

Scrambling across more than one clause b oundary

Although it is clear that scrambling across one clause b oundary is p ossible it is not crystal

clear whether scrambling across multiple clause b oundaries is grammatical Most p eople

nd that scrambling across more than one clausal b oundary is hard to understand as

illustrated in b

eyse Minholul a nanun nwuka sensayngnimkk

S S S

Itop whonom teachernom Minhoacc

pyenayhasintako malhayssnunci kwungkumhata

like excessivelycomp saidwhether wonder

I wonder who said that the teacher likes Minho excessively

b Minho lul nanun nwuka sensayngnimkkeyse t

i S S S i

pyenayhasintako malhayssnunci kwungkumhata

The generalization on the unb oundednes s of scrambling must b e taken with care Pro cessing of a

scrambled sentence b ecomes prop ortionately harder as the numb er of arguments increases Here I ascrib e

this to a p erformance factor analogous to the pro cessing dicultyofcenter emb edded sentences in English

However there is a clear dierence b etween generating a center emb edded sentence and a scrambled sentence

from a formal p oint of view In the case of center emb edding once weacceptCFG as a grammar formalism

for natural language there is no elegantway of ruling out center emb edded structure in general In the case

of scrambling however the Tree Adjoining Grammar TAG formalism the formal p ower of whichismildly

contextsensitivebetween context free and context sensitive but closer to context free grammar makes a

clear prediction ab out acceptable and unacceptable scrambling Becker et al shows that the language

L f NP NP NP NP V V j apermutationg cannot b e generated byaTAG that contains only

elementary trees ob eying the coo ccurrence restrain ts They also show that under the condition that the verbs

of the emb edded clauses sub categorize for two NPs one of which is an empty sub ject PRO and the other

an S L f NP NP V V j k N and a p ermutation g w NP NP NP NP NP V V V V V

k k

cannot b e generated byaTAGwhich ob eys coo ccurrence constraints Given these ndings it maybethe

case that some long distance scrambled sentences are imp ossible even on comp etence grounds In this thesis

I simply ignore all these considerations and assume that scrambling is unb ounded

a which is in the base order consists of three clauses Scrambling of the most deeply

emb edded ob ject to sentence initial p osition as shown in b is judged to b e unacceptable

by most sp eakers Nevertheless the restriction on longdistance scrambling indicated by

examples like b seems only apparent Consider

cakiney cip ey nanun Minhoka t totwuki

i S S S i

selfgen houseloc Itop Minhonom thiefnom

tulesstanunkesul anunci kwungkumhata

brokeincompacc knowwhether wonder

I wonder whether Minho knows that his house has b een broken into

In the longdistance scrambled phrase cakiney cipey has crossed two clause b ound

aries S and S Nevertheless the sentence is almost p erfect or at least sounds much

b etter than b The contrast b etween b and indicates that scrambling is un

b ounded in principle and that the unacceptabilityofb is due to reasons other than

syntax The grammaticalityofb through dbelow further supp orts the claim that

scrambling is unb ounded In the examples b old face characters indicate coreference

a sensayngnimkkeyse nayka PRO naycwukkaci i pro jectlul

S S S

teachernom Inom PRO next weekby this pro jectacc

S S S

machikilo yaksokhankesul kiekhakokyesinta

nishten promisecompacc is rememb ering

The professor rememb ers the fact that I promised to nish this pro ject

bynextweek

b naycwukkaci sensayngnimkkeyse nayka PRO t i pro jectlul

i S S S i

machikilo yaksokhankesul kiekhakokyesinta

c i pro ject lul sensayngnimkkeyse nayka PRO naycwukkaci t

i S S S i

machikilo yaksokhankesul kiekhakokyesinta

d i pro ject lul naycwukkaci sensayngnimkkeyse nayka PRO t t

j i S S S i j

machikilo yaksokhankesul kiekhakokyesinta

a is the base order sentence with twoemb edded clauses S and S The sub ject of S

is a PRO controlled by the sub ject of S b through d are scrambled counterparts

In bandc the quasiargument naycwukkaci and the ob ject argument iproject

lul b elonging to S havescrambled across S and S resp ectively d shows that

multiple scrambling of b oth the quasiargument and the ob ject argument across the two

clause b oundaries is p ossible

Verb yaksokhata to promise is a socalled control verb and it seems more plausible to treat the

sub ject of clause S as PRO rather than pro Which category I cho ose b etween pro and PRO however

do es not aect the p oint here

By quasiargument I refer to an element the sub categorizatio n status of which is not clear suchasto

Boston as in I am going to Boston soon

Scrambling of clausal elements

Not only phrasal but also clausal elements can b e scrambled as illustrated in b and

b

a Minhoka Youngheeney cipey totwuki tulesstako

S

Minhonom Youngheegen houseloc thiefnom broke incomp

malhayssta

said

Minho said that Younghees house had b een broken into

b Youngheeney cipey totwuki tulesstako Minhokat malhayssta

S i i

a Youngheenun Minhoka cakiney cipey totwuki

S S S

Youngheetop Minhonom selfgen houseloc thiefnom

tulesstanunkesul anunyako mwulessta

broke incompacc knowqm asked

to Younghee asked whether Minho knew that his house had b een broken in

b cakiney cipey totwuki tulesstanunkesul Youngheenun

S i S

Minhokat anunyako mwulessta

S i

In b the clausal argument is lo cally scrambled In b the most deeply emb edded

clause S is longdistance scrambled across S and S

Assumptions

I summarize some theoretical assumptions I make whichhave b een controversial in the

debates on scrambling in Korean They include scrambling as movement scrambling as

adjunction the lack of rightward scrambling and relations b etween scrambling and topi

calization

Scrambling as movement

I assume that scrambling is movement as opp osed to basegeneration I briey review

Hale s prop osal whichisarepresentative of basegeneration analyses of scrambling

discuss some problems in applying his analysis to the scrambling languages at issue and

discuss facts which run counter to a basegeneration analysis

Given the assumption that scrambling is a syntactic phenomenon as opp osed to a PF

phenomenon it has b een controversial whether it is movement or basegenerated The

most inuential prop osal in this debate is the congurationality parameter by Hale

Before Hales prop osal linguists had asso ciated the sup ercial characteristics listed in

with the term noncongurational

This assumption will b e further justied in Ch in relation to long distance scrambling of whphrases

and negative p olarity items

free word order

use of discontinuous expressions

free or frequent pronoun drop

lackofNPmovement transformation

lack of pleonastic NPs eg it there iletc

use of a rich case system

complex verb words or verbcumAUX systems

Hales main concern is to derive the characteristics listed in one of which is free

word order from an interaction b etween at structure and such grammatical principles

as government abstract caseassignment and role assignment He assumes that there

are only two core linguistic typ es to b e dened along the hierarchical dimension of Xbar

theory namelytwobar languages and onebar languages Twobar languages utilize the

endo centric PS rule schemata and ellipses represent the p ositions of sp eciers

and complements

X X

X X

Onebar languages utilize only the PSschema Twobar langauges are termed cong

urational and onebar languages noncongurational

t as a relation which holds b etween the head of a category and Hale denes governmen

its immediate sisters In a conguration like there are two distinct domains in which

goverment op erates

X

Z

Z

NP X

c

c

NP X

The leftmost NP is governed byX while the rightmost is governed by X An imp ortant

prop erty of congurational structures represented by is that in such a structure govern

ment can function to distinguish among the arguments of the lexical head X By contrast

in noncongurational language whose phrase structures are at as depicted in

government as dened ab ove cannot serve to partition a structure into distinct subphrasal

domains of government and hence it cannot serve to distinguish among the arguments of

X

X

a

a

a

a

NP NP X

A direct consequence of the congurationality parameter outlined ab ovetofreeword

order ie scrambling is that wedonothave to app eal to a scrambling rule as found

in Ross or movement to account for free word order If wetake the NP that is

hierarchically closer to the head to b e the ob ject of a clause then the NP which is sister

to X must b e the ob ject and the one whichissistertoX must b e the sub ject in not

vice versa On the other hand in a tree structure like it is imp ossible to x the order

of the sub ject and the ob ject in this wayIn neither of the two NPs is hierarchically

closer to the verb than the other Hence it follows that either of the two NPs can b e the

ob ject and therefore an ob ject may b e basegenerated sentence initially or mediallyThus

under Hales prop osal all the p ossible word orders of a clause in scrambling languages are

basegenerated

A diculty in adopting Hales prop osal with regard to scrambling is that the phrase

structure represented in and predicts that free word order obtains only clause

internally cf Saito and that longdistance scrambling is necessarily a phe

nomenon distinct from lo cal scrambling However this is inconsistent with the facts of

Korean German and Turkish where lo cal and long distance scrambling are the same phe

nomenon A way to accommo date b oth the basegeneration of scrambling and a uniform

analysis for lo cal and long distance scrambling would b e to p osit a phrase structure like

by whichanyword order of a sentence regardless of whether it is simple or complex

is basegenerated

S W

The phrase structure given in however completely ignores the notion of lo calityfor

role assignment and do es not have explanatory p ower

In addition to the problem involving the derivation of long distance scrambling facts

concerning incorp oration and idiom formation suggest that there is some asymmetry b e

tween internal arguments and external arguments which can b e easily accommo dated by

p ositing hierarchical asymmetry b etween them in the phrase structure In idiom formation

we nd only idioms consisting of a complement and its sub categorizing transitiveverb as

in and or a complement and its sub categorizing unaccusativeverb as in and

miyekkukul mekta

seaweed soupacc eat

literal eat soup made of seaweed

idiomatic fail in an exam

See Kim for the denition of unaccusativeverbs in Korean

engdengilul pwutita

hipacc put on

literal put ones hip on the o or

idiomatic sit down

nokchoka toyta

melted candlenom become

literal b ecome a melted candle

idiomatic b e totally exhausted

ttongi thata

shitnom b e burnt

literal shit is burnt

idiomatic to b e extremely anxious

However we do not nd an idiom consisting of an unergativeverb and its external argu

ment The same pattern holds for incorp oration as well ie it is p ossible to incorp orate

a complement of a transitiveverb and an unaccusativeverb to its sub categorizing verb

while it is imp ossible to incorp orate the external argument of a transitiveverb and an

unergativeverb to the verb as illustrated by the contrast b etween which illustrates

the incorp oration of the complement to the transitiveverb hulita to exude and which

illustrates the imp ossibili ty of the incorp oration of the external argument to the unergative

verb wusta to laugh

Kimi nwunmwulhuliessta

Kimnom tearexuded

Kim brokeinto tears

sonyenwusessta

ylaughed bo

A b oy laughed

The asymmetry b etween an internal and an external argument in idiom formation and

incorp oration can b e easily explained if we assume a corresp onding hierarchical asymmetry

in the phrase structure b etween an internal and an external argument

Saito also rejects the at phrase structure analysis for Japanese by arguing that Japanese has

category VP in the phrase structure where the sub ject is VPexternal and a complement is VPinternal

Saito assumes it to b e universal that a verb assigns a role directly to its ob ject but assigns a role to the

sub ject comp ositional ly with its complement He claims adopting Hasegawa that some of Marantzs

arguments for the externalinternal asymmetry are directly reected in Japanese There are idioms

consisting of a transitiveverb and its ob ject but none consisting of a transitiveverb and its sub ject

The semantic role of the sub ject often dep ends on the choice of ob ject but the semantic role of the ob ject

is determined only by the lexical prop erties of the verb and indep endentl y of the choice of a sub ject From

these facts Saito concludes that Japanese sentences must haveVPatthelevel of representation where role

assignmenttakes place and that given the Pro jection Principle stated b elow Japanese must haveVPat

every syntactic level

Representations at each syntactic level ie LF and D and Sstructure are pro jected from the

lexicon in that they observe the sub categorizatio n prop erties of lexical items Chomsky

Scrambling as Adjunction

I assume that scrambling is uniformly adjunction as opp osed to substitution along the

lines of Saito and Web elhuth Web elhuth

This assumption is based on the following grounds One of the characteristics of scram

bling is multiple scrambling ie scrambling of more than one argument b elonging to the

same argument structure cf section Abstracting away from the actual prop erties of

scrambling if we assume that scrambling is substitution then there havetobemultiple

sp ecier p ositions available as landing sites of scrambling As Heyco ckandKroch

argue if sp ecier p ositions may b e licensed by predication by agreement and p ossibly by

caseassignment and if we further assume that sp ecier p ositions available for scrambled

arguments are all licensed by the same relation they havetobelicensedby either all

agreement or all caseassignment since there is only one predication relation at Sstructure

Neither agreement nor caseassignment seems to b e a viable option however Concerning

agreement aside from the fact that Korean do es not have agreement of a pronominal nature

and therefore AgrP is not empirically motivated there havetobemultiple agreement pro

jections which are hierarchically ordered Since mutliply scrambled arguments may o ccur

in any order this will require the hierarchical order of the agreement pro jections to b e as

variable as the order of the arguments Variable ordering of agreement pro jections however

vitiates the purp ose of p ositing multiple functional pro jections for which the hierarchical

ordering has b een crucial in explaining linguistic phenomena in a predictable way The

same argument applies to licensing of sp ecier p ositions by case assignment Namelyifwe

assume that there are multiple functional pro jections and that each pro jection is asso ciated

with only a particular case they have to b e reordered according to the order of arguments

to assign relevant case to the arguments This is undesirable for exactly the same reason

as variable ordering of agreement pro jections

An alternative prop osal would b e that scrambling is substitution on some o ccasions

and adjunction on others This is in fact the line which Maha jan takes for Hindi

However this prop osal has problems which I will discuss in detail in Ch and I reject this

option

Leftward scrambling vs Rightward scrambling

Korean is known as a strict verbnal language along with Japanese Therefore scrambling

of an element to the right of the verb has not b een widely considered However as p ointed

out by Ahn and Cho e rightward scrambling seems to b e p ossible even

though somewhat marginal as in through

Note that my argument for distingui shi ng external and internal arguments in the phrase structure is similar

to Saitos argument for p ositing VPeven though the executions of the idea are dierent I distinguish

external and internal arguments by simply p ositing a hierarchical structure VPinternally in terms of VP

internal sub ject hyp othesis while Saito distinguis hes them by p ositing a sub ject p osition to b e VPexternal

and an ob ject p osition VPinternal

Cho e calls rightward scrambling Korean Inversion

t sakwalul hwumchiessta John i

i i

appleacc stole Johnnom

John stole the apple

nanun t coahaynelul

i i

Itop like youacc

I likeyou

Minhoka t malhayssta Youngheeka cakilul coahantako

i i

Minhonom t said Youngheenom selfacc likecomp

i

Minho said that Younghee likes him

In the sub ject is scrambled to the right of the verb in the ob ject and in

the clausal complement Despite the fact that these instances of rightward scrambling are

p ossible there are some clear contrasts b etween rightward and leftward scrambling

First leftward scrambling is ne in general regardless of the typ e of the sentence in

terrogative declarative cf and On the other hand rightward volved eg in

scrambling across the verb in an interrogativesentence results in an ungrammaticalityas

in and

Minho lul Johni t coahani

i i

Minhoacc Johnnom likeqm

Do es John like Minho

nwukwu lul Johni t coahani

i i

whoacc Johnnom likeqm

Who do es Minho like

In and the prop er name Minho and the whphrase nwukwu resp ectively are

scrambled leftward And the sentences are grammatical

Johni t coahani Minholul

i

Johnnom likeqm Minhoacc

Do es John like Minho

t Youngheelul coahani nwu ka

i i

whonom Youngheeacc likeqm

Who likes Younghee

In and the prop er name Minho and the whphrase nwukwu resp ectively are

scrambled rightward and are ungrammatical

Second rightward scrambling of an emb edded argument to the p osition b etween the

emb edded and the matrix verb in a complex sentence is imp ossible as in while an

emb edded argumentmay b e freely scrambled leftward to any p osition

There is no obligatory syntactic whmovement in Korean and the instances of apparent whmovement

are considered as sub cases of scrambling

A more accurate description of the grammatical status of sentences and would b e to say that

the scrambled elements do not feel as a part of the sentences from which they originate Rather they feel

as the initial elements of the subsequentsentences in the discourse It is interesting to note that even other

scrambling languages suchasTurkish Umit Turan pc and Hungarian cf Kiss whichallowan

argument to o ccur in p ostverbal p osition freely do not allow p ostverbal o ccurrences of whphrases

nanun Minhoka t coahantako Younghee lul sayngkakhanta

i i

Itop Minhonom likecomp Youngheeacc think

I think that Minho likes Younghee

In the ob ject of the emb edded clause is scrambled to the p osition right after the

emb edded verb and the sentence is ungrammatical However as noted by Cho e

rightward scrambling of an emb edded argumenttosentence nal p osition ie the p osition

following the matrix verb is much more acceptable than the typ e exhibited in

nanun Minhoka t coahantako sayngkakhanta Younghee lul

i i

Itop Minhonom likecomp think Youngheeacc

I think that Minho likes Younghee

Finally rightward scrambling diers from leftward scrambling with resp ect to proso dy

In righ tward scrambling there is a denite pause b etween the predicate and the p ost verbal

element while there is no such pause in leftward scrambling The pause in rightward

scrambling is accompanied by the feeling of an afterthought whichwe do not havefor

leftward scrambling

On the basis of the dierences describ ed ab ove I tentatively conclude that leftward

scrambling diers from rightward scrambling leaving the explanation of the dierences to

future research I will consider only leftward scrambling in this thesis and will continue to

use the term scrambling to refer to leftward scrambling

Scrambling Topicalization and Leftdislo cation

Arguments in Korean maybemarked with the socalled topic marker nun as in

Scrambling across a topic marked phrase is p ossible as in

Kimun yekwen sincang wuntongul cekkuk cicihanta

Kimtop feminism movementacc hard supp ort

Kim strongly supp orts the feminism movement

yekwen sincang wuntong ul Kim un t cekkuk cicihanta

i i

feminism movementacc Kimtop hard supp ort

Kim strongly supp orts the feminism movement

There are at least two questions related to a topicmarked phrase and scrambling First

what is the p osition o ccupied by a topicmarked phrase Do es it o ccupy the same kind

of p ositions as other nontopicmarked arguments or do es it o ccupy a sp ecial p osition

eg SPEC of TopicPhrase Second is scrambling across a topicmarked phrase the

same as scrambling across a casemarked one In this thesis I will assume that a topic

marked phrase do es not o ccupy a sp ecial p osition and that scrambling across a topicmarked

phrase is no dierent from that across a nontopicmarked phrase I will briey justify these

assumptions I also briey discuss the so called leftdislo cation which has b een assumed

The kind of contrast b etween and that we observe in Korean also seems to exist in Turkish

Beryl Homan in p ersonal communication which is another verbnal scrambling language

See Erku for the distinction b etween rightward scrambling and p ost verbal afterthoughtinTurkish

to b e a sub class of topicalization and used to argue for the distinctness of scrambling and

topicalization I argue that topicalization is distinct from leftdislo cation and therefore

arguments based on left dislo cation to distinguish topicalization from scrambling are not

valid

Topicalizati on

Nun which is generally called a topic or thematic particle has two functions i to

mark the theme of the sentence ii to mark an element whichiscontrasted with some

other element either present or understo o d in the sentence It is generally p erceived that

a topicmarked elementinsentence initial p osition receives the theme reading cf and

a topicmarked element in sentence medial p osition the contrastive reading cf

Kimun yekwuensincang wuntongul cekkuk cicihanta

Kimtop feminism movementacc hard supp ort

As for Kim he strongly supp orts the feminism movement

Kimi yekwuensincang wuntongun cekkuk cicihanta

Kimnom feminism movementtop hard supp ort

Kim strongly supp orts the feminism movement but not other things

Given this sup ercial dichotomybetween the theme and contrastive readings of a topic

marked phrase according to its p osition in a sentence p eople have argued that it is only the

sentenceinitial topicmarked phrases whic h are derived by topicalization and that they

occupy the highest SPEC p osition in the phrase structure

Below however I argue that it is only a subset of all sentenceinitial topicmarked

phrases which receive the theme reading suggesting that the p osition o ccupied by a topic

marked phrase is not a go o d indicator for there b eing a topicalization movement even if

we accept the view that only a sentence initial topicmarked phrase with the theme reading

has undergone topicalization movement Furthermore an element can scramble across a

sentence initial topicmarked phrase with the theme reading contradicting the view that

a topic phrase o ccupies the highest p osition in the phrase structure My discussion b elow

heavily draws from the observations made in Kuno for Japanese

As Kuno observes for Japanese when the subject noun phrase in sentence initial p osi

tion is followed by nun if it is either generic or anaphoric b oth the thematic and the

contrastiveinterpretation result as illustrated in

Johnun Bostoney kassta

Johntop Bostonloc went

i theme Sp eaking of John he went to Boston

ii contrast As for John he went to Boston but not other p eople

When the sub ject is marked with dative case ie exp eriencer sub ject as in the

contrastive reading is much more prominent than the theme reading even when the sub ject

is anaphoric as in

Kimeykeynun cacenkeka kkok philyohata

Kimdattop bicyclenom really is in need of

As for Kim he is really in need of a bicycle as opp osed to other p eople

When either an indenite sub ject or a nonsubject noun phrase is followed by nun usually

only the contrastive reading results as in and

Boston eynun Johni t kassta

i i

Bostonloctop Johnnom went

contrast As for Boston John went there

manhun salamtulun partyey kassta

many p eoplepltop partyloc went

Many p eople went to the party but not everyone

Furthermore scrambling across a topicmarked phrase with the theme reading overrides

the theme reading and leaves the contrastive reading as the salient reading of the topic

y the contrast in the force of theme reading of the marked phrase This is illustrated b

topicmarked phrase in and

yemsonun congilul coahanta

goattop pap eracc like

theme Sp eaking of goats they like pap ers

congi lul yemsonun t coahanta

i i

pap eracc goattop like

Pap ers goats like but not other animals

As illustrated in at least three factors are involved in determining the reading

of a topicmarked phrase namely the grammatical function the information status and

the p osition of the topicmarked phrase The readings which a topicmarked phrase receives

according to these criteria are summarized in table

anaphoricgeneric sub ject nonsub jectindenite sub ject

sentence initial contrastive theme contrastive

sentence medial contrastive contrastive

Table Available readings of a topicmarked phrase

Accepting the view that only topicmarked phrases with the theme reading are derived

by topicalization movement the surface p osition of a topicmarked p osition is not a go o d

indicator for the topicalization movement Rather the topicalization at issue is sensitive

to the grammatical function and the information status of the topicmarked phrase as

well as its surface p osition However this is not what we nd in topicalization in other

languages In particular topicalization in German is blind to the grammatical function

and the information status of the topicalized element Moreover even if there is a sp ecial

p osition o ccupied by a topic phrase it cannot b e the outermost highest p osition in the

phrase structure otherwise scrambling across a topic phrase should b e imp ossible Given

this discrepancy b etween syntactic topicalization in German and the socalled topicalization

in Korean which is signalled by the presense of the topic morpheme it seems undesirable to

analyze topicmarked phrases with the theme reading as b eing derived by a topicalization

movement Instead the topic morpheme can b e attached to any grammatical category and

the particular reading we get from a topicmarked phrase is via conspiracy among various

factors which are summarized in table

Leftdislo cati on

Another main argument for distinguishing topicalization from scrambling is that topical

ization can license a resumptive pronoun while scrambling cannot as illustrated by the

contrast b etween topicalization and scrambling cf Saito for similar

discussion in Japanese

Jihonun Minhoka kulul coahanta

Jihotop Minhonom heacc likes

As for Jiho Minho likes him

Jiholul Minhoka kulul coahanta

Jihoacc Minhonom heacc likes

Jiho Minho likes him

I will argue b elow that topicalization with no resumptive pronoun as in and

topicalization with a resumptive pronoun as in which is called leftdislo cation are

distinct phenomena Therefore the ability to license resumptive pronouns is not a prop

erty distinguishing topicalization from scrambling ev en though it is a dierence b etween

scrambling and leftdislo cation

Dierences b etween Topicalizatio n and Leftdislo cati on

A rst dierence b etween topicalization and leftdislo cation is that while b oth denite

and generic NPs can b e topicalized only denite NPs can b e leftdislo cated Nonsp ecic

indenite NPs can b e neither topicalized nor leftdislo cated

Topicalization of a nonsp ecic indenite NP

nwukwuinka nun Minhoka t coahayssta

i i

someonetop Minhonom liked

Someone Minho liked

Leftdislo cation of nonsp ecic indenite NP

nwukwuinka nun Minhoka kukunyelul coahayssta

i

someonetop Minhonom hesheacc liked

Someone Minho liked herhim

As denoted by in the use of a resumptive pronoun is somewhat marginal Nevertheless the

contrast b etween and is clear

However nonsp ecic indenites can b e scrambled as will b e discussed in Ch

Topicalization of a generic NP

sakoa nun Minhoka t coahanta

i i

appletop Minhonom likes

Apples Minho likes

Leftdislo cation of a generic NP

sakoanun Minhoka kukestulul coahanta

appletop Minhonom itplacc likes

Apples Minho likes them

Topicalization of a denite NP

Jiho nun Minhoka t coahayssta

i i

Jihotop Minhonom liked

Jiho Minho liked

Leftdislo cation of a denite NP

Jiho nun Minhoka kulul coahayssta

i

Jihotop Minhonom heacc liked

Jiho Minho liked him

bled across a leftdislo cated element while there is no such Second no element can b e scram

restriction for topicmarked phrases

Scrambling across a topicmarked element

a Minho eykey Youngheenun t pyencilul ssessta

i i

Minhodat Youngheetop letteracc wrote

To Minho Younghee wrote a letter

b na eykey Minho nun Youngheeka t t sokayhaycwuessta

j i j i

Idat Minhotop Youngheenom intro ducegave

TomeYounghee intro duced Minho

Scrambling across a leftdislo cated element

a Minho eykey Youngheenun kunyeka pyencilul ssessta

i

Minhodat Youngheetop shenom letteracc wrote

To Minho Younghee she wrote a letter

b na eykey Minho nun Youngheeka t kulul sokayhaycwuessta

j i j

Idat Minhotop Youngheenom heacc intro ducegave

TomeYounghee intro duced Minho

Third leftdislo cation is strictly a matrix clause phenomenon while topicalization is p ossible

in b oth matrix and emb edded clauses

Topicalization in an emb edded clause

naykaMinho nun Youngheeka t coahantako malhayssta

i i

Inom Minhotop Youngheenom likecomp said

I said that Minho Younghee likes

Leftdislo cation in an emb edded clause

naykaMinho nun Youngheeka kulul coahantako malhayssta

i

Inom Minhotop Youngheenom heacc likecomp said

I said that MinhoYounghee likes him

Fourth leftdislo cation is not sensitive to relative island eects regardless of the grammatical

function of the leftdislo cated element while topicalization of an ob ject out of a relative

clause results in ungrammaticality

Topicalization of a sub ject out of a relative clause

Minho nun t t ipko issnun os i telepta

i S Rel i j j

Minhotop is wearingrel clothnom is dirty

Minho the cloth that he is wearing is dirty

Leftdislo cation of a sub ject out of a relative clause

Minhonun kuka t ipko issnun os i telepta

S Rel j j

Minhotop henom is wearingrel clothnom is dirty

Minho the cloth that he is wearing is dirty

Topicalization of an ob ject out of a relative clause

Minho nunnayka t t coahanun yeca lul anta

j Rel i j i

Minhotop Isc nom likerel womanacc know

Minho I know the woman who likes him

Leftdislo cation of an ob ject out of a relative clause

Minhonunnayka t kulul coahanun yeca lul anta

Rel i i

Minhotop Isc nom heacc likerel womanacc know

oman who likes him Minho I know the w

The dierences b etween topicalization and leftdislo cation describ ed ab ove suggest that

they are distinct phenomena The presence of island eects in topicalization and its absence

in leftdislo cation indicate that the former is movement whereas the latter is basegeneration

The fact that leftdislo cation is a matrix clause phenomenon while topicalization is not

and the fact that scrambling across a leftdislo cated element is imp ossible or at least quite

marginal whereas scrambling across a topicmarked element is ne indicate that the p osition

o ccupied by a leftdislo cated element is higher than the p osition o ccupied by a topicmarked

element Once we accept the view that leftdislo cation and topicalization are dierent the

characteristics of leftdislo cation cannot b e used to justify the sp ecial status of the p osition

o ccupied by a topicmarked phrase

In summary I argued that scrambling across a topicmarked phrase is no dierentfrom

scrambling across a casemarked phrase I indirectly justied this view by arguing that a

topicmarked phrase do es not o ccupy a sp ecial p osition and that leftdislo cation is distinct

from topicalization and therefore characteristics of leftdislo cation cannot b e an indication

of the sp ecial status of the p osition o ccupied by a topicmarked phrase

The Organization of the Thesis

In Ch and Ch based on binding facts and scop e reconstruction I claim that scrambling

is b est analyzed as Amovement Scrambling either creates a binding relation whichdoes

not obtain in the base order or destroys a binding relation which obtains in the base order

Ascrambled element undergo es optional reconstruction for scop e interpretation All these

prop erties are consistent with those of standard Amovement

In Ch I prop ose that scrambling is a consequence of casedriven movement On the

basis of case and word order p ossibilities in event nominal clauses I rst establish that

in Korean nominative case is licensed by infl and accusative case by a complex category

formed by the head raising of verbtoinfl Under the VPinternal Sub ject Hyp othesis

all the arguments havetomove out of VP to b e assigned case As long as the case licensing

conditions are met arguments may b e arranged in any order and therefore scrambling is

a consequence of casedriven movement

The combination of the assumption that scrambling is adjunction with the prop osal that

scramblingisAmovement leads to the conclusion that adjoined p ositions are Ap ositions

contrary to the view in Chomsky that adjoined p ositions are A p ositions In Ch I

defend the conclusion that adjoined p ositions are Ap ositions in Korean on the basis of facts

involving case assignmenttoadverbials binding by a nominative adjunct NP in multiple

nominative constructions and absence of island eects in scrambling out of a scrambled

clause

In Ch I examine island eects and discourse constraints on scrambling I argue that

islandho o d of various clause typ es is determined by the selectional prop erties of the clause

t I also argue that the relevant discourse as argued by Cinque for whmovemen

notion characterizing the scramblability of an element is presupp ositionality as dened

in Diesing rather than sp ecicity as various authors including Moltmann

Maha jan and Enc advo cate

Chapter

Scrambling and Binding

In this chapter I consider the interaction b etween scrambling and binding I fo cus on two

typ es of binding relation binding of a pronoun byaquantier which I call pronoun

binding and binding of an rexpression by a pronoun sub case of the principle C

In section I summarize diagnostics for AA movement In section I examine the

interaction b etween binding and scrambling in Korean concentrating on the distribution

of reconstruction eects Reconstruction is obligatory if the binder is a sub ject Otherwise

reconstruction is imp ossible I argue that the limited distribution of reconstruction eects

is due to the sp ecial status of the sub ject in binding Explaining reconstruction eects

this way correctly captures the identical b ehavior of lo cal and long distance scrambling In

section I review some previous analyses of scrambling and claim that none of these can

accommo date the data discussed in

Diagnostics for AA movement

Amovement passive raising

Most diagnostics for Amovement rely on binding facts When elements undergo Amovement

binding relations are entirely determined on the basis of the surface structure as illustrated

in to Amovement either creates a binding relation which do es not obtain in the

base order or destroys the binding relation which obtains in the base order In all the

examples in this chapter the coreference relation is indicated by b old face and antecedent

trace relation by coindexation

This chapter heavily draws up on my jointwork with Beatrice Santorini Lee and Santorini and

Rob ert Frank and Owen RambowFrank et al

Amovement licenses reconstruction with regard to scop e interpretations though This is illustrated by

the scop e ambiguityofsomeone in a ie someone can take a scop e over either the emb edded predicate

or the matrix predicate

a Someone is likely t to win the election

i i

I consider scop e reconstruction to b e distinct from reconstruction with regard to binding in that the former

is a characteristic of Amovement while the latter is a characteristic of A movement I justify this view in

Ch

a It seems to his mother that every b oy is intelligent weak crossover

b Every b oy seems to his mother t to b e intelligent no weak crossover

i i

a It seems to every b oy that his mother is intelligent

b His mother seems to every b oy t to b e intelligent

i i

a It seems to him that Johns mother is intelligent

b Johns mother seems to him t to b e intelligent no strong crossover

i i

a It seems to himself that John is charming

b John seems to himself t to b e charming

i i

Note that the contrast in acceptabilitybetween a and b has often b een referred

ween a and to to show the absence of weak crossover eect and and the contrast b et

b the absence of strong crossover eect in Amovement No matter howwe call them

however they fall under the more general description that binding relations are entirely

determined on the basis of Sstructure after Amovement

A movement whmovement topicalization

Elements which undergo A movement do not aect binding relations That is an element

whichmoves to an A p osition reconstructs to its base p osition

a Every girl loves her parents

b Her parents every girl loves t

i i

a Her mother loves every girlweak crossover

b Who do es her mother lovetweak crossover

i i

a He likes John s mother

b Whose mother do es he liket strong crossover

i i

I use the word reconstruction as a cover term for any analysis in which the premovementposi

tion of a moved elementplays the relevant role As Heyco ck p oints out there have b een three

main typ es of approach to reconstruction for binding a a literal reconstruction lowering of a moved

phrase or some part of a moved phrase at LF eg Riemsdijk and Willia ms b a resort to the

traces left bymovement to account for reconstruction eects on the basis of the Sstructure congura

tion eg Barss c a prop osal that the Binding Conditions apply at some level or levels b efore

whmovement eg Riemsdijk and Williams There are p otential problems in treating reconstruction

eects as a diagnostic for A movement Not only A movement but also some instances of Amovement

exhibit reconstruction eects as shown in a and b b elow

a Pictures of himself seem to John t to b e ugly

i i

b Pictures of himself b other t John

i i

a is an instance of raising Amovement and yet the raised phrase containing the reexive himself seems

to reconstruct to b e b ound by the antecedent John b is a psychverb construction If we assume that

tence in English always involves Amovement following Belletti and Rizzi it is an a psychverb sen

instance of Amovement which licenses reconstruction For the time b eing I ignore examples like likea

and b and continue to assume that reconstruction for binding is a diagnostic for A movement for the

present discussion

a John hates himself

b Himself John hates t

i i

An A moved element licenses a parasitic gap as in a while an Amoved element

do es not as in b

a Which article did you le t without reading pg

i i i

b Which article was led t without PRO reading pg

i i i

Scrambling and Binding

I examine the b ehavior of scrambling with regard to twotyp es of binding binding of a

pronoun by a quantier pronoun binding and binding of an rexpression by a pronoun

ie sub case of Principle C The data suggest that scrambling can b e b est analyzed as

Amovement

In applying the ab ove diagnostics to studies of scrambling p eople have concluded that

scrambling can b e interepreted as an instance of A movementby lo oking at only a few

examples which exibit reconstruction eects However it might b e the case that there are

more factors involved in reconstruction eects other than its b eing A movement T oavoid

the problem of misgeneralizing the data I consider the entire paradigm of binding p ossi

bilities in double ob ject constructions namely binding byeach of three sentence elements

ie sub ject direct ob ject DO and indirect ob ject IO of each of the remaining two el

ements for a total of six cases In the cases where binding is p ossible in the base order

the b ound elementisscrambled past the binder so that scrambling p otentially disrupts the

binding relationship When the binding do es not obtain in the base order the p otential

binder is scrambled past the p otential bindee so that the former ccommands the latter and

therefore binding could b e p ossible For a ditransitiveverb sentence I assume that the base

order is Sub jectIODO

Even though reconstruction in general refers to the case where the binding relation which

obtains in the base order is retained after movement reconstruction can b e sub divided into

two categories namely optional and obligatory reconstruction Authors such as Saito

and Web elhuth argue that reconstruction asso ciated with scrambling is optional To

see whether reconstruction we observe in scrambling is indeed optional or obligatoryI

interpret the reconstruction data in the following way Lets take pronoun binding There

are four logically p ossible combinations of grammaticalityofasentence b efore and after

movement under the intended coreference as in to

a It seems to his mother that every b oy is intelligent

b Every b oy seems to his mother t to b e intelligent

i i

The most extensively examined facts in studies of scrambling are those involving anaphor binding

However I do not discuss anaphor binding data here since it has b een widely observed that the distribution

of the socalled reexive pronoun caki self and the recipro cal pronoun selo eachotherisnotsubjectto

the currently accepted binding theory Hong and Lee Furthermore if the theory of anaphor

binding advanced by Reinhart and Reuland is on the right track all the data whichhave b een discussed as

evidence for reconstruction for the purp ose of anaphor binding b ecome irrelavant For studies on scrambling

and anaphor binding in Korean assuming the eligibil ity of the standard binding theory for Korean I refer

the reader to Lee and Cho b in which the authors argue that anaphor binding patterns exactly

like pronoun binding as discussed in this chapter

a It seems to every b oy that his mother is intelligent

b His mother seems to every b oy t to b e intelligent

i i

a Every girl loves her parents

b Her parents every girl loves t

i i

a Her mother loves every girl

b Who do es her mother lovet

i i

In the sentence is ungrammatical in the base order but b ecomes acceptable after

movement in the sentence is grammatical in the base order but b ecomes ungram

matical after movement in the sentence is grammatical b oth in the base order and

after movement in the sentence is ungrammatical in the base order and remains

ungrammatical after movement

In addition there are three p ossible interpretations we can give ab out a particular

instance of reconstruction reconstruction a always takes place b optionally takes place

es place The p ossible interpretations we can give for each combination and c never tak

in to is given in Table

p

Ex Data Interpretation possible

Base Order After movt Always Optional Never

p p

ok

p

ok

p p

ok ok

p

Table The interpretation of reconstruction eects

Ifasentence is ungrammatical in the base order but b ecomes grammatical after

movement as in we can conclude that there is no reconstruction However we

cannot decide whether reconstruction never takes place or optionally takes place

Ifasentence is grammatical in the base order but b ecomes ungrammatical after

movement as in we can conclude that there is no reconstruction Furthermore

we can draw the stronger conclusion that reconstruction never takes place Otherwise

the sentence could have b een remedied bymovement

If a sentence is grammatical in the base order and still grammatical after movement

as in we can conclude that there is reconstruction However we cannot decide

whether reconstruction always takes place or optionally takes place

Ifasentence is ungrammatical in the base order and continues to b e ungrammatical

after movement as in we can conclude that there is reconstruction Furthermore

we can draw the stronger conclusion that reconstruction always takes place

Although the interpretation of reconstruction in Table is based on pronoun binding

data the table happ ens to equally applicable to Principle C data In the following two

subsections I examine the reconstruction eects of scrambling In all the examples a

sentences are in the base order and b and c examples are in scrambled order Finally a

note on the judgments is in order Judgments on binding relations esp ecially in a scrambled

sentence are sub ject to fairly wide individual variations The judgments I give for each

exampleismeant to b e indicative of the contrast b etween the scrambled and unscrambled

versions of the sentences In the interpretation of reconstruction A means reconstruction

takes place always O optionally and Nnever

Lo cal scrambling and Pronoun Binding

For pronoun binding a quantied expression is the binder and a p ossessive pronoun is the

bindee If binding do es not hold the pronoun is unable to b e interpreted as a b ound

variable and hence the coindexed reading is ungrammatical

Binding by IO quantier Reconstruction Imp ossible

Binding by IO quantier N

Bindee in the DO N

a Kim pancangi nwukwueykeyna pro iwusul sokayhayssta

Kim district chairnom everyonedatuq progen neighboracc intro duced

neighb or The district chair Kim intro duced everyone to his

b Kim pancangi pro iwus ul nwukwueykeyna t sokayhayssta

i i

c pro iwus ul Kim pancangi nwukwueykeyna t sokayhayssta

i i

Bindee in the Sub ject N O

a pro ap ecika nwukwueykeyna yongtonul cwunta

progen fathernom everyonedatuq moneyacc gives

His father gives everyone money

b nwukwu eykeyna pro ap ecikat yongtonul cwunta

i i

Binding byDOquantier N O

There are at least three items which can b e identied as a b ound pronoun in Korean caki self the

overt pronoun kukukes heit and the empty pronoun proFor human entities caki is highly preferred

with a sub ject antecedent and ku with a dativeantecedent For human entities kukes is used regardless of

the grammatical function of the antecedent The empty pronoun pro can o ccur in anyenvironment in which

overt pronouns can o ccur Throughout this thesis I use these three lexical items interchangeably to facilitate

the naturalness of the examples I also limit the range of the antecedents of a b ound pronoun to singular

universal quantiers which end with the sux naasin nwukwuna everyone enu haksayngina every

student This is to abstract away from some dicourse eects which are accompanied by plural universal

quantied expressions suchasmotwn haksayng all students These plural quantiers are often discourse

linked According to some of my informants the contrast in binding p ossibiliti es b etween a base word order

sentence and its scrambled counterpart b ecomes weaker if we replace the singular quantiers with plural

quantiers

Bindee in the IO N O

a Kim pancangi pro iwuseykey nwukwunalul sokayhayssta

Kim district chairnom progen neighbordat everyoneacc intro duced

The district chair Kim intro duced everyone to his neighb or

b Kim pancangi nwukwuna lul pro iwuseykey t sokayhayssta

i i

c nwukwuna lul Kim pancangi pro iwuseykey t sokayhayssta

i i

Bindee in the Sub ject N O

a pro chinkwuka nwukwulul paypanhayssni

progen friendnom whoacc b etrayedQ

y Who did his friend b etra

b nwukwu lul pro chinkwukat paypanhayssni

i i

Binding by Sub ject quantier A O

Bindee in the IO A O

a nwukwunakacaki chinkwueykey kominul thelenohnunta

everyonenom self s frienddat problemacc tell

Everyone tells hisher friend problems

b caki chinkwu eykey nwukwunakat kominul thelenohnunta

i i

Bindee in the DO A O

a nwukwunakacaki uymwulul chwungsilhi ihaynghayssta

everyonenom self s dutyacc faithfully carriedout

Everyone carried out hisher duty faithfully

b caki uymwu lul nwukwunakat chwungsilhi ihaynghayssta

i i

Sub j binder IODO binder IODO binder

IODO bindee IODO bindee Sub j bindee

A O N N O

Table Interpretation of pronoun binding with regard to reconstruction

Lo cal scrambling and Principle C

For Principle C the p otential binder is a pronoun and the p otential bindee a coindexed

rexpression If the pronoun ccommands the rexpression and there are no reconstruction

eects then a Principle C violation results

Binding by IO N

Bindee in the DO N O

a Youngheeka kueykey Minswuuy sacinul poyecwuessta

Youngheenom himdat Minswugen pictureacc showed

Younghee showed him Minswus picture

b YoungheekaMinswuuy sacin ul kueykey t poyecwuessta

i i

c Minswuuy sacin ul Youngheeka kueykey t poyecwuessta

i i

Bindee in the Sub ject N

a Minswuuy tongsayngi kueykey sacinul poyecwuessta

Minswugen brothernom himdat pictureacc showed

Minswus brother showed him a picture

b ku eykey Minswuuy tongsayngi t sacinul p oyecwuessta

i i

Binding byDON

Bindee in the IO N

a naykaMinswuuy pwumoeykey kulul tolyep onayssta

Inom Minswugen parentdat heacc returned

I returned him to Minswus parents

b nayka ku lul Minswuuy pwumoeykey t tolyep onayssta

i i

c ku lul naykaMinswuuy pwumoeykey t tolyep onayssta

i i

Bindee in the Sub ject N

a Minswuuy pwumonimi kulul pangmwunhayssta

visited Minswugen parentsnom heacc

Minswus parents visited him

b ku lul Minswuuy pwumonimi t pangmwunhayssta

i i

Binding by Sub ject A

Bindee in the IO A

a kuka Minswuuy ap ecieykey nay sacinul poyecwuessta

henom Minswugen fatherdat my pictureacc showed

He showed Minswus father my picture

b Minswuuy ap eci eykey kukat nay sacinul p oyecwuessta

i i

Bindee in the DO A

a kuka Minswuuy emmalul coahanta

henom Minswugen motheracc like

He likes Minswus mother

b Minswuuy emma lul kukat coahanta

i i

Sub j binder IODO binder IODO binder

IODO bindee IODO bindee Sub j bindee

A N N

Table Interpretation of Principle C wrt reconstruction

Sub j binder IODO binder IODO binder

IODO bindee IODO bindee Sub j bindee

Principle C A N N

Pronoun binding A O N N O

Conclusion A N N

Table Interpretation of reconstruction wrt binding

Summary

Table is derived by combining table and table

The conclusion at the b ottom row of the table is drawn by taking the stronger conclusion

of the principle C and the pronoun binding data The weaker conclusions are compatible

with the stronger ones but not vice versa The table shows that what determines the

o ccurrence of reconstruction eects is neither the landing site of scrambling eg b efore or

after sub ject the rst and the third columns represent the instances of scrambling across

a sub ject yet the rst column says that reconstruction always takes place and the third

never nor the nature of the scrambled element eg whphrase or rexpression The single

factor determining the o ccurrence of reconstruction eects is the grammatical function of

the binder Namely reconstruction is obligatory when the binder is a sub ject Otherwise

reconstruction is imp ossible

Concerning this rather p eculiar distribution of reconstruction eects the question is

what the reconstruction eects are due to Following the standard diagnostic that movement

to an A p osition licenses reconstruction is it the case that scrambling moves an element

to an A p osition if the binder is a sub ject and to an Ap osition otherwise It seems

unlikely that the prop erty of an elements b eing b ound by another element of a certain

typ e can imp ose restrictions on the typ es of movement esp ecially considering the general

convention in the theory that movement is restricted on the basis of the domain to which

movementtakes place or the inherent prop erties of the moved element With this line of

reasoning I assume that scrambling is Amovement and ascrib e the limited distribution

of reconstruction eects to a sp ecial prop erty of the sub ject in binding not to scrambling

b eing A movement

A p ossible analysis for the absence of WCO eects with regard to scrambling ie creation of pro

noun binding is to assume that the trace left b ehind by scrambling is a null epithet along the lines of

Lasnik and Stowell This line was taken by Cho However I reject this p osition for the follow

ing two reasons

Although it accounts for why scrambling do es not induce WCO it do es not explain whyscrambling

ever destroys binding relation given that a null epithet must b e A b ound and therefore can undergo

reconstruction

Scrambling do es not create SCO either while a null epithet is sub ject to SCO

Sp ecial status of sub jects

In Frank et al we state the sp ecial status of sub jects as sub ject binding generaliza

tion

Sub ject Binding Generalization If X in sub ject p osition binds Y at

Dstructure then X binds Y at all levels of representation

Note that sub ject in the generalization refers to only Dstructure sub ject and therefore

do es not include derived sub jects such as Sstructure sub ject in passives and unaccusative

constructions For the present discussion Dstructure sub ject may b e understo o d as the

external argument in the argument structure

Frank et al also denes binding in terms of coindexation and marking stated

in instead of the standard binding condition which employs the notion of ccommand

and coindexation

X binds Y at LF i X and Y are coindexed and Y is marked byX

In I assume that LF is the level at which binding applies or equivalently marking

is checked The conditions for marking are stated in

X marks Y at level L i

i X is a sub ject at Dstructure or

ii X ccommands Y In this case if X and Y have conicting marking

relations at dierent levels the marking relation established at a later

level sup ercedes the previous ones

i instantiates the sub ject binding generalization in and implies that a marking

relation established b etween a sub ject and another argument at Dstructure is indelibl e

tation Otherwise marking is determined by ccommand through all levels of represen

relation whichmaychange at dierentlevels cf ii The marking relation whichis

established at a later level overrides the one established in the previous levels

I apply this mechanism to some of the examples in the previous section First consider

example from the data concerning principle C rep eated here as At Dstructure

the sub ject pronoun ku ccommands the coindexed Rexpression Minswu and therefore the

former marks the latter according to i The marking in this case is indelibl e at

A notion analogous to marking can b e found in marking prop osed by Lasnik and Saito for the

formulation of ECPYoshimura in fact formulates a pronoun binding condition in terms of marking

In Frank et al we assumed that the level at which binding is checked is the NPstructure prop osed

by Riemsdijk and Williams However the binding relations which obtain in sentences suchasini

below cannot b e accommo dated bysuch a prop osal

i Which picture of himself do es John think Bill likes t

i i

In i the reexive pronoun contained in the moved whphrase may b e b ound either by the matrix sub ject

John or by the emb edded sub ject Bil lHowever an NPstructure analysis wrongly predicts that the reexive

can b e b ound only bytheemb edded sub ject

By sp ecifying that the X is a Dstructure sub ject I am excluding cases involving passives

later levels of the grammar since it is done by the sub ject and hence scrambling of the NP

Minswuuy emmab cannot change the marking relation The sentence is ruled out

by the Condition C

a kuka Minswuuy emmalul coahanta

henom Minswugen motheracc like

He likes Minswus mother

b Minswuuy emma lul kukatcoahanta

i i

Consider another example binvolving pronoun binding rep eated here as b elow

In the pronoun pro is marked at Dstructure by the nonsub ject quantier nwukwu

eykeyna according to ii After scrambling of the quantier at Sstructure however the

quantier do es not ccommand ie mark the pronoun Since the marking established

at Sstructure overrides the one established at Dstructure the quantier do es not bind the

pronoun and the sentence is ruled out under the intended reading

yssta Kim pancangi pro iwus ul nwukwueykeyna t sokayha

i i

Kim district chairnom pro neighboracc everyonedatuq intro duced

District chair Kim intro duced everyone to his neighb or

Thus far I have incorp orated the sp ecial b ehavior of a sub ject regarding reconstruction

into the binding condition adopting Frank et al The new binding condition states

that a binding relation established at Dstructure with a sub ject binder is retained at all

levels of representation regardless of the surface conguration

In the rest of this section I briey remark on the notion of sub ject and discuss some data

which indicate that nominative casemarked NPs are not necessarily sub jects contrary to

what I argued in Heyco ck and Lee Instead an argumentwhich carries the external

role o ccupies the sub ject p osition at Dstructure cf Williams Grimshaw

Sub jects

On the basis of the facts presented in section and it app ears that when an

oblique NP is b ound by a nominativeNP the oblique NP obligatorily reconstructs and

therefore a nominative NP is the sub ject no matter how nominativecasemay b e assigned

However nominative arguments of certain predicates most likely those which carry theme

and experiencer roles to discharge do not trigger reconstruction even when they are the

binder Consider

procakiuy emmaeykey nwukwunaka choikota

proself gen motherdat everyoneuqnom the b estdec

Intended meaning Everyone is the greatest to his mother

In the nominative quantier nwukwunaka everyonenom is the p otential binder of

the pronoun contained in the dative argument If we assume that the nominativeNPina

sentence is always the sub ject and that is a result of scrambling the dative argument

has scrambled across the nominative argument its ungrammaticality is unexp ected since

reconstruction is obligatory in the case of binding by sub jects Instead if we assume that

the order in directly reects its syntactic argument structure and that the dative

argument exp eriencer is the Dstructure sub ject the ungrammaticality of is easily

explained The pronoun pro is not marked at anylevels of representation and therefore

is not b ound by the quantier nwukwuna Binding p ossibilities in and further

indicate that the argumentwhich carries the external role in a sentence is the Dstructure

sub ject of the sentence

nwukwuna ka procakiuy emmaeykey t choikota

i i

everyoneuqnom proselfgen motherdat the b estdec

Everyone is the greatest to his mother

a nwukwueykeyna procakiuy casiki choikota

everyonedatuq proselfgen childnom the b estdec

His child is the b est to everyone

b procakiuy casik i nwukwueykeyna t choikota

i i

proselfgen childnom everyonedatuq the b estdec

The grammaticality of and can b e explained in the following way The dative

arguments exp eriencer are the external arguments and the Dstructure sub jects of each

sentence In scrambling of the nominative argument whichisaquantier created

the pronoun binding at Sstructure In a the dative quantier which the Dstructure

sub ject marks the pronoun contained in the nominative argument at Dstructure This

marking is retained after scrambling of the nominative argument across the dative quantier

causing the reconstruction eects as in b

Long distance scrambling and binding

My treatment of binding in terms of marking do es not distinguish cases of lo cal scrambling

from those of long distance scrambling as long as the binding domain is not limited to a

single clause Therefore it predicts a uniform b ehavior of lo cal and long distance scrambling

in Korean in which the binding domain is the whole ro ot clause To recapitulate the

However constructions similiar to in German indicate that the surface case of an argumentisa

b etter indicator of the sub jectho o d of the argument than the role which the argument carries in that

language Consider i b elow which is discussed in Lee and Santorini

i da seiner Mutter jeder gefallt

that hisgen motherdat everyonenom pleases

that everyones mother likes him

If the exp eriencer argument seiner Mutter which is marked dative not the theme argument jeder whichis

marked nominative is the sub ject and therefore o ccupies the structurally highest p osition at Dstructure its

grammaticality is unexp ected since under this assumption the pronoun is not ccommanded by the quantier

anywhere in the course of derivation On the other hand if the theme argument jeder is the Dstructure

sub ject and therefore marks the pronoun seiner contained in the exp eriencer argument its grammaticality

is easily accommo dated At the moment I have no clear idea ab out what the right explanation for the data

For a detailed discussion of the German data see Lee and Santorini

For a detailed discussion of anaphor binding in Korean within the GB framework the reader is referred

to Yang

descriptive generalization on lo cal scrambling reconstruction is obligatory when the binder

is a sub ject otherwise there is no reconstruction This generalization is extended to long

distance scrambling In the interpretation of reconstruction A means reconstruction takes

place always O optionallyandNnever

The relevant data to b e considered are the ones where the p otential binding relation

obtains b etween one of the matrix arguments and one of the emb edded arguments The

cases in which the p otential binding relation holds b etween two arguments b elonging to

the same emb edded clause and then one of the two scrambles out of the clause is not

relevant since they can always b e reduced to lo cal scrambling There is always a p ossible

derivation in which the scrambled element undergo es lo cal scrambling rst and then long

distance scrambling whichisschematically represented in

scrambledNP t pro t

S i S i i

Long distance scrambling and Principle C

Binding by the matrix sub ject of the bindee in the emb edded ob ject A

a kuka Youngheeka Minswuuy pwumonimul manna p oasstako

Minswugen parentsacc metcomp henom Youngheenom

sayngkakhanta

think

He thinks that Younghee met Minswus parents

b Minswuuy pwumonim ul kukaYoungheekat manna p oasstako

i i

sayngkakhanta

Binding by the matrix IO of the bindee in the emb edded ob ject O N

a nanun kueykey nayka Minswuuy pwumonimul cal tolp okeysstako

Itop hedat Inom Minswugen parentsacc well take care ofcomp

yaksokhayssta

promised

I promised him that I would take go o d care of Minswus parents

b nanun Minswuuy pwumo lul kueykey pwumoeykey naykat cal

i i

tolp okeysstako yaksokhayssta

Binding by the emb edded ob ject of the bindee in the matrix ob ject N

a nanun Minswuuy pwumoeykey nayka kulul cal tolp okeysstako

Itop Minswugen parentdat Inom heacc well takecareofcomp

yaksokhayssta

promised

I promised Minswus parents that I would takegoodcareofhim

b nanun ku lul Minswuuy pwumoeykey naykatcal tolp okeysstako

i i

yaksokhayssta

Long distance scrambling and Pronoun Binding

Binding by the matrix sub ject of the bindee in the emb edded ob ject A O

a nwukwuka nayka cakiuy emmalul hyungp oasstako sayngkakhani

whonom Inom selfgen momacc sp oke ill ofcomp thinkqm

Who thinks that I sp oke ill of his mother

b cakiuy emma lul nwukwuka naykat hyungp oasstako sayngkakhani

i i

Binding by the matrix IO of the bindee in the emb edded ob ject N

a nanun enu haksayngeykeyna nayka kuuy pwumolul

Itop every studentdatuq hegen parentacc

mannap okeysstako yaksokhayssta

meet and seecomp promised

ery student to meet his parents I promised ev

b nanun kuuy pwumo lul enu haksayngeykeyna naykat

i i

manna p okeysstako yaksokhayssta

Binding by the emb edded ob ject of the bindee in the matrix ob ject O N

ananun kuuy pwumoeykey nayka enu haksayngina cal tolp okeysstako

Itop hegen parentdat Inom every student well takecareofcomp

yaksokhayssta

promised

I promised his parents that I would take care of every student

b nanun enu haksayngina kuuy pwumoeykey naykat cal tolp okeysstako

i i

yaksokhayssta

ve the data on long distance scrambling conrms As illustrated by the examples ab o

the prop osal that reconstruction eects in scrambling are not due to its b eing A movement

under the limited circumstances but due to the sp ecial status of sub ject in binding At

tributing the reconstruction eects to the sp ecial status of sub ject easily explains some

data involving long distance scrambling out of a controlled clause which cannot b e easily

explained byany purely congurational account of binding andor scrambling

First consider and

a nanun enu haksayngeykeyna PRO cakiuy immwulul cwungsilhihalako

Itop every studentdatuq selfgen dutyacc faithfully docomp

seltukhayssta

p ersuaded

I p ersuaded every student to do his duty faithfully

b nanun cakiuy immwu lul enu haksayngeykeyna PRO t

j j

cwungsilhi hatolok seltukhayssta

In a the sub ject of the emb edded clause is PRO whichiscontrolled by the dative

argument of the matrix clause Binding of the pronoun pronoun caki in the emb edded clause

by the matrix dative argument nwukwueykeyna is grammatical The binding relation

survives after scrambling of the emb edded ob ject across the matrix dative argument as in

b ie reconstruction

Now consider a and b

a nanun enu haksayngeykeyna PRO kuuy pwumolul

Itop every studentdatuq hegen parentacc

manna p okeysstako yaksokhayssta

come and visitcomp promised

I promised every student to meet his parents

b nanun kuuy pwumolul enu haksayngeykeyna PROt

j j

manna p okeysstako yaksokhayssta

In a the sub ject of the emb edded clause is PRO which is controlled by the matrix

contained in the emb edded ob ject is b ound by the matrix dative argu sub ject Pronoun ku

ment In this case scrambling of the emb edded ob ject across the matrix dative argument

as in b destroys the binding relation ie no reconstruction

The contrast in grammaticalitybetween b and b is unexp ected under an

analysis in which the nature of the p osition to which scrambling takes place determines

the AA nature of scrambling hence the distribution of reconstruction eects In b oth

sentences scrambling has taken place to exactly the same lo cation and therefore wewould

exp ect that reconstruction eects o ccur either in b oth sentences or in neither sentence

contrary to the fact However under the current analysis in which reconstruction eects

are incorp orated into the binding condition in terms of marking the contrast nds a

simple explanation In a the pronoun contained in the emb edded ob ject is marked

at Dstructure bythePRO sub ject which is in turn controlled by the matrix dative quanti

er Scrambling cannot aect this marking and the long distance scrambled counterpart

b is grammatical as exp ected In a the pronoun contained in the emb edded ob

ject is marked at Dstructure by the matrix dative argument Since the marking in this

case is by a nonsub ject scrambling can change the marking relation In its scrambled

counterpart b the marking relation has indeed changed and the intended binding is

not acceptable

Finally I discuss a p otential problem for the current treatment of reconstruction eects

Consider

casini Johnul miwuehanta

selfnom Johnacc hates

lit Himself hates John

can b e ruled out by one of the following twoways The anaphor casin in the sub ject

p osition is unb ound ie principle A violation Or the Rexpression John is b ound by the

anaphor ie principle C violation If we scramble the ob ject NP across the sub ject as in

the sentence b ecomes acceptable

John ul casini t miwuehanta

i i

The current analysis predicts to b e ungrammatical It is b ecause the marking

of John by the anaphor casin at Dstructure should b e retained through all levels hence

causing a principle C violation Contrary to this prediction the sentence is go o d I dont

understand what the right solution for this problem is and leave it as an op en problem

ApparentParasitic Gaps

Examples suchasbc and b where scrambled whphrases bind two gaps have

b een analyzed as parasitic gap constructions cf Ho ji and Saito for Japanese

Lee a for Korean leading to the conclusion that scrambling is an instance of A

movement If this is indeed the case then it p oses a problem for the claim that scram

bling is in principle Amovement In this section however I argue that the socalled par

asitic gaps in this language are not real parasitic gaps but that they are empty pronouns

b ound by scrambled phrases ie instances of creation of pronoun binding by scrambling

Main evidence for this argument comes from the absence of sub jacency eects in licensing

the gaps at issue

a Minhoka nwukwunkaka gap yesp oki ceney

i

b efore Minhonom someonenom p eep intonmz

etten pyenci lul cciep elyessni

i

whichletteracc tore upqm

Which letter did Minho tear up b efore anyone could p eep into it

b etten pyenci lul Minhokanwukwunkaka gap yesp oki ceney t

i i i

cciep elyessni

c Minhoka etten pyenci lul nwukwunkaka gap yesp oki ceney t

i i i

cciep elyessni

is slightly marginal as indicated by which is probably due to the awkwardness of this construc

tion in a null context However the contrast in acceptabili tybetween and is clear Furthermore

sounds p erfectly natural in a prop er discourse context Consider the following conversation

A seysangey e casinul miwuehanun salam i eti isse

i i

on earth selfACC hateREL p ersonNOM where existQM

Where on earth is the p erson who hates himself

B John ul casini t miwuehanunkel

i i

JohnACC selfNOM hateassertive

John hates himself

The order given in B which is the same as is p erfectly natural as a resp onse to an utterance A

Somehow a sounds b etter than a although the gap in question and its antecedent are in the

same conguration in b oth sentences

Larson to app ear using the Light Predicate Raising idea claims that Adjunct Parasitic Gaps result

from complex predicate formation or argument sharing and do not havetodowithAdep endencies

crucially

a t hanp en gap p on salam i nwukwu eykey panhayssni

i j i j

once seerel p ersonnom whodat had a crush onqm

lit Who did the p erson who saw him once have a crush on

j j

b nwukwu eykey t hanp en gap p on salam i t panhayssni

j i j i j

No sub jacency eects

The sub jacency test proves that the gaps in question in and are not real parasitic

gaps or at least dier from parasitic gaps in English

Chomsky notes that the distribution of parasitic gaps is sensitive to sub jacency

as illustrated by the contrast in grammaticalitybetween and

this is the man John interviewed t b efore reading the b o ok you gaveto pq

i i

this is the man John interviewed t b efore hearing ab out the plan to sp eak to pg

i i

The parasitic gap is contained in a relative clause in and in a complement clause of

the head noun plan in Considering that a relative clause is a strong island while a

complemen t clause of a noun is a weak island for sub jacency the contrast in grammaticality

between and is easily explained if we assume that the distribution of parasitic

gaps is sensitive to sub jacency

As will b e discussed in Ch islandho o d of various clauses for scrambling in Korean is

generally weaker than that for whmovement in English Nevertheless relative clauses are

clear islands for scrambling cf Fukui for similar b ehavior in Japanese Applying

the sub jacency test to a sentence in which the gap at issue is contained inside tworelative

clauses we nd that the gaps in question do not ob ey sub jacency cf Saito for

Japanese however Consider

Johni etten pyenci lul t t gap ponayn yeca lul

i REL k REL j i j

Johnnom which letteracc sendrel womanacc

ciltwuhanun cakipwuin eykey t poye cwuessni

k i

b ejealousofrel self s wifedat showedqm

lit Whichletter did John show to his wife who is jealous of the woman

i

who sentt

i

In the gap in question t is contained in relative clause rel which is in turn

i

contained in another relative clause rel Still the sentence is grammatical indicating that

the gaps in question are not sensitive to islands and therefore they are not of the same

nature as parasitic gaps in English

Note that in b oth cases the clauses containing the parasitic gaps are contained in another island an

adjunct clause Since a parasitic gap typically o ccurs in an island to b egin with having another island

aside from the island whichcontains the parastic gap is crucial in testing whether a parasitic gap ob eys

sub jacency

Apparent parasitic gaps as empty pronouns

Given the fact that the distribution of apparent parasitic gaps in Korean diers from that of

parasitic gaps in English and that Korean is a prodrop language which allows an ob ject

or PP to b e dropp ed in addition to a sub ject it seems reasonable to conclude that the gaps

in question are empty pro In fact creation of pronoun binding byscrambling discussed

in section is in favor of this conclusion The gaps in question in bc b and

are empty pronouns b ound by the scrambled phrase in Ap ositions

I conclude this section by giving one more piece of evidence that the gaps at issue are

empty pronouns rather than parasitic gaps The gaps can b e discourse b ound b esides co

varying with the p otential antecedent gap Discourse b oundness is clearly a characteristic

of a pronoun not a parasitic gap Consider b which is rep eated here as

nwukwu eykey t hanp en gap p on salam i t panhayssni

j i j i j

whodat once seeREL p ersonnom had a crushqm

Who did the p erson who sawhim once have a crush on t

j j j

Who did the p erson who sawsomeone once have a crush on t

j k j

In the referent of the gap in question gap may b e consistent with that of the

j

scrambled argument nwukwueykey b ound reading or it could refer to an entity in the

discourse suchasJohn Maryetc

Previous analyses

In this section I review three previous analyses of scrambling They include Maha jan

Web elhuth Web elhuth and Saito These analyses are divided into two

groups dep ending on how they capture A and A movementcharacteristics of scrambling

Maha jan prop oses that the landing site of scrambling can b e strictly divided into Ap osition

and A p osition and that A and A movement prop erties of scrambling are due to the

landing site b eing an A and A p osition resp ectively On the other hand Web elhuth and

Saito prop ose that there is a third p osition typ e which shares prop erties of b oth A and

A p osition I will argue that none of these analyses can successfully accommo date the

Korean data discussed in this chapter

Maha jan

Maha jan attempts to completely reduce scrambling to the standard typ es of move

ment That is it is either Amovement derived by substitution to SPEC p ositions or

A movement derived by adjunction to a maximal pro jection I briey summerize his anal

ysis and discuss some inadequacies of his analysis with resp ect to his own data and the

data discussed in this chapter

Assuming a highly articulated phrase structure along the lines of Pollo ck and

Chomsky as in Maha jan argues that scrambling is a nonunitary

I am grateful to Sabine Iatridou for p ointing this out to me

Web elhuth gives a detailed critique of Maha jans analysis

phenomenon It is Amovement when it is substitution into the SPEC p ositions of functional

categories while it is A movement when it is an adjunction to maximal pro jections

AGR P

s

H

H

H

SPEC AGR

s

Q

Q

AGR TP

s

b

b

b

SPEC T

Q

Q

T AUXP

H

H

H

SPEC AUX

H

H

H

AUX AGR P

o

H

H

H

SPEC AGR

o

Q

Q

VP AGR

o

b

b

b

SUBJ V

Z

Z

OBJ V

The main motivation for p ositing such an elab orate phrase structure lies in the fact that

Hindi has b oth sub ject agreement and direct ob ject agreement as illustrated in and

resp ectively

raam roTii khaataa thaa

Ramm breadf eatm imp b em pst

Ram habitually ate bread

raamne roTii khaayii thii

Ramm breadf eatf p erf b ef pst

Ram had eaten bread

In addition to the morphological motivation based on agreement Maha jan justies the

SPEC p ositions by arguing that scrambling to a SPEC p osition is casedriven That is

NPs that are not structurally case marked VP internally must scramble to a SPEC p osition

where they receive case via SPECHEAD agreement cf Maha jan Under the

system based on such a phrase structure multiple scrambling of Amovement prop erties

Maha jan uses the term LrelatednonLrelated p osition instead of AA p osition and argu

ment shiftadjunction to XP instead of AA movement Since the distinction do es not aect the discussion

below I continue to frame it in terms of the AA distinction for exp ository convenience

According to Maha jan in Hindi structural case assigned by SPECHEAD agreement is not lexically

casemarked An NP which is marked with an overt particle such ne and ko is lexicall y casemarked

is p ossible due to the availabilityofnumerous SPEC p ositions and that of A movement

prop erties due to multiple adjunctions

Despite the conceptual elegance of the claim that scrambling can b e identied with either

substitution or adjunction which in turn corresp onds to the standard A or A movement

his system has some problems Consider and and in Maha jan

uskii bahin sabko pyaar kartii thii

their sisterf SUB everyone DO love dof imp b ef pst

Their sister loved everyone

sabko uskii bahin t pyaar kartii thii

i i

The verb in is imp erfective and the sentence shows sub ject agreement Therefore the

sub ject must have b een scrambled to the SPEC AGRs p osition to b e assigned case without

disrupting the base order Since there is no more SPEC p osition available to the left of SPEC

AGRs scrambling of the direct ob ject sabko across the sub ject o ccupying SPEC AGRs

p osition is predicted to b e A movement Contrary to this prediction however scrambling

of the ob ject as in exhibits an Amovement prop erty The scrambled ob ject sabko

everyoneDO whic hisaquantier binds the pronoun contained in the sub ject phrase

Given this problem Maha jan pc suggests two p ossible ways out One is to generate

an empty SPEC p osition ab ove SPEC AGRs whichwould b e headed by an emptyhead

which can license an inherent case The other is to case mark the sub ject lower down in the

tree say SPEC AGRo p osition which can b e motivated by the fact that in Hindi sub ject

and ob ject agreement are identical in morphology and in complementary distribution and

therefore sub ject and ob ject are structurally case marked from the same p osition Whichever

solution we adopt howeveritcontradicts the main spirit for p ositing extra functional

pro jections As for the rst option of p ositing an empty SPEC headed by an empty head

it do esnt haveany morphological justication and more imp ortantly is not indep endently

motivated As for the second option if b oth sub ject and ob ject are case assigned from the

same p osition then it is not clear why the multiple functional pro jections are necessary to

b egin with

In addition to the problem discussed ab ove Maha jan needs a stipulation which vitiates

the asso ciation of the AA prop erty of scrambling with substitutionadjunction distinction

Consider the examples b elow which are and in Maha jan resp ectively

raamne mohanko apnii kitaab lOTaaii

i j

ij

RamSUB MohanIO self s b o okfDO returnp erff

Ram returned self s book to Mohan

i j

ij

kitaab mohanko t lOTaaii raamne apnii

k j k i

j i

RamSUB self s b o okfDO MohanIO returnp erff

is in the base order and is its scrambled counterparts In the reexive

pronoun apnii contained in the DO can b e b ound either by the sub ject or by the IO Scram

bling of the DO across the IO as in destroys the binding by IO while maintaining

the binding by sub ject Destruction of binding in however is not exp ected under

his analysis In principle adjunction to anyintermediate functional pro jection should b e

p ossible the order in mayhave b een derived by adjunction as well as substitution

oftheDOtoanintermediate function pro jection therefore we never exp ect scrambling

to destroy a binding relation which obtains in the base order Noting this shortcoming of

his analysis Maha jan fn stipulates that leftward adjunction to pro jections

lower than IP is ruled out A consequence of this stipulation however is that scrambling

beyond IP is always A movement adjunction and scrambling within IP is always A

movement substitution whichmakes the substitutionadjunction distinction redundant

as a to ol to distinguish the twotyp es of scrambling

Even though Maha jan do es not make it explicit there is some indication in his data

that distribution of reconstruction eects is not so much due to the nature of the p osition to

whichscrambling takes place as due to the grammatical function of the binder ie binding

by a sub ject cf section for Korean for which he needs another stipulation Consider

which is another scrambled counterpart of which is in Maha jan

apnii kitaab raamne mohanko t lOTaaii

k i j k

i
j

self s b o okfDO RamSUB MohanIO returnp erff

In the direct ob ject containing the reexive pronoun has scrambled across the sub ject

In this case the sub ject still binds the reexive pronoun despite the fact that the latter is not

ccommanded by the former But the binding by the IO is destroyed For the binding by the

sub ject the scrambled DO has to undergo reconstruction But this reconstruction cannot

b e to its Dstructure p osition otherwise destruction of binding by the IO is unaccounted

for To explain the binding fact in c Maha jan assigns it the derivation and

adds another stipulation that reconstruction is p ossible only to a variable p osition

apnii kitaab raamne t mohanko t lOTaaii

k i j k

i
j

k

In the trace of the scrambled ob ject indicated byt is an Ab ound trace anaphor

k

and the intermediate trace t isanA b ound trace variable Therefore the scrambled DO

k

reconstructs to the p osition o ccupied byt and this is why binding by the IO is imp ossible

k

even after reconstruction

To summarize so far Maha jans analysis of scrambling has the following problems

Despite the existence of multiple SPEC p ositions his system cannot accommo date

the Amovementproperty of scrambling in which an ob ject has scrambled across a

sub ject in SPEC IP cf

To capture the fact that scrambling to a p osition lower than SPEC IP do es not

license reconstruction Maha jan stipulates that adjunction is p ossible only to a max

imal pro jection higher than IP This stipulation vitiates the asso ciation of substitu

tionadjunction with AA typ e scrambling movement which had given conceptual

elegance to his system

To explain the limited distribution of reconstruction eects in examples such as

Maha jan assumes that scrambling of a DO to sentence initial p osition in a ditransitive

sentence has to b e in two steps scrambling rst to the p ostion immediately preceding

the IO Amovement and then scrambling to sentence initial p osition preceding the

sub ject A movement And he stipulates that reconstruction is p ossible only to a

variable p osition

Inow turn to data in Korean which Maha jans system cannot accommo date In sec

tion I discussed some data involving long distance scrambling out of a controlled

clause The examples are rep eated here as and

a nanun enu haksayngeykeyna PRO cakiuy immwulul cwungsilhi

Itop every studentdatuq selfgen dutyacc faithfully

halako seltukhayssta

docomp p ersuaded

I p ersuaded every student to do his duty faithfully

b nanun cakiuy immwu lul enu haksayngeykeyna PRO t cwungsilhi

j j

hatolok seltukhayssta

a nanun enu haksayngeykeyna PRO kuuy pwumolul

Itop every studentdatuq hegen parentacc

manna p okeysstako yaksokhayssta

come and visitcomp promised

I promised every student to visit his parents

b nanun kuuy pwumolul enu haksayngeykeyna PROt

j j

manna p okeysstako yaksokhayssta

The matrix dative argument which is a quantier binds the pronoun contained in the em

b edded ob ject in b oth a and a which are in the base order Scrambling the

emb edded ob ject across the matrix dative argument as in b and b results in

dierence in grammaticality Binding is retained in b reconstruction while it is

destroyed in b This contrast in grammaticalityhowever cannot b e explained in Ma

ha jans system If we supp ose that any scrambling b eyond the IP in which the scrambled

element originates is A movement then b oth examples are predicted to b e equally go o d

since the scrambled phrase can undergo reconstruction But b is ungrammatical The

ays landing site of scrambling in these examples maybeinterpreted in the following twow

Supp ose that scrambling to any p osition lower than SPEC IP is always Amovement and

scrambling to any p osition higher than SPEC IPAmovement regardless of whether it

is lo cal or long distance scrambling Then b oth examples are predicted to b e equally bad

since the scrambled elements cannot reconstruct Nevertheless b is grammatical

On the other hand my analysis which attributes reconstruction eects in scrambling to

sp ecial status of sub ject in binding adequately captures the contrast as explained in detail in

section Despite their identical conguration the crucial dierence b etween a and

a is that the emb edded sub ject in aiscontrolled by the matrix dative argument

while the emb edded sub ject in aiscontrolled by the matrix sub ject Therefore in

a the element which binds the pronoun contained in the emb edded ob ject is the

emb edded PRO sub ject rather than the matrix dative argument Since binding by a sub ject

at Dstructure retains through all levels of representation the binding is still acceptable

after scrambling as in b In a the binder is the matrix dative argument and the

binding is destroyed after scrambling

Web elhuth

Web elhuth assumes that scrambling is adjunction to either IP or to VP Noting

that scrambling in German has prop erties of b oth A and A movement he prop oses that

scrambling is movement to a mixed p osition whichisneitheranAoranAp osition

The main evidence supp orting the claim that scrambling is movement to a mixed p o

sition comes from examples in which a scrambled element exhibits prop erties of A and

A movement at the same time Consider b elow

Peter hat jeden Gast ohne e anzuschauen seinem Nachbarn t vorgestellt

i i i

Peter has every guest without tolo okat his neighbor intro duced

Peter intro duced every guest to his neighb or without lo oking at him

In the scrambled phrase jeden Gast can bind the pronoun seinem Amovement and

can license the parasitic gap in the adjunct clause A movement simultaneously On the

basis of this example Web elhuth concludes that a scrambled phrase o ccupies a third typ e of

p osition namely a mixedposition which shares the prop erties of b oth A and A p osition

The analysis of scrambling as movement to a mixed p osition makes a clear prediction

with resp ect to reconstruction That is reconstruction is always optional and therefore

scrambling can only enhance the grammaticality of the sentence in the base order How

ever this prediction is not b orne out when we consider the Korean data discussed in the

previous section Reconstruction is obligatory when the binder is a sub ject Otherwise

reconstruction is imp ossible In fact Lee and Santorini and Frank et al show

that German patterns exactly like Korean with regard to the twotyp es of binding which

I examined in this thesis Therefore Web elhuths analysis is not adequate for German

either

Aside from the problem involving reconstruction discussed ab ove my analysis shares

the following prop erties with Web elhuths

The prop erty of the landing site of long distance scrambling is no dierent from that

of lo cal scrambling

An adjoined p osition created by scrambling can have prop erties of an Ap osition

contrary to the standard assumption that adjoined p ositions are A p ositions

In fact the following quote from Web elhuth has led me to prop ose the

ve adjoined argumenthyp othesis whichderives the second prop ertyabo

The binding theory in Chomsky is stated as a theory of Abinding ie it denes

binding constraints b etween two argument p ositions Since the notion of A

binding is basic in this system it categorizes argument p ositions against all

others in particular against b oth op erator and adjoined p ositions As wehave

seen ab ove this classication of p ositions is to o coarse since it cannot express the

correct generalization that adjoined p ositions pattern with b oth The latter

fact went unnoticed probably b ecause the relevant data in its supp ort are only

available in a language with more overt adjunction than English the language

Chomskys BT was based on The theory resulting from the two assumptions

that wehave just sp elled out leaves ro om for another typ e of p osition namely

one that is neither an argument p osition nor an op erator p osition ie mixed

p osition

I b elieve that Web elhuths reasoning in the ab ove quote is right Namely Amovement

prop erties of scrambling are due to the availability of an adjoined p osition as an Ap osition

in scrambling languages rather than due to the multitude of functional pro jections

Saito

Saito notes the following facts with regard to scrambling in Japanese

Lo cal scrambling creates pronoun binding and anaphor binding Amovement

Lo cal scrambling exhibits reconstruction eects for anaphor binding and strong crossover

eects A movement

Long distance scrambling always b ehaves likeAmovement

From the b ehavior of lo cal scrambling he concludes that scrambling is to a nonop erator

A p osition the prop erties of which are identical to a mixed p osition in Web elhuth

for the present purp oses Both binding and reconstruction are p ossible from suc h a p osition

To accommo date the pure A movement prop erties of long distance scrambling he argues

that a nonop erator A p osition cannot b e licensed at LF following Tada Instead

one of the following three things has to happ en to the p osition at LF a the p osition

completely disapp ears ie reconstructs b it is reanalyzed as an op erator p osition c

it is reanalyzed into an Ap osition When a scrambled p osition is reanalyzed as an A

p osition it has to form an Achain with its trace which ob eys the generalization in

cf Saito

EachlinkofanAchain must b e sub jacent ie No barrier can intervene

between two memb ers of a single Achain

The pure A movement prop erties of long distance scrambling follow from the fact that a

long distance scrambled element out of a nite clause can never form an Achain with its

trace due to the existence of a barrier b etween the two ie the intervening CP

Saitos analysis makes exactly the same prediction as Web elhuths namely reconstruc

tion is optional and hence scrambling only improves the grammaticalityofasentence

However the fact is that reconstruction is obligatory when the binder is a sub ject Oth

erwise reconstruction is imp ossible and therefore a grammatical sentence in its base order

can b ecome ungrammatical through scrambling

Contrary to Saitos claim that long distance scrambling is always A movement Yoshimura notes

that long distance scrambling b ehaves identically to lo cal scrambling with regard to pronoun binding bya

quantier Ueyama also notes that long distance scrambling may create anaphor binding dep ending

on the prop erty of the matrix verb

In summary none of the analyses discussed here have correctly characterized the crucial

parameter for determining the reconstruction eects in binding namely the grammatical

function of the binder sub ject These analyses capture A and A movementcharacteristics

of scrambling in the language under their scrutiny dierently Maha jan prop oses that

landing site of scrambling can b e strictly divided into Ap osition and A p osition and that

A and A movement prop erties of scrambling are due to the landing site b eing an A and A

p osition resp ectively As discussed this prop osal needed two ma jor stipulations concerning

adjunction and reconstruction sites to capture the limited distribution of reconstruction

eects let alone the inadequacy of the system in accounting for Amovement prop erties of

scrambling On the other hand Web elhuth and Saito prop ose that there is a third p osition

typ e which shares prop erties of b oth A and A p osition Under such an analysis sentences

can only b e improved by scrambling New binding can always b e created by exploiting

Amovement while old binding mayalways b e retained through A movement However

the data involving binding byanobjectshow that sentences can b ecome ungrammatical by

scrambling

Chapter

Scrambling and Scop e

This chapter examines the interaction b etween scrambling and scop e interpretations of wh

phrases and negativepolarityitem amwu N any N As in raising section scrambled

elements optionally reconstruct for scop e interpretations and undergo further LFmovement

section and section Scop e reconstruction in scrambling has an imp ortantim

plication for the theory of scrambling namely that scrambling involves movement not

basegeneration

AA distinction and scop e

While A moved elements reconstruct for binding as discussed in Ch Amoved elements

reconstruct for scop e interpretation

Scop e reconstruction

As discussed in May a quantier whichismoved to an Ap osition optionally under

go es reconstruction or quantier lowering for its scop e interpretation Consider

Someone is likely t to win the game

i IP i

In the quantied NP someone may takescopeover the matrix verb which presupp oses

that there is a particular individual who is talked ab out who is likely to win the game Or

ts it may b e in the scop e of the matrix verb with no such presupp osition May accoun

for this ambiguityby assuming that in LF someone may raise and adjoin to the matrix

IP for the wide scop e reading or it may reconstruct to its Dstructure p osition and then

adjoin to the emb edded IP for the narrow scop e reading

Scop e reconstruction as a diagnostic for Amovement implies that it is distinct from

reconstruction for binding which isanAmovementcharacteristic cf Ch However

Cinque claims that the twotyp es of reconstruction pattern together BelowI

attempt to justify the view that scop e reconstruction and reconstruction for binding are

distinct drawing on Williams I also argue that the scop e ambiguityin which

can b e taken as evidence that an A moved element undergo es scop e reconstruction has an

explanation whichdoesnotinvolve reconstruction

Howmany patients do you think that every do ctor in the hospital can visit

in an hour

Scop e reconstruction vs Reconstruction for binding

Consider a and its topicalized counterpart b which are b and a in

Williams resp ectively

a I think Bob didnt see manyofmy friends

b Manyofmy friends I think Bob didnt see

According to Williams aisambiguous with regard to the interpretation of the nu

meral quantier many of my friends and the negation while there is no suchambiguityin

b If an A moved element reconstructs for scop e interpretations the absence of scop e

ambiguity in b is unexp ected Moreover the absence of scop e reconstruction eects in

bcontrasts with the presence of the reconstruction eects for binding in topicalization

as in b

a I think Bob didnt see a picture of himself

b A picture of himself I think Bob didnt see t

i i

On the basis of the contrast in reconstruction eects b etween and we ex

p ect that an A moved elementcontaining b oth an anaphor and a quantier will exhibit

reconstruction eects with regard to anaphor binding but not with regard to scop e inter

pretation This exp ectation is b orne out as illustrated in

I think Bob didnt see t Many pictures of himself

i j i j

In himself whichiscontained in the moved phrase is contrued with Bob indicating

that there is reconstruction for binding However the narrow scop e reading of the quantied

expression many pictures is not available suggesting that there is no reconstruction for

scop e

Another fact which indicates that reconstruction for scop e and binding is distinct in

volves the interpretation of the quantier each AsKroch observes the quantier

each unlike every and al l o ccurs most naturally only where there is a p otential scop e

ambiguity to b e resolved ie it must distribute over something This is illustrated by

the dierence in acceptabilityforsentences in taken from Williams ex

p

a Each patient left cf All the patients left

b Each patient saw his own chart

c Each patient saw a dierentdoctor

In b the quantier binds the pronoun whose reference varies in accordance with the

sub ject of the sentence ccon tains an indenite sub ordinate to each whose reference

varies as the sub ject varies The interesting fact ab out b and c is that they

involve fundamentally dierent relations The former involves a binding relation b etween a

quantier and a pronoun in Ap ositions while the latter involves a scop e relation b etween

two quantiers each and the indenite If reconstruction eects for scop e interpretation and

binding involve the same principlemechani sm then we exp ect that the binding relation in

b and the scop e relation in c remain the same after the same kind of movement

This exp ectation however is not b orne out as illustrated by the contrast in grammaticality

between a and b ex a and b in Williams resp ectively

a What each patient sawwas his own chart

b efore movement Each patient saw his own chart

b What each patientsawwas a dierent do ctor

b efore movement Eachpatientsaw a dierent do ctor

The pronoun binding relation is maintained in a while the quantier scop e relation is

destroyed in b

Another example which indicates that clefting A movement do es not maintain the

scop e interpretations which exists in the canonical sentence is given in taken from

ex in Carlson

one ate a tomato a Every

b It was a tomato that everyone ate

aisambiguous b etween Each p erson ate hisher tomato and A tomato is shared

by all b however has only the reading that a tomato is shared by all

The contrast in acceptabilitybetween c and b and the lack of a reading in

b whichisavailable in a suggest that scop e reconstruction and reconstruction for

binding are distinct

Reconstruction vs QR

Consider and its Italian counterpart in whichistaken from Kro ch and

Cinque resp ectively

Howmany patients do you think that every do ctor in the hospital can visit

in an hour

Quanti pazienti p ensi che ognuno dei medici riesca a visitare t

i i

howmany patients do you think that every one of the do ctors can visit

in unora

in an hour

How manypatients do you think that every one of the do ctors can visit in an hour

In b oth and themoved whphrases can have scop e either over or under the

y a family of universal quantier phrases That is the questions can b e satised either b

answers like I think that Dr Rossi can visit in one hour Dr Bianchi and so on orby

just one numb er like Only Giorgi and Longobardi explain the ambiguity

of such sentences by p ositing the optional reconstruction of the moved whphrase how

many patientsquanti pazienti to their Dstructure p osition cf Kro ch for a detailed

explanation If this account is correct it constitutes evidence against the view that scop e

reconstruction is particular to Amovement

However I argue that the scop e ambiguity in examples like and can have

an alternative account in terms of QR of the emb edded quantiers to the matrix clause

following the suggestion of AnthonyKroch pc and that the QR account is b etter moti

vated than the reconstruction account First there are some data which indirectly supp ort

the QR account They include the cases in which a quantier in an emb edded clause takes

its scop e over an insitu matrix element

Consider whichistaken from Fo dor and Sag ex p

This pro ducer b elieves that every actor in our company is to o fat to app ear

in public

According to Fo dor and Sag the universal quantier every actor in the emb edded clause

can take scop e over the matrix verb believeeven though the more natural reading is the

other way around If wechange the quantier to eachhowever the wide scop e reading

of the quantier is favored as illustrated by whichisFo dor and Sag ex

p

This pro ducer b elieves that each actor in our company is to o fat to app ear

in public

a quantier in an emb edded clause interacts with a quantier in the More imp ortantly

matrix clause as illustrated in

Someone thinks every candidate has a chance

In the wide scop e reading of every is p ossible esp ecially with parallel stress on the

two quantiers AnthonyKroch and Michael Hegarty pc Again if wechange every to

each as in the wide scop e reading of each b ecomes more prominent

Someone thinks each candidate has a chance

There are apparently problematic data for the claim that the wide scop e reading of

every in is due to QR of every not due to reconstruction of the A moved phrase how

many patients Consider the italian example taken from Cinque

Quanti pazienti p ensano che ognuno dei medici riesca a vusutarli

i i

howmany patients think that every one of the do ctors can visit them

in unora

in one hour

How manypatients think that every one of the do ctors can visit them in an hour

This is contrary to Aoun and Hornstein who argue that any nite clause is a scop e island

I am grateful to Rob ert Frank Michael Hegarty Caroline Heyco ckandAntony Kro ch for giving me

scop e judgments on various examples in this section

Angelika Kratzer in the talk given at Penn in Novemb er argues that the scop e interaction b etween

the matrix verb and the quantier in the emb edded clause in examples like and is distinct from

scop e interaction b etween two quantiers Instead the scop e ambiguity in and hinges on the

presupp osition in the mo del Even if Kratzers claim is correct scop e interaction b etween two quantier

NPs in examples like and indicates that QR from the emb edded clause to the matrix clause is

p ossible

According to Cinque and Longobardi is unambiguous The wide scop e reading of

the universal quantier ognuno is not available In which is an English counterpart

of the wide scop e reading of the emb edded quantier every heart surgeon is hardly

available

Howmany patients will say every heart surgeon in the hospital is the b est

one around

If a quantier in an emb edded clause can raise to the matrix clause as I argued for the

scop e ambiguity of and the absence of the wide scop e reading of the emb edded

quantiers in and is rather surprising The asymmetry in scop e interpretations

between and on the one hand and and on the other seems to favor

the reconstruction account However b elowIgiveMays account for suchcontrasts

which do es not involve reconstruction

Consider the examples in and

a What did everyone bring t

i i

b Who t brought everything

i i

talk to t a Who did everyone

i i

b Who t talked to everyone

i i

May observes that the a sentences with a quantied NP in sub ject p osition and

whtrace in ob ject p osition are ambiguous having either a singlequestion or family

ofquestions reading On the other hand the b sentences with a quantier in ob ject

p osition and whtrace in sub ject p osition are unambiguous having only a singlequestion

reading May analyzes the results in terms of a path theory of scop e relations Mays

account rests on the following three p oints

i A moved elements generate a path to their trace

ii Paths may not cross

iii A familyofquestions reading is p ossible for WH and Q only when Q adjoins

to the highest S in the S containing WHtrace

Given i iii a familyofquestions reading will b e p ossible in sen tences like a

and a in which Q ccommands WH in underlying form The relevant LF representation

involves no crossing paths as schematically representedin

WH Q e t

i S k S k i S

However when WH ccommands Q in underlying form as in b and b the repre

sen tation necessary for a familyofquestions reading will involve crossing paths

It is worth noting that if we use the quantier each instead of every in the wide scop e reading of

the quantier each b ecomes clear as shown in i

i How many patients think each of the heart surgeons is the b est

WH Q t e

i S k S i k S

No familyofquestions reading will thus b e p ossible for suchsentences The only well

formed LFrepresentations for and b will b e ones in which the paths are com

pletely nonintersecting

WH t Q e

i S i VP k VP k

S

The quantierwhphrase conguration in and according to Mays analysis

is represented in and that of and is represented in

WH Q e t

i S k Smatrix Semb edded k i S

e WH t Q

k VP k i Smatrix i Semb edded VP

S

In the paths for the whphrase and the quantier do not cross the quantier is ad

joined to the matrix clause to which the whphrase has moved and the familyofquestions

reading is available On the other hand in the quantier phrase is not adjoined to

the matrix clause whichcontains the trace of the moved whphrase and therefore the

familyofquestions reading is imp ossible

To summarize scop e reconstruction is indep endent of reconstruction for binding The

former is an Amovementcharacteristic and the latter an A movementcharacteristic

Sloan gives the following list of examples whichcontradict b oth the reconstruction account

and Mays account for the contrast in scop e ambiguitybetween and

a Who did everyone see t

i i

b Who do you think everyone sawt

i i

c Who do es everyone think you sawt

i i

d Who do es everyone think t sawyou

i i

e Who do you think everyone saw Mary kiss t

i i

f Who did everyone see Bills picture of t

i i

In each of these examples everyone ccommands the whtrace and therefore everyone must b e able to take

scop e over the whphrase under the reconstruction account Furthermore since everyone in each example

will b e able to adjoin to the matrix IP without forming a crossing path the wide scop e reading of the

quan tier should b e p ossible However only a and b are ambiguous and cf allow only the reading

where who has scop e over everyone Sloan explains this contrast byhyp othesizing g b elow

g A quantier can b e interpreted as wide wrt a whterm in matrix COMP if the quantier

i ccommands the whtrace and ii is within the governing category of the whtrace

Applying g to the examples af only in a and b is the quantier everyone within the governing

category for the purp ose of binding of the whtrace Sloans solution predicts the following

h Who does everyone expect to win is ambiguous

i How many patients are likely to every doctor to die is unambiguous

j In How many patients does someonemost people think that everyone saw

someone cannot havescopeover how many patientseven though everyone can

LFmovement

Another characteristic which distinguishes AmovementfromA movement is that A moved

elements at Sstructure cannot move further at LF cf Uriagereka and Lasnik Ch

Aoun et al while Amoved elements can Consider and

Q Who thinks who is likely t to b e late

i i

A Mary thinks Bill is likelytobelateandTom thinks Sue is likely to b e late

A John thinks who is likely to b e late

A John

In the whphrase who in the emb edded clause has moved to an Ap osition The only

p ossible reading of the sentence is a multiple question reading in which the emb edded wh

phrase takes scop e over the matrix clause as illustrated by the acceptable answer A

and the unacceptable answers A and A This multiple question interpretation

can b e obtained by p ositing LFmovementoftheemb edded whphrase to the matrix clause

A whphrase which o ccupies an A p osition at Sstructure however cannot undergo

LFmovement Consider and

John b oughtt Q Who t knows what

IP j i i CP j

A Mary knows what John b ought

A Mary knows that John b ought apples

In the whphrase what o ccupies SPECCP at Sstructure As illustrated by the

acceptable answer A and the unacceptable answer A the whphrase what in

the emb edded clause in cannot raise at LF to take matrix scop e contrary to the

Amoved whphrase in

To summarize an Amoved element optionally reconstructs for scop e interpretations

and undergo es further LFmovement

Scrambling of whphrases

In this section I show that scrambled phrases optionally reconstruct and undergo LF

movement just like standard Amovement

Licensing condition on whphrases

As discussed by Cho e and Kim among others in Korean a whword has two

interpretations one as an indenite NP and the other as a real whword as shown in

table

Aoun et al captures this dierence in LFraising b etween an element in Ap osition and one

in A p osition in terms of the generalizatio n stated b elow

WhR meaning whraising as opp osed to QR aects whphrases in Ap osition

whword whinterpretation indenite NP interpretation

nwukwu who someone

mwues what something

eti where somewhere

encey when sometime

enu N whichN some N

Table Interpretation of a p otential whword

For a p otential whword to b e interpreted as a whphrase it has to b e within the

scop e of a question morepheme suchasni which also licenses the sentence as a question

Consider the contrast in interpretation b etween a and b In a there is

no question morpheme and the p otential whword mwues can only b e interpreted as an

indenite quantier and the whole sentence is interpreted as declarative On the other

hand b which is identical to a except that there is question morpheme nican

be interpreted as either a whquestion or a yesnoquestion In the former the p otential

whword is interpreted as a whphrase and in the latter as an indenite quantier

a Maryka mwuesul sassta

Marynom somethingacc b oughtdec

Mary b ought somethingWhat did Mary buy

b Maryka mwuesul sassni

Marynom whatsomethingacc b oughtqm

What did Mary buyDid Mary buy something

Although a whword and a question morpheme must o ccur in the same clause at Dstructure

for the sentence to b e interpreted as a whquestion there is an exception When the matrix

verb is one of the so called bridge v erbs suchasmalha say or sayngkakha think a

question morpheme asso ciated with the matrix clause can license a whword in the emb edded

clause as illustrated in and hereafter I ignore the indenite NP interpretation

of a whword

Minswunun Youngheeka mwueslul mekesstako sayngkakhani

Minswutop Youngheenom whacc atecomp thinkqm

What do es Minswu think that Younghee ate

Minswunun Youngheeka mwueslul mekesstako malhayssni

Minswutop Youngheenom whacc atecomp saidqm

What did Minswu saythatYounghee ate

A whword which o ccurs b eyond the ccommand domain of a question morpheme at D

structure cannot b e licenced by the question morpheme as in

nwuka Minswuka sihemul poassnunci anta

dec whonom Minswunom examacc to okqm know

Who knows if Minswu to ok the exam

is the licensing condition on whwords which correctly rules in grammatical sentences

b and rules out ungrammatical sentence

Licensing condition on whphrases

Forapotential whword to b e interpreted as a whphrase it has to b e

within the ccommand domain of a question morpheme at LF

Note that is a necessary not a sucient condition since even if a whword is within

the ccommand domain of a question morpheme it can still b e interpreted as an indenite

NPasinb

Another question concerning the interpretation of a whphrase is how an op erator

variable relation is established for scop e For this I follow the standard convention that a

whphrase raises at LF to form an op eratorvariable relation with its trace In particular I

adopt Saito s prop osal and assume that a whword moves to COMP p osition which

is o ccupied by the question morpheme

Scop e reconstruction

As has b een observed by Saito Saito for Japanese long distance scrambling

of a whphrase b eyond the clause whichcontains the licensing question morpheme do es not

aect the whquestion interpretation This is illustrated by the identical interpretation of

a in the base order and binascrambled order

a nanun Minswuka mwuesul ceyil coahanunci anta

Itop Minswunom whatacc b est likeqm knowdec

I know what Minswu likes b est

ul nanun Minswuka t ceyil coahanunci anta b mwues

i i

whatacc Itop Minswunom b est likeqm knowdec

I know what Minswu likes b est

The interpretation of b indicates that the scrambled element reconstructs to its base

p osition and then moves to the COMP o ccupied by the question morpheme qmtotake

scop e over the emb edded clause Examples in also suggest that a scrambled wh

element undergo es reconstruction for scop e interpretation

On the other hand Kim argues that whphrases are no dierent from other quantiers

and hence undergo QR and are adjoined to IP or VP at LF Which prop osal I assume do es not makeany

dierence for the present purp oses

The same p oint has b een made for Japanese by Saito and Deprez

a nanun motwuka Minhoka nwukwulul coahantako

S S S

Itop allnom Minhonom whoacc likecomp

sayngkakhanunci kwungkumhata

thinkqm wonderdec

I wonder who everyone thinks that Minho likes

b Minhoka nwukwulul coahantako nanun motwukat

S i S S i

Minhonom whoacc likecomp Itop allnom

sayngkakhanunci kwungkumhata

thinkqm wonder

I wonder who everyone thinks that I like

The most deeply emb edded complement clause of a S whichcontains the whphrase

has b een scrambled to sentence initial p osition in bbeyond the ccommand domain

of the question morpheme nunciEven though its acceptability is slighly degraded b

maintains the indirect question interpretation which obtains in the base order sentence

a The scrambled clause reconstructs to its base p osition from which the whphrase

nwukwu raises to the COMP p osition o ccupied by nunci

Reconstruction of a scrambled element for scop e interpretations is optional as illustrated

by the examples in

a nenun Minswuka nwukwulul coahanunciani

youtop Minswunom whoacc likeqm knowqm

Do you know who Minswu likes

b nwukwulul nenun Minswuka t coahanunciani

i

whoacc youtop Minswunom likeqm knowqm

Who do you know Minswu likesDo you know who Minswu likes

a which is in the base order has t wo question morphemes ni which is asso ciated

with the matrix clause nunci which is asso ciatd with the emb edded clause The whword

in the emb edded clause nwukwu is licensed only by the emb edded question morpheme and

the whole sentence is interpreted as a yesno question However after scrambling of the

whword to sentence initial p osition as in b the whole sentence can b e interpreted as a

whquestion as wellasayesno question These twointerpretations can b e easily explained

by p ositing an optional reconstruction of the scrambled phrase For the yesno question

interpretation the scrambled whword reconstructs to its base p osition and then raise to

the COMP p osition of the emb edded clause For the whquestion interpretation the

b is an instance of scrambling out of a whisland As will b e discussed in detail in the next chapter

a whcomplement clause do es not constitute an island for scrambling However for an opp osing view see

Cho e

With a fo cal stress on the whphrase nwukwu I can marginally get the whquestion interpretation of

the whole sentence

I am assuming that a moved element reconstructs to its Dstructure p osition and then moves from there

to the emb edded COMP for scop e following May May and Saito However as Sabine

Iatridou pc p oints out to me an alternativeway of explaining the reconstruction eects is to directly

reconstruct to the emb edded COMP assuming that the movement takes place successive cyclically and that

the moved elementmoves through the emb edded COMP

scrambled whword raises to the COMP p osition of the matrix clause without undergoing

reconstruction This optional reconstruction we observe in scrambling with regard to wh

phrase scop e interpretation is consistent with the b ehavior of standard Amovement which

I discussed in section

Before moving to the next topic I review Watanab e s account for Japanese data

similar to b and b and argue that his analysis makes a wrong prediction for the

interpretation of the data likeb Watanab e assumes that whphrases in Japanese have

the structure shown in

DP

a

a

a

a

SPEC D

b

b

Operator D QP

whword

He calls the invisible op erator in SPEC DP the purewhoperator and the head of the

DP do es not haveany phonological content He argues that there is an obligatory invisible

whmovement at Sstructure in Japanese which is analogous to an overt whmovementin

English and by which the covert sp ecier of the DP Op moves to SPEC CP CP is

headed by the question morpheme ka The moved covert sp ecier must bind a variable to

avoid violating the ban against vacuous quantication stated in

Ban against vacuous quantication An op erator must bind a variable

Under this system lo cal scrambling of a whphrase to sentence initial p osition can b e

schematically represented as in

dp t qm whop

i IP i i

In DP is a scrambled whword t is its trace and whop isthecovert sp ecier of

i i i

DP All these three elements share the same index Rememb er that movement of the covert

i

bling sp ecier to SPEC CP is obligatory regardless of the absence or presence of scram

Now consider some Japanese examples and their schematic representations b elow These

are a and b in Watanab e in the order given The

explanation for each example is also his

dono hon o Maryga Johnga toshokankara t karidasita ka

i i

whichbookacc Marynom Johnnom libraryfrom checked out qm

siritagatteiru

wanttoknow

Mary wants to know whichbookJohnchecked out from the library

dp t ka whop

IP CP i IP CP IP i i

In and its schematic representation the whop erator can bind the trace of the

scrambled DP Therefore the sentence is relatively go o d although it is slightly degraded due

to sub jacency violation

Johnga dono hono toshokankara karidasita to

i

Johnnom whichbookacc libraryfrom checkedout comp

Maryga minnaga t omotteiru ka siritagatteiru

i

Marynom allnom think qm wanttoknow

Mary wants to know whichbookeveryb o dy thinks that John checked out

from the library

dp t ka whop

IP CP i j IP CP IP j i

In and its representation nothing is b ound by the whop erator and hence the

ban against vacuous quantication is violated That is why is worse than

Maryga t yonda to sono hono Johnga t itta

i j i IP j

Marynom read comp that b o okacc Johnnom said

John said that Mary read that b o ok

t dp t

IP CP i j IP i IP j

In the nonwh scrambled phrase sono hon which is representedasDP in

i

do es not bind its own trace According to Watanab e this is a violation of the ban against

vacuous quantication since he assumes that scrambling is A movement and the sentence

is unacceptable

A question arises with regard to the contrast in grammaticalitybetween and

That is why is b etter than despite the fact that b oth of them equally violate

the ban against vacuous quantication To account for this contrast Watanab e stipulates

the following The whop erator and the trace of the scrambled phrase in forms a

nonreal chain while the scrambled phrase and its trace in form a real chain And

then he prop oses

Ban against vacuous quantication

The head of a non trivial A chain must bind something

Proviso The violation counts less signicantly for nonreal chains

This solution however leaves it completely unexplained why there is such a distinction

between a chain for a whop erator and a chain for scrambling Rather the contrast in

acceptabilitybetween and seems to b e due to dierence in the degree of the

pro cessing dicutly of the twosentences

More imp ortantly his analysis cannot account for the ambiguity of the sentences such

as b Consider a and b which are due to Naoki Fukui pc which are

Japanese counterparts to a and b resp ectively

a anatawaTaro oga nanio katta ka sitteimasuka

youtop Taro onom whatacc b ought qm knowqm

Do you knowwhatTaro b ought

b nani o anatawaTaro oga t katta ka sitteimasuka

i i

whatacc youtop Taro onom b ought qm knowqm

Do you knowwhatTaro b oughtWhat do you know whether Taro b ought

Just as in Korean in a the whphrase nani which is in situ takes scop e only over

the emb edded clause After scrambling of the whphrase to sentence initial p osition as in

b the whphrase takes either the emb edded clause or the matrix clause scop e even

though the latter interpretation is slightly weak as indicated by the question mark

The representations of a and b under Watanab es system are given in

a anatawaTaro oga nani o katta ka whop sitteimasuka

i i

b nani o Taro oga t katta ka whop sitteimasuka

i i i

For Watanab e the overt whphrases in a and b are coindexed with the covert

whop erators whichhavemoved to the emb edded SPEC CP Since the scop e of overt

whphrase is the domain of the covert whop erator the scrambled whphrase in bis

predicted not to take matrix clause scop e there is no whop erator in the matrix SPEC

CP which is coindexed with it Nevertheless the scrambled whphrase takes scop e over the

matrix clause as well as over the emb edded clause contrary to the prediction

To summarize the data suchasb and b suggest that the scop e of a whphrase

is determined by the lo cation of a question morpheme rather than by the covert whop erator

whichWatanab e prop oses

LFmovement of scrambled whphrases

Besides undergoing optional reconstruction a scrambled elementcanmove at LF Consider

rep eated here as and its scrambled counterpart

Minswunun Y oungheeka mwueslul mekesstako sayngkakhani

Minswutop Youngheenom whatacc atecomp thinkqm

What do es Minswu think that Younghee ate

Minswunun mwues lul Youngheeka t mekesstakosayngkakhani

i i

Minswutop whatacc Youngheenom atecomp thinkqm

What do es Minswu think that Younghee ate

In the whphrase mwues in the emb edded clause takes scop e over the matrix clause

This fact can b e explained by assuming that the insitu whphrase raises to the matrix

COMP at LF Scrambling of the whphrase as in do es not aect the wide scop e

interpretation of the whphrase If scrambling is A movement then the wide scop e in

terpretation of the scrambled whphrase is unexp ected since an elementinanAp osition

cannot move further at LF as discussed in section More examples are given b elow

nunciani nwuka Minhoka etten wuntongul coaha

whonom Minhonom which sp ortacc likeqm knowqm

Who knows which sp ort Minho likes

In out of the two whphrases only the one in the matrix clause nwukwu takes the

matrix scop e as illustrated by a p ossible answerasin and an imp ossible answer as

in

Youngheeka Minhoka etten wuntongul cohahanunci ala

Youngheenom Minhonom whatacc likeqm knowdec

Younghee knows what Minho likes

Youngheeka tennislul Minhoka coahanunci ala

Youngheenom tennisacc Minhonom likeqm knowdec

Younghee knows whether Minho likes tennis

is a scrambled counterpart of Incontrast with the scrambled whphrase

etten wuntong can take either matrix or emb edded scop e Hence either or can

b e the answer to the question The availability of the wide scop e reading of the scrambled

whphrase indicates that a scrambled elementcanmoveatLF

cohahanunci ani nwuka etten wuntongul Minhoka

whonom whichsportacc Minhonom likeqm knowqm

Scrambling of NegativePolarityItems

Interpretation of a scrambled negative p olarity item NPI suggests that a scrambled ele

ment reconstructs for scop e purp oses Here I use the term scop e of NPIs in a nonstandard

sense In the case of quantiers and whphrases which are standard scop eb earing elements

they take scop e over other elements However for NPIs at issue it is not the case that

NPIs take scop e over other elements but that they have to b e within the scop e of negative

op erators That is scop e reconstruction of NPIs is to refer to the fact that NPIs reconstruct

to b e within the scop e of a negative morpheme not to the fact that they reconstruct to

take scop e over other elements

Licensing of NPIs

For an NPI to b e licensed in Korean it has to b e within the ccommand domain of a

trigger such as negation cf Ladusaw Linebarger Laka similarly to the

licensing of whphrases discussed in the previous section Consider the examples b elow

a amwuto Minswulul coahaci annunta

anyone Minswuacc likenmz negdo

not like Minswu lit Anyone do es

b Minswuka amwuto coahaci annunta

Minswunom anyone likenmz negdo

Minswu do es not like anyone

The whphrase in the emb edded clause marginally takes the matrix scop e with a fo cul stress on it

though

The p osition o ccupied by the scrambled whphrase in is ambiguous It can b e either within or

beyond the emb edded clause b oundary Since I am assuming that a whphrase moves to the closest COMP

for scop e based on the facts in baseorder sentences the scrambled phrase is to b e analyzed as b eing in a

p osition b eyond the emb edded clause for the matrix scop e interpretation

Lee discusses negative p olarity items in Korean in more detail

a amwuto Minswulul coahanta

anyone Minswuacc like

b Minswuka amwuto coahanta

Minswunom anyone like

ab illustrate that NPIs can o ccur in sub ject as well as in ob ject p osition ab

show that an NPI cannot b e licensed without its trigger An NPI which is generated outside

the ccommand domain of a p otential licenser at Dstructure results in an ungrammatical

sentence as in

amwuto Youngheeka Minswulul coahaci annuntako sayngkakhanta

anyone Youngheenom Minswuacc like negdocomp think

lit Anyone thinks that Younghee do es not like Minswu

A licensing condition on NPI which is relevant for the present purp ose is stated in

Licensing condition on NPI

An NPI must b e ccommanded by its trigger within a CP which contains

b oth at Dstructure

Reconstruction of a scrambled NPI

As in the case of whphrases scrambling of an NPI b eyond the ccommand domain of its

tence licenser do es not aect the grammaticality of the sen

amwuto nanun t Minswulul coahaci annuntako sayngkakhanta

i i

anyone Itop Minswuacc likenmz negdocomp think

lit Anyone I think t do es not like Minswu

i i

amwuto nanun Minswuka t coahaci annuntako sayngkakhanta

i i

anyone Itop Minswunom like negdocomp think

lit Anyone I think Minswu do es not liket

i i

In and the NPI whichistheemb edded sub ject and ob ject resp ectively has b een

longdistance scrambled b eyond the ccommand domain of the emb edded clause negation

Nevertheless the sentences receive the same interpretation as the one in the baseorder

bled Grammaticality of and and their interpretations indicate that the scram

NPI reconstructs to its Dstructure p osition for its licensing

Reconstruction of a scrambled NPI is optional Consider the examples in

ananun Maryka amwuto miwehantako sayngkakhaci annunta

Itop Marynom anyone hatecomp think negdo

lit I dont think Mary hates anyone

b amwuto nanun Maryka t miwehantako sayngkakhaci annunta

i i

anyone Itop Marynom hatecomp think negdo

Many p eople nd and to b e marginal compared to their unscrambled counterparts This is

probably due to the following reason amwuto has b oth existential indenite and presupp ositional reading

in its base p osition As I will argue in section of Ch only a presupp osed element can undergo

scrambling That is the only reading available after scrambling is the one in which the NPI is presupp osed

Hence those who try to get the indenite reading for and would nd them to b e unacceptable

Nevertheless they are p erfectly acceptable under the presupp ositio nal reading

In a the NPI in the emb edded clause cannot b e licensed by the negation in the

matrix clause since they do not meet condition Scrambling of the NPI to sentence

initial p osition as in b however enables the NPI to b e licensed by the matrix clause

negation The creation of NPI licensing byscrambling we observeinb indicates that

reconstruction is optional Otherwise the sentence should remain ungrammatical

To summarize facts concerning NPI licensing suggest that scrambled elements option

ally reconstruct for scop e interpretations which is consistent with the b ehavior of typical

Amovement

Implications

Scrambling as movement

It has b een controversial whether scrambling involves movement or is basegenerated Scop e

reconstruction facts discussed in this chapter are imp ortant in this regard since they are clear

evidence that long distance scrambling involves movement The question then is whether

or not lo cal scrambling is the same syntactic phenomenon as long distance scrambling

Although this issue is not easy to settle there are some arguments in favor of the same

analysis for b oth typ es of scrambling First in Ch I showed that b oth typ es of scrambling

b ehave the same with regard to binding Second as I will discuss in Ch b oth lo cal and

long distance scrambling are sub ject to the same discourse conditions That is only a

presupp osed element can undergo scrambling In the absence of evidence to the contrary

and given the two facts mentioned ab ove it seems reasonable to conclude that lo cal and

longdistance scrambling are the same syntactic phenomenon and therefore lo cal as well as

longdistance scrambling involves movement

Reconstruction for binding and scop e

In Ch I argued that reconstruction eects with regard to binding in scrambling are due to

the sp ecial status of sub jects in binding while there is no such restriction for reconstruction

with regard to scop e interpretation involving whphrases and negative p olarity items This

leads to the conclusion that reconstruction for binding is indep endent of reconstruction for

scop e interpretation in scrambling

Chapter

Scrambling as casedriven

Amovement

In this chapter I prop ose that scrambling in Korean is casedriven obligatory movement

This is consistent with the Amovement prop erties of scrambling discussed in the previous

twochapters and the economy principle under whichmovement is considered as the

last resort This prop osal crucially diers from the widely accepted view that scrambling

is optional cf Fukui

In section I establish that scrambling is a consequence of casedriven obligatory

movement based on an analysis of case and word order p ossibilities in event nominal clauses

In section I prop ose the Case Assignment Rule which incorp orates the notion of

index to case assignment to ensure that each argument is assigned the right kind of case

eg sub ject is assigned nominative case and ob ject accusative case not vice versa in a

transitiveverb sentence In section I discuss apparent problems to my prop osal In

section I argue that longdistance scrambling is casedriven just like lo cal scrambling

and examine some p otential problems for this prop osal Finally in section I discuss

the implications of the current prop osal on the theory of scrambling

Scrambling as casedriven adjunction to IP

Case and word order p ossibiliti es in event nominal clauses to b e discussed in section

indicate that nominative case is assigned by a functional head and accusative case bya

complex category consisting of a lexical head with feature stative and a functional head

erbraising to infl Assuming the VPinternal sub ject hyp othesis all whichisformedbyv

arguments havetomove out of VP and are adjoined to IP to b e assigned case Scrambling

is due to the fact that arguments may b e arranged in any order for the purp ose of case

assignment since b oth nominative and accusative case assigners are in the same p osition

after verb raising whichismotivated by accusative case licensing

Case and word order p ossibilities in event nominal clauses

Some nouns typically event nouns have their own argument structure and arguments

o ccurring in an NP can b e marked only genitive in Korean as illustrated in

a Kim kyoswuuy wencahaykuy yenkwu

Kim professorgen nuclear weap ongen research

Prof Kims researchonnuclear weap ons

b Kim kyoswuka wencahaykuy yenkwu

Kim professornom nuclear weap ongen research

c Kim kyoswuka wencahaykul yenkwu

Kim professornom nuclear weap onacc research

a Kim kyoswuuy wencahaykuy yenkwuhanta

Kim professorgen nuclear weap ongen do research

Prof Kim do es researchonnuclear weap ons

b Kim kyoswuka wencahaykuy yenkwuhanta

Kim professornom nuclear weap ongen do research

c Kim kyoswuka wencahaykul yenkwuhan ta

Kim professornom nuclear weap onacc do research

The head of the examples in is the eventnounyenkwu research while the lexical

head of the examples in is the verb yenkwuhata to research The main dierence

between the two cases is that in all the arguments havetobemarked with genitive

case while in they havetobemarked with verbal case

As discussed by Iida Shibatani and Kageyama Sells Miyagawa

for Japanese and Cho and Sells for Korean when the eventnounisfollowed byan

asp ect morpheme suchastocwung during cikhwu right after tangsi when etc

the arguments exhibit additional case p ossibilities as illustrated in

By verbal case I refer to nominative and accusative case as opp osed to genitive case

Strings relevant for the present discussion are indicated by square brackets The matrix clause cencayngi

ilenassta a war to ok place is added to show that the sub ject of the nominal clause Kim kyoswu in this

case do es not have to coincide with the sub ject of the matrix clause cencayng

At rst glance an eventnounfollowed by an asp ect morpheme is analogous to the combination of an

event noun plus lightverb hata ie lightverb construction However there is a crucial dierence b etween

the two namely no particles can intervene b etween an event noun and an asp ect morpheme cf a while

various particles can freely intervene b etween an event noun and the lightverb as noted in Sells

cf b

a yenkwulultoman hwu

researchaccevenonly after

after evenonly research

b yenkwulultoman hata

researchaccevenonly do

do evenonly research

Furthermore an adverb and a nonevent noun can combine with the lightverb as in c while they cannot

combine with an asp ect markerasind

c keyullichwung sil hi yokhata

negligentlyfai thful ly curs edo

neglectdo faithfullycu rse

dkeyullichwun gsi lh i yokhwu

negligentlyfai thful l ycu rseafter

a Kim kyoswuuy wencahaykuy yenkwucwung cencayngi ilenassta

Kim Profgen nuclear weap ongen researchduring warnom to ok place

During Prof Kims research on nuclear weapons a war to ok place

b Kim kyoswukawencahaykuy yenkwucwung cencayngi ilenassta

Kim Profnom nuclear weap ongen researchduring

c Kim hyoswukawencahaykul yenkwucwung cencayngi ilenassta

Kim Profnom nuclear weap onacc researchduring

d Kim kyoswuuy wencahaykul yenkwucwung cencayngi ilenassta

Kim Profgen nuclear weap onacc researchduring

The arguments may all b e marked with genitive case as in a or all with verbal case

as in c Also the sub ject can b e marked with nominative case and the ob ject with

The combination shown in d however genitive case mixed case array as in b

is totally unacceptable where the sub ject is marked genitive and the ob ject accusative in

that order

Dep ending on the kind of case an argument b ears the arguments exhibit dierent de

grees of word order freedom Only the p ermutation of nominative sub ject and accusative

ob ject is allowed as in c

awencahaykuy Kim kyoswuuy yenkwucwung

nuclear weap ongen Prof Kimgen researchduring

bwencahaykuy Kim kyoswuka yenkwucwung

nuclear weap ongen Prof Kimnom researchduring

c wencahaykul Kim kyoswuka yenkwucwung

nuclear weap onacc Prof Kimnom researchduring

dwencahaykul Kim kyoswuuy yenkwucwung

nuclear weap onacc Prof Kimnom researchduring

An analysis

The key to the analysis of the data describ ed ab ove is to come up with an adequate case

licensing condition A case licensing condition which accommo dates the whole range of the

data can b e informally stated as in

A mixed case array in general sounds rather marginal compared to a purely verbal or a purely nominal

case arrayasJamesYo on pc p oints out

An anonymous reviewer of Language Research judged b oth b and c to b e marginal marking

them with This indicates that the judgment of the data is sub ject to individ ual variation Nevertheless

p eople seem to agree on the contrast b etween abc on the one hand and d on the other and my

goal is to account for this contrast

a Genitive case is licensed by the event noun

b Accusative case is licensed by the complex category resulting from head

movementoftheevent noun to the p osition of the asp ect morpheme

c Nominative case is licensed by the asp ect morpheme

The o ccurrence of genitive case with a bare head noun as in a suggests that geni

tive case is licensed by the head noun The contrast in grammaticalitybetween examples

bc and bc indicates that the presence of an asp ect morpheme is crucial for b oth

nominative and accusative case licensing However the ungrammaticalityofd in which

the ob ject is marked accusative in the presence of the asp ect morpheme suggests that the

mere presence of an asp ect morpheme is not sucient for accusative case licensing and

requires the accusative case licensing condition in b

In giving a precise formulation of case licensing condition in I assume that lexical

categories have feature F and functional categories F following Grimshaw In

addition I assume the following

a An asp ect morpheme is a functional head with feature F and has an

indep endent pro jection in the phrase structure

b An event noun has feature stative

Incorp orating the ab ove assumptions into the informal case licensing condition we

have the case licensing condition in

a Genitive case is licensed via head government byanX category with

feature N V

b Accusative case is licensed via head government byacomplexX category

with feature stativeF

c Nominative case is licensed via head gov ernment byanX category with

feature F

The category with feature N V in a is a noun the event noun in the present

discussion Feature F in c comes from either an asp ect or a tense morpheme in

the case of verbal clauses A complex category with feature stative F in b is formed

by the combination of an event noun with feature stative and an asp ect morpheme with

feature F Head government and its related notion relativized minimality are dened as

in and

Head Government X headgoverns Y i

i X fANPVINFLAspg

ii X mcommands Y

iii No barrier intervenes

iv Relativized Minimality is resp ected

Relativized MinimalityXgoverns Y only if there is no Z such that

iZisatypical p otential governor for Y

ii Z ccommands Y and do es not ccommand X

Asp F N V

a

a

a

a

NP AspF N V

H

H

H

NPGEN N cwung

P

P

P

P

NPGEN NN V

Figure Derivation of GENGEN combination

AspPF N V

X

X

X

X

X

NP NOM Asp F N V

i

P

P

P

P

AspF N V NP

P

P

P

P

P

N NP cwung

P

P

P

P

NPGEN NN V t

i

yenkwu

Figure Derivation of NOMGEN combination

Case p ossibilities of the examples in can b e explained in the following way in terms

of the case licensing condition in

In Figure the genitive case on b oth the sub ject and the ob ject is licensed bythehead

noun and the asp ect morpheme do es not participate in case licensing at all

In Figure the genitive case on the ob ject is licensed by the head noun and the nomi

nativecaseonthesubjectby the asp ect morpheme after the sub ject moves to the p osition

AspPadjoined p osition in this case governed by the asp ect morpheme with feature F

In Figure a complex category with feature F stative needs to b e formed for accusative

case licensing and the only way to form this complex category is via head movementof

the event noun to the p osition o ccupied by the asp ect morpheme After head movement

the ob ject moves out of NP to b e assigned accusativecaseby the newly formed complex

category AspF stative The sub ject moves out of NP to b e assigned nominative case by

the asp ect morpheme which happ ens to have a complex feature F stative as a result of

head movement

In Grimshaw not only lexical but also functional categories have categorial feature sp ecication

with regard to NV and INFL has feature N V If we assume that Asp also has feature N V then

an event noun which Iassumetohave feature N V cannot form a legitimate extended pro jection with

Asp due to their conicting features One way of making the current system compatible with the extended

pro jection system in Grimshaw is assume that an event noun is category neutral with resp ect to NV

and therefore can combine with either a N V or a V V category as Grimshaw herself suggests to

AspPF N V

X

X

X

X

X

NP NOM AspPF N V

i

X

X

X

X

X

NP ACC Asp F N V

j

X

X

X

X

X

NP AspF N V

b b

b b

b b

NP N Asp N

k

c

c

t NP yenkwu cwung N

i

t t

j k

Figure Derivation of NOMACC combination

Asp F N V

a

a

a

a

NP AspF N V

H

H

H

NPGEN N cwung

P

P

P

P

NPACC NN V

Figure Derivation of GENACC combination

Figure is a representation of the genacc combination which is ungrammatical The

ungrammaticality is due to the conicting demands on the p osition of the head noun for

genitive and accusative case licensing For the genitive case licensing on the sub ject the

head noun has to stay in situ while for the accusative case licensing on the ob ject it has

to move to the p osition of the asp ect morpheme to form a complex category with feature

stative F Since these two conicting demands cannot b e satised simultaneously the

string is ruled out

The characteristics of my analysis are summarized as follows First head movementis

a necessary condition for accusative case licensing which requires a complex category with

feature F stative Second after head movement of the head noun b oth the nominative

case licenser with feature F and the accusative case licenser with feature F stative are

in the same p osition This leads us to exp ect that a nominative sub ject and an accusative

ob ject can b e arranged in any order for the purp ose of case licensing This exp ectation

is met as evidenced by the grammaticality of the minimal pair strings c and c

In this analysis scrambling is a consequence of casedriven movement and therefore it is

predicted that if there is no casedriven movement there is no scrambling either This

account for lightverb constructions in Korean and Japanese

prediction is indeed b orne out as evidenced bya in which the order p ermutation

of the two arguments marked with genitive case results in ungrammaticality Genitive

case is licensed by the head noun without the arguments having to move out of the NP

Hence there is no way to get the p ermuted word order Finally an implicit assumption

in my analysis is that when there is movement it is the moved element the head of the

chain not its trace the tail of the chain that is resp onsible for case licensing This

assumption has an interesting consequence on scrambling when combined with the prop osal

by Heyco ck and Kro ch namelyany licensing relation satised by the head of a chain

at Sstructure cannot in addition license a trace and the trace has to delete unless it is

indep endently licensed I will discuss this consequence in detail in section

Extension to verbal clauses

The case licensing conditions and the analysis of scrambling given in the previous section

can b e easily extended to verbal clauses ie a clause the lexical head of whichisaverb

as opp osed to a noun As I have b een assuming all along under the VPinternal sub ject

hyp othesis the Dstructure representation of a verbal clause lo oks roughly like Figure

except that the no de lab els Asp NPN and N are replaced by INFL VPV and V

resp ectively abstracting away from other unsp ecied pro jections suchasCP Assuming the

case licensing condition in averb has to raise to INFL to form a complex category

with feature stative F feature stative is due to the verb and feature F due to

e out of VP to b e INFL for accusative case licensing Both the sub ject and the ob ject mov

assigned case resulting in a representation like Figure

Since b oth the nominative case licensor INFLF and the accusative case licensor INFL

stative F are in the same p osition the sub ject and the ob ject may b e arranged in any

order giving rise to scrambling eects A question arises concerning how to ensure the

sub ject is assigned nominative case and the ob ject accusative case and not vice versa I

address this question and prop ose a solution for it in section

In Lee b I argued that scrambling among genitive phrases is p ossible assuming basegeneration of

arguments in any order I ascrib ed the marginality of to the antiambiguity condition on scrambling

which will b e discussed at length in the next section I supp orted my argumentby giving examples such

as i where scrambling among genitive phrases is slightly more acceptable than The only dierence

between and i is that the adverbial phrase mikwukeyseuy is added in i

i wencahayk uy Kim kyoswuuymikwukeyseuy t yenkwucwung

i i

nuclear weap ongen Prof Kimgen Americalocgen researchduring

during Prof Kims researchonnuclear weap ons in America

However I retract that for the following reasons The nature of unacceptabili ty of is somewhat dierent

from that caused by the antiambiguity constraint in the sense that unacceptabil ity of an example which

violates the antiambiguity constraint can b e improved signicantly if a prop er discousre context is provided

while the unacceptabili ty of examples such as remains prettymuch constant in almost anycontext

y a long intonation break b etween each genitive The improved acceptability ofiseemstobeachieved b

marked phrase If there is no intonation break the sentence sounds signicantly degraded

Assuming that a ro ot clause in Korean is a CP as Cho e argues a question arises whether a verb

raises all the way up to COMP or stops at INFL As far as my analysis on case licensing is concerned a

verb has only to raise to INFL However taking up Cho e s prop osal Whitman argues that a

verb raises to COMP on a par with verb raising to COMP in German even though the two languages dier

in that verb raising in German is visible while in Korean it is string vacuous Even if we assume that a verb

IP

H

H

H

NP NOM IP

i

H

H

H

NP ACC I

j

a

a

a

a

VP IF N V

b

b

b

NP V V I

k

c

c

t NP V

i

t t

j k

Figure Sstructure representation of a verbal clause

Comparison with Miyagawa

Miyagawa also argues that scrambling is closely related to the existence of verb

raising on the basis of case and word order p ossibiliti es in Japanese which are identical to

the Korean data discussed in section In this section I compare Miyagawas system

to the one I prop ose here

As far as genitive and nominative case licensing is concerned there is no dierence

between Miyagawas and my analysis The two analyses diverge in the way accusative case

licensing is done and in the role of a trace in case licensing Miyagawa assumes the accusative

case licensing condition stated in and the GovernmentTransparency Corollary which

isawayofallowing a trace the tail of a chain to participate in case licensing

Accusative case is licensed bytwo steps

a Case feature assignmentbyaACC nounverb at Dstructure

b Case realization at Sstructure via governmentby a functional head

Asp INFL

GovernmentTransparency Corollary Baker

A lexical category which has an item incorp orated into it governs everything

which the incorp orated item governed in its original structural p osition

Under and obligatory verb raising to INFL which Miyagawa also assumes an

ob ject may b e assigned accusative case either in its Dstructure p osition or in IPadjoined

p osition An ob ject is assigned the ACC feature in its Dstructure p osition cf a If

the ob ject do es not move out of VP accusative case is realized via GovernmentTransparency

Corollary the verb raises to INFL and INFL can then govern the ob ject p osition If

the ob ject moves out of VP and is adjoined to IP accusative case is realized via direct

governmentby INFL In this system scrambling is due to the fact that an ob ject can b e

assigned accusative case either in its Dstructure p osition or in IPadjoined p osition

raises to COMP the current analysis of case licensing and scrambling can still b e maintained

Similarities and dierences b etween Miyagawas and my system can b e summarized as

follows First b oth systems assume obligatory verb raising to INFL In my system it is

motivated by accusative case licensing while in Miyagawas system there is no obvious

motivation for it Second scrambling is a consequence of obligatory casedriven movement

in my system while it is a consequence of the optionality ofobjectmovement for accusative

case realization Third in my system only the head of a chain licenses case while in

Miyagawas either the head or the tail of a chain can license case via GovernmentTrans

parency Corollary Finallyinmy system the case licensing condition in is enough

to account for the whole range of data in section while the case licensing condition

whichMiyagawa assumes cannot accommo date the same range of data In particular the

ungrammaticalityofd rep eated here as cannot b e explained by Miyagawas

case licensing condition alone

wencahayk ul Kim kyoswuuy t yenkwucwung

i i

nuclear weap onacc Prof Kimnom researchduring

during Prof Kims researchonnuclear weap ons

The genitive case on the sub ject is assigned by the head noun in situ For the accusative

case on the scrambled ob ject ACC is assigned by the head noun at Dstructure and

verned by Asp at Sstructure Namely the the accusative case can b e realized by b eing go

string is predicted to b e go o d in Miyagawas system contrary to the fact and therefore

Miyagawa needs an extra mechanism to rule out such a string

Implications

I discuss implications of the case licensing condition and the prop osed analysis of scrambling

as a consequence of casedriven obligatory movement of arguments

Separation of case assignment from role assignment

In my analysis verbal case assignment is completely disso ciated from role assignment

roles are assigned VPinternally under the strict sisterho o d condition while verbal case

In his unpublished work Miyagawa indep endentl y argues that scrambling is a consequence of case

driven movement as I claim in this thesis This earlier view of Miyagawas is summarized Miyagawa

fn and given b elow

In an earlier version of this pap er I suggested that the ob jective case marker in Japanese is

realized outside of VPby adjoining to the pro jection of I This has the advantage that we

can simplify the statement for the environmentinwhich Case realization is licensed Case

nominative and accusative may b e realized if it o ccurs in a p osition immediately dominated

by the pro jection of a functional category suchasInAspScrambling then is simply an

instance in which the ob ject NP moves in front of the sub ject NP to realize Case instead of

the p osition after the sub ject NP This analysis also allows us to suggest the following gener

alization regarding overt and Abstract case

Overt case marker must b e realized outside of VP directly dominated by pro jection of I

while Abstract case must b e realized within VP

While I b elieve that this analysis is promising it also has a numb er of conceptual problems

and I will not pursue it in this pap er

is assigned VPexternally in any order as long as case licensing conditions are met As

Miyagawa has already noted for Japanese this dierence b etween role and case

assignment is resp onsible for b oth the congurational and the noncongurational asp ects

of Korean It is congurational with regard to role assignment It is noncongurational

with resp ect to case assignment which results in scrambling

Adjoined arguments

Under case licensing condition Case can b e assigned to an IPadjoined p osition as

well as a SPEC IP p osition since either p osition can b e governed by INFL In fact a

SPEC p osition do es not enjoy a sp ecial status compared to an adjoined p osition The

nondistinctness of an adjoined p osition from a SPEC p osition in Korean is b oth empirically

and theoretically wellmotivated

Empirically as will b e discussed in detail in Ch accusativenominative case can b e

assigned to timeplace adverbial as well as to an argument in Korean And case assignment

to an adverbial is sub ject to exactly the same case licensing condition as that to an argu

ment Assuming the standard view that timeplace adverbials are adjoined to an IP or

I case assignment to an IPadjoined argument is exp ected Theoretically Ho ekstra

argues that there is no need to distinguish b etween adjuncts and sp eciers indep endently

of agreement rather a sp ecier is an adjunct that agrees with the head If we assume that

Ho ekstra is right and that Korean do es not have an agreement of a pronominal nature

all sp eciers in Korean are indep endent of agreement and therefore there is no distinction

between adjuncts and sp eciers

The mismatchbetween case and role assignment has b een implicit in the case of exceptional case mark

ing ECM Case assignment is completely disso ciated from role assignment in the Minimalist Framework

prop osed in Chomsky

Note that there is an asymmetry b etween a nominal not followed by a functional category with feature

V N and a verbal clause with regard to case assignment and role assignment In a nominal clause the

head noun has the ability to assign b oth role and case while in a verbal clause the head verb has only role

assigning abilityAnthony Kro ch pc suggests that it is unlikely that there is such an asymmetry b etween

verbs and nouns and that role and case assignments are completely disso ciated in b oth cases The idea

can b e instantiated along the following lines There is an abstract category pro jection which constitutes the

core argument structure which is common in nominal and verbal pro jections and is lower in the hierarchy

than the pro jections of N and V roles are assigned by the head and arguments raise out of their argument

structure to b e assigned genitivecaseby a noun or accusativecasebyaverb maintaing the traditional idea

that accusative case is assigned bya verb not bythecombination of a verb and a functional category This

alternative seems to b e b oth conceptually more elegant and to have a b etter p otential to cover the wider

range of data o ccurring in Japanese including case p ossibili ti es in purp osive expressions which are discussed

in Sells and illustrated b elow

wa kyonen itta i Johnga Americani eigoo BENKYOOsi ni

Johnnom Americato Englishacc studydovstem purp osive top last year went

John went to America last year in order to study English

ii Johnga Americani eigoo BENKYOO ni wa kyonen itta

Johnnom Americato Englishacc study purp osive top last year went

In ii the arguments exhibit verbal case array despite the fact that there is no clear functional category

of a verbal nature A way of analyzing this data consistentwithAnthony Kro chs suggestion is to assume

the abstract argument structure and an abstract verbal pro jection which is resp onsible for verbal case

assignment Under my system an obvious way of accommo dating such data is to assume that the purp osive

particle ni is a functional category with feature F V N just like other asp ect morphemes discussed in

this chapter

Obligatoriness of scrambling

In my system scrambling is a consequence of casedriven obligatory movement contrary to

the widely accepted view that scrambling is optional Obligatoriness of scrambling is a wel

come result under the economy principle outlined in Chomsky and Chomsky

The basic economyofderivation assumption is that op erations are driven by necessity

they are last resort applied if they must b e not otherwise

The prop erties of scrambling discussed in Ch along with other recent studies of scram

bling cf Web elhuth Deprez Maha jan Miyagawa Saito

Homan and Turan further supp ort the view that scrambling is forced As Fukui

argues under the economy approach optional movement will b e p ermissible only if its ap

plication is costless Alternatively optional movement will not aect the interpretation

ofasentence if we reasonably assume that a change in interpretation is accompanied by

a certain cost However as discussed in Ch scrambling aects binding relations and

consequently the interpretation of the sentence leading to the conclusion that scrambling

cannot b e optional

Thetaindex and case assignment

The case lincensing mechanism which I prop osed in the previous section leaves the question

of how to ensure that each argument is assigned appropriate case ie sub ject is assigned

nominative case and ob ject accusative case not vice versa in an unergative transitive

I argue that case assignmentissensitivetothe role which an argument verb sentence

carries Before I prop ose a revised case licensing condition I sketch the factors involved

in case licensing in Korean and intro duce the notion of indexing

Prop erty of the predicate and hierarchy

There are two imp ortant factors involved in case assignment in Korean ie the category of

the predicate eg verbs adjectives and the hierarchyofthe role carried by an argument

Adjectives vs Verbs

As Kim notes adjectives in Korean unlike in languages like English do not app ear

under a higher copula verb Instead they are directly inected for tense asp ect and mo dal

itycfNavaho Anderson Mohawk Postal and Japanese Kuno In

this resp ect adjectives are almost indistinguishable from verbs in Korean However there

are two crucial dierences b etween adjectives and verbs First the morpheme nun which

is identied as the present tense marker can only b e suxed to verbs not to adjectives as

shown in and

I am grateful to James Yo on for directing my attention to this issue

As will b ecome clear later this should not b e confounded with the claim that it involves inherent case

assignment

Adjective

a Minhoka ttokttokhata

Minhonom smartdec

Minho is smart

b Minhoka ttokttokhanta

Minhonom smartpresdec

Verb

a Minhoka wusta

Minhonom laughdec

Minho laughs

b Minhoka wusnunta

Minhonom laughpresdec

Second adjectives do not have accusative case assigning ability regardless of the number

of arguments they select for while verbs do have such an ability For instance all the

arguments of an adjectival predicate mwusepta to b e afraid of are marked nominative as

in while the ob ject of a verbal predicate alta to know is marked accusative as in

Minhoka holangikalul mwusepta

Minhonom tigernomacc b e afraid of

Minho is afraid of a tiger

Minhoka Marylulk a anta

Minhonom Maryaccnom know

Minho knows Mary

Following the tradition in Korean and Japanese linguistics cf Kuno b Kang

I call adjectives stative and verbs stative predicates However I would like to p oint

out that the stative predicate distinction I assume here should b e distinguished from the

states and activities predicate distinction in Dowty

As discussed in section a noun also can b ear the feature stative eg event nouns suchasyenkwu

research are stative and result nouns suchaschayksang desk are stative

The following illustrates some of the states and activities predicate classication whichDowty

prop oses

I StatesStatives

A Intransitive Adjectives

With individ ual s as sub jects be tal l big green American quadrilateral

With prop osition s as sub jects be true false likely doubtful

C TransitiveandTwoplace phrasal adjectives

like similar identical relatedtoNP

These are the symmetric predicates of Lako and Peters

proud jealous fond of NP

IIActivities

There are at least two diagnostics by whichwe can distinguish stative predicates from

stative predicates in Korean First only stative predicates verbs can o ccur with

the progressive forming morpheme ko issta as illustrated by the grammaticality of in

which the predicate is stative and the ungrammaticality of in which the predicate

is stative cf Kim

Kimi ikosul hyanghayoko issta

Kimnom this placeacc toward comeprog

Kim is coming toward this place

Kimi yongkamhakoissta

Kimnom braveprog

Kim is b eing brave

Second while stative predicates are compatible with the present p erfect tense which

is formed bycombining the verb ro ot with the verb e ota stative predicates are not

This is illustrated by the grammaticality of containing a stative predicate and the

ungrammaticality of containing stative predicate

hankwukmintulun ssalul cwusikulo meke oassta

have eaten Koreanstop riceacc main mealinst

Koreans have eaten rice as main meal

Kimi yongkamhaye oassta

Kimnom have b een brave

Kim has b een brave

Theta hierarchy

In addition to its sensitivity to the category of the p otential caseassigner ie selecting

predicate the case assigned to an argument is sensitivetothe hierarchy among the

arguments b elonging to the same argument structure Among the arguments selected for

by the same verb ie stative predicate the argument which carries the highest role

in a hierarchy is assigned nominative case and the others accusative case The situation

never arises in which an argument with a higher role is assigned accusative case and an

argument with a lower role nominative case Examples in illustrate the imp ossible

case array of the arguments of the ditransitiveverb senmwulhata to give as a present

A Adjectives all adjectival and predicate nominal activities are volitional

Intransitive bebrave greedy

Twoplace phrasal be rude nice polite obnoxious to NP

Note that in Dowtys classication some adjectives b elong to the category of activity predicates while in

my classication they b elong to stative predicates

Dowty lists ve criteria for distingui shi ng statives from nonstatives a only nonstatives

o ccur in the progressive b only nonstatives o ccur as complements of force and persuade c only non

statives can o ccur as imp eratives d only nonstatives coo ccur with the adverbs deliberately careful ly e

only nonstatives app ear in pseudocleft constructions These criteria however are not directly applicabl e

to mystative predicate distinction except for the one on pregressive formation

a Maryka Minhoeykey chaykul senmwulhayssta

Marynom Minhodat bookacc gave as a present

Mary gave Minho a b o ok as a present

b Maryka Minholul chaykul senmwulhayssta

Marynom Minhoacc bookacc gave as a present

cMarylul Minhoeykey chayki senmwulhayssta

Marynom Minhodat booknom gave as a present

Table gives a schematic representation of p ossible case arrays according to the cat

egory of the predicate and the hierarchy of the argument

stative nom acc acc

stative nom nom nom

Table Possible case array of arguments

The numb ers at the top rowinTable represent the hierarchy of the arguments selected

by the same predicate in a decreasing order b eing the argument carrying the highest

role The rst column of the table represents the category of the predicate stative b eing

verbs and stative adjectives According to the table all the arguments selected by

an adjective stative are marked nominative and the arguments selected byaverb

stative are marked accusative except for the argument which carries the highest role

which is marked nominative

Arguments with inherent case

From my discussion on case assignment so far a question arises concerning arguments with

inherent case In particular Table do es not sayanything ab out such arguments I

argue that an argument with an inherent case is assigned structural case at Sstructure

and is sub ject to exactly the same condition as arguments with no inherent case

Sub jects exp eriencer of transitive adjectives may b e marked with dative case whichI

assume to b e an inherent case as in

Minhoeykey holangikamwusepta

Minhodat tigernom b e afraid of

Minho is afraid of a tiger

Nominative case can b e assigned to the dative casemark ed sub ject as in

Minhoeykeyka holangika mwusepta

Minhodatnom tigernom b e afraid of

Minho is afraid of a tiger

Accusative case can b e assigned to a dative casemarked ob ject in a ditransitiveverb sen

tence as in

I assume that inherent case is assigned at the same level that role is assigned and structural case

nominative and accusative is assigned at Sstructure

Maryka Minhoeykeylul chaykul senmwulhayssta

Marynom Minhodatacc bookacc gave as a present

Mary gave Minho a b o ok as a present

Genitive case can b e assigned to a dative casemarked argument as in

Kim chongcanguy swusek haksayngeykeyuy colep canguy swuye

Kim presidentgen rank one studentdatgen graduation certicategen award

President Kims award of a graduation certicate to the b est student

To summarize arguments marked with an inherent case can b e assigned structural

case as well and therefore do not constitute an exception to the general pattern of case

assignment summarized in Table For the cases where inherently case marked arguments

are not marked with overt structural case I assume that they are due to PF case particle

deletion

Thetaindex

his For role assignment I assume the mechanism prop osed in Fukui whic

summarized b elow marking takes place under strict sisterho o d as dened in

and are sisters if they are dominated by the same no de

An argument structure a grid in the sense of Stowell is more than just an

unordered list of roles it is structured according to the closeness of a role to the

predicate This is represented by the linear order of the role in a grid ie the lefthand

role is higher than the one to its right For example in is higher than to

i i

the lexical head to whichthe grid is asso ciated and is the highest role

grid h i

i i n

The discharge of the roles takes place sequentially from right to left under the strict

sisterho o d without skipping over a non marked p osition This mo de of role discharge is

schematically representedin

The argument structure I am assuming here is close to the prominence theory of argument structure

advo cated in Grimshaw in which argument structure represents the argumentlicensin g capacityof

a predicate without sp ecifying anysemantic information ab out its arguments except for their relative

prominence hierarchy Crucially I am not adopting the Uniformity of Theta Assignment Hyp othesis

prop osed by Baker and stated b elow

The UNIFORMITY of THETA ASSIGNMENT HYPOTHESIS UTAH

Identical thematic relationships b etween items are represented byidentical

structural relationship s b etween those items at the level of Dstructure

V

Q

Q

V

Q

Q

V

Q

Q

V

Q

Q

V

In the only p osition whichgets marked directly by the V is the sister of the verbal

head V and all the other roles in a given grid are assigned comp ositionally from the

b ottom up under the sisterho o d relation

Besides the accountof role assignment describ ed ab ove I make an additional assump

tion that once a role is assigned to an argument the argument carries a numerical index

ie index which is identical to the numb er of the role in the grid

Thetaindexing algorithm

a roles are discharged in a b ottomup fashion according to the grid of the

lexical head

b The numb er asso ciated with each role index in the grid is inherited by the

argument assigned the role

grid of the verb is h agent goal theme i The complement senmwul In b elow the

carries index the dative argument Younghee and the sub ject Minho

Minho ka Younghee eykey senmwul ul cwuessta

Minhonom Youngheedat presentacc gave

Minho gaveYounghee a present

According to the thetaindexing algorithm the argument which is assigned the highest

role always carries thetaindex

Case assignmentrule

Case licensing condition is rep eated as in

a Genitive case is licensed via head government byanX category with

feature N V

b Accusative case is licensed via head government byacomplexX category

with feature stativeF

c Nominative case is licensed via head government byanX category with

feature F

index and the case pattern for Korean summarized in Table I pro Incorp orating

p ose the Case Assignment Rule in

Kang prop osed the case assignment rule given b elow

Case AssignmentRule

a Assign genitive case if an argumentisgoverned byanX category with

feature N V

b Assign accusative case if an argument whose index is not is governed by

an X category with feature stative F

c Assign nominative case if an argument is assigned neither genitive nor

accusative case and is governed byanX category with feature F

asays that arguments which are governed by a noun are assigned genitive case b

says that all the arguments of a stative predicate ie verb except for the one which

carries the highest role is assigned accusative case csays that the argumentofa

stative predicate which carries the highest role and all the arguments of a stative

predicate ie adjective are assigned nominative case

is particularly interesting b ecause it combines the apparently conicting views

that nominative case is assigned by default cf Kang Kim and that nomi

native case is assigned by infl Yim Ahn and Yoon Whitman The

condition that nominative case is assigned to an argument which is assigned neither gen

itive nor accusative case instantiates the idea that nominative case is the default case in

tby Korean At the same time the condition that nominative case is assigned via governmen

a category with feature F implements the idea that nominative case is assigned by infl

In short the view that nominative case is the default case and the view that nominative

case is assigned by infl are not mutually exclusive In fact b oth views are correct

Examples

Ditransitiveverb sentences

The goal argument of ditransitiveverb cwuta give can b e marked dative accusativeor

the combination of dative and accusative as in in which the numb ers subscripted to

the arguments are their indices

Generalized case marking in Korean Kang

a ACC case is assigned to NPs which are sisters of stative V

b NOM case is assigned to all noncasemarked NPs default

There can arise a question concerning whether nominative on complements of transitive adjectives eg

mwusepta to b e afraid of is indeed assigned by infl as I argue here or by a complex category with feature

stative F parallel to accusative case on complements of transitiveverbs I assume that they are assigned

by infl just like the nominative case b orne by sub ject arguments based on the following fact in English

Transitive adjectives in English like to beafraid and to befond lack the accusative case assigning ability

and their complements are assigned accusative case which is assumed to b e the default case in English via

dummy of insertion In analogy to English adjectives whichdonothave a caseassigning ability I assume

that adjectives in Korean do not have a caseassigning ability and nominative case which is the default

case in Korean is assigned to them under governmentby infl

See Ahn and Maling and Kim for dativeaccusati ve case alternation of the goal argument

and Yo on and Yo on for case stacking of datacc and datnom

Minho ka Younghee eykeyluleykeylul senmwul ul cwuessta

Minhonom Youngheedataccdatacc presentacc gave

Minho gaveYounghee a present

Aschematic phrase structure representation of is given in in which the numb ers

asso ciated with the NPs are indices

C

P

P

P

P

CIFVstat IP

Z

Z

NP IP cwuessta

H

H

H

NP dat IP

Z

Z

NP I

H

H

H

I VP

H

H

H

t VP NP

b

b

b

t NP V

c

c

V t NP

t t

In NP and NP are governed by the complex category IVF stative and there

fore assigned accusative case NP which carries index cannot b e assigned accusative

case even though it is governed by IVF stative and is assigned nominativecaseby

iii

Examples like in which more than one argument is assigned accusative case

supp ort my prop osal that case is not assigned by SPEChead agreement since under such

a view it is necessary to p osit more than one pro jection of AGR to accommo date multiple

O

accusative case Positing more than one pro jection of the same head however runs counter

to the main motivation for p ositing an indep endent pro jection in the phrase structure Each

pro jection in the phrase structure is assumed to have its hierarchical status distinct from

all other pro jections and the hierarchyofeach pro jection plays a crucial role in explaining

various phenomena For multiple pro jections of the same head however each pro jection

do es not corresp ond to distinct hiearchical status defeating the motivation for p ositing

distinct pro jections

Passives

Consider which is the passivecounterpart of Following Kang I assume

that a passive predicate formed bytheci auxiliary as that in is stative

Minho kaeykeykaeykeylul senmwul iul cwueciessta

Minhonomdatnomdatacc presentnomacc givepasspstdec

Minho was given a present

C

P

P

P

P

P

CIFVstative IP

H

H

H

cwueciessta NP dat IP

Z

Z

NP IP

H

H

H

I VP

b

b

b

t NP V

c

c

V t NP

t t

In there is no X category with feature either N V for genitive case or F

stative for accusative case and therefore all the arguments are assigned nominative case

according to iii

As Kang and Hong show lexical passives formed by ihiliki axation

exhibit two distinct case p ossibiliti es as illustrated in ex a and b in Kang

ku namwukakacikalul calliessta

the treenom branchnomacc cutpasspstdec

A branch of the tree was cut

In one argument of the lexical passivemay b e assigned either nominativeorac

cusative Concerning these two distinct case p ossibilities I assume that a lexical passiveis

ambiguous b etween verbal and adjectival cf Levin and Rappap ort b oth of which

are formed in the lexicon and will b e discussed in detail in Ch With this assumption

if the passive predicate is adjectival with feature stative then all of its arguments will

b e assigned nominative case while if it is verbal with feature stative then the argument

which carries the highest role ku namwu in this case is assigned nominative case and

others accusative case

Passives in Korean are sub divided into two categories One is socalled lexical passives which are formed

by inxing one of the morphemes i hi li ki between the verb stem and the tense morpheme and

the other socalled ci passives which are formed by adding the auxiliary verb ci after the verb stem as

in kala ciessta is changed Ci passives are stative and lexical passives are either stative adjectival

passive or stative verbal passive

ECM constructions

Exceptional case marking ECM constructions in Korean app ears to p ose some problems

for the Case Assignment Rule in

nayka Minhoul chencaylako sayngkakhanta

Inom Minhoacc geniuscopcomp think

I think Minho to b e a genius

If we assume the standard analysis of exceptional case marking the emb edded sub ject Minho

is assigned accusativecaseby the matrix verb somehow However this mo de of exceptional

case marking is ruled out in my analysis The matrix verb has raised to the matrix INFL

and it is the trace of the raised verb whichgoverns the exceptionally casemarked element

contradicting my assumption that only the head of a chain assigns case

However it has b een argued in Hong that the accusative NP in is in fact an

argument of the matrix predicate rather than the emb edded giving rise to a representation

in

naykaMinholul PRO chencaylako sayngkakhanta

S i S i

If Hong is right then case assignment to the arguments in the sentences like will b e

no dierent from that in other sentences The socalled exceptionally case marked element

will carry a index assigned by the matrix verb and b e assigned case accordingly

Empty pro sub ject

The Case Assignment Rule predicts that there is always at least one nominative case

marked argument in a clause the argument which carries index This prediction seems

to b e b orne out in general except for some cases suchas in which there is no overt

nominative casemarked argument

kwukpangpwueyse choisin mwukilul taylyang kwuiphayssta

Defense deptloc most recentweap onacc large quantity purchased

The Defense Dept purchased the most recentweap ons in large quantities

Despite its surface form in which there is no missing argument intuitively sentence

feels like there is a missing sub ject which receives the agent role of the verb and the miss

ing sub ject refers to entities related to the lo cative phrase kwukpangpwueyse A translation

which reects this intuition is something like People in the Defense Department purchased

the most recent weapons in large quantitiesTaking this intuition seriouslywemay b e able

to p osit a pro sub ject in whichislinked to the lo cative phrase If this is the case

then is not a real counterexample to the case assignment rule but the nominative

NP is realized as pro

Apparent problems

There are apparent problems for the prop osal that obligatory verb and infl raising to

comp induce obligatory scrambling of arguments for case purp oses They include arguments

without an overt case morpheme co ordination in event nominal clauses case assignmentin

innitival clauses and the apparent imp ossibility of nominative argument scrambling I will

examine each phenomenon in turn and argue that they do not constitute real problems

Arguments without overt case morphemes

Ihave b een implicitly assuming that case licensing is realized by an o ccurrence of an overt

case particle Given this assumption a question arises concerning the cases where an

argument is not suxed byanovert case particle as in

Minho ku chayk sasse

Minho sub j the b o ok ob j b ought

Minho b ought the b o ok

The question is what kind of Case is b orne by the bare arguments arguments with no overt

case particles suxed to them assuming that the Case Filter is universal At least two

answers present themselves

a The bare arguments are a result of case particle deletion at PF

b The bare arguments b ear abstract Case as opp osed to morphologically realized

case

If we assume that the absence of an overt case particle is due to case particle deletion

at PF we exp ect that reversing the order of arguments is equally p ossible in This

exp ectation however is not b orne out Consider in which the order of arguments in

is reversed

ku chayk Minho sasse

Certainly is not as acceptable as the case in which arguments are marked with overt

een ku case Toconvey the intended reading a clear intonation break is necessary b etw

chayk and Minho

If we assume that the absence of an overt case particle is due to abstract Case assignment

a subsequent question arises concerning how abstract Case is assigned Considering the

unacceptability of examples like it do esnt seem that abstract Case is assigned in the

same wayasovert case An obvious hyp othesis would b e that abstract case b oth nominative

and accusative is assigned VPinternally by the verb This hyp othesis however results

in a contradiction when combined with the assumption that case is assigned by the head

ofachain at Sstructure and that verb raising is obligatory After verb raising to comp

the raised verb cannot govern the VPinternal arguments and therefore cannot assign case

unless we adopt the GovernmentTransparency Corollary

Another alternative is that abstract case is assigned VPinternally by the verb and

abstract case assignmenttakes place at the same level of grammar as role assignment

This alternative is consistentwithmy analysis of scrambling and overt case assignment

The only o dd thing ab out this prop osal is that in general abstract Case is assumed to b e

assigned at Sstructure and role at Dstructure At the moment I am not sure what the

correct analysis is for arguments without overt case particles and leave this question op en

for further research

Co ordination in event nominal clauses

Event nominal clauses can b e co ordinated as in which is due to James Yo on

Johni swuhakul yenkwu Maryka thongsalonul kongpwucwung

XP XP

Johnnom mathacc research Marynom syntaxacc researchduring

During Johns study of math and Marys study of syntax

lo oks like an NP co ordination ie XP NP and app ears to p ose a problem for my

prop osal that the functional feature F is necessary for nominative and accusative case

licensing The head noun of the rst conjunct is not followed byanaspectmarker whichI

argue to b e resp onsible for verbal case marking Nevertheless the arguments exhibit verbal

case array ie the sub ject is marked nominative and the ob ject accusative However if

we assume that the category of the rst conjunct is in fact AspP and contains an abstract

asp ect marker as in do es not p ose a problem for the current prop osal

AspP

h

h

h

h

h

h

h

h

h

AspP AspP

H H

H H

H H

nom AspP NP nom AspP NP

l i

H H

H H

H H

NP acc Asp NP acc Asp

j m

X X

X X

X X

X X

X

X

NP Asp Asp NP

H

b b

Q

H

Q H

b b

b b

NP Asp Asp N N N N NP

k n

c c

c c

t yenkwu t kongpwu cwung N N NP NP

i l

t t t t

j k m n

Alternativelywe could assume that is an instance of rightno de raising as James

Yo on p oints out to me in p ersonal communication

In fact there is indirect evidence that is not an instance of NP co ordination

ie XP NP In general NPs can b e co ordinated by using the conjunctive particle kwa

as illustrated in

Johnuy swuhakuy yenkwuwa Maryuy thongsalonuy kongpwu

Johngen mathgen researchand Marygen syntaxgen study

Johns study of math and Marys study of syntax

However the use of the conjuntive particle wa to co ordinate the category XP in results

in an ungrammatical string as in

For other ways of co ordinating NPs refer to Cho and Morgan

Johni swuhakul yenkwuwa Maryka thongsalonul kongpwucwung

Johnnom mathacc researchandand Marynom syntaxacc studyduring

If the category of the co ordinated elements in is really an NP in the absence of the

asp ect marker the unacceptability of is surprising while it is easily explained if we

assume that there is a zero asp ect marker in the rst conjunct The unacceptability of

contrasts with the acceptabilityof

Johnuy swuhakuy yenkwuwa Maryuy thongsalonuy kongpwucwung

Johngen mathgen researchand Marygen syntaxgen studyduring

minimally diers from in that all arguments are marked with genitive case

as opp osed to verbal case ie the head noun is not raised to the p osition o ccupied by

the asp ect marker Therefore the category of the co ordinated elements is NP and the

co ordination by particle wa is grammatical

Co ordination p ossibilitie s in event nominal clauses other than and are given

b elow in the examples kuliko is another co ordination word which can b e used for co ordi

nation of any category

i Johni swuhakuy yenkwuwakuliko Maryka thongsalonuy kongpwucwung

wakuliko thongsalonuy yenkwucwung ii Johni swuhakuy yenkwu

The unacceptability of i is exp ected under my analysis Given that the sub ject is

marked nominative the category of the co ordinate must b e AspP Therefore co ordination

by kuliko is ne but co ordination by wa is ruled out Note that the marginal acceptability

of the co ordination by kuliko simply reects the degraded acceptability of the mixed case

array b efore co ordination cf c Concerning ii it seems to b e b etter treated as

rightno de raising or acrosstheb oard rather than simple co ordination Regardless of the

exact nature of co ordination in ii however it is clear from the nominativecaseon

the sub ject that the head noun has b een incorp orated into the asp ect marker by the time

the co ordination has taken place Hence the category of the co ordinate cannot b e an NP

explaining the unacceptability of co ordination by wa

To conclude the co ordination in is not a counterexample to my claim that a

functional category with the feature F participates in verbal case licensing and that the

licensing conditions on verbal case require obligatory scrambling of arguments out of their

domain

Case assignment in innitival clauses

Obligatory casedriven scrambling out of VP which can result in p ermuted word orders

crucially hinges on the existence of verb raising to comp Therefore nominativeaccusative

case assignment in a clause the category of whichisnotaCP will run counter to this

prop osal Apparently there exist such cases in Korean namely case assignment to a causee

in causative constructions It has b een generally assumed that the categorial status of the

clausal complement of a causativeverb is IP or sometimes even VP Nevertheless arguments

of the complemen t clause of a causativeverb in Korean can b e assigned nominative and

accusative case as illustrated in

Youngheenun Minhoka tayhakipsilul phokihakey mantulessta

Youngheetop Minhonom college entrance examacc give upcomp made

Younghee made Minho give up taking the college entrance exam

The emb edded clause of the causativeverb mantulta make in is innitival yet

the sub ject and the ob ject are marked with nominative and accusative case resp ectively

Therefore examples like app ears to run counter to my claim

However I argue that this is not a real counterexample and make the following assump

tions First there is an abstract infl in innitival clauses as has b een standardly assumed

for innitival clauses in general Second the innitival clause in is either CP or C

contrary to the standard assumption that innitival clauses are either IP or I I further

assume that the verb and the abstract infl in an innitival clause raise to comp just as

in nite clauses as schematically represented in cf Stowell for raising of tense

op erator to comp at LF

CP

b

b

b

SPEC C

b

b

b

IP CI V

abs

b

b

b

NP nom IP

i

b

b

b

NP acc I

j

H

H

H

VP I

abs

b

b

b

t NP V

c

c

t V NP

i

t t

j

If the assumptions stated ab ove are correct then a sentence like is not a counterex

ample to my prop osal In fact the assumptions explain why there exists such a dierence

between English and Korean in the case p ossibilities of a causee in causative constructions

The categorial status of the complement clause of a causativeverb is an IP in English while

it is a CP in Korean

Heyco ck sp eculates that there are no true small clauses in Korean on the basis of her study of

the same construction

It has b een widely noted in the literature cf Lee b Bratt and references cited there that

the causee in p eriphrastic causative constructions in Korean can b e marked accusative and dativeaswell

as nominative For the cases where the causee is marked accusative and dative the categorial status of the

complement clause mightbeanIP

Scrambling of nominative arguments

Certain stative predicates ie transitive adjectives in Korean sub categorize for nominative

complements Consider and

Minhoka caki citokyoswukaul mwusepta

Minhonom selfgen advisornomacc is afraid of

Minho is afraid of his advisor

nayka Minhokalul cohta

Inom Minhonomacc b e fond of

I am fond of Minho

The theme arguments citokyoswuka in and Minhoka in apparently cannot b e

scrambled as illustrated in and

caki citokyoswu ka Minhokat mwusepta

i i

Minho ka naykat cohta

i i

If I am right in claiming that verb and infl raising induces scrambling for case purp oses

thereby arranging arguments in any order the unacceptability of the examples such as

and is problematic Nominative case is assigned by IF in comp to the scrambled

arguments in any order

ts is only appar Concerning this problem I will argue that scrambling of argumen

ently blo cked A close examination of a wide range of data indicates that this blo cking

eect on nominative argument scrambling is due to the antiambiguity condition ad

vanced in Kuno which I take to b e a discourse condition I will rst briey review

Miyagawa s prop osal for a similar problem in Japanese and then prop ose my expla

nation of the data

Inlowering account

Discussing p otential constructions in Japanese in which the predicate sub categorizes for

either an accusative or a nominative complement Miyagawa notes that if

the complementismarked nominative scrambling of the complement is almost imp ossible

which is analogous to the Korean examples in and This is illustrated b elow

the judgments are also Miyagawas

Johnga nihongooga hanaseru

Johnnom Japaneseaccnom sp eakcanpresent

John can sp eak Japanese

nihongo ga Johnga t hanaseru

i i

Miyagawa explains the unacceptability of the scrambled sentence by p ositing In

lowering adopting Takezawas analysis of caseassignment in p otential constructions Ac

cording to Takezawa the nominative case on the ob ject NP in is assigned b y the

lowered In as illustrated in

Miyagawa assumes Inlowering to account for the unacceptabili ty of the examples like while

maintaining Inraising to account for scrambling

nihongoga hanaseru t

IP VP i i

Japanesenom sp eakcanpresent

Assuming that this analysis is correct the unacceptability of the scrambled phrase in

is easily explainable The scrambled ob ject NP has adjoined to IPHowever this IP adjoined

p osition is not a Case realization p osition b ecause the In has lowered to V and therefore

do es not govern the moved NP

Even though the Inlowering account seems to work well for the cases which Miyagawa

discusses there are some problems in applying it to the Korean examples I discuss Going

back to examples and the arguments mayhavedatnom case array in addition

to the nomnom case array as illustrated in

Minhoeykeyka caki citokyoswuka mwusepta

Minhodatnom selfgen advisornomacc is afraid of

Minho is afraid of his advisorTo Minho his advisor is frightening

ta naeykey Minhoka coh

Idat Minhonomacc b e fond of

I am fond of MinhoTo me Minho is nice

When the arguments exhibit datnom case array scrambling of the nominative comple

ments is p erfectly acceptable as shown in and

okcaki citokyoswu ka Minhoeykey t mwusepta

i i

okMinho ka naeykey t cohta

i i

If the Inlowering account is correct the acceptability of and is unexp ected

since the scrambled nominative complement is adjoined to IP and is not governed by the

lowered In as in and Another fact which casts doubts on the Inlowering

account is that if the exp eriencer arguments sub jects are marked with the topic particle

nun the acceptability of the sentences improves signicantly as in and

caki citokyoswu ka Minhonun t mwusepta

i i

Minho k ananun t cohta

i i

Prop osal Antiambiguity constraint

Concerning the apparent unacceptability of and I will argue that it is due to

the antiambiguity constraint on scrambling whichI take to b e a discourse constraint

along the line of Kuno Kuno argues that the unacceptabilityofscrambling

in sentences like ab ove and bbelow ex in Kuno is due to the

antiambiguity device which is stated as taken from Saito

Historically one of the ma jor arguments whichhave supp orted the view that nominative arguments

cannot b e scrambled comes from a certain asymmetry in the b ehavior of oating numeral classiers NC

hereafter which are asso ciated with sub jects and ob jects Based up on the observation by Kuro da

Saito and Miyagawa explain the contrast in acceptabilitybetween b and d by

assuming that sub jects can not b e scrambled while ob jects can

a Gakuseiga sannin hono katta

Studentnom clp erson b o okacc bought

Three students b ought b o oks

b Gakuseiga hono sannin katta

Studentnom b o okacc clp erson b ought

c Gakuseiga hono nisatu katta

Studentnom b o okacc clb o ok b ought

A student b oughttwobooks

d hono Gakuseiga nisatu katta

b o okacc Studentnom clbook bought

In a the NC sannin is asso ciated with the sub ject NP and the ob ject NP cannot intervene b etween them

as shown in b In c the NC nisatu is asso ciated with the ob ject NP Contrary to b however the

intervention of the sub ject NP b etween them do es not result in an unacceptabili tyasshown in d

Saito derives this contrast by p ositing asymmetry b et ween nominative and accusative case assign

ment which has the eect of preventing a nominative NP from b eing scrambled but allows an accusative

NP to b e freely scrambled Assuming that an NC is in mo dication relation to its host NP and that they

are generated adjacenttoeach other at Dstructure b is derived by rst scrambling the ob ject and then

scrambling the sub ject across the scrambled ob ject leaving the NC b ehind as shown in e However the

derivation e is illegiti mate since the sub ject NP cannot b e scrambled hence b is ungrammatical On

the other hand d is derived by scrambling the accusative ob ject leaving the NC b ehind as in f This

derivation is legitimate hence d is grammatical

e Gakusei ga hon o t sannin t katta

i j IP i VP j

studentnom b o okacc cl bought

f hon o Gakuseiga t nisatu katta

i IP VP i

bookacc studentnom cl bought

The contrast b etween b and d remains unexplained in my analysis Nevertheless some facts in Korean

ukushima lead me to b elieve that the account given similar facts are observed also for Japanese in F

by Saito is not on the righttrack That is the constrast b etween b and d disapp ears if we replace the

NC by nonnumeral oating quantiers as illustrated in g j

g yehaksayngi motwu isakenul mokkyekhayssta

female studentnom all this eventacc witnessed

All female students witnessed this event

h yehaksayngi i sakenul motwu mokkyekhayssta

i siptaytuli taypwupwun Michael Jacksonul coahanta

teenagerplnom most Michael Jacksonacc like

Most teenagers like Michael Jackson

j siptaytuli Michael Jacksonul taypwupwun coahanta

In g and i the oating quantiers motwu and taypwupwun are asso ciated with their sub jects Inter

vention of the ob ject b etween the sub jects and the oating quantiers as in h and j do not result in

ungrammaticalityincontrast to b ab ove Of course we need to explain wh y there is such a dierence

between h and j on the one hand and b on the other Nevertheless the acceptability of h and j

casts doubts on Saitos claim that the ungrammaticality of b is due to the imp ossibi li ty of scrambling of

nominative arguments

In general the greater the likeliho o d of ambiguous interpretation the more

dicult it is to switch the word order of two NPs marked with the same

grammatical formative eg particle

a Taro oga tenisuga zyoozuda

Taro onom tennisnom go o d atis

Taro o is go o d at tennis

b tenisu ga Taro oga t zyoozuta

i i

Kuno do es not make explicit what the nature of the antiambiguity device is I interpret it

as a discourse condition That is scrambling is asso ciated with a particular discourse func

tion such as presupp ositionality which will b e discussed in Ch and the functorargument

structure of a scrambled sentence has to b e inferrable from the discourse Otherwise the

scrambled interpretation is disfavored even though the given scrambling is p ossible syntac

tically If this explanation is correct we exp ect that nominative argument scrambling is

p ossible in an appropriate discourse context Furthermore scrambling of an oblique argu

ment eg accusative and dative argument will b e hard if the intended functorargument

structure of the scrambled sentence is not easily inferrable even though scramblabilityof

an oblique argumenthasnever b een put into a question Below I will show that these

exp ectations are indeed b orne out

Scrambling of nominative arguments

Consider examples b elowinwhich scrambling of nominative arguments is p erfectly acceptable

catongchaka sako siphessnuntey computer ka aituli t kkok

i i

carnom buycomp wantedbut computernom kidsnom really

philyohatako hayse computerlul sassta

is in need ofcomp saytherefore computeracc b ought

I wanted to buy a car but a computer my kids said that they really need

and therefore I b ought a computer

In the nominative complement computerka has b een scrambled across the nominative

sub ject aituliyet the sentence is p erfectly acceptable The preceding sentence catongchaka

sako siphessnuntey facilitates the intended scrambling

Even though I argue that nominative argument scrambling is p ossible I am unsure as to whether or not

scrambling of adjunct nominative NPs in multiple nominative constructions is equally p ossible If scrambling

is blo cked in multiple nominative constructions it may b e due to the same ordering constraints imp osed on

ordering of multiple mo diers in general

A question arises concerning whether computerka in has undergone lo cal or longdistance scram

bling The sentence containing the scrambled phrase is a complex sentence in which either the matrix or

the emb edded sub ject is pro dep ending on howwe analyze the sentence The two p ossible representations

are given in a and b b elow

a computer kaaituli pro t kkok philyohatako hayse

i i

b computer kapro aituli t kkok philyohatako hayse

i i

The nominative complement has undergone longdistance scrambling in a while it is lo cally scrambled in

b Which representation is the correct one however dep ends on ones theory of the distributio n of pro

ecey wulinun kangwentoey issnun han mokcangey kassessta

Rel

yesterdaywetop Kangwen provincein existrel a pastureloc went

Yesterdaywewent to a pasture whichisinKangwen Province

kulentey ku mokcang i Minhoka t caknyenkkaciman hayto

i i

by the way that pasturenom Minhonom last yearonly until

koaswuweniesstako malhayssta

was an orchardcomp said

By the way that pasture Minho said was an orchard even until last year

In theemb edded nominative sub ject ku mokcangi has longdistance scrambled across

the matrix nominative sub ject Minhoka and the sentence sounds more natural than its

canonical order counterpart to myear More examples involving nominative argument

scrambling are given b elow

Lo cal scrambling of nominative complement

ton i naykat sampayk wen issta

i i

moneynom Inom three hundred wen unit of Korean currency exist

As for moneyIhave three hundred wen

Longdistance scrambling of a whsub ject of a transitiveverb

nwukwu ka Minhonun t nay cacenkelul hwumchiekassnunci ani

i i

whonom Minhotop my bicycleacc stolewhether knowqm

Do es Minho know who stole my bicycle and ran away

Longdistance scrambling of sub ject NP of a transitiveverb

cakiuy kachwul i Youngheenun t emmaeykey khun chwungkyekul

i i

nom Youngheetop motherdat big sho ckacc selfgen elop e

cwulilanunkesul alko issessta

would givethatacc knew

Younghee knew that her eloping would give a big sho ck to her mother

Scrambling of oblique arguments

In addition to the cases in which scrambling of nominative arguments sounds p erfectly nat

ural with an appropriate discourse context there are cases in which scrambling of accusative

and dative arguments is unacceptable

is the base order sentence in which the psychverb koylophita b other takes the

clausal and the accusative argument is a scrambled counterpart of The ac

cusative complement nalul has scrambled across the clausal argument and the sentence is

highly marginal

According to Huang the correct representation would b e a while Suh argues that the correct

representation is b The analysis wecho ose however do es not aect my claim

Chelswuka Youngheelul salanghantanun kesi nalul koylophiessta

Chelswunom Youngheeacc lovemod thatnom Iacc b othered

The fact that Chelswu loves Younghee b othered me

na lul ChelswukaYoungheelul salanghantanun kesi t koylophiessta

i i

Scrambling of the accusative argument Youngheelul of the emb edded clause across the

scrambled matrix accusative argument em nalul in is completely out as illustrated

by

Younghee lul na lul Chelswuka t salanghantanun kesi t koylophiessta

j i j i

Youngheeacc Iacc Chelswunom lovemod thatnom b othered

The marginality of and the unacceptability of is quite unexp ected since in

bling of an accusative argument is p erfectly grammatical However under general scram

the antiamibiguity constraint the marginality of the sentences nds an easy explanation

Scrambling of an accusative argument across another accusative argumentmakes it hard to

identify the functorargument relation of each clause In addition a discourse context which

accommo dates the instance of scrambling as in is not easily available explaining the

severe unacceptability of compared to

b and b illustrate that scrambling of a dative argument across another dative

argument is as bad as scrambling of a nominative argument across another nominative

argument

a emmaka ap ecieykey Minhohantey yongtonul moscwukey hayssta

momnom fatherdat Minhodat moneyacc neggivece made

Mom made father not give money to Minho

b emmaka Minho hantey ap ecieykey t yongtonul moscwukey hayssta

i i

a Chelswuka Youngheeeykey nwukwukaYouleeeykey cangmikkosul

Chelswu nom Youngheedat whonom Youleedat rosesacc

senmwulhayssnunci mwulessta

gaveasapresentwhether asked

Chelswu asked Younghee who gave roses to Youlee as a present

b Youlee eykey ChelswukaYoungheeeykey nwukwukat cangmikkosul

i i

senmwulhayssnunci mwulessta

To summarize the usual unacceptabilityofscrambling nominative arguments is only

apparent and is due to the antiambiguity constraint The diculty of scrambling of

oblique argument across another oblique argument of the same sort further supp orts the

claim

Long distance scrambling

I argue that long distance scrambling is no dierent from lo cal scrambling in that it is

equally casedriven The only dierence b etween them is the landing site of movement

ie inter clausal vs intra clausal

I will rst sketchhow long distance scrambling can b e explained in a manner parallel

to lo cal scrambling I then address the question of why long distance scrambling is not

sub ject to the same kind of lo cality constraints as standard Amovement The answer lies

in understanding the nature of Atraces Assuming that an Atrace is an anaphor sub ject

to the Binding Principle A it is not surprising that there exists long distance movement

of Anature in the same language

Derivation of longdistance scrambling

Consider which is an instance of long distance scrambling Numb ers asso ciated with

the arguments are their indices Arguments the indices of which are marked are the

arguments of the emb edded verb

i chayk ul Kim kyoswu ka motwu ka t ilkeyahantako malhayssta

i i

this b o okacc Prof Kimnom everyonenom must readcomp said

Prof Kim said that everyone has to read this b o ok

In the long distance scrambled emb edded ob ject is adjoined to the matrix IPItis

assigned accusativecaseby the IVF stative complex asso ciated with the matrix clause

bling implies that there is no correlation b etween the My analysis of long distance scram

sub categorization frame of a verb and case assignment ie Case do esnt have to b e assigned

to an argumentby its sub categorizing verb This is anticipated by the disso ciation of case

assignment from role assignment discussed in section and is further supp orted by

facts concerning case assignmenttoanadverbial which will b e discussed in detail in Ch

The relevantpoint is that Case is assigned not only to an argument but also to an adverbial

in Korean indicating that case assignment has nothing to do with the sub categorization

frame of a verb

Absence of lo cality constraints on scrambling

Even though I have b een arguing that scrambling is casedriven Amovement there is a clear

dierence b etween standard Amovements and scrambling While standard Amovementis

sub ject to the strict lo cality conditions as illustrated by the ungrammaticalityofb

scrambling is not Long distance scrambling out of a nite clause do esnt lead to an un

grammaticalityasin

a John seems t to b e intelligent

i i

b John seems that it is considered t to b e intelligent

i i

caki tongsayng ul Minhoka t haktayhayssta

i i

self s younger siblingacc Minhonom mistreated

His younger sibling Minho mistreated

However if we assume that an Atrace is like an anaphor and therefore is sub ject

to the same lo cality conditions for anaphor binding the lack of strict lo cality condition on

scrambling in Korean is not surprising Even though anaphor binding in English is sub ject to

lo cality conditions like the sp ecied sub ject condition SSC as in and the nominative

island condition NIC as in anaphor binding in Korean is sub ject to neither of

these conditions as illustrated in and Yang Progovac and Franks

Hong

John exp ects Mary to like herself himself

i j j i

John thinks that himself is a genius

i i

Younghee ka Minho eykey cakicasin sub jul chingchanhakey mantulessta

i j i

Youngheenom Minhodat selfselfacc praisece made

Younghee made Minho help herself himself

i j i j

Minho ka cakicasin i checaylako sayngkakhanta

i i

Minhonom selfselfnom geniuscopcomp think

Minho thinks that himself is a genius

i i

a Youngheeka caki lul pipanhaysstako malhayssta Minho k

i i

Minhonom Youngheenom selfacc criticizedcomp said

Minho said that Younghee criticized self

i i

Given that long distance binding into a nite clause is p ossible in Korean the existence

of long distance scrambling of Anature is exp ected

Achain

A question remains concerning howlongscrambling is derivedDoesittake place successive

cyclically through intermediate traces or do es it take place in one fell swo op Whatever op

tion we adopt our current understanding of the theory p oses a problem for an Amovement

analysis of scrambling Supp ose the movement takes place out of the emb edded VP to the

matrix IP in one fell swo op then sub jacency or a lo cality condition of a similar sort will b e

violated On the other hand if the movementtakes place successive cyclicallyintermediate

traces b ecome problematic For instance the derivation of rep eated here as

will b e represented as in

eyahantako malhayssta ka t ilk i chayk ul Kim kyoswu ka motwu

i i

this b o okacc Prof Kimnom everyonenom must readcomp said

Prof Kim said that everyone has to read this b o ok

kat ilkeyahantako malhayssta motwu ichayk ul Kim kyoswu kat

i i i

The problem lies in the fact that the intermediate trace t in is in a case assignable

p osition and therefore the movement from the intermediate p osition to the landing site

of the matrix clause do es not constitute prop er Amovement hence is a violation of the

condition on Achains stated in

A maximal Achain has exactly one Casemarked p osition

n

namely and exactly one marked p osition namely Chomsky

n

Yang attributes the extended binding domain in Korean to the lackofAGR in this language

I am agnostic ab out which is the b etter of these two options even though the option

that the movementtakes place in one fell swo op seems more reasonable in that a sentence

involving long distance scrambling sounds likeaweak sub jacency violation comparable

to whmovementofacomplement out of a weak island in English In addition allowing

intermediate traces leads to highly unconstrained derivations Empirically there are some

facts which call the currentformulation of Achain into question in particular the Case

uniqueness condition As Yo on and Yo on argue and as wesaw in section an

argument with an inherent case can b e assigned structural case nominativeaccusative

in Korean The existence of suchmultiply casemarked arguments indicates that the

condition on Achains as currently formulated cannot b e maintained Finally examples like

due to Rob ert Frank pc in English suggest that the condition on Achains is to o

narrowly dened

Whom do you b elievet is smart

i i

The whmoved phrase whom in has originated in the p osition t towhich nom

i

inative case is assigned However the case whichisovertly realized is accusative not

nominative indicating that the phrase is exceptionally case marked by the matrix predi

ement Whether wetake the exceptional case marking to havetaken cate believe b efore mov

place through movement or insitu in it is clear that the whmoved phrase carries

two Cases namely nominative and accusative indicating that the chain condition which

requires anywellformed chain to have only one element with Case cannot adequately cover

all grammatical sentences

Where do es long distance scrambling diverge from lo cal scram

bling

Ihave assumed that when there is movement it is the moved element not its trace whichis

resp onsible for case licensing cf section This assumption results in a very interesting

dierence b etween lo cal and long distance scrambling when combined with the prop osal

in Heyco ck and Kro ch namely that any licensing relation satised by the head of

achain at Sstructure cannot in addition license a trace Consequentlygiven minimalist

assumptions along the lines of Chomsky Chomsky a trace that has had all

of the licensing conditions in which it participates preempted by the head of its chain must

delete unless it is indep endently licensed

Assuming that the Principle of Full Interpretation requires that a predicateargument

relationship should b e identied within the same clause ie the same extended pro jection

the trace of a lo cally scrambled elementmust b e deleted while that of a longdistance

scrambled element cannot Case licensing is satised by the moved argument and therefore

the trace has to delete as far as case licensing condition is concerned For the condition on

predicateargument identication the trace of a lo cally scrambled argumen t is not necessary

since the predicate and the moved arguments are within the same clause However the

trace of a longdistance scrambled argument needs to remain for the predicateargument

In fact under my system all arguments with inherent case obligatoril y scramble out of their domain

to b e assigned structural case The cases in which an argument with inherent case is not marked with an

overt nominativeaccus ative case morpheme are ascrib ed to casedeletion at PF

identication since the moved argument is not in the same clause as its sub categorizing

predicate This dierence b etween lo cal and long distance scrambling in terms of the

status of their trace captures the intuition whichmany p eople havehad That is lo cal

scrambling can b e easily handled in terms of basegeneration while long distance scrambling

cannot primarily due to the lo cality condition on role assignement cf Hale

My analysis of b oth lo cal and long distance scrambling as casedriven Amovement cap

tures the identical prop erties of lo cal and long distance scrambling with regard to binding

At the same time the particular view on a trace advanced by Heyco ckandKroch

which I adopt here adequately captures the intuition that a long distance scrambled argu

ment leaves a trace b ehind while a lo cal scrambled elementdoesnot

Deriving the parametric dierence b etween English and

Korean

The claim that scrambling is a consequence of casedriven obligatory movement of arguments

on a par with standard Amovement leads to the question of why English do es not have

scrambling In this section I showhowtoderive this parametric dierence b etween the two

languages

I assume that the case licensing condition in applies to English as well as Korean

The only dierence b etween the two languages is the level at which accusativecaseis

licensed In Korean accusative case is licensed at Sstructure while in English it is done

at LF as Chomsky prop oses Nominative case is licensed at Sstructure in

vert scrambling in English and Korean b oth languages The absence and presence of o

resp ectively follow from the dierence in the level at which accusative case is licensed

In Korean b oth a sub ject and an ob ject move out of VP at Sstructure and the moved

arguments can b e arranged in any order for the purp ose of case licensing giving rise to

scrambling eects In English only a sub ject moves out of VP at Sstructure since only

nominative case is licensed at Sstructure resulting in the constant sub jectverbob ject

order Movement of an ob ject at LF for accusative case licensing is invisible and therefore

do es not aect the surface word order

A question in turn arises namely what induces the dierence b etween English and

Korean in the level at which accusative case is assigned I argue that it is reduced to the

level at whichverb raising to INFL takes place Recall that the precondition of accusative

case assignmentisverb raising to INFL Verb raising takes place at Sstructure in Korean

overt raising but at LF in English covert raising cf Chomsky This

dierence is reduced to the nature of INFL in the two languages That is the Vfeature of

INFL in Korean is strong while that in English is weak cf Pollo ck Chomsky

This of course do es not exclude the p ossibili ty that the trace of a lo cally scrambled elementmightbe

required by an indep endent principle in which case the trace cannot delete Note that the skeleton structure

after trace deletion is compatible with one which do es not p osit functional pro jections and head movement

as advo cated in Sells abstracting away from the no de lab el

Roughly sp eaking overt movement in Chomsky corresp onds to Sstructure movement and covert

movement to LFmovement

Chapter

The Adjoined Argument

Hyp othesis

Throughout this thesis I have assumed that scrambling is adjunction I have also argued

that scrambling is like standard Amovement in that it is casedriven Ch and ex

hibits prop erties of Amovement with regard to binding Ch Combining these two

wereach the conclusion that adjoined p ositions are Ap ositions in Korean From this I

hyp othesize that adjoined p ositions can b e Ap ositions in Korean and call this the adjoined

argument hypothesis In this chapter I attempt to strengthen this hyp othesis by discussing

various phenomena They include case assignmenttoadverbials section binding bya

nominative adjunct NP in multiple nominative constructions section and absence of

island eects in scrambling out of a scrambled clause section

Case assignment to adverbials

In this section I discuss some data involving case assignmenttoadverbials and argue that

it is sub ject to the same case licensing conditions for arguments and therefore constitutes

further evidence that adjoined p ositions can b e Ap ositions in Korean

As rst noted by Maling and further elab orated in Cho a in Korean ac

cusative case can b e assigned to adverbial NPsPPs in particular to durationfrequency

adverbials Some examples are given b elow

Inhoka i chaykul twu p enuli ilkessta

read Inhonom this b o okacc two timesaccnom

Inho read this b o ok twice

Inhoka i chaykul sey sikantonganuli ilkessta

Inhonom this b o okacc three hourforaccnom read

Inho read this b o ok for three hours

The frequency adverbial twu pen in and the duration adverbial sey sikantongan in

are marked with accusative case despite the fact that they are not arguments

Citing Audrey Li s work Maling gives examples in Chinese which illustrate that not only an

Not only accusative but also nominative and genitive case can b e assigned to adverbials

as illustrated in

kili seoulkkacika hemhata

roadnom seoulup tonom bad

The road is bad up to Seoul

cinantali oltule pwutongsan kyengkika kacang cecohata

last monthnom this year real estate businessnom most b e sluggish

The real estate business was most sluggish last month in this year

ceng triouy mikwukeyseuy thukpyel kongyen

Jung Triogen Americalocgen sp ecial p erformance

Jung Trios sp ecial p erformance in America

ceng triouy sam nyenmanuy thukpyel kongyen

Jung Triogen three yearingen sp ecial p erformance

Jung Trios sp ecial p erformance in three years

The data involving case assignmenttoadverbials illustrated ab ove raise the question of

how Case is assigned to adverbials In particular are adverbials sub ject to the same Case

w I argue that they are indeed sub ject to the same Assignment Rule as arguments Belo

Case Assignment Rule fo cusing on nominative and accusativecase

Sensitivity to the stative distinction

Recall the Case Assignment Rule prop osed in Ch rep eated here as

Case AssignmentRule

i Assign genitive case if an argumentisgoverned byanovert X category

the feature of which is compatible with N V

ii Assign accusativecaseifanargumentwhose index is not is governed by

an overt X category with feature stative F

iii Assign nominative case if an argument is assigned neither genitive nor

accusative case and is governed byanX category with feature F

argument NP but also an adverbial NP requires abstract Case as in a and b b elow The crucial dierence

between Chinese and Korean is that accusative case in Chinese cannot b e assigned to an argument and an

adverbial simultaneously in the same sentence as illustrated by the ungrammaticality of c while sucha

situation is p ossible in Korean as shown in and in the text

a Ta nian le shu

he read asp book

He reads a b o ok

b Tanianle sange xiaoshi

he read asp three hours

He reads for three hours

cTa nian le shu sange xiaoshi

he read asp b o ok three hours

He reads a b o ok for three hours

If adverbials are sub ject to Case Assignment Rule just like arguments we exp ect case

assignmenttoadverbials to b e sensitive to the stative distinction of the predicate In

particular adverbials in intransitiveverb stative sentences are marked with accusative

case and those in transitive adjectival stative sentences nominative case This exp ec

tation is b orne out as illustrated in

catongchaka swuici anko sey sikantonganuli talyessta

carnom stopnmz notconj three hourforaccnom ran

The car ran for three hours without stopping

olhay tule hankangi sey penuli elessta

this year entercont Han rivernom three timeaccnom froze

The Han river froze three times this year

olhay tule sonakpika sey penuli oassta

this year entercont big rainnom three timeaccnom came

It rained heavily three times this year

naykacakny eney moktoni sey p eniul philyohayssta

Inom last yearloc a lot of moneynom three timenomacc was in need of

I was in need of a lot of money three times last year

nayka Inhoka sam nyentonganiul cohassta

Inom Inhonom three yearfornomacc b e fond of

I was fond of Inho for three years

The predicates in are stative The predicate in is unergativeintransi

tive and those in and are unaccusative The adverbials in these sentences are all

marked accusative and cannot b e marked nominative On the other hand the predicates in

and are stative transitive adjectives and the adverbials in these sentences

can only b e marked with nominative case Note also that the predicates in and

in which adverbials are marked accusative are stative and those in and in

whichadverbials are marked nominative are stative

Passivization Test

The passivization test also conrms the hyp othesis that Case on adverbials is of the same

nature as that on arguments as already p ointed out by Maling

Assuming that adverbials are sub ject to the same case assignment rules as arguments sentences like

and where the predicates are unaccusative and yet they license accusative case constitute

counterexamples to Burzios generalizatio n stated in i cf Chomsky

Averb with an ob ject Casemarks its ob ject i it marks its sub ject

Examples like also contradict the view that case on adverbials is assigned via case agreement

with the internal arguments of the verb In these examples there are no internal arguments with whichthe

adverbials can agree with

Consider and which are the passivecounterparts of and resp ectively

i chayki twu p eni ilkhiessta

this b o oknom twicenom was read

This book was read twice

i chayki sey sikantongani ilkhiessta

this b o oknom three hourfornom was read

This b o ok was read for three hours

In and b oth the complement and the adverbial which are accusative in the

activesentences are marked nominative This is exactly what we exp ect if the verb assigns

accusative case directly to the argument and the adverbial and passivization deprives the

verb of accusative case assigning ability

The accusativenominative case alternation of adverbials describ ed ab ove sharply con

trasts with the b ehavior of semantic case the use of which is constant regardless of the

voice of the sentence For example the instrumentcase lo in aismaintained in its

passive counterpart b

lo callassta a naykakokilul khal

Inom meatacc knifeins cut

I cut the meat with a knife

b kokika khalloka calliessta

meatnom knifeinsnom was cut

The meat was cut with a knife

Assuming that b oth arguments and adverbials are sub ject to the same case assignment

mechanism we can revise Case Assignment Rule as in to accommo date Case

assignmenttoadverbials

Generalized Case AssignmentRule

i Assign genitive case if an NPPP is governed byanovert X category

the feature of which is compatible with N V

ii Assign accusative case if an NPPP whose index is not is governed by

an overt X category with feature stative F

iii Assign nominative case if an NPPP is assigned neither genitive nor

accusative case and is governed byanX category with feature F

The only dierence b etween and is that the word arguments in is

replaced by NPPP in Rule implies that any NP and PP is assigned structural

As I will discuss later the adverbials can b e marked accusativeaswell as nominative in the passives I will

argue that adverbials are marked accusativeinverbal passives where the predicates are stative and nomi

native in adjectival passives where the predicates are stative in the sense of Levin and Rappap ort

Note that the accusative case b orne by the adverbials discussed here diers from inherent case b orne by

weekwhich are discussed in Larson bare NP adverbials in English as in I wil l do it next

case

Let us consider rep eated here as and see how the arguments and the

adverbial are assigned Case under the Generalized Case Assignment Rule

Inho ka i chayk ul twu p enuli ilkessta

Inhonom this b o okacc two timesaccnom read

Inho read this b o ok twice

In the numb er subscripted to each argumentisthe index of the argument and it

is not assumed that an adverbial is assigned a role There are three elements to which

Case needs to b e assigned is the Sstructure conguration of

CP

b

b

b

SPEC C

P

P

P

P

CIFVstat IP

Z

Z

NP IP

Z

Z

NP IP

Z

Z

NP I

adv

H

H

H

VP I

b

b

b

V t NP

c

c

t NP V

t t

In all of the three NPs are governed by the complex head IFVstat Among

the three NP and NP are assigned accusative case according to ii and NP

adv

nominative case according to iii

Case Assignment Rule overgenerates since there are some PPs which cannot b e marked with

accusative case and therefore cannot b e treated as a sub case of case marker deletion at PF For instance

the PP adverbial khallo with a knife in cannot b e marked with accusative case although the same

phrase can b e marked with nominative case with dierent predicates as in i

kokilul calukinun khalloka cohta

meatacc cutnmztop knifeinstnom go o d

Cutting meat with a knife is easy

Even though an instrumental PP cannot b e marked with accusative case there are instances in whichthe

intrumental particle alternates with accusative case as in ii

nayka Minholul chinkwulolul samassta

Inom Minhoacc friendinstacc to ok

I to ok Minho as a friend

Adjectival vs Verbal Passives

As noted by Maling adverbials in lexical passives maybemarked accusativeaswell

as nominative as in

i chayki twu p enuli ilkhiessta

this b o oknom twiceaccnom was read

This book was read twice

i chayki sey sikantonganuli ilkhiessta

this b o oknom three hourforaccnom was read

This b o ok was read for three hours

The case alternation on adverbials in lexical passives seems to b e a puzzle If a passive

verb is stative then we exp ect only nominative case whereas if it is stative only

accusative case I argue however that the case alternation is due to the fact that lexical

passives are ambiguous b etween an adjectival passive ie stative and a verbal passive

ie stative in the sense of Levin and Rappap ort

Levin and Rappap ort note that passives in English are divided into adjectival

that the and verbal passives They argue along with Chomsky and Marantz

essential prop erty of passive morpheme is the suppression of the external role Axing

the passivemorphemetoaverb prevents the verb from assigning its external role An

adjectival passive is formed from a verbal passiveby a category conversion rule which

changes the category VN into VN A verbal passive still maintains the abilityto

assign case to its internal argument while an adjectival passive do es not They list three

diagnostic environments for adjectival passives First negativeprexun attaches only to

adjectives eg unfriend ly unhappy but not to verbs Therefore passive participles that

are prexed with unasin unshaven unmarked untouched are categorially adjectival and

never verbal Second a number of verbs in English suchasseem remain sound select

adjectival but not verbal complements A passive participle app earing as the complement

to suchaverb is therefore taken to b e adjectival but not verbal Third only adjectives

may o ccur as prenominal mo diers

The diagnostics Levin and Rappap ort provide to distinguish adjectival passives from

verbal passiv es in English are not applicable to Korean however Nevertheless wecan

apply the two diagnostics which I describ ed in Ch to distinguish adjectival predicates

from verbal ones in Korean Only verbs can coo ccur with the progressive forming auxiliary

ko issta or present p erfect tense forming auxiliary e ota Applying these two diagnostics

we exp ect that only an adverbial marked with accusative case is compatible with a lexical

passive a in a progressive form and b in a present p erfect tense This exp ectation is

b orne out

ku chayki taycwungtuleyuyhay swu sipnyentonganuli ilkhikoissta

the b o oknom publicby several yearforaccnom readpassprog

The b o ok has b een b eing read by the public for several decades

The adjectival and verbal passive distinction here is comparable to the direct and adversity passive

distinction in Maling and Kim A direct passive absorbs the accusative case assigning abilityofthe

predicate and an adversity passive adds a b enefactivemalefacti ve sub ject argument but do es not change

the case assigning ability of the predicate

i nolayka semintulsaieyse swupayknyentonganuli

this songnom common p eopleamong hundreds of yearsforaccnom

pwullieoassta

singpassperf

This song has b een sung by common p eople for hundreds of years

and show that the progressive and the present p erfect forming auxiliaries are not

compatible with nominativeadverbials while b eing p erfectly compatible with accusative

ones This indicates that the nominativeaccusative case alternation on adverbials in lexical

passives is due to the ambiguous nature of lexical passive predicates as either adjectives

stative or verbs stative

Finally Levin and Rappap ort s claim that adjectival passives are derived from

verbal passives via a category conversion rule correctly predicts the absence of a passivefor

an adjectival predicate even if it is transitive as illustrated by the ungrammaticalityof

and which are the p otential passivecounterparts of and resp ectively

caknyeney moktoni philyohayciessta

last yearloc large amountofmoneynom is in needpasspast

A large amountofmoneyw as needed last year

Inhoka cohacinta

Inhonom b e fond ofpasspres

Inho is liked by someone

To summarize I have argued that adverbials are sub ject to the same case licensing

conditions as arguments Distribution of accusative case marked adverbials in intransitive

and unaccusativeverb sentences which is surprising under the standard assumption on ac

cusative case assignment nds an easy explanation under the Generalized Case Assignment

Rule Nominativeaccusative case alternation on adverbials in lexical passives is due

to the ambiguous nature of a lexical passive predicate as a verb and an adjective

Implications

If we adopt the standard assumption that adverbials are generated in adjoined p ositions

from the claim that Case assigned to an adverbial is of the same nature as that assigned to

an argument it follows that Case is assigned to a basegenerated adjoined p osition This

supp orts my claim that Case is assigned to a scrambled element which is adjoined

Of course at least two alternative conclusions may b e drawn from the fact that ar

guments and adverbials are sub ject to the same case licensing conditions One is adver

bials o ccupy complement p ositions on a par with complements as has b een assumed in

is p erfectly grammatical in the reading that Inho is b ecoming go o dnice which is due to the

lexical ambiguity of the predicate coh and is irrelevant for the present discussion Note that transitive

adjectival sentences in English suchasIamafraid of John and IamfondofMary do not have passive

counterparts either

This conclusion is consistent with Cho and Sells s claim that b oth casemarked NPs and adverbials

are verbal mo diers and therefore are not distinguishe d in the phrase structure

McConnellGinet Larson and Cho a The other is adverbials o ccupy

SPEC p ositions and case is uniformly assigned under SPEChead agreement Whichever

alternativewe take wewould need to p osit multiple pro jections of the same category for

cases in which there is more than one element of the same case Whether the adverbials

are in complement p ositions or in SPEC p ositions the heads which license these p ositions

would have to b e of the same category since the same case will b e licensed by the heads of

the same category and feature However I do not adopt these alternatives and continue to

assue that adverbials are adjoined

Binding by an adjunct

Another fact which indicates that adjoined p ositions in Korean b ehavelike Ap ositions

comes from binding by an adjunct The main data involve binding by an adjunct nominative

NP in socalled multiple nominative constructions

Multiple nominative constructions MNC hereafter are a widely discussed topic in Ko

rean and Japanese linguistics Here I briey sketchsomecharacteristics of the constructions

which are minimally necessary for my argument here In Korean and Japanese a clause

mayhavemultiple nominativecasemarked NPs or PPs only one of which is sub catego

rized for by the predicate of the clause Consider and

engi kilta pwukpankwuka mwunmyengkwukkaka yecaka swumy

north hemispherenom civilized countrynom womannom life spannom long

For the North hemisphere for civilize d countries for women life spans are long

ku samnyentongani cencayngi kacang simhayssta

the three yearsduringnom warnom most was severe

The war was most severe during the three years

In there are four nominative NPs The predicate kilta long selects for one theme

argument Only the innermost nominativeNPswumyengi life spannom is sub categorized

for by the predicate The remaining three nominative NPs are therefore adjuncts assuming

that only selected elements are arguments The extra nominative phrases can b e PPs as

well as NPs as illustrated by ku samnyentongan the three yearsduring in In

principle there is no upp er limit in the numb er of nominative phrases as long as certain

semanticpragmatic conditions roughly an ab outness condition are met As Hong

p oints out in general MNCs are b etter with individual level than with stage level predicates

probably b ecause the former refer to a p ermanent feature of an entity predicated of and

provide a b etter characterization of it compared to the latter which refer to a temp orarily

Saito in the talk given in the MidAtlantic Workshop on East Asian Linguistics February Univer

sityofDelaware indep endently argues on the basis of ECP facts in English that time and place adverbials

are complements of the verb and are distinct from reason and manner adverbials

Uniform mo de of case assignment via SPEChead agreement has most recently b een advo cated in

Chomsky

For more detailed discussions of the topic I refer the reader to Yang Kuno b Saito

Yim Yo on Heyco ck and Lee among many others

acquired feature This recursive nature of the o ccurrence of nominative NPs indicates that

the nonselected nominative NPs in these constructions are adjuncts and o ccupy adjoined

p ositions and makes implausible an analysis in which they o ccupy the sp ecier of saya

Topic Phrase

The relevance of the multiple nominative constructions for the present discussion is that

an adjunct nominative NP in MNCs can participate in binding contrary to the standard

assumption that binding is p ossible only from an Ap osition Consider the examples in

a Minho ka caki uy ap ecika paykmancangcaita

Minhonom selfgen fathernom millionairecop

Minhos father is a millionaire

b Minho wa Younghee ka selo ka macwu p oassta

Minhoand Youngheenom each othernom facetoface saw

Minho and Younghee lo oked at each other

c Minho wa Younghee ka selo ka selolul chasko issta

Minhoand Youngheenom each othernom each otherACC is lo oking for

Minho and Younghee are lo oking for each other

In the ab ove examples the outermost nominativeNPsMinhoka andMinhowa Younghee

ka are adjuncts Nevertheless they bind the reexive pronoun caki in a and the

recipro cal pronoun selo each other in b and c leading to the conclusion that

adjoined p ositions b ehave like Ap ositions with regard to binding

An alternative account for the binding facts in whichwould lead to an exactly

opp osite conclusion ie adjunct nominative NPs are A elements is to analyze the reex

iverecipro cal pronouns as resumptive pronouns which are b ound by the adjunct nominative

NPs which are op erators However this alternative analysis turns out to b e untenable when

we consider the Principle C

kuka Minhouy ap ecika paykmancangcaita

henom Minhogen fathernom millionairecop

lit As for him Minhos father is a millionaire

This generalization do es not exclude the p ossibil ty that a stage level predicate can participate in an

MNC

i Minhoka atuli kachwulhaysse

Minhonom sonnom ran away from home

As for Minho his son ran away from home

The predicate in i kachwulhata is a stage level predicate and yet it participates in the MNC

With regard to licensing of the whole NP adjunct NPs in the presentcontext in multiple case construc

tions various linguis ts haveobserved that certain nouns are relational in the sense that they always havean

implicit p ossessor cf Yo on Crosslinguistica ll y relational nouns include b o dy parts and kinship

terms Nouns with relational interpretations may b e thoughtofashaving an op en p osition for an implicit

p ossessor or an unsaturated argument structure cf Yo on Note that it is not p ossible to analyze the

sub ject arguments in as relational nouns with implicit p ossessors since the reexive pronoun itself is

the p ossessor in a

kutuli Minhowa Youngheeka macwup oassta

theynom Minhoand Youngheenom lo oked at each other

lit As for them Minho and Younghee lo oked at each other

and are parallel to a and b resp ectively except that the outermost

nominative NPs are pronouns and the expressions b ound by them are rexpressions If the

outermost nominative NPs are indeed op erators o ccupying an A p osition then the binding

theory is not applicable to the p osition Therefore we exp ect that the sentences

and are grammatical under the coreferential reading b etween the pronoun and the

rexpressions However the sentences are ungrammatical This can b e easily explained if we

assume that the adjunct nominative pronouns bind the rexpressions which is a violation

of the principle C

Islandho o d of scrambledclauses

Another fact whichisinfavor of the adjoined argumenthyp othesis involves the lackof

islandho o d of scrambled argument clauses

As I will discuss in Ch in detail an adjunct unselected by the predicate clause is

a rather strong island while a complement clause is not an island or at most a very weak

island for long distance scrambling Relevant examples are given b elow

Scrambling out of a prop ositional complement clause

lul Youngheeka nwuka t coahantako malhayssni a ku namca

i i

that manacc Youngheenom whonom likecomp saidq

lit That man Younghee said who likes t

i i

b ilensikulo Minhoka nwuka t sakilul chintako malhayssni

i i

This way Minhonom whonom cheatingacc docomp saidq

lit This way Minho said that who cheats t

i i

Scrambling out of an adjunct clause

a caki tongsayng ul Minhonun amwulato t ttaylimyen

i i

self s brothernom Minhotop anyone hitif

pro kamantwuci ankeysstako malhayssta

revengecomp said

Self s brother Minho said that if anyone t hits then he would revenge

i i

b ilen sikulo Minhonun pro nalmata t kongpwuhamyen

i i

this wayinst Minhotop every day studyif

pro iltungul halkesilako malhayssta

rank oneacc will docomp said

This way Minho said that if I study t everyday then I will b e numb er one

i i

Examples in and illustrate that long distance scrambling of an argument and an

adjunct out of a selected complement clause is grammatical or slightly marginal while long

distance scrambling of an argument and an adjunct out of an unselected adjunct clause is

ungrammatical or pretty bad

Under the standard assumption the categorial status of an adjoined clause is an adjunct

regardless of its selectional status as a selected complement or an unselected adjunct

Therefore it is predicted that an adjoined complement clausephrase constitutes an island

just like an unselected adjunct clausephrase cf Ross On the other hand under

the adjoined argumenthyp othesis there is no categorial distinction b etween an adjoined

element and an element o ccupying a complement p osition Therefore it is exp ected that

there is no dierence in islandho o d b etween an insitu complement clause and a scrambled

complement clause which is adjoined This exp ectation is b orne out supp orting the adjoined

argumenthyp othesis

Scrambling out of a scrambled complement clause

The verb yaksokhata to promise in a which is in the base order takes three arguments

ie a sub ject a dative argument and a clausal complement In b the ob ject argument

of the emb edded clause S has b een scrambledtosentence initial p osition In c

the adjunct ilensikulo of the emb edded clause S has b een scrambled to sentence initial

p osition Both b and c are p erfectly acceptable

Scrambling out of the in situ complement clause

a nayka Kim silcangeykey PRO ilen sikulo

S S

Inom Kim dept headdat this wayinst

chaki epmwukyehoikul silhaynghakeysstako yaksokhayssta

next quarter business planacc carry outcomp promised

I promised the dept head Kim that I would carry out the next quarters

business plan this way

b chaki epmwukyehoik ul nayka Kim silcangeykey PRO ilen

i S S

sikulo t silhaynghakeysstako yaksokhayssta

i

c ilen sikulo nayka Kim silcangeykey PRO t chaki

j S S j

epmwukyehoikul silhaynghakeysstako yaksokhayssta

Now Consider the examples in

Since I assume that even a canonical word order sentence is derived by obligatory scrambling it is

p ossible that even a complement clause in the canonical order is scrambled and adjoined to IP just likeany

other NP argument Even if this is the case the argumentgiven in this section can b e maintained

Scrambling out of a scrambled complement clause

a nayka PRO ilen sikulo chaki epmwukyehoikul silhaynghakeysstako

S S i

Kim silcangeykey t yaksokhayssta

i

b chaki epmwukyehoik ul nayka PRO ilen sikulo t

j S S j

silhaynghakeysstako Kim silcangeykey t yaksokhayssta

i i

c ilen sikulo nayka PROt chaki epmwukyehoikul

k S S k

silhaynghakeysstako Kim silcangeykey t yaksokhayssta

i i

In the complement clause of a has b een lo cally scrambled across the dative argu

ment Kim silcangb is derived by scrambling the complementNPchaki epmwukyehoik

ed by scrambling the adverbial ilen out of the scrambled clause in a cisderiv

sikulo out of the scrambled clause in a Both b and c are p erfectly accept

able In fact for some reason c is easier to pro cess than c The acceptability

of b and c indicates that a scrambled complement clause do es not constitute an

island for scrambling as predicted by the adjoined argumenthyp othesis

Before moving on to next topic I would like to remark on the derivation of b and

c In addition to the derivation I describ ed ab ove there is an alternative derivation for

these sentences That is the sentences maybederived by rst longdistance scrambling the

accusative argument or the adjunct of S to sentence initial p osition and then scrambling

the rest of clause S across the matrix dative argument remnant scrambling A schematic

representation of this alternativederivation is given b elow

step chaki epmwukyehoik ul t

j S S j

step chaki epmwukyehoik ul t t

j S S j i i

If the derivation in is allowed then my argument that a scrambled clause do es not

constitute an island cannot b e maintained since there is no scrambling out of a scrambled

clause As Michael Hegarty pc p oints out however this alternative derivation can b e

indep endently ruled out by the strict cycle condition stated in cf Chomsky

and reinstantiated for the substitution op eration in Chomsky

No rule can apply to a domain dominated by a cyclic no de A in suchawayto

aect solely a prop er sub domain of A dominated by a no de B which is also a

cyclic no de

says that rules cannot return to earlier stages of the cycle after the derivation has

moved to larger more inclusive domains In the second movement remnant scram

bling takes place within the domain which is aected by the rst movement and hence

violates the strict cycle condition

A similar situation arises in some cases involving topicalization in English There are

two p ossible derivations for the ungrammatical string c One is by rst topicalizing

the VP in the emb edded clause as in a and then p erforming whmovement out of the

topicalized VP The other is byrstmoving the whphrase to sentence initial p osition and

then topicalizing the emb edded VP containing the trace of the whmovement

a I know that buy a b o ok John never will t

i i

b What do you know that Joh never will buy t

j j

c What do you know that buy t John never will t

j j i i

Similarly the ungrammatical string cbelow whichistaken from Lasnik and Saito

can b e derived either by topicalization rst and then whmovement out of the topi

calized phrase or by whmovement rst and then remnant topicalization

a I think that pictures of Picasso Johnwanted

i

b What do you think that John wanted pictures of t

j j

c What do you think that pictures of t Johnwanted t

j j i i

and are schematic representations of the two p ossible derivations for c

and c

step topic t

S S i i

topicalization within the emb edded clause

step what t t

S j S top j i i

whmovement out of the topicalized phrase

step what t

S i S i

longdistance whmovement

step what t t

S i S top i j j

remnant topicalization within the emb edded clause

In step of the derivation violates either Kuno as internal constituent eect

or the adjunct island condition and therefore the sentences will b e correctly ruled out In

the only way to rule out the derivation is by resorting to the strict cycle condition

In summary the ungrammaticalityofsentences c and c and the p ossible

derivations for them indicate that the strict cycle condition has to b e considered a legiti

mate condition in the grammar Therefore is an illegitimate derivation for scrambled

sentences bc

Islandho o d of extrap osed clausesphrases in English

The absence of island eects in scrambling out of a scrambled clause contrasts with move

ment out of an extrap osed adjoined complement clausephrase in English

tences in with the pleonastic ob ject it demonstrating what Consider the sen

has b een called extrap osition structures cf Rosenbaum Postal and Pullum

Rothstein

a John regrets it that Bill red Susan

CP

b John resents it that Bill criticized me

CP

c John hates it that Bill red Susan

CP

d John p ointed it out that Bill hired Susan

CP

Lasnik and Saito incorp orate the strict cycle condition into their system as principle of strict

cycle and excludes all derivations similar to in their discussion of topicalizati on in Ch

In the ab ove examples it has b een generally assumed that the CPs have b een extrap osed

from the direct complement p osition whichislledby it Alternatively the CP comple

ments are basegenerated in a VPadjoined p osition and the direct ob ject p ostion is reserved

for the pleonastic it

Assuming that the extrap osed complements in are adjoined we exp ect that ex

traction out of them will exhibit island eects comparable to extraction out of an unselected

adjunct clause This is indeed the case

Extraction out of a nonextrap osed factive complement clause

a Who do es John regret that Bill red t

i i

b Who do es John resent that Bill criticized t

i i

c Who did John p oint out that Susan hired t

i i

Extraction out of an extrap osed factive complement clause

a Who do es John regret it that Bill red t

i i

b Who do es John resent it that Bill criticized t

i i

c Who did John p oint it out that Susan hired t

i i

Extraction out of an adjunct clause

a What did they cancel the show b ecause everyone sawt

i i

b Who did John sho ot p o ol while talking to t

i i

c What did John watchamovie b efore he ate t

i i

The extraction facts in illustrate that an extrap osed complement clause is

as strong an island as an adjunct clause Note that extraction out of an insitu factive

complement clause results in weak island eects cf Cinque

According to the pro jection principle as dened in Chomsky however the pleonastic it cannot

o ccupy the direct ob ject p ositon Consider principle i quoted from Chomsky p

i If sub categorizes the p osition then marks

Consider also the following quote from Chomsky p

Let us call such expressions arguments as distinct from idiom chunks nonargument

it as in it is certain that John wil l win or existential there as in therearebelievedtobe

unicorns in the garden terms which assume no role

The assumption that pleonastic it has no role combined with i leads to the conclusion that pleonastics

cannot o ccur in strictly sub categorized p ositions as p ointed out in Postal and Pullum However

Postal and Pullum convincingl y argue that pleonastic it in the examples in indeed o ccupies the

sub categorized direct ob ject p osition and I adopt their view here

Contrary to my assumption Cinque treats an extrap osed clause as a weak island in parallel with

a factive island on the basis of the data given in a and b b elow

a Towhom is it time to sp eak t

i i

b How is it time to b ehavet

i i

Out of the extrap osed clause extraction of an argument is grammatical as in a while extraction of an

Consider another example in which is taken from Lasnik and Saito

the explanation is also theirs

what did you givet to John a b o ok ab out t

i j i j

Given the principle of the strict cycle this example is derived as follows the Dstructure

ob ject abook about what is rst adjoined to VP and then what is moved to SPEC CP

The second movementinvolves extraction out of an adjoined phrase and hence Ross

subsumed this example under his generalization that adjunction structures are islands for

movement

Why is there a complementadjunct distinction

Even though the absence of island eects in scrambling out of a scrambled clause is con

sistent with the adjoined argumenthyp othesis a question arises concerning the contrast in

islandho o d b etween an adjoined complement clause and a nonselected adjoined clause If

islandho o d is determined on purely structural grounds and if there is no structural distinc

tion b etween an adjoined and a nonadjoined p osition as entailed by the adjoined argument

hyp othesis we exp ect there to b e no contrast in islandho o d b etween a complement and an

adjunct clause either

adjunct is ungrammatical as in b which is indicativeofaweak island However it seems to me that in

a and b the allegedly extrap osed clause is not really extrap osed but it is the insitu complementofthe

noun time The expletive it is inserted to satisfy the pro jection principle or the principle of predication

just like the expletive it in c b elow

c It is obviouscertai nl i kel y that John is sp eaking to the president

d To whom is it obvious that John is sp eaking t

i i

There are real problematic examples for my assumption that extrap osed clauses are adjuncts however

Consider eh whicharetaken from Pullum

e Which commitment has Jo e quit b ecause we cannot keep t

i i

f Which commitment will Jo e quit if we cannot keep t

i i

g Which commitment would it b e useful for us to keep t

i i

h Which commitment would it b e useful if wekept t

i i

The grammaticality of g and h which is an instance of extraction out of extrap osed clauses contrasts

with the ungrammaticality of e and f which is an instance of extraction out of true adjunct clauses At

the momentIhave no explanation for the grammaticality of g and h

However Lasnik and Saito ascrib e the ungrammaticali ty of the example to the crossing eect

discussed by Kuno and Robinson Pesetsky They argue that an A binder is not a barrier

ie island for movement on the basis of the fact that extraction out of a topicalized phrase as in b is

not as bad as extraction out of a sub ject phrase as in a

a Who do you think that pictures of t are on sale

i i

b Who do you think that pictures of t John wanted t

i i j j

They attribute the marginality of b to Kuno a s internal constituent eect However most native

sp eaker informants I have consulted with judge a and b equally unacceptable and I will takebtobe

ungrammatical due to a violation of the strict cycle condition

Given this apparent problem I suggest that there are two factors involved in determin

ing the barrierho o d of a clausephrase The selectional and the structural prop erties of the

element Of these two the selectional prop erties of the element are fundamental and can

not b e parameterized while the structural prop erties of an element can b e parameterized

Islandho o d of an element in terms of its selectional prop erties has b een incorp orated into

the denition of barrier by Cinque as stated in and

A single denition of barrier for bindingb oundin g

Every maximal pro jection that fails to b e directly or indirectly selected in the

canonical direction by a category nondistinct from V is a barrier for binding

A single denition of barrier for government

Every maximal pro jection that fails to b e directly selected by a category

nondistinct from V is a barrier for government

As for the structural prop erties of an element in languages like English any elementinan

adjoined p osition all adverbials and arguments adjoined via movement is an adjunct and

constitutes a barrier for movement while in languages like Korean the structural distinction

between an argument and an adjunct is blurred and all that matters for barrierho o d are

the inherent status of the category as an argumentadjunct This explains why an adjoined

complement is not an island in Korean while it is in English

A remaining question is why a sub ject clause in Korean do es not exhibit as severe barrierho o d as a

sub ject clause in English

Chapter

Constraints on Scrambling

In this chapter I discuss various constraints on scrambling They include island eects on

long distance scrambling and discourse constraints on various p ermuted word orders

In section I examine the islandho o d of various clause typ es with resp ect to scram

bling It will b e shown that in general islandho o d of various clause typ es is determined

by the selectional prop erties of the clause as argued by Cinque for whmovement

In section I examine discourse constraints on scrambling I argue that the relevant

discourse notion constraining word order is presupp ositionality in the sense dened in

Diesing rather than sp ecicity as argued by Moltmann Maha jan

and Enc

Island eects on scrambling

In this section I examine the islandho o d of various typ es of clauses with regard to long

distance scrambling in Korean whcomplement clause prop ositional complement clause

complement clause of a noun pure complex NP sub ject clause relative clause and adjunct

clause It will b e shown that clauses which are not selected byaverb eg adjunct and

relative clauses are strong islands while those which are selected byaverb eg various

complement clauses are either weak islands or not islands Barrierho o d of each category

with regard to scrambling in Korean is consistent with Cinque s denition of a barrier

which hinges on the notion of selection Even though I have only considered scrambling of

arguments so far I will consider scrambling of b oth arguments and adjuncts in this section

t as closely as p ossible This is to compare the b ehavior of scrambling and whmovemen

without necessarily committing to an analysis which treats scrambling of adjuncts in the

same wayasscrambling of arguments

I divide the data into two sub categories Scrambling of denite phrases in section

and scrambling of whphrases in section At the end of each section I present the

results of the questionnaire survey I conducted with native sp eaker informants

Due to the lackofovert syntactic whmovement in Korean movement of a whphrase has b een treated

as a sub case of scrambling

Scrambling of denite expressions

Consider through Examples in a are scrambling of a complement in b

scrambling of a manner adverbial and in c scrambling of a reason adverbial

Scrambling out of a prop ositional complement clause

a ku namca lul Youngheeka nwuka t coahantako malhayssni

i i

that manacc Youngheenom whonom likecomp saidq

That man Younghee said who likes t

i i

b ilensikulo Minhoka nwuka t sakilul chintako malhayssni

i i

This way Minhonom whonom cheatingacc docomp saidq

This way Minho said that who cheats t

i i

wuka t ipsiey c ikesttaymwuney Minhoka n

i i

thisb ecause Minhonom whonom entrance examloc

tteleciesstako malhayssni

failedcomp saidqm

For this reason Minho said that who failed in the entrance exam t

i i

AsImentioned in the b eginning of this chapter scrambling is constrained by discourse contexts There

fore scrambling without an appropriate discourse context sounds rather marginal compared to syntactic

whmovement The same situation is observed in topicalization in English Long distance topicalizati on in

English without any discourse context sounds awkward even though it is a p erfectly grammatical pro cess

as illustrated in ii

i Who do es John think that Mary likes t

i i

ii John Susie thinks that Mary likes t

i i

Scrambling out of the complement clause of a noun

a ku nyesek eykey nanun Youngheeka t holttak ppacie isstanun

i i

that guydat Itop Youngheenom completely is fallen in lovemod

sasili anmitecinta

factnom notb elieve

With that guy I cannot b elieve the fact that Younghee is fallen in lovet

i i

b ilensikulo nanun Minhoka t nalul kimanhaysstanun sasili

i i

this way Itop Minhonom Iacc cheatedmod factnom

anmitecinta

notb elieve

This way I cannot b elieve the fact that Minho cheated me t

i i

c kyewu ilen iyulo nanun Minhoka t nalul miwuehantanun

i i

top Minhonom Iacc hatemod only this reasonwith I

sasili anmitecinta

factnom notb elieve

Merely for this reason I cannot b elieve the fact that Minho hates me t

i i

Scrambling out of a whcomplement clause

a ku chayk ul Minhonun nwuka t hwuchiekassnunci anta

i i

that b o okacc Minhotop whonom stealwhether know

That b o ok Minho knows who stole t

i i

b ilensikulo Minhonun nwuka t sakilul chinunci anta

i i

this wayinst Minhotop whonom cheatingacc dowhether know

This way Minho know who cheats t

i i

c ilen iyulo Minhon un nwuka t haykotanghayssnunci anta

i i

this reasoninst Cheslwutop whonom got redwhether know

For this reason Minho knows who got red t

i i

Scrambling out of a sub ject clause

a ku namca lul Youngheenun amwulato t coahantanun sasili

i i

that manacc Youngheetop anyone likemod factnom

nollaptako malhayssta

surprisingcomp said

The man Younghee said that the fact that anyone likes t is surprising

i i

b ilen sikulo Youngheenun nwukwunkakat salanamulswu issesstanun

i i

this wayinst Youngheetop someonenom survive couldmod

sasili nolaptako malhayssta

factnom surprisingcomp said

This way Younghee said that the fact that someone could survivet is surprising

i i

un c ilen iyulo Youngheenun naykat cakilul miwuehantan

i i

this reasoninst Youngheetop Inom selfacc hatemod

kesi nolaptako malhayssta

factnom surprisingcomp said

For this reason Younghee said that the fact that I hate her t is surprising

i i

Scrambling out of a relative clause

a ku namca eykey nanun Youngheeka t e ssun pyenci lul

i i j j

that mandat Itop Youngheenom writerel letteracc

mollay hwumchiep oassta

furtively lo oked at

I furtively lo oked at the letter whichYounghee wrote to the man

b kulen sikulo nanun e t kwuenlyekey apwuhanun salam ul

i j i j

that wayinst Itop the p o wer attermod p ersonacc

kyengmyelhanta

despise

That way I despise a p erson who atters the p ower t

i i

c ilen myengmokulo nanun e t phoklyekul hayngsahanun salam ul

i j i j

this reasoninst Itop violenceacc resort tomod p ersonacc

kyengmyelhanta

despise

For this reason I despise a p erson who resorts to violence t

i i

Scrambling out of an adjunct clause

a caki tongsayng ul Minhonun amwulato t ttaylimyen

i i

self s brothernom Minhotop anyone hitif

pro kamantwuci ankeysstako malhayssta

revengecomp said

Self s brother Minho said that if anyone t hits then he would revenge

i i

b ilen sikulo Minhonun pro nalmata t kongpwuhamyen

i i

this wayinst Minhotop every day studyif

pro iltungul halkesilako malhayssta

rank oneacc will docomp said

This way Minho said that if I study t everydaythenIwillbenumb er one

i i

yen c ilen mokcekulo Minhonun naykat mokumwuntongul ham

i i

this purp oseinst Minhotop Inom fundraisingacc doif

motwuka hyep cohalkesilako malhayssta

everyonenom helpwillcomp said

With this purp ose Minho said that if I did fundraising t

i i

then everyone would help

The following is the summary of the data

Scrambling of an ob ject out of a prop ositional complement whcomplement sub ject

clause or complement clause of a noun is grammatical

Scrambling of a mannerreason adverbial out of a prop ositional complement wh

complement clause complement clause of a noun or sub ject clause is slighly marginal

as indicated by

Scrambling of an ob ject out of a relative or adjunct clause is pretty bad as indicated

by

Scrambling of a mannerreason adverbial out of a relative or adjunct clause is un

grammatical as indicated by or

Table is the result of the questionnaire survey I conducted to examine island eects

on scrambling There were a total of ten informants The numb ers on the left hand side

of each column are the numb er of informants who accept the scrambled sentence The

I thank the following informants for participatin g in my questionnai re survey HeeRhak Chae Daeho

Chung JeongShik Lee SungKi Suh HaeHak Yo on EunJung Yo o DongIn Cho ChangBong Lee Hyo

Kon Kim So oKyung Huh NoJo o Kim MyungKwan Park JinYoung Choi Mo onYurl Jung SeoYoung

Chae JongCheol Park Wonchul Park Inhye Kang JeeIn Kim Hyun Ahn I am particularly grateful to

JinYoung Choi DongIn Cho and HeeRhak Chae for their help in conducting the surveyTo calculate

the results in Table and Table in the next section I have included only the resp onses of the rst

informants in the list of whom are sp ecializin g in syntax

numb ers on the righthand side of each column are the degree of go o dnessbadness of the

scrambled sentence go o d marginal bad ungrammatical The judgments

of the informants are consistent with the summary given ab ove Clauses which are not

selected byaverb eg adjunct and relative clauses are strong islands while various typ es

of complement clauses are either not islands or very weak islands

whcomp propcomp subject relative adjunct

object

manner

reason

Table Degree of acceptability wrt scrambling of denite expressions

Scrambling of whexpressions

a examples illustrate scrambling of a complement b of a manner adverbial and c of

a reason adverbial

Scrambling out of a prop ositional complement clause

a nwukwu lul Youngheenun Minhoka t michitolok salanghantako

i i

whoacc Youngheetop Minhonom madly lovecomp

malhayssni

saidqm

Who did Younghee say that Minho loves t madly

i i

b ettehkey Minhonun naykat sayngkeylul yucihaykantako

i i

how Minhotop Inom livingacc maintaincomp

malhayssni

saidqm

How did Minho say that I make a living t

i i

c way Minhonun naykat haykolul tanghaysstako malhayssni

i i

why Minhotop Inom reacc aectedcomp saidqm

Why did Minho say that I got red t

i i

Scrambling out of a pure complex NP complement clause

a nwukwu lul nenun Minhoka t coahantanun sasiley nollassni

i i

whodat youtop Minhonom likesmod factat surprised atqm

who are you surprized by the fact that Minho loves t

i i

b ettehkey nenun Minhoka tonul p elesstanun sasiley nollassni

i

how youtop Minhonom moneyacc earnedmod factat surprized atqm

How are you surprized by the fact that Minho made money t

i i

c way nenun Youngheeka t tayhakipsilul phokihaysstanun

i i

why youtop Youngheenom college examacc gave upmod

sasiley nollassni

factloc surprised

Why are you surprised by the fact that Younghee decided not to take the college

i

entrance exam t

i

Scrambling out of a whcomplement clause

a mwues ul Youngheenun nwuka t hwumchiekassnunci ani

i i

whatacc Youngheetop whonom stolewhether knowqm

What do es Younghee know who stole

b ettehkey Minhonun Youngheeka t sayngkyelul yucihanunci ani

i i

how Minhotop Youngheenom livingacc maintainwhether knowqm

How do es Minho know whether Younghee makes a living t

i i

oungheeka t mikwukulo ttenassnunci ani c way Minhonun Y

i i

why Minhotop Youngheenom Americadir leftwhether knowqm

Why do es Minho know whether Mary left for the US t

i i

Scrambling out of a sub ject clause

anwukwu ul Minhonun Youngheeka t seltukhaynayn kesi

i i

whatacc Minhotop Youngheenom p ersuadedmod thatnom

taytanhatako malhayssni

remarkablecomp saidqm

Who did Minho say that the fact that Younghee p ersuaded t is remarkable

i i

b ettehkey sensayngnimkkeysen pro t younge hoihoalul yensuphanun

i i

how teachertop sp oken Englishacc practicemod

kesi kacang hoykoacekilako malssumhasiessni

thatnom most ecientcomp saidqm

How did the teacher say that practicing sp oken English t is most ecient

i i

mokumwuntongul hanun kesi c Minhonun Youngheeka way

Minhotop Youngheenom why fundraisingacc domod thatnom

elisektako malhayssni

b e stupidcomp saidqm

Did Minho say that why Younghee do es fundraising was stupid

Scrambling out of a relative clause

a nwukwu eykey nenun Youngheeka t ssun pyencilul

i i

whodat youtop Youngheenom writerel letteracc

mollay hwumchiep oassni

furtively lo oked atqm

To whom did you lo ok at the letter whichYounghee wrote t

i i

b ettehkey Youngheenun e t namul to oacwunun salam ul

i j i j

w Youngheetop othersacc helprel p ersonacc ho

conkyenghani

resp ectqm

How do es Younghee resp ect a p erson who helps others t

i i

c nenun e Minholul way ttaylin salam eykey hanguyhayssni

j j

youtop Minhoacc why hitrel p ersonto argued withqm

Did you argue with the p erson who hit Minho why

Scrambling out of an adjunct clause

a nwukwu eykey nenun Youngheeka t malul pwutilttaymata

i i

whodat youtop Youngheenom sp eak towhenever

yaki oluni

get upsetqm

To whom do you get upset whenever Younghee sp eaks t

i i

b ettehkey Minhonun pro nalmata t kongpwuhamyen

i i

how Minhotop every day studyif

pro iltungul halkesilako malhayssni

rank oneacc dowillcomp saidqm

How did Minho say that if I study t everyday then I will b e numb er one

i i

c Minhonun nayka way mokumwuntongul hamyen

top Inom why fundraisingacc doif Minho

motwuka hyep cohalkesilako malhayssni

everyonenom helpwillcomp saidqm

Did Minho say that if I did fundraising why then everyone would help

The following is the summary of the data

Scrambling of an ob ject out of a prop ositional complement whcomplement or sub

ject clause is grammatical Scrambling of a mannerreason adverbial out of a wh

complement clause is also grammatical

Scrambling of an ob ject out of a sub ject clause and an adjunct clause is slightly

marginal as indicated by Scrambling of a mannerreason adverbial out of a

prop ositional complement clause is also slighly marginal

Scrambling of an ob ject out of the complement clause of a noun or out of a relative

clause is pretty bad as indicated by Scrambling of a manner adverbial out of a

sub ject clause is prettybad

Scrambling of a mannerreason adverbial out of a complement clause of a noun ie

factive island is ungrammatical as indicated by or

The reason adverb way why within a sub ject relative or adjunct clause is ungram

matical even without scrambling

Table is the result of the questionnaire survey The metho d of arranging the table

is the same as that in table Ns at the b ottom of a column for sub ject relative

and adjunct mean that an o ccurrence of the reason whphrase way in these clause typ es

is nonsensical This must b e due to the fact that the three clause typ es are islands for LF

whmov ement A whphrase o ccurring in these clause typ es has to move to COMP of the

matrix clause searching for a question morpheme cf section in Ch

whcomp propcomp subject relative adjunct

object

manner

reason ns ns ns

Table Scrambling of whexpressions

Summary

The following observation can b e made on the basis of the data in section and sec

tion

Scrambling of a complementisworse than scrambling of a mannerreason adverbial

out of the same clause typ e cf the adjunctargument asymmetry in whmovement

Rizzi Cinque

There is a clear contrast in islandho o d b etween a sub categorized eg complement

and a nonsub categorized eg adjunct and relative clause The former do es not

constitute an island while the latter constitutes a strong island Islandho o d of sub ject

clauses is somewhere b etween the two

There is a subtle dierence in island eects b etween scrambling of a denite and a

whexpression At the moment it is not clear to me what causes such a dierence

One clear dierence b etween islands for whmovement in English and those for scram

bling in Korean is that a whcomplement clause is not an island for the latter while

it is a weak island in the sense of Cinque for the former

Discourse constraints on scrambling

In this section I discuss discourse constraints on scrambling I examine the b ehavior of

scrambling in terms of three discourse notions namely referentiality sp ecicity and pre

supp ositionality I argue that the notion which adequately characterizes the elements un

dergoing scrambling is presupp ositionality cf Diesing rather than sp ecicityas

argued by Moltmann Enc and Maha jan In this section a scrambled

order only refers to a noncanonical order

tiality and scrambling Referen

The notion of referentiality has b een the topic of muchrecent discussion in relation to long

whmovement cf Pesetsky Rizzi Cinque Kro ch Referentiality

is closely related to discourselinking in the sense of Pesetsky As Ellen Prince

pc p oints out in its broadest sense the term referential is taken as evoking any entity

For a discussion of the dierences b etween reference and discourse reference refer to Heim

Phrase at issue sp ecicity others referentiality

John is lo oking for a Norwegian sp ecic referential

but she didnt show up new lecard

John is lo oking for a Norwegian nonsp ecic referential

but hell never nd one new lecard

I wish I were a Norwegian predicative nonreferential

ie had the Norwegian prop erty no lecard

ANorwegian can swim well nonsp ec generic referential

ie norwegians in general new lecard

The rst Norwegian to come in attributive referential

will win the prize new lecard

The rst Norwegian to come in referential referential

lo oks just likemyhusband old lecard

Table Indeniteness and Referentiality

cf Prince including a class entity Referentiality is distinct from the notion of

sp ecicity or deniteness For instance in the indenite NP a Norwegian is non

sp ecic but referential since it must evoke a discourse entity and is subsequently referred

to by the pronoun she

John is lo oking for aNorwegianandshe has to b e very tall

Table is due to Ellen Prince pc and gives an idea ab out the relation b etween refer

entiality and other notions

Rizzi argues following Chomsky that lexically selected adverbials

measure phrases and nominal parts of idioms receive nonreferential

roles and are therefore nonreferential

Mary dresses wel l

John weighs lbs

Imadeheadway on this pro ject

Cinque adds quanticational phrases suchasevery museum and no museum to non

referential expressions since quanticational expressions are nonreferential in nature

Examining the b ehavior of scrambling with regard to referentialitywe nd the following

pattern a scrambling of lexically selected adverbials and measure phrases is ungrammat

ical b scrambling of predicativ e expressions which are another typ e of nonreferential ex

pressions is ungrammatical c scrambling of nominal parts of idioms is in general marginal

but can b e improved byproviding an appropriate discourse context and d scrambling of

quanticational expressions is ne I will discuss each case b elow

Measure phrases

Verbs suchasnemta to b e b eyond and nakata to weigh sub categorize for amount com

plements and scrambling of amount complements is imp ossible as shown in and

a i chayki chen weniulto nemnunta

this b o oknom wennomacceven is more than

This b o ok costs more than wen unit of Korean currency

b chen wen iulto i chayi t nemnunta

i i

a Minhoka p oundna nakanta

Minhonom p oundas muchasweigh

Minho weighs as much as p ounds

b p oundna Minhokat nakanta

i i

Lexically selected adverbials

Scrambling of lexically selected adverbials do es not lead to obvious syntactic ungrammati

cality but changes the truthconditional semantics of the baseorder sentence This indicates

that lexically selected adverbials in princicple cannot b e scrambled

ipnunta a Kimi osul mesisskey

Kimnom clothesacc well wears

Kim dresses well

b mesisskey Kimi osul t ipnunta

i i

The lexically selected adverbial mesisskey is interpreted in a particular wayasshown in

the translation Scrambling of the adverbial as in b changes the meaning of the

sentence which I indicate by The predicate in the canonical order sentence ais

interpreted generically but that in the scrambled order sentence bisinterpreted as an

instantaneous action mo died by the scrambled adverbial

PredicativeNPs

Predicates suchastoyta to b ecome and pwulkoahata to b e nothing but sub categorize for

predicative complements Scrambling of the predicative complements is pretty marginal

a Minhoka ilkay piseey pwulkoahata

Minhonom onederog secretaryloc is nothing but

Minho is nothing but a secretary

b ilkaypiseey Minhokat pwulkoahata

i i

a Minhoka taythonglyengulo toyessta

Youngswunom presidentdir became

Minho b ecame the president

b taythonglyeng ulo Minhokat toyessta

i i

The same restriction holds for predicative nominals in small clauses as in

a nayka Inholul chinkwulomantulesstasamassta

Inom Inhoacc frienddir madeto ok

I made Inho a friendI to ok Inho as a friend

b nayka chinkwulo Inholul t mantulesstasamassta

i i

c chinkwulo nayka Inholul t mantulesstasamassta

i i

Idiom chunks

Complement NPs in idioms suchasmiyekkwukul mekta fail in exam and nwuntokul tulita

keep ones eyes on cannot b e scrambled as illustrated in and

a Minhoka ipsieyse miyekkukul mekessta

Minhonom entrance exam failed

Minho failed an entrance exam

b miyekkuk ul Minhoka ipsieyse t mekessta

i i

a Minhoka Youngheeeykey nwuntokul tuliessta

Minhonom Youngheedat kept his eyes on

es on Younghee Minho kept his ey

b nwuntok ul MinhokaYoungheeeykey t tuliessta

i i

Even though scrambling of the nominal part of an idiom in general yields an ungram

matical sentence when it is preceded byacontext which enables us to anticipate the use

of the idiom the acceptability of the scrambled sentence improves signicantly This is

illustrated by the contrast b etween b and b

a Minhoka kimchikwukpwute masinta

Minhoka kimchi soupfrom drink

Minho thinks of getting things in advance

b kimchikwukpwute Minhokat masinta

i i

The idiomatic reading in the base order sentence a disapp ears in the scrambled order

sentence b However if the scrambled sentence is preceded by a prop er discourse

context then the idiomatic reading is still available as in b

The unacceptability of b merely indicates that the idiomatic reading in the base order is not

available in the scrambled order Sentence b is p erfectly acceptable under the literal reading Minho

drinks kimchi soup rst

a ttek cwul salamun sayngkakcito annuntey

Rel

rice cakegiverel p ersontop think negdo

Minhoka kimchikwukpwute masinta

Minhonom think of getting things in advance

Before the p erson who can oer things decides Minho thought ab out getting it

b ttek cwul salamun sayngkakcito annuntey

Rel

kimchikwukpwute Minhokat masinta

i i

In the expression ttek cwul salamun syangkakcito annuntey precedes the sentence

containing the idiom whichistypically used in combination with the idiom The sentence

containing the idiom has only the idiomatic reading in this case Scrambling of the nominal

part of the idiom as in b do es not aect the idiomatic reading of the sentence

Quanticational expressions

Most quanticational expressions undergo b oth lo cal and longdistance scrambling as il

lustrated in and

enu haksayngina Kim kyoswukat colepsayng hoansonghoiey

i i

every studentuq Prof Kimnom graduating studentfarewell partyto

chotayhayssta

invited

Prof Kim invited every student to the farewell party for graduating students

enu haksayngina Minhonun Kim kyoswukat colepsayng

i i

every studentuq Minhotop Prof Kimnom graduating student

hoansonghoiey chotayhaysstakomalhayssta

farewell partyto invitedcomp said

Minho said that Prof Kim invited every student to the farewell party for

graduating students

and are examples of lo cal and long distance scrambling of the universally quan

tied expressions enu haksayngina resp ectively

To summarize this section some nonreferential expressions cannot b e scrambled eg lex

ically selected adverbials measure phrases and predicative NPs while others can eg quan

ticational expressions Nominal parts of idioms can b e scrambled provided that there is

a prop er discourse context preceding the scrambled sentence In addition nonnominal

clausal complements which are not referential can b e scrambled From this I conclude that

referentiality is the right notion characterizing the nature of scrambled elements

Sp ecicity and scrambling

Anumb er of authors including Moltmann Enc Maha jan hav e argued

that only sp ecic elements can b e scrambled In this section however I will argue that this

claim cannot b e maintained I follow Moltmann and dene sp ecicity as in

A sp ecic NP refers to an entity which the sp eaker assumes to b e familiar to the

addressee in the context of communication A nonsp ecic NP refers to an

entity which is relevant in the universe of discourse

Scrambling of Indenites

Most arguments in supp ort of the claim that only sp ecic elements can b e scrambled involve

scrambling of indenites

Enc argues that the accusative case particle in Turkish marks the sp ecicityof

the entity Therefore the ob ject in whichismarked accusative is sp ecic while the

bare ob ject in is nonsp ecic

Ali bir kitabi aldi

Ali one b o okacc b ought

A b o ok is such that Ali b ought it

Ali bir kitap aldi

Ali one b o ok b ought

Ali b ought some b o ok or other

Scrambling of the sp ecic ob ject in is acceptable as in while scrambling of the

nonsp ecic ob ject in is unacceptable as in

Bir kitab i Ali t aldi

i i

Bir kitap Ali t aldi

i i

Moltmann also argues for German that only sp ecic elements can b e scrambled

For instance the complementofaverb that imp oses deniteness eects cannot b e scrambled

in German as illustrated in and which are and of Moltmann

resp ectively

a weil Hans wohl ein Freund von Bill ist

b ecause Hans presumably a friend of Bill is

b ecause Hans is presumably a friend of Bill

b weil Hans ein Freund von Bill wohl t ist

i i

a weil Maria wohl eine Schwester hat

b ecause Maria presumably a sister has

b ecause Maria presumably has a sister

b weil Maria eine Schwester wohl t hat

i i

Diesing convincingl y argues that Encs notion of sp ecicity is b etter characterized as

presupp osition al ity which will b e discussed in the next section

Sp ecicity apparently aects scrambling in Korean also In the insitu indenite

ob ject yumyeng violinist is ambiguous b etween sp ecic and nonsp ecic After scrambling

of the indenite ob ject however only the sp ecic reading remains as shown in ab

Minhoka pro lotte hoteleyse yumyeng violinistlul p oasstako calanghayssta

Minhonom lotte hotelloc famous violistacc sawcomp said proudly

Minho said proudly that he saw a famous violinist at Hotel Lotte

a Minhoka yumyeng violinist lul lotte hoteleyse t p oasstako calanghayssta

i i

b yumyeng violinist lul Minhoka lotte hoteleyse t p oasstako calanghayssta

i i

Minho said proudly that he saw a sp ecic violist at Hotel Lotte

Likewise in the indenite ob ject is either sp ecic or nonsp ecic and only the non

sp ecic reading remains after scrambling as in ab

Minhoka cantipateyse chaykul ilkessta

Minhonom grassloc b o okacc read

Minho read a sp ecic or nonsp ecic b o ok on the grass

ul cantipateyse t ilkessta a Minhoka chayk

i i

b chayk ul Minhoka cantipateyse t ilkessta

i i

Minho read a sp ecic b o ok on the grass

In summary examples like b and b seem to indicate that sp ecicity is the

relevant discourse notion constraining scrambling

In Korean the indenitenes s of an NP is not morphological ly marked in general unlike English

Therefore calling the ob ject NP yumyeng violinist in an indenite is strictly sp eaking misleading

Rather the crucial asp ect of the phrase for the present discussion is that it can b e either sp ecic or non

sp ecic in the base order but can only b e sp ecic after scrambling just like indenites in other scrambling

languages

The dierence in the numb er of p ossible readings of an NP b efore and after scrambling mightbetaken

as evidence for scrambling b eing movement in addition to the scop e reconstruction discussed in Ch

Lenerz discussing scrambling in German argues that the base order of a sentece is felicious in every

context ie default word order while a scrambled order is felicitous only in particular discourse contexts

According to Lenerzs diagnostic is the base word order and is its scrambled counterpart Dis

cussing the canonical p osition of lo catives in existential sentences in Japanese as in a and b Kuno

also gives a similar argument for identifying the base order of a sentence in terms of the numb er of readings

available for a topicmarked NP When a sub ject NP is followed by the topic particle wa ifitiseither

generic or anaphoric b oth the thematic and the contrastiveinterpretation result On the other hand if

a nonsub ject NP is followed by wa ordinarily only the contrastiveinterpretation results Based on this

Kuno argues that if a topicmarked NP in sentence initial p osition has only the contrastive reading then it

is derived by scrambling movement

a Teiburuno ueni koppukaaru

tablegen toploc cupnom exist

There are cups on the table

b koppuka teiburuno ueni aru

cupnom tablegen toploc exist

Counterexamples

However there are a numb er of examples in Korean which run counter to the claim that

only sp ecic elements can b e scrambled

Lee a notes that the indenite quantier nwukwuinka someone can only b e non

sp ecic Nevertheless it freely undergo es scrambling as in

nwukwuncinun moluciman nwukwuinka lul Minhoka t salanghako issta

i i

whonmztop dont knowbut someoneacc Minhonom love

I dont know who he is but someone Minho loves

More examples whichinvolve scrambling of nonsp ecic arguments are given b elow

caki chinkwutulcwunguy myechmyeng ul Minhoka t

i i

self friendsamonggen someclacc Minhonom

naysayngil pathiey chotayhal kesita

my birthdaypartyloc will invite

Some of his friends Minho will in vite to my birthdayparty

hako manhun mwulkencwungeyse yangmal han kyelley lul Minhoka

i

numerousmod thingamongloc apairofsocksacc Minhonom

naeykey t senmwullo cwuessta

i

Idat presentas gave

Of all the things he could havegottenapairofsocks Minho gave to me as a present

The scrambled phrase in can b e only interpreted as nonsp ecic which is due to the

inherently nonsp ecic expression myech myeng some numb er of p eople Nevertheless

the sentence is p erfectly acceptable Likewise the scrambled phrase yangmal han kyel ley a

pair of so cks in is nonsp ecic

ttokkatun sasil lul nawa Minhoka t talukey kiswulhayssta

i i

samemod factacc Iand Minhonom dierently describ ed

The same fact Minho and I describ ed dierently

In which is due to Ellen Prince pc the scrambled phrase tookkatnun sasil is

ambiguous b etween b eing sp ecic and nonsp ecic In the sp ecic reading the hearer knows

what the fact is and in the nonsp ecic reading the hearer do esnt know what the fact is

Generic expressions can also b e scrambled even though they are not sp ecic as in

and

sakoa lul Minhoka t coahanta

i i

applesacc Minhonom like

Apples Minho likes

inkan ul sikincongi t meknunta

i i

human b eingacc cannibalsnom eat

Human b eings cannibals eat

In summary examples suchas suggest that the claim that only sp ecic

expressions can b e scrambled cannot b e maintained In scrambled expressions

are nonsp ecic In the scrambled expression can b e b oth sp ecic and nonsp ecic

In generic expressions are scrambled

Presupp ositionality and scrambling

I will argue that the discourse notion which correctly characterizes the prop erty of elements

undergoing scrambling is presupp ositionality as dened in Diesing

Presupp ositionality

I summarize the notion of presupp ositionality discussed in Diesing Citing Milsark

Diesing distinguishes twotyp es of determiners namely strong and weakWeak determiners

can app ear with a sub ject NP in thereinsertion contexts while strong determiners cannot

as shown in

a There isare asomea fewmany y ies in my soup

b There isare theeveryallmost y ies in my soup

Another dierence b etween strong and weak determiners is that strong determiners pre

supp ose the existence of the entities they are applied to Weak determiners are ambigu

ous b etween a presupp ositional reading and a nonpresupp ositional reading in which they

merely assert the existence of whatever entities they are applied to The ambiguityofweak

quantiers is illustrated b elow

a There are some ghosts in my house

unstressed some asserts existence of ghosts

b Some ghosts are in the pantry the others are in the attic

presupp oses the existence of ghosts

In a the nonpresupp ositional or cardinal reading of the determiner some is shown If

there are ghosts the sentence is true If ghosts turn out not to exist the sentence will b e

false In b the determiner is stressed and the presupp ositional reading is most salient

This sentence presupp oses the existence of ghosts This presupp ositional reading unlike

the cardinal reading can b e paraphrased as a partitive three of the ghosts

ve only the presup Strong determiners on the other hand are unambiguous They ha

p ositional reading The cardinal reading is not p ossible for the sentences in

a Every ghost roasted marshmallows

b Most ghosts sleep late

Explaining the data

I reexamine the data discussed in the previous sections in relation to referentiality and

sp ecicity and argue that all the data can b e accommo dated if wehyp othesize that only

presupp osed elements can b e scrambled

Diesing assumes that if there is no entityreferredtoby a strong quantier then the truth value of the

sentence is undened along the tradition of Fregean logic Therefore the absence of ghosts in b and

ab leaves the truth value of the sentences undened However according to the Russellian view of

presupp osition a failure to satisfy the presupp ositi on of an expression or assertion simply leads to the falsity

of the sentence cf Levinson

Let us rst consider the data discussed in section Fivetyp es of expressions are con

sidered namely lexically selected adverbials measure phrases predicative nominals nomi

nal parts of idioms and quanticational phrases Among these quanticational phrases can

b e presupp ositional as discussed ab ove while lexically selected adverbials measure phrases

and predicative nominals are attributive and cannot b e presupp osed Therefore the non

scramblability of the latter three typ es of phrases and the scramblability of quanticational

phrases are easily explained under the hyp othesis that only presupp osed elements can b e

scrambled For the scramblability of nominal parts of idioms we can give the following

explanation In general nominal parts of idioms are not presupp osed and therefore cannot

be scrambled cf b and b However when the sentence containing an idiom is

preceded byacontext whichtypically coo ccurs with the idiom the o ccurrence of the id

iom is presupp osed and therefore scrambling of the nominal part of the idiom is felicitous

cf b

Turning to the examples whichinvolve scrambling of nonsp ecic elements they can

also b e explained byhyp othesizing that presupp ositionality is the necessary condition for

scrambling The scrambled phrase nwukwuinka in is always presupp ositional as well

as nonsp ecic Furthermore the preceding sentence gives a clear indication of the pre

supp osition of the entity referred to by the scrambled phrase The scrambled phrase caki

chinkwutulcwunguy myechmyeng some of his friends in is partitive and parititives

are always presupp ositional The scrambled phrase yangmal han kyel ley a pair of so cks

e and the preceding sentence accommo dates the presupp ositional in is also partitiv

reading of the scrambled phrase The scrambled phrase ttokkatun sasil the same fact in

is denite which presupp oses the existence of the entity referred to Note that sp ecic

elements are necessarily presupp osed under the denition of sp ecicitygiven in and

therefore constitute a subset of presupp osed elements

In conclusion the relevant discourse notion characterizing the elements whichcanbe

scrambling is presupp ositionality That is the sentence presupp oses the existence of the

element which undergo es scrambling Before I close this section I would like to mention

another fact which might at rst glance seem problematic for the prop osal that only pre

supp osed elements can b e scrambled This is scrambling of a prop ositional complement as

illustrated in b

a Kim kyoswuka Minhoka chencaylako mitnunta

Kim profnom Minhonom geniuscopcomp believe

Prof Kim b elieves that Minho is a genius

b Minhokachencaylako Kim kyoswukat mitnunta

i i

In b the prop ositional complementisscrambled The question is if the scrambled

prop ositional complement is presupp osed as I have b een arguing If so what would b e the

nature of the presupp osition I argue that the scrambled clause is indeed presupp osed

What the sentence presupp oses in this case is the existence of Prof Kims b elief When the

sen tence is in the base order as in a the sentence is ambigous b etween presupp osing

the existence of Prof Kims b elief and not having such a presupp osition When the sentence

is scrambled as in b only the presupp ositional reading survives We observe a similar

fact in topicalization of a prop ositional complement in English illustrated b elow

I am grateful to Eric Fosler for providing his intuition on this matter

a Sue b elieves that the sky is green

b That the sky is green Sue b elieves t

i i

The base order sentence aisambiguous in the same wayasa Namely the

sentence either presupp oses the truth of Sues b elief or it do esnt The topicalized sentence

b however has only the presupp osed reading More sp ecically the b elief that the

sky is green contrasts with other prop ositions which are the candidates for Sues b elief

In short scramblability of a prop ositional complement do es not constitute a counterex

ample to the claim that only presupp osed elements can b e scrambled

Biblio graphy

Ahn Ahn HeeDon Multidominance structure scrambling and the theory of

grammar in Korean In Papers from the th regional meeting of the Chicago Linguistic

Society

Ahn and Yo on Ahn HeeDon and Yo on HangJin Functional categories in

Korean In Kuno Susumu et al editors Harvard studies in Korean linguistics III pages

Hanshin Publishing Company Seoul

Aoun et al Aoun J Hornstein N and Lightfo ot D Some asp ects of wide

scop e quantication Journal of Linguistic Research

Aoun and Hornstein Aoun Joseph and Hornstein Norb ert Quantier typ es

Linguistic Inquiry

Baker Baker Mark Incorporation A theory of grammatical function changing

University of Chicago Press Chicago

Barss Barss Andrew Paths connectivity and featureless empty categories

In Constituent structure Proceedings of the GLOW ConferenceForis Press Dor

drecht

Becker et al Becker Tilman Joshi Aravind and Rambow Owen Long

distance scrambling and Tree Adjoining Grammars In Proceedings of the th Conference

of the European Chapter of the Asso ciation of Computational Linguistics

Belletti and Rizzi Belletti Adriana and Rizzi Luigi Psychverbs and

theory Natural language and linguistic theory

Bratt Bratt Elizab eth Owen The structure of the Korean causative Pap er

presented at the LSA annual meeting Philadelphi a

Carlson Carlson G A unied analysis of the English bare plural Linguistics

and Philosophy

Cho a Cho DongIn a Multiple accusative constructions and verb movement

Pap er presented at the Student Conference in Linguistics

Cho and Morgan Cho JaeOhk and Morgan Jerry The interaction of syntax

and morphology in Korean VP co ordination In Kuno Susumu et al editors Harvard

studies in Korean linguistics II pages Hanshin Publishing Company Seoul

Cho b Cho JaiHyong b Three typ es of scrambling reconstruction crossover

and anaphor binding To app ear in the pro ceedings of the th International Conference

on Korean Linguistics

Cho Cho JaiHyoung Scrambling and crossover in Korean Presented at the

Harvard Workshop on Korean Linguistics

Cho and Sells Cho YoungMee and Sells Peter A lexical account of phrasal

suxes in Korean Ms Stanford University

Cho e Cho e HyonSo ok SuccessiveCyclic Rightward Movement in Korean

In Kuno Susumu et al editors Harvard studies in Korean linguistics II pages

Hanshin Publishing Company Seoul

Cho e Cho e HyonSo ok Restructuring parameters and complex predicates A

transformation approach PhD thesis MIT

Cho e Cho e HyonSo ok Fo cus and topic movement in Korean and licensing

Cho e Cho e JaeWo ong Pitchaccent and qwh words in Korean In Kuno

Susumu et al editors Harvard studies in Korean linguistics pages Hanshin

Publishing Company Seoul

Chomsky Chomsky Noam Conditions on transformations In Anderson

Stephen and KiparskyPaul editors A Festschrift for Morris Hal le MIT Press Cam

bridge MA

Chomsky Chomsky Noam Lectures in government and binding Studies in

ht generative grammar Foris Dordrec

Chomsky Chomsky Noam Know ledge of language its nature origins and

use Praeger New York

Chomsky Chomsky Noam Barriers Linguistic Inquiry Monograph MIT

Press Cambridge MA

Chomsky Chomsky Noam Economy of derivation and representations In

Freidin Rob ert editor Principles and parameters in comparative grammar pages

MIT Press Cambridge MA

Chomsky Chomsky Noam A minimalist program for linguistic theory

Cinque Cinque Guglielmo Types of Abar dependencies MIT Press

Deprez Deprez Viviane On the typology of syntactic positions and the nature

of chains Move to the specier of functional projections PhD thesis MIT

Diesing Diesing Molly The syntax of roots of semantic partition PhD thesis

University of Massachussetts Amherst

Dowty DowtyDavid Wordmeaning and Montague grammar D Reidel

Publishing Company Dordrecht Holland

Enc Enc Murv et The semantics of sp ecicity Linguistic Inquiry

Erku Erku Feride Discourse pragmatics and wordorder in Turkish PhD

thesis University of Minnesotta

Fillmore Fillmore Charles J The case for case In Bach Emmon and Harms

Rob ert T editors Universals in linguistic theory pages Holt Rinehart and Win

ston New York

Fo dor and Sag Fo dor Janet and Sag Ivan Referential and quanticational

indenites Linguistics and Philosophy

Frank et al Frank Rob ert Lee YoungSuk and Rambow Owen Scram

bling as nonop erator movement and the sp ecial status of sub jects In Barbiers S

den Dikken M and Levelt C editors Proceedings of the ThirdLeiden Conferencefor

Junior Linguists pages

Fukui Fukui Naoki Atheory of category projection and its applicationsPhD

thesis MIT

Fukui Fukui Naoki Deriving the dierences b etween English and Japanese

English Linguistics

Fukui Fukui Naoki Parameters and optionality To app ear in Linguistic

Inquiry

Fukui and Sp eas Fukui Naoki and Sp eas Margaret Sp eciers and pro jection

MIT working papers in linguistics

Fukushima Fukushima Kazuhiko GeneralizedFloating Quantiers PhD the

sis The University of Arizona

Giorgi and Longobardi Giorgi Alessandra and Longobardi Giusepp e The

syntax of Noun PhrasesCambridge University Press Cambridge Great Britain

Grimshaw Grimshaw Jane Argument Structure MIT Press Cambridge

MA

Grimshaw Grimshaw Jane Extended Pro jection Ms Brandeis University

Hale Hale Ken Preliminary remarks on congurationality Proceedings of

NELS

Hasegawa Hasegawa N Alexical interpretive theory with emphasis on the role

of subject PhD thesis University of Washington

Heim Heim Irene File change semantics and the familiarity theory of de

niteness

Heyco ck Heyco ck Caroline The structure of the Japanese causative Technical

rep ort MSCIS Department of Computer and Information Science Universityof

Pennsylvania

Heyco ck Heyco ck Caroline antiReconstruction and Referentiality Ms

Yale University

Heyco ck and Kro ch Heyco ck Caroline and Kro ch Anthony Verb movement

and co ordination in the Germanic languages Evidence for a relational p ersp ectiveon

licensing Presented in the Germanic Language Workshop Tromso Norway

Heyco ck and Lee Heyco ck Caroline and Lee YoungSuk Sub jects and pred

ication in Korean and Japanese In Ho ji Ha jime editor JapaneseKorean linguistics

pages CSLI Stanford

Ho ekstra Ho ekstra Eric Licensing Conditions on Phrase Structure PhD

thesis University of Groningen

Homan and Turan Homan Beryl and Turan Umit Deniz Scrambling in

Turkish In Lee YoungSuk OMaley Mary and Turan Umit Deniz editors Penn

Review of Linguistics

Ho ji Ho ji Ha jime LF constraints and congurational structure in Japanese

PhD thesis University of Washington

Hong Hong KiSun Discourse binding of the Korean Reexive caki In Kuno

S et al editors Harvard Studies in Korean Linguistics II pages Hanshin

Publishing Company Seoul Korea

Hong Hong KiSun Sub jecttoOb ject Raising in Korean In Dziwirek

al Relations A crosstheoretical perspective pages Katarzyna et al editors Grammatic

CSLI

Hong Hong KiSun The passive construction and case in Korean In Berkeley

Linguistic Society

Huang Huang ChengTeh James On the distribution and reference of empty

pronouns Linguistic Inquiry

Iatridou Iatridou Sabine Ab out AgrP Linguistic Inquiry

Iida Iida Masayo Case assignmentby nominals in Japanese In Wechsler

Steve and Zec Draga editors Working PapersinGrammatical Theory and Discourse

Structure CSLI Stanford

Kang Kang YoungSe Korean Syntax and Universal Grammar PhD thesis

Harvard University

Kang Kang YoungSe Stativity and nominative case marking in Korean In

Kuno S et al editors Harvard studies in Korean Linguistics II pages Hanshin

Publishing Company Seoul Korea

Kim Kim So oWon The QP status of whphrases in Korean and Japanese

In Proceeding of the th West Coast Conference on Formal Linguistics pages

Stanford University

Kim Kim YoungJo o The syntax and the semantics of Korean casePhD

thesis Harvard University

Kim Kim YoungJo o Unaccusativity in Korean

Kiss Kiss Katalin Congurationality in Hugarian DReidel Publishing Com

pany Dordrecht

Ko opman and Sp ortiche Ko opman Hilda and Sp ortiche Dominique The

p osition of sub jects Lingua

Kro ch Kro ch AnthonyS The semantics of scope in English PhD thesis

MIT

Kro ch Kro ch Anthony S Amount quantication referentiality and long wh

movement Ms University of Pennsylvania

Kuno and Robinson Kuno S and Robinson J Multiple wh questions Lin

guistic Inquiry

Kuno Kuno Susumu The p osition of lo catives in existential sentences Lin

guistic Inquiry

Kuno a Kuno Susumu a Constraints on internal clauses and sentential sub ject

Linguistic Inquiry

Kuno b Kuno Susumu b The structure of the Japanese language The MIT

Press

Kuno Kuno Susumu A further note on Tonoikes intrasub jectivization hy

p othesis MIT working papers in Linguistics

t and Kuro da Kuro da ShigeYuki What can Japanese say ab out governmen

binding In Barlow M et al editors WCCFL

Ladusaw Ladusaw W Polarity sensitivity as inherent scoperelationsPhD

thesis University of Texas at Austin

Laka Laka Miren Itziar Negation in syntax On the nature of functional

categories and projections PhD thesis MIT

Larson Larson Richard K BareNp Adverbs Linguistic Inquiry

Larson Larson Richard K On the double ob ject construction Linguistic

Inquiry

Lasnik and Saito Lasnik Howard and Saito Mamoru On the nature of prop er

government Linguistic Inquiry

Lasnik and Saito Lasnik Howard and Saito Mamoru Move Conditions

on its applications and output MIT Press Cambridge Massachusetts

Lasnik and Stowell Lasnik Howard and Stowell Tim Weakest crossover

Linguistic Inquiry

Lee Lee Chungmin Issues in Korean Anaphora Hanshin Publishing Compay

Lee a Lee Chungmin a Numeral classiers classiers and sp ecicity in Korean

Pap er presented at the LSA Summer Institute on Korean SyntaxSemantics Santa Cruz

Lee a Lee KapHee a On parasitic gaps in Korean In Kuno Susumu et al

editors Harvard studies in Korean linguistics III pages Hanshin Publishing

Company Seoul

Lee b Lee YoungSuk b The Korean causative A Tree Adjoining Grammar

analysis Technical rep ort MSCIS University of Pennsylvania Dept of Computer

and Information Science

Lee Lee YoungSuk An investigation of scrambling in Korean Proceedings of

the th International ConferenceonKorean Linguistics

Lee b Lee YoungSuk b Case array and word order variation in nominal clauses

In Ho ji Ha jime editor JapaneseKorean Linguistics CSLI Stanford University

Lee Lee YoungSuk Twotyp es of negative p olarity items in Korean Presented

at NELS

Lee and Santorini Lee YoungSuk and Santorini Beatrice Resolving We

belhuths paradox evidence from German and Korean In Proce edings of the Tilburg

Workshop on Scrambling

Lenerz Lenerz J Zur Abfolge nominaler Satzglieder im DeutschenVerlag

Gunter Narr Tubingen

Levin and Rappap ort Levin Beth and Rappap ort Malka The formation of

adjectival passives Linguistic Inquiry

Levinson Levinson Stephen Pragmatics Cambridge University Press

LondonNew York

Linebarger Linebarger Marcia Negativepolarity and grammatical represen

tation Linguistics and Philosophy

Maha jan Maha jan Ano op On the AAbar distinction scrambling and weak

crossover and binding in Hindi Ms MIT

Maha jan Maha jan Ano op The AAbar distinction and movement theory

PhD thesis MIT

Maling Maling Joan Adverbials and structural case in Korean In Kuno

Susumu et al editors Harvard studies in Korean linguistics III pages Hanshin

Publishing Company Seoul

Maling and Kim Maling Joan and Kim So oWon Case assignment in the

inalienable p ossession construction in Korean Journal of East Asian Linguistics pages

Marantz Marantz Alec On the natureofgrammatical relations Linguistic

Inquiry Monograph MIT Press Cambridge MA

May May Rob ert The grammar of quantication PhD thesis MIT

May May Rob ert Logical Form Its structure and derivation Linguistic

Inquiry Monograph MIT Press Cambridge MA

McCawley McCawley James D English as a VSO language Language

McConnellGinet McConnellGinet Sally Adverbs and Logical Form Lan

guage

Milsark Milsark G Existential sentences in English PhD thesis MIT

Miyagawa Miyagawa Shigeru Lightverbs and the ergativehyp othesis Lin

guistic Inquiry

Miyagawa Miyagawa Shigeru Functional category and case assignment Ms

Ohio State University

Miyagawa Miyagawa Shigeru Case realization and scrambling To app ear in

the pro ceedings of the Tilburg Workshop on Scrambling Holland

Moltmann Moltmann Friederike Scrambling in German and the sp ecicity

eect Ms MIT

Pesetsky PesetskyDavid Paths and Categories PhD thesis Massachusetts

Institute of Technology

Pesetsky PesetskyDavid Whinsitu movement and unselective binding In

Reuland E and ter Meulen A G B editors The representation of indeniteness

pages MIT Press Cambridge MA

Pollo ck Pollo ck JeanYves Verb movement Universal Grammar and the

structure of IP Linguistic Inquiry

Postal and Pullum Postal Paul and Pullum Georey Expletive noun phrases

in sub categorized p ositions Linguistic Inquiry

Prince Prince Ellen F Toward a taxonomyofgivennew information In Cole

Peter editor Radical pragmatics pages Academic Press New York

Progovac and Franks Progovac Ljiljana and Franks Steven Relativized SUB

JECT for reexives In Proceedings of NELS Amherst University of Massachusetts

Pullum Pullum Georey Implications of English extrap osed irrealis clauses

In Proceedings of ESCOLvolume

Riemsdijk and Williams Riemsdijk Henk van and Williams Edwin Anin

troduction to the theory of grammar MIT Press Cambridge

Riemsdijk and Williams Riemsdijk Henk van and Williams Edwin S NP

structure The Linguistic Review

Rizzi Rizzi Luigi Relativized minimality Linguistic Inquiry Monograph

MIT Press Cambridge MA

Rosenbaum Rosenbaum Peter S The grammar of English predicate comple

ment constructions MIT Press Cambridge MA

Ross Ross John Rob ert Constraints on variables in syntax PhD thesis

MIT

Ross Ross John Rob ert Three batons for cognitivepsychologyInWeimer

and Palermo editors Cognition and symbolic processesLawrence Erlbaum Asso ciates

Hillsdale NJ

Rothstein Rothstein Susan Pleonastics and the interpretation of pronouns

Ms

Saito Saito Mamoru Case marking in Japanese A preliminary studyMs

MIT

Saito Saito Mamoru Some Asymmetries in Japanese and Their Theoretical

Consequences PhD thesis MIT

Saito Saito Mamoru Scrambling as semantically vacuous Amovement In

ase structure Baltin Mark and Kro ch Anthony S editors Alternative conceptions of phr

pages University of Chicago Press Chicago

Saito Saito Mamoru Long distance scrambling in Japanese Journal of East

Asian Linguistics

Sells Sells Peter More on lightverbs and theta marking Ms Stanford

University

Sells Sells Peter Korean and Japanese Morphology from a Lexical Persp ective

Ms Stanford University

Shibatani and Kageyama Shibatani Masayoshi and Kageyama T Word for

mation in a mo dular theory of grammar Language

Sloan Sloan Kelly QuantierWH interaction In Cheng Lisa and Demirdash

Hamida editors MIT working papers in linguistics vol MITWPL Cambridge MA

Stowell Stowell Tim The tense of innitives Linguistic Inquiry

Stowell Stowell Timothy Origins of phrase structure PhD thesis MIT

Suh Suh SungKi In Proceedings of ESCOLvolume Ohio State

University Columbus OH

Tada Tada Hiroaki Scramblings Ms MIT

Ueyama Ueyama Ayumi Scrambling in Japanese as a uniform chain Pre

sented at the Tilburg workshop on scrambling

Uriagereka and Lasnik Uriagereka H and Lasnik H A Course in GB Syntax

MIT Press Cambridge MA

Watanab e Watanab e Akira Whinsitu sub jacency and chain formation

Ms MIT

Web elhuth Web elhuth Gert Syntactic saturation phenomena and the modern

Germanic languages PhD thesis UniversityofMassachusetts Amherst

Web elhuth Web elhuth Gert Scrambling without functional heads Ms Uni

versity of North Carolina

Whitman Whitman John String vacuous INFL to COMP Pap er given at

GLOW collo quium

Williams Williams Edwin S Predication Linguistic Inquiry

Williams Williams Edwin S A reassignment of the functions of LF Linguistic

Inquiry

Yang Yang DongWhee Hankwukeuy tayyonghoaAnaphorainKorean

Hankwuk yenkwuwen Korea ResearchCenter Seoul

Yang Yang InSeok Korean syntax case marking complementation and rel

ativization PhD thesis UniversityofHawaii

Yim Yim YoungJae Multiple sub ject constructions In Kuno Susumu

et al editors Harvard studies in Korean linguistics pages Hanshin Publishing

Company Seoul

Yo on Yo on James Some queries concerning the syntax of multiple sub ject

constructions in Korean In Kuno Susumu et al editors Harvard studies in Korean

linguistics pages Hanshin Publishing Company Seoul

Yo on Yo on James The grammar of inalienable p ossession constructions in

korean Mandarin and French In Kuno Susumu et al editors Harvard Studies in

Korean Linguistics III pages Hanshin Publishing Company Seoul Korea

Yo on Yo on James The grammar of inalienable p ossession constructions in

Korean Mandarin and French In Proceedings of the th Conference of the Northeast

Linguistic Society pages Carnegie Mellon University

Yo on and Yo on Yo on James and Yo on Jeongmi Morphosyntactic mis

matches and the functioncontent distinction In K Deaton M Noske and Ziolkowski

M editors The proceedings of Chicago Linguistics Society

Yo on and Yo on Yo on James and Yo on Jeongmi Chain condition ambigu

ityofgovernment and derivational grammars In Sherer Tim editor Proceedings of

the North Eastern Linguistic Society annual meeting pages Graduate student

linguistic asso ciation University of Massachusetts Amherst

Yoshimura Yoshimura Noriko Parasitic pronouns Pap er presented at the st

Southern California Conference on Japanese and Korean Linguistics