Optimizing Structure in Context: Scrambling And

Home , Scrambling (linguistics)

OPTIMIZING STRUCTURE IN CONTEXT

SCRAMBLING AND INFORMATION STRUCTURE

a dissertation

submitted to the department of linguistics

and the committee on graduate studies

of stanford university

in partial fulfillment of the requirements

for the degree of

doctor of philosophy

HyeWon Choi August

CopyrightbyHyeWon Choi

I certify that I have read this dissertation and that in

my opinion it is fully adequate in scop e and qualityas

a dissertation for the degree of Do ctor of Philosophy

Joan Bresnan

Principal Advisor

I certify that I have read this dissertation and that in

my opinion it is fully adequate in scop e and qualityas

a dissertation for the degree of Do ctor of Philosophy

Peter Sells

I certify that I have read this dissertation and that in

my opinion it is fully adequate in scop e and qualityas

a dissertation for the degree of Do ctor of Philosophy

Henriette de Swart

I certify that I have read this dissertation and that in

my opinion it is fully adequate in scop e and qualityas

a dissertation for the degree of Do ctor of Philosophy

Tom Wasow

Approved for the University Committee on Graduate

Studies iii

Abstract

This dissertation examines the free word order or scrambling phenomena in German

and Korean from the p ersp ective of constraintinteraction in Optimality TheoryTo

overcome the problems raised in singlecomp onent analyses in explaining word order

variation I prop ose an interface approachinwhich the constraints from several

dierent comp onents of grammar participate comp ete and interact with one another

That is various word orders are considered to b e motivated and constrained by

interactions among syntactic semantic and discourse principles of these languages

As the constraints from dierent mo dules of grammar are highly conicting I utilize

Optimality Theory to demonstrate how the constraints interact and resolve conicts

among one another In this approach each scrambled variant ie a sentence with a

particular word order is conceived of as the optimal output which instantiates the

syntactic semantic and discoursecontextual information given in the input

Irstdevelop the phrase structural constraints in German and Korean referred

to as canonwhich are resp onsible for the mapping from the argumentstructure and

grammaticalfunction information to the phrase structure conguration whichinturn

reects the surface word order Then I examine the semantic and discourse eects

of scrambling and prop ose a mo del of information structure based on the two cross

classifying discourse features New and Prom to capture the complex interactions

of topic and fo cus on word order The semantic eect of sp ecicity is also handled iv

in terms of information structure by means of semantic restrictions on discourse fea

ture assignment Based on this information structure I prop ose two information

structuring constraints new and prom which are the mapping constraints b etween

information structure and phrase structure as the ma jor driving forces of scrambling

Finally I demonstrate the interaction and conictresolution among these constraints

in the German and Korean scrambling data by prop osing a particular ranking for

each language v

Acknowledgments

Once in my life I went to see a fortune teller right b efore I came to Stanford She

said no worries go ahead and you will meet great teachers in the United States She

was right indeed I have b een blessed with great teachers at Stanford

Among the greatest are my dissertation committee memb ers My deep est grati

tude go es to my principal advisor Joan Bresnan for her insightful ideas invaluable

comments and discussions and untiring p ersonal encouragement I wish I could b e

such an advisor to my future students I must also thank for her nal push otherwise

Imay not have made it I havealways relied on Peter Sells not only for his prompt

comments but also for various academic and nonacademic advice He has b een no less

accessible although he was on sabbatical and the late nightoverseas email exchanges

with him have b een a nice source of refreshment I have b eneted a great deal from

the discussion with Henriette de Swart She was the p erson who rst got me inter

ested in topic and fo cus phenomena and her precise comments and criticisms help ed

me greatly with sharp ening my unorganized ideas Finally I am deeply grateful to

Tom Wasow for his interest in my topic careful reading and invaluable comments I

learned a lot from his own interest and knowledge on word order variation

I am also indebted to Ivan Sag Elizab eth Traugott and Gert Web elhuth who

was visiting Stanford in mynalyear of dissertation writing for their interest in my

work helpful comments and encouragement Also sp ecial thanks are due to my dear

professors in Korea SungHwan Lee and SeungHyuk Park at Ewha and IkHwan Lee vi

at Yonsei for rst letting me know ab out linguistics and moreover getting me caught

in it

Ihave b een privileged to havesuchwonderful friends and fellow students at Stan

ford I rst thank the classmates in myyear for making our rst few years so en

joyable Lynn Cherny Chris Manning Yo okyung Kim HinrichSchutze and Hadar

ShemTov Sp ecial thanks to Chris Manning for b eing such a go o d friend always

listening to me and helping me with every little thing

Heartfelt thanks go to my ocemates in Cordura It has b een such a pleasure

to share the space with them almost likehaving myown sisters around I thank Tracy

Holloway King for answering all my silly little question from formatting to writing

tips and esp ecially for pro ofreading my dissertation so carefullyTalking to Marga

Eugenia Nino has always given me comfort and relief I am also grateful to her

for letting me share her X terminal when mine went dead Finally I thank Rachel

Nordlinger for constantly encouraging and cheering me up and also for making our

oce such a fun place to b e in

I am esp ecially grateful to my German informants Susanne Riehemann and Mar

tina Faller for their prompt and patient resp onses through email or in p erson to my

endless questions I knowfrommyown exp erience how timeconsuming and mind

b oggling it can b e to b e an informant I am also indebted to other fellow linguists

for their intellectual and moral supp ort and also for their helpful comments and dis

cussions they are Alex Alsina Arto Anttila Jennifer Arnold Emily Bender Miriam

Butt Youngmee Yu Cho Vivienne Fong Eunjo o Han Jongb ok Kim Rob Malouf

and Changyong Sohn And sp ecial thanks to Eunjin Oh for ling this dissertation for

Among the family memb ers my rst thanks go to my mother Jong Lim Kim and

my father Duksan Choi They havealways b een encouraging and supp ortiveofmy

pursuing an academic career which not all Korean parents are to their daughters I vii

truly appreciate what they have done for me and I dedicate this dissertation to them as

a small token of my gratitude I thank my dear son Hongso on for putting up with mom

and dads absence in his second months of his life and I am grateful and indebted

to my motherinlaw Nam Ki Kim for her taking the b est care of him in that p erio d

Ihave appreciated having myhusband Sunhyuk Kim as a true friend and companion

throughout my life at Stanford He has b een supp ortive and understanding but also

critical as a true friend would b e Writing and also nishing our dissertations at the

same time was not the most pleasant exp erience weve had together but I am proud

that we survived it viii

Contents

Abstract iv

Acknowledgments vi

Intro duction

Overview

Theoretical Assumptions

Optimality Theory

Lexical FunctionalGrammar

Organization

Phrase Structure and Congurationality

Phrase Structure in German

Verb Second and Clause Structure

S and VP

Canonical Order and Phrase Structural Constraints

Canonical Word Order

Phrase Structural Constraints canon

Phrase Structure in Korean

Evidence for VP in Korean ix

Co ordination and Clause Structure

Summary

Sp ecicityFo cus and Information Structure

Semantic and Discourse Eects of Scrambling

Sp ecicity Eect

Antifo cus Eect

ContrastiveFo cus Eect

Information Structure

Information Packaging

Topic and Tail

ContrastiveFo cus and CompletiveFocus

Topic and ContrastiveFocus

Scrambling and Information Structuring Constraints

FeatureBased Information Structure

Information Structuring Constraints

Sp ecicity and Information Structure

Sp ecicity Eect Revisited

Sp ecicity and Information Status

Summary

Optimality Theory and Scrambling in German

Optimality Theory and Scrambling

Inputs and Information Structure

Outputs and Candidate Set

ConstraintsandRanking

Deriving SemanticDiscourse Eects in Scrambling

Information Context Typ es x

AntiFo cus Eect

ContrastiveFo cus Eect

Sp ecicity Eect

Summary

Order and Morphology Korean

Information Structuring byMorphology

nunTopic or ContrastiveFocus

nun as a Prom marker

Information Structuring byOrder

Reading Distribution and Relative Order

Scrambling motivated byOldness

Scrambling motivated by Prominence

Ambiguity Resolution byFurther Constraints

OptimalityTheoretic Interaction b etween Morphology and Order

ConstraintsandRanking

Cho osing b etween Topic and ContrastiveFocus

Summary

Conclusion

Summary and Concluding Remarks

Remaining Issues xi

Chapter

Intro duction

Overview

This dissertation examines free word order or scrambling phenomena in German

and Korean Word order in German or Korean is free in that fairly extensive alter

native constituent orders are p ossible Yet it is not totally free b ecause alternative

orders are not in absolute free variation they are also restricted and constrained so

that only certain orders are p ossible in certain semantic and discourse contexts

Variation is a dicult problem to deal with in a principled manner It is esp ecially

problematic for the theories which are based on certain principles governing a single

domain or comp onent of grammar Variation means more than one form is p ossible

Under standard assumptions of grammar where principles or rules are not violable

variation often requires either optional application of those principles or rules or

p osing some abstract or invisible elements or features whichwould satisfy those prin

ciples in the alternative cases Also if only a single comp onent or mo dule of grammar

is resp onsible for a certain phenomenon it is hard to understand whyvariation should

happ en Principles whichgovern one comp onent of grammar are usually dened to

yield one grammatical form and thus they have a common interest and therefore

CHAPTER INTRODUCTION

seldom conict with each other

Interestinglyseveral dierent comp onents of grammar have claimed that they are

resp onsible for surface word order Syntaxisanobvious one It has b een argued that

grammatical features such as case agreement and thematic roles are licensed in cer

tain phrase structural congurations so that the surface order is determined by those

features Recently it has b een claimed that word order variation or scrambling is

also driven bysuch grammatical features Maha jan Miyagawa Miyagawa

to app ear That is Case or agreement motivates scrambling a phrase is scrambled

so that its Case can b e licensed in a certain Sp ec p osition or a phrase is scrambled

so that its strong Agr feature can b e checked o Even apart from the problem

that more than one strong Case or Agreement p osition should b e p osed b ecause

scrambling unlike ob ject shift Holmb erg Vikner is not limited to a

single Sp ec p osition the problem of optionality of scrambling still remains hard to

handle It has to b e said that an insitu phrase do es not scramble b ecause Case can

optionally b e licensed in the base p osition the Case licensed in the base p osition is

dierent from that licensed in a scrambled p osition eg weak versus strong de Ho op

or the CaseAgr feature can optionally b e either strong or weak These

claims however are hard to test esp ecially when there is no overt evidence suchas

morphological case or agreement distinctions

Discoursepragmatics is another area in whichword order has long b een researched

Here it has b een argued that discoursecontextual information suchasgivennew and

familiarnovel determines the relative order among the comp onentelements Sgall et

al Li and Thompson Givon In most of the Praguean and func

tionalist analyses given information is argued to precede new information in the

sentence Sgall et al refer to this as Communicative Dynamism CD Con

stituents with more CD contain the newest most prominent information and app ear

CHAPTER INTRODUCTION

after constituents with less CD ie with older information However this is not al

ways the case Not all sentences are aligned this way and not all languages allow

rearrangement of the elements according to this typ e of information Although this

approach has an intuitive app eal the problem is that it is not always true at least

in German and Korean word order as we will see in this dissertation

Recently it has also b een argued that a certain ordering variant is semantically

driven de Ho op Diesing Diesing That is a phrase ie an indenite

phrase is scrambled to receive a sp ecic strong or presupp ositional interpreta

tion Diesing fairly strongly claims that scrambling as well as ob ject shift

is an instance of semanticallydriven movement as result of interpretation condi

tions applying in the syntaxsemantics mapping whichinducemovementAsisthe

case with other approaches optionality is a problem here to o a denite phrase can

b e in the base p osition regardless of its strength or presupp ositionality Moreover a

scrambled indenite phrase do es not always receive a strongpresupp ositional reading

see chapter for detailed discussion

To summarize none of the ab ove approaches are quite right although eachof

them has some truth in it I argue that the problem lies in that they try to explain

the phenomena in a single domain or comp onent of grammar I argue instead that

scrambling is due to the interplay of several dierent comp onents of grammar rather

than to a dominantcontrol byany one particular mo dule Sp ecically I argue that

scrambling is motivated and constrained by the interactions of syntax and discourse

and further constrained by semantics indirectly as illustrated in

CHAPTER INTRODUCTION

Syntax

word order

Semantics Discourse

The basic idea is that each comp onent of grammar has its own interests represented in

some form of principles or constraints and scrambling happ ens when the signals from

dierentcomponents mismatch For example in syntax one of its ma jor concerns is

to realize the argumentstructural or grammaticalfunctional information in terms of

phrase structure for instance sub ject should ccommand other phrases in German

and Korean On the other hand it is one of discoursepragmaticss ma jor concerns

to realize discoursefunctional information in terms of order topic should precede

fo cus Finally in semantics it is an issue whether a sp ecic or a nonsp ecic entity

has a discoursereferent and this indirectly inuences as to what can b e a topic or a

fo cus These principles or constraints in dierent comp onents of grammar could send

a conicting signal as to how to realize all the information and then the grammar

nally decides which surface structure would b e the b est way to compromise all the

information In a sense grammar divides its lab or into several dierent comp onents

in deciding its surface word order Reinhart

As a concrete to ol to demonstrate how the principles from dierent mo dules of

the grammar interact and resolve conicts among one another I utilize Optimality

Theory Optimization seems to b e the right concept to capture the idea of instan

tiation of the p otentially conicting information Esp ecially the notion of violable

constraints and the ranking of the constraints are quite crucial in the multicomp onent

CHAPTER INTRODUCTION

interface approach b ecause if constraints were not violable there could b e no com

promise and reconciliation and if constraints were not ranked it would b e hard to

know whichword order is the b est form p ossible for the given syntactic semantic

and discourse information

In the next section I briey lay out the theoretical assumptions I take throughout

this thesis I adopt a grammar of parallel copresent representations likeLFGtoshow

the activeinteractions b etween dierent comp onents of grammar involved in free word

order phenomena I rst present the core ideas and basic op erating mechanismsin

Optimality Theory and then the organization of the grammatical structures in LFG

Theoretical Assumptions

Optimality Theory

Optimality Theory Prince and Smolensky to app ear seems to b e an appropriate to ol

to deal with word order phenomena which I argue are a consequence of interactions

of several dierenttyp es of information formulated in the form of constraints from

dierent sectors of grammar b ecause it is designed to formally manage such p otential

conicts among constraints

In Optimality Theory a grammar is a function mapping each linguistic input to

its correct structural description or parse or output For example it maps an un

derlying phonological string to its proso dic parse Prince and Smolensky to app ear

Inputs consist of raw materials from which the candidate outputs are built see section

for more detailed discussion of input and output representations assumed in this

thesis In syntax inputs are usually taken to b e some skeletal structure containing

predicateargument information Grimshaw Grimshaw in press Legendre et al Legendre et al b Bresnan tenseasp ect sp ecication and certain

CHAPTER INTRODUCTION

semantic and pragmatic information Grimshaw in press Legendre et al Leg

endre et al b Grimshaw and SamekLo dovici

For every input GEN for generator a crosslinguistically universal function

generates the universe of p ossible candidate outputs or parses ie the candidate

set eg roughly a combination of Sstructure and LF in GB Legendre et al b

Pesetsky or a combination of functionalstructure and constituentstructure

in LFG Bresnan

Universal Grammar also provides a set of wellformedness constraints on outputs

Interestingly unlike principles of more familiar typ e in the previous theories the

constraints in OT are violable and actually many p ossible outputs violate one or

more constraints These constraints are in general highly conictingOTprovides

an explicit way to determine how these highly conicting and also violable constraints

interact with each other and resolve the conicts constraints are ranked According

to OptimalityTheory the grammar of a particular language ranks the universal

constraints in a languageparticular dominance hierarchy the grammar declares that

one constraint is more imp ortant than another and thus satisfying the former is more

imp ortant than satisfying the other As such Optimality Theory provides a general

means for constructing particular grammars from universal constraints

The grammatical forms or the optimal outputs among the candidates are se

lected according to howwell they satisfy the wellformedness constraints This pro cess

of selecting the optimal outputs is called EVAL These forms are not forced to exhaus

tively satisfy the entire constraint set In other words failure to satisfy a constraint

do es not necessarily result in ungrammaticality Rather a grammatical output is the

one that satises the constraints as b est as it can or optimal lyGiven the ranking of

the constraints the optimal output is the form which for every pairwise comp etition

involving it b est satises the highestranking constraint on which the comp etitors

conict Grimshaw in press This optimal output is the correct or grammatical

CHAPTER INTRODUCTION

parse of the given input

Supp ose that wehave a candidate set of A B and C and a universal set of con

straints C C and C Supp ose that the particular language L under consideration

has the following grammar The dominance hierarchy is represented by C C

states that C is ranked higher than C

GrammarL

C C C

Now lets assume that the candidates violate the constraints in the following pattern

candidates C C C

The constraints in the tableau are lined up from left to right according to their rank

ing the leftmost constraint is the highest When a candidate violates a constraint

C then it incurs a mark to the constraint Evaluation b egins with the highest

constraint The comp eting candidates A B and C are evaluated against the highest

constraintC and candidate A fails to satisfy it in Although it do es not vio

late any of the remaining lowerranked constraints it is dropp ed out of comp etition

b ecause it violates the highest constraint Candidate B and candidate C enter the

next comp etition Now candidate C violates the next highest constraintCand

thus it loses to candidate B Notice that candidate B also violates a constraint ie

C However it is still the optimal output which is represented as a smiling face

in this thesis b ecause it b est satises the highestranking constraintonwhich the comp etitors conict

CHAPTER INTRODUCTION

The core ideas in Optimality Theory ie constraints are violable and ranked and

optimal outputs do not necessarily satisfy all the constraints in the grammar provide

an insightful solution to the variation problem in word order Any candidate can

in principle b e an optimal output as long as it do es not violate a higher constraint

Dep ending on the input the constraint violation will varyandthus EVAL can yield

a dierent output The information in the input for word order can vary b ecause it

contains discoursecontextual information as well as grammatical information There

fore each alternative order can b e thought of as the optimal realization of the given

information in the input

Lexical Functional Grammar

This section outlines the basic theoretical assumptions of LFG a unicationbased

theory without movement Bresnan and Kaplan Bresnan Sells Bres

nan a The grammatical representation in LFG consists of several dierent

parallel structures no structure is derived from another structure and the relation

ship b etween structures is dened by a mapping function Esp eciallyLFG sepa

rates information ab out grammatical functions from that ab out the phrase structure

and grammatical functions are not necessarily dened via phrase structure p ositions

Structural relationships are represented as constituentstructure while grammatical

functions are represented at functionalstructure

The surface word order is enco ded by constituentstructure Cstructures repre

sent the surface precedence and dominance relations in a language The X theory in

LFG accommo dates crosslinguistic typ ological variation It not only generates reg

ular pro jecting categories eg V V VP including functional categories suchasIP

and CP King Kro eger Bresnan and Mchomb o Bresnan a but

also pro duces a category such as S which has no xed categorial head exo centric and

pro jects no higher category This category S can dominate either a congurational or

CHAPTER INTRODUCTION

noncongurational at structure In the latter case grammatical functions are

not determined by the phrase structural conguration but by the case and agreement

morphology Bresnan a Notablyelements in cstructure are all basegenerated

there is no movement The eects of movement are handled by the fact that dierent

cstructure p ositions may corresp ond to the same fstructure by general principles

unication

Unlike cstructure which enco des dominance and precedence relations functional

structures enco de grammatical function and other information relevant to the syntax

Also unlike cstructures fstructures are comp osed of attributevalue matrices At

tributes can b e suchentities as grammatical functions SUBJ OBJ COMP tense

TENSE and nominal features CASE NUM GND Each attribute must b e as

signed an appropriate value Values are supplied by the lexical entries of the items and

by functional annotations on the cstructure Fstructures can app ear within other

fstructures as the value of one of the outer fstructures attributes For example

the fstructure of a sentence containing a complement clause has an attribute COMP

whose value is an fstructure corresp onding to the complement clause Consider the

simple example of cstructure and fstructure representationsin

a S

SUBJ

NP VP

OBJ

Mary V NP

met John

CHAPTER INTRODUCTION

b pred meetxy

subj

Mary

obj

John

tense Past

The annotations under the phrase structure no des state how this information maps

onto information in the fstructure The indicates the information corresp onding

to that no de while the indicates the fstructure corresp onding to the mother no de

The annotation SUBJ ab ove the NP states that the NP therein is the sub ject

of the fstructure corresp onding to its mother ie of the fstructure corresp onding

to S numb ered as Likewise with OBJ The annotation ab ove the VP

and the V states that the VP and also the V is the head of its fstructure which

is identical to that of S Fstructures are constrained by general principles

suchas completeness and coherencewhich roughly require that every function

has its predicate and a predicate has all its functions

Consider nowhowascrambled structure is represented in the cstructure and f

structure I assume that scrambling has an adjunction structure since it is not limited

to a single p osition and multiple scrambling is also p ossible Lee Compare the

cstructure for the ob ject in the canonical p osition in and that for the ob ject

which is adjoined to VP in

OBJ

NP V

CHAPTER INTRODUCTION

OBJ

NP VP

As mentioned ab ove every element is basegenerated in this framework Therefore

the adjoined ob ject is also basegenerated there as shown in there are no move

ment and no traces The functional annotation ab ove the NP indicates that it is the

fstructure OBJ just as that in is In this case some other principles in the lan

guage such as case principles not the phrase structural conguration determine the

functional annotation In other words although and dier in the cstructure

they have the same fstructure as in

pred

obj

If an ob ject can b e basegenerated in an adjoined p osition howwould it b e dis

tinguished from an adverb for example whichmay also b e basegenerated in such

a p osition An adverb will b e annotated dierently from an ob ject it carries the

ADJUNCT function as shown in

ADJUNCT

Adv VP

OBJ

NP V

As dened in Bresnan a ADJUNCT is distinguished from SUBJ or OBJ in

that only the latter are argument functions Thus I dene scrambling as follows

CHAPTER INTRODUCTION

Scrambling is an adjunction to a maximal pro jection of a phrase which carries

an argument grammatical function

In addition to cstructure and fstructure there are other mo dules of grammar

such as argumentstructure Alsina and semanticstructure Halvorsen

Halvorsen and Kaplan Dalrymple et al Argumentstructure represents

information ab out predicateargument relations Although research has b een fo cused

on the mapping b etween the astructure and fstructure eg Lexical Mapping Theory

LMT see Alsina and Mchomb o Bresnan and Kanerva and Bresnan

and Moshi there is no reason in principle that this structure should not

interact directly with cstructure Actually I argue in chapter that word order

which is represented at the cstructure is inuenced by the information ab out the

thematic roles as well as by the information ab out the grammatical functions In other

words there is an activeinteraction b etween the astructure and the cstructure in

addition to one b etween the fstructure and the cstructure The prop erties of the

semanticstructure are b eginning to b e explored but it can b e assumed that basic

semantic information is represented there

In addition to the alreadyexisting structures intro duced ab ove I prop ose an

other structure in this dissertation iinformationstructure This structure contains

discoursecontextual information of eachelement represented by the crossclassifying

features New and Prom see chapter for motivation of these features Just as

any other structures of grammar the istructure also interacts with other structures

as an indep endent mo dule and the information owbetween the structures will b e

dened by a relevant mapping mechanism

As discussed in section word order variation cannot b e easily explained bya

single domain or comp onent of grammar but should rather b e captured byinterac

tions of several dierent mo dules of grammar LFG ts well for this purp ose b ecause

it consists of parallel copresent structures Moreover as Bresnan shows in her

CHAPTER INTRODUCTION

OT analysis of the head movement phenomena in English the mapping principles b e

tween anytwo structures of grammar eg cstructure and fstructure and astructure

and fstructure can b e easily converted to the OT constraints Since no structure

is derived from another structure it is natural that all these mapping constraints

between mo dules comp ete and interact simultaneously This simultaneous constraint

comp etition b etween dierent mo dules of grammar is what I argue happ ens in the

word order variation phenomena in German and Korean Therefore I prop ose the

following as the interstructure interactions involved in the scrambling phenomena

argumentstructure functionalstructure

constituentstructure

semanticstructure informationstructure

Organization

This dissertation is organized as follows

Chapter examines basic phrase structures in German and Korean Although

German and Korean are wellknown scrambling or free wordorder languages a

certain sentential structure or string order of constituents is considered to b e un

marked or canonical The purp ose of this chapter is to investigate what the un

marked canonical phrase structural descriptions in German and Korean are and how

they can b e achieved in terms of a set of phrase structural constraints in the spirit of

Optimality Theory

In chapter I investigate the free ordering or scrambling phenomena I show

CHAPTER INTRODUCTION

that each alternative structure denotes a slightly dierentinterpretation and clas

sify these meaningrelated eects into two distinct categories semantic sp ecicity

eect and discourse antifo cus eect and p oint out that these two eects are not

unrelated which is demonstrated in the contrastive fo cus eect Taking the interpre

tational dierences as a clue to the ordering variation in these languages I develop

two information structuring constraints within the featurebased mo del of informa

tion structure Finally I argue that the sp ecicity eect can also b e subsumed under

the general informationbased approach

In chapter utilizing the core ideas in Optimality Theory Prince and Smolensky

to app ear Grimshaw Grimshaw in press I showhowvarying scrambled struc

tures are derived bytheinteractions among the constraints develop ed in the previous

chapters Here each alternativesentential structure of a sentence is viewed as the

optimal output which enco des the syntactic eg argumentstructure grammatical

functional structure and also discoursepragmatic eg information structure infor

mation provided in the input in the b est p ossible wayby means of phrase structure

andor proso dic structure It is shown that while searching for the optimal output

in each context the semantic and discourse eects ie the sp ecicityantifo cus and

contrastivefo cus eects often asso ciated with scrambling naturally follow from the

constraint comp etition from dierent mo dules of grammar

Chapter examines scrambling phenomena in Korean also from the p ersp ective

of information structuring It is argued that scrambling in Korean is also motivated

and constrained bytheinteraction of the information structuring constraints and the

phrase structural constraints Sp ecial attention is paid to the dierence in information

enco ding b etween the scrambling of the regularly casemarked phrases and that of

the phrases marked with the socalled topic marker nunItwillbeshown that the

dierence results from the interaction b etween constituent order and morphological marking

CHAPTER INTRODUCTION

Finallychapter summarizes the ma jor arguments of this dissertation and briey

discusses several remaining issues for future research

Chapter

Phrase Structure and

Congurationality

This chapter examines basic phrase structures in German and Korean As is well

known German and Korean are scrambling or free wordorder languages which

allow more than one phrase structural description p er sentence and thus havevar

ious ordering p ossibilities of the constituents Even in these scrambling languages

however a certain sentential structure or string order of constituents is considered to

b e unmarked default neutral or canonical The purp ose of this chapter is to

investigate what the unmarked canonical phrase structures are in German and Ko

rean and how they can b e achieved in terms of a set of phrase structural constraints

in the spirit of OptimalityTheory

The free word order prop erty of these languages has often given rise to the con

gurationality controversy as it is one of the ma jor noncongurational prop erties

describ ed by Hale However I argue in this chapter that b oth German

and Korean have congurational structures in the sense that they have a VP pro jec

tion which structurally distinguishes sub ject from nonsub ject arguments only the

latter are generated within VP Accordingly the free word order is characterized

CHAPTER PHRASE STRUCTURE AND CONFIGURATIONALITY

not by alternative at phrase structures but by alternative adjoining structures

where scrambled elements are adjoined to higher maximal pro jections

I prop ose that German and Korean have the same phrase structural congurations

for their basic clause structure ie S see the discussion in section Yet they

dier in that German has a CP pro jection whichevokes the V eect in matrix

clauses while Korean do es not haveany functional pro jections ab ove the simple clause

structure S Thus Korean is uniformly S in b oth matrix and emb edded clauses

The organization of this chapter is as follows In section I briey discuss Ger

man phrase structures fo cusing on topicalization phenomena including VPfronting

Based on the German facts in section which will also apply to Korean I prop ose

a set of phrase structural constraints in section which will b e resp onsible for the

default or canonical phrase structural descriptions Then in section I provide

detailed discussion of the phrase structural conguration in Korean particularly of

the lack of functional categories in this language

Phrase Structure in German

I assume with den Besten Web elhuth and Vikner

among others that German has a congurational phrase structure despite the fact

that it entertains fairly free word order p ossibilities I argue that German has a

VP pro jection which includes all the internal arguments but excludes the external

argument sub ject This is in contrast to the claims by Haider and others that

German should b e describ ed as having a noncongurational base comp onent The

main argument for a VP pro jection comes from the fact that German demonstrates

atypical b ehavior of verb phrase languages a verb and its ob jects are often involved

in syntactic op erations but not a verb and its sub ject I will primarily discuss

topicalization or fronting as a constituency test for German

CHAPTER PHRASE STRUCTURE AND CONFIGURATIONALITY

Verb Second and Clause Structure

Although most elements in German can freely switch p ositions with one another

esp ecially in the Middle Field which is roughly the area b etween the nite verb and

if any the nonnite verbs at the end of a matrix clause one p osition is strictly

xed for a particular element of the sentence that of the nite verb Namely the nite

verb must b e placed in the second p osition in matrix declarative clauses ie right

after the rst constituent of the sentence This phenomenon is usually called Verb

Second The rst constituent immediately b efore the nite verb can b e practically

anything it may b e a sub ject an ob ject an adverbial or even a nonnite verb or a

verb phrase as illustrated in

Der Kurier sollte nachher dem Spion den Brief zustecken

the courierNom should later the spyDat the noteAcc slip

The courier was supp osed to slip the spy the note later

a Der Kurier sollte nachher dem Spion den Brief zustecken

b Nachher sollte der Kurier dem Spion den Brief zustecken

c Dem Spion sollte der Kurier nachher den Brief zustecken

d Den Brief sollte der Kurier nachher dem Spion zustecken

e Zustecken sollte der Kurier nachher dem Spion den Brief

Although the sub jectinitial clause in a is conceived of as b eing in the default or

unmarked order any of the other elements can b e topicalized ie can app ear in

the initial p osition What is remarkable here is that no matter what kind of phrase

or word is in the initial p osition in the second p osition is the nite verb eg sol lte

should in The invariabilityoftheverb second p osition b ecomes more notable

Interrogative and imp erative sentences mayhave the nite verb in the rst p osition

CHAPTER PHRASE STRUCTURE AND CONFIGURATIONALITY

when it is compared with the exibility of order among the elements after the nite

verb except for the nonnite verb in the nal p osition Examples with the adverb

nachher later in the initial p osition are illustrated in

a Nachher sollte der Kurier dem Spion den Brief zustecken

b Nachher sollte der Kurier den Brief dem Spion zustecken

c Nachher sollte den Brief der Kurier dem Spion zustecken

d Nachher sollte dem Spion der Kurier den Brief zustecken

e Nachher sollte dem Spion den Brief der Kurier zustecken

f Nachher sollte den Brief dem Spion der Kurier zustecken

All p ossible p ermutations arise of the three arguments ie der Kurier the courier

dem Spion the spy and den Brief the note in Therefore the word order

in matrix clauses can b e generalized as follows the rst p osition is lled byany

constituent except for the nite verb then the second p osition is lled by the nite

verb and the remaining part of the sentence contains the nonverbal elements in free

order

In sharp contrast with matrix clauses emb edded clauses do not showanyV

eect The nite verb in fact should not b e p ositioned in the second p osition as

shown in a Instead it must b e in the sentencenal p osition as shown in b

a Ich glaub e da der Kurier sollte nachher dem Spion den Brief zustecken

I b elieve that the courier should later the spy the note slip

I b elieve that the courier was supp osed to slip the spy the note

b Ich glaub e da der Kurier nachher dem Spion den Brief zustecken sollte

I b elieve that the courier later the spy the note slip should

I b elieve that the courier was supp osed to slip the spy the note

CHAPTER PHRASE STRUCTURE AND CONFIGURATIONALITY

This matrixemb edded asymmetry as to the nite verbs p osition makes it almost a

consensus among researchers on German phrase structure that the nite verb and the

emb edding complementizer eg da that are in complementary distribution b ecause

they comp ete for a single p osition ie C the head of the ComplementizerPhrase

Holmb erg Platzack Vikner Vikner That is since the C

p osition is lled with a complementizer in emb edded clauses the nite verb cannot

b e lo cated there thus there is no V On the other hand in matrix clauses which

are not headed by a complementizer the nite verb can b e p ositioned in C In other

words the matrix clause in German is viewed as a CP pro jection just as the embedded

clause is The fact that only one constituent is p ossible b efore the nite verb in matrix

clauses can also b e explained by the CP analysis ie by the fact that there is only

one Sp ec p osition available b efore C ie Sp ecCP

The complementary distribution b etween a complementizer and a nite verb is

further supp orted by the following data First consider the example in

Ich glaub e der Kurier sollte nachher dem Spion den Brief zustecken

I b elieve the courier should later the spy the note slip

I b elieve the courier was supp osed to slip the spy the note

If the complementizer da that is missing V arises even in an emb edded clause as

demonstrated in This to o can b e predicated if we assume that the nite verb is

generated in C whenever p ossible That is if C is not lled by a complementizer the

p osition is available for the nite verb Since C is now o ccupied by a nite verb the

Sp ec p osition is also available and is lled by the sub ject in this case

WhenCislledbyacomplementizer the Sp ec p osition cannot b e lled byanything in Ger

man This is unlike the situation in Swedish for example whose complementizer som that allows

a whphrase to b e generated in the Sp ec This seems to b e due to an intrinsic prop erty of the

complementizer in German which prevents its Sp ec from b eing lled

CHAPTER PHRASE STRUCTURE AND CONFIGURATIONALITY

In addition the CP analysis is also supp orted by the fact that the order among

the elements after the complementizerinCinemb edded clauses is just as free as

it is in matrix clauses after the nite verb also in C Compare the examples in

with those in

a da der Kurier dem Spion den Brief zustecken sollte

b da der Kurier den Brief dem Spion zustecken sollte

c da den Brief der Kurier dem Spion zustecken sollte

d da dem Spion der Kurier den Brief zustecken sollte

e da dem Spion den Brief der Kurier zustecken sollte

f da den Brief dem Spion der Kurier zustecken sollte

Hence by p ositing CP for b oth matrix and emb edded clauses the Verb Second

eect in matrix clauses and also its absence in emb edded clauses is neatly captured

This gives us the following phrase structural description for matrix and embedded

clauses in German leaving the structural sister of C undetermined for now a is

the structure for a matrix clause and b for an emb edded clause

a CP

YP C

Nachher C X

sollte

der Kurierzustecken

CHAPTER PHRASE STRUCTURE AND CONFIGURATIONALITY

b CP

C X

der Kurierzustecken sollte

As describ ed ab ove the head of CP is lled with the nite verb sol lte should in a

while it is lled with the complementizer da that in b Also the topicalized

elementtakes the Sp ecCP p osition in the matrix clause Finally the elements in

X the heretofore undetermined category are fairly free in order b oth in the matrix

and emb edded clauses as wehave seen ab ove any order is p ermissible provided that

the verb is in the nal p osition Nevertheless I will argue that this category is also

hierarchically structured I will examine the internal structure of this category in the

next section

S and VP

Evidence adduced in the literature in favor of a hierarchical phrase structure in Ger

man and also in Korean unfortunately is not always free of theoryinternal reason

ings For example binding theories based on the ccommand relationship frequently

assume that the success of a binding relation automatically entails that the binder is

in a structurally higher p osition than the bindee and hence these theories generate

higher functional pro jections as demanded by the relevant binding data Case and

agreement theories recently intro duced in the Minimalist Program Chomsky

b eing based on the notion of syntactic checking of morphological features generate

I do not generate Sp ecCP in emb edded clauses for the reason mentioned in the previous fo otnote

CHAPTER PHRASE STRUCTURE AND CONFIGURATIONALITY

anumb er of functional categories suchasAgrSPAgrOPTP etc even for morpho

logically covert case or agreement features However such evidence is often irrelevant

in theories likeLFG and HPSG in which binding facts or caseagreement marking

are not necessarily explained by phrase structural congurations Therefore in this

section I will rely mainly on classic constituency tests such as fronting to examine

the internal structure of the basic clause

In addition it should b e noted that in the current framework alternative p osi

tioning p ossibilities of a phrase do not require generation of additional pro jections

b ecause scrambling is taken to b e the adjunction of a phrase to a maximal pro jection

rather than to its placement in Sp ec of a functional pro jection eg in Sp ecAgrOP

Maha jan Jonas and Bobaljik Miyagawa to app ear

NonEndo centric Category S

Since Pollo ck and Chomsky it is often assumed that basic clause struc

ture is presumably universally highly articulated with functional pro jections such

as AgrSPTP and AgrOP Due to the lackofany theoryneutral evidence phrase

structural or morphological I do not assume any functional pro jections other than

CP for German nor for Korean as will b e argued in section I argue instead

that the basic clause structure emb edded under C in German is a simple S ie an

exo centric internalsub ject category Kro eger King Bresnan a

Some Germanic languages such as Icelandic and Yiddish actually showovert syn

tactic evidence for another verbal pro jection other than VP and CPAswe will see

below in these languages a nite verb can b e placed in an intermediate p osition b e

tween the head of CP and the head of VPwhich is often regarded as evidence for IP

or AgrP Rognvaldsson and Thrainsson Diesing Reinholtz Vikner

Consider the examples in and from Icelandic and Yiddish resp ectively Vikner

CHAPTER PHRASE STRUCTURE AND CONFIGURATIONALITY

a af hverju Helgi oft hefdi lesidpessa bok

why Helgi often had read this book

why Helgi had often read this b o ok

b af hverju Helgi hefdi oft lesidpessa bok

why Helgi had often read this book

why Helgi had often read this b o ok

a ven Avrom derib er hot geleyent dos dozike bukh

when Avrom therefore has read this book

when Avrom has therefore read this b o ok

b ven Avrom hot derib er geleyent dos dozike bukh

when Avrom has therefore read this book

when Avrom has therefore read this b o ok

The nite verb hefdi had in the Icelandic example b is in the intermediate p osition

following the sub ject Helgi and preceding the adverb oft often which marks the VP

boundary It is certainly not in C b ecause C or Sp ecCP is o ccupied by the wh

phrase af hverju why This shows that the nite verb is in a p osition ab oveVPbut

belowCP eg in I The Yiddish examples in demonstrate the same p oint

However German do es not showanysuchovert syntactic evidence in favor of the

existence of IP As noted in the previous section the nite verb in German is lo cated

in the sentencenal p osition in emb edded clauses unless it on rare o ccasions lls

the C p osition in the absence of an overt complementizer

a Ich glaub e da der Kurier sollte nachher dem Spion den Brief zustecken

I b elieve that the courier should later the spy the note slip

I b elieve that the courier was supp osed to slip the spy the note

CHAPTER PHRASE STRUCTURE AND CONFIGURATIONALITY

b Ich glaub e da der Kurier nachher dem Spion den Brief zustecken sollte

I b elieve that the courier later the spy the note slip should

I b elieve that the courier was supp osed to slip the spy the note

One could argue that IP in German is headnal just likeVP hence the nite verb

sol lte in b is in fact in I not in V but we cannot see it overtly Vikner

Although this is a plausible hyp othesis p ositing another pro jection is against the

general economy principle of phrase structure generation see Haider and

references therein for similar ideas and is unnecessary to explain the data in the

present framework Also not assuming a theory of syntactic caseagreement feature

checking the mere presence of rich verbal inections eg tense and agreement

do es not necessarily lead to the generation of TP or AgrPs Therefore I assume

that German has S generated under CP and do es not haveintermediate functional

pro jections

Following Kro eger King and Bresnan a I dene S

as a universally available exo centric category having no xed categorial head and

max

pro jecting no higher category Namely there is no category X such that SX

and there is no S Smay dominate either congurational or noncongurational

at structures The congurational version of S which consists of a sub ject and

an XP predicate eg a VP can b e conceived of as b eing parallel in spirit to the

internalsub ject categories such as SmallClause in Chung and McCloskey

max

and McCloskey V in Ko opman and Sp ortiche or VP in Yoon

in that it contains a separate maximal predicate pro jection eg VP These

prop osals are distinguished from other internalsub ject hyp otheses which assume that

the sub ject is generated within VP in Sp ecVP as a sister to V Kitagawa

Kuro da Sp eas Guilfoyle et al In the latter it is hard to explain why

the nonmaximal intermediate pro jection ie V behaves like a maximal pro jection

CHAPTER PHRASE STRUCTURE AND CONFIGURATIONALITY

in various syntactic phenomena including topicalization

In what follows I will argue that German has a congurational S That is S is

further structured into smaller constituents which captures the sub jectob ject asym

metries in this language Crucially whether there is a separate maximal pro jection

eg VP which includes a verb and its ob jects but excludes the sub ject will b e

tested

Topicalization and VP

One common test for constituency is fronting This test is based on the claim that

fronting or movement only aects constituents The main evidence in favor of the

sub jectob ject asymmetry in German also comes from a fronting phenomenon ie

topicalization As wehave seen in the previous section any constituent can b e top

icalized or prep osed to the initial p osition Since the initial p osition is a Sp ec ie

Sp ecCP it allows only one constituent For example in a the sub ject der Junge

the b oy is topicalized and in b the ob ject dem Mann the man is topicalized

a Der Junge hat dem Mann geholfen

the b oyNom hat the manDat help ed

The b oy help ed the man

b Dem Mann hat der Junge geholfen

the manDat hat the b oyNom help ed

The b oy help ed the man

If any constituent can topicalize in German and if there exists a constituentwhich

is smaller than S but bigger than each individual argument phrase a VP for in

stance then this constituent should b e able to topicalize to o This is indeed the case Consider the examples in

CHAPTER PHRASE STRUCTURE AND CONFIGURATIONALITY

a Dem Mann geholfen hat der Junge

the manDat help has the b oyNom

The b oy help ed the man

b Der Junge geholfen hat dem Mann

the b oyNom help has the manDat

The b oy help ed the man

Not surprisingly the nonnite verb geholfen help can b e topicalized together with

the ob ject dem Mann as shown in a In contrast however the nonnite verb may

not take its sub ject along as illustrated in b Thiersch Web elhuth

These facts suggest that German has a structural asymmetry b etween the sub ject

and the ob jects such that the verb and its ob jects or internal arguments comp ose

a constituent ie a VP but the verb and its sub ject do not

As a matter of fact not only the whole VP but also part of it can b e topical

ized in German which is often referred to as remnant VP topicalization This is

In fact there are a couple of constructions which allow the topicalization of the verb and its

sub ject Web elhuth Uszkoreit These include passive and unaccusative constructions

in which the sub ject is more a typ e of internal argument ie the themepatienttyp e of argument

Some examples are shown in i Web elhuth

i a

Zwei Manner erschossen wurden wahrend des Vochenendes

two menNom shot were during the weekend

Twomenwere shot during the weekend

Ein Unfall passiert ist hier schon lange nicht mehr

an accidentNom happ ened has here for long not more

No accident has happ ened here for a long time

InterestinglyWeb elhuth argues that German do es not have NPmovement which means

that unlike in English the themepatient argument remains in situ ie within the VP in passive

and unaccusative constructions That is the grammatical function change in these constructions

is not derived bymovement or by a dierent phrase structure but by a dierent case marking

pattern

This claim is supp orted by the fact that the direct ob ject p osition in an activesentence is also

the default p osition for the sub ject of the passivecounterpart as shown in ii

CHAPTER PHRASE STRUCTURE AND CONFIGURATIONALITY

demonstrated in Web elhuth and den Besten Uszkoreit Pollard

a Den Brief zustecken sollte der Kurier nachher dem Spion

the noteAcc slip should the courierNom later the spyDat

b Dem Spion zustecken sollte der Kurier nachher den Brief

the spyDat slip should the courierNom later the noteAcc

c Nachher dem Spion zustecken sollte der Kurier den Brief

later the spyDat slip should the courierNom the noteAcc

This partial VP topicalization might cast doubt on the validity of topicalization as

a constituency test Yet even in partial VP topicalization the sub ject may not b e

included in the topicalized domain as shown in Uszkoreit

a Der Kurier zustecken sollte den Brief nachher dem Spion

the courierNom slip should the noteAcc later the spyDat

b Der Kurier den Brief zustecken sollte nachher dem Spion

the courierNom the noteAcc slip should later the spyDat

Following Web elhuth and den Besten I assume that the remnant VP topical

ization is in fact topicalization of the maximal pro jection VP not of V or V The

partiality results from the fact that some memb ers of VP can scramble out of it and

weil dem Mann ein Buch geschenkt wurde

b ecause the manDat a b o okNom given was

b ecause a b o ok was given to the man

Therefore the apparentcounterexamples in i can b e accounted for by still treating the topicalized

phrase as a VPThismay suggest that the VP in German is actually a b oundary distinguishing

an agent argument from nonagent arguments rather than a sub ject from nonsub ject arguments

See the distinction of external and internal arguments in Williams See also Bresnan

and Zaenen for an illustration of howsuch a distinction can b e derived from the thematic

rolefunction linking For the purp oses of the current discussion though I will continue to assume that the VP distinguishes the sub ject from nonsub jects

CHAPTER PHRASE STRUCTURE AND CONFIGURATIONALITY

that only the remaining part is topicalized That is the topicalized partial VP is the

result of the removal of the scrambled phrases from the whole VPFor example the

sentence c can b e illustrated as in where the direct ob ject den Brief the

note is scrambled out and adjoined to VP

VP C

Nachher dem Spion zustecken C S

sollte NP VP

der Kurier NP VP

den Brief

This claim that remnant VP topicalization is a result of scrambling is further

supp orted by the ungrammaticality of the examples in Uszkoreit

a Den Brief nachher sollte der Kurier dem Spion zustecken

the noteAcc later should the courierNom a spyDat slip

b Den Brief dem Spion sollte der Kurier nachher zustecken

the noteAcc the spyDat should the courierNom later slip

c Nachher dem Spion sollte der Kurier den Brief zustecken

later the spyDat should the courierNom the noteAcc slip

The examples in show that without the accompanying verb no combination of

the remaining elements can b e fronted even if they all b elong to the VP This is

exp ected if we assume that partial VP topicalization is the topicalization of the VP

CHAPTER PHRASE STRUCTURE AND CONFIGURATIONALITY

after scrambling As such the examples in cannot b e regarded as VP topical

ization in which the verb is scrambled out b ecause verbal elements in German as in

many other languages including Korean cannot scramble Therefore it follows that

the fronted items in are not a VP constituent but a nonconstituent collection

of some memb ers of the VP and thus they cannot topicalize Recall that only one

constituent is allowed in the initial p osition

Note that this assumption do es not prevent the verb from topicalizing alone as

illustrated in

Zustecken sollte der Kurier nachher dem Spion den Brief

slip should the courierNom later the spyDat the noteAcc

The courier was supp osed to slip the spy the note

This example can also b e explained as VP topicalization where all memb ers of the

VP except the verb are scrambled out

Thus I argue that German indeed has a VP pro jection which do es not include

the sub ject We can then explain the sub jectnonsub ject asymmetry by assuming

that the sub ject is generated as the rst daughter of S ie NPS while the other

arguments are generated within VP Web elhuth Hence the undetermined

category X in has turned out to b e S whichinturncontains a hierarchical VP This structure is illustrated in

CHAPTER PHRASE STRUCTURE AND CONFIGURATIONALITY

C S

da NP VP

der Kurier VP V

Adv VP sollte

nachher V

NP V

dem Spion NP V

den Brief zustecken

Since the C p osition is lled by the complementizer da that as it is an embedded

clause the nite verb sol lte cannot b e p ositioned there Hence it is in the original

V p osition Since topicalization is not p ossible in emb edded clauses all elements are

within S ie the sub ject der Kurier the courier is the rst daughter of S and the

rest are within VP Note that I assume that the indirect or dative ob ject dem Spion

the spy is structurally higher than the direct or accusativeobjectden Brief the

note This is due to the observation by Lenerz Abraham and others

that the indirect ob ject precedes the direct ob ject in the default or unmarked order

as I will discuss in section

I assume a hierarchical organization of VP where the indirect ob ject ccommands the direct

ob ject However it could equally well b e at Although binding facts are often used as evidence

for hierarchical structure it is not clear whether they really constitute genuine theoryneutral

evidence As argued by Choi and Bresnan b some binding relations ie op erator

binding of pronominals are not based on the hierarchical notion of ccommand but on grammatical

function ranking and linear precedence Whether hierarchical or at it would not makemuch

CHAPTER PHRASE STRUCTURE AND CONFIGURATIONALITY

Canonical Order and Phrase Structural Con

straints

Based on the German facts discussed in section this section motivates a set

of phrase structural constraints which will constrain the basic phrase structural

description so as to give the canonical phrase structure in as the b est output

in terms of Optimality Theory see section for the basic assumptions in OT

Since the main concern of this thesis is scrambling I will concentrate on the basic S

structure within which scrambling is represented by adjunction In order not to b e

sidetracked by the V eect I will fo cus on emb edded clause structure

Canonical Word Order

The basic clause structure prop osed in in section yields the socalled un

marked or default order of a sentence For a ditransitive sentence for instance it

provides the linear order of Sub jectIndirect Ob jectDirect Ob ject among the

nonverbal elements of the sentence Consider the following example

da Hans dem Schuler das Buch gegeb en hat

that Hans the StudentDat the b o okAcc given has

that Hans gave the student the b o ok

The order of the arguments in the clause given in is of course not the only

p ossible order which expresses the event that the sentence is describing As men

tioned earlier any nonverbal element can scramble therefore all six p ermutations

of sub ject direct ob ject and indirect ob ject are p ossible as illustrated in The

dierence in our discussion as long as the indirect ob ject precedes the direct ob ject in the default order See also Bresnan

CHAPTER PHRASE STRUCTURE AND CONFIGURATIONALITY

free order within S is derived by scrambling ie alternative adjunction of phrases

either to VP or to S

a da Hans dem Schuler das Buch gegeb en hat

b da Hans das Buch dem Schuler gegeb en hat

c da dem Schuler Hans das Buch gegeb en hat

d da das Buch Hans dem Schuler gegeb en hat

e da dem Schuler das Buch Hans gegeb en hat

f da das Buch dem Schuler Hans gegeb en hat

All the clauses in have more or less the same truthconditional meaning

Then what do es it mean that some order of a clause is unmarked or canonical while

others are notW eb elhuth denes the unmarked order as the word order

the nativespeaker accepts as most natural in a context where none of the referents

of the noun phrases involved are known to the hearer or where all these referents are

equally well known Lenerz and also Abraham dene the unmarked

order as the one which is not constrained by fo cus That is for example in the

unmarked order in a anyelement of the sentence can b e fo cused However in a

scrambled order in b through f only a particular element can b e fo cused a

more detailed discussion of the fo cus constraintisprovided in chapter

In sum the unmarked or canonical order is the order whichisnotcontextually

restricted or constrained In other words the unmarked order is context neutral

This can b e interpreted as meaning that the unmarked order or the phrase struc

ture which results in that order is determined only by sentenceinternal information

ie the morphological or syntactic prop erties of the comp onent elements of the sen

tence and is not inuenced by sentenceexternal factors such as discoursecontextual information

CHAPTER PHRASE STRUCTURE AND CONFIGURATIONALITY

In Optimality Theory grammar is conceived of as a mapping from inputs to out

puts and a set of universally available wellformedness constraints determine through

EVAL which one of the outputs is optimal for the given input Viewing the canon

ical order or structure from this p ersp ective the canonically ordered structure is

the one whichischosen as the optimal output over all other p ossible phrase struc

tural descriptions based purely on sentenceinternal information That is there is a

set of constraints involved in the mapping of the sentenceinternal grammatical

information to its phrase structural realization which the canonical structure do es

not violate at all or violates minimally but which noncanonical structures violate

more seriously In other words this set of constraints will favor nonscrambled

or nonadjoined canonical structures over scrambled ones other conditions which

could otherwise aect the decision b eing equal I will call this set of constraints

canon short for canonical which will b e discussed in the following section

Phrase Structural Constraints canon

Flat or Hierarchical

The canonical phrase structures clearly dier from language to language Some lan

guages mayhave congurational phrase structures while others have noncongu

rational or at structures Some mayhave functionalcategorial pro jections while

others may not As prop osed in section German has highly articulated cong

urational phrase structures For example an emb edded ditransitive clause will have

the following canonical phrase structure S

Although I limit my discussion to the constraints relevanttoemb edded clauses there are more

phrase structural constraints than those involved in giving the canonical S which will b e discussed

shortly For example the V eect must b e handled for matrix clauses See Grimshaw in press

for other examples of phrase structural constraints eg those which account for the V eect in

English whquestions

CHAPTER PHRASE STRUCTURE AND CONFIGURATIONALITY

NP VP

Hans VP V

V hat

NP V

dem Schuler NP V

das Buch gegeb en

In Optimality Theory GEN whichprovides syntactic structures in addition to

other grammatical structures such as proso dic structures and phonological structures

generates in principle all p ossible phrase structures universally available according

to general guidelines of universal principles eg X theory Therefore given X theory

or the theory of phrase structures adopted here Kro eger King Bresnan

a it can generate the completely at structure in for example as b elow

in addition to the structure in

That at structures are p ossible means that phrase structural information is not used to map

argument roles to grammatical functions in certain languages Bresnan a Austin and Bresnan

to app ear In that case morphological information such as case marking is resp onsible for the

determination of grammatical functions An example of the constraint which deals with the mapping

between case marking and grammatical functions is presented b elow Bresnan

i Case to fstructure Alignment

Obliqueness of case aligns with obliqueness of function reverse syntactic rank

A nominative cstructure constituent is an fstructure Sub j and accusative cstructure con

stituent is an fstructure Ob j etc

CHAPTER PHRASE STRUCTURE AND CONFIGURATIONALITY

NP NP NP V V

Hans dem Schuler das Buch gegeb en hat

Based on the discussion in section the more articulated structure in wins

over the at structure in in German In Optimality Theory this means that

the phrase structural description in violates some higher constraint whichthe

phrase structural description in do es not violate The phrase structure could

violate other constraints and yet still b e optimal as long as the constraints it violates

are lower in rank than the one the structure in violates

Following Bresnan I prop ose that the choice b etween and in Ger

man results from comp etition b etween two conicting constraints One is economy

of expression and the other is endocentricity alignmentwhicharegiven

below in and resp ectively

Economy of Expression

Do not pro ject

Endo centricity Alignment

a Heads align A cstructure head is an fstructure head

b Complements of lexical categories align The cstructure complementof

a lexical category is an fstructure complement

The rst constraint economy of expression is a general constraint against building

unnecessary categorial structures It prefers simpler phrase structures unless more

complicated structures are required for some other reason

The second constraint endocentricity alignmentincontrast requires build

ing more hierarchical structures The rst requirement states that a fstructure head

CHAPTER PHRASE STRUCTURE AND CONFIGURATIONALITY

eg a verbal PRED is the head of its cstructural pro jection eg VP and the second

requirement states that a fstructure complementsuch as an OBJ is a cstructure

complement ie a structural sister of the head of a lexical category such as V There

fore this constraint will prefer structures whichhave a discrete VP which consists of

the head V and the complements such as ob jects as its structural sisters The fact

that German has evidence for VP as discussed in section requires that German

rank the endocentricity alignment constraint higher than the economy of

expression constraint In contrast socalled noncongurational languages would

rank them the other way around In addition a directionality constraint eg head

final is required to generate the right branching rather than the leftbranching

structures

Ranking

endocentricity economy

Why then would German not have the more articulated structure involving an IP

pro jection for exampleWh y do es the structure in still win over the structure in

CHAPTER PHRASE STRUCTURE AND CONFIGURATIONALITY

NP I

Hans I S

hat NP VP

VP V

NP V

dem Schuler NP V

das Buch gegeb en

Notice that although German ranks economy of expression lower than endo

centricity alignmentwhentwo structures equally satisfy the second constraint

then the grammar will prefer the one which minimally violates the rst constraint

For example unless German has another constraintwhichfavors the IP structure over

S eg one that requires that the inected verb b e in I would b e preferred over

b ecause the latter unnecessarily builds up one more pro jection ie IPwhich

violates economy of expression This is illustrated in

candidates endocentricity economy

CHAPTER PHRASE STRUCTURE AND CONFIGURATIONALITY

canon Grammatical Relations or Thematic Roles

The constraints intro duced ab ovecannotyet distinguish the unmarked order from

scrambled orders Recall that in a ditransitive clause Sub jectIndirect Ob ject

Ob ject is the unmarked or canonical order as shown in

da Hans dem Schuler das Buch gegeb en hat

that Hans the StudentDat the b o okAcc given has

that Hans gave the student the b o ok

This is the order whichiscontext indep endent and not sub ject to the fo cus constraint

Lenerz Abraham Thus it should b e determined only by sentence

internal information such as argument structure case marking and grammatical

relations among the elements of the sentence

In a nonderivational framework likeLFG which is organized into copresent par

allel structures ie argumentstructure functionalstructure and constituent

structure the information owbetween dierenttyp es of structures can b e regarded

as b eing controlled by a set of mapping or corresp ondence constraints Argument

structure is where the predicateargument information ie thematic relations of the

predicate with its arguments is represented functionalstructure is where the in

formation ab out the elements grammatical relations such as sub ject ob ject and

adjuncts etc is represented nally constituentstructure is equivalent to phrase

structure in that the information ab out category and constituency of phrases is rep

resented

The constituent ordering phenomena can b e conceived of as b eing derived from

the interplaybetween the cstructure and the other parallel structures in terms of

See Bresnan for the organization and op eration of grammar based on the OT ideas

and also for examples of the application of categorization and alignment constraints to the verb

movement phenomenon in English

CHAPTER PHRASE STRUCTURE AND CONFIGURATIONALITY

corresp ondence constraints Supp ose initially that the unmarked phrase structure in

German is determined by the interaction b etween fstructure and cstructure Namely

information ab out the grammatical relations determines the constituent order and

hierarchy among the elements of the sentence

Let me rst assume with Bresnan c and Mohanan and Mohanan

among others that there is a universal hierarchy of grammatical functions cf the

Accessibility Hierarchy in Keenan and Comrie whichisgiven in

Functional Hierarchy

SUBJ OBJ OBJ OBL ADJUNCT

Based on this functional hierarchy I prop ose the phrase structural constraints in

which will capture the canonical or unmarked order describ ed earlier in this chapter

This set of constraints controls the mapping b etween the grammatical functional

information and the constituent structure I call them canon short for canonical

order based on grammatical functions

canon

a cn SUBJ should b e structurally more prominent than eg ccommand

nonSUBJ functions

b cn NonSUBJ functions align reversely with the cstructure according

to the functional hierarchy

These constraints in conjunction with the endocentricity alignment constraint

intro duced in the previous subsection have the eect of generating the canonical

structure ie Sub ject as the rst daughter of S and Direct Ob ject as the closest sister

of V or the closest daughter of V to V and then Indirect Ob ject as the next closest

The distinction b etween OBJ and OBJ can b e equated with that b etween direct accusative

and indirect dative ob jects in a ditransitive clause for present purp oses

CHAPTER PHRASE STRUCTURE AND CONFIGURATIONALITY

daughter of V ie SUBJ IOBJ DOBJ V thus yielding the canonical order

S VP

of Sub jectIndirect Ob jectDirect Ob ject as shown in example Note that

the reverse alignment requirement cn puts the higher function Direct Ob ject OBJ

in lower in the cstructure This will also put adverbs ADJUNCT b efore all

nonSUBJ arguments The idea of the reverse alignment of nonSub ject functions

originates from Mohanan and Mohanan in their discussion of Hindi

word order although they assume a nonhierarchical structure for Hindi

However a closer examination of German word order indicates that the canonical

order may not b e solely determined by the grammatical functional information That

is it also seems to b e aected by the thematic roles of the arguments A version of

the thematic hierarchy is presented in Kiparsky Mohanan Bresnan

Bresnan and Kanerva Bresnan and Zaenen Bresnan c

Thematic Hierarchy

Agent Beneciary Exp eriencerGoal Instrument PatientTheme

Lo cative

According to the fo cus constraint test for unmarked order by Lenerz and

Abraham ie defaultordered arguments are not restricted in fo cus assign

ment in certain constructions the sub ject nominative argument which is function

ally highest do es not necessarily precede ob ject or nonsub ject arguments which are

functionally lower in default order First compare and Abraham

da der Chirurg dem Mann hilft

that the surgeonNom the manDat helps that the surgeon helps the man

CHAPTER PHRASE STRUCTURE AND CONFIGURATIONALITY

a da die Kur dem Mann hilft

that the cureNom the manDat helps

that the cure helps the man

b da dem Mann die Kur hilft

that the manDat the cureNom helps

that the cure helps the man

Although b oth and have the same verb helfen help as their predicate they

show dierent patterns in terms of argument structure takes an agent sub ject

and a patient ob ject and it is clear here that the canonical order is Sub jectOb ject

The reverse order is considered fairly marked On the other hand takes a non

agent themeinstrument sub ject and a b eneciary ob ject In this case neither

a or b is restricted in fo cus assignment which means that b oth are unmarked

default orders

Abraham rep orts that a certain group of verbs called symmetrical

verbs suchasahneln resemble begegnen meet treen meet gegenub erstehen stand

opp osite to also show a similar pattern with resp ect to the fo cus assignment Not

unexp ectedly psych verbs suchas gefal len please storen irritate schaden b e to

the detrimentof behagen please also do not necessarily prefer the sub ject to precede

other arguments den Besten Abraham Web elhuth as illustrated

a weil dem Jungen das Buch gefallen hat

b ecause the b oyDat the b o okNom please hat

b ecause the b o ok pleased the b oy

CHAPTER PHRASE STRUCTURE AND CONFIGURATIONALITY

b weil das Buch dem Jungen gefallen hat

b ecause the b o okNom the b oyDat please hat

b ecause the b o ok pleased the b oy

Notice again that the sub ject das Buch the b o ok here is the theme while the oblique

argument dem Jungen the b oy is the exp eriencer That is although das Buch is

functionally higher than dem Jungenitislower thematically

Finally also in passive constructions the theme sub ject mayormay not precede

other eg goal arguments in the default order as shown in Lenerz

a da dem Albrecht die Torte geschickt worden ist

that the AlbrechtDat the cakeNom sent b een has

that the cakewas sent to Albrecht

b da die Torte dem Albrecht geschickt worden ist

that the cakeNom the AlbrechtDat sent been has

that the cakewas sent to Albrecht

Here again the sub ject is lower than the nonsub ject argument in terms of thematic

hierarchy

The ab ove examples show that the thematic hierarchyaswell as the grammatical

functional hierarchyplays a crucial role in determining the canonical order Therefore

I prop ose another canonicalorder constraint based on the thematic hierarchy I will

call this constraint canon

canon

Align elements according to the thematic hierarchy

Note that this is a corresp ondence constraint mapping b etween argumentstructure

and constituentstructure

CHAPTER PHRASE STRUCTURE AND CONFIGURATIONALITY

As a matter of fact canon also can explain the canonical order of the ditransitive

clause rep eated here in Since the sub ject Hans is an agent the indirect ob ject

dem Schuler is a goal and the direct ob ject das Buch is a theme all the arguments

are aligned according to the thematic hierarchy in

da Hans dem Schuler das Buch gegeb en hat

that Hans the studentDat the b o okAcc given has

that Hans gave the student the b o ok

That is in regular transitive or ditransitive clauses b oth canon and canon

predict the same canonical order among the arguments

However in the constructions demonstrated in through ab ove the the

matic roles of the arguments do not match their grammatical functions in terms of

hierarchy And in these examples the canonical order seems to b e either the one

determined by canon or the one determined by canon In other words the

canonical order realized in constituentstructure is determined b oth by the interac

tion with functionalstructure and with argumentstructure

One way to capture this joint eect of two constraints is to think of them as tied

constraints which are not ranked with resp ect to each other so that they participate

equally in the decision For ease of exp osition though I will only use canon

which is based on the functional hierarchy and simply call it canon for the rest of

the discussion b ecause as far as active ditransitive sentences are concerned which

are the ma jor concern of this thesis canon and canon do not dier in their

predictions

I do not explore here the exact application of these tied constraints See Pesetsky

Broihier Anttila and Smolensky for discussion of tied constraints

CHAPTER PHRASE STRUCTURE AND CONFIGURATIONALITY

canon and Canonical Order

Now let us see how this new set of constraints canon distinguishes the canonical

order from scrambled noncanonical orders canon is rep eated in

canon

a cn SUBJ should b e structurally more prominent than eg ccommand

nonSUBJ functions

b cn NonSUBJ functions align reversely with the cstructure according

to the functional hierarchy

Note that these constraints are dened in terms of structural prominence Struc

tural prominence can b e either linear precedence or structural dominance in the hi

erarchy Recall that whether or not the structure is hierarchical is determined by

the endocentricity alignment constraint and whether it is rightbranching or

leftbranching is determined bythedirectionality constraints Since German has

hierarchical structures which are rightbranching as discussed ab ove a structurally

more prominent ccommanding element also precedes a structurally less prominent

element In eect the rst part cn requires that the sub ject precede all nonsub ject

arguments and the second part cn requires that the indirect ob ject precede the

direct ob ject when there are two nonsub ject arguments Therefore in a ditransi

tive sentence which has three arguments these sub constraints together pro duce the

following subhierarchies among the arguments Bakovic

a SUBJ IOBJ

b SUBJ DOBJ

c IOBJ DOBJ

Thus if the relative order in any pair of these arguments is reversed canon is

violated Also canon has a cumulative violation eect so two violations is more serious than one violation

CHAPTER PHRASE STRUCTURE AND CONFIGURATIONALITY

First consider the phrase structure in

NP VP

Hans VP V

V hat

NP V

dem Schuler NP V

das Buch gegeb en

This structure is exactly like the one describ ed bythecanon constraints The sub ject

Hans is the rst daughter of S the direct ob ject das Buch the b o ok is the closest

sister to V and the indirect ob ject dem Schuler the student is the next closest Thus

Sub ject ccommands all nonsub ject arguments and then the nonsub ject arguments

are reversely aligned so that the lowerranked Indirect Ob ject ccommands the higher

ranked Direct Ob ject That is this structure do es not violate canon

Then what ab out other scrambled structuresLet us no w compare the scrambled

structures given in and with the canonical one given in First in

the direct ob ject das Buch the b o ok is adjoined to VPthus disturbing the

subhierarchy and order of Indirect Ob jectDirect Ob ject and hence violating the

second part of canon ie that the nonsub ject functions should b e reversely aligned So this structure incurs one mark

CHAPTER PHRASE STRUCTURE AND CONFIGURATIONALITY

NP VP

Hans VP V

NP VP hat

das Buch V

NP V

dem Schuler gegeb en

In on the other hand the direct ob ject das Buch is adjoined to S

NP S

das Buch NP VP

Hans VP V

V hat

NP V

dem Schuler gegeb en

This structure not only violates the subhierarchy and order of Indirect Ob ject

Direct Ob ject but also disturbs that of Sub jectDirect Ob ject That is it now

violates the constraint that Sub ject should ccommand nonSub jects in addition

to the constraint of reverse alignment among nonSub ject functions Namelythis

CHAPTER PHRASE STRUCTURE AND CONFIGURATIONALITY

structure gets one mark for cn and another mark for cn Therefore the

structure in is worse than that in

To summarize the canon constraints have the eect of favoring the canonical

structure which has the order of Sub jectIndirect Ob jectDirect Ob ject over other

scrambled structures other things b eing equal Thus if any part of this canonical

order is disturb ed canon will b e violated In addition this constraint has a cumula

tive violation eect This way canon not only has the eect of favoring the canonical

structure but also of dierentiating among dierent scrambled structures it punishes

some scrambled structures more severely than others eg making worse than

Note that canon actually consists of two separate sub constraints ie cn

and cn This split will play a crucial role in explaining the fact that in German

scrambling over a sub ject is muchworse than scrambling over a nonsub ject argu

mentWeb elhuth We will see how this works later in chapter provides

examples of the canon violations in the various orders of a ditransitivesentence

candidates canon

a Sub ject IOb ject DOb ject

b Sub ject DOb ject IOb ject

c IOb ject Sub ject DOb ject

d DOb ject Sub ject IOb ject

e IOb ject DOb ject Sub ject

f DOb ject IOb ject Sub ject

Notice here that in e and f b oth Indirect Ob ject and Direct Ob ject violate cn

by ccommanding Sub ject which is considered to b e two separate violations thus

incurring two marks

This constraint can b e conceived of as having a similar eect to the minlink constraintin

Legendre et al a in the sense that it favors shorter movement if p ossible Chomsky

CHAPTER PHRASE STRUCTURE AND CONFIGURATIONALITY

Nowwehave established the phrase structural constraints canon for German

which together with endocentricity alignment yields the basic clause structure

and canonical word order of Sub ject IOb ject DOb ject in this language In the

S VP

last section of this chapter I will present arguments for the basic phrase structure in

Korean As mentioned earlier in this chapter Korean also shares the congurational

S structure and the canonical order among elements with German thus the phrase

structural constraints which apply to German will also apply to Korean The only

dierence lies in the way Korean applies the two sub constraints of canonThisis

contingent on the issue of scrambling over the sub ject and will b e discussed in chapter

in comparison with German

Phrase Structure in Korean

Korean like German has several prop erties considered diagnostic of noncongura

tional languages Hale including free word order as well as a rich case

system Word order in Korean is fairly free except that the verb is xed in the

clausenal p osition as shown in which has led some p eople to prop ose a non

congurational structure such as eg Chung

Maryka Johneykey phyencilul ssuessta

MaryNom JohnDat letterAcc writePstDcl Mary wrote John a letter

CHAPTER PHRASE STRUCTURE AND CONFIGURATIONALITY

a Maryka Johneykey phyencilul ssuessta

b Maryka phyencilul Johneykey ssuessta

c Johneykey Maryka phyencilul ssuessta

d Johneykey phyencilul Maryka ssuessta

e phyencilul Maryka Johneykey ssuessta

f phyencilul Johneykey Maryka ssuessta

NP NP NP V

Maryka Johneykey phyencilul ssuessta

MaryNom JohnDat noteAcc writePstDcl

However Korean also demonstrates several of the classic phenomena of a verb

phrase language just like German A verb and its ob jects are often involved in syn

tactic op erations but not a verb and its sub ject I examine VP topicalization and the

fo cus domain eect see Cho e and Bratt for more detailed arguments

From the results of these tests I p osit an underlying congurational S structure for

Korean to o In the last section I argue against functional pro jections in Korean see

also Kim and maintain the uniform S structure for b oth matrix and embedded clauses in Korean

CHAPTER PHRASE STRUCTURE AND CONFIGURATIONALITY

Evidence for VP in Korean

VP Topicalization

Topicalization in Korean is a case of fronting where a constituentisfronted and

marked by the topic marker nun As demonstrated in a VPwhich includes

all nonsub ject arguments but excludes the sub ject can b e fronted

a ttuykinun Maryka ppalli hanta

runNmlTop MaryNom fast doPrsDcl

As for running Mary do es it fast

b chaykul ilkkinun Maryka yelsimhi hayssta

b o okAcc readNmlTop MaryNom diligently doPstDcl

As for reading a b o ok Mary did it diligently

c emenieykey phyencilul ssukinun Maryka mayil hayyahanta

motherDat letterAcc writeNmlTop MaryNom everyday doshouldPrsDcl

As for writing her mother a letter Mary should do it everyday

The VP gap is lled bya proverb ha in or it can optionally b e lled bya

reduplicated verb form The fronted VP is interpreted as a topic of the sentence

ie something that the sentence is ab out

As exp ected however the sub ject and the verb cannot b e fronted together leaving

ob jects b ehind This is shown in

Later in chapter I argue that topicho o d in Korean is not in fact directly enco ded by the

socalled topic marker nun see Choi in press for a similar prop osal This however is not crucial

for the purp oses of this chapter

Some examples sound b etter with reduplicated verbs the reason for whichisnotvery clear to me

CHAPTER PHRASE STRUCTURE AND CONFIGURATIONALITY

a Maryka ttuykinun ppalli hanta

MaryNom runNmlTop fast doPrsDcl

As for Marys running she do es it fast

b Maryka ilkkinun yelsimhi chaykul hayssta

MaryNom readNmlTop diligently b o okAcc doPstDcl

As for Marys reading she did a b o ok diligently

c Maryka ssukinun mayil emenieykeyphyencilul hayyahanta

MaryNom writeNmlTop everyday motherDat letterAcc doshouldPrsDcl

As for Marys writing she should do her mother a letter everyday

It is interesting to note that even when the fronted item ie the sub ject and verb is

changed to a more general or generic expression such as lions eating the sentence

do es not improve signicantly This is shown in

sacaka mekkinun sayngkokilul hanta

lionNom eatNmlTop raw meatAcc doPrsDcl

As for lions eating they do raw meat

This indicates that the fronted items should constitute a syntactic constituentand

that the imp ossibility of fronting a sub ject and a verb in is not due to semantic

awkwardness b ecause a semantically plausible unit cannot topicalize either as shown

in The examples in and contrast with the examples in which

shows that a verb and its ob jects form a constituent while a verb and a sub ject do

not

Remnant topicalization in Korean do es not app ear to b e as free as in German Although direct

ob ject verb can quite easily topicalize leaving the indirect ob ject b ehind as illustrated in ia

indirect ob ject verb cannot do so as easily

CHAPTER PHRASE STRUCTURE AND CONFIGURATIONALITY

VP Fo cus

Another construction whichinvolves nunmarking is the fo cus construction When

an element is marked with nun in its base p osition it usually carries contrastive

fo cus For example when the ob ject chayk b o ok in is marked with nunit

carries contrastive fo cus and thus the sentence implies something like Mary reads

BOOKS but she do es not read NEWSPAPERS as shown in a

a Maryka chaykun ilknunta

MaryNom b o okTop readPrsDcl

Mary reads b o oks but reads nothing else

b Maryka chaykun ilknunta

MaryNom b o okTop readPrsDcl

Mary reads b o oks but do es nothing else

Interestingly the domain of fo cus can b e expanded to the VP when the ob ject is

fo cused For instance the whole VP chaykul ilknunta read b o oks in can b e

fo cused so that the sentence can mean something like Mary READS BOOKS but

she do es not WRITE PAPERS as shown in b Imp ortantlyhowever the fo cus

cannot b e pro jected to the whole sentence That is the sub ject Mary in cannot

i a

phyencilul ssukinun Maryka mayil emenieykey hayyahanta

letterAcc writeNmlTop MaryNom everyday motherDat doshouldPrsDcl

As for writing a letter Mary should do it to her mother everyday

emenieykey ssukinun Maryka mayil phyencilul hayyahanta

motherDat writeNmlTop MaryNom everyday letterAcc doshouldPrsDcl

As for writing to her mother Mary should do a letter everyday

See chapter for discussion of the double function of nun one as a topic marker and one as a

fo cus marker

CHAPTER PHRASE STRUCTURE AND CONFIGURATIONALITY

b e included in the fo cus domain

A similar pattern app ears with a verb fo cused When a verb is marked with nun

the verb can carry contrastive fo cus by itself as shown in Additionallyany

part of the VP including the entire VP can b e fo cused but never the sub ject Kang

a Maryka ttuykinun hanta

MaryNom runNmlTop doPrsDcl

Mary do es run

b Maryka chaykul ilkkinun hayssta

MaryNom b o okAcc readNmlTop doPstDcl

Mary did read a b o ok

c Maryka emenieykey phyencilul ssukinun hayssta

MaryNom motherDat letterAcc writeNmlTop doPstDcl

Mary did write her mother a letter

So for instance c in the ab ove examples can mean one of the following

a Mary WROTE a letter to her mother but she did not SEND it to her

b Mary WROTE A LETTER to her mother but she did not SEND

A PRESENT to her

c Mary WROTE A LETTER TO HER MOTHER but she did not SEND

A PRESENT TO HER FATHER

This fo cus pro jection is comparable to the fo cus asso ciation phenomenon with such adverbs as

only and even in English

i a I only ate CABBAGE

b I only ate CABBAGE

See Jackendo Selkirk Ro oth and Diesing for discussions of this phenomenon

CHAPTER PHRASE STRUCTURE AND CONFIGURATIONALITY

However it is hard to get a reading like for c

MARYWROTE A LETTER TO HER MOTHER

but JOHN did not SEND A PRESENT TO HER FATHER

As such the fo cus pro jection with nunmarking shows that the VP is some sort of

a domain which the fo cus can pro ject within but not b eyond and thus argues for a

separate syntactic pro jection

Phonological Supp ort for VP

There is also phonological evidence in favor of the existence of a VP in Korean in ad

dition to the syntactic evidence shown in the previous sections Phonological phrasing

often hinges on syntactic phrase structures Cho has presented phonological

phrase formation in Korean as phonological evidence for the existence of VPThe

evidence comes from a contrast b etween sub ject and nonsub ject arguments in form

ing phonological phrases in the application of the ObstruentVoicing rule see Cho

for details of phonological phrase formation

ObstruentVoicing in Korean is a rule whichchanges plain obstruents ie p

t k and c into their voiced counterparts ie b d g and j resp ectively

when an obstruent o ccurs b etween voiced segments suchasvowels or nasals This rule

applies not only within words but also across words within a phonological phrase

In the following examples a voiced segment is underlined and a phonological phrase

is represented with square brackets in the following examples

a Cwunik a kongul cayppalli cap a

CwuniNom ballAcc quickly catchPrs

Cwuni catches balls quickly

CHAPTER PHRASE STRUCTURE AND CONFIGURATIONALITY

b kongul Cwunik a cayppalli cap a

ballAcc CwuniNom quickly catchPrs

Balls Cwuni catches quickly

The crucial contrast o ccurs in the rst consonantoftheadverb cayppal li quickly

whichisvoiced when directly preceded by the ob ject in a and not voiced when

preceded by the sub ject in b This indicates that the ob ject kongul ballAcc

o ccurs within the same phonological phrase as the adverb cayppal li quicklyand

that the sub ject Cwunika CwuniNom do es not o ccur within the same phrase as

cayppal li quickly In other words the VP is a necessary syntactic constituent for

explaining phonological phrasing in Korean

To summarize a VP exists in Korean which includes all nonsub ject arguments

and excludes the sub ject Its presence has b een attested by topicalization and fo cus

pro jection Also there is phonological evidence which comes from phonological phrase

formation in Korean In the next subsection I will argue following Cho and Sells

that no functional pro jections are necessary in Korean

Co ordination and Clause Structure

Ihave fo cused on the sub jectob ject asymmetry in section and concluded that

Korean has a VP This alone would give us a clause structure like without

additional evidence for p ositing a more articulated structure

Maryka chaykul ilkessta

MaryNom b o okAcc readPstDcl

Mary read a b o ok

See Bratt for more detailed arguments in favor of a VP in Korean

CHAPTER PHRASE STRUCTURE AND CONFIGURATIONALITY

NP VP

Maryka NP V

chaykul ilkessta

Recentlyhowever there have b een several prop osals in the literature on Korean

as well as in that on Japanese to treat verbal inectional morphemes such as tense

and mo o d elements as discrete syntactic phrasal heads esp ecially of such functional

categories as IP and CP or TP and MP for tense and mo o d resp ectively They

are motivated by theories of phrase structure based on head movement and extended

X theory Chomsky Pollo ck Under this assumption the sentence in

would b e represented as in

IP C

VP I ta

NP VP ess

Maryka NP V

chaykul ilk

For this typ e of phrase structure for Korean see Whitman Yo on and Yoon and

Yo on among many others Some linguists Ahn Ahn and Yo on suggest that

the honoric marker si and the negative marker anh are also syntactic heads in Korean ie Agr

and Neg resp ectively Arguments against this prop osal are detailed in Han and Sells

Therefore I will not discuss Agr and Neg here and only concentrate on I and C

I use VP following Yo on However it is in essence equivalenttoSinthecurrent

framework as a category whichcontains a sub ject and a separate verbal maximal pro jection

CHAPTER PHRASE STRUCTURE AND CONFIGURATIONALITY

In this typ e of analysis only the verb stem is generated under the V no de and the

tense and the mo o d morphemes are generated separately under I and C The verb

stem ilk read is supp osed to form a unit with the tense and mo o d morphemes ess

Past and ta Declarative by Head Movement ie V to I to C Whitman Ahn

and Yo on Ahn or by Phrasal Axation Yo on Yo on In other

words the morphological pro cess of axation is taken over by a syntactic mechanism

A lexicalist view of morphology including the lexical integrity principle prohibits

a phrase structure suchasandthesyntactic takeover of morphological pro cesses

In such a view words which feed syntactic phrase structures are unanalyzable units

and thus stems axes and other b ound morphemes are combined in the lexicon and

notinsyntax In other words the principles and constraints which are involved in

morphological comp osition are dierent from those which constrain syntactic op era

tions

An alternative to the syntactic building view on morphology is a basegeneration

approach within a lexical constraintbased framework Kro eger for Tagalog and

King for Russian In these works verb movement is replaced by optional gen

eration of the head category in either its source or target p osition It is a compromise

between the two approaches illustrated ab ove in the sense that it adopts the extended

X theory but still maintains the lexical integrity principle is an example of a

phrase structural representation of this view

The checking theory Chomsky could b e viewed as analogous to the approach

intro duced here in that verbs are inserted in the syntax fully inected although they still need to

move to the appropriate functional heads to checkotherelevant features

CHAPTER PHRASE STRUCTURE AND CONFIGURATIONALITY

IP C

VP I ilkessta

NP VP

Maryka NP V

chaykul

In what follows I rst intro duce morphophonological evidence which shows that

verbal inections in Korean are lexical they combine with stems in the lexicon and

thus verbs are inserted in their fully inected form in syntax Cho and Sells

Next I review the syntactic arguments for the existence of the functional categories

IP and CP in Korean and conclude that there is no crucial evidence for sucha

distinction The main arguments come from co ordination

Verbal Inections are Lexical

Though it is uncontroversial that the tense and mo o d morphemes in Korean are

phonologically dep endent on the preceding word or morpheme their morphologi

calorsyntactic status is still in question The ma jor arguments for the syntactic

treatment are based on suchsyntactic generalizations as the phrasal distribution of

these morphemes and their scop e if one assumes that these morphemes are attached

to syntacticallyformed phrases rather than to the basegenerated verb stems their

phrasal or sentential scop e falls out from the phrase structure One easy resolution to

the problem of the phonological dep endency on the one hand and the phrasal scop e

on the other would b e to assume that these morphemes are clitics However Cho and

Sells argue that the mere fact that these elements play an imp ortant role in

CHAPTER PHRASE STRUCTURE AND CONFIGURATIONALITY

syntax and semantics should not necessarily lead to the conclusion that they should

b e handled by p ostsyntactic morphology or cliticization see Poser and Kan

erva for similar cases Moreover the Korean suxes show strong phonological

evidence that they are lexically attached

One of the prop erties which distinguish axes from clitics is the morphological

sensitivity of certain lexical rules Zwicky and Pullum Sado ck One such

rule is Co da Neutralization Cho and Sells demonstrate that the axation

of tense and mo o d morphemes in Korean shows lexical rather than p ostlexical

behavior in the sense of Lexical Phonology Kiparsky Mohanan

The phonological rule Co da Neutralization neutralizes a co da continuant to a stop a

palatal to a dental and a laryngeal to a plain stop in syllable co da p osition However

when a continuant is syllabied lexically as the onset of the following sux it escap es

the application of Co da Neutralization Lexical syllabication takes as its domain the

ro ot and the suxes either derivational or inectional but never covers two separate

lexical words

For example lexically derived nouns suchas wusum laughter show that Co da

Neutralization is blo cked The s in is lexically syllabied as the onset of the

following syllable and therefore Co da Neutralization do es not apply

wusum laughNml wusum vs wudum

Similarly Co da Neutralization fails to apply b etween a verb stem and its axes as

shown in The s in again is resyllabied as the onset of the following

syllable and thus not neutralized to d This indicates that a verb stem and its tense

morpheme also form a lexical word

wusessta laughPstDcl wusetta vs wudetta

Acknowledging the lexical prop erty of the inectional morphemes in Korean Yoon

CHAPTER PHRASE STRUCTURE AND CONFIGURATIONALITY

calls them phrasal axes Yo on denes phrasal axes as b ound mor

phemes whichcombine syntactically with phrases like clitics but which app ear as an

ax on the p eripheryhead of the constituent they sub categorize syntactically cf

Sado ck Lap ointe Therefore on the surface they will app ear as

suxes on the verbal ro ot In short a phrasal ax is a clitic which has lexical prop

erties which is an anomalous category from a morphological p oint of view It seems

to me that the phrasal ax analysis is not a real solution to the problem but only

an attempt to give a name to the problem itself it simply says that a certain b ound

morpheme is neither a regular ax nor a clitic Following Cho and Sells I will

assume that verbal inections in Korean are regular axes showing lexical prop

erties and handle the phrasal scop e problem as in co ordination without assuming

phrasal heads for those morphemes

Co ordination

Yo on presents co ordination patterns in Korean as strong arguments that tense

and mood are syntactically indep endent from the verb stem and hence that only the

uninected verbal ro ot is inserted as the head of VP The generalization regarding

verbal co ordination in Korean as well as noun phrase co ordination see Cho and

Sells is that if the verbal inections are not overtly sp ecied in nonnal

conjuncts information from those of the nal conjunct is distributed to the non

nal conjuncts Cho and Morgan Yoon and Yo on Yo on whereas

if nonnal conjuncts are also sp ecied with inections there will b e no distribution

from the nal conjunct For example in where the verbs in nonnal conjuncts

uniformly lack tense and mo o d inections and only the verbs in the nal conjuncts

are sp ecied with them the tense and mo o d of the nal conjuncts ie past and

declarative resp ectively in this case are distributed to the nonnal conjuncts and

CHAPTER PHRASE STRUCTURE AND CONFIGURATIONALITY

thus are interpreted as having the same tense and mo o d

a Maryka nolaylul tut ko pwulu essta

V V

MaryNom songAcc listenConj singPstDcl

Mary listened to and sang a song

b Maryka nolaylul pwulu kochwumul chwu essta

VP VP

MaryNom songAcc singConj danceAcc dancePstDcl

Mary sang a song and danced a dance

c Maryka nolaylul pwulu ko Sueka chwumul chwu essta

VP VP

MaryNom songAcc singConj SueNom danceAcc dancePstDcl

Mary sang a song and Sue danced a dance

Putting aside the lexical prop erty of the inectional morphologyasyntactic head

analysis could easily capture the phrasal scop e of tense and mo o d in the co ordination

data By p ositing tense and mo o d as separate syntactic heads ie I and C whose

pro jections congurationally scop e over co ordinated Vs VPs and VPs the distri

bution of tense and mo o d would naturally fall out A phrase structural representation

of the VP co ordination in c for instance is exemplied in

In fact a sentence with verb co ordination do es not sound very natural as marked with in

a for example unless the two or more verbs usually go together as if they were a semantic unit

eg mekko masi eat and drink enjoyentertain oneself takkko ssul wip e and sweep clean etc

a actually tends to b e interpreted as Mary listened to a song and sang after it and thus sounds

much b etter when ttala after is inserted b efore pwulu sing Therefore I would rather analyze

a as VP co ordination where the second ob ject is a null pronoun However this distinction is

not crucial to the present discussion

CHAPTER PHRASE STRUCTURE AND CONFIGURATIONALITY

IP C

VP I ta

VP ko VP ess

NP VP NP VP

Maryka NP V Sueka NP V

nolaylul pwulu chwumul chwu

Morphological b onding is usually assumed to b e done in the syntax in this typ e of

analysis Head movement ie verb movement from V to I to C Whitman

among others however is problematic for explaining the co ordination data illus

trated in If Head verb movementisinvolved in it must take place only

in the nal conjunct b ecause the verb in the nonnal conjunct is not inected

which systematically violates the ATB AcrosstheBoard restriction on co ordinate

structure The insertion of fullyinected verbs would not work either with a phrase

structure like as in the basegeneration view Kro eger King If

verbs are inserted in their fully inected form the verb in the nal conjunct in c

for instance could b e basegenerated in C Then the conjunct would b e asymmetric

ie of a VP and a CPwhich violates the general identity constraint on co ordina

tion Yo on alternatively prop oses phrasal axation which simply states that

the verb stem in the nal conjunct combines with the inectional morphemes in the

syntax However as mentioned in the previous section this do es not seem to b e so

much a solution to the problem as a mere restatementofit

CHAPTER PHRASE STRUCTURE AND CONFIGURATIONALITY

In addition to the tenseless co ordination given in tensed co ordination is also

p ossible is the tensed counterpart of However the tensed counterparts

in are not exact equivalents of the tenseless sentences in In the former the

tense in each conjunct even though it is marked with the same morpheme ess Past

is not necessarily considered to refer to the same tense although the actual events

could happ en all at the same time and thus each conjunct is interpreted more as an

individual and indep endentevent On the other hand in the tenseless co ordination

the tense of the nal conjunct is shared in the sense that the conjoined whole

is interpreted as denoting a single unied event although in fact eachevent could

happ en overalongtimeinterval

a Maryka nolaylul tuless ko pwuluess ta

IP IP

MaryNom songAcc listenPstConj singPstDcl

Mary listened to and sang a song

b Maryka nolaylul pwuluess kochwumul chwuess ta

IP IP

MaryNom songAcc singPstConj danceAcc dancePstDcl

Mary sang a song and danced a dance

c Maryka nolaylul pwuluess ko Sueka chwumul chwuess ta

IP IP

MaryNom songAcc singPstConj SueNom danceAcc dancePstDcl

Mary sang a song and Sue danced a dance

This eect is captured by the phrasal scop e in the phrasal head analysis Yoon

argues that when the nonnal conjuncts are sp ecied with tense and mo o d the

co ordination is necessarily clausal ie of IP or CP In a tensed co ordination the

As mentioned in the previous fo otnote the verb co ordination example in a is not very go o d

and is actually worse than its tenseless counterpart This may b e b ecause when the verbs are tensed

it is harder to consider them as a single event

When the verb of a nonnal conjunct is inected down to mo o d the conjunction marker ko is

not allowed A conjunction adverb kuliko and is instead used As a matter of fact it is problematic

CHAPTER PHRASE STRUCTURE AND CONFIGURATIONALITY

nal conjunct is also regarded as an IP even though it is not necessarily accompanied

by all its arguments as in a or b The missing arguments in the nal

conjuncts are considered to b e prodropp ed This is supp orted by the fact that the

sub ject of the rst conjunct cannot b e a nonreferential nontopical NP b ecause a

nontopical elementwould not allowa prodrop for the second sub ject McCloskey

King The phrase structural representation for c for example would

b e the following

IP C

IP ko IP ta

VP I VP I

NP VP ess NP VP ess

Maryka NP V Sueka NP V

nolaylul pwulu chwumul chwu

By treating tensed co ordination as co ordination of two IPs structurally prohibits

the tense of the nal conjunct from distributing to the nonnal conjunct while it

allows the mo o d to distribute

Unlike the syntactic head approach a lexicalist approachasinLFGwould not

structurally distinguish co ordination of tensed clauses from that of untensed clauses

Therefore it would b e a Sco ordination in either case as shown in a is a

to view twosentences connected by kuliko as a case of co ordination b ecause the rst sentence is

punctuated by a p erio d Rather I consider it as two separate sentences the second of which starts

with a conjunctive adverb

CHAPTER PHRASE STRUCTURE AND CONFIGURATIONALITY

tenseless co ordination structure and b is a tensed one

a S

S S

NP VP NP VP

Maryka NP V Sueka NP V

nolaylul pwuluko chwumul chwuessta

b S

S S

NP VP NP VP

Maryka NP V Sueka NP V

nolaylul pwuluessko chwumul chwuessta

Since verbs are inserted in their fully inected form abiding by the lexical integrity

principle axation is not executed in the syntax and therefore there is no need

for functional heads for morphological purp oses unless required by other syntactic

considerations Hence the phrasal scop e or distribution of the tense and mo o d needs

to b e handled in some other way

The standard LFG treatment of co ordination puts the information from each

conjunct as a memb er of a set in the fstructure Bresnan et al Kaplan and

Maxwell this is carried out by the functional annotation as shown in

CHAPTER PHRASE STRUCTURE AND CONFIGURATIONALITY

As is the general convention in LFG refers to the cstructure mother no de and

to the daughter no de

S S CONJ S

The set representations of the fstructures for examples a and b are shown in

a and b resp ectively

pred sing

subj Mary

obj song

tense

mood

pred dance

subj Sue

obj dance

tense Pst

mood Dcl

CHAPTER PHRASE STRUCTURE AND CONFIGURATIONALITY

pred sing

subj Mary

obj song

tense Pst

mood

pred dance

subj Sue

obj dance

tense Pst

mood Dcl

Note that although there is no distinction at the constituent structure level b etween a

tensed and an untensed co ordination clause the distinction app ears at the functional

structure level whatever value is lacking in the fstructure of the rst conjunct is

shared with that in the fstructure of the second conjunct

Interestingly this sharing or distribution of values is already builtin in the

analysis of co ordination in LFG Kaplan and Maxwell following Bresnan et al

extend the functionapplication device so that it is dened for sets of functions

and treat a set formally as the generalization of its functional elements As originally

formulated by Kaplan and Bresnan the functionapplication equation f a

v holds if and only if f denotes an fstructure which yields the value v when applied

to the attribute a For example in the second fstructure in a f tense

Pst The extended denition of functionapplication says that if s denotes a set of

functions s a v holds if and only if v is the generalization of all the elements of s

applied to a and the generalization is dened to b e the greatest lower b ound in the

subsumption ordering on the fstructure lattice see Kaplan and Maxwell

CHAPTER PHRASE STRUCTURE AND CONFIGURATIONALITY

for the denitions of functionapplication for sets and of generalization This causes

prop erties attributed externally to a co ordinate structure to b e uniformly distributed

across its elements without requiring additional grammatical sp ecications

So for example if we assume that every ro ot clause S should have tense and

moodandthus annotate the highest S in as tense and mood then the

values of these attributes for the whole co ordinate structure eg the set s in a

will b e determined by the generalization of the values of tense and mood of the

element fstructures eg f andf in a By the denition of generalization

the values for tense and mood in the co ordinate S are Pst and Dcl resp ectively

Also if the values of these attributes are sp ecied dierently in each fstructure the

generalization will not yield any common values hence no distribution Therefore the

distribution or sharing eect of the nonsp ecied values and also the nondistribution

of the sp ecied values are b oth captured in this analysis

Wehave just seen that the distribution eect in co ordinate structures can b e

dealt with in a nonstructural way Thus co ordination do es not really constitute

evidence for the phrasal head analaysis of the inectional morphemes in Korean

Moreover given the morphophonological evidence which shows that the inectional

morphemes should b e pro cessed in the lexicon it seems undesirable to p osit functional

pro jections such as IP and CP or TP and MP in Korean Therefore concluding

from the discussion in section the clause structure in Korean is uniformly S

b oth in matrix and emb edded clauses and it has a congurational VP structure

as argued earlier Although I havenotprovided detailed discussion of canonical

ordering constraints for Korean the discussion for German order given in section

also applies to Korean in that canon yields the default order of Sub jectIndirect

Yo on presents more co ordinationrelated phenomena including negative p olarity licensing

scop e of negation and scrambling as evidence for the phrasal head analysis However Choi

shows that those phenomena can b e explained without assuming such phrase structural dierences

see Choi and also Kim for detailed arguments

CHAPTER PHRASE STRUCTURE AND CONFIGURATIONALITY

Ob jectDirect Ob ject There is in fact a dierence b etween German and Korean in

their ranking of the sub constraints of canon I will discuss this in chapter and

chapter

Summary

In this chapter I have examined the basic clause structures in German and Ko

rean and concluded that b oth have a congurational S structure ie Sub ject

IOb ject DOb jectV This is prop osed as the canonical structure which

is solely determined by grammatical information The canonical structure is a re

sult of the mapping b etween the grammatical information in a and fstructure

and the phrase structural realization in cstructure whichiscontroled byasetof

OT constraints called canon

Positing a congurational canonical structure in relatively free word order lan

guages requires one to give an explanation for the range of alternativeword orders

b ecause the canonical structure only gives a subset of the surface word orders In

other words if the canon constraints were all and only the constraints involved in

phrase structural descriptions and word order scrambling would never o ccur b ecause

canon would always make the canonicallyordered sentence win over noncanonical

sentences However this is not the case of course In chapter I will develop a set

of information structuring constraints which I argue to b e the main motivation for

alternative orders or scrambling

Chapter

Sp ecicity Focus and Information

Structure

Having established the canonical structures in German and Korean in chapter I

will investigate the free ordering or scrambling phenomena in this chapter Despite

the fact that these languages are frequently called free word order languages word

order is hardly really free That is each alternative structure denotes a slightly

dierentinterpretation and thus alternative orders are not in free variation in a

strict sense Taking the interpretational dierences as a clue to the ordering variation

in these languages I identify constraints which prompt this variation in constituent

order

I will rst review the meaningrelated eects of the alternative orders in section

I classify them into two distinct categories semantic sp ecicity eect and

discourse antifo cus eect and p oint out that these two eects are not unrelated

which is demonstrated in the contrastive fo cus eect In section I prop ose a

mo del of information structure cf Vallduv and showhow it can capture

the discourserelated facts describ ed in section Based on the newlyestablished

CHAPTER SPECIFICITY FOCUS AND INFORMATION STRUCTURE

featurebased information structure in section two information structuring con

straints are develop ed which I argue are resp onsible for the interpretational eects of

scrambling describ ed in the earlier sections Finally in section I argue that the

sp ecicity eect can also b e subsumed under the general informationbased approach

Semantic and Discourse Eects of Scrambling

It has b een observed that there are certain semantic or discourserelated eects as

so ciated with scrambling These eects can largely b e summarized by the following

two generalizations One is that a denite or sp ecic NP can scramble while an in

denite or nonsp ecic NP cannot Maha jan Moltmann Diesing de

Ho op The other is that a nonfo cal or topical NP can scramble whereas a fo cal

NP cannot Lenerz Abraham Moltmann Web elhuth I will

call the former the sp ecicity eect and the latter the antifo cus eect However

these generalizations are challenged by a third eect ie the contrastive fo cus eect

Lenerz Abraham Moltmann whichcontradicts b oth of the gener

alizations b ecause in this case a nonsp ecic fo cal NP can scramble In this section

I discuss these interpretationrelated eects in scrambling and review the previous

analyses of these phenomena

Sp ecicity Eect

Some sort of deniteness eect is frequently rep orted to b e asso ciated with scram

bling opp osite to what holds for the thereconstruction in English Milsark

That is a scrambled element should b e denite or sp ecic and thus a nonsp ecic or

In this chapter I mainly discuss German data to keep the arguments consistently presented

Although not explicitly brought in the Korean facts can b e considered parallel esp ecially when the

arguments are regularly casemarked The sp ecial nunmarking case is discussed in chapter

CHAPTER SPECIFICITY FOCUS AND INFORMATION STRUCTURE

indenite NP cannot b e in a scrambled p osition Maha jan Moltmann We

belhuth Abraham

This contrast is illustrated in the distinction in German b etween b and b

Web elhuth

a weil er wohl das Buch gelesen hat

b ecause he probably the b o ok read has

b ecause he has probably read the b o ok

b weil er das Buch wohl gelesen hat

b ecause he the b o ok probably read has

b ecause he has probably read the b o ok

a weil er wohl ein Buch gelesen hat

b ecause he probably a b o ok read has

b ecause he has probably read a b o ok

b weil er ein Buch wohl gelesen hat

b ecause he a b o ok probably read has

b ecause he has probably read a b o ok

A denite ob ject NP eg das Buch the b o ok in a which in the default order

follows the adverb wohl probably and precedes the verb in emb edded clauses can

scramble over the adverb wohl as in b In contrast an indenite ob ject NP ein

Buch a b o ok cannot scramble over the same adverb wohl as illustrated in b

Similarly a denite direct ob ject can scramble over an indirect ob ject while an

indenite direct ob ject cannot as illustrated in the following Abraham see

chapter for discussion of the canonical order b etween two ob jects

CHAPTER SPECIFICITY FOCUS AND INFORMATION STRUCTURE

a Ich hab e meinem Bruder den Brief geschickt

I havemy brotherDat the letterAcc sent

I sentmy brother the letter

b Ichhabe den Brief meinem Bruder geschickt

I have the letterAcc my brotherDat sent

I sent the letter to my brother

a Ich hab e meinem Bruder einen Brief geschickt

I havemy brotherDat a letterAcc sent

I sentmy brother a letter

b Ichhabe einen Brief meinem Bruder geschickt

I have a letterAcc my brotherDat sent

I sent a letter to my brother

In the denite direct ob ject NP den Brief the letter can b e in a scrambled

p osition to the left of the indirect ob ject NP meinem Bruder my brother as shown

in b On the other hand the nonsp ecic indenite NP ein Brief a letter cannot

b e in that alternative p osition as shown in b

It has also b een noted however that it is not only denite NPs which can scram

ble but also indenite NPs under certain sp ecial interpretations these sp ecial in

terpretations include sp ecicreferential partitive and generic readings de Ho op

Diesing Moltmann Therefore the distinction lies in a semantic

dierence rather than in a morphological contrast Consider the following Dutch

examples de Ho op

a dat de p olitie een kraker gisteren opgepakt heeft

that the p olice a squatter yesterday arrested has

referential reading

CHAPTER SPECIFICITY FOCUS AND INFORMATION STRUCTURE

b dat de p olitie twee krakers gisteren opgepakt heeft

that the p olice two squatters yesterday arrested has

partitive reading

c dat de p olitie krakers altijd oppakt

that the p olice squatters always arrests

generic reading

d dat de p olitie tien krakers altijd oppakt

that the p olice ten squatters always arrests

generic collective reading

In a the indenite NP een kraker a squatter scrambles over the adverb gisteren

yesterday and in this case it can only receive a sp ecicreferential reading and

cannot receive an existential reading which is p ossible in the canonical p osition In

b also the scrambled cardinal indenite NP tweekraker two squatters has a

partitive reading which is roughly equivalenttotwo of the squatters and again the

regular existential cardinal reading is absent here Finally the scrambled indenite

NPs in c and d have generic readings in which the ob ject krakers is interpreted

as something like squatters in general This reading is not p ossible in the original

p osition

Also in connection with similar generalizations regarding the interpretations of in

denite sub jects in certain syntactic p ositions eg in a VPinternal sub ject p osition

de Ho op and Diesing generalize these sp ecicreferential partitive and

generic readings as strong and presupp ositional resp ectivelyFor example in Ger

man an indenite sub ject receives a strong or presupp ositional reading when it is

p ositioned b efore a particle suchas ja doch but it has a weak or existential reading

when it is p ositioned after the particle Diesing Also in Dutch a similar

meaning dierence arises b etween a sub ject in an erconstruction and that in the

CHAPTER SPECIFICITY FOCUS AND INFORMATION STRUCTURE

regular sub ject p osition Reuland This is demonstrated in and

resp ectively

a weil ja do ch Linguisten Kammermusik spielen

since PRT linguists chamber music play

since there are linguists playing chamber music

b weil Linguisten ja do ch Kammermusikspielen

since linguists PRT chamber music play

since in general linguists playchamber music

a Fred denkt dat er twee ko eien op het dak liggen

Fred thinks that there twocows on the ro of lie

Fred thinks that there are twocows lying on the ro of

b Fred denkt dat twee ko eien op het dak liggen

Fred thinks that twocows on the ro of lie

Fred thinks that two sp ecic cows are lying on the ro of

The sub ject Linguisten linguists in a which is lo cated after the particle ja doch

receives an existential reading while the sub ject lo cated b efore the particle in b

receives a generic reading ie a strong reading Similarly the cardinal sub ject twee

koeien twocows in a which is p ositioned after er there has an existential

reading whereas that in b which is in the clauseinitial p osition receives a sp ecic

or partitive reading

De Ho op argues that the distinction of strong versus weak interpretation

comes from an abstract structural Case distinction She argues that an ob ject is

interpreted as strong or as a generalized quantier of typ e hhe titi if and only

Although it seems fairly plausible to argue that this semantic distinction may follow from the

CHAPTER SPECIFICITY FOCUS AND INFORMATION STRUCTURE

if it b ears strong Case otherwise it receives a weak reading which is a predicate

mo dier typ e reading of typ e hhe ti he tii According to her strong Case is a

structural Case licensed at Sstructure while weak Case is a default Case licensed

at Dstructure Therefore an indenite NP which receives its Case strong Case at

Sstructure such as a scrambled NP receives a strong reading while an indenite NP

which has case weak Case at Dstructure eg in default p osition receives a weak

reading

Diesing on the other hand explains this semantic distinction ie presup

p ositional versus existential by arguing that the VPwhich is demarcated by ja doch

for instance in is a b oundary for existential closure and that an NP inside VP is

mapp ed to the nuclear scop e while an NP outside VP is mapp ed to the restrictive

clause This is summarized bytheMapping Hypothesis as follows Diesing

a Material from VP is mapp ed into the Nuclear Scop e

b Material from IP is mapp ed into a Restrictive Clause

By virtue of b eing external to VP an indenite NP is semantically mapp ed into the

restrictive clause and thereby receives a presupp ositional reading On the other hand

an indenite NP whichisinternal to VP is mapp ed to the nuclear scop e and b ound

within the VP by an existential op erator thus receiving an existential reading

Further pursuing this line of a direct syntaxsemantics mapping approach Diesing

argues that scrambling or ob ject shift Holmb erg Vikner is an

instance of semanticallydrivenmovement a result of interpretation conditions ap

plying in the syntaxsemantics mapping whichinducemovementFollowing Heim

morphological case distinction in languages suchasTurkish nominative default case versus ac

cusative Enc and Finnish partitiveversus accusative it b ecomes much more abstract in

languages suchasDutch German and Korean where no overt case distinction exists b etween weak

and strong NPs Even the distinction based on abstract Case marking do es not seem highly plausi

ble in German and Korean b ecause scrambling in these languages is not limited to a single p osition

It is not very desirable to say that there is more than one strongCase p osition and even if so

variation b etween two strongCase p ositions is hard to explain

CHAPTER SPECIFICITY FOCUS AND INFORMATION STRUCTURE

Diesing assumes that denites like indenites intro duce vari

ables but that these are familiar entities and thus incompatible with existential

binding which requires novel variables Therefore denite ob jects cannot bybound

by the existential closure and thus must move out of the VP

In short b oth de Ho op and Diesing assume that the semantic

distinction comes from a syntactic distinction whether it is Case or a certain phrase

structural scop e eg VP Namely they argue that there is a onetoone mapping

between syntax and semantics From this p ersp ective scrambling is viewed as a

syntactic mechanism to enco de a certain typ e of semantic meaning of NPs ie strong

presupp ositional or sp ecic meaning

However this raises an immediate question with regard to the optionality of scram

bling of denite NPs As wehave seen ab ove inwhich is rep eated here as a

denite NP mayormay not scramble

a Ich hab e meinem Bruder den Brief geschickt

I havemy brotherDat the letterAcc sent

I sentmy brother the letter

b Ichhabe den Brief meinem Bruder geschickt

I have the letterAcc my brotherDat sent

I sent the letter to my brother

According to Diesing a denite NP must move out of VP to avoid existential

closure b ecause it is a familiar entity In other words it must necessarily scramble

However it can stay in the canonical p osition as shown in a Moreover there

Diesing argues that the existential closure condition can b e regarded as part of a general

requirement that the relative scop e of NPs ie the Scop e Condition b e syntactically determined

eg by the ccommand relation cf Moltmann However although some examples esp ecially

those whichinvolve scop eaecting adverbs or negations can b e explained by the direct syntax

semantics mapping examples like are harder to explain with scop e interactions See Vallduv for an example of a discourse explanation of scop e phenomena

CHAPTER SPECIFICITY FOCUS AND INFORMATION STRUCTURE

is no distinct semantic dierence in strongness or presupp ositionality b etween the

insitu denite NP in a and the scrambled one in b The denite NP den Brief

the letter denotes a sp ecic referential thus strong and presupp ositional entity

regardless of its relative p osition in the sentence

It is also true however that the denite NP den Brief enco des a slightly dierent

reading dep ending on the p osition although it is not a dierence in sp ecicity the

scrambled NP in b receives more of a topiclikeinterpretation whereas the one in

a receives more of a fo cuslikeinterpretation Interestingly Diesing

also notes the topicfo cus dierence in interpretation and argues that the familiar

status of denite NPs can b e overridden by fo cus which signals novel information

is b eing or ab out to b e presented and in this case denite NPs can remain in situ

This fact indicates that the scramblability of a phrase is fairly strongly aected by

discourse factors such as topic and fo cus

Similarly this discourse inuence is also exhibited with indenite NPs to o Con

trary to our initial observation that indenite NPs esp ecially nonsp ecic weak or

existential NPs cannot scramble an indenite NP can actually b e out of its canonical

p osition when it is contrastively fo cused and moreover in that case it can maintain

its existential reading Moltmann We will discuss this issue in more

detail in section

Therefore it can b e concluded that the semantic distinction of sp ecicityalone

cannot explain all asp ects of scramblability of a phrase but rather discourse prop erties

such as topicality and fo cality should b e taken into account As a matter of fact the

topicfo cus eect of scrambling has b een frequently noted in the literature which will

b e discussed in the sections to follow

CHAPTER SPECIFICITY FOCUS AND INFORMATION STRUCTURE

Antifo cus Eect

Scrambling also exhibits what I call the antifo cus eect Lenerz Abraham

Web elhuth Moltmann which can b e roughly summarized as the

scrambled elementmust necessarily b e unfo cused Web elhuth As Lenerz

p oints out the data b elow in and demonstrate an asymmetry in

fo cality of constituents b etween the default order sentence and ones with a scrambled

order That is in the default order no fo cus constraint holds ie any elementofa

sentence argument or adjunct can b e fo cused On the other hand in a scrambled

order a fo cus constraint prevents a scrambled phrase from b eing fo cused Here

fo cus can b e interpreted as new information it is used to refer to an elementwhich

corresp onds to a whphrase in the previous question This element usually receives

proso dic prominence as well For this reason in all the following examples a fo cused

element even if not contrastive will b e represented in upp er case

A discourse context which easily oers a fo cus construction is a question and

answer pair In most cases the fo cus or the new information of a sentence is the

part which corresp onds to the question phrase Let us rst lo ok at the question and

answer pairs in and Lenerz

a Wann hast du das Buch gelesen

when haveyou the b o ok read

When did you read the b o ok

b Ich hab e GESTERN das Buch gelesen

I haveyesterday the book read

I read the b o ok yesterday

b Ich hab e das Buch GESTERN gelesen

I have the book yesterday read

I read the b o ok yesterday

CHAPTER SPECIFICITY FOCUS AND INFORMATION STRUCTURE

a Was hast du gestern gelesen

what haveyou yesterday read

What did you read yesterday

b Ich hab e gestern das BUCH gelesen

I haveyesterdaythebook read

I read the b o ok yesterday

b Ich hab e das BUCH gestern gelesen

I have the book yesterday read

I read the b o ok yesterday

In since the question is ab out wann when the b o okreading event happ ened

the time adverb phrase gestern yesterday will b e the fo cus in the following answers

b and b On the other hand in the question is ab out was what was

read in the reading eventofyesterday das Buch the b o ok will b e the fo cus in the

following answers b and b In b oth and the b sentences are in

the canonical order and the b sentences are in a scrambled order in whichthe

ob ject scrambles over the adverb

As Lenerz p oints out in the default order either the adverb gestern yes

terday in b or the ob ject das Buch the b o ok in b can b e fo cused In other

words no fo cus constraint applies in a default ordered sentence In contrast how

ever a fo cus constraint holds in scrambled sentences That is while the insitu adverb

gestern in b can b e fo cused the scrambled ob ject das Buch in b cannot b e

fo cused To put it in other words the nonfo cused ob ject phrase das Buch in can

scramble as shown in b However if it is fo cused as in the ob ject phrase

cannot scramble over the adverb as shown in b

A ditransitive construction shows a similar eect Lenerz as demon

strated in and In the indirect NP dem Kassierer the cashier is the

CHAPTER SPECIFICITY FOCUS AND INFORMATION STRUCTURE

fo cus b ecause the question solicits the information ab out wem to whom I gave

the money Therefore the nonfo cused direct ob ject NP das Geld the money can

scramble as in b

a Wem hast du das Geld gegeb en

whom haveyou the money given

Who did you give the money

b Ich hab e dem KASSIERER das Geld gegeb en

I havethe cashier the money given

I gave the cashier the money

b Ich hab e das Geld dem KASSIERER gegeb en

I have the money the cashier given

I have the money to the cashier

In on the other hand the direct ob ject NP das Geld is the fo cus of the sentence

b ecause the question is ab out was what I gave to the cashier Thusinthiscase

the direct ob ject das Geldwhichisnow fo cused cannot scramble as illustrated in

a Was hast du dem Kassierer gegeb en

what haveyou the cashier given

What did you give the cashier

b Ich hab e dem Kassierer das GELD gegeb en

I havethe cashier the money given

I gave the cashier the money

b Ich hab e das GELD dem Kassierer gegeb en

I have the money the cashier given

I gave the money to the cashier

CHAPTER SPECIFICITY FOCUS AND INFORMATION STRUCTURE

To put it in terms of the fo cus constraint in the default order as in b and b

either the indirect ob ject dem Kassierer the cashier or the direct ob ject das Geld

the money can b e fo cused whereas in scrambled sentences such as in b and

b only the nonscrambled phrase dem Kassierer can b e fo cused

We can summarize the discussion here that a scrambled phrase should b e defo

cused and maynotbeinterpreted as new information However this generalization

is undermined by the contrastive fo cus eect which will b e discussed in the next

section Putting the contrastive fo cus eect aside for the moment we are left with

two separate generalizations regarding the semantic or discourse eects of scrambling

To recapitulate the rst is that a scrambled phrase should b e sp ecic denite or

indenite and the second is that a scrambled phrase should b e defo cused or topi

cal How can we reconcile these two generalizationsIn terestingly there havebeen

attempts to unify these two apparently distinct eects whichwe will see in what

follows

First Web elhuth explains the sp ecicity eect from the fo cusoriented

p ersp ective He argues that it can b e assumed that an unsp ecic indenite phrase

is inherently fo cused Web elhuth If we understand that fo cus here is

new information as wehave assumed b efore then his remark can b e rephrased as an

unsp ecic indenite phrase is inherently new information This idea is in fact fairly

intuitive in the sense that an unsp ecic indenite phrase is normally not referential

and is hence interpreted as unfamiliar or unidentiable in discourse therefore b eing

interpreted as new cf Heim Enc A nonsp ecic NP normally has a

function of intro ducing a new entity in the discourse Once we assume that the

sp ecicity eect can b e automatically captured by means of the antifo cus constraint

ie a scrambled phrase should b e unfo cused or defo cused

CHAPTER SPECIFICITY FOCUS AND INFORMATION STRUCTURE

a weil er wohl ein Buch gelesen hat

b ecause he probably a b o ok read has

b ecause he has probably read a b o ok

b weil er ein Buchwohl gelesen hat

b ecause he a b o ok probably read has

b ecause he has probably read a b o ok

In for example which is rep eated from section the scrambled sentence

b is ungrammatical b ecause the scrambled indenite NP ein Buch a b o ok is

inherently fo cused by b eing indenite thus violating the antifo cus constrainton

scrambled elements

Diesing on the other hand prop oses to unify the eects from the

other p ersp ective she tries to derive the fo cus eect from her treesplitting mapping

hyp othesis She notes the following phenomenon in individual predicate sentences fo

cusing the sub ject leads to favoring the existential reading as in a while fo cusing

the adjective leads to the generic reading as in b

a FIREMEN are available

b Firemen are AVAILABLE

The interpretive preferences in are explained by her semantic partition in that

the fo cus part of the sentence corresp onds to the nuclear scop e of the logical repre

sentation and the unfo cused p ortion corresp onds to the restrictive clause To push

for direct semanticsyntactic structure mapping by means of treesplitting in the syn

tax she further claims that fo cusing the sub ject causes the sub ject to lower into the

Sp ecVP at LF

Whatever the mechanical details of each prop osal may b e an imp ortant thing

to note is that they share the following p oints b oth prop osals agree that the fo cus

CHAPTER SPECIFICITY FOCUS AND INFORMATION STRUCTURE

interpretation and the existential reading b elong to one group and the nonfo cal or

topical interpretation and the sp ecicstrongpresupp ositional reading b elong to an

other group furthermore they also agree that the former typ e of NP cannot scramble

whereas the latter can However this generalization is not necessarily correct As

will b e discussed in the next section there are counterexamples which contradict

these generalizations That is certain focused phrases can also scramble I call this

phenomenon the contrastive fo cus eect

ContrastiveFo cus Eect

Moltmann provides the following interesting examples which indicate

that fo cused phrases esp ecially contrastively fo cused phrases can scramble or b e

p ositioned out of their default place Moreover she argues that when contrastively

fo cused indenite NPs scramble they can retain their existential or weak readings

We will call this the contrastive fo cus eect This is illustrated in

a weil Hans ein BUCH dem Mann gegeb en hat nicht eine ZEITUNG

b ecause Hans a b o okAcc the manDat given has not a newspap er

b ecause Hans gave a b o ok to the man not a newspap er

b weil Hans BUCHER dem Mann gegeb en hat nicht ZIGARETTEN

b ecause Hans b o oksAcc the manDat given has not cigarettesAcc

b ecause Hans gave b o oks to the man not cigarettes

In a the scrambled NP ein Buch a b o ok is fo cused ie having a high pitch accent

and b eing interpreted as new information note that contrastively fo cused elements

are also represented in upp er case Unlike the regular fo cus case however the fo cused

It seems that sp eakers vary in accepting the sentences in I will discuss this issue in the

next chapter

CHAPTER SPECIFICITY FOCUS AND INFORMATION STRUCTURE

phrase ein Buch is compared with another phrase ie eine Zeitung a newspap er and

thus receives a contrastiveinterpretation Moreover unlike the unfo cused scrambled

phrases discussed in section the fo cused indenite NP ein Buch in a need

not b e interpreted as sp ecic ie a certain sp ecic b o ok the sp eaker has in mind

Likewise in b the scrambled NP Bucher b o oks is fo cused and contrasted with

Zigaretten cigarettes and here again Bucher books does not havetobespecic

partitive or generic

Moltmann further notes that adverbs suchasnur only or sogar even

whichhave b een noted as asso ciated with fo cus Jackendo Ro oth

allow indenite NPs to scramble and also to b e interpreted nonsp ecicallyThisis

illustrated in

a weil Hans NUR ein BUCH dem Mann gegeb en hat

b ecause Hans only a b o okAcc the manDat given has

b ecause Hans only gave a b o ok to the man

b weil Hans SOGAR ein BUCH dem Mann gegeb en hat

b ecause Hans even a b o okAcc the manDat given has

b ecause Hans even gave a b o ok to the man

Here again ein Buch a b o ok is interpreted as a nonsp ecic random b o ok not as

a sp ecic b o ok Why do the phrases mo died by suchadverbs as only or even

pattern together with contrastively fo cused phrasesIn terestingly Dik et al

treat this asso ciation with fo cus case also as contrastive fo cus This will b e discussed

in more detail in next section

Abraham notes a similar fo cus eect He argues that if a def

element go es b efore a def element it has to carry fo cal stress He provides the

generalizations in regarding the deniteness and fo cality of scrambled elements

Abraham Likemany others Abraham do es not distinguish contrastive

CHAPTER SPECIFICITY FOCUS AND INFORMATION STRUCTURE

fo cus from normal fo cus thus dening his F as denoting carrying normal or

contrastivesentential accentpitch and his F as denoting the topic in its usual

meaning Abraham

a IOdefF DOdefF

b IOdefF DOdefF

c DOdefF IOdefF

d DOdefF IOdefF

As also p ointed out by others the default order sentences as illustrated in a and

b do not showany fo cus constraint b oth IO and DO can b e either F or

F However scrambled sentences where the direct ob ject precedes the indirect

ob ject as in c and d for instance are sub ject to a fo cus constraint That

is an indenite phrase should b e fo cused when it is scrambled as shown in c

Otherwise the scrambling is not p ossible as shown in d In other words c

describ es exactly the cases as in where contrastively fo cused indenites are able

to scramble

This contrastive fo cus eect or constraint is in fact a serious problem for the

previous two generalizations Simply this new contrastive fo cus eect contradicts

b oth the sp ecicity and the antifo cus eects The sp ecicity constraint requires

that nonsp ecic indenite phrases not scramble but the examples in and

indicate that they can On the other hand the antifo cus constraint requires that

fo cused phrases should not scramble but and again show that they can

We can see this contradiction in some of Abrahams constraints to o Earlier

in his presentation he prop osed the constraints on scrambling demonstrated in a

Abraham puts a for c This slight marginality seems to b e due to the fact that the

scrambling of indenite NPs requires a more complex context than the normal scrambling case ie

acontrastive fo cus context What matters in the current discussion is the distinction b etween c

and d and it is very clearly represented there

CHAPTER SPECIFICITY FOCUS AND INFORMATION STRUCTURE

and b Abraham

a DOdef IOdef

b DOF IOF

a is equivalent to the sp ecicity constraintwhich is discussed in section and

b is parallel to the antifo cus constraint discussed in section However b oth

of these constraints contradict another constraint of his whichwehave just seen in

c Ironically the combination of the negative constraints a and b turns

out to b e a p ositive constraint ie c

These new facts regarding the contrastive fo cus eect of scrambling lead us to

conclude that neither the sp ecicity constraint or the antifo cus constraintisthe

correct generalization to describ e the semantic or discourse eect asso ciated with

scrambling Furthermore the existence of a third constraint ie the contrastive

fo cus eect which requires information b oth ab out the sp ecicity and fo calityofa

scrambled element shows that the ab ovetwo eects should not b e treated separately

sp ecicity and fo cality seem to aect each other Also the contrastive fo cus eect

further demands that we should more closely examine discourse notions suchasfocus

topic and contrastive fo cus b ecause these certainly seem to b e crucial factors which

determine the scramblability of phrases Wewillinvestigate this problem in the next

section

Information Structure

A sentence can b e internally analyzed in several dierentways according to which

asp ect of it is under discussion Just as a sentence can b e structured in terms of

its elements categorial features and constituency or in terms of its comp onents

CHAPTER SPECIFICITY FOCUS AND INFORMATION STRUCTURE

semantic prop erties it can also b e structured in terms of its constituents discourse

contextual functions At this discoursepragmatic dimension a sentence can b e par

titioned into topic and comment Gundel Dahl Kuno Reinhart

presupp ositionbackground and fo cus Chomsky Jackendo Dahl

Ro chemont Prince Ward Lambrecht or theme and

rheme Firbas Halliday I will call the partition of this sort information

structure following Halliday Vallduv and Lambrecht

The information structure of a sentence is roughly sp eaking a reection of

discoursecontextual information in that sentence Dep ending on how it relates to

the discourse a sentence can havevarious information structural descriptions Vall

duv along with Lambrecht has established information structure as

an indep endent comp onent of grammar As a separate level of linguistic represen

tation information structure interacts with other comp onents of grammar suchas

syntax semantics and proso dy

As wehave seen in the previous section information structural notions suchas

topic and fo cus play a crucial role in scrambling Even the semantic eect of sp ecicity

seems to b e overridden by some informational phenomena eg contrastive fo cus as

demonstrated in In this section we will explore what information structure is

and how it relates to scrambling

Information Packaging

Vallduv takes information structure or information packaging to b e

a sub domain of pragmatics which deals with problems such as illo cution discourse

structure reference resolution implicature and empathy He argues that the suc

cessful interpretation of a sentence requires not only the interpretation of the truth

conditional meaning or the semantic meaning in its narrow sense but also the inter

pretation of the informational meaning According to him the latter is enco ded by

CHAPTER SPECIFICITY FOCUS AND INFORMATION STRUCTURE

information packaging Information packaging reects the sp eakers b eliefs ab out

how the information ts the hearers knowledgestore Vallduv In other

words it not only indicates what part of the sentence constitutes information but

also instructs where and how that information should b e stored

Vallduvs information structure consists of three primitives ie fo

cus link and tail which enco de the information status of each elementofthe

sentence He rst partitions a sentence into two parts ie focus and ground

Ground is the part that anchors the sentence to the previous discourse or the hearers

mental world whereas fo cus is the informative part that makes some contribution

to the discourse or the hearers mental world This distinction is parallel to the

binary partition of givennew or topicfo cus in the literature Then he further divides

ground into link and tail This sub division of background information is crucial in

Vallduv ch to his accountofword order in Catalan according to him link

elements are leftdetached and tail elements are rightdetached in Catalan

S ffo cus groundg

ground flink tailg

Since for Vallduvthe role of information packaging is to optimize the entry of

information into the hearers knowledge store his primitives are dened towards that

goal So for example the link indicates where the informative part of the sentence

ie fo cus should go within the hearers knowledgestore ie a lo cus of information

entry and the tail indicates how it ts there Using the Heimian le card analogy

he compares the hearers knowledgestore to a pile of le cards and information

packaging to up dating those le cards The fo cus is the information which the sp eaker

wants the hearer to up date his le cards with The link instructs the hearer as to

which card he should keep the record of the information on and the tail further

sp ecies where exactly in that le card p ointedtoby the link the hearer should keep

CHAPTER SPECIFICITY FOCUS AND INFORMATION STRUCTURE

the record and it also indicates whether the information should simply b e added or

replaced In more traditional terms however wecaninterpret link as topic Gundel

Reinhart or theme Firbas Halliday Kuno whichis

roughly sp eaking what the sentence is ab out and tail as the rest of the ground or

the given information which is somewhat inconspicuous in the sentence

Let us lo ok at the examples in and Vallduv Both b and

b more or less carry the same truthconditional meaning They mean that there

was a giving event at some past p oint of time where the giver is Mary the given

is a shirt and the recipient is Harry This is briey illustrated in However

sentence b and sentence b carry information of two distinct sorts Although

they share the informative part ie the fo cus of the sentence they dier in what

eachsentence is ab out Fo cused elements are represented in upp er case and topics

in small capitals

a What ab out MaryWhat did she giv e to Harry

b Mary gave a SHIRT to Harry

Link Focus

a What ab out HarryWhat did Mary giv e to him

b To harry Mary gave a SHIRT

Link Focus

givexyz xMary ya shirt zHarryTensePast

Since b oth b and b are the answers to the same question ie What did

Mary give to Harryhim it is easy to detect what the fo cus of the sentence is for

b oth of them it is the part which corresp onds to the question phrase whatiea

shirt The link of the sentence is signaled byawhat about phrase This phrase

has the function of highlighting an element among the given information In

the preceding discourse singles out Mary as the elementwhich the answer should pay

attention to and thus Mary is the link in b Namely Mary is what the sentence

CHAPTER SPECIFICITY FOCUS AND INFORMATION STRUCTURE

is talking ab out On the other hand in Harry is singled out as the link thus

b is ab out to Harry Finally the tail is the part of the sentence whichis

neither fo cus nor link ie inconspicuous old information Thus the tail in b is

gave to Harry and that in b is Mary gave

Nowthatwehave an example of what information structure lo oks like lets return

to our original problem ie how the semantic and discourse eects of scrambling

can b e explained and see how the idea of information structure can apply to them

Putting the sp ecicity eect aside for the momentwhich will b e discussed in section

let us rst consider the two fo cusrelated eects To recall the rst is the anti

fo cus eect that a nonfo cused or topical element can scramble and a fo cused element

cannot scramble The second quite contrarily is that a contrastively fo cused element

can scramble I will rst discuss the antifo cus eect in the next section and

then the contrastive fo cus problem in the section following that

Topic and Tail

The antifo cus constraint requires that a fo cused element cannot scramble while a

nonfo cused or defo cused elementcan In Vallduvs information structure

intro duced ab ove the distinction b etween fo cus and nonfo cus is quite clear The

fo cused element is the fo cus ie new and informative information and the non

fo cused or defo cused element is the ground ie old or given information The

ground is further classied into link and tail Therefore the antifo cus constraintcan

b e reinterpreted as while link and tail can scramble fo cus cannot This prediction

holds as will b e shown b elow

In the discussion of the antifo cus eect earlier in this chapter wehave only paid

attention to the fo cusnonfo cus distinction and have not really taken notice of the

linktail distinction Although it is often the case that nonfo cusedness or defo cused

ness is equated with topicality Abraham Web elhuth Moltmann

CHAPTER SPECIFICITY FOCUS AND INFORMATION STRUCTURE

caution should b e taken ab out what topicality really means Topic sometimes refers

to any old information esp ecially when the binary distinction of topic and fo cus is

the only available classication On the other hand topic can refer to a subset of

old information namely only the salient or prominent part as in the case of link

in Vallduv In what follows I will use the what about X phrase following

Vallduv to identify the link or the topic in its narrow sense of the

sentence

Let us rst consider the tail case We can assume that some given old information

is the tail when there has b een no particular attention paid to it for instance when

it is not highlighted with a what about phrase or no pitch accentisgiven to it Let

us lo ok at the question and answer pair in also taken from Lenerz

a Wem hast du das Geld gegeb en

to whom haveyou the money given

To whom did you give the money

b Ich hab e dem Kassierer das Geld gegeb en

I havethe cashier the money given

I gave the cashier the money

b Ichhabe das Geld dem Kassierer gegeb en

I have the money the cashier given

I gave the money to the cashier

Das Geld the money in is a tail it is nonsalient old information It is not

the fo cus the fo cus is dem Kassierer here nor is it the link there is no link in this

context As a tail das Geld can indeed scramble In b it scrambles over the

indirect ob ject dem Kassierer the cashier

What ab out the linkAs exp ected a link can also scram ble Consider the example

belowin

CHAPTER SPECIFICITY FOCUS AND INFORMATION STRUCTURE

a Wie stehts mit dem GeldW em hast du das Geld gegeb en

how stands it with the money to whom haveyou the money given

What ab out the moneyT owhomdidyou give the money

b Ichhabe das Geld dem Kassierer gegeb en

I have the money the cashier given

I gave the money to the cashier

Being in the wie stehts mit phrase ie the German counterpart to what about das

Geld the money is presented as the link ie salient or prominent old information

As shown in b this prominent old information is out of its canonical p osition

Therefore the antifo cus constraints holds in that the nonfo cused part of the sentence

can scramble that is link or tail can scramble

The next question to ask is if the further distinction of link and tail in the ground

elements makes any dierence in the scrambling phenomena that is if prominence

counts in scrambling Vallduv argues that in Catalan link b ehaves dierently

from tail in that the former go es to the right side of the sentence while the latter go es

to the left side to make ro om for fo cus so the threeway distinction in his information

structure works very well there In scrambling also link seems to b ehave dierently

from tail here it is the degree of scramblability rather than the direction found in

Catalan

There are twoways link diers from tail First link scrambles more often than

tail although tail can stay in the base p osition link tends to scramble whenever

p ossible Notice that in the tail example in b oth the nonscrambled sentence in

a and the scrambled one in b are equally available Lenerz whereas in

the link example in the nonscrambled sentence is not available as an answer in

that context Second only link can scramble over the sub ject Scrambling over the

sub ject is fairly marked in German but when it is p ossible it is only link a very

CHAPTER SPECIFICITY FOCUS AND INFORMATION STRUCTURE

prominently presented element that can b e p ositioned b efore the sub ject Compare

and

a Wie stehts mit dem GeldW em hat Hans das Geld gegeb en

how stands it with the b o ok to whom has Hans the money given

What ab out the moneyT o whom did Hans give the money

b Ich glaub e da das Geld Hans dem Kassierer gegeb en hat

I b elieve that the money Hans the cashier given has

I b elieve that Hans gave the money to the cashier

a Wem hat Hans das Geld gegeb en

to whom has Hans the money given

To whom did Hans give the money

b Ich glaub e da das Geld Hans dem Kassierer gegeb en hat

I b elieve that the money Hans the cashier given has

I b elieve that Hans gave the money to the cashier

Using the what about phrase again to identify the link of the sentence in

the link das Geld can scramble over the sub ject Hans However when das Geld is

interpreted as a tail as in the scrambled sentence b ecomes muchworse tail

cannot scramble over the sub ject

Thus let me close this section with the following generalization as to the scram

blability of the ground elements

Scrambling of Ground elements

a Ground elements b oth link and tail can scramble

b Link more easily scrambles than tail

CHAPTER SPECIFICITY FOCUS AND INFORMATION STRUCTURE

In other words as the antifo cus constraint predicts nonfo cused or old elements are

allowed to b e out of their canonical p ositions In addition it app ears that prominent

elements achieve more freedom in scrambling than nonprominent elements

ContrastiveFo cus and CompletiveFocus

Now let us turn our attention to the fo cused part of the sentence As mentioned

rep eatedly the antifo cus constraint indistinguishably prohibits any fo cus from scram

bling Wehave observed that acting against the antifo cus constraint however an

element that is a contrastive fo cus can scramble and b e out of its canonical p osition

This seems to indicate that there are in fact two distinct typ es of fo cus one a kind

which prohibits scrambling and the other one whichallows it Lenerz Molt

mann I argue in fact that the prop erty which distinguishes these twotyp es

of fo cus is also prominence I examine the nature of the two kinds of fo cus in this

section

The idea that more than one typ e of fo cus exists is by no means new Contrastive

fo cus has b een noted as distinguishable from a purely newinformation typeoffocus

by several linguists It has also b een p ointed out that contrastivefocusmay not carry

really new information Herring distinguishes presentational focus

and contrastive focus and argues that the information status of contrastive fo

cus and presentational fo cus diers in that arguments presented for the rst time

eg as participants in a narrative are completely new while contrastively fo cused

arguments are already explicitly or implicitly present in the discourse context She

takes an entityintro duced by the thereconstruction in English as a typical example of

presentational fo cus and that intro duced byanitcleft construction as an example of

contrastive fo cus SimilarlyRochemont and Ro chemont and Culicover

argue that contrastive focus unlike presentational focus do es not haveto

b e totally new information Using the notion of contextconstruability whichis

CHAPTER SPECIFICITY FOCUS AND INFORMATION STRUCTURE

dened to b e under discussion or to have a semantic or discourse antecedentinthe

discourse Ro chemont Ro chemont and Culicover they dene pre

sentational fo cus as not cconstruable but they dont require such a restriction for

contrastive fo cus Finally Dik et al also dierentiate completive focus

and contrastive focus and argue that the latter case often involves some pre

supp osed alternatives while the former simply lls in the information gap b etween

the sp eaker and the addressee

Contrastive fo cus as sketched ab ove is often characterized as b eing not entirely

new although it certainly carries new or informative information otherwise it

would not b e called a fo cus at all The main reason that contrastive fo cus is

often cited as not b eing so new as presentational or completive fo cus is that b eing

contrastive the fo cused element is compared with or even opp osed to something

else this comparison or opp osition may b e either explicit or implicit or stated or

predicted Halliday Thus a set of p ossible candidates for the role played by

the element which is b eing contrasted Chafe or a set of alternatives cf

Ro oths CSet is created This set of alternatives whether already present

in the discourse or accommo dated at the time of the utterance contexualizes the

information conveyed by the ob ject in fo cus hence making it less new in a sense

Consider the following example in Dik et al Supp ose that a p erson

Aasked another p erson B the following question What did John buy and that the

examples in are answerstoit

a John b oughtaToyota

b John b oughtaTOYOTA

c John b oughtaTOYOTA not a VOLKSWAGEN

Person B can giveananswer as in a to the question simply providing the in

formation that p erson A lacks at the moment ie that it is aToyota that John

CHAPTER SPECIFICITY FOCUS AND INFORMATION STRUCTURE

bought This is a regular newinformation fo cus ie a presentational or completive

fo cus whose ma jor function is to ll in the informational gap b etween the sp eaker

and the hearer On the other hand the answer in b puts the information in

contrastivefocusby means of a heavy accenton Toyota or with some extralinguistic

devices such as gestures or facial expressions Alternativelycontrastive fo cus can

b e more explicitly expressed by an accompanying alternative phrase as in c An

answer like b or c implies that the information that the sp eaker is providing

is contrary to hisher andor the hearers exp ectation For example it can imply

that the sp eaker thought that knowing his taste John would buy a Volkswagen for

example instead of a Toyota or the sp eaker thought that John would buy a Volkswa

gen b ecause he usually likes German cars and so on In other words in the sp eakers

mind and also in the hearers mind as the sp eaker conceives of it there is a set of

alternatives which includes a Volkswagen as a more plausible candidate Hence the

existence of p otential alternatives makes the currently fo cused item prominent so

that the fo cus ie contrastive fo cus gets emphasis and evokes a contraexp ectation

eect

Dik et al prop ose several dierenttyp es of contrastive fo cus select

ing restricting expanding replacing and parallel Not surprisingly all these sub

typ es either explicitly or implicitly preexistent or accommo dated assume a set of

alternatives Examples of eachtyp e are illustrated as follows

Selecting

a Did Andrew buy cho colate or our

b He b ought CHOCOLATE

Restricting

Dik et al actually do not consider the parallel fo cus to have a set of alternatives b ecause it is

not presupp osed at the time of utterance However I b elieve that the parallel fo cus also involves a

set of alternatives it is accommo dated at the time of utterance if not already presupp osed ie the

elements in parallel fo cus immediately form a contrast set

CHAPTER SPECIFICITY FOCUS AND INFORMATION STRUCTURE

a Since Andrew b oughtcho colate and our he can makea cake

b No he only b ought CHOCOLATE

Expanding Focus

a Since Andrew b oughtcho colate he will b e happy

b Yes but he also b ought FLOURsohecanmake a cake

Replacing Fo cus

a Andrew went to New Mexico

b No he wenttoUTAH not NEW MEXICO

Parallel Fo cus

Andrew b oughtaSTARSHIP but Peter b ought a PLANET

Note that although the existence of a set of alternatives may b e necessary to evoke

a contrastive fo cus it is not b e sucient For instance the ve kinds of contrastive

fo cus intro duced in are not necessarily conceived of as equally contrastiveor

prominent even though all of them evoke sets of alternatives It app ears that avail

ability of alternatives do es contribute to prominence but it do es not guarantee it

Now with the distinction of completive and contrastive fo cus in mind let us return

to one of the dilemmas we had with resp ect to the fo cusrelated eects of scrambling

ie the apparent contradiction b etween the two generalizations one that a fo cused

element cannot scramble and the other that a contrastively fo cused element can

Compare and again

a weil Hans das BUCH dem Mann gegeb en hat

b ecause Hans the b o okAcc the manDat given has

b ecause Hans gave the b o ok to the man

CHAPTER SPECIFICITY FOCUS AND INFORMATION STRUCTURE

b weil Hans ein BUCH dem Mann gegeb en hat

b ecause Hans a b o okAcc the manDat given has

b ecause Hans gave a b o ok to the man

a weil Hans das BUCH dem Mann gegeb en hat nicht die ZEITUNG

b ecause Hans the b o ok the manDat given has not the newspap er

b ecause Hans gave the b o ok to the man not the newspap er

b weil Hans ein BUCH dem Mann gegeb en hat nicht eine ZEITUNG

b ecause Hans a b o ok the manDat given has not a newspap er

b ecause Hans gave a b o ok to the man not a newspap er

Note that the surface forms in and are identical except for the phrases in

parentheses Now that wehavetwo distinct typ es of fo cus in hand though we

can see that the fo cus in the examples in is a completive fo cus or a purely new

information typ e of fo cus and the fo cus in the examples in is a contrastivefocus

ie a replacing kind in Dik et als classication In the latter a set of alternatives

to the fo cused elementisevoked in each case which includes die Zeitung and eine

Zeitung resp ectivelythus making the fo cused item das Buch or ein Buch prominent

Thus the acceptability of the ab ove scrambled structures dep ends on the contextual

information ie whether or not the context is contrastive enough or whether or not

the scrambled element is presented prominently enough

It is interesting to note that Dik et al include as contrastive the restricting and

the expanding fo cus whichinvolve fo cus particles suchas only alsoandeven upp er

b ounded and lowerb ounded scalars resp ectively in Horn This class of words

In fact a contrastive fo cus usually receives a stronger pitch accent than a completive fo cus al

though I put b oth of them in upp er case This may b e b ecause in a contrastive fo cus its prominence

contributes to the pitch accent in addition to its newness doing so

CHAPTER SPECIFICITY FOCUS AND INFORMATION STRUCTURE

often called fo cus adverbs or fo cussensitive particles has b een claimed to havean

intimate connection with fo cus This phenomenon has b een called asso ciation with

fo cus Jackendo Ro oth When an element is mo died by only

or even then it presupp oses that there already exist some alternatives to it only

restricts the set of alternatives bypicking only one memb er out of it and even

expands the set by adding a member to it Recall that in section even a

nonsp ecic phrase can scramble if it is mo died with one of these fo cus adverbs

The examples are rep eated here in

a weil Hans NUR ein BUCH dem Mann gegeb en hat

b ecause Hans only a b o okAcc the manDat given has

b ecause Hans gave only a b o ok to the man

b weil Hans SOGAR ein BUCH dem Mann gegeb en hat

b ecause Hans even a b o okAcc the manDat given has

b ecause Hans gaveeven a b o ok to the man

If we assume following Dik et al that the fo cus asso ciated with adverbs such

as only and even is contrastive fo cus as distinct from completive fo cus then the

set of examples in section which showed anomalous b ehavior with resp ect to

scrambling can naturally fall under the cover of contrastivefocus

The discussion of the fo cusrelated eects of scrambling in this section leads us to

the following generalization regarding the fo cused elements

Scrambling of Fo cus elements

a Completive fo cus cannot scramble

b Contrastive fo cus can scramble

Neeleman provides similar examples involving the fo cus adverbs in Dutch They also

showunusual scrambling b ehavior

CHAPTER SPECIFICITY FOCUS AND INFORMATION STRUCTURE

Therefore the apparent contradiction noted b etween the antifo cus eect and the

contrastive fo cus eect can b e understo o d as the result of partial observation of the

fo cus phenomena the former only refers to completive fo cus whereas the latter to

contrastivefocus

What is more interesting is that it is prominence which dierentiates the two

typ es of fo cus Namely prominence is what counts in scrambling in fo cus elements

b ecause only the prominent fo cus can scramble Recall that we describ ed link or

topic also as prominent in the previous section which implies that link and contrastive

fo cus share the same prop erty I argue this is indeed the case Lets turn to next

section for supp orting evidence

Topic and ContrastiveFocus

Topic also can b e conceived of as contrastive in standing out among other p otential

topical elements in the discourse That is a topic is compared either with other

alternatives within the sentence ie the tail elements or by b eing implicitly or ex

plicitly contrasted with other topics outside of the sentence Consider the following

examples and again rep eated from section

a What ab out MaryWhat did she giv e to Harry

b Mary gave a SHIRT to Harry

Link Focus

a What ab out HarryWhat did Mary giv e to him

bTo harry Mary gave a SHIRT

Link Focus

In for example the fact that this sentence is ab out Mary among the ground

items can b e interpreted as that the sentence is not ab out the tail elements eg

In this sense the current notion of topic is more likethatofcontrastive or shifted topic rather

than that of continuous or continuing topic Herring Aissen Continuing topics need not

b e prominent and only refer to an entity which is already intro duced into the discourse They are

more like tail in the current system

CHAPTER SPECIFICITY FOCUS AND INFORMATION STRUCTURE

Harry or a giving event On the other hand the fact that is ab out Mary can

also mean that the sentence is not ab out other topics eg the topic to Harry in

cf linkcontrast in Vallduv contrastive topic in Szab olcsi

Iprovide two cases where topic and contrastivefocusbehavealike One is the syn

tactic op eration topicalization in English and the other is the morphological marking

by the socalled topic marker nun in Korean I argue that these are mechanisms

which enco de prominence of elements one syntactically and the other morpholog

ically

Topicalization and Fo cus Movement

It is often noted that topicalization in English is not a uniform phenomenon in terms

of the information status of the fronted item Namely the fronted topicalized el

ement is sometimes interpreted as topic but sometimes interpreted as fo cus

Thus Gundel for example separates a Topic topicalization from a Fo cus

topicalization Chafe likewise notes these twotyp es of topicalization and

calls the rst a topicalization with twofociofcontrast the rst fo cus b eing topic

and the second fo cus b eing fo cus as in and the second a topicalization with

a single fo cus of contrast as in Prince also distinguishes the former

Topicalization from the latter Fo cus Movement The following are some exam

ples The small capitals represent topics and the large capitals represent fo cus b oth

contrastive and completive

This is true with German topicalization to o although the syntactic prop erties of topicalization

are slightly dierent in that the nite verb is necessarily in the second p osition Recall the discussion

of Verb Second in chapter

She actually identies one more typ e ie YiddishMovement which only o ccurs in a dialect of English

CHAPTER SPECIFICITY FOCUS AND INFORMATION STRUCTURE

a Beans I dont LIKE Ross

b The play John saw YESTERDAY Chafe

The fronted elements in beans and the play are the topic of each sentence They

are what the sentences are ab out and thus are presented more prominently than the

tail elements in the sentence eg I in a and John in b The part following

the topic is the comment to it and therefore the comment can include some new

information or fo cus suchas not like in a and yesterday in b

In contrast the fronted element ie Fido in is fo cus ie new information

The elements to its right is the ground material whichmay include a topic eg their

dog or they may b e all tail

FIDO they named their dog Prince

I argue that this fronted fo cus is contrastive fo cus ie prominent fo cus The fo cus in

is much more emphatic or prominent than the one for instance in

They named their dog FIDO

Several p eople havepointed out that the fo cus enco ded byafronted item is dierent

in its informational imp ort from that enco ded by a nonfronted counterpart Ward

Vallduv Ward for example shows that in cases like the

fronted phrase actually refers to two discourse elements one a set or scale and the

other a sp ecication of a value or an element in that set or scale This is exactly

the prop ertywe used to characterize contrastive fo cus a set of alternatives and a

value in that set In this alternative set is the set of dog names and the value is

Fidowhich is an unexp ected one In other words in there is a strong sense of

comparison or contrast of the value Fido with other alternativevalues thus making

this value really stand out among the alternatives This standingout prop ertyis

prominence In contrast in the value Fido is not necessarily compared or

contrasted with other values although it could b e

CHAPTER SPECIFICITY FOCUS AND INFORMATION STRUCTURE

To summarize topicalization in English not only enco des the topicho o d of the

fronted element but also expresses the contrastive fo calityIfwe assume that topic

and contrastive fo cus share the prop erty of b eing prominent in discourse English

topicalization can b e regarded as a uniform phenomenon namely an op eration of

enco ding prominence

NunMarking in Korean

Morphological marking by the socalled topic marker nun in Korean is another case

where topic and contrastive fo cus are group ed together That is a phrase marked

with nun is always either a topic or a contrastive fo cus but never a tail or a completive

fo cus Namely nun is a prominence marker in Korean

The ax nun in Korean has traditionally b een called the topic marker b ecause

when it is attached to the sub ject it enco des the topicho o d of the sub ject The

topic marker nun is distinguishable in its topicenco ding prop erty from the regular

nominative case marker kaAkamarked phrase is discourseneutral This distinction

is illustrated in a and b

a Swunika Inholul mannassta

SwuniNom InhoAcc met

Swuni met Inho

b Swuninun Inholul mannassta

SwuniTop InhoAcc met

As for Swuni she met Inho Topic

The nominativecase marked sub ject Swunika in a simply functions as the sub ject

of the sentence and is not restricted in terms of discourseinformational prop erties

However the nunmarked sub ject Swuninun in b expresses the fact that the

sentence is ab out Swuni ie Swuni is the topic of the sentence

CHAPTER SPECIFICITY FOCUS AND INFORMATION STRUCTURE

In contrast when nun marks a nonsub ject element as in b it has a dier

ent function It expresses the fact that the phrase whichismarked with nun is a

contrastivefocus

a Swunika Inholul mannassta

SwuniNom InhoAcc met

Swuni met Inho

b Swunika Inhonun mannassta

SwuniNom InhoTop met

Swuni met Inho but nob o dy else ContrastiveFocus

The ob ject Inhonun in b is implicitly compared with other alternative p eople

and the sentence implies that the situation is not necessarily truefor those alternatives

When the ob ject is marked with the regular accusative case marker lul as in a

this reading is not evoked

The distinction b etween topic and contrastive fo cus shown in and cannot

simply b e reduced to that of grammatical relations ie sub ject versus nonsub ject

Word order also plays a role a and b are the scrambled variants of b and

b resp ectively

a Inholul Swuninun mannassta

InhoAcc SwuniTop met

As for Inho Swuni but nob o dy else met him ContrastiveFo cus

b Inhonun Swunika mannassta

InhoTop SwuniNom met

As for Inho Swuni met him Topic

Interestingly when the nunmarked sub ject is scrambled over by another phrase in the

sentence as in a it loses its topic reading but instead achieves a contrastivefocus

CHAPTER SPECIFICITY FOCUS AND INFORMATION STRUCTURE

reading On the contrary when the nunmarked ob ject scrambles over another phrase

as in b it loses its contrastive fo cus reading and b ecomes the topic of the sentence

This problem will b e discussed in greater detail in chapter in connection with the

information structuring constraints presented in the later part of this chapter

What is imp ortant with this morphological marking in Korean is that even though

the reading of a nunmarked phrase maychange dep ending on its relative p osition in

the sentence as demonstrated in nun always enco des either topic or contrastive

fo cus but nothing else That is a nunmarked phrase whether sub ject or non

sub ject or whether sentenceinitial or not cannot b e tail or completive fo cus In

Korean literature Cho e there has b een a controversy as to whether nun should

b e classied as a topic marker or as a contrast marker This problem is easily solved

if we assume that b oth topic and contrastive fo cus are prominent nun in Korean is

the prominence marker

To summarize topic is an elementofasentence which is singled out and then

talked ab out among several p otential alternatives in the discourse These alternatives

are already anchored in the discourse ie they are ground or oldknown informa

tion Contrastive fo cus likewise is singled out among several p otential alternatives

However in this case the alternative are new information ie p otentially fo cal

items I have called this prop erty of b eing singled out among p otential alternatives

prominent Both topic and contrastive fo cus are prominent discourse elements

Scrambling and Information Structuring Con

straints

To recapitulate the discussion in the previous section I have argued that scram bling cannot b e prop erly understo o d without a deep er understanding of information

CHAPTER SPECIFICITY FOCUS AND INFORMATION STRUCTURE

structure While examining the antifo cus eect and the contrastive fo cus eect in

Vallduvs framework of information packaging I have prop osed a further di

vision of fo cus according to prominence Moreover it was suggested that Topic and

ContrastiveFo cus share the prominence prop erty

Based on the ndings in the previous section I prop ose a featurebased information

structure which can systematically capture the crossclassicatory nature of the four

information units In the following part a set of constraints are intro duced which

participate in the mapping b etween information structure and phrase structure I

argue that these information structuring constraints are the ma jor motivations for

alternative phrase structural descriptions of a sentence namely scrambling

FeatureBased Information Structure

Crossclassifying Features

The scrambling facts in German call up on us to make some revisions to Vallduvs

information packaging system intro duced in section As shown in Valludv

do es not further classify fo cus into any subgroups

S ffo cus groundg

ground flink tailg

To capture the distinct b ehavior of contrastive fo cus in scrambling therefore the

notion of fo cus should b e nergrained Initially I prop ose that fo cus b e divided into

completive fo cus and contrastive fo cus just as the ground is divided into link and

Vallduv argues that contrastive fo cus or fo cuscontrast in his terms is derivable

in his system Using the lecard analogy again Vallduv argues that there are actually twoways of

keeping the record on a le card one is to simply add the information and the other is to substitute

the information for an already existent record In our terms the former is completive fo cus and the

latter is contrastive fo cus In eect Vallduv also has two kinds of fo cus by letting the two

dierent recordkeeping mechanisms addition and substitution derive the informational dierences

CHAPTER SPECIFICITY FOCUS AND INFORMATION STRUCTURE

tail Hence I suggest the basic information structure in byaddingcontrastive

fo cus to Vallduvs system I hereafter return to the more traditional term

topic instead of link b ecause I am not necessarily assuming the information storage

function implied in the term The new information structure can b e illustrated as

S ffo cus groundg

ground ftopic tailg

fo cus fcompletivefocuscontrastivefocusg

With the further division in fo cus added to the system wenowhave four dierent

typ es of information units ie topic tail completive fo cus and nally contrastive

fo cus As shown in topic and tail are group ed together as ground materials

ie old or given information in discourse while completive fo cus and contrastive

fo cus are classied as fo cus materials ie new or novel information That is the

characteristic which distinguishes topic and tail on the one hand and completive

fo cus and contrastive fo cus on the other hand is the discoursenewness I will

call this feature New Thus the former are marked New and the latter marked

New Then as wehave observed in the previous section b oth ground and fo cus

can b e further partitioned in terms of discourseprominence which will b e called

Prom here Topic and contrastive fo cus are prominent while tail and completive

fo cus are not So the former are Prom and the latter Prom

The two information partitioning features New and Prom therefore givethe

following crossclassication of the four information typ es

CHAPTER SPECIFICITY FOCUS AND INFORMATION STRUCTURE

New New

Topic Tail ContrastiveFocus CompletiveFo cus

Prom Prom

Each information typ e is then represented by a pair of features in the new feature

based system which is illustrated in

Topic ContrastiveFocus Tail CompletiveFocus

Prom

New

One of the advantages of this featurebased information structure is that it can cross

refer to more than one distinct informational typ e For example topic and tail can b e

group ed together as b eing New as Vallduvdoesby calling them ground Also

we can crossrefer to topic and contrastive fo cus together as Prom elements This

crosscutting prop erty of the system is absolutely crucial in explaining the discourse

eects of scrambling as we will see shortly In the discrete primitivebased system

suchasVallduvs however it is not easy to explain why topic and con

trastive fo cus should b ehavealike for example As mentioned ab ove topic b elongs

to ground and contrastive fo cus b elongs to fo cus and in Vallduvs system they

are only two distinct kinds of information Now with the new featurebased system

we can easily refer to topic and contrastive fo cus together as Prom items Atthe

same time their dierences the one b eing old information and the other b eing new

information can easily b e captured bythevalues of the New feature ie the former

is New and the latter New

CHAPTER SPECIFICITY FOCUS AND INFORMATION STRUCTURE

Restrictions on Feature Marking and Undersp ecication

SofarIhave implicitly assumed that every element of a sentence equally participates

in the information structuring as an indep endententity and accordingly that each

element has an equal chance to b e sp ecied with the discourse features New and

Prom Of course this is to o simplistic a view of information structuring

First of all some lexical items may not function as an indep endent discourse

entity p erhaps due to their intrinsic morphological prop erties and hence cannot b e

marked with those features Complementizers particles and probably some adverbs

are examples of such items and they may not hold discourse features indep endently

Even a more contentful element is not always sp ecied with these features as an

indep endent unit if it is part of a bigger information unit For example recall the

fo cus pro jection examples discussed in chapter In the Korean example in the

verb ssu will b e sp ecied NewProm according to our information structure since

it is marked with the contrastive fo cus marker nun

Maryka emenieykey phyencilul ssukinun hayssta

F F

MaryNom motherDat letterAcc writeNmlTop doPstDcl

Mary did write her mother a letter

As noted b efore this sentence can b e interpreted in several dierentways dep ending

on how far up the fo cus in the verb is pro jected within the VP This is shown in

a Mary WROTE a letter to her mother but she did not SEND it to her

b Mary WROTE A LETTER to her mother

but she did not SEND A PRESENT to her

c Mary WROTE A LETTER TO HER MOTHER

but she did not SEND A PRESENT TO HER FATHER

CHAPTER SPECIFICITY FOCUS AND INFORMATION STRUCTURE

This indicates that other elements in the VP ie emenieykey to mother and phyenci

lul the letter are not sp ecied by themselves in terms of the discourse features and

that those of the verb ssu ie NewProm are pro jected onto them If the

pro jection is carried all the way up to the VP as in c the VP will havethe

following feature marking

emenieykey phyencilul ssukinun

NewProm

New Prom

On the other hand it is also p ossible that none of the internal elements have inde

p endent feature markings but that only the outer unit which contains the elements

do es This kind of situation happ ens when a VP for example is presented as new

information as a whole Consider the following example

a What did Mary do yesterday

b She wenttoscho ol

In b the whole VP went to school is the fo cus the new information and the inter

nal elements do not necessarily have indep endent feature marking This means that

the contained elements are dep endent on the larger unit and they are informationally

neutral with resp ect to each other This is shown in

went to scho ol

NewProm

Information structuring is thus sub ject to several dierenttyp es of constraints

These constraints may b e morphological whichprevent certain lexical items from

b eing assigned any discourse features They maybesyntactic whichmake a certain

phrase form an information unit Or they maybesemantic which prohibit a certain

element from receiving a sp ecic feature due to semantic incompatibility In these

CHAPTER SPECIFICITY FOCUS AND INFORMATION STRUCTURE

cases some elements in the sentence may not b e able to b e given indep endent feature

marking It is b eyond the scop e of this dissertation to explore the exact feature

marking mechanism However I will argue later that the sp ecicity eect is one of

the cases where a semantic restriction on information feature marking is applied In

what follows for the purp oses of scrambling I will assume a at information structure

in which all nonverbal arguments are equally sp ecied with New and Prom In

other words scrambling of a phrase is p ossible when the phrase has an indep endent

information status

Information Structurin g Constraints

Now let us recall the scrambling b ehavior of each information typ e which is gener

alized as in and

Scrambling of Ground elements

a Ground elements b oth topic and tail can scramble

b Topic more easily scrambles than tail

Scrambling of Fo cus elements

a Completive fo cus cannot scramble

b Contrastive fo cus can scramble

Given the newly prop osed information structure based on crosscutting features I

prop ose that the scrambling b ehavior can b e reduced to two generalizations The

rst is a requirement on the New information ie topic and tail and the other

is on the Prom information ie topic and contrastive fo cus I will call these two

Information Structuring Constraints They are dened in

CHAPTER SPECIFICITY FOCUS AND INFORMATION STRUCTURE

Information Structuring Constraints

a new ANew element should precede a New element

b prom A Prom element should precede a Prom element

The rst constraint new immediately explains the rst generalization in a that

a ground element topic or tail whichisNew in our information structure can

b e out of its canonical p osition unless it already precedes the New elements in

the sentence in the canonical conguration In other words if topic or tail do es

not precede fo cus in the canonical structure this constraintgives preference to a

scrambled structure in which the former precedes the latter On the other hand

the second constraint prom explains the distinction b etween completive fo cus and

contrastive fo cus expressed in this constraint p ermits only a contrastivefocus

ie a Prom element to scramble Also prom is resp onsible for the contrast

between topic Prom and tail Prom in their scrambling p ossibilities describ ed

in b The information structuring constraints new and promworking together

yield the following consequences Firstly a completivefocuswhichisNewProm

is the least likely element to b e out of its canonical p osition b ecause neither new nor

prom motivates it to scramble Also they yield another consequence that topic

whichisNewProm is most likely to scramble b ecause b oth new and prom

endorse its scrambling As exp ected however the alternative ordering is restricted

by the phrase structural constraints canon Recall that canon would act against

any noncanonically ordered structures We will see this interaction in chapter and

in great detail

Note that these constraints are dened simply as a precedence relation and not as

a hierarchical notion like ccommand The underlying motivation b ehind this is to

make the constraint general enough to apply to languages with various phrase struc tures including at ones rather than just to languages with highly congurational

CHAPTER SPECIFICITY FOCUS AND INFORMATION STRUCTURE

structures Therefore in actual application the precedence relation can b e realized in

several dierent forms dep ending on the language If the phrase structural constraints

of a language are very minimal eg b ecause the constraint economy of expres

sion outranks the the endocentricity alignment constraints as suggested by

Bresnan so that the language is noncongurational having a at sentence

structure then the precedence entails linear precedence Therefore the constraints

in would have the eect of swapping two constituents to change the linear order

if necessaryIncontrast if the phrase structural constraints of a language are dened

and ranked such as to prefer highly articulated rightbranching phrase structures

such as those in German then the precedence results from the ccommand relation

In congurational languages this constraintwould make a constituent scramble ie

b e adjoined to a higher maximal pro jection or ob jectshift ie b e generated in a

higher Sp ec p osition

The information structuring constraints in short tend to realign the order among

the constituentelements of the sentence according to each elements information

status In fact the idea that discoursefunctional notions suchasnewandoldor

topic and fo cus are resp onsible for word order has a long historyFor example the

idea that old information precedes new information which is the basis of the rst

constraint newisbynomeansnovel in researchonword order in the Prague Scho ol

or in other functionalist work Sgall et al Li and Thompson among many

others It has b een argued in many places that topic precedes comment theme

precedes rheme old information precedes new information etc However it has

also b een noted that the mere oldnew distinction hardly captures all the word order

Whether an element out of its canonical p osition is adjoined or generated in a Sp ec p osition

will b e further constrained by other relevant considerations of the language which is not our main

concern here It can b e noted however that generating something in a Sp ec p osition will b e more

costly than adjoining it b ecause the former involves generating a new pro jection which is against

the general constraintofeconomy of expression

CHAPTER SPECIFICITY FOCUS AND INFORMATION STRUCTURE

generalizations Givon Ch for example argues that the Praguean view on

word order based on the oldnew distinction do es not always work and p oints out

that in many languages more imp ortant or urgent information tends to b e uttered

earlier in a sentence and that this information do es not have to b e old This notion

of imp ortance or urgency is analogous to the notion of prominence in our feature

based information structure see also Payne and Austin in press for similar

prop osals

The dierence in the current information structuring constraints lies in that they

are OT constraints Each constraint is violable and interacts with other constraints

in terms of the ranking among them Moreover these constraints also interact or

rather comp ete with phrase structural constraints so the word order is no longer

a matter of simply lining up arguments according to a single typ e of information

whether syntactic or discoursefunctional It is viewed rather as a complex interface

phenomenon

Before closing this section let me intro duce another information structuring con

straint in addition to the twointro duced ab ove This one is a proso dic structural

requirement on new information Although this constraint is not directly related to

word order itself it has an indirect eect on the antifo cus eect of scrambling

Proso dic Constraints

a N Put a high pitch accent on a New element

b X Do not place anypitch accent

The rst constraint N says to put a high pitch accent on a newinformational

element This constraint is to capture the fact that fo cus elements completiveor

contrastive have high pitch accents or proso dic prominence while topic and tail do

not Topical elements can also havesomeamountofpitchaccent but it is smaller

CHAPTER SPECIFICITY FOCUS AND INFORMATION STRUCTURE

compared to fo cused elements so I will assume for the purp oses of the present discus

sion that only fo cal elements havepitch accents to capture the categorical distinction

between old and new elements in terms of proso dic prominence Obviously this view

of sentential intonation is greatly simplied and by no means comprehensive but it

is serviceable for the purp oses of the scrambling facts see Jackendo Prince

and Steedman and the references therein for more detailed discussion

on sentential intonation in terms of information status of elements of the sentence

The second constraint X in contrast says not to put a high pitch accenton

anything This constraint therefore is naturally in conict with the N constraint

Again the conict is resolved by the ranking b etween them As exp ected N is

ranked higher than X

Ranking

N X

This ranking will have the eect of favoring a fo cused element with a high pitch accent

rather than one without Actually almost all fo cused elementsat least in German

seem to get pitch accents abstracting away from the default sentence intonation If

this is true the two constraints could b e collapsed into one constraintwhichwould

guarantee the biuniqueness of the new information and the high pitch accent eg by

making the N constraint bidirectional Vivienne Fong p c all and only New

An alternativeviewwould b e to say that Prom also receives a pitch accent ie P Then

a contrastive fo cus NewProm will have the highest pitch reinforced byN and then a

completive fo cus and a topic will have some and nally a tail will not haveany pitch accent

Although it seems quite plausible I do not pursue this line of approach in order to consistently

present the data provided by Lenerz and Abraham who equate fo cus and high pitch

accent Also in Korean the only way to distinguish topic and contrastive fo cus in some cases eg

in the sentenceinitial p osition is the pitch accent as will b e discussed in chapter Thus I will

maintain the ab ove assumption that only New elements are pitchaccented

Even without anycontextual information a sentence has a default intonation pattern in which

some nal constituent of the sentence is usually stressed eg the very last phrase in English or the

penultimate phrase in Dutch cf Reinhart I do not deal with the default intonation here

CHAPTER SPECIFICITY FOCUS AND INFORMATION STRUCTURE

information should get a high pitch accent However suchconvergence do es not

always o ccur For example in Czech new information do es not and actually should

not carry stress in certain p ositions Mirjam Fried p c Considering the universal

nature of constraints in Optimality Theory I therefore keep these two constraints

separate although it do es not makemuch dierence either way in German

Sp ecicity and Information Structure

Our discussion has so far b een concentrated on the fo cusrelated eects of scram

bling It has b een argued that the discourse notions such as newness New and

prominence Prom play crucial roles in scrambling Discourseold New and

discourseprominent Prom elements scramble In this sense scrambling is viewed

as a discoursesyntax mapping phenomenon in which discourse information suchas

the information status of eachelement of a sentence is realized syntactically in terms

of dierentword orders ie scrambling

In this section we will turn our attention to the semantic eect of scrambling

ie the sp ecicity eect see section In short I will argue that the sp ecicity

eect can also b e subsumed under the general scheme of understanding scrambling as

a discoursesyntax mapping phenomenon This claim is in contrast with the direct

semanticssyntax approach of de Ho op and Diesing

Sp ecicity Eect Revisited

Earlier in this chapter it was p ointed out that the scramblability of a phrase is also

aected by a semantic prop erty namely its sp ecicity As describ ed in section

while a sp ecic phrase can easily scramble a nonsp ecic phrase with its existential

interpretation cannot scramble This semantic eect is illustrated in the grammatical

distinction b etween a and b in which the former involves a scrambled denite

CHAPTER SPECIFICITY FOCUS AND INFORMATION STRUCTURE

NP and the latter involves a scrambled indenite NP

a Ichhabe den Brief meinem Bruder geschickt

I have the letterAcc my brotherDat sent

I sent the letter to my brother

b Ichhabe einen Brief meinem Bruder geschickt

I have a letterAcc my brotherDat sent

I sent a letter to my brother

Diesing and de Ho op nd answers to this problem in the direct syntax

semantics mapping see section for details of their analyses The basic idea is

that a certain syntactic prop erty asso ciated with the scrambled p osition whether it

is strong Case as in de Ho op or the syntactic scop e VP of the existential

op erator as in Diesing is resp onsible for the licensing of the semantic prop erty

of sp ecicity Simply nonsp ecicity cannot b e licensed in the scrambled p osition

whichiswhy a nonsp ecic NP cannot scramble as shown in b

In contrast I argue that a nonsp ecic NPs general inability to scramble is not

due to its incompatibility with certain syntactic prop erties asso ciated with partic

ular phrase structural p ositions but to its intrinsic discourse prop erty which pre

vents it from b eing interpreted as old or New information In other words I

argue that the solution to the sp ecicity problem in scrambling lies in the semantics

discoursepragmatics interface rather than in the semanticssyntax interface This

approach enables us to give a uniform account of the scrambling b oth of sp ecic and

of nonsp ecic phrases as an information structurally motivated phenomenon

As p ointed out earlier in this chapter the distinction is not b etween denite and indenite

phrases but b etween the socalled strongpresupp ositional and weakexistential phrases so

that an indenite NP with the former typ e of interpretation can also scramble I will continue to call the strongpresupp ositional phrases sp ecic regardless of their deniteness

CHAPTER SPECIFICITY FOCUS AND INFORMATION STRUCTURE

This approach is supp orted bytwo pieces of evidence already mentioned earlier in

this chapter Firstly denite phrases are immune to the semanticssyntax mapping

hyp othesis

a Ich hab e meinem Bruder den Brief geschickt

I havemy brotherDat the letterAcc sent

I sentmy brother the letter

b Ichhabe den Brief meinem Bruder geschickt

I have the letterAcc my brotherDat sent

I sent the letter to my brother

Although Diesing claims that denite phrases must scramble out of the VP

that is not the case as shown in a As Diesing herself partially p oints out a

denite phrase can also stay in the canonical p osition dep ending on its information

status Diesing treats the insitu denite phrases as an exceptional case whichis

limited to the sp ecial fo cus interpretation case However as Lenerz shows

these denites can b e interpreted as nonfo cal entities in the base p osition Lo ok at

the example in

a Wem hast du den Brief geschickt

whomDat haveyou the letterAcc sent

To whom did you send the letter

b Ich hab e meinem Bruder den Brief geschickt

I havemy brotherDat the letterAcc sent

I sentmy brother the letter

The denite phrase den Brief the letter in b is a New information b ecause

it was already mentioned in the preceding sentence a and the fo cus ie the

CHAPTER SPECIFICITY FOCUS AND INFORMATION STRUCTURE

New information is meinem Bruder my brother Still den Brief can b e in the

base p osition as shown in Thus a denite phrase in its canonical p osition cannot

b e explained even with the provision of the sp ecial fo cus case The optionality

demonstrated in is explained by the interaction b etween new and canon in the

current analysis as I will showinchapter

Secondlyeven a nonsp ecic phrase can scramble under particular discourse cir

cumstances ie in a contrastive fo cus context Recall the contrastive fo cus eect

discussed in section

weil Hans ein BUCH dem Mann gegeb en hat nicht eine ZEITUNG

b ecause Hans a b o ok to the man given has not a newspap er

b ecause Hans gave a b o ok to the man not a newspap er

The example cannot b e accounted for by the semanticssyntax mapping analyses

which are based on the idea that the semantics of an indenite NP is determined

by its p osition According to such analyses the scrambled indenite NP ein Buch

should receive a strong or presupp ositional reading However with contrastivefocus

ein Buch in b ears an existential interpretation

To sum up the relationship b etween the sp ecicity of a phrase and its scramblabil

ity it is neither the case that every sp ecic phrase scrambles b ecause denite phrases

need not scramble nor the case that every scrambled phrase is sp ecic b ecause con

trastively fo cused phrases need not b e sp ecic In short as shown in the ab ovetwo

cases the sp ecicity eect is often overridden by discourse eects So no onetoone

mapping obtains b etween sp ecicity and scramblability although it is still a fairly

strong tendency that a scrambled phrase is sp ecic

In next section I will examine what the intrinsic discoursepragmatic prop erties of

nonsp ecic phrases are and more generallyhow the sp ecicityofanentityisrelated

to its information status in the discourse I argue that this connection b etween

CHAPTER SPECIFICITY FOCUS AND INFORMATION STRUCTURE

sp ecicity and information status is indirectly resp onsible for the limited scrambling

options of nonsp ecic NPs

Sp ecicity and Information Status

As discussed in information feature marking is not free Information structure can

b e hierarchically structured so that the internal elements may not always b e assigned

features Or certain elements may not b e compatible with a sp ecic feature for mor

phological or semantic reasons and thus end up with no feature sp ecication I argue

in this section that the sp ecicity eect is a result of a mismatchbetween semantic

sp ecicity and discourse newness which leaves a nonsp ecic NP undersp ecied with

the discourse features This makes a nonsp ecic NP informationally dep endentonthe

larger information unit which accounts for the observation of the socalled predicate

mo dier interpretation de Ho op

Semantic deniteness referentiality or sp ecicity is often discussed in relation

to discourse notions such as familiarity or oldness It is often argued that a sp e

cic or referentiallyanchored NP is a discourseold familiar or discourselinked en

tity whereas a nonsp ecic or referentially nonanchored NP is a discoursenew entity

Heim Enc Pesetsky Heim for example argues that b oth

indenite and denite NPs intro duce variables but that they dier in that inde

nite NPs are new entities while denite NPs are familiar entities SimilarlyEnc

argues that sp ecicitycaneventually b e reduced to discoursefamiliarityIn

languages whichhave morphological means of realizing this distinction suchasEn

glish and German the former is usually realized as a denite NP while the latter

as an indenite NP In fact an indenite NP can also denote a sp ecic strong or

presupp ositional entityasarguedby de Ho op and Diesing We will

simply treat this kind of indenite NPs just like denite NPs and group them together here as sp ecic

CHAPTER SPECIFICITY FOCUS AND INFORMATION STRUCTURE

Given that the feature New in our information structure is also based on the

discourseoldnessnewness we can make parallel assumptions ab out sp ecicity that

sp ecic phrases are New while nonsp ecic phrases are New Let me initially

prop ose the hyp othesis in

a A sp ecic phrase is New

b A nonsp ecic phrase is New

This app ears to b e a quite plausible assumption since a nonsp ecic indenite NP

usually intro duces a new entityinto the discourse as such it b ecomes new information

or fo cus in the discourse On the other hand a sp ecic denite phrase usually refers

to an entity which is already existent in the discourse therefore this old entity often

functions as given or ground information The example in is a clear case which

illustrates this p oint

a Was hat Hans dem Schuler gegeb en

whatAcc has Hans the studentDat given

What did Hans give to the student

b da Hans dem Schuler ein BUCH gegeb en hat

that Hans the studentDat a b o okAcc given has

that Hans gave the student a book

The sp ecic phrase dem Schuler is a familiar entity in the discourse and thus used

as ground New information here On the other hand the nonsp ecic phrase ein

Buch is an unfamiliar unknown entityandthus functions as a fo cus New

However this is not always the case As wewillseebelow there are some contexts

which disturb this nice semanticsdiscourse linking Sometimes a sp ecic phrase is

forcedtobeinterpreted as if it were a new entity or sometimes a nonsp ecic phrase

CHAPTER SPECIFICITY FOCUS AND INFORMATION STRUCTURE

is forced to b e interpreted as if it were old information Let us consider these cases

in the following small sections

Sp ecic New Entity

One context in which a sp ecic phrase is forced to b e interpreted as a new entity

is when it is intro duced in the discourse as an answer to a whquestion In terms of

our information structure it is the case where the sp ecic phrase is interpreted as a

fo cus either completiveorcontrastive Consider the example in

a Was hat Hans dem Schuler gegeb en

whatAcc has Hans the studentDat given

What did Hans give to the student

b da Hans dem Schuler das BUCH gegeb en hat

that Hans the studentDat the b o okAcc given has

that Hans gave the student the b o ok

Diesing notes a similar situation and describ es it as a context where a

denite NP is sub ject to a fo cused or contrastiveinterpretation which presumably

overrides the familiar status of the variable intro duced by the NP by signaling that

novel information is b eing or ab out to b e presented The information status of the

sp ecic phrase is ambivalent it still enco des a familiar entity otherwise it would

b e presented as an indenite phrase but the currentcontext forces it to b ehaveas

if it were not known in the discourse ie as an unfamiliar entity This new denite

phrase such as the one in may b e analogous to a reactivated dormant lecard

in Vallduvs lecard analogy Being a denite phrase it has a preexistent

le card already created for it However this lecard maynothave b een used for a

while dormant and thus in the current discourse it can b e freshly referred to ie

CHAPTER SPECIFICITY FOCUS AND INFORMATION STRUCTURE

reactivated as if it were a new entity

So the inherently familiar prop erty of a sp ecic phrase can b e overridden in a

particular context such as the one in and then it is interpreted as a temp orary

New entity Casting this problem of overriding from the viewp oint of Optimality

Theorywe think of this case as one where a certain constraint is violated Sup

p ose that the semanticsdiscourse linking hyp otheses prop osed ab ove are OTstyle

constraints which are universal and also violable

specificity

a sp A sp ecic phrase should not b e New

b sp A nonsp ecic phrase should not b e New

These constraints are the ones which control the feature marking in the information

structure They determine which elementmayormay not b e marked with which

feature Then we can conceive of the new sp ecic entity case as one in whichthe

rst part of the specificity constraint sp is violated Thus a sp ecic phrase

which is supp osed to b e New is forced to b e interpreted as a New entityin

this context

Nonsp ecic Old Entity

Now let us consider the reverse case Can a nonsp ecic phrase b e interpreted as

an old entityaswellSimply put the answ er is no A nonsp ecic phrase cannot

be New In OT terms this would mean that the second part of the specificity

constraint sp is so strong or highly ranked that it would never b e violated

This problem may really involvetwo dierenttyp es of discourseoldness for example discourse

oldness and heareroldness Prince So the new sp ecic phrase would b e classied as a

discoursenew and hearerold entitySeeWard and Birner for the application of this notion

to the there construction in English

CHAPTER SPECIFICITY FOCUS AND INFORMATION STRUCTURE

A nonsp ecic phrase refers to an entity which is not referentially anchored in

discourse namelyanentitywhich is unfamiliar or unknown in discourse Usuallyit

has the function of intro ducing a new entityinto the discourse rather than referring

to one which is already in the discourse This is the intuition that Web elhuth

tries to capture when he states that a nonsp ecic indenite NP is inherently fo cused

In Vallduvs lecard analogy this means that there is no lecard which has

b een created for it For example consider

a I b ought abook

b Where did you buy itthe b o ok

As usual a nonsp ecic indenite NP abook in a intro duces a new entityinthe

discourse whose reference is unknown or undetermined In a normal context it is

then immediately referentially anchored and therefore is replaced by a denite NP or

a pronoun as shown in b Now the denite NP the book refers to a sp ecic b o ok

which the p erson in a b ought and is thus treated as known

On the other hand in a context like abook still remains unidentiable even

after the initial intro duction

a I b ought abook

b Where did you buy abook

As in a nonsp ecic indenite NP abook in a intro duces a new entityof

a b o ok kind in the discourse However in this case it is still not referentially

anchored in the second sentence b It seems that the sp eaker in b p ostp ones

the discourseanchoring pro cess of the newlyintro duced entity a b o ok and do es not

treat it as an indep endententity but instead handles the eventofbuyingabookas

an unanalyzable whole Namely the sp eaker is only interested in the buying event

not in the individual b o ok Compare b with b So the question in b is

more likely to b e interpreted as Where did you do the b o okbuying activity

CHAPTER SPECIFICITY FOCUS AND INFORMATION STRUCTURE

Now recall the discussion of the feature sp ecication in the information structure

in section It was noted there that not all elements of the sentence are sp ecied

with discourse features New or Prom Some are undersp ecied One case in which

this undersp ecication o ccurs is when an element is contained in a larger informational

unit for example a VP In this case only the larger unit is sp ecied with the discourse

features but not each comp onentelement Not b eing indep endently sp ecied the

contained elements are informationally dep endent on the larger unit

This is the situation which describ es the example in b The VP buy a book

is presented as a single informational unit in b actually as New since the

b o okbuying information is already intro duced in the previous sentence in a

However the indenite NP abook whichiscontained in the VP do es not hold a

feature marking itself Thus it is informationally dep endent on the containing VP

structure I argue that this informational dep endency of a nonsp ecic NP makes it

reinterpreted almost like a predicate mo dier rather than an indep endententity

This diers from de Ho ops prop osal that the predicate mo dier reading is

derived from the weak Case assignment ie from a syntactic prop erty I argue

that the informationbased approach has an advantage b ecause a nonsp ecic NP is

not interpreted as a predicate mo dier if it achieves indep endent information status

When an indenite NP intro duces a new entity in the discourse as in the rst sentence

in a it is marked New itself and interpreted as a regular argument That is

the semantic interpretation relies on the information status

Now let us return to the sp ecicity problem and see how the indenite NP is

interpreted there

a Wem hat Hans ein Buch gegeb en

whoDat has Hans a b o okAcc given

To whom did Hans give a book

CHAPTER SPECIFICITY FOCUS AND INFORMATION STRUCTURE

b da Hans dem SCHULER ein Buch gegeb en hat

that Hans the studentDat a b o okAcc given has

that Hans gave the student a book

b da Hans ein Buch dem SCHULER gegeb en hat

that Hans a b o ok the studentDat given has

that Hans gave the student a book

In b the nonsp ecic NP ein Buch cannot b e interpreted as New b ecause

its initial intro duction has already happ ened and the fo cus has shifted to another

element dem Schuler It cannot b e interpreted as New by itself either b ecause it

is not referentially anchored sp Not b eing able to b e informationally marked it

b ecomes part of the larger information unit ie ein Buch gegeb en just as in

New

b It is hence interpreted as a dep endent part of the predicate Since ein Buch

is not informationally indep endentandthus not holding any discourse feature itself

scrambling is not even an issue Recall that the information structuring constraints

which motivate scrambling do not apply to an elementwhich is not discoursemarked

To summarize a nonsp ecic NP cannot scramble as an old element b ecause when

it is not referentially anchored it b ecomes informationally dep endent As mentioned

at the b eginning of this section this phenomenon can b e understo o d in OT as a

constraint which is hardly violable Namely the second part of the specificity

constraint which controls the relationship b etween semantic sp ecicity and discourse

oldness in the information structure is so strong that it cannot b e violated

sp A nonsp ecic phrase should not b e New

Note that this do es not prevent a nonsp ecic NP from scrambling as a new element

as long as the information structuring constraints p ermit this If it is a contrastive

fo cus the prom constraintwould allow its scrambling although the new constraint

CHAPTER SPECIFICITY FOCUS AND INFORMATION STRUCTURE

would not Exactly how these constraints will interact will b e examined in the next

chapter

Therefore with the provision of the specificity constraints as a restriction on

the information feature marking I argue that scrambling can b e uniformly explained

as a discoursesyntax mapping phenomenon without recourse to the direct syntax

semantics mapping mechanisms of Diesing or de Ho op The sp ecicity

eect results from a semanticsdiscourse mismatchinwhich a nonsp ecic NP is forced

to b e interpreted as an old entitywhich is restrained by a semanticsdiscourse

mapping constraint sp

Summary

In this chapter I have examined alternative ordering phenomena or scrambling and

attempted to nd motivations for the variation Having investigated various semantic

and discourserelated eects asso ciated with scrambling ie the sp ecicityantifo cus

and contrastive fo cus eects I have prop osed a view of information structure based

on the crosscutting discourse features New and Prom as a generalization to ol to

capture the meaningrelated eects of scrambling I argued that the alternative struc

tural descriptions or the scrambled variants are motivated by a need to instantiate the

information structure in terms of phrase structure I have prop osed two information

structuring constraints new and prom as ma jor driving forces for the alternations

In other words the information structuring constraints in grammar try to realign

the elements of the sentence according to their information status Therefore they are

in tension with the phrase structural constraints intro duced in chapter which try

to align the constituents according to their grammatical information see the relevant

discussion in chapter Basically the information structuring constraints act against

the phrase structural constraints which force all elements of the sentencetobein

CHAPTER SPECIFICITY FOCUS AND INFORMATION STRUCTURE

their canonical p ositions and motivate and p ermit these elements to b e out of

their canonical p ositions if they meet these informational requirements Therefore

these constraints comp ete with each other and yield more complicated results which

are the various scrambled structures In Optimality Theory the interaction or the

conict resolution among dierent constraints is carried out by the ranking among

them Now in chapters and I will prop ose particular rankings of these constraint

namely the grammars of German and Korean As will b e shown the two languages

are almost identical in their word order p ossibilities and only dier in one asp ect

the sub ject eect This dierence will b e handled with a slightchange in the rankings

of the constraints in each language

Chapter

Optimality Theory and Scrambling

in German

In chapter I have argued that the socalled canonical word order in scrambling

languages is determined by a set of phrase structural constraints which control the

mapping from one syntactic structure which contains information such as predicate

argument relations or grammatical functions to another syntactic structure which

represents information such as categorization and constituency These constraints

which I call canon therefore favor one particular phrase structural description over

others unless some other factors in the language such as discoursepragmatic infor

mation give preference to other structural descriptions This particular structure or

the consequent order is the socalled canonical or unmarked order

In chapter I have examined what asp ects of language inuence and motivate

the choice of noncanonical orders Based on the fact that alternative structures

are not equivalent in their interpretation I haveinvestigated semantic and discourse

factors asso ciated with scrambling and prop osed that they b e b etter accounted for in

terms of information structure cf Vallduv I have presented a mo del of

information structure which is based on two crosscutting discourse features New

CHAPTER OPTIMALITY THEORY AND SCRAMBLING IN GERMAN

and Prom and identied two information structuring constraints new and prom

as ma jor driving forces for scrambling These constraints control the mapping from

the information structure to the phrase structure

In this chapter utilizing the core ideas in Optimality Theory Prince and Smolen

sky to app ear Grimshaw Grimshaw in press I will showhowvarying scrambled

structures are derived bytheinteractions among the constraints develop ed in the pre

vious chapters Here each alternative structure of a sentence is viewed as the optimal

output which enco des the syntactic eg grammaticalfunctional structure and also

discoursepragmatic eg information structure information providedintheinputin

the b est p ossible wayby means of phrase structure andor proso dic structure It is

shown that while searching for the optimal output in eachcontext the semantic and

discourse eects ie the sp ecicityantifo cus and contrastivefo cus eects often

asso ciated with scrambling naturally follow from the constraint comp etition from the

dierent mo dules of grammar

This chapter is organized in two parts In section the basic OT layout for

the scrambling phenomena is illustrated Section then demonstrates the OT

derivations for various ordering p ossibilities in dierentcontexts and thus shows how

the semantic and pragmatic restrictions on scrambling are captured by Optimality

Theoretic interactions

Optimality Theory and Scrambling

In this section I lay out the basic OptimalityTheoretic set up for scrambling In

particular I discuss the input and output representations and also the ranking of

the constraints involved in scrambling whichhavebeendevelop ed in chapter and

chapter

Optimality Theory is designed in sucha way that for a given input there is one

CHAPTER OPTIMALITY THEORY AND SCRAMBLING IN GERMAN

optimal output which violates fewer highlyranked constraints than the other can

didate outputs and thus wins over them In this theory only this output is considered

grammatical with resp ect to the given input For the free word order phenom

ena under discussion now the comp etition can b e thought of as arising among the

alternative sentential structures diering in phrase structure eg string order ie

the scrambled variants of a sentence Here each alternative structure of a sentence

is viewed as the optimal output in a given context which reconciles the given infor

mation structure and other syntactic information in the b est way

Inputs and Information Structure

In Optimality Theory a grammar is a function which maps inputs to outputs Inputs

for syntax are the basic building blo cks such as lexical items and the information

ab out how they are related to each other from whichsyntactic structures are built

For example Grimshaws input for an extended pro jection contains a

lexical head and its argument structure an assignment of lexical heads to its argu

ments and also tense and asp ect sp ecications Similarly Legendre et al and

Legendre et al b also use some form of predicateargument structure for their

inputs b is an example input for a sentence like a

a Mary saw John

b seex y x Maryy J ohn P ast

It is not clear however whether the nonargument elements of a sentence suchas

adverbs would b e included in this typeofinputbutitisobvious that the input for

the scrambling phenomena needs to include adjuncts as well as arguments b ecause

adverbs also play a crucial role in constituent ordering Therefore I suggest following

Bresnan that an input is a set of lexical heads including adjuncts and an

undersp ecied or skeletal fstructure indicating how the heads are to b e related to

CHAPTER OPTIMALITY THEORY AND SCRAMBLING IN GERMAN

each other b is an example of the undersp ecied fstructure for a sentence like

a Mary saw John yesterday

b pred seexy

Mary

John

yesterday

tense Past

The only ma jor dierence b etween and is that the latter can include non

argument phrases Otherwise the dierences are primarily notational

In addition I argue that the input contains information ab out each elements

information status As argued in chapter the information structure of a sentence

is absolutely crucial in determining the constituent structure eg linear order in

scrambling languages Accordingly I prop ose that each elementintheskeletal f

structure is marked with the discourse features New and Prom according to its

information status as I have prop osed in the previous chapter The featurebased

information structure is rep eated here as

Prom Prom

New Topic Tail

New ContrastiveFocus CompletiveFo cus

As a matter of fact it has b een prop osed in several places that the input materials

should contain certain semantic or pragmatic information which includes deniteness

sp ecication Grimshaw in press abstract prominence sp ecication Legendre et

CHAPTER OPTIMALITY THEORY AND SCRAMBLING IN GERMAN

al and scop e information Legendre et al b Interestingly Grimshaw

and SamekLo dovici suggest that inputs include information ab out which

phrases are fo ci and which are coreferent with the topic The following in are

some examples

a sing xx topic x J ohn

b g iv ex y z x topic z focus x Johny pr esent z Mary

Grimshaw and SamekLo dovicis prop osal is very similar to the one I pro

p ose here in the sense that the input contains such discoursefunctional information as

topic and fo cus The dierence lies in that in the current system as illustrated in

the information structure is represented with the crossclassifying features ie New

and Prom which enables us not only to recognize the existence of such information

typ es as topic and fo cus but also to capture the similarities and dierences among

dierent information typ es For example topic and fo cus usually b ehave dierently

since they have distinct values for the New feature That is topic is New and

fo cus completiveorcontrastive is New However topic and contrastive fo cus

sometimes b ehavealike since they share a feature ie Prom although they still

dier in the value of the feature New

Now lets return to the scrambling data and see what their input representation

would lo ok like Consider the German sentence in which has a ditransitive clause

emb edded in it As in chapter I will concentrate on the emb edded clause which

has more freedom in terms of word order

From now on I will only fo cus on the nonverbal elements of a sentence ie sub ject direct ob ject

and indirect ob ject etc b ecause verbal elements do not scramble although they may topicalize

This may b e b ecause verbal elements cannot adjoin

CHAPTER OPTIMALITY THEORY AND SCRAMBLING IN GERMAN

Ich glaub e da Hans dem Schuler das Buch gegeb en hat

I b elieve that Hans the StudentDat the b o okAcc given has

I b elieve that Hans gave the student the book

The emb edded clause in has three nonverbal elements in it German word order

is indeed free so that we can have all six p ossible p ermutations of these three items

Uszkoreit This is illustrated in

a da Hans dem Schuler das Buch gegeb en hat

b da Hans das Buch dem Schuler gegeb en hat

c da dem Schuler Hans das Buch gegeb en hat

d da das Buch Hans dem Schuler gegeb en hat

e da dem Schuler das Buch Hans gegeb en hat

f da das Buch dem Schuler Hans gegeb en hat

Although they dier in order the clauses in have the same predicateargument

structure or basic fstructure This basic argumentstructural information can b e

represented in the input as in

As exp ected some orders are more marked than others Also not all sp eakers accept all the

orders demonstrated here some never allowany orders in which sub ject is not in the initial p osition

Generally sp eaking the orders towards the b ottom in are worse or more marked This markedness

issue will b e briey discussed from the OT p ersp ective in the concluding chapter

The input represented by the undersp ecied fstructure also includes any information brought

by the lexical entries of arguments such as case and agreement features in addition to the predicate

argument information Tokeep the representations simple I do not list these features in the inputs

CHAPTER OPTIMALITY THEORY AND SCRAMBLING IN GERMAN

pred geb enxyz

Hans

dem Schuler

das Buch

tense Past

However they are not necessarily equivalent in their discoursecontextual meaning In

the present framework this means that they mayhave dierent input representations

in terms of information structure In other words the elements in the input mayhave

dierent feature markings with resp ect to New and Prom

Supp ose for example that a sentence has a fo cus but do es not haveanyspe

cic topic This means that there is some New elementinthissentence but

there is no Prom element In other words all the elements have equal status

in terms of prominence and thus no one element is more prominent than another

This information structure is the backgroundfo cus Dahl Chafe or the

presupp ositionop enprop ositionfo cus typ e Jackendo Prince Ward

or the tailfo cus typ e in Vallduv This typ e of information

structure usually o ccurs as an answer to a whquestion as shown in

I assume that every sentence has a fo cus ie new information following Vallduv Oth

erwise a sentence would consist only of old information whichwould simply b e a rep etition of the

already known information

One could argue that fo cus is also prominent b ecause the new information is also distinct

from the nonnew information in the rest of the sentence However this is not the sense of prominence

I am using here b eing prominent by b eing the informative part of a sentence is termed New

here Only contrastive fo cus is prominent or Prom in this system See chapter for a review of

the prop erties of these features

Note that the upp er case representation in the examples do not necessarily mark sp ecial em

phasis To b e consistent with the data provided by Lenerz Abraham and Web elhuth

I represented all fo cused elements alike whether completiveorcontrastive in upp er case

CHAPTER OPTIMALITY THEORY AND SCRAMBLING IN GERMAN

a Was hat Hans dem Schuler gegeb en

whatAcc has Hans the studentDat given

What did Hans give to the student

b Hans hat dem Schuler das BUCH gegeb en

Hans has the studentDat the b o okAcc given

Hans gave the student the b o ok

In this case the input will b e marked as in which do es not dier in the feature

Prom everything is marked as Prom by default but only diers in the feature

New For ease of presentation I will hereafter simply mark the discourse features

under each lexical item

I do not makeany claims here ab out how exactly the information structure istructure should

be internally structured other than that it contains Prom and New as primitives This topic

certainly needs further research see also the discussion in section in chapter Also an

input could b e represented as the corresp onding pair of the undersp ecied fstructure and the

istructure linked together by a mapping function similar to what exists b etween the cstructure

and the fstructure However I do not pursue this issue here

CHAPTER OPTIMALITY THEORY AND SCRAMBLING IN GERMAN

pred geb enxyz

Hans

New

Prom

dem Schuler

New

Prom

das Buch

New

Prom

tense Past

Here the sentence b is simply partitioned into the oldgiven part and the informa

tivenew part Halliday as b eing an answer to a question like a The direct

ob ject das Buch in b is the new information b ecause it is the part corresp ond

ing to the whphrase in a and the rest of the sentence is the given information

Therefore the former is marked New and the latter New in this system On the

other hand since no one item is presented as a standingout or prominent element

eg with stress or with a certain phrase like what about in the previous discourse

nothing is marked Prom Therefore is the input that is given expression in

the utterance of b

Now supp ose that a certain element the sub ject for instance is presented promi

nently This can b e achieved by putting a slightly higher pitch accentonitorby

putting it in a phrase like what about as in

CHAPTER OPTIMALITY THEORY AND SCRAMBLING IN GERMAN

a Wie stehts mit HansW as hat Hans dem Schuler gegeb en

how stand it with Hans what has Hans the studentDat given

What ab out HansWhat did Hans ga ve to the student

b Hans hat dem Schuler das BUCH gegeb en

Hans has the studentDat the b o okAcc given

Hans gave the student the b o ok

pred geb enxyz

Hans

New

Prom

dem Schuler

New

Prom

das Buch

New

Prom

tense Past

Signaled by the preceding sentence Wie stehts mit Hans What ab out Hans in

a the sub ject Hans is understo o d now as prominent It is more prominent than

other elements of the sentence in b Hence it is marked Prom and the other

elements are marked Prom Also just as in the whquestion signals that the

fo cus of the following sentence is das Buch the b o ok Accordingly das Buch gets the

New marking This results in the information structure of the topicfo custail

CHAPTER OPTIMALITY THEORY AND SCRAMBLING IN GERMAN

typ e in Vallduv Hans is the topic ie prominent old information das Buch

is the completive fo cus ie nonprominent new information and the rest is the

tail ie nonprominent old information

Outputs and Candidate Set

Outputs are all the p ossible structural descriptions or parses of a given input which

are generated by GEN GEN instantiates an input by assigning it syntactic and other

eg proso dic structures In a parallel structure framework likeLFG GEN will

pro duce pairs of corresp onding cstructures and fstructures fully sp ecied with sub

stitution of sp ecic functions for the undersp ecied GFs Bresnan Since

GEN pro duces all universally p ossible structures according to X theory for in

stance it not only generates acceptable parses in a particular language but also

generates unacceptable ones It is the inputs job to collect the relevant outputs

from among those in the universal set of structures and to put them in the candidate

set So the input determines what structures comp ete in a candidate set

Grimshaw in press suggests that the outputs in a candidate set have non

distinct logical forms although this has b een denied by some for scop e considerations

Legendre et al b Under the current assumptions the candidate outputs can

b e conceived of as those which share the same undersp ecied fstructure or argument

structure such as the one given ab ove rep eated here as in

The pro cess of this full sp ecication of functions is carried out by dierentmechanisms dep ending

on the language it may b e determined by phrase structural congurations thematic roles or

morphological case marking Bresnan and Zaenen Legendre et al Since this is not the

ma jor concern in the present discussion I will assume that the function sp ecication is executed by

relevant mapping functions

CHAPTER OPTIMALITY THEORY AND SCRAMBLING IN GERMAN

pred geb enxyz

Hans

dem Schuler

das Buch

tense Past

In this sense we can assume that all the scrambled variants including the default

ordered one in a b elong to the same candidate set

a da Hans dem Schuler das Buch gegeb en hat

b da Hans das Buch dem Schuler gegeb en hat

c da dem Schuler Hans das Buch gegeb en hat

d da das Buch Hans dem Schuler gegeb en hat

e da dem Schuler das Buch Hans gegeb en hat

f da das Buch dem Schuler Hans gegeb en hat

The actual syntactic output representations will b e pairs of corresp onding c

structures and fstructures for a given undersp ecied fstructural input following

Bresnan I list only a few examples here for the input given in

CHAPTER OPTIMALITY THEORY AND SCRAMBLING IN GERMAN

pred geb enxyz

Hans

New

Prom

dem Schuler

New

Prom

das Buch

New

Prom

tense Past

In Optimality Theory GEN in principle generates all p ossible structures univer

sally available Legendre et al Therefore it can pro duce for instance com

pletely at structures for any sentence in Lets take b as an example which

is instantiated as a at structure in In the following pairs of cstructures and

fstructures the mapping b etween them is represented by the corresp onding numbers

in each representation

a S

NP NP NP V V

Hans das Buch dem Schuler gegeb en hat

In the fstructures b elow I use IOBJ and DOBJ short for Indirect Ob ject and Direct Ob ject

resp ectively to avoid any confusion instead of OBJ and OBJ which are more standard terms See

Zaenen et al for discussion of OBJ and OBJ

CHAPTER OPTIMALITY THEORY AND SCRAMBLING IN GERMAN

b pred geb enxyz

Hans

subj

New

Prom

dem Schuler

iobj

New

Prom

das Buch

dobj

New

Prom

tense Past

As assumed in chapter however this parse will b e ruled out by the constraint

interaction b etween economy of expression and endocentricity alignment

Since German ranks endocentricity alignment higher than economy of ex

pressionmorehierarchical structures which pro ject VPs such as the one in or

are favored over at structures Bresnan see chapter for arguments in

favor of congurational structures for German and Korean yields the order in a the canonical order and generates the order in d

CHAPTER OPTIMALITY THEORY AND SCRAMBLING IN GERMAN

a S

NP VP

Hans VP V

V hat

NP V

dem Schuler NP V

das Buch gegeb en

pred geb enxyz b

Hans

subj

New

Prom

dem Schuler

iobj

New

Prom

das Buch

dobj

New

Prom

tense Past

CHAPTER OPTIMALITY THEORY AND SCRAMBLING IN GERMAN

a S

NP S

das Buch NP VP

Hans VP V

V hat

NP V

dem Schuler gegeb en

pred geb enxyz b

Hans

subj

New

Prom

dem Schuler

iobj

New

Prom

das Buch

dobj

New

Prom

tense Past

Note that the example outputs presented ab ove all maintain the information given

by the input they all faithfully parse the input information GEN however can

also generate unfaithful outputs such as the one in in which some of the input

CHAPTER OPTIMALITY THEORY AND SCRAMBLING IN GERMAN

information ie the discourse features for dem Schuler and for das Buch is not parsed

Compare it with the faithful parse in

a S

NP S

das Buch NP VP

Hans VP V

V hat

NP V

dem Schuler gegeb en

b pred geb enxyz

Hans

subj

New

Prom

iobj

dem Schuler

das Buch

dobj

Prom

tense Past

These unfaithful parses are usually not even considered to b e comp eting with faith

ful parses b ecause in most cases faithful parses win This ltering of unfaithful

parses is carried out by the faithfulness constraint which is usually ranked so high

CHAPTER OPTIMALITY THEORY AND SCRAMBLING IN GERMAN

in the grammar that it can rarely b e outranked by other constraints in the grammar

The faithfulness constraintisgiven in Bresnan

faithfulness Constraint

All and only the attributes of input lexical heads app ear in the output

In addition to the syntactic instantiations GEN also provides proso dic instantia

tions by assigning varying intonational patterns to each output I will call these pairs

of syntactic and proso dic structures the surface structures Concentrating only on

the nonverbal elements of the sentence lets supp ose that GEN mayormay not put

a high pitch accent on eachelement Then since a ditransitivesentence has three

nonverbal arguments each distinctlyordered variant will have eight proso dically dif

ferent p ossible outputs b ecause eachofthethreeelements mayormay not get a high

pitch accent ie Consider the rst sentence of the six scrambled variants in

as an example The eight dierent proso dic p ossibilites for this syntactic variant

are shown in An element with a high pitch accent is represented in upp er case

in the examples

Legendre et al b argue that although unfaithful outputs are usually disfavored to faithful

parses since they violate the fairly highranked faithfulness constraints they could win over faithful

parses and b ecome optimal when the faithfulness violations they incur are outranked by other vio

lations of higherranked constraints by faithful parses See also Grimshaw for an application

of the idea of faithfulness violations

Although I overlap the proso dic structure on the phrase structure as the surface structure here

for ease of exp osition manyhave argued that it is necessary to assume a separate level of proso dic

structural representation Selkirk Nesp or and Vogel Inkelas Then the surface

structure will b e represented also as a pairing of the cstructure and the proso dic structure linked

by a mapping function b etween them

CHAPTER OPTIMALITY THEORY AND SCRAMBLING IN GERMAN

a da Hans dem Schuler das Buch gegeb en hat

b da Hans das Buch dem Schuler gegeb en hat

c da dem Schuler Hans das Buch gegeb en hat

d da das Buch Hans dem Schuler gegeb en hat

e da dem Schuler das Buch Hans gegeb en hat

f da das Buch dem Schuler Hans gegeb en hat

a da Hans dem Schuler das Buch gegeb en hat

b da HANS dem Schuler das Buch gegeb en hat

c da Hans dem SCHULER das Buch gegeb en hat

d da Hans dem Schuler das BUCH gegeb en hat

e da HANS dem SCHULER das Buch gegeb en hat

f da HANS dem Schuler das BUCH gegeb en hat

g da Hans dem SCHULER das BUCH gegeb en hat

h da HANS dem SCHULER das BUCH gegeb en hat

Since wehave six distinctlyordered variants for one ditransitivesentence eachof

which has eight dierent proso dic patterns we end up with x p ossible candi

dates for each ditransitive sentence

Now the outputs b elong to the same candidate set and comp ete with one an

other to b e selected as the optimal output These outputs are the structural instanti

ations of the input whichcontains the discoursecontextual information represented

in the discourse features in the information structure in addition to the syntactic in

formation Note that these parses are all syntactically go o d acceptable outputs

wehave assumed that the ltering of unacceptable syntactic structures is already

taken care of The outputs now comp ete to b e the b est t for the given context

In other words in scrambling languages the b est output will b e the one which re

alizes the information structure of the input in the b est way p ossible with a sp ecic

CHAPTER OPTIMALITY THEORY AND SCRAMBLING IN GERMAN

structural description

Constraints and Ranking

Outputs are evaluated by a set of ranked universal constraints In Optimality Theory

constraints are violable So an output which violates some constraints can b e an

optimal one unless it violates more highlyranked constraints than other candidates

A grammar of a particular language is considered to b e a particular ranking of these

universal constraints In this section I will examine the constraints which are involved

in scrambling and the ranking among them for German

As argued in chapter and chapter scrambling is constrained not only by syn

tactic constraints but also by discoursepragmatic constraints Just as there are con

straints whichcontrol the mapping relations b etween the undersp ecied fstructure

and the phrase structure there are also constraints whichcontrol those b etween the

information structure and the surface structure The former are phrase structural

constraints canon and the latter are information structuring constraints See sec

tion and for a review of these constraints These constraints are rep eated

from chapter and chapter in and b elow

canon

a cn SUBJ should b e structurally more prominent than eg ccommand

nonSUBJ functions

b cn NonSUBJ functions align reversely with the cstructure according

to the functional hierarchy

SUBJ DOBJ IOBJ OBL ADJUNCT

Information Structuring Constraints

a new ANew element should precede a New element

b prom A Prom element should precede a Prom element

CHAPTER OPTIMALITY THEORY AND SCRAMBLING IN GERMAN

Let us briey review how these constraints are interpreted The phrase structural

constraints in basically align elements according to syntactic information ie

grammatical functions of the elements see section in chapter for an alternative

which is based on the thematic hierarchy On the other hand the information struc

turing constraints align elements according to their information status ie whether

they are new or prominent information Therefore the grammar has two dierent

sources constraining the constituent order eachofwhichhasitsown interest ie

to realize certain information in terms of the phrase structure Naturally they are

potentially in conict

Optimality Theory provides a very go o d to ol for resolving suchpotential conicts

In fact the theory itself is a pro cess of conict resolution in a sense The OT con

straints are universal and very general so as to b e available to all languages so there

will b e constraints which are contradictory to each other Fortunately constraints

are also violable which means that they are ready to negotiate with each other The

grammar resolves conicts by giving some constraints preference over others That

is constraints are ranked Therefore when conicts arise a higherranked constraint

is given preference to realize its interest rst So dep ending on which ranking a

language takes it will pro duce dierentword order p ossibilities

Letusnow see what the ranking of the constraints in and is in German

Consider canon rst As discussed in chapter it in eect generates the following

subhierarchies if there are three elements in a sentence Bakovic Whenever

a subhierarchy is violated it incurs a mark Thus a total of three marks is

p ossibly generated

a SUBJ IOBJ

b SUBJ DOBJ

c IOBJ DOBJ

CHAPTER OPTIMALITY THEORY AND SCRAMBLING IN GERMAN

The violation pattern for each order is presented in rep eated from chapter

candidates canon

a Sub ject IOb ject DOb ject

b Sub ject DOb ject IOb ject

c IOb ject Sub ject DOb ject

d DOb ject Sub ject IOb ject

e IOb ject DOb ject Sub ject

f DOb ject IOb ject Sub ject

For example b incurs a mark b ecause it violates the subhierarchycnamely

it violates the canonical order b etween the indirect ob ject and the direct ob ject Also

c incurs a mark In this case it violates the subhierarchy a ie the canonical

order b etween the sub ject and the indirect ob ject is disturb ed Interestinglyhowever

c is regarded as muchmoremarked than b although b oth incur only one mark

Consider the actual example in

a da Hans das Buch dem Schuler gegeb en hat

that Hans the b o okAcc the studentDat given has

that Hans gave the student the b o ok

b da dem Schuler Hans das Buch gegeb en hat

that the studentDat Hans the b o okAcc given has

that Hans gave the student the b o ok

Given no context b is muchworse than a We can nd a clue to this in

the fact that a is a reordering among the ob jects while b is a reordering

which includes the sub ject Generally scrambling over the sub ject is very marked in

CHAPTER OPTIMALITY THEORY AND SCRAMBLING IN GERMAN

German

Now recall that the phrase structural constraints canon actually consist of two

sub constraints as rep eated in The rst subpart cn is a constraintabout

sub ject versus nonsub jects and the second cn is one ab out nonsub jects

canon

a cn SUBJ should b e structurally more prominent than eg ccommand

nonSUBJ functions

b cn NonSUBJ functions align reversely with the cstructure according

to the functional hierarchy

The facts shown ab ove indicate that a violation of cn is more serious than that of

cn In other words cn is ranked higher than cn

Ranking

cn cn

In actual application the violation of a subhierarchywhichinvolves the sub ject eg

a or b causes a more serious problem than the violation of a subhierarchy

whichdoesnotinvolve the sub ject eg c The grammatical distinction b etween

a and b can thus b e captured by the following tableau Since b violates

the higherranked constraint it loses to a

candidates cn cn

a Sub ject DOb ject IOb ject

b IOb ject Sub ject DOb ject

The violation pattern based on the new ranking is given in

Some dialects do not allow examples like b where sub ject is scrambled over at all

CHAPTER OPTIMALITY THEORY AND SCRAMBLING IN GERMAN

candidates cn cn

a Sub ject IOb ject DOb ject

b Sub ject DOb ject IOb ject

c IOb ject Sub ject DOb ject

d DOb ject Sub ject IOb ject

e IOb ject DOb ject Sub ject

f DOb ject IOb ject Sub ject

Let us now consider the ranking b etween the information structuring constraints

There are two information structuring constraints which are rep eated here in

Information Structuring Constraints

a new ANew element should precede a New element

b prom A Prom element should precede a Prom element

I prop ose that the prom constraint b e higher than the new constraint

Ranking

prom new

This ranking explains the contrastive fo cus eect see section and in chapter

Recall that a contrastivefo cused element can scramble as shown in again

weil Hans das BUCH dem Mann gegeb en hat nicht die ZEITUNG

b ecause Hans the b o okAcc the manDat given has not the newspap er

b ecause Hans gave the b o ok to the man not the newspap er

Acontrastive fo cus is a NewProm element canon would not prompt this

scramlbing b ecause it acts against all noncanonical structures The constraint new

would not encourage this scrambling either It is in fact supp osed to discourage it

CHAPTER OPTIMALITY THEORY AND SCRAMBLING IN GERMAN

b ecause contrastive fo cus is also New Therefore this leads to the conclusion that

the constraint prom is resp onsible for it The fact that the prom constraint causes

scrambling in spite of the new constraintshows that prom is ranked higher than

new

The example in also shows that prom is also higher than cnIfcn were

higher the scrambling should not b e p ossible If b oth new and cn are lower than

promhow are these two orderedI argue that new and cn are not ranked with

resp ect to each other Or to put it in dierentwords b oth rankings are p ossible

Ranking

a new cn

b cn new

This ranking is supp orted by the following data from Lenerz Consider the

order b etween the indirect ob ject dem Schuler and the direct ob ject das Buch in

a Wem hat Hans das Buch gegeb en

whoDat has Hans the b o okAcc given

To whom did Hans give a book

Several dierentways to interpret the unranked constraints have b een prop osed Broihier

summarizes three p ossible interpretations of tied constraints which are discussed originally byPe

setsky class notes cited by Broihier See also Anttila and Smolensky

i a Branching Version of Tied Constraints

Satisfaction of any one of the tied constraints suces for satisfaction of the entire blo ck

of tied constraints

b Po oled Violation Version of Tied Constraints

The marks from the tied constraints are combined into one column and evaluation pro

ceeds otherwise as normal

c Reordering Version of Tied Constraints

The candidates are run through each p ossible reordering of the tied constraints in parallel

See also Tesar a b for discussion of learnability problems with tied constraints

CHAPTER OPTIMALITY THEORY AND SCRAMBLING IN GERMAN

b Ich glaub e da Hans dem SCHULER das Buch gegeb en hat

I b elieve that Hans the studentDat the b o okAcc given has

I b elieve that Hans gave the student a book

b Ich glaub e da Hans das Buch dem SCHULER gegeb en hat

I b elieve that Hans the b o okAcc the studentDat given has

I b elieve that Hans gave the b o ok to the student

In this context das Buch is New and dem Schuler is New Neither of them is

Prom so the constraint prom is not relevant Thus the order comp etition here

is a result of the interplaybetween new and cn As shown in b oth orders

are p ossible b says that cn is stronger than new while b says that the

opp osite is true

Finally consider the ranking b etween cn and the prom constraint The relevant

fact to test is whether a Prom element can scramble over a sub ject From the

sp eakers I consulted the answer is yes whichgives the following ranking

Ranking

prom cn

When a nonsub ject eg das Buch in is presented prominently eg as the topic

of the sentence it can scramble over the sub ject ie Hans Lo ok at the example in

a Wie stehts mit dem BuchW em hat Hans das Buch gegeb en

how stands it with the b o ok who has Hans the b o okAcc given

What ab out the b o okT o whom did Hans give the b o ok

It might b e the case that b and b are not really equivalent and there is some subtle

dierence b etween them which implies that the ranking b etween new and cn should b e xed I

leave this question op en for future research

CHAPTER OPTIMALITY THEORY AND SCRAMBLING IN GERMAN

b Ich glaub e da das Buch Hans dem SCHULER gegeb en hat

I b elieve that the b o okAcc Hans the studentDat given has

I b elieve that Hans gave the b o ok to the student

However as noted earlier some sp eakers do not accept scrambling over sub jects at

all For those the ranking will b e the reverse For the rest of the chapter I will

describ e the grammar which allows this scrambling and thus assume the ranking in

With all subrankings xed wehave the nal ranking in for German

Constraint Ranking German

new

prom cn

Information structuring constraints are intermingled with phrase structural constraints

in the ranking given in This predicts that there will b e complicated interactions

between the two comp onents of the grammar ie b etween syntax and discourse If

both cn and cn were ranked higher than the information structuring constraints

then scrambling would not happ en b ecause canon insists on the canonical order

If the ranking were the reverse ie b oth information structuring constraints were

ranked higher than cn and cn scrambling would b e much freer and the canonical

order would b e more often ignored b ecause the information structuring constraints

would strongly motivate rearranging the relative order of the constituents according

to their information status The intermingled ranking in implies that in this

Certainly there may b e more factors that inuence word order Semantic scop e is one An

element which has wider scop e tends to precede one with narrow scop e Diesing Another

factor may b e parsing considerations A long and complicated phrase tends to follow others Hawkins

Certain factors can blo ck otherwise p ossible scrambling Casemarking is such a factor For

CHAPTER OPTIMALITY THEORY AND SCRAMBLING IN GERMAN

language no one comp onent of grammar exclusively determines the word order In

this sense German is dierent from English in which it the syntactic comp onent

seems to play a dominant role and is also dierent from Czech where it is claimed

that the discoursepragmatic comp onent dominantly determines the order cf the

Praguean view of word order In German scrambling is fairly common but not

unrestricted either

The ranking predicts several things In particular the fact that cn is fairly high

in ranking predicts that scrambling is much more common among ob jects than with

the sub ject which is true in this language From the p ersp ective of information status

it predicts that topic whichisNewProm in our system tends to precede other

information typ es while completivefocuswhich is NewProm tends to follow

others and thus to b e immediately preverbal ie in the last nonverbal p osition in

SOV languages

Before we discuss the actual data let me briey mention the proso dic constraints

As discussed in chapter there is a proso dic requirement on a certain typ e of infor

mation ie New information Although there is also a counterconstraintwhich

prevents any element from having a pitch accent the ranking b etween them in in

German allows all fo cus elements to receive high pitch accents See in chapter

for more discussion

Proso dic Constraints

a N Put a high pitch accent on a newinformation element

b X Do not place any pitch accent

Ranking

N X

example in Korean if casemarkers which are optional in the default order are dropp ed scrambling

is not p ossible One explanation may b e that casemarking is the only means to tell grammatical functions apart when the canonical phrase structure is disturb ed

CHAPTER OPTIMALITY THEORY AND SCRAMBLING IN GERMAN

Therefore this ranking will have the eect of favoring a fo cused element with a high

pitch accent rather than one without

In discussing the output representations wehave calculated that there are the

dierent output structures taking the proso dic variation into account As seen in

the previous section each scrambled variant has eight dierent proso dic variants

Supp ose that the indirect ob ject dem Schuler is a fo cus ie New The rst eight

out of outputs are demonstrated in

candidates N X

a Hans dem Schuler das Buch

b HANS dem Schuler das Buch

Hans dem SCHULER das Buch c

d Hans dem Schuler das BUCH

e HANS dem SCHULER das Buch

f HANS dem Schuler das BUCH

g Hans dem SCHULER das BUCH

h HANS dem SCHULER das BUCH

The rst constraint N requires that the elementmarked New should havea

high pitch accent Since dem Schuler is New the outputs in which dem Schuler

do es not have a high pitch in violate this constraint ie a b d and f

Since this constraint is ranked higher than the other constraint X the four outputs

which violate the rst constraint are immediately ruled out The second constraint

in contrast requires that no elementshouldhave a high pitch accent and thus all

but the rst candidate a violate this constraint Moreover the outputs whichhave

more than one stressed elementhave more than one violation Now among the four

surviving candidates ie c e g and h from the rst constraint the candidate

CHAPTER OPTIMALITY THEORY AND SCRAMBLING IN GERMAN

c incurs the fewest violations with resp ect to the second constraint Hence c is

the optimal output in this context

Given the ranking b etween the two proso dic constraints the optimal output will

always b e the candidate which has all and only New elements with high pitch

accents Thus also for the second eight candidates whichhave a dierent constituent

order from the rst eight only the one with dem Schuler pitchaccented will survive

Likewise with the remaining candidates Therefore the comp etition b oils down

to six candidates ie the collection of the output which has the pitchaccented dem

Schuler from each distinctlyordered candidate group The relevant candidates are

illustrated in the tableau in

candidates prom

a Hans dem SCHULER das Buch

b Hans das Buch dem SCHULER

c dem SCHULER Hans das Buch

d das Buch Hans dem SCHULER

e dem SCHULER das Buch Hans

f das Buch dem SCHULER Hans

From now on for simplicitywe will only consider the six candidates which survive

the proso dic variation comp etition

CHAPTER OPTIMALITY THEORY AND SCRAMBLING IN GERMAN

Deriving SemanticDiscourse Eects in Scram

bling

Nowwehave set up the basic groundwork for the OT accountofscrambling Here

scrambled variants are considered to b e candidate outputs which comp ete to b e the

optimal one for the given input represented in the skeletal fstructure enriched with

discourse features The candidate outputs are evaluated through a set of constraints

b oth syntactic and discoursepragmatic The output which violates the leasthighly

ranked constraints is the optimal output

I will mainly discuss the emb edded ditransitive clause which has b een intro duced

earlier in the chapter ie da Hans dem Schuler das Buch gegeben hat that Hans gave

the student the b o ok Each tableau whichfollows represents EVAL ie the pro cess

of selecting the b est output for the given input context therefore one tableau p er

input context In all the following tableaux the candidate outputs are the distinct

variants diering in b oth syntactic and proso dic structures as discussed in section

even though I will mostly concentrate on the six distinct orders for the purp ose

of simplicity The fo cused element whether completiveorcontrastive is represented

in upp er case

Information Context Typ es

AllFo cus or Neutral Context

Given the information structure with the two features assumed here ie New and

Prom we can exp ect the following typ es of informational contexts First of all the

whole sentence can b e delivered unpartitioned This means that no elementofthe

sentence is either newer or more prominentthanany other element One such case

is when a sentence is newly presented in the discourse which can arise when the

CHAPTER OPTIMALITY THEORY AND SCRAMBLING IN GERMAN

sentence is intro duced as an answer to a question like What happened This is what

is often referred to as allfo cus Vallduv or neutral description Kuno

An example context is illustrated in

a Was ist passiert

what is happ ened

What happ ened

b Ich glaub e da Hans dem Schuler das Buch gegeb en hat

I b elieve that Hans the studentDat the b o okAcc given has

I b elieve that Hans gave the student the b o ok

As shown in b when a sentence is presented as an allfo cus or neutral descrip

tion it takes the socalled default order Let us now see how it will b e analyzed in

the OT account

In this context I assume that the sentence as a whole receives the feature marking

ie New not each individual element With resp ect to the feature Prom the

input will b e marked Prom in a normal situation If the question What happened

is delivered in a very emphatic manner eg with heavy stress on every item then it

will invoke Prom marking in the input for the answer sentence

Alternatively one could assume that each element is uniformly marked in this situation all as

NewProm

Hans dem Schuler das Buch

NewProm NewProm NewProm

Fo cus

Either way the tableau will generate the same result b ecause the inputs in i and are parallel in

the sense that they do not distinguish any one element from another and maketheinternal elements

informationally neutral with resp ect to each other

CHAPTER OPTIMALITY THEORY AND SCRAMBLING IN GERMAN

The input the sentence b in this context is given in For ease of exp osition

I simply list the nonverbal constituents and put the informationstructural feature

markings b elow each item Below the feature markings are the information typ es are

lab elled whichisgiven for claricational purp oses

Hans dem Schuler das Buch

NewProm

Focus

This input feeds the following EVAL pro cess expressed in the tableau in

candidates prom cn new cn

a Hans dem Schuler das Buch

b Hans das Buch dem Schuler

c dem Schuler Hans das Buch

d das Buch Hans dem Schuler

e dem Schuler das Buch Hans

f das Buch dem Schuler Hans

First lo ok at the information structuring constraints prom and new As exp ected

neither constraint applies in this situation Since no individual elementhasitsown

feature marking in the input which means that they are all neutral with resp ect

to each other the constraints simply cannot applyWecaninterpret this situation

as having none of the outputs in violate these constraints hence no marks for

prom or new In short unless a dierence in the feature marking exists in the input

the information structuring constraints are not imp ortant in determining the order

among constituents To put it in other words as long as the given context confers

CHAPTER OPTIMALITY THEORY AND SCRAMBLING IN GERMAN

the same information status either by not assigning features to individual elements

or by assigning the same features to all individual elements the context b ehaves as

if it were neutral with resp ect to word order In other words scrambling happ ens

only when there exist informational dierences among the comp onent elements

In the neutral context therefore the phrase structural constraints canon alone

cn and cn determine the b est constituent order of the sentence And this situ

ation is exactly the one where we should exp ect the unmarked or canonical order

to take place This prediction is b orne out As wehave seen in section the rst

output where the sub ject precedes the indirect ob ject which in turn precedes the

direct ob ject do es not violate cn or cn b ecause it ob eys all the subhierarchies b e

tween anytwo elements as prescrib ed by the constraints ie SUBJIOBJDOBJ

In contrast the second output b incurs one violation mark for cn b ecause the

direct ob ject das Buch precedes the indirect ob ject dem Schuler The output c on

the other hand incurs one violation mark for cn b ecause here the indirect ob ject

dem Schuler precedes the sub ject Hans In d the order b etween the sub ject and

the direct ob ject and also that b etween the indirect ob ject and the direct ob ject are

violated hence there is one mark for cn and another for cn In e the order

between the sub ject and the indirect ob ject and also that b etween the sub ject and

the direct ob ject are violated In this case it incurs two marks for cn Note that

b oth b and c and d and e incur the same numb er of violations However c

is worse than b and e is worse than d b ecause c and e involve a greater vio

lation of the higherranked constraint cn Finally output f incurs three violations

two for cn and one for cnwhich is the most severe violation of all In this output

the order is totally reversed from the canonical order Therefore not unexp ectedly

the rst output a which is in the default order is chosen as the optimal output in

this context

CHAPTER OPTIMALITY THEORY AND SCRAMBLING IN GERMAN

GroundFo cus Context

The next p ossible information context is the GroundFocus typ e This usually hap

p ens in a normal question and answer pair The elementwhich corresp onds to the

question phrase is the fo cus and the rest is the ground This is illustrated in

a Wem hat Hans das Buch gegeb en

whoDat has Hans the b o okAcc given

To whom did Hans give a book

b Ich glaub e da Hans dem SCHULER das Buch gegeb en hat

I b elieve that Hans the studentDat the b o okAcc given has

I b elieve that Hans gave the student a book

b Ich glaub e da Hans das Buch dem SCHULER gegeb en hat

I b elieve that Hans the b o okAcc the studentDat given has

I b elieve that Hans gave the b o ok to the student

Since the question is ab out wem who the completive fo cus in the sentence b or

b is dem Schuler the student Therefore dem Schuler is marked New and the

rest are marked New in the input Also since no element is presented prominently

in this context all elements are marked Prom This is shown in

Hans dem Schuler das Buch

NewProm NewProm NewProm

Ground Focus Ground

It might b e also p ossible that the ground elements together form an informational unit I do

not pursue this p ossibility in this dissertation The information unit formation should b e sub ject to

certain constraints to o syntactic or other See the brief discussion in section in chapter In

what follows I assume that each element is assigned an individual feature marking unless prohibited

by some other constraint For example see the discussion of the sp ecicity eect in section in

chapter and in this chapter

CHAPTER OPTIMALITY THEORY AND SCRAMBLING IN GERMAN

In this context the prom constraintwould not playany role b ecause every element

has the same status with resp ect to Prom Therefore the new constraintand

canon together decide the optimal output The violation patterns for cn and cn

for each output are just like the ones in the tableau for the allfo cus example and

it will remain the same for all the other tableaux to o b ecause the violations of the

phrase structural constraints are not contextsensitive

candidates pm cn nw cn

Hans dem SCHULER das Buch a

b Hans das Buch dem SCHULER

c dem SCHULER Hans das Buch

d das Buch Hans dem SCHULER

e dem SCHULER das Buch Hans

f das Buch dem SCHULER Hans

Unlike the previous case in whichthenew constraintalsowas not involved in

the optimality decision this input causes some violations for the new constraint A

violation o ccurs whenever a New element precedes a New element The rst

output a has one violation mark b ecause the fo cus dem Schuler which is New

precedes das Buch whichisNew Likewise f causes one violation mark In

contrast the second output b do es not violate this constraint b ecause the New

dem Schuler is the last element following the other elements For the same reason

d do es not violate the constraint either c and e on the other hand havetwo

violations b ecause the fo cus dem Schuler is in the initial p osition preceding two

other New elements

CHAPTER OPTIMALITY THEORY AND SCRAMBLING IN GERMAN

Note that the new constraintand cn are not ranked with resp ect to each other

and thus b oth rankings are p ossible Lets rst assume that new is ranked higher

than cn This yields b as the optimal output

candidates pm cn nw cn

a Hans dem SCHULER das Buch

b Hans das Buch dem SCHULER

c dem SCHULER Hans das Buch

d das Buch Hans dem SCHULER

e dem SCHULER das Buch Hans

f das Buch dem SCHULER Hans

On the other hand if the ranking is the reverse a is chosen as the optimal

output as illustrated b elow

candidates pm cn cn nw

a Hans dem SCHULER das Buch

b Hans das Buch dem SCHULER

c dem SCHULER Hans das Buch

d das Buch Hans dem SCHULER

e dem SCHULER das Buch Hans

f das Buch dem SCHULER Hans

CHAPTER OPTIMALITY THEORY AND SCRAMBLING IN GERMAN

As noted earlier this result is exactly what needs to b e captured Consider again

the question and answer pair discussed earlier rep eated here as

a Wem hat Hans das Buch gegeb en

whoDat has Hans the b o okAcc given

To whom did Hans give a book

b Ich glaub e da Hans dem SCHULER das Buch gegeb en hat

I b elieve that Hans the studentDat the b o okAcc given has

I b elieve that Hans gave the student a book

b Ich glaub e da Hans das Buch dem SCHULER gegeb en hat

I b elieve that Hans the b o okAcc the studentDat given has

I b elieve that Hans gave the b o ok to the student

It has often b een noted in the literature that a question like a invokes two p ossible

answers diering in order such as b and b Lenerz This optionality

noted in this context is nicely captured in the OT account given the constraint

interaction prop osed here

TopicFo cusTail Context

Finally there are also cases in which some elements are presented more prominently

than others In combination with the distinction in newness this creates a Topic

Fo cusTail typ e of information structure One common waytoinvoke prominence in

discourse is to highlight an elementby putting heavy stress on it or by putting it in

a phrase like what about Vallduv Consider the following context

a Wie stehts mit dem BuchW em hat Hans das Buch gegeb en

how stands it with the b o ok who has Hans the b o okAcc given

What ab out the b o okT o whom did Hans give the b o ok

CHAPTER OPTIMALITY THEORY AND SCRAMBLING IN GERMAN

b Ich glaub e da das Buch Hans dem SCHULER gegeb en hat

I b elieve that the b o okAcc Hans the studentDat given has

I b elieve that Hans gave the b o ok to the student

As b efore dem Schuler is the fo cus in this context to o b eing the phrase whichcor

resp onds to the whword in the question Unlike the case presented b efore however

the direct ob ject das Buch is now presented prominently Das Buch receives more

attention than other elements in the discourse since the discourse is now set up to

talk ab out das Buch Namely das Buch is the topic This context is represented

in the input as follows

Hans dem Schuler das Buch

NewProm NewProm NewProm

Tail CompFocus Topic

The indirect ob ject dem Schuler is marked New and the sub ject Hans and the

direct ob ject das Buch are marked New On the other hand the ob ject das Buch

is marked Prom and the rest are marked Prom Therefore das Buchwhich

is marked NewProm is the topic of the sentence dem Schuler b eing marked

NewProm is the fo cus ie a completive fo cus nally Hans is the tail marked

NewProm

This typ e of input provokes all three constraints to come in to play b ecause the

elements involved are distinct not only in newness but also in prominence So the

As mentioned earlier some sp eakers do not like the order in which the sub ject is preceded by

some other element in the sentence In that case cn is ranked higher than prom One more source

of unacceptabilitymay b e the wayinwhich a sp eaker classies the discourse context Sp eakers can

vary in their judgments of the given context That is if the sp eaker do es not consider das Buch in

to b e prominent enough in that context shehe would not mark das Buch as Prom and

then das Buch would not b e able to scramble Hence the order in b is not p ossible for such

sp eakers For more discussion of sp eaker variation see section

CHAPTER OPTIMALITY THEORY AND SCRAMBLING IN GERMAN

prom constraint also do es its job

candidates prom cn new cn

a Hans SCHULER das Buch

b Hans das Buch SCHULER

c SCHULER Hans das Buch

d das Buch Hans SCHULER

e SCHULER das Buch Hans

f das Buch SCHULER Hans

Rememb er that the information structuring constraint Prom demands that a Prom

element precede a Prom element Since the direct ob ject das Buch is Prom in

this context it should precede all other elements Hence any order in which das Buch

is not in the rst p osition violates this constraint b and e incur one violation and

a and c incur two violations Since prom is the highest constraint of all the

choice is narrowed down to d and e With resp ect to the next highest constraint

ie cn d evokes one violation and e evokes two Hence d is the optimal

output

In the direct ob ject as a topic scrambles all the way to the initial p osition

Recall that topic is the most scramblable information typ e b ecause it is simultaneously

New and Prom Thus b oth the new constraintandtheprom constraint

facilitate its scrambling Notice the contrast in das Buchs scramblabilitybetween

and As a tail in it cannot scramble over the sub ject but as a topic

in it can Also note that while it can stay in situ as a tail in the direct

ob ject necessarily scrambles as a topic See section in chapter for discussion

of the dierences b etween topic and tail

CHAPTER OPTIMALITY THEORY AND SCRAMBLING IN GERMAN

Let us nowseehow this OT account of scrambling captures the pragmatic and

semantic eects whichhave b een discussed in chapter These eects havebeen

identied as the antifo cus eect the sp ecicity eect and the contrastivefocus

eect It will b e shown in the sections that follow that once scrambling is viewed

as the pro cess of optimizing structure for a given context these eects naturally

follow from the theory without sp ecial arrangements Also the sub ject eect in

scrambling ie that a sub ject is much harder to scramble over than other elements

are is naturally captured by the high ranking of cn

AntiFo cus Eect

Wehave seen in the previous section that a certain question context ie one in which

the indirect ob ject is in question generates two p ossible answers one in the canonical

order and the other in a scrambled order rep eated here in However this

optionality do es not always happ en Lenerz The question context illustrated

below in allows only the canonicallyordered sentence as the p ossible answer as

shown in b but disallows a scrambled one as in b Compare and

a Wem hat Hans das Buch gegeb en

whoDat has Hans the b o okAcc given

To whom did Hans give a book

b Ich glaub e da Hans dem SCHULER das Buch gegeb en hat

I b elieve that Hans the studentDat the b o okAcc given has

I b elieve that Hans gave the student a book

b Ich glaub e da Hans das Buch dem SCHULER gegeb en hat

I b elieve that Hans the b o okAcc the studentDat given has

I b elieve that Hans gave the b o ok to the student

CHAPTER OPTIMALITY THEORY AND SCRAMBLING IN GERMAN

a Was hat Hans dem Schuler gegeb en

whatAcc has Hans the studentDat given

What did Hans give to the student

b Ich glaub e da Hans dem Schuler das BUCH gegeb en hat

I b elieve that Hans the studentDat the b o okAcc has

I b elieve that Hans gave the student the b o ok

b Ich glaub e da Hans das BUCH dem Schuler gegeb en hat

I b elieve that Hans the b o okAcc the studentDat given has

I b elieve that Hans gave the b o ok to the student

The dierence b etween and lies in that the scrambled element in b is

a nonfo cused phrase whereas the one in b is a fo cused one Thus it has b een

prop osed in the literature as a constraining condition on scrambling that a fo cused

element cannot scramble Lenerz Abraham Moltmann Web elhuth

This is what I called the antifo cus eect in chapter

Let us examine how the OT account prop osed here explains this phenomenon

It is accounted for in a fairly straightforward manner in this framework Not unex

p ectedlythe new constraint can easily handle this problem Consider rst the input

representation for

Hans dem Schuler das Buch

NewProm NewProm NewProm

Ground CompFocus

Just as in the previous question and answer pair the elementwhich corresp onds to

the question word is the fo cus of the sentence hence marked New in the input The

direct ob ject das Buch the b o ok is New in this context Then the sub ject Hans

CHAPTER OPTIMALITY THEORY AND SCRAMBLING IN GERMAN

and the indirect ob ject dem Schuler are marked New With no other contextual

clues no elements are assumed to b e prominent Therefore all elements are marked

Prom

This input provides the tableau in Just as in the previous case only the six

outputs whichhave survived the proso dic constraints are listed in the tableau Thus

the fo cus das Buch has a high pitch accent in all the outputs in

candidates prom cn new cn

a Hans dem Schuler das BUCH

b Hans das BUCH dem Schuler

c dem Schuler Hans das BUCH

d das BUCH Hans dem Schuler

e dem Schuler das BUCH Hans

f das BUCH dem Schuler Hans

The prom constraint do es not apply in this case b ecause all the elements are marked

Prom Therefore the decision is to b e made by the interaction b etween the new

constraintandcanon cnwhich is higher than cn and the new constraint quite

easily narrows down the choice to a and b b ecause all the others violate this high

constraint and thus are quickly eliminated from the comp etition Since das Buch

is New here the new constraint demands that it follow other elements There

fore the outputs where das Buch is not preceded by the other elements violate this

constraint Between a and b b is discarded b ecause here the New element

das Buch precedes a New element dem Schuler causing a violation of the new

constraint Therefore a is the optimal output That is this context cho oses a as

the only optimal output and only this output is grammatical in this context This

result matches the fact that the question context in has only one p ossible answer

CHAPTER OPTIMALITY THEORY AND SCRAMBLING IN GERMAN

It would b e interesting at this p oint to compare the tableau here with that in the

previous case which has two optimal outputs ie for Candidates a and b

are rep eated b elow from each tableau a and b are extracted from and

a and b are from in the previous section

candidates prom cn new cn

a Hans dem Schuler das BUCH

b Hans das BUCH dem Schuler

Hans dem SCHULER das Buch a

b Hans das Buch dem SCHULER

In the rst case ie a and b a is denitely the b est choice In a whichis

in the default order the New element das Buch is already in the nal p osition of

all the elements hence not violating the new constraint Moreover this is the only

candidate that do es not violate the phrase structural constraint canon either So

the canonicallyordered sentence is free of violation in this context In contrast the

candidate b violates b oth the new constraintandcn In this order das Buch is

scrambled over dem Schuler and therefore not only disturbs the canonical order of

the sentence but also disob eys the desirable order b etween old and new elements

Hence this context denitely cho oses candidate a over candidate b as the optimal

output thus giving us only one grammatical output

In comparison the choice is not as clear in the second case In the candidate a in

the default order the New element dem Schuler is not in the last p osition preced

ing the New element das Buchandthus violating the new constraint However

the candidate b do es not violate the new constraintby scrambling the New

elementover the New element ie das Buch over dem Schuler In other words

scrambling saves this candidate from violating the information structuring constraint

CHAPTER OPTIMALITY THEORY AND SCRAMBLING IN GERMAN

New although it forces it into violating the phrase structural constraint cnIn

terestingly the scrambling candidate b still survives b ecause in this grammar the

informationstructuring new constraint is ranked in a tie with the phrase structural

constraint cnThisiswhywe can havetwo grammatical outputs in this context

ContrastiveFo cus Eect

So far wehave considered contexts whichinvoke purely newinformation typeoffocus

ie completive fo cus This is a nonprominent fo cus ie the fo cus which is marked

NewProm in our system However our system also allows for prominent fo cus

ie that of NewProm whichIhave lab eled contrastive fo cus in chapter What

is interesting ab out this information typ e is that one part of it ie New encourages

it to fol low other elements while the other part ie Prom encourages it to precede

them In a sense the information structure of this discourse function contains an

internal conict Lets see how this conict is resolved in what follows

A contrastive fo cus can b e evoked in several dierent contexts Simply stressing

the question word suchasinWAS hat Hans dem Schuler gegeben WHAT did Hans

give to the student may suce to make the fo cus prominent ie contrastive

Alternatively as Dik et al illustrate one of the contrastive contexts ie

selecting expanding restricting replacing and parallel may facilitate making fo cus

prominent by means of evoking relevant sets of alternatives see in chapter

for more discussion

For instance a replacing context such as in will grant prominence to the fo

cused element das Buch the b o ok in the answer by comparing it with an alternative

ie die Zeitung the newspap er In this case the fo cused phrase can scramble

CHAPTER OPTIMALITY THEORY AND SCRAMBLING IN GERMAN

a Was hat Hans dem Schuler gegeb endie Zeitung

what has Hans the studentDat given the newspap er

What did Hans give to the studentThe newspap er

b Ich glaub e da Hans das BUCH dem Schuler gegeb en hat

I b elieve that Hans the b o ok the student given has

nicht die ZEITUNG

not the newspap er

I b elieve that Hans gave the b o ok to the student not the newspap er

As argued in chapter a contrastive fo cus is represented as NewProm in the

input ie as prominent new information Therefore for the ab ovecontext the input

will b e marked as follows

Hans dem Schuler das Buch

NewProm NewProm NewProm

Topic Tail ContFocus

As mentioned ab ove a contrastive fo cus is a very puzzling case b ecause while its new

ness New discourages it from scrambling its prominence Prom encourages

it to do so given the information structuring constraints prop osed in this chapter As

exp ected Prom wins and thus the fo cused phrase scrambles b ecause the prom

constraint outranks the new constraint in this language This is illustrated in the

tableau in

Note that the sub ject Hans is also marked Prom I assume that unlike other elements of

the sentence the sub ject can easily b e the topic of the sentence even without a high pitch or b eing

presented in the what about phrase In other words unless something else is presented as the topic

sub ject tends to b e assigned Prom although Prom is also an option From now on I will only

consider cases where the sub ject is the topic of the sentence

CHAPTER OPTIMALITY THEORY AND SCRAMBLING IN GERMAN

candidates prom cn new cn

a Hans dem Schuler das BUCH

Hans das BUCH dem Schuler b

c dem Schuler Hans das BUCH

d das BUCH Hans dem Schuler

e dem Schuler das BUCH Hans

f das BUCH dem Schuler Hans

Since das Buch is a prominentelement the prom constraint instructs it to scramble

and thereby to precede all nonprominentelements of the sentence However by doing

so it will violate the New constraint Compare b and d Neither violates the

prom constraint If das Buch scrambles over the sub ject Hans as in d however

it not only adds another mark to the new constraint but also violates cn On

the other hand if it stops in the middle as a compromise it do es not violate the

higher constraint cn Therefore the optimal case would b e the one in which das

Buch scrambles over dem Schuler butnotover Hans This result indeed obtains as

conrmed in Lenerz Moltmann

Compare this contrastive fo cus case with the topic case discussed in the previous

subsection There wehave seen the direct ob ject scramble over the sub ject This

fact conrms our initial prediction that topic is the most scramblable entity It

should b e noted also that I do not claim that in all cases where a direct ob ject is the

contrastive fo cus should it scramble over an indirect ob ject It scrambles over the

indirect ob ject only when it is more prominent than the latter If the indirect ob ject

were also prominent then the direct ob ject would not b e motivated to scramble and

actually would b e punished by canon if it did When that is the case then the

Many sp eakers nd a still pretty go o d in this context Sp eaker variation is briey discussed later in this section

CHAPTER OPTIMALITY THEORY AND SCRAMBLING IN GERMAN

contrastively fo cused phrase also stays in its canonical p osition

Sp ecicity Eect

Wehave considered the scrambling of denite phrases so far As often noted however

indenite phrases or weak phrases do not necessarily b ehavelike denite or strong

phrases As discussed in chapter a weak phrase cannot scramble Lenerz

Abraham Moltmann de Ho op Diesing Web elhuth which

has b een referred to as the sp ecicity eect in chapter It has also b een noted

that an exception exists to the sp ecicity eect which o ccurs when contrastivefocus

is involved Lenerz Abraham Moltmann I devote this section to

discussing the problems related to indenite phrases

A NonSp ecic Phrase as CompletiveFo cus

If a nonsp ecic indenite phrase is intro duced in the discourse as a completivefocus

ie as nonprominent new information it b ehaves just like a denite phrase First

consider the following context

a Was hat Hans dem Schuler gegeb en

whatAcc has Hans the studentDat given

What did Hans give to the student

b Ich glaub e da Hans dem Schuler ein BUCH gegeb en hat

I b elieve that Hans the studentDat a b o okAcc given has

I b elieve that Hans gave the student a book

The indenite phrase ein Buch a b o ok in b is intro duced as an answer to the

question phrase was what in a Since there is no sign that it is prominent ein

CHAPTER OPTIMALITY THEORY AND SCRAMBLING IN GERMAN

Buch is a completive fo cus hence marked NewProm

Hans dem Schuler ein Buch

NewProm NewProm NewProm

Topic Tail CompFo cus

Now since other elements are old information in this context as illustrated in the

input then we can easily predict that the fo cus will follow all other elements

Undoubtedly this prediction is b orne out This is illustrated in the tableau in

candidates prom cn new cn

a Hans dem Schuler ein BUCH

b Hans ein BUCH dem Schuler

c dem Schuler Hans ein BUCH

d ein BUCH Hans dem Schuler

e dem Schuler ein BUCH Hans

f ein BUCH dem Schuler Hans

The information structuring constraint prom and the phrase structural constraint

cn immediately rule out all outputs where the sub ject is scrambled over Therefore

the comp etition is b etween a and b The second constraint New nowchecks

if the fo cus ein Buch is followed byanyNew element and marks any candidate

Briey recall the discussion in section and in chapter that the feature marking in the

input is also not unrestricted Esp ecially nonsp ecic indenite NPs are restrained from having the

New feature as shown in b

i specificity

a sp A sp ecic phrase should not b e New

b sp A nonsp ecic phrase should not b e New

However this constraint do es not apply in this case b ecause the context assigns New to the

indenite NP ein Buch in

CHAPTER OPTIMALITY THEORY AND SCRAMBLING IN GERMAN

in which it is Candidate a and candidate c survive this constraint they are the

only candidates where the New element ein Buch follows all New elements

Therefore naturally a comes out as the optimal output since it do es not violate

cn as well while b do es In short when interpreted as completive fo cus a non

sp ecic indenite phrase is exp ected to stay in its canonical p osition just as a sp ecic

phrase do es Therefore the completive fo cus context do es not cause a problem since

it do es not motivate a nonsp ecic phrase to scramble

If the indirect ob ject is the topic then it may aect the relative order b etween

the sub ject and the indirect ob ject itself but it would not change the fact that

the fo cused direct ob ject ein Buch should stay in its canonical p osition Given the

constraints intro duced in this chapter a completively fo cused direct ob ject sp ecic

or not always has to stay in its canonical p osition b ecause rst neither of the

information structuring constraints New or Prom would motivate it to scramble

since the completive fo cus gets the weakest marking in terms of scramblabilityie

NewProm second the canon constraint cnwould always prefer it actually

anything to stay in the canonical p osition

It is notable that the present OT account has the interesting consequence that it

has the option to allow the immediate preverbal elementtobeinterpreted as fo cus

Due to the new constraint the fo cus element tends to b e p ositioned as the rightmost

element of the nonverbal elements and the rightmost p osition is the preverbal p o

sition in an SOV language like German or Korean Not surprisinglyseveral p eople

have prop osed that the preverbal p osition is the fo cus p osition in these languages

Krifka Kim Jo following Szab olcsi and Kiss So the

fo cused phrase is either basegenerated or moved to the preverbal p osition in these

analyses Reinhart makes a slightly dierent prop osal that a nonfo cused

CHAPTER OPTIMALITY THEORY AND SCRAMBLING IN GERMAN

elementmoves or scrambles to allow the leftover element to b e fo cused Our new

constraint can capture this intuition

However fo cus is not necessarily immediately preverbal It can stay in the ele

ments canonical p osition This is shown in This is also true with denite NPs

as wehave seen ab ove

a Wem hat Hans das Buch gegeb en

whoDat has Hans the b o okAcc given

To whom did Hans give the b o ok

b Ich glaub e da Hans einem SCHULER das Buch gegeb en hat

I b elieve that Hans a studentDat the b o ok given has

I b elieve that Hans gave a student the b o ok

b Ich glaub e da Hans das Buch einem SCHULER gegeb en hat

I b elieve that Hans the b o ok a studentDat given has

I b elieve that Hans gave the b o ok to a student

The current prop osal explains this optionality of fo cus generation bytheinteraction

between the new constraintandcanon New motivates fo cus to b e preverbal while

canon encourages it to remain in situ Therefore our prop osal has an advantage

over the preverbal fo cus accounts in that while it can certainly accommo date the

observation that fo cus tends to b e generated preverbally it also allows it to b e

generated in the elements canonical p osition

She actually argues that the ob ject in Dutch scrambles to makethe verb fo cused This is not

necessarily true a nonverbal element eg argument or adjunct can also b e fo cused excluding the

verb from its fo cus domain if the ob ject scrambles over it

CHAPTER OPTIMALITY THEORY AND SCRAMBLING IN GERMAN

A NonSp ecic Phrase as Old Information

A problem arises when a nonsp ecic phrase is not intro duced in the discourse as a

fo cus or new information One suchcontext is when the fo cus of the sentence falls on

some other element so that an indenite NP is forced to b e interpreted as a taillike

element This is illustrated in

a Wem hat Hans ein Buch gegeb en

whoDat has Hans a b o okAcc given

To whom did Hans give a book

b Ich glaub e da Hans dem SCHULER ein Buch gegeb en hat

I b elieve that Hans the studentDat a b o ok given has

I b elieve that Hans gave the student a book

b Ich glaub e da Hans ein Buch dem SCHULER gegeb en hat

I b elieve that Hans a b o ok the studentDat given has

I b elieve that Hans gave a b o ok to the student

Here the indirect ob ject dem Schuler is the fo cus ie New and thus the non

sp ecic indenite NP ein Buch along with the sub ject Hans is not interpreted as

New What is it thenIf it really w ere a tail and hence marked New then it

should b e able to scramble just as a sp ecic New element can However this is

not the case as shown in This is what has b een called the sp ecicity eect of

scrambling

Recall the discussion in section in chapter where it was argued that the sp eci

city eect in scrambling should b e handled by the semanticsdiscoursepragmatics

mapping principles which constrain the feature marking in the input rather than

the syntaxsemantics ones cf de Ho op Diesing Based on the

observation that a nonsp ecic indenite NP usually intro duces a new entityinto

CHAPTER OPTIMALITY THEORY AND SCRAMBLING IN GERMAN

the discourse while a sp ecic indenite NP refers to an entitywhich is already

existent in the discourse cf Heim Enc Web elhuth I prop osed

the specificity constraints which are basically resp onsible for the linking b etween

the semantic notion of sp ecicity and the pragmatic notion of discourseoldness or

discoursefamiliarity and thus control the feature assignment of the relevant phrases

in the input

specificity

a sp A sp ecic phrase should not b e New

b sp A nonsp ecic phrase should not b e New

Ihave further argued that the rst constraint sp can b e violated while the sec

ond constraint sp cannot Namely a sp ecic phrase can b e interpreted as a new

New entity if it is forced to by the context One suchcontext is when it is

interpreted as a fo cus as in

a Was hat Hans dem Schuler gegeb en

whatAcc has Hans the studentDat given

What did Hans give to the student

b Ich glaub e da Hans dem Schuler das BUCH gegeb en hat

I b elieve that Hans the studentDat the b o okAcc has

I b elieve that Hans gave the student the b o ok

The sp ecic NP das Buch here is interpreted as if it were a newlyintro duced entity

although it has b een referentiallyanchored in the discourse That is why it is rep

resented as a denite NP rather than an indenite NP This temp orary newness is

achieved by its capability of violating the specificity constraint sp See for

detailed discussion of this problem cf Diesing Vallduv

CHAPTER OPTIMALITY THEORY AND SCRAMBLING IN GERMAN

In contrast a nonsp ecic indenite NP cannot really b e interpreted as an old

entityby itself As shown in if a nonsp ecic NP fails to b e referentially anchored

usually by b eing replaced by a denite NP or a pronoun as in b after its initial

intro duction into the discourse it is not interpreted as an indep endententity any

more but rather b ehaves as if it were part of the predicate de Ho op

a I b ought a b o ok

b Where did you buy itthe b o ok

b Where did you buy a b o ok

That is if a nonsp ecic NP is forced to b e interpreted as a New entity whichit

cannot b ecause of the sp constraint it constitutes another information unit with the

predicate By b eing part of the larger unit it avoids the problem of b eing assigned

the New feature as an individual although the larger unit is assigned New In

other words a nonsp ecic NP has an indep endent life in the discourse only as a new

entity but once it has lost its newness it loses its indep endence to o

As noted in the previous chapter these constraints dier from other constraints

wehave seen so far eg canon promand new in that they are not concerned

with the mapping b etween inputs and outputs but rather are concerned with the

mapping within input markings In other words a nonsp ecic NP cannot havea

New marking in the input due to its inherentsemantic incompatibility with this

discourse prop ertyNew whereas a sp ecic NP mayhave a New marking in

the input Again this distinction dep ends on the constraint strength

Therefore in a context like the nonsp ecic NP ein Buch is not marked with

any discourse features as illustrated in

a Wem hat Hans ein Buch gegeb en

whoDat has Hans a b o okAcc given

To whom did Hans give a book

CHAPTER OPTIMALITY THEORY AND SCRAMBLING IN GERMAN

b Ich glaub e da Hans dem SCHULER ein Buch gegeb en hat

I b elieve that Hans the studentDat a b o ok given has

I b elieve that Hans gave the student a book

b Ich glaub e da Hans ein Buch dem SCHULER gegeb en hat

I b elieve that Hans a b o ok the studentDat given has

I b elieve that Hans gave a b o ok to the student

Hans dem Schuler ein Buch

NewProm NewProm

Topic CompFocus

Note that ein Buch is not assigned a Prom feature either It follows from the fact

that the NP is now part of a larger information unit and thus cannot b e assigned

any feature individually This predicts that a nonsp ecic indenite NP cannot b e

interpreted as a topic either This seems to b e a desirable result For example

a nonsp ecic NP cannot b e put in a what about phrase unless it is interpreted as

generic or sp ecic as shown in

What ab out a b o ok

candidates prom cn new cn

Hans dem Schuler ein BUCH a

b Hans ein BUCH dem Schuler

c dem Schuler Hans ein BUCH

d ein BUCH Hans dem Schuler

e dem Schuler ein BUCH Hans

f ein BUCH dem Schuler Hans

CHAPTER OPTIMALITY THEORY AND SCRAMBLING IN GERMAN

Now since ein Buch is not marked New it do es not violate the new constraint

even if it follows the New element dem Schuler Therefore it is not restricted with

resp ect to the new constraint However the phrase structural constraints canon

forbid it to b e out of its canonical p osition In other words no element can scramble

out of its canonical p osition unless it has an individual feature assignment Conse

quently the rst output the canonically ordered one is the optimal output And

this is why the scrambled variant in b is bad

A NonSp ecic Phrase as ContrastiveFo cus

Finally consider what happ ens if a nonsp ecic phrase is a contrastive fo cus Just

as in the denite NP cases a context like ie a replacing context provokes a

contrastive fo cus see section for more discussion on contrastive fo cus Inter

estingly in this case a nonsp ecic indenite phrase suchas ein Buch can scramble

a Was hat Hans dem Schuler gegeb eneine Zeitung

what has Hans the studentDat given a newspap er

What did Hans give to the studentA newspap er

b Ich glaub e da Hans ein BUCH dem Schuler gegeb en hat

I b elieve that Hans a b o ok the student given has

nicht eine ZEITUNG

not a newspap er

I b elieve that Hans gave a b o ok to the student not a newspap er

Recall that this has b een the most troublesome case for previous analyses which

are either based on sp ecicity de Ho op Diesing or on fo cality Lenerz

Web elhuth see chapter for detailed discussion This case runs against

b oth of these accounts ie a scrambled phrase should b e sp ecic or a scrambled

CHAPTER OPTIMALITY THEORY AND SCRAMBLING IN GERMAN

phrase should b e unfo cused b ecause the indenite NP ein Buch in is neither

sp ecic nor unfo cused

In the current account this is not a sp ecial case Just like the denite phrase the

indenite NP ein Buch is marked NewProm since it is contrastively fo cused in

this context

Hans dem Schuler ein Buch

NewProm NewProm NewProm

Topic Tail ContFocus

The nonsp ecic phrase ein Buch b eing New now is not constrained by the speci

ficity constraint Rememb er that the specificity constraintonlycontrols the cases

where a nonsp ecic phrase is marked New

candidates prom cn new cn

a Hans dem Schuler ein BUCH

b Hans ein BUCH dem Schuler

c dem Schuler Hans ein BUCH

d ein BUCH Hans dem Schuler

e dem Schuler ein BUCH Hans

f ein BUCH dem Schuler Hans

The prom constraint narrows down the comp etition to b etween b and d Because

d violates the cn constraint however b is the optimal output And this explains

the problematic contrastive fo cus case The nonsp ecic NP ein Buch maintains its

New marking b ecause it is not restricted bythespecificity constraint and as

a Prom element it can b e scrambled

CHAPTER OPTIMALITY THEORY AND SCRAMBLING IN GERMAN

Sp eaker Variation

Before closing this section let me briey add a few words ab out sp eaker variation

in the acceptability of scrambled sentences Not surprisingly not all sp eakers readily

accept using the scrambled order in contrastive fo cus contexts such as in Also

not all contrastivecontexts are conceived of as equally contrastive That is some

sp eakers consider a certain context more contrastive than others I have argued in

chapter that socalled fo cus adverbs suchasnur only or sogar even trigger a

contrastive fo cus context by restricting or expanding the set of alternatives Dik

et al The examples are illustrated in

a weil Hans NUR ein BUCH dem Mann gegeb en hat

b ecause Hans only a b o okAcc the manDat given has

b ecause Hans gave only a b o ok to the man

b weil Hans SOGAR ein BUCH dem Mann gegeb en hat

b ecause Hans even a b o okAcc the manDat given has

b ecause Hans gaveeven a b o ok to the man

Although many p eople accept b oth a and b equally some prefer a to

b or vice versa

Likewise in the same environment eg in a replacing context some p eople would

accept the scrambled order as in b but others would not likeitasmuch

a weil Hans dem Schuler ein BUCH gegeb en hat nicht eine ZEITUNG

b ecause Hans the student a book given has not a newspap er

b ecause Hans gave a b o ok to the student not a newspap er

b weil Hans ein BUCH dem Schuler gegeb en hat nicht eine ZEITUNG

b ecause Hans a b o ok the student given has not a newspap er

b ecause Hans gave a b o ok to the student not a newspap er

CHAPTER OPTIMALITY THEORY AND SCRAMBLING IN GERMAN

It may b e b ecause to those sp eakers this context is not contrastive enough or the

fo cused element eg ein Buch is not prominent enough Naturally sp eakers can vary

in their judgments of the context However it is remarkable that almost all sp eakers

agree that the scrambled NPs in and are certainly more contrastive or more

prominent than the insitu counterparts

Therefore I argue that sp eaker variation arises b ecause there are dierences in

howspeakers p erceive a context or the informational status of an element That is

it is up to sp eakers judgment whether a certain context is contrastive or not or

whether a certain element is prominent enough or not For example one can p erceive

that a restricting context is contrastive while an expanding context is not or vice

versa Likewise with a replacing context Some could see the context for as

not contrastive at all while others may p erceive it as contrastiveeven without the

overt replacing phrase nicht eine Zeitung In particular prominence seems more

likely to b e sub ject to sp eaker variation than newness It seems quite plausible

that the prominence judgmentwould b e more sub jective than the newness judgment

However once a sp eaker judges a certain context as contrastive or an elementas

prominent then shehe will accept the scrambled sentences Therefore I claim that

sp eaker variation o ccurs b ecause sp eakers can vary in classifying the context and thus

co de the input dierently not b ecause the mapping from inputs to outputs varies

Summary

This chapter has explored what kinds of principles motivate and also restrict scram

bling in terms of Optimality Theory Using the featurebased information structure

develop ed in chapter as the ma jor part of input representation I have argued that

each scrambled variant is the b est structural description of a particular discourse

contextual information in the input Ihave prop osed that the mapping from each

CHAPTER OPTIMALITY THEORY AND SCRAMBLING IN GERMAN

input to its optimal output in scrambling is constrained by a relatively small num

berofsyntactic and discoursecontextual constraints ie cn cn prom and new

which will b e further restrained by the semantic constraints specificity on the fea

ture marking in the input representation I prop osed the following ranking for the

German scrambling phenomena

Constraint Ranking German

new

prom cn

In addition to having captured the oftenobserved semantic and discourse eects

asso ciated with scrambling such as the sp ecicityantifo cus and contrastivefocus

eects rep orted in the literature the OT analysis prop osed in this chapter predicts

the following generalizations

a An element in its canonical p osition can b e any of the four information typ es

ie topic tail contrastive fo cus or completivefocus

b A topic NewProm is never scrambled over by other elements

c A completive fo cus NewProm never scrambles ie a scrambled ele

ment cannot b e a completivefocus

d A nonsp ecic element can scramble only as a contrastivefocus

e The sub ject in German is hard to scramble over b ecause the phrase structural

constraint cn is a strong constraint

Although they require more thorough testing the generalizations in seem to

match nativespeakers intuitions

Chapter

Order and Morphology Korean

Wehave observed in chapter and chapter that scrambling ie the reordering

of constituentsofasentence out of their canonical p ositions is a ma jor resource for

enco ding information structure in German In particular scrambling is argued to

b e motivated bytwo information structuring constraints one is the new constraint

whichmotivates a New element to precede a New element the other is the

prom constraint which imp els a Prom element to precede a Prom element

see chapter for the motivations for these information structuring constraints

This chapter examines scrambling phenomena in Korean also from the p ersp ective

of information structuring It is argued that scrambling in Korean is also motivated

and constrained by the interaction of the information structuring constraints devel

op ed in chapter and the phrase structural constraints motivated in chapter

Sp ecically sp ecial attention is paid to the dierence in information enco ding b e

tween the scrambling of regularly casemarked phrases and that of phrases marked

with the socalled topic marker nun It will b e shown that the dierence results from

the interaction b etween constituent order and morphological marking

CHAPTER ORDER AND MORPHOLOGY KOREAN

Information Structuring by Morphology

nunTopic or ContrastiveFo cus

Topicho o d in Korean is traditionally b elieved to b e enco ded by morphology ie by

the socalled topic marker nun cf wa in Japanese For example unlike the regular

nominative case marker kawhich do es not in itself endow a sub ject phrase with any

sp ecial information status the topic marker nun makes a phrase the topic of the

sentence This contrast is exemplied in

a Swunika Inholul mannassta

SwuniNom InhoAcc meetPstDcl

Swuni met Inho neutral

b Swuninun Inholul mannassta

SwuniTop InhoAcc meetPstDcl

As for Swuni she met Inho topic

The sub ject phrase Swunimarked with ka in a is neutral in the sense that it

is not informationally restricted in this default conguration it can b e interpreted

as new or old prominent or nonprominent information dep ending on the context

In the information structure prop osed in this thesis this means that the kaphrase

can b e marked New or New and Prom or Prom On the other hand

the sub ject phrase in b whichismarked with the socalled topic marker nun

The information status of a regularly casemarked phrase will b e determined by its relative

p osition in the sentence as will b e shown b elow

Analogous to its Japanese counterpart ga Kuno it has b een prop osed that ka

in Korean also induces only newinformation readings ie a neutral description reading or an

exhaustive listing reading cf a neutral or fo cus reading in Kim the neutral description

reading is analogous to our completive fo cus and the exhaustive listing to our contrastive fo cus

As we will see b elow however a kaphrase can also enco de old information which distinguishes it

from the Japanese counterpart ga see Haig for a comparative discussion of the Korean and Japanese sub ject phrases

CHAPTER ORDER AND MORPHOLOGY KOREAN

is interpreted as the topic or theme Kuno of the sentence The sub ject

Swuni here is presented as a distinct entity from the rest of the sentence it is what

the sentence is ab out and what the rest of the sentence is commenting on In the

currently assumed information structure Swuni is marked NewProm

The informational restrictedness of a nunphrase comes out in a context suchas

the one b elow In a neutral description Kuno or an allfo cus Vallduv

context as in ie one which requires that all elements of the sentence b e

equally presented as new information only the kamarked sub ject as in b but not

the nunmarked sub ject as in b is allowed

a ecey mwusun ili issessni

yesterdaywhat eventNom b ePstInt

What happ ened yesterday

b Swunika Inholul mannassta

SwuniNom InhoAcc meetPstDcl

Swuni met Inho neutral

b Swuninun Inholul mannassta

SwuniTop InhoAcc meetPstDcl

As for Swuni she met Inho topic

The nun marking in b in contrast to the regular case marking in b makes the

sub ject phrase Swuninun stand out as old and prominent information among other

elements of the sentence and partitions the sentence as such so an allfo cus context

which requires equal information status for all elements of the sentence would not

allowa nunphrase in it The sub ject kaphrase Swunika here is allowed in this

context b ecause it is not informationally restricted in this default conguration

The following context shows that nun as a topic marker is not compatible with a

tail element either which is old but not prominent information

CHAPTER ORDER AND MORPHOLOGY KOREAN

a eceynun Swunika ecey nwukwulul mannassni

yesterdayTop SwuniNom yesterday whoAcc meetPstInt

What ab out yesterdayWho did Swuni meet y esterday

b Swunika Inholul mannassta

SwuniNom InhoAcc meetPstDcl

Swuni met Inho

b Swuninun Inholul mannassta

SwuniTop InhoAcc meetPstDcl

As for Swuni she met Inho

In ecey yesterday is presented as the topic of the sentence Therefore the

sub ject Swuni in b and b is not conceived of as a topic although it is old

information b ecause it is already mentioned in the previous sentence Instead it is

a tail ie NewProm in the current information structure Since the nominative

case marker is informationally neutral as argued ab ove the sub ject phrase marked

with the regular nominativecasemarker in b is allowed in this context However

the sub ject phrase marked with nunwhich is informationally very restricted is not

allowed in this context as shown in b That is nun cannot mark tailho o d

Information enco ding by nun is not limited to topic In addition to the topic

reading describ ed ab ove the same morpheme nun enco des a fo cus reading more

sp ecically a contrastive fo cus reading Consider the example in

a Swunika Inholul mannassta

SwuniNom InhoAcc meetPstDcl

Swuni met Inho neutral

It is known that the nominative case marker ga in Japanese cannot b e used in this context

Kuno Haig That is ga cannot mark anyNew element ie neither topic nor tail

CHAPTER ORDER AND MORPHOLOGY KOREAN

b Swunika Inhonun mannassta

SwuniNom InhoTop meetPstDcl

Swuni met Inho but not othersSwuni met at least Inho

contrastivefocus

Just like the nominative case marker ka the accusative case marker do es not itself

function as an information enco der thus letting the ob ject phrase in a b e infor

mationally neutral in the default p osition In contrast when nun is attached to

the ob ject phrase as in b the phrase acquires a sp ecial status This time the

nunphrase is interpreted as a contrastivefocusAsshown in the translations b

implies that Swuni met Inho but she did not meet other p eople or Swuni met

Inho at least but we do not know whether she met other p eople to o Namelythe

entity referred to bythe nunphrase is implicitly compared with other p otential

alternatives and thus gains prominence relative to other entities see chapter for

the dierence b etween completivefocusandcontrastive fo cus

Not only an ob ject phrase but also a sub ject phrase is interpreted only as a

contrastive fo cus Let us examine the existential construction b elow

a chayksangwiey chayki issta

deskon b o okNom b eDcl

There isare a b o oks on the desk

b chayksangwiey chaykun issta

deskon b o okTop b eDcl

There isare a b o oks on the desk but nothing else contrastive fo cus

In the existential construction as shown in the base p osition of the sub ject is

not sentenceinitial it follows the lo cativephrase When the sub ject is marked with

See Kuno for detailed arguments for the base order of the Japanese existential construc

tion The same arguments hold for Korean

CHAPTER ORDER AND MORPHOLOGY KOREAN

nun in this p osition it always has a contrastive reading

As a contrastive fo cus therefore a nunphrase cannot app ear in a context which

requires for instance a completive fo cus A presentational construction is suchan

example Consider the examples b elow

a yeysnaley han mauley Swunilanun aika salassta

pastin one village Swuninamed childNom livePstDcl

Once up on a time there lived a child named Swuni in a village

b yeysnaley han mauley Swunilanun ainun salassta

pastin one village Swuninamed childTop livePstDcl

Once up on a time there lived a child named Swuni but not others

in a village

The main function of a presentational construction is to intro duce a new entityinto

the discourse Hence it is exp ected that the entitywhichistobeintro duced is purely

new information which is classied as completive fo cus in the current framework

A nunphrase which is informationally more loaded since it expresses contrastiveness

and thus carries prominence as well as newness is very awkward in this context

as shown in b

nun as a Prom marker

Wehave observed ab ove that nun in Korean sometimes expresses the topichood of a

phrase and sometimes enco des the contrastive fo cality of it The fact that a single

morpheme nun marks twovery dierent functions ie topic and contrastivefocus

has b een a source of controversy in Korean literature It has even b een suggested

due to the lackofany satisfactory explanation that there are actually two distinct

but homophonous morphemes one for topic and another for contrastive fo cus See

CHAPTER ORDER AND MORPHOLOGY KOREAN

Choi for a literature review on the debate on the double functions of nun See

also Lee Lee and Kim

Not surprisinglyhowever nun marking in Korean is not the only case where topic

and contrastive fo cus are enco ded by a single mechanism Recall the information

enco ding by topicalization in English discussed in section in chapter The

socalled topicalized phrase is often interpreted as a contrastive fo cus as well as a

topic Consider the examples again which are rep eated from chapter

a The play John sawyesterday

b FIDO they named their dog

Both the play in a and Fido in b are equally topicalized to the sentenceinitial

p osition from their canonical p ositions which are immediately after the verb However

the play in a is interpreted as a topic ie prominent old information whichthe

phrases that follow comment on and provide new information ab out In contrast

Fido in b is new and moreover prominent information see the discussion in

section while the rest is given information

The featurebased information structure prop osed in this thesis provides a straight

forward answer to this puzzle It is in fact designed to capture the oftennoted fact

that topic and contrastive fo cus are somehow enco ded by a single mechanism eg by

a sp ecial morpheme in Korean and by a syntactic op eration in English Recall the

crossclassication of the four information typ es rep eated b elowin

Prom Prom

New Topic Tail

New ContrastiveFocus CompletiveFo cus

As illustrated in topic and contrastive fo cus share the prominence prop erty

CHAPTER ORDER AND MORPHOLOGY KOREAN

although they dier in that the former expresses old information while the latter

expresses new information Both are Prom in this system

Therefore if we assume that the ma jor function of the socalled topic marker

is in fact to mark prominence of the phrase it is attached to ie nun in Korean

is a morphological enco der of prominence then it naturally follows that it will

always express topic or contrastive fo cus Moreover why nun would not express any

other typ e of information is also explained Since the prominence marking is an

inherent morphological prop erty nun would never enco de tail or completivefocus

which are nonprominentProm by denition Recall the examples in

and in the previous section Therefore the information structure system based

on the crossclassifying features as in naturally gives a uniform analysis to the dual

function of nun That is nun in Korean is a Prom marker

nun Prom

Accordinglya nunphrase will b e interpreted as topic or contrastive fo cus dep ending

on the value of the feature New whichwillbegiven by other mechanisms of the

grammar if New then topic if New then contrastive fo cus In the next section

I will examine what determines the value for the feature New so as to systematically

dierentiate the two readings

Information Structuring by Order

Reading Distribution and Relative Order

Our initial observation of the distribution of the two readings enco ded by nun in sec

tion suggests that the informational interpretation of a phrase can b e also aected

by the phrase structural conguration Recall that a sub ject nunphrase enco des a

topic reading while an ob ject nunphrase enco des a contrastive fo cus reading The

CHAPTER ORDER AND MORPHOLOGY KOREAN

relevant examples are rep eated in It is immediately clear however that the

grammatical function eg sub ject ob ject etc is not the crucial factor b ecause a

sub ject nunphrase can also receive a contrastive fo cus reading as shown in

a Swuninun Inholul mannassta

SwuniTop InhoAcc meetPstDcl

As for Swuni she met Inho topic

b Swunika Inhonun mannassta

SwuniNom InhoTop meetPstDcl

Swuni met Inho but not others contrastivefocus

chayksangwiey chaykun issta

deskon b o okTop b eDcl

There isare a b o oks on the desk but nothing else contrastivefocus

The phrase structural conguration of elements can b e interpreted in two dierent

ways ie a sp ecic p osition that the element in question is lo cated in or its relative

order whichmay also come from phrase structural conguration as argued in chapter

with resp ect to other elements of the sentence In fact there have b een a couple of

prop osals along the rst line of approach For example following the HeimDiesing

typ e of view of the syntaxsemantics mapping Han in press argues that a VP

internal nunphrase has a contrastive reading while a VPexternal nunphrase has a

topic reading Kim also makes a similar prop osal

There is evidence however whichshows that it is not a sp ecic p osition that

an elementisinwhich determines its information status First of all an element

which is lo cated in the same phrase structural p osition can b e interpreted dierently

Consider the sub ject nunphrase in b

CHAPTER ORDER AND MORPHOLOGY KOREAN

a chinkwutul cwung nwuka Inholul mannassni

friendPl among whoNom InhoAcc meetPstInt

Who among his friends met Inho

b Swuninun ama Inholul mannassul keya

SwuniTop probably InhoAcc meetPstDcl

Swuni but not others probably met Inho

At least Swuni probably met Inho contrastivefocus

As illustrated in a when the context demands it the sub ject nunphrase in its

sentenceinitial p osition can b e interpreted as a contrastive fo cus as shown in b

One could argue that the sentenceinitial p osition can b e ambiguously either VP

external or VPinternal and that the sub ject nunphrase is actually within VP eg

in Sp ecVP in b while that in a is outside VPHowever even though we

put the sentential adverb ama probably after the sub ject to mark the VP b oundary

Diesing the nunphrase still receives a contrastive fo cus reading

Moreover when the ob ject phrase scrambles over the sub ject nunphrase as in

the sub ject is unambiguously interpreted as a contrastive fo cus The topic

interpretation is not available in this case

Inholul Swuninun ama mannassul keya

InhoAcc SwuniTop probably meetPstDcl

As for Inho probably Swuni but not others met him contrastive fo cus

This example shows that it is the relative order among the elements ie b etween

the sub ject and the ob ject here which determines the informational status of the

elements It should b e noted that the sub ject nunphrase Swuninun in receives

acontrastive fo cus reading and loses its topic reading not b ecause it is lo cated in a

dierent phrase structural p osition but b ecause some other element of the sentence

CHAPTER ORDER AND MORPHOLOGY KOREAN

ie the ob ject Inholulchanges its lo cation in the sentence so that the relative order

between them is reserved

A similar eect is noted in the doublesub ject construction The rst sub ject in

a double or multiplesub ject construction is often argued to b e VPexternal Kim

a Seoulun salami manhta

SeoulTop p ersonNom manyDcl

As for Seoul it has many p eople topic

b salami Seoulun manhta

p eopleNom SeoulTop manyDcl

Sp eaking of p eople Seoul but not others places has many of them

contrastivefocus

The external nunphrase Seoulun in a as exp ected receives a topic reading

However when the internal nominative phrase salami scrambles over it the nun

phrase loses its topic reading and has a contrastive fo cus interpretation

Ihave prop osed in earlier chapters that scrambling is a ma jor information en

co der in German and is motivated and constrained bytwo information structuring

constraints which are rep eated in

Information Structuring Constraints

a newANew element should precede a New element

b prom A Prom element should precede a Prom element

It app ears that Korean also makes use of constituent order as an information enco ding

mechanism in addition to the morphological marking I will argue in the remainder

of this section that the same information structuring constraints new and prom

in are the ma jor drivingforces for constituent reordering in Korean to o It is

CHAPTER ORDER AND MORPHOLOGY KOREAN

an interesting result which shows the universality of the constraints prop osed here

although German and Korean are typ ologically unrelated languages their word order

is constrained by the same set of constraints

Scrambling motivated by Oldness

The fact that nun is a Prom marker provides us with an interesting device for

testing the motivation of scrambling That is as argued in the previous section a

nunmarked phrase always yields a topic reading or a contrastive fo cus reading and

nothing else ie a phrase marked with nun will always carry the Prom feature

with it Supp ose that wehavetwo nunphrases in a sentence Then it can b e inferred

that scrambling of a nunphrase over the other is not motivated by prominence since

b oth of the nunphrases do not dier in their Prom prop erty they need not switch

their p ositions to adjust their relative order in accordance with the prom constraint

Given the information structuring constraints in wecanhyp othesize that

scrambling of a nunphrase is motivated bythenew constraint In other words

scrambling is prompted to allowaNew element to precede a New element

This hyp othesis is well supp orted by the following data Consider the examples in

a Swuninun ama Inhonun mannassul keya

SwuniTop probably InhoTop meetPstDcl

As for Swuni she probably met Inho but not others

Swunitopic Inhocontrastive fo cus

b Inhonun Swuninun ama mannassul keya

InhoTop SwuniTop probably meetPstDcl

As for Inho Swuni but not others probably met him

Swunicontrastive fo cus Inhotopic

CHAPTER ORDER AND MORPHOLOGY KOREAN

Supp ose that b oth a and b describ e events where Swuni met Inho that is

Swuni is the sub ject and Inho is the ob ject since b oth phrases are marked with

nun their grammatical functions are not dierentiated by the morphological case

marking Unlike the insitu ob ject nunphrase in a the scrambled ob ject Inho

nun in b cannot b e interpreted as a contrastive fo cus it must b e interpreted as

a topic On the other hand the sub ject nunphrase Swuninun in b whichis

scrambled over by the ob ject phrase can only b e interpreted as a contrastivefocus

not as a topic This is also in contrast with the insitu sub ject phrase in a In

other words the scrambled ob ject Inhonun is NewProm whereas the the left

over sub ject phrase Swuninun is NewProm Since b oth the sub ject and the

ob ject are Prom inherently as argued ab ove what these examples show is that the

nunphrases are aligned according to their newness in the scrambled example b

That is the scrambling in b is triggered bythenew constraint

Similarly the sub ject nunphrase chaykun b o ok in the existential construction

also acquires a topic interpretation when it is scrambled over the lo cative phrase while

the leftover lo cative phrase tosekwaneynun in the library receives a contrastive

fo cus reading This is shown in

a tosekwaneynun chaykun issta

libraryinTop bookTop b eDcl

In the library there are at least b o oks but mayb e not other things

librarytopic b o oktopic

b chaykun tosekwaneynun issta

b o okTop libraryinTop b eDcl

As for b o oks they are at least in the library but mayb e not in other places

librarycontrastive focus booktopic

Here again since b oth the elements share the Prom feature the scrambling in

b can b e regarded as b eing motivated by their dierence in the New feature

CHAPTER ORDER AND MORPHOLOGY KOREAN

the New element scrambles over the New element

To recapitulate as dened in the current information structure based on the New

and Prom features the only dierence b etween topic and contrastive fo cus is that

topic is New and contrastive fo cus is New they are identical in terms of the

Prom feature ie b oth are Prom Hence the fact that the scrambled nunphrase

in b and in b has a topic reading while the nunphrase which is scrambled

over receives a contrastive fo cus reading shows that scrambling has rearranged the

elements of the sentence in accordance to the new constraint which requires a New

element to precede a New element In other words the new constraint is fairly

active in Korean as well as in German as a constraint on the constituent order I will

show in the next section that the prom constraint is also eective in this language

Scrambling motivated by Prominence

It is wellknown that Korean along with Chinese and Japanese do es not have wh

movement That is a whphrase in Korean stays in situ in syntax whereas that in

English or German has to b e prep osed to the sentenceinitial p osition eg Sp ecCP

with the exception of the echoquestion case or the multiple whquestion construction

So as illustrated b elow leaving a whphrase in situ in English for instance as in

a is bad while it is p erfectly go o d in Korean as shown in a

a Did Mary meet whom

b Who did Mary meet

a Swunika nwukwulul mannassni

SwuniNom whoAcc meetPstInt Who did Swuni meet

CHAPTER ORDER AND MORPHOLOGY KOREAN

b nwukwulul Swunika mannassni

whoAcc SwuniNom meetPstInt

Who did Swuni meet

What is interesting is that a whphrase in Korean whichdoesnothave to b e prep osed

for any syntactic reason can actually b e scrambled This is shown in b This

indicates that whphrases in Korean are no more restricted in their scramblability than

any other regular phrases and that their p osition in a sentence is also determined by

their information status

In addition to nunphrase scrambling scrambling of a whphrase serves as a re

vealing test to show the motivation for scrambling A whphrase can b e argued to

inherently carry the New feature b ecause it always expresses new information

regardless of its p osition in a sentence The whphrase nwukwulul in b oth a and

b solicits new information as to who Swuni met for example Therefore wecan

predict that its scrambling is not triggered bythenew constraint Moreover the

new constraint is supp osed to discourage scrambling of a New element it would

place a New element in front of a New element

Thus we can nowhyp othesize that scrambling of a whphrase is motivated by

the prom constraint namely it is prompted to b e placed in a noncanonical p osition

to realize its Prom feature This predicts that a scrambled whphrase should

be interpreted as a contrastive fo cus ie NewProm rather than simply as a

completive fo cus ie NewProm This prediction is b orne out Let us return to

the examples in The scrambled sentence in b is indeed interpreted slightly

dierently from the canonicallyordered sentence in a The question b implies

that there is a set of p eople that Swuni mighthave met and asks WHO among them

While the scrambling of whphrases is as free as that of other phrases it seems to b e fairly restricted and often imp ossible in German See Moltmann for a brief discussion of the facts

CHAPTER ORDER AND MORPHOLOGY KOREAN

Swuni actually met That is the p erson who Swuni actually met receives prominence

in distinction to other p eople she mighthave met So b can b e b etter translated

as Who is it among the p eople that Swuni met

Likewise the scrambling of a nominativecase phrase in the existential construc

tion shows a similar eect The whphrase in b solicits prominent new informa

tion

a chayksangwiey mwuesi issni

deskon whatNom b eInt

What is there on the desk

b mwuesi chayksangwiey issni

what deskon b eInt

What is it that is on the desk

That the scrambled question phrase expresses contrastive fo cus rather than comple

tive fo cus is supp orted by the examples in

a mwuesi chayksangwiey issni

what deskon b eInt

What is it that is on the desk

b chayksangwiey chayki isse

deskon b o okNom b eDcl

There is a b o ok on the desk

b chayki chayksangwiey isse

b o okNom deskon b eDcl It is a b o ok that is on the desk

CHAPTER ORDER AND MORPHOLOGY KOREAN

The question in a requests prominent and new information as argued ab ove

b is an existential sentence in the default conguration where the nominative

phrase is p ositioned after the lo cative phrase As is usually the case an existential

construction presents purely new information in the discourse so the chayk b o ok

in b in the default p osition is interpreted as completive fo cus Therefore b

is not appropriate as an answer to the question in a b ecause while the question

requests a contrastive fo cus the answer provides a completivefocusthus causing an

informational clash In contrast the scrambled phrase chayk in b is interpreted

as a contrastive fo cus thanks to scrambling just as the whphrase in a is Hence

only the scrambled sentence b can b e a prop er answer to the scrambled question

in a

Ambiguity Resolution byFurther Constraints

Wehavenow established that scrambling in Korean is motivated bythenew con

straintortheprom constraint Although I have presented examples whichinvolve

either a nunphrase or a whphrase in order to control the informational environment

scrambling of course is not limited to these typ es of phrases In other words when

a regular phrase is scrambled it is exp ected that ambiguity arises as to whether it

is scrambled to realize its oldness or to realize its prominence I will showinthis

subsection howambiguity is resolved by the constraints wehave already develop ed

in the previous chapters

Proso dic Constraints

Scrambling of an ob ject for example yields a couple of dierentinterpretations Consider the examples in

CHAPTER ORDER AND MORPHOLOGY KOREAN

a Swunika Inholul mannassta

SwuniNom InhoAcc meetPstDcl

Swuni met Inho

b Inholul Swunika mannassta

InhoAcc SwuniNom meetPstDcl

As for Inho Swuni met him topictail

It is Inho among other p eople who Swuni met contrastive fo cus

The ob ject phrase Inho with the regular case marking as in a is informationally

neutral in the sense that it can b e interpreted as old or new or prominent or non

prominent information dep ending on the context and thus on the information status

of the other elements of the sentence Given no other context the insitu phrase is

usually interpreted as a completive fo cus providing new information to the discourse

When the ob ject phrase is scrambled as in b however it clearly partitions

the sentence into two parts Moreover it partitions the sentence in twovery dierent

ways Firstly the scrambled ob ject can b e interpreted as old New information

Prom or Prom dep ending on the information status of the sub ject and then

the rest of the sentence b ecomes new New information Secondly it can b e in

terpreted as new information actually prominent new information ie contrastive

fo cus otherwise it would not have scrambled in the rst place and in that case the

rest of the sentence is marked as old information Roughly sp eaking scrambling can

partition a sentence either into a TopicComment structure or intoaFo cusGround

structure

Notably though it is never interpreted as a completive fo cus NewProm whichisthe

most common reading for the insitu phrase This result is exactly what is exp ected given the

two information structuring constraints new and prom If the scrambled phrase is interpreted as

new information rather than old information it follows that the phrase is necessarily Prom

otherwise the scrambling would not b e prompted at all

CHAPTER ORDER AND MORPHOLOGY KOREAN

The ambiguity in information structuring is fortunately resolved by dierent

proso dic patterns Not surprisinglyspeakers often note that a high pitch accent

is used when the scrambled phrase is interpreted as a contrastive fo cus but not

necessarily so when it is interpreted as old information This contrast is shown in

a Inholul Swunika mannassta

InhoAcc SwuniNom meetPstDcl

As for Inho Swuni met him yesterday topic

b INHOLUL Swunika mannassta

InhoAcc SwuniNom meetPstDcl

It is Inho among other p eople who Swuni met contrastive fo cus

A similar proso dic contrast is noted in the interpretation of a nunphrase in the

identical conguration Namelyonlywhena nunphrase is interpreted as new in

formation ie as a contrastive fo cus do es it acquire a high pitch accent This is

illustrated in

a Swuninun Inholul mannassta

SwuniTop InhoAcc meetPstDcl

As for Swuni she met Inho topic

Prince makes a similar observation regarding the proso dic dierences b etween a Topical

ized sentence and a Fo cusmoved or Yiddishmoved sentence

If the scrambled elementisinterpreted as prominent old information ie a topic rather than

nonprominent information ie a tail then it receives a certain degree of pitch accent though it is

muchweaker than that which a new element receives It may b e the case that prominence as well

as newness triggers high pitch accents However for simplicity of discussion I will assume that only

a New elementhasapitch accent

CHAPTER ORDER AND MORPHOLOGY KOREAN

b SWUNINUN Inholul mannassta

SwuniTop InhoAcc meetPstDcl

Swuni but not others met Inho

At least Swuni met Inho contrastivefocus

This fact leads some linguists like Cho e to argue that the topic marker nun in

a and the contrastive marker nun in b are in fact two separate morphemes

the topic marker is nun and the contrastive fo cus marker is NUN

Now recall the proso dic constraintN whichwas intro duced in chapter as part

of the information structuring constraints It requires that new information havea

high pitch accent It was argued for German in chapter that this constraintis

ranked higher than the economy typ e of constraintX so that it will always place a

pitch accent on fo cused elements b oth completive and contrastive The constraints

and the ranking b etween them are listed b elow

Proso dic Constraints

a N Put a high pitch accent on a newinformation element

b X Do not place anypitch accent

Ranking

N X

Supp ose that Korean has the same ranking b etween these two proso dic constraints as

in Then it predicts that the scrambled phrase in will receive a high pitch

accent only when it is interpreted as a contrastive fo cus but not as old information

Similarly the interpretation of the insitu nunphrase in will b e dierentiated in

a parallel way Therefore it is not necessary to assume two separate morphemes for

nun

CHAPTER ORDER AND MORPHOLOGY KOREAN

Hence the ranking b etween the two proso dic constraints in further disam

biguates the underdetermined interpretation of a scrambled phrase which has not

b een fully distinguished by the phrase structural conguration

specificity Constraints

Scrambling of a bare noun in Korean provides a more interesting case Nouns in

Korean can b e multiply ambiguous b ecause morphological sp ecication of deniteness

as well as for plurality is optional in this language Therefore a bare noun chayk

b o ok for example can mean several dierent things as shown in

chayk a b o ok the b o ok b o oks the b o oks

Putting aside the plurality problem for simplicity of discussion I will concentrate on

the deniteness problem

An insitu bare noun chayk book can be interpreted either as a b o ok or the

b o ok as shown in

Swunika ecey chaykul ilkessta

SwuniNom b o okAcc yesterday readPstDcl

Swuni read a book yesterday Nonsp ecic

Swuni read the book yesterday Sp ecic

However if it is scrambled as in a esp ecially when it do es not have a pitch

accent it is uniformly interpreted as the b o ok Lee Interestingly when it is

pitchaccented as in b the ambiguity remains

a chaykul Swunika ecey ilkessta

b o okAcc SwuniNom yesterday readPstDcl

The novel Swuni read yesterday

CHAPTER ORDER AND MORPHOLOGY KOREAN

b CHAYKUL Swunika ecey ilkessta

b o okAcc SwuniNom yesterday readPstDcl

It is athe b o ok that Swuni read yesterday

Given the previous discussion ab out the proso dic constraints this implies that the

scrambled ob ject in a is New while that in b is New Furthermore it

shows that when the bare noun chaykul is interpreted as a New entity it receives

a sp ecic reading while when it is interpreted as a New entity it can receive either

a sp ecic or a nonsp ecic reading

Likewise the scrambled kaphrase in a without apitch accentisinterpreted as

a sp ecic b o ok whereas that in b with apitch accent is not necessarily interpreted

as a sp ecic b o ok

chayksangwiey chayki issta

deskon b o okNom b eDcl

There is a b o ok on the desk

a chayki chayksangwiey issta

b o okNom deskon beDcl

The b o ok is on the desk

b CHAYKI chayksangwiey issta

b o okNom deskon b eDcl

It is athe b o ok that is on the desk

This also indicates that when a scrambled phrase is interpreted as a New entity

it necessarily receives a sp ecic reading

This fact reminds us of the specificity constraints whichhave b een discussed in

earlier chapters as constraints whichcontrol the discourse feature assignmentinthe

CHAPTER ORDER AND MORPHOLOGY KOREAN

input Recall that the rst constraint ab out sp ecic phrases can b e easily violated

whereas the second constraint cannot

specificity

a sp A sp ecic phrase should not b e New

b sp A nonsp ecic phrase should not b e New

Therefore in eect that a nonsp ecic phrase should not b e marked New since

b is a highly ranked constraint although a sp ecic phrase do es not havesuch

a restriction ie it can b e marked either New or New since a is easily

violable To put it in dierentwords a New element can b e either sp ecic or

nonsp ecic while a New element is necessarily sp ecic So going backtothe

examples ab ove the scrambled phrase in a and in a cannot b e interpreted as

a nonsp ecic entity b ecause it is New whereas the scrambled phrase in b and

in b is not constrained as such b ecause it is New

Therefore if we assume that the second constraint sp is also an active constraint

also in Korean we can nicely capture the sp ecicity distribution of the scrambled

phrases illustrated ab ove It is interesting to note that in Korean where the mor

phological marking of sp ecicitydeniteness is very minimal the sp ecicityinter

pretation of an undersp ecied phrase heavily relies on its information status This

further supp orts the strong link b etween semantic sp ecicity and the discourse notion

of oldness

OptimalityTheoretic Interaction b etween Mor

phology and Order

Scrambling involving a nunphrase is an interesting case of information enco ding

b ecause part of its information status is determined by the morphological marking

CHAPTER ORDER AND MORPHOLOGY KOREAN

and the other part is determined by the phrase structural conguration As I have

argued ab ove nun is a Prom marker which carries the feature Prom as part of

its morphological prop erty Whether it is interpreted as New or New is then

determined by its relative p osition with resp ect to other elements of the sentence

Constraints and Ranking

Although not explicitly mentioned it has b een the underlying assumption for Korean

to o that there is a certain default phrase structural conguration for a sentence

which yields the canonical word order among its elements and scrambling has b een

viewed as a mechanism to reorder them This canonical order is the one whichisused

in the neutral context ie when no contextual information is available or when all

the elements are presented equally in their information status eg as new information

by default cf neutral description in Kuno or allfo cus in Vallduv

As argued in chapter Korean also has a congurational structure where the sub ject

is generated as the most structurally prominent element then the indirect ob ject and

nally the direct ob ject in a ditransitive clause Just as in German this canonical

structure will b e chosen as the most optimal one by the phrase structural constraints

canon

a cn SUBJ should b e structurally more prominent than eg ccommand

nonSUBJ functions

b cn NonSUBJ functions align reversely with the cstructure according

to the functional hierarchy

Accordingly these phrase structural constraints canon will generate the following

order as default unless the discoursecontext demands otherwise

CHAPTER ORDER AND MORPHOLOGY KOREAN

Swunika Inhoeykey chaykul cwuessta

SwuniNom InhoDat b o okAcc givePstDcl

Swuni gave Inho a b o ok

If a sentence is used in a certain discourse context and the constituents of the sentence

are dierentiated as such in their information status then the information structuring

constraints suchasnew and prom in will b e involved and interact with canon

to yield the optimal word order in that sp ecic context

Information Structuring Constraints

a newANew element should precede a New element

b prom A Prom element should precede a Prom element

In chapter I have argued that these constraints are ranked as follows in German

Constraint Ranking German

new

prom cn

It was argued particularly that the sub contraints of canon split in German and that

cn the sub ject constraint is ranked higher than cn the nonsub ject constraint to

capture the fact that the sub ject is harder to scramble over than nonsub ject elements

are Furthermore cn was argued to b e ranked b etween the prom constraintand

the new constraint whichshows that only a Prom element can scramble over the

sub ject while a New element cannot

However in Korean the sub ject do es not app ear to b ehave distinctly from other

elements of the sentence It is as easily scrambled over Consider the examples in

and b elow Let us rst lo ok at the ordering b etween the nonsub ject elements in

CHAPTER ORDER AND MORPHOLOGY KOREAN

a Swunika nwukwueykey nonmwunul cwuessni

SwuniNom whoDat thesisAcc givePstInt

To whom did Swuni give her thesis

b Swunika nonmwunul Inhoeykey cwuessta

SwuniNom thesisAcc InhoDat givePstDcl

Swuni gave her thesis to Inho

b Swunika Inhoeykey nonmwunul cwuessta

SwuniNom InhoDat thesisAcc givePstDcl

Swuni gave Inho her thesis

If the context requires that the indirect ob ject is presented as New and the direct

ob ject is presented as New as shown ab ove then the default order of the indirect

ob ject followed by the direct ob ject as in b violates the new constraint This

triggers the New direct ob ject nonmwunul thesis to scramble over the New

indirect ob ject Inhoeykey to Inho which generates the output in b The exam

ple b shows that the new constraintisranked higher than the nonsub ject phrase

structural constraint cn b ecause the satisfaction of new yields a go o d sentence in

spite of the violation of cn

Ranking

new cn

However the default ordered sentence in b is also p ossible in this context which

indicates that the default order is selected despite the undesirable order b etween the

two ob ject phrases in terms of newness In other words a sentence can b e go o d when

cn is satised and the new constraint is violated This result gives the opp osite

ranking b etween new and cn

CHAPTER ORDER AND MORPHOLOGY KOREAN

Ranking

cn new

Therefore we can conclude that new and cn in Korean are unranked with resp ect

to each other just as is the case in German

Now consider the sub ject case in In this context the sub ject is New and

the indirect ob ject is New There are no other contextual clues that would signal

either element as b eing Prom So the prom constraint do es not play a role here

a nwuka Inhoeykey chaykul cwuessni

whoNom InhoDat b o okAcc givePstInt

Who gave Inho a b o okDid anybody give Inho a book

b Inhoeykey Swunika chaykul cwuessta

InhoDat SwuniNom b o okAcc givePstDcl

Swuni gave Inho a b o ok

b Swunika Inhoeykey chaykul cwuessta

SwuniNom InhoDat b o okAcc givePstDcl

Swuni gave Inho a b o ok

Unlike in the German counterpart simple oldness can trigger the indirect ob jects

scrambling over the sub ject in Korean The New indirect ob ject Inhoeykey to

Inho scrambles over the New sub ject Swunika Swuni in b Just likethe

nonsub ject case the default order is also p ossible which shows that new and cn

are not ranked

Ranking

a cn new

b new cn

CHAPTER ORDER AND MORPHOLOGY KOREAN

Therefore we can conclude that the phrase structural sub constraints cn and cn

in Korean do not split as two separate ranked constraints Hence I will refer to

them collectively as canon from nowon canon is unranked with resp ect to new

in Korean

Finally the examples in showthattheprom constraintisranked the highest

in Korean as well as in German If the context demands a contrastive fo cus as in

then the contrastive fo cus Swunikawhich is NewProm is ordered b efore

the tail InhoeykeywhichisNewProm as shown in b The other order is

not p ossible as illustrated in b This shows that a violation of prom as in b

is more serious than the violation of new in b resulting in the ranking in

a chinkwutul cwung nwuka Inhoeykey chaykul cwuessni

friendPl among whoNom InhoDat b o okAcc givePstInt

Who among his friends gave Inho a book

b Swunika Inhoeykey chaykul cwuessta

SwuniNom InhoDat b o okAcc givePstDcl

Swuni gave Inho a b o ok

b Inhoeykey Swunika chaykul cwuessta

InhoDat SwuniNom b o okAcc givePstDcl

Swuni gave Inho a book

Ranking

prom new

Thus is the nal ranking b etween the information structuring constraints new

and prom and the phrase structural constraint canon

Constraint Ranking Korean

CHAPTER ORDER AND MORPHOLOGY KOREAN

new

prom

canon

Cho osing b etween Topic and ContrastiveFocus

Having identied the constraints involved in information enco ding in Korean and the

ranking among them let me now explain the reading distribution of a nunphrase in

terms of Optimality Theory

Let us b egin with the sentence in which the sub ject phrase is marked with nun as

Swuninun Inholul mannassta

SwuniTop InhoAcc meetPstDcl

Swuni met Inho

Recall the discussion in the preceding sections that the sub ject nunphrase can b e

interpreted as a topic or a contrastive fo cus in the canonical conguration whereas

when it is scrambled over by another element of the sentence it loses its topic reading

and only receives a contrastive fo cus interpretation

a Swuninun Inholul mannassta

SwuniTop InhoAcc meetPstDcl

As for Swuni she met Inho topic

Swuni but not others met Inho contrastivefocus

b Inholul Swuninun mannassta

InhoAcc SwuniTop meetPstDcl

Inho Swuni but not others met contrastivefocus

Now let me showhow the OT analysis can capture this generalization ab out the readings

CHAPTER ORDER AND MORPHOLOGY KOREAN

Even without anycontextual clues we know that the sub ject Swuni will b e inter

preted either as a topic or a contrastive fo cus b ecause nun automatically marks Swuni

as Prom in the input as in b ecause Prom is its invariable morphological

prop erty The rest of the discourse features will b e lled in from the context

pred mannaxy

Swuni

subj

Prom

obj

Inho

tense past

The candidates are all the scrambled variants of the defaultorder sentence Since the

sentence in question contains two scramblable elements ie the sub ject Swuninun

and the ob ject Inholulwehavetwo candidate outputs And the ranking among the

constraints discussed in the previous subsection will generate the tableau in for

each input

candidates prom new canon

a Swuninun Inholul

b Inholul Swuninun

Now let us rst supp ose that the sub ject Swuni is the given information ie

New and the ob ject Inho is new information New This automatically endows

the sub ject Swuni with topicho o d b ecause the morphology nun has already given it

the Prom feature Given the ranking among the constraints we can predict that

Swuninun as a topic would never b e scrambled over by Inholul b ecause b oth of its

features ie New and Prom urge it to precede other elements The example

CHAPTER ORDER AND MORPHOLOGY KOREAN

in shows that the ob ject phrase cannot scramble over the sub ject topic phrase

even when it is also interpreted as Prom its New feature makes it lose to the

topic phrase

a SwuninunSwunik a chinkwutul cwung nwukwulul mannassni

SwuniTop SwuniNom friendPl among whoAcc meetPstInt

What ab out SwuniWho among her friends did Swuni meet

b Inholul Swuninun mannassta

InhoAcc SwuniTop meetPstDcl

Swuni met Inho

Therefore regardless of the information status of the ob ject Inholul whether it is a

completive fo cus or a contrastive fo cus the sub ject will precede the ob ject This is

demonstrated in the two tableaux in and

Swuninun Inholul

NewProm NewProm

Topic CompFoc

candidates prom new canon

a Swuninun Inholul

b Inholul Swuninun

Swuninun Inholul

NewProm NewProm

Topic ContFoc

CHAPTER ORDER AND MORPHOLOGY KOREAN

candidates prom new canon

a Swuninun Inholul

b Inholul Swuninun

On the other hand if the sub ject Swuninun is new information or fo cus then

Swuni will have the New marking thus b eing a contrastive fo cus In this case the

only thing which can precede a contrastive fo cus is a topic b ecause a topic is not only

Prom but also New Therefore if Inho is simply a tail as in then Inho

cannot precede SwunisoSwuni will have a contrastive fo cus reading in the initial

p osition as shown in

a chinkwutul cwung nwuka Inholul mannassni

friendPl among whoNom InhoAcc meetPstInt

Who among his friends met Inho

b Swuninun Inholul mannassta

SwuniTop InhoAcc meetPstDcl

At least Swuni met Inho contrastivefocus

b Inholul Swuninun mannassta

InhoAcc SwuniTop meetPstDcl

Swuni met Inho

The tableau in exactly predicts this situation

Swuninun Inholul

NewProm NewProm

ContFoc Tail

CHAPTER ORDER AND MORPHOLOGY KOREAN

candidates prom new canon

a Swuninun Inholul

b Inholul Swuninun

Because the prom constraint is ranked higher than the new constraint in this lan

guage a tail cannot scramble over a contrastive fo cus

In contrast if Inho is also prominentandthus b ecomes a topic then Inho can

precede SwuninunsoSwuni will have a contrastive fo cus reading in the leftover

p osition as shown in b

a Inhonunc hinkwutul cwung nwuka Inholul mannassni

InhoTop friendPl among whoNom InhoAcc meetPstInt

What ab out InhoWho among his friends met Inho

b Inholul Swuninun mannassta

InhoAcc SwuniTop meetPstDcl

As for Inho Swuni but not others met him contrastive fo cus

Note that the tableau actually gives the two outputs in this context b ecause new

and canon are tied and thus b oth rankings of the constraints are p ossible Still

this is the only context in which the ob ject Inholul can scramble over the sub ject

Actually candidate a is not very go o d in this context This may b e b ecause of a general

principle that a topic should b e predicated of a contentful enough comment or the predicate

should b e contrastive enough See Principle of Contrastiveness in de Ho op See also the

discussion in Reinhart In a the comment to the topic Inho is the simple verb mannassta

it is to o weak to supp ort the topic in a sense Consider the examples in i

i a

pinun onta

rainTop comePrsDcl

It is raining b

CHAPTER ORDER AND MORPHOLOGY KOREAN

Swuninun In other words if a nunphrase is scrambled over by another element it

necessarily receives a contrastive fo cus reading

Swuninun Inholul

NewProm NewProm

ContFoc Topic

candidates prom new canon

a Swuninun Inholul

Inholul Swuninun b

Overall the tableaux ab ove tell us that a sub ject topic phrase cannot b e scrambled

over In other words if a sub ject nunphrase is scrambled over it must b e interpreted

asacontrastive fo cus This is exactly what we initially observed in That is

the scrambled sentence b only receives a contrastive fo cus reading while the de

fault sentence can havebothinterpretations Therefore the OT analysis successfully

captures the reading distribution of the sub ject nun sentence

Now let us take the case of the ob ject nunphrase as the next example The

comp etition in this case lies b etween the two orders in a and b

a Swunika Inhonun mannassta

SwuniNom InhoTop meetPstDcl

Swuni met Inho

pinun ekswukathi onta

rainTop heavily comePrsDcl

It is raining cats and dogs

Interestingly the sub ject nunphrase in ia with the simple predicate to come down is almost

exclusively interpreted as a contrastive fo cus while the same nunphrase in ib with the more

contrastive predicate to come down heavily can b e interpreted as a topic

CHAPTER ORDER AND MORPHOLOGY KOREAN

b Inhonun Swunika mannassta

InhoTop SwuniNom meetPstDcl

Swuni met Inho

As is the case with the sub ject nunphrase the ob ject Inhonun also carries the

Prom feature as its morphological prop ertyasshown in the input in

pred mannaxy

Swuni

Inho

Prom

tense past

First consider the cases in which the sub ject phrase is interpreted as old infor

mation and the ob ject nun phrase is thus interpreted as new information namely

as a contrastive fo cus Since the ob ject phrase is now NewProm its scrambla

bility dep ends on the prominence of the sub ject phrase if the sub ject is Prom

ie if the sub ject is a tail then the ob ject nunphrase can scramble over it This is

illustrated in

a chinkwutul cwung nwukwulul Swunika mannassni

friendPl among whoAcc SwuniNom meetPstInt

Who among her friends did Swuni meet

b Inhonun Swunika mannassta

InhoTop SwuniNom meetPstDcl

Swuni met Inho but not others contrastivefocus

CHAPTER ORDER AND MORPHOLOGY KOREAN

Although the ob ject Inhonun is New which prevents it from scrambling its sup e

riorityover the sub ject in terms of prominence makes the scrambling p ossible Recall

the prom constraint is ranked higher than the new constraint and also than canon

Swunika Inhonun

NewProm NewProm

Tail ContFoc

candidates prom new canon

a Swunika Inhonun

b Inhonun Swunika

In contrast if the sub ject Swunika is Prom as well and thus receives a topic

reading then scrambling of the ob ject is prohibited as exp ected

a SwuninunSwunik a chinkwutul cwung nwukwulul mannassni

SwuniTop SwuniNom friendPl among whoAcc meetPstInt

What ab out SwuniWho among her friends did Swuni meet

b Inhonun Swunika mannassta

InhoTop SwuniNom meetPstDcl

Swuni met Inho

The ob ject nunphrase is not distinct from the sub ject phrase in terms of prominence

ie b oth are Prom and moreover the new constraint and canon encourages it

to remain in situ

Swunika Inhonun

NewProm NewProm

Topic ContFoc

CHAPTER ORDER AND MORPHOLOGY KOREAN

candidates prom new canon

a Swunika Inhonun

b Inhonun Swunika

On the other hand if the ob ject Inhonun is old information ie New and

thus is interpreted as a topic then it can always scramble as exp ected

a Inhonunn wuka Inholul mannassni

InhoTop whoNom InhoAcc meetPstInt

What ab out InhoWho met Inho

b Inhonun Swunika mannassta

InhoTop SwuniNom meetPstDcl

As for Inho Swuni met him topic

This is illustrated in the tableaux in and

Swunika Inhonun

NewProm NewProm

CompFoc Topic

candidates prom new canon

a Swunika Inhonun

b Inhonun Swunika

Swunika Inhonun

NewProm NewProm

ContFoc Topic

CHAPTER ORDER AND MORPHOLOGY KOREAN

candidates prom new canon

a Swunika Inhonun

Inhonun Swunika b

Therefore the OT account illustrated ab ove oers the desirable result that a non

sub ject nunphrase can always scramble over the sub ject if it is interpreted as a topic

but it mayormay not scramble dep ending on the Prom status of the sub ject if it

is interpreted as a contrastivefocus

Summary

To summarize the constituent order in Korean is also inuenced by the information

status of eachelement of the sentence and therefore constrained by the same set of

constraints whichgovern German word order ie the information structuring con

straints new and prominteracting with the phrase structural constraints canon

In this chapter I have illustrated that a slight dierence in ranking of the two sub

constraints of canon with the assumption that nun is a morphological enco der of

prominence nicely explains the scrambling data involving nunphrases in Korean

Constraint Ranking Korean

new

prom

canon

Together with the analysis of the German data demonstrated in chapter the

discussion in this chapter further supp orts the OT analysis of word order based on

The candidate a is not very felicitous in this context Again this is the case in which the

topic is not supp orted bya contentful enough predicate See the previous fo otnote

CHAPTER ORDER AND MORPHOLOGY KOREAN

the notion of candidate comp etition of the dierentlyordered sentences and the in

teraction of violable constraints from b oth syntax and discourse The OT analysis

prop osed here can provide a systematic accountof variation and optionality in the or

dering phenomena which alternative analyses based only on one mo dule of grammar

have had diculty with

Chapter

Conclusion

This dissertation has investigated scrambling phenomena in German and Korean from

the p ersp ective that the dierent ordering p ossibilities are motivated and constrained

byinteractions of syntactic and discoursepragmatic constraints of these languages

In particular I have utilized Optimality Theory to demonstrate how these constraints

from dierent mo dules of the grammar interact and resolve conicts among one an

other to yield the optimal output a sentence with a particular word order in a

given context

I rst summarize the conclusions reached from this research I then discuss some

issues raised in this dissertation which remain as areas for future research

Summary and Concluding Remarks

Despite the various word order p ossibilities in these languages a certain order is

regarded as the unmarked or default one Under the assumption that a surface word

order is a reection of a phrase structure I examine the basic clause structures in

German and Korean and conclude that b oth have a congurational S structure ie

Sub ject IOb ject DOb jectVchapter I prop ose this as the canonical

S VP

CHAPTER CONCLUSION

structure for these languages whichiscontextneutral and thus is solely determined

by grammatical information The canonical structure is dened to b e the result

of the mapping b etween the grammatical information in argumentstructure and

functionalstructure and its phrase structural realization in constituentstructure I

argue that this mapping is controlled by a set of OT constraints called canon

Positing a congurational canonical structure in relatively free word order lan

guages requires an explanation for the range of alternativeword orders b ecause the

canonical structure only yields a subset of the surface word orders I prop ose that

there is another set of constraints which require word order to b e determined by dis

course information such as newness and prominence of each element I identify

these information structuring constraints as new and prom which I argue to b e

the ma jor motivation for alternative orders or scrambling chapter As a means

of capturing the generalizations b ehind the various semantic and discourserelated

eects asso ciated with scrambling I prop ose an information structure based on two

crossclassifying discourse features New and Prom I also claim that the assign

ment of these features in the input is further restrained by the semantic prop erties

of an element eg sp ecicity and prop ose the specificity constraints which handle

the internal mismatchbetween semantic sp ecicity and discourse oldness Therefore

the sp ecicity eect is not treated as a direct syntaxsemantics mapping but is

subsumed under the general syntaxdiscourse interface

Since the phrase structural constraints and information structuring constraints

try to align the elements of a sentence based on dierenttyp es of information ie

syntactic and discoursal resp ectively they are in p otential conict I prop ose that

scrambling arises when the information structuring constraints acting against the

phrase structural constraints which require that eachelement of the sentence b e in

its canonical p osition allow an argument to b e out of its canonical p osition if it

meets certain informational requirements Using Optimality Theorythisinteraction

CHAPTER CONCLUSION

or conict resolution among the dierent constraints is carried out by the ranking

among them I prop ose the following rankings of the ab ove constraints namely the

grammars for German and Korean resp ectively chapter and chapter

Constraint Ranking German

new

prom cn

Constraint Ranking Korean

new

prom

canon

The two languages are almost identical in their word order p ossibilities and only

dier in one asp ect the sub ject eect ie the sub ject is harder to scramble over

than the other arguments are This dierence is handled with a slightchange in

the rankings of the constraints in each language That is in German the canon

constraints split into two and the sub ject constraint cn is ranked higher than the

nonsub ject constraint cn The fact that the sub ject constraint is separated from the

other phrase structural constraint and is actually ranked b etween the two information

structuring constraints prom and new in German reects an interesting prop ertyof

German sub jects the sub ject in German precedes all nonprominent elements only

a prominentelement can precede the sub ject This gives the sub ject more chances

to b e interpreted as the topic thus making it b ehave as if it had dual roles one as a

grammatical function and the other as a discourse function

As such the OT analysis prop osed here can provide a systematic accountof

variation within the ordering phenomena Variation here is not presented as free

CHAPTER CONCLUSION

nor as a result of some optional op eration Eachscrambled variant is viewed as a

necessary and also the most appropriate structure for instantiating all the relevant

information in the given context Thus the current OT analysis explains why and how

variation o ccurs First each order is the optimal structural description for the given

grammatical and discoursal information in the input since the discoursal information

of a sentence changes dep ending on the context dierent orders arise Also each

variant is an output which satises the constraints in the b est p ossible way Since the

constraints are ranked according to their relativeweight or imp ortance given the

input the output is the automatic result of constraint comp etition

The observation that discourse information such as fo cus and topic reanalyzed

as the combinations of the New and Prom features in the current analysis as well

as syntactic information plays a role in determining word order has a long history

cf the Prague Scho ol In this sense the main contribution of this dissertation may

not lie in that it has uncovered new principles or constraints on scrambling which

have never b een discussed b efore but lies rather in that it has shown an example of

how those constraints interact with each other to yield the optimal solution This

is a desirable consequence which conrms the core ideas in Optimality Theory the

OT constraints are prop osed to b e universal although ranked dierently in dierent

languages and thus the fact that some constraints have precedents in linguistic theory

further supp orts their universality

Moreover since the constraints not only within a single mo dule but also ones

from dierent mo dules of grammar ie syntax and discourse are in comp etition

the current OT framework provides a mo del of interface theory cf Reinhart

In other words interface is viewed here as the simultaneous comp etition of the con

straints from several dierent sectors of the grammar This interface approachis

easily captured by the framework of LFG Since LFG is organized in terms of par

allel copresent structures not in terms of one structure derived from another the

CHAPTER CONCLUSION

simultaneous constraint comp etition can b e conceived of as a natural op erational

mechanism in this framework In particular adding the informationstructure to

the already existing framework in an explicit fashion provides a way to unite the

formal asp ects of language with the functional asp ects which is necessary for the

explanation of word order phenomena illustrates the intermo dule interactions

involved in scrambling

argumentstructure functionalstructure

canon

constituentstructure

new prom

semanticstructure informationstructure

specificity

The syntactic inuence on word order is realized in the form of the canon constraints

whichgovern the mapping b etween from argumentstructure and the functional

structure to the constituent structure On the other hand the discoursal eect

on word order is instantiated as the mapping b etween the informationstructure

and the constituentstructure governed bythenew and prom constraints Finally

the semantic inuence on scrambling is indirectly accomplished via its control over

the informationstructure by means of the specificity constraints It may also

b e the case that there is a direct mapping b etween the semanticstructure and

the constituentstructure whichmay p ossibly involve the cases of scop e eects

realized in word order Although I have not pursued this issue in this dissertation

the current framework do es not prevent it from b eing incorp orated as yet another

CHAPTER CONCLUSION

factor or constraintonword order phenomena

Remaining Issues

In this section I briey discuss several issues raised in this dissertation whichneed

further research

First the information structure istructure should b e further investigated eg

howitisinternally structured and how it as an indep endent grammatical represen

tation relates to other parts of grammar

pred geb enxyz

Hans

New

Prom

dem Schuler

New

Prom

das Buch

New

Prom

tense Past

Although I have simply added the discourse features New and Prom to the functional

structure as an input representation as exemplied here in the information

structure should b e thought of as an indep endent grammatical representation Firstly

as briey mentioned in chapter the discourse feature assignment can b e further re

strained bysyntactic morphological and semantic factors These interactions with

CHAPTER CONCLUSION

other grammatical information cannot b e eciently captured if the discourse infor

mation is conned to the fstructure Furthermore the fstructure representation is

noncongurational and unordered However the fo cus pro jection phenomena and

also the nonsp ecic NP problem may require a more hierarchical organization of in

formation structure This hierarchical organization mayalsobedierent from the

cstructural hierarchy

Next the notion of candidates in OT should b e further rened I have assumed

in chapter that the comp eting candidates in syntax are those that share the skeletal

fstructure roughly sp eaking the candidate sentences have more or less the same

meaning and contain identical lexical items However lexical choice itself is gov

erned by information structural eects Consider the following examples In this

dissertation I have only considered cases in which the candidates have the xed lexi

cal choice so that the answers are given in the full sentence form and no noun phrases

are replaced by the pronouns as in b However b with the old information

replaced by a pronoun is certainly a go o d and often preferred answer in this context

a Was hat Hans gekauft

whatAcc has Hans b ought

What did Hans buy

b Ich glaub e da Hans ein Buch gekauft hat

I b elieve that Hans a b o okAcc b oughthas

I b elieve that Hans b ought a book

b Ich glaub e da er ein Buch gekauft hat

I b elieve that heNom a b o okAcc b oughthas

I b elieve that he b ought a b o ok

Korean p oses a variant of this problem of lexical choice it can drop any phrases

which can b e understo o d from the context Consider the sub jectdrop case in b

CHAPTER CONCLUSION

a Swunika nwukulul mannassni

SwuniNom whoAcc meetPstInt

Who did Swuni meet

b Swunika Inholul mannassta

SwuniNom InhoAcc meetPstDcl

Swuni met Inho

b Inholul mannassta

InhoAcc meetPstDcl

She met Inho

The problem of pronominalization and prodrop must b e taken into account in a full

study of the optimization of lexical choices in context

Finally let me briey mention the p otential applicability of the current mo del to

statistical and corpusbased work on word order Hob erg The currentanal

ysis makes an interesting prediction ab out the realworld frequency distribution of

the scrambled variants The information structure theory prop osed in this disserta

tion pro duces a nite numb er of p ossible informational environments or discourse

contexts given the limited number of elements in a sentence Each element can b e

assigned New and Prom and the combinations of these feature assign

ments yield a nite numb er of discoursecontexts each of which is represented in an

OT tableau in the current theoryEach tableau in turn yields one or more if there

are tied constraints optimal output which dier dep ending on the input Counting

the number of eachvariants o ccurrence as an optimal output we can predict which

variant will app ear the most frequently and which one the least

This prediction is under the assumption that each discoursecontext has an equal chance of

o ccurrence which in realitymay not b e true As such some sort of statistical normalization would

b e required to derive more accurate quantitative predictions

CHAPTER CONCLUSION

The quantitative implication of the current theory also oers an interesting sp ec

ulation on the theory of markedness Nativespeakers usually rep ort that one variant

sounds b etter than another These intuitions are indep endently conrmed by large

corp ora where the preferred or unmarked variant is usually the more frequentone

cf Anttila Ihave dened the unmarked or canonical structure to b e the

one which o ccurs in the neutral context in which the one which is solely determined

by the grammatical information is the optimal one However even though the neu

tral context is only one of the manycontexts that the current information structure

pro duces the canonical structure is chosen as the optimal output in more than the

neutral context and is actually the most frequent optimal choice in the OT tableaux

given the ranking of the constraints Of course the choice of the unmarked output

will still b e indirectly aected by the grammatical information b ecause the canon

constraints are part of the determining factor Yet still the unmarked variantasthe

most frequent optimal choice incorp orates the discoursecontextual considerations as

well and thus reects the languageuse factor in its denition

Bibliography

Abraham Werner Word order in the middle eld of the German sentence In

Werner Abraham and Sjaak de Meij eds Topic Focus and Congurationality

Ahn HeeDon Light verbs VPmovement negation and clausal architecture

in Korean and English Do ctoral dissertation University of WisconsinMadison

Ahn HeeDon and HangJin Yoon Functional categories in Korean In Susumu

Kuno IkHwan Lee John Whitman SungYun Bak YoungSe Kang and Young

jo o Kim eds Harvard Studies in Korean Linguistics III Seoul Korea

Hanshin

Aissen Judith Topic and fo cus in Mayan Language

Alsina Alex Predicate Composition A Theory of Syntactic Function Alterna

tions Do ctoral dissertation Stanford University

Alsina Alex and Sam Mchombo Ob ject asymmetries and the Chichewa

applicative construction In Sam Mchomb o ed Theoretical Aspects of Bantu

Grammar Stanford CA CSLI Publications

Anttila Arto Deriving variation from grammar To app ear in Hinskens F

R van Hout and L Wetzels eds Variation Change and Phonological Theory

Benjamins

Austin Peter in press Word order in a free word order language the case of Jiwarli

Language

Austin Peter and Joan Bresnan to app ear Noncongurationality in Australian

ab original languages Natural Languages Linguistic Theory

Bibliography

Bakovic Eric J A markedness subhierarchy in syntax Presented at the MIT

Workshop on OptimalityinSyntactic Theory

Bratt Elizab eth Owen Argument Composition and the Lexicon Lexical and

Periphrastic Causatives in Korean Do ctoral dissertation Stanford University

Bresnan Joan ed The Mental Representation of Grammatical Relations

Cambridge MA MIT Press

Bresnan Joan Exercises and problem sets Lexical functional grammar ms

Stanford University

Bresnan Joan a Lexicalfunctional syntax ms Stanford University

Bresnan Joan b Linear order syntactic rank and empty categories On weak

crossover In Mary Dalrymple Ronald M Kaplan John T Maxwell and Annie

Zaenen eds Formal Issues in LexicalFunctional Grammar Stanford

CA CSLI Publications

Bresnan Joan c Morphology comp etes with syntax Explaining typ ological

variation in weak crossover eects Presented at the MIT Workshop on Optimality

in Syntax

Bresnan Joan Optimal syntax Notes on pro jection heads and optimality

ms Stanford University

Bresnan Joan and Jonni Kanerva Lo cativeinversion in Chichewa A case

study of factorization in grammar Linguistic Inquiry

Bresnan Joan and Ron Kaplan LexicalFunctional Grammar A formal sys

tem for grammatical representation In Joan Bresnan ed The Mental Repre

sentation of Grammatical Relations Cambridge MA MIT Press

Bibliography

Bresnan Joan Ron Kaplan and Peter Peterson Co ordination and the owof

information through phrase structure ms Stanford University and XeroxPARC

Palo Alto

Bresnan Joan and Sam A Mchomb o The lexical integrity principle Evidence

from Bantu Natural Language Linguistic Theory

Bresnan Joan and Lioba Moshi Ob ject asymmetries in comparativeBantu

syntax Linguistic Inquiry

Bresnan Joan and Annie Zaenen Deep unaccusativityinLFG In Proceed

ings of the Fifth Biennial ConferenceonGrammatical Relations Stanford CSLI

Publications

Broihier Kevin Optimality theoretic rankings with tied constraints Slavic

relatives resumptive pronouns and learnability ms Department of Brain and

Cognitive Sciences MIT

Chafe Wallace Givenness contrastiveness deniteness sub jects topics and

point of view In Charles Li ed Subject and Topic New York Academic

Press

Cho JaeOhk and Jerry Morgan The interaction of syntax and morphology in

Korean VP co ordination In Susumu Kuno John Whitman IkHwan Lee Sung

Yun Bak and YoungSe Kang eds Harvard Studies in Korean Linguistics II

Seoul Korea Hanshin

Cho Youngmee Yu Syntax and phrasing in Korean In Sharon Inkelas and

Draga Zec eds The PhonologySyntax Connection Stanford CA CSLI Publications

Bibliography

Cho Youngmee Yu and Peter Sells A lexical account of inectional suxes

in Korean Journal of East Asian Linguistics

Cho e HyonBae Uli Malpon Our Grammar Seoul Korea Chungumsa

Cho e Hyon So ok Remarks on congurationality parameters In Susumu

Kuno John Whitman IkHwan Lee and YoungSe Kang eds Harvard Studies

in Korean Linguistics I Seoul Korea Hanshin

Cho e Hyon So ok Fo cus and topic movement in Korean and licensing In

Katalin E Kiss ed Discourse Congurational Languages Oxford Ox

ford University Press

Choi HyeWon Weak crossover in scrambling languages Precedence rank and

discourse ms Presented at the Annual Meeting of the Linguistic So cietyof

America New Orleans

Choi HyeWon Disp ensing with functional categories in Korean A reply to

Yo on ms Stanford University

Choi HyeWon in press Topic and fo cus in Korean The information partition by

phrase structure and morphologyInJapaneseKorean Linguistics

Choi So oYoung Cwucehwawacwukyek cosa Cosa nun kwaka lul

cwungsimulo Topicalization and sub ject particles Fo cusing on the particles nun

and ka Language Research

Chomsky Noam Deep structure surface structure and semantic interpreta

tion In D Steinb erg and L Jacob ovits eds Semantics Cambridge

Cambridge University Press

Chomsky Noam BarriersCambridge MA MIT Press

Bibliography

Chomsky Noam Some notes on economy of derivation and representation

In Rob ert Freidin ed Principles and Parameters in Comparative Grammar

Cambridge MA MIT Press

Chomsky Noam A minimalist program for linguistic theory MIT Occasional

Pap ers in Linguistics No

Chomsky Noam A minimalist program for linguistic theory In Kenneth Hale

and Samuel Jay Keyser eds The View from Building Essays in Linguistics

in Honor of Sylvain Bromberger Cambridge MA MIT Press

Chung Chan Korean auxiliary verb constructions without VP no des In

Susumu Kuno John Whitman YoungSe Kang IkHwan Lee Joan Maling and

Youngjo o Kim eds Harvard Studies in Korean Linguistics V Seoul

Korea Hanshin

Chung Sandra and James McCloskey Government barriers and small clauses

in Mo dern Irish Linguistic Inquiry

Dahl Osten Topiccomment structure revisited In Osten Dahl ed Topic

and Comment Contextual Boundedness and Focus Papers in Text Linguistics

Hamburg Helmut Buske

Dalrymple Mary J Lamping and V Saraswat LFG semantics via constraints

XeroxPARC Palo Alto CA

de Ho op Helen Case Conguration and Noun Phrase Interpretation Do ctoral

dissertation University of Groningen

den Besten Hans A case lter In A Belleti L Brandi and L Rizzi eds

Theory of Markedness in Generative Grammar Scuola Normale Sup eriore di Pisa

Bibliography

den Besten Hans Some remarks on the ergativehyp othesis In GAGL

den Besten Hans The ergativehyp othesis and free word order in Dutchand

German In J Toman ed Issues in the Grammar of German Dordrecht Foris

Diesing Molly Verb movement and the sub ject p osition in Yiddish Natural

Language and Linguistic Theory

Diesing Molly Indenites Cambridge MA MIT Press

Diesing Molly Semantic variables and ob ject shift ms Presented at the th

Comparative Germanic Syntax Workshop Harvard

Dik Simon et al On the typ ology of fo cus phenomena In Teun Ho ekstra

Harry van der Hulst and Michael Mo ortgat eds Perspectives on Functional

Grammar Dordrecht Holland Foris

Enc Murv et The semantics of sp ecicity Linguistic Inquiry

Firbas Jan On dening the theme in functional sentence analysis In Fran

tisek Danes ed Travaux Linguistiques de PragueVol Universityof

Alabama Press

Givon Talmy Mind Code and Context Essays in Pragmatics Hillsdale NJ

Lawrence Erlbaum Asso ciates

Grimshaw Jane Minimal pro jection heads and optimalityTechnical Rep ort

RuCCS TR Center for Cognitive Science Rutgers University Piscataway NJ

Grimshaw Jane Optimizing lexical choice Expletives and opaque clitics as

cases of minimal violation ms Rutgers University Presented at the Linguistics

Collo quium at Stanford University

Bibliography

Grimshaw Jane in press Pro jection heads and optimality Linguistic Inquiry

Grimshaw Jane and Vieri SamekLo dovici Optimal sub jects and sub ject

universals Presented at the MIT Workshop on OptimalityinSyntax

Guilfoyle Eithne Henrietta Hung and Lisa Travis Sp ec of IP and Sp ec of VP

Two sub jects in MalayoPolynesian languages Natural Language and Linguistic

Theory

Gundel Jeanette The Role of Topic and Comment in Linguistic TheoryDoc

toral dissertation UniversityofTexas at Austin

Haider Hub ert Empty categories and some dierences b etween English and

German Wiener Linguistische Gazette

Haider Hub ert Studies on phrase structure and economy msUniversityof

Stuttgart

Haig John H Japanese ga and Korean kai ms UniversityofHawaii Pre

sented at the Second International Conference on Korean Linguistics

Hale Ken Preliminary remarks on congurationality In J Pustejovsky and

Peter Sells eds Proceedings of the North Eastern Linguistics Society

Hale Ken Warlbiri and the grammar of noncongurational languages Natural

Language and Linguistic Theory

HallidayMichael A K Notes on transitivity and theme in English Journal

of Linguistics

Halvorsen PerKristian Semantics for LexicalFunctional Grammar Linguistic

Inquiry

Bibliography

Halvorsen PerKristian and Ronald M Kaplan Pro jections and semantic

description in LexicalFunctional Grammar In Proceedings of the International

Conference on Fifth Generation Computer Systems FGCS Tokyo

Institute for New Generation Systems

Han Chunghye in press A syntactic account of the ambiguity of the topic marker

nun in Korean In JapaneseKorean Linguistics

Han Eunjo o Honorication in Korean ms Stanford University

Hawkins John A A Performance Theory of Order and Constituency Cam

bridge Cambridge University Press

Heim Irene The Semantics of Denite and Indenite Noun Phrases Do ctoral

dissertation University of Massachussetts Amherst

Herring Susan C Information structure as a consequence of word order typ e

In Proceedings of the Berkeley Linguistics SocietyVol

Hob erg U Die Wortstel lung in der geschriebenen deutschen Gegen

wartsspracheMunchen Hueb er

Holmb erg Anders WordOrder and Syntactic Features in Scandinavian Lan

guages and English Do ctoral dissertation UniversityofStockholm

Horn Larry On the Semantic Properties of Logical Operators in EnglishDoc

toral dissertation UCLA

Inkelas Sharon Prosodic Constituency in the Lexicon Do ctoral dissertation

Stanford University

Jackendo Ray Semantic Interpretation in Generative GrammarCambridge MA MIT Press

Bibliography

Jo MiJeung Fixed WordOrder and the TheoryofthePreVerbal Focus Po

sition in Korean Do ctoral dissertation UniversityofWashington Seattle WA

Jonas Dianne and Jonathan David Bobaljik Sp ecs for sub jects The role of

TP in Icelandic In Jonathan David Bobaljik and Collin Philips eds Case and

Agreement IVol MIT Working Pap ers in Linguistics

Kanerva Jonni Morphological integrity and syntax The evidence from Finnish

p ossessive suxes Language

Kaplan Ronald M and Joan Bresnan LexicalFunctional Grammar A for

mal system for grammatical representation In Joan Bresnan ed The Mental

Representation of Grammatical Relations Cambridge MA MIT Press

Kaplan Ronald M and John T Maxwell Constituent co ordination in Lexical

Functional Grammar In Mary Dalrymple Ronald M Kaplan John T Maxwell

and Annie Zaenen eds Formal Issues in LexicalFunctional Grammar

CSLI Publications

Keenan Edward and Bernard Comrie Noun phrase accessibilityanduniversal

grammar Linguistic Inquiry

Kim Alan HyunOk The Grammar of Focus in Korean Syntax and its Typo

logical Implication Do ctoral dissertation University of Southern California

Kim JongBok The Grammar of Negation ALexicalist ConstraintBased

Perspective Do ctoral dissertation Stanford University

Kim Kwangsup Where do the constrastive and fo cus readings come fromIn

Ha jime Ho ji ed JapaneseKorean Linguistics CSLI Publications

Bibliography

King Tracy Holloway Conguring Topic and Focus in Russian Do ctoral

dissertation Stanford University Published by CSLI Publications

KiparskyPaul Lexical Morphology and Phonology In Linguistic So ciety

of Korea ed Linguistics in the Morning Calm Seoul Korea Hanshin

KiparskyPaul Some consequences of Lexical PhonologyInCEwen and J

Anderson eds Phonology Year BookVol Cambridge Cambridge

University Press

KiparskyPaul Morphology and grammatical relations ms Stanford Univer

sity

Kiss Katalin E Structural relations in Hungarian Linguistic Inquiry

Kitagawa Y Subject in Japanese and English Do ctoral dissertation University

of Massachusetts at Amherst

Ko opman Hilda and Dominique Sp ortiche The p osition of sub jects Lingua

Krifka Manfred Scop e inversion under the risefall contour in German ms

UniversityofTexsas Austin and Stanford University

Kro eger Paul Phrase StructureandGrammatical Relations in Tagalog Stan

ford CA CSLI Publications

Kuno Susumu Functional sentence p ersp ective A case study from Japanese

and English Linguistic Inquiry

Kuno Susumu The Structure of the Japanese LanguageCambridge MA MIT Press

Bibliography

Kuno Susumu The scop e of the question and negation in some verbnal

languages In Papers from the th Regional Meeting of the Chicago Linguistics

Society

Kuro da ShigeYuki Whether we agree or not A comparativesyntax of English

and Japanese Lingvuisticae Investigationes

Lambrecht Knud Information Structure and SentenceForm Cambridge

Cambridge University Press

Lap ointe Steven Two analyses of Korean verb inections In Proceedings of

the Eastern States Conference on Linguistics

Lap ointe Steven Korean verb markers and Autolexical TheoryInProceedings

of FLSM UniversityofMichigan Working Pap ers in Linguistics

Lee Chungmin Indenites case markers classiers and quantiers in Ko

rean In Susumu Kuno IkHwan Lee John Whitman SungYun Bak YoungSe

Kang and Youngjo o Kim eds Harvard Studies in Korean Linguistics III

Seoul Korea Hanshin

Lee Hyuno o The Korean topic construction and the semantics of bare NPs

ms University of California Los Angeles

Lee YoungSuk Scrambling as CaseDriven Obigatory Movement Do ctoral

dissertation UniversityofPennsylvania

Legendre Geraldine William Raymond and Paul Smolensky An Optimality

Theoretic typ ology of case and grammatical voice systems In Proceedingsofthe

Berkeley Linguistics SocietyVol

Bibliography

Legendre Geraldine Paul Smolensky and Colin Wilson a When is less more

Faithfulness and minimal links in whchains Presented at the MIT Workshop on

OptimalityinSyntax

Legendre Geraldine Colin Wilson Paul Smolensky Kristin Homer and William

Raymond b Optimalityand whextraction In Jill N Beckman Laura Walsh

Dickey and Suzanne Urbanczyk eds Papers in Optimality Theory

Amherst MA GLSA University of Masachusetts

Lenerz Jurgen Zur Abfolge nominaler Satzglieder im DeutschenTubingen

TBL Verlag Gunter Narr

Li Charles and Sandra Thompson Sub ject and topic A new typ ology of lan

guage In Charles Li ed Subject and Topic New York NY Academic

Press

Maha jan Ano op K The AAbar Distinction and Movement Theory Do ctoral

dissertation MIT

McCloskey James Clause structure ellipsis and prop er government in Irish

Lingua

Milsark Gary Toward an explanation of certain p eculiarities of the existential

construction in English Linguistic Analysis

Miyagawa Shigeru Noncongurationality within a congurational structure

msMIT

Miyagawa Shigeru to app ear Against optional scrambling Linguistic Inquiry

Mohanan K P The Theory of Lexical Phonology Dordrecht Reidel

Bibliography

Mohanan K P and Tara Mohanan Issues in word order in South Asian

languages Enriched phrase structure or multidimensionality In Miriam Butt

Tracy Holloway King and Gillian Ramchand eds Theoretical Perspectives on

WordOrder in South Asian Languages Stanford CA CSLI Publica

tions

Mohanan Tara Arguments in Hindi Do ctoral dissertation Stanford Univer

sity Published by CSLI Publications

Moltmann Friederike Scrambling in German and the sp ecicity eect ms

MIT

Neeleman Ad Scrambling as a Dstructure phenomenon In Norb ert Corver

and Henk van Riemsdijk eds Studies on Scrambling Movement and Non

Movement Approaches to FreeWordOrder Phenomena Berlin Mouton

de Gruyter

Nesp or Marina and Irene Vogel Prosodic Phonology Dordrecht Foris

Payne Doris L Information structuring in Papago narrative discourse Lan

guage

PesetskyDavid Whinsitu Movement and unselective binding In Eric J

Reuland and Alice G B ter Meulen eds The Representation of Indeniteness

MIT Press

PesetskyDavid Class notes MIT

Platzack Christer COMP INFL and Germanic word order In Lars Hellan

and K K Christensen eds Topics in Scandinavian Syntax Dordrecht

Foris

Bibliography

Pollard Carl and Ivan A Sag HeadDriven Phrase StructureGrammar

Chicago University of Chicago Press

Pollard Carl J On head nonmovement ms CarnegieMellon University

Pollo ck JeanYves Verb movement universal grammar and the structure of

IP Linguistic Inquiry

Poser William Cliticization to NP and Lexical PhonologyInProceedings of

West Coast ConferenceinFormal Linguistics CSLI Publications

Prince Alan and Paul Smolensky to app ear Optimality Theory Constraint Inter

action in Generative Grammar Cambridge MA MIT Press

Prince Ellen F Topicalization fo cusmovement and Yiddishmovement In

Proceedings of the Berkeley Linguistics SocietyVol

Prince Ellen F On the syntactic marking of presupp osed op en prop ositions In

Papers from the th Regional Meeting of the Chicago Linguistics Society Parases

sion Papers

Prince Ellen F The ZPG letter Sub jects deniteness and informationstatus

In Sandra Thompson and William Mann eds Discourse Description Diverse

Analyses of a Fundraising Text Amsterdam John Benjamins

Reinhart Tanya Pragmatics and linguistics An analysis of sentence topics

ms Distributed by the Indiana University Linguistics Club

Reinhart Tanya Interface strategies Technical Rep ort OTSWPTL

Research Institute for Language and Sp eech Utrecht University Utrecht the Netherlands

Bibliography

Reinholtz Charlotte Verb second in Mainland Scandinavian A reanalysis In

Aaron Halp ern ed Proceedings of the th West Coast ConferenceonFormal

Linguistics

Reuland Eric Indenite sub jects In Proceedings of the North Eastern Lin

guistics Society GLSA

Ro chemont Michael S Focus in Generative Grammar Amsterdam John

Benjamins

Ro chemont Michael S and Peter W Culicover English Focus Constructions

and the Theory of GrammarCambridge Cambridge University Press

Rognvaldsson Eirkur and Hoskuldur Thrainsson On Icelandic word order

once more In Joan Maling and Annie Zaenen eds Syntax and Semantics

Modern Icelandic Syntax Academic Press

Ro oth Mats E Association with Focus Do ctoral dissertation Universityof

Massachusetts Amherst

Ro oth Mats E A theory of fo cus interpretation Natural Language Semantics

Sado ck Jerrold Noun incorp oration in Greenlandic A case of wordformation

Language

Sado ck Jerrold Autolexical Syntax A Theory of Paral lel Grammatical Repre

sentations Chicago University of Chicago Press

Selkirk Elisab eth O Phonology and SyntaxCambridge MA MIT Press

Sells Peter Lectures on Contemporary Syntactic Theories Stanford CSLI Publications

Bibliography

Sells Peter Korean and Japanese morphology from a lexical p ersp ective

Linguistic Inquiry

Sgall Petr EvaHajicova and Jarmila Panevova The Meaning of the Sentence

in its Semantic and Pragmatic Aspects Prague Academia

SmolenskyPaul On the internal structure of the constraints comp onent Con

of UG Handout of a talk given at UCLA April

Sp eas Margaret Adjunctions and Projections in Syntax Do ctoral dissertation

MIT

Steedman Mark J Structure and intonation Language

Szab olcsi Anna The semantics of topicfo cus articulation In T Jannsen and

M Stokho j eds Formal Methods in the Study of Language Amster

dam MathematischCentrum

Tesar Bruce a Computational Optimality Theory Do ctoral dissertation Uni

versity of Colorado at Boulder

Tesar Bruce b Errordriven learning in OT via the ecient computation of

optimal forms Presented at the MIT Workshop on Optimality in Syntax

Thiersch C A note on scrambling and the existence of VP Wiener Linguis

tische Gazette

Uszkoreit Hans Linear precedence in discontinuous constituents Complex

fronting in German In Georey J Huck and Almerindo E Ojeda eds Syntax

and Semantics Discontinuous Constituency Academic Press

Vallduv Enric The Informational ComponentNewYork Garland

Bibliography

Vallduv Enric Information packaging A survey ms Centre for Cognitive

Science and Human Communication ResearchCentre University of Edinburgh

Vikner Sten Verb Movement and the Licensing of NPPositions in the Ger

manic Languages Do ctoral dissertation University of Geneva

Vikner Sten Verb Movement and Expletive Subjects in the Germanic Lan

guages Oxford University Press

Ward Gregory L The Semantics and Pragmatics of Preposing Do ctoral

dissertation UniversityofPennsylvania

Ward Gregory L and Betty B Birner Deniteness and the English existential

Language

Web elhuth Gert German is congurational Linguistic Review

Web elhuth Gert Syntactic Saturation Phenomena and the Modern Germanic

languages Do ctoral dissertation University of Massachusetts Amherst

Web elhuth Gert Principles and Parameters of Syntactic Saturation New

York NY Oxford University Press

Web elhuth Gert and Hans den Besten Remnant topicalization and the con

sitituent structure of the Germanic SOVlanguages Presented at GLOW Venice

Italy

Whitman John Topic mo dality and IP structure In Susumu Kuno IkHwan

Lee John Whitman SungYun Bak YoungSe Kang and Youngjo o Kim eds

Harvard Studies in Korean Linguistics III Seoul Korea Hanshin

Williams Edwin Predication Linguistic Inquiry

Bibliography

Yo on James HyeSuk Tense co ordination and the clausal structures of English

and Korean In Susumu Kuno IkHwan Lee John Whitman Joan Maling Young

Se Kang and Youngjo o Kim eds Harvard Studies in Korean Linguistics V

Seoul Korea Hanshin

Yo on James HyeSuk Korean verbal inection in Checking Theory In Heidi

Harley and Colin Phillips eds The MorphologySyntax ConnectionVol

MIT Working Pap ers in Linguistics

Yo on James HyeSuk and Jeongme Yo on Morphosyntactic mismatches and

the functioncontent distinction In K Deaton M Noske and M Ziolkowski

eds Papers from the th Regional Meeting of Chicago Linguistics Society

Chicago Linguistics So ciety

Zaenen Annie Joan Maling and Hoskuldur Thrainsson Case and grammatical

functions The Icelandic passive Natural Language and Linguistic Theory

Zwicky Arnold M and Georey K Pullum Cliticization vs inection English

nt Language