sign in sign up
<<
Home , Empty category

Empty categories in LFG

Judith Berman

University of Stuttgart

Pro ceedings of the LFG Conference

University of California San Diego

Miriam Butt and Tracy Holloway King Editors

CSLI Publications

httpwwwcslistanfordedupublications

I am grateful to Christian Fortmann Gert Web elhuth Hub ert Haider Joan Bresnan Miriam Butt Steve Berman

and Werner Frey for helpful comments and discussion This researchwas supp orted by the DFG Graduiertenkolleg

Linguistische Grundlagen fur die Sprachverarb eitung at the University of Stuttgart

LFG JBerman Empty Categories in LFG

This pap er is concerned with the question whether there is any necessity and evidence for empty

categories sp ecically traces in German Following the analysis of weak crossover in Bresnan

b and Choi it is shown that the German weak crossover data can b e captured correctly

if it is assumed that a topicalized constituent has to b e linked with an in its lo cal

domain its minimal clause only in the case of long distance dep endencies The empty category is

indep endently motivated by a lo cality requirement on function sp ecication which is empirically

supp orted by the fact that free word order in German is restricted to the lo cal clause It is further

shown that the empty category cannot o ccupy the canonical p osition of the antecedent Instead it

is claimed based on work byFrey that the sp ecier p osition of the functional category is

the only p osition in which the empty category is licensed The resulting analysis not only accounts

for the weak crossover data but is also shown to capture some restrictions on extraction and have

implications for the structure of certain innitival constructions in German

Constraints on cstructure

In Bresnan b a sp ecic economy principle is prop osed to prevent the o ccurrence of unneces

sary constituents in cstructure namely economy of expression

Economy of expression all cstructure no des are optional and are not used unless required

for expressivity or completeness

To illustrate this principle Bresnan compares the topicalization structures of English and Russian

based on King which are given in

IP IP

DP IP DP IP

the old b oat Staruyu lo dku

DP I

DP VP my

we I

V DP dali

sold e

The old b oat we sold Staruyu lo dku my pro dali

old acc b oat acc we nom sold pl

While in Russian the syntactic functions are morphologically marked in English a congurational

language the functions are dened structurally So in order to identify the function of a topicalized

phrase in English it must b e linked to an empty category in the canonical p osition but in Russian

a trace would violate economy of expression since the functional information is already given by

the case

LFG JBerman Empty Categories in LFG

In German syntactic functions are also morphologically marked so we exp ect parallel to Russian

that there is no empty category in the unmarked p osition of a topicalized or scrambled constituent

This is illustrated by the topicalization structure in

FP

DP F

DF

den alten Kahn

F VP

hab en

DP VP

wir

V

verkauft

Den alten Kahn hab en wir verkauft

the old b oat acc have pl we nom sold

Following Haider simple sentences in German are analyzed as containing only one functional

pro jection called FP whichemb eds a VP In the topicalized ob ject the old b oat is lo cated

in the sp ecier p osition of the functional pro jection a p osition for discourse functions see section

Evidence for analyzing topicalization sentences like without a trace comes from the pattern of

op erator in suchsentences This is discussed in the next section

Weak crossover

The sentences in illustrate the basic facts of op erator binding in German sentences with clause

internal topicalization

a Jeder Wer mag seine Mutter

i i i

everyone nom who nom likes his mother acc

Everyone likes his motherWho likes his mother

b Jeden Wen mag seine Mutter

i i i

everyone acc who acc likes his mother nom

Everyone his mother likesWho do es his mother like

c Seine Mutter mag jeder

i i

his mother acc likes everyone nom

His mother everyone likes

Including noninsitu whphrases

LFG JBerman Empty Categories in LFG

d Seine Mutter mag jeden

i i

his mother nom likes everyone acc

His mother likes everyone

As we will see this pattern can be accounted for in terms of the analysis of op erator binding in

Bresnan b which is presented in the next subsection

Bresnan b

In Bresnan b op erator binding is restricted by the socalled prominence principle

Prominence principle A binder excludes from its domain any element more prominent

than it

The domain of a binder is informally dened as follows Bresnan

The domain of a binder is the minimal clause or predication structure containing it formally

mo delled as the minimal fstructure

At each level of representation the relation of prominence is determined by prop erties sp ecic to

that level These are given by the following denitions cf Bresnan

a outranks  if and  b elong to the same fstructure and is more prominent than 

on the functional hierarchyor outranks some  that contains 

b Syntactic rank a relation of prominence among expressed or unexpressed syntactic

arguments based on the functional hierarchy SUBJ  OBJ  OBJ  OBL  COMPL



a The linear order of and  is determined by the right edges of their lexical expres

sions reduces to simple linear order in the string whenever the corresp ondence between

arguments and the no des that express them is onetoone

b The lexical expression of an consists of any cstructure word or b ound form

that lexically carries the semantic features of the argument together with all nonterminal

phrase structure no des that also corresp ond to the argument

Thematic prominence a relation of prominence among arguments asso ciated with the same

argument structure inherited from the underlying semantic or role structure eg Dowty

Jackendo Bresnan and Zaenen Alsina

What results from these distinct notions of prominence are the following levelsp ecic instantiations of the prominence principle

LFG JBerman Empty Categories in LFG

Resulting domain constraints on pronominal binding

i The domain of excludes any  that outranks in fstructure

ii The domain of excludes any  that precedes in cstructure

iii The domain of excludes any  that is thematically more prominent than in astructure

Consequently the eect of the prominence principle may dier at each level of representation

Moreover crosslinguistic dierences in the distribution of op erator binding maybeaccounted for

by assuming that languages observe dierentcombinations of the levelsp ecic constraints Bresnan

b prop oses the following illustrativeweak crossover typ ology based on the domain constraints

at fstructure i and cstructure ii

English Palauan i ii

German Korean i v ii

Chichewa Kiswahili i

Malayalam Hindi ii

Choi

The idea that op erator binding in languages like Korean and German is constrained by the dis

junction of the fstructure and cstructure constraints is due to Choi She showed that this

directly accounts for the p ossibilities of scrambling in Korean and German her German examples

are given in

a dass jeder seine Mutter mag

i i

that everyone nom his mother acc likes

that everyone likes his mother

b dass jeden seine Mutter mag

i i

that everyone acc his mother nom likes

that his mother likes everyone

c dass seine Mutter jeder mag

i i

that his mother acc everyone nom likes

that everyone likes his mother

d dass seine Mutter jeden mag

i i

that his mother nom everyone acc likes

that his mother likes everyone

Only in d is the b ound reading of the not available and it is only in this example that the pronoun is excluded from the op erators binding domain in b oth cstructure and fstructure In

LFG JBerman Empty Categories in LFG

in b binding is excluded in fstructure but not in cstructure so the b ound reading



is correctely predicted

As a of the sentences in with those in makes clear the pattern of op erator

binding is the same in b oth scrambling and clauseinternal topicalization Thus Chois account will

also apply directly to the latter

This analysis succeeds only b ecause no trace of the scrambled or topicalized constituent is assumed

If there were a trace in the canonical p osition of the antecedent in b and b the sentences

would have the following structures

a Jeden Wen mag seine Mutter e

i i i i

b weil jeden seine Mutter e mag

i i i

In b oth of these structures the pronoun precedes the right edge of the op erator so by it precedes

the op erator and is therefore excluded from its domain in cstructure Thus with a trace the b ound



reading would b e wrongly ruled out



These results in fstructure follow only on the assumption that in i stands for not just the outer fstructure

eg corresp onding to the DP no de of the phrase seine Mutter in but also all fstructures it may contain in

particular the one corresp onding to the p ossessive pronoun seine The earlier formulation of the fstructure condition

on op erator binding given in Bresnan a is less ambiguous in this resp ect it reads The pronominal

binding domain of an op erator O is restricted to the set of fstructure elements that O outranks There is also no

ambiguity in the formulation of the cstructure constraint ii due to the nature of linear orders



That astructure also plays a role is seen by the p ossibility of backwards binding According to

FrankLeeRambow and Frey backwards binding is p ossible only for sub jects this is illustrated bythe

following sentences from FrankLeeRambow b a

i Ich glaub e dass der Jorg seinem Vater jeden gezeigt hat

i i

I think that art J nom his father dat everyone acc shown has

i i

I think that J has shown his father everyone

i i

ii Ich glaub e dass der Jorg seinen Vater jedem gezeigt hat

i i

I think that art J nom his father acc everyone dat shown has

i i

I think that J has shown everyone his father

i i

To account for this Bresnan b suggested that in anysentence in German the domain constraints must b e

satised in at least twoofthethreelevels of representation in i the constraint is satised only at fstructure and

in ii only at astructure assuming the thematic hierarchy  b eneciary  exp eriencergoal  instrument 

theme  lo cative in Bresnan and Zaenen However Choi rep orts iii as an acceptable example

of backwards binding from the dative ob ject to the accusative ob ject and I share this judgement

iii dass man seine Tanzpartnerin jedem vorstellte

i i

that they his dance partner acc everyone dat intro duced

that they intro duced to everyone his dance partner

To account for this it might b e prop osed that op erator binding in German is constrained by the disjunction of the

astructure and cstructure constraints at least in the dialects of those sp eakers who accept these data However

since this issue is not directly relevant to the concerns of this pap er for simplicity I ignore such data as iiii and

thus put aside the role of astructure

LFG JBerman Empty Categories in LFG

Weak crossover involving long distance dep endencies

In examples like whichinvolve long distance dep endencies the topicalized op erator here is the

direct ob ject of the emb edded verb the pronoun cannot b e understo o d as b ound by the op erator

Jeden Wen sagte seine Mutter hab e sie getrostet

i i i

everyone acc who acc said his mother nom has she consoled

Everyone his mother said she consoledWho did his mother say she consoled

i i i i

The analysis presented ab ove fails to account for this Although the binding condition is not satis

ed in fstructure since the pronoun is contained in an element that outranks the op erator still

the op erator precedes the pronoun and according to Chois analysis that is sucient to sanction

binding

Bresnan b discusses an analogous example in Hindi She observes that on the assumption

that function sp ecication by morphology is only lo cally accessible there must be an empty ca

tegory inside the lo cal clause in order to assign the correct argument relation between

the op erator and the emb edded verb Note that since this empty category is required in order to

satisfy completeness it do es not violate economy of expression Applying this prop osal to German

must b e analyzed as having a structure like the following

Jeden Wen sagte seine Mutter hab e sie e getrostet

i i i i

Here the right edge of the op erator follows the pronoun so by the denition of linear order the

pronoun precedes the op erator and so is excluded from its binding domain Thus the presence of

the trace correctly predicts the absence of the b ound reading in

One piece of evidence that morphology identies syntactic functions only lo cally in German is

that free word order is restricted to the lo cal clause Within the lo cal clause almost every order

is grammatical but outside the clause there is only one p osition for long distance dep endencies

available namely FPSp ec see section In other words in German long distance scrambling is

ungrammatical

Now consider the long distance dep endency in Here in contrast to the b ound reading is

p ossible This shows that the p osition of the empty category cannot b e in the canonical p osition of

the antecedent indicated in bye since then by the pronoun would precede the op erator

i

Jeden Wen sagte sie hab e seine Mutter e getrostet

i i i i

everyone acc who acc said she have his mother nom consoled

Of everyone she said his mother consoled him Who did she sayhis mother

i i i i i

consoled

Again the analogous fact obtains in Hindi and to account for it Bresnan b prop osed based

on work by Maha jan that the trace of the extractee may be freely ordered this was also

prop osed by Haider for German Then the trace may o ccur b efore the sub ject as shown in

and the p ossibility of binding is correctly predicted

LFG JBerman Empty Categories in LFG

Jeden Wen sagte sie hab e e seine Mutter getrostet

i i i i

I accept the idea that in long distance dep endencies there must be an empty category inside the

lo cal domain However rather than allowing it to b e freely ordered we will see that it is p ossible

to lo calize the empty category unambigously by building on work by Frey The resulting

analysis will b e shown to makeseveral correct predictions that the freeorder theory do es not make

First the relevantpoints of Freys analysis of op erator binding will b e very briey presented

An Unambiguous Position for the Empty Category

Frey working within the general framework of GB showed that the assumption usually

made in GB that an op erator has to ccommand a pronoun at dstructure in order to bind it

makes wrong predictions Wehave seen examples of this in rep eated here for convenience

a Jeden Wen mag seine Mutter e

i i i i

everyone accwho acc likes his mother nom

b weil jeden seine Mutter e mag

i i i

b ecause everyone acc his mother nom likes

Frey prop osed that there is a sp ecial relation between the sub ject argument and the functional

head of the sentence the p osition we have lab elled F in which enables an op erator b earing

the sub ject argumenttobindany other argument ccommanded by F Since by denition F in fact

ccommands all no des dominated by its maximal pro jection FP this accounts for the p ossibility

of binding in cf Frey But as we have shown ab ove we account for these cases

by how the prominence principle op erates in German therefore we have no need to followFreys

assumptions ab out the role of F in op erator binding For the pattern of binding in long distance

dep endencies on the other hand Frey makes a prop osal that can be protably adapted to the

present analysis

Sp ecically Frey prop oses that an extracted binder or bindee has to be reconstructed into the

structurally highest p osition of its chain within the lo cal domain For example in a conguration

like the binder or the bindee has to b e reconstructed into the lowest IPSp ecp osition

XP t t t V Frey

Vmax IP V max IP V max

This is the p osition we are calling FPSp ec

It should b e noted that Frey do es not assume an LF level of representation so reconstruction should

not b e understo o d in a literal sense rather he uses the term to refer to the p osition in the chain

to which binding principles apply

Supp ose we rene Freys prop osal by claiming that in general the empty category needed in the

case of long distance dep endencies can only o ccur in the lo cal FPSp ec Then we get the right results

for the weak crossover data as seen in Compare these with and

LFG JBerman Empty Categories in LFG

a Jeden Wen sagte seine Mutter e hab e sie getrostet

i i i FP i

b Jeden Wen sagte sie e hab e seine Mutter getrostet

i i FP i i

Note that on Freys GBbased analysis there are additional traces within the lo cal VP namely in

the base p osition or scrambled p osition of the extracted constituent But on the present analysis

these additional traces are not p ermitted by economy of expression given morphological marking

in German

In short my prop osal is that there is exactly one empty category in long distance dep endencies

and it o ccupies the lo cal FPSp ec Why should this p osition b e singled out

Since wehave adopted Bresnans assumption that morphology can identify syntactic functions only

lo cally it follows that the trace is only used as a last resort to identify the syntactic function of

the extracted element Now recall from the cstructure given in that FPSp ec is a designated

cstructural p osition for discourse functions In fact it is the only p osition in German which is

asso ciated with a sp ecic grammatical function In contrast inside the German VP there is no

structurefunction asso ciation In other words in German the functional pro jection is conguratio

nal while the VP is not The claim is now that an empty category is only licensed in congurational

that is functionally unambiguous p ositions Note that on this hyp othesis we exp ect evidence



for empty categories inside the VP in English but not in German and this is what wend

Restricting the empty category to this p osition has several advantages First this is an unambiguous

p osition Thus with an empty category in this p osition the sentence is much easier to pro cess

computationally than if the empty category can app ear anywhere inside the VP

Second having the empty category in the lo cal FPSp ec means that the structural prominence

obtained by the antecedent in cstructure is reected inside its lo cal domain In other words an

extracted element is still marked as prominent within its lo cal domain at cstructure This is

necessary to account for examples like b



Third the lo calization of the empty category accounts for certain restrictions on extraction As

we will see FPSp ec is the only p osition which allows a constituenttoleave the lo cal clause

Long Distance Dep endencies

In general extraction is p ossible in German out of a nite or nonnite argument clause but

imp ossible out of an clause Both adjuncts and arguments are extractable constituents

The examples in illustrate these facts

a Wer glaubst Du kommt morgen

who nom think you nom comes tomorrow

Who do you think is coming tomorrow



This was p ointed out to me byGertWeb elhuth



This was suggested to me by Joan Bresnan

LFG JBerman Empty Categories in LFG

b Wie lange hat er b ehauptet geschlafen zu hab en

howlong has he claimed slept to haveinf

How long did he claim to have slept

c Welchen Zug hat sie sich b eeilt damit sie no ch erreicht

which train acc has she re hurry sothat she still reaches

Which train did she hurrysothatshewould make

KaplanZaenen account for long distance dep endencies with outsidein functional uncer

tainty They sub divide the conditions on functional uncertainty into the b ottom which contains

the p otential functions at the end of the uncertainty path and the b o dywhichcontains the func

tions in the middle of the path Following KaplanZaenen we can formulate the equation in



to account for long distance dep endencies in German

DF fCOMPjXCOMPg GF

The antecedent o ccupies the sp ecier of the functional category of the matrix clause which is lled

by a discourse function The b o dy contains nite and nonnite argument clauses and the b ottom

contains arguments as well as adjuncts So this equation captures the fact that it is grammatical

to extract out of nite and nonnite argument clauses but ungrammatical to extract out of an

adjunct clause



In KaplanZaenens analysis of English COMP is excluded from the b ottom functions They formulate the

following equation for long distance dep endencies in English DF fCOMPjXCOMPg GFCOMP In German

analogous restrictions were observed byWeb elhuth Only verbs which also sub categorize for a nominal ob ject

allow their sentential to b e topicalized

i Maria argert das

Maria acc annoys that nom

That annoys Maria

ii Dass sie den Bus verpasst hat argert Maria

That she the bus missed has annoys Mary acc

That she missed the bus annoys Maria

iii Maria argert sichdas

Maria nom annoys re that

iv Dass sie den Bus verpasst hat argert sich Maria

That she the bus missed has annoys re Maria nom

In Berman this distinction is captured by assuming that every nite argument clause is obligatorily left

dislo cated and that in accordance with the conditions on topicdrop in German only the sub ject and accusative

ob jectpronoun can b e dropp ed but not the dativegenitive and prep ositional ob jectpronoun

v Dass sie den Bus verpasst hat das argert Maria

That she missed the bus that annoys Maria

vi Dass sie den Bus verpasst hat darub er argert sich Maria

That she missed the bus Maria is annoyed ab out that

On this analysis COMP do esnt have to b e excluded from the b ottom functions

LFG JBerman Empty Categories in LFG

Further restriction on long distance dep endencies

But cannot account for certain word order restrictions in the clause that is extracted out of In

particular the nite verb must b e the rst word in the extraction clause as shown by the examples

in

a Diesen Teilnehmer glaub e ich kennt niemand

FP

this participant acc think I knows nob o dy nom

This participant I think nob o dy knows

b Diesen Teilnehmer glaub e ich niemand kennt

FP

this participant acc think I nob o dy nom knows

This participant I think nob o dy knows

c Diesen Teilnehmer glaub e ich dass niemand kennt

FP

this participant acc think I that nob o dy nom knows

This participant I think that nob o dy knows

On my analysis a is assigned the following structure

Diesen Teilnehmer glaub e ich e kennt niemand

i FP i



As this structure makes clear the emb edded clause has socalled verbsecond word order the verb

o ccupies the head of the functional pro jection and the empty category o ccupies FPSp ec Thus

satises the lo cality condition on function sp ecication But if FPSp ec is lexically lled as in

b the lo cality condition cannot b e satised Extraction out of a dassclause as in c is also

ungrammatical since generally in standard German the sp ecier of a lexically lled

cannot also b e lled This means there is no available p osition for the required trace



Reis argues that extraction out of a verbsecondclause is not p ossible at all in German In her view all

these cases are Verbrst Integrated Parentheticals However as Hub ert Haider p ointed out to me there are some

distinctions b etween parentheticals and the extractionconguration For example true parentheticals as determined

by their p osition in the string do not p ermit certain particles such as denn as in i or whinsitu as in ii

i Wo hab e Karl sagte sie denn es versteckt

where has sub jn Karl said she PRT it hidden

Where did Karl so did she sayhideit

ii Womit soll Karl hat wer b ehauptet den Tresor geknackt hab en

withwhat issupp osed Karl has who claimed the safe cracked has

What is Karl supp osed who claimed to havecracked the safe with

In constrast the following sentences are ne which implies they do not involve parentheticals

iii Wo sagte sie denn hab e Karl es versteckt

So where did she say Karl hid it

iv Womit hat wer b ehauptet soll Karl den Tresor geknackt hab en

Who claimed that Karl is supp osed to have cracked the safe with what

So I assume that German has both extraction from verbsecond clauses as well as verbrstparenthetical expressions

LFG JBerman Empty Categories in LFG

In dialects suchasBavarian in contrast which p ermit a doubly lled COMP as in a extraction

out of a dassclause is p ossible more precisely extraction out of the sp ecier of the dassclause

as in b

a Ich sage nicht wen dass der Hans liebt

I say not who acc that the Hans nom loves

Im not saying who Hans loves

b Wen sagst du e dass der Hans liebt

i i

who acc say you that the Hans nom loves

Who are you saying that Hans loves

But also in Bavarian if the sp ecier p osition is lexically lled creating a whisland extraction is

ungrammatical

Den Hans sage ich nicht wer dass liebt

the Hans acc say I not who nom that knows

Hans Im not saying who loves

My analysis of extraction in German can b e formalized by using insideout functional uncertainty

as prop osed by Bresnan c for topicalization in English and adopting the distinction made by

KaplanZaenen b etween b o dy and b ottom This results in annotated cstructures for long distance

dep endencies like

FP

XP F

DF

F VP

FP

XP F

e

fCOMPjXCOMPgGF DF

Multiply emb edded clauses

Since the empty category is required for function sp ecication only inside the lo cal domain of the

extractee economy of expression prevents intermediate traces in the case of extraction out of a

multiply emb edded clause However when the sp ecier of an intermediate clause is o ccupied by

another constituent as in b the sentence is in fact ungrammatical

LFG JBerman Empty Categories in LFG

a Den Peter glaub e ich sagte der Hans gestern e hat

i i

the Peter acc think I said the Hans nom yesterday has

die Maria eingeladen

the Maria acc invited

Peter is who I think Hans said yesterday Maria invited

b Den Peter glaub e ich gestern sagte der Hans e hat

i i

the Peter acc think I yesterday said the Hans nom has

die Maria eingeladen

the Maria acc invited

To account for this I prop ose that the empty category in the lo cal FPSp ec must b e linked with the

nearest p otential antecedent that is the next higher lexically lled FPSp ec This can b e thought

of as a kind of economy requirement

In b the nearest p otential antecedent is gestern But identifying the fstructure of gestern

and the empty category would violate completeness since the verb einladen would lack an ob ject of

the appropriate sort In a on the other hand the nearest p otential antecedent is den Peter in

the matrix FPSp ec since the intermediate FPSp ecp osition is not lled and this is an appropriate

ob ject of einladen

Wecanformulate the necessary condition by restricting the b o dypath so that the b o dy do es not

pass an fstructure containing a discourse function this is similiar to the way Dalrymple

formulates restrictions on anaphoric binding The resulting equation is given in

f XCOMP COMPg GF DF

DF

The Structure of Innitives

To conclude it may b e noted that my analysis also has consequences for the structure of innitives

As the examples in show op erator binding is p ossible only with socalled coherent innitives

as in a but not with incoherent innitives as in b

a Wen versucht seine Mutter zu trosten

i i

who acc tries his mother nom to console

Who is his mother trying to console

i i

b Wen zwingt seine Mutter Maria t zu heiraten

i i FP i

who acc forces his mother nom Maria acc to marry

Who is his mother forcing Mary to marry

i i

Sentences with coherent innitives have prop erties of mono clausal sentences cf Bech

Haider and others This means that the op erator in a is inside its lo cal domain

However sentences with incoherent innitives are biclausal so the op erator in b is in the matrix

LFG JBerman Empty Categories in LFG

clause outside its lo cal domain This suggests that an empty category must b e present in incoherent

innitives but not in coherent ones If the prop osal that the empty category must o ccur in FPSp ec

is right then the grammaticality contrast in provides indep endent supp ort in addition to

distributional prop erties for the analysis of incoherent innitives as FPs and coherent innitives

as nonfunctionalpro jections

Binding by a lowerranked op erator is also p ossible in the socalled third construction illustrated

by the following examples

weil jeden seine Mutter versucht hat zu trosten

b ecause everyone acc his mother nom tried has to console

b ecause his mother has tried to console everyone

In this case the innitive is extrap osed but one argument of the innitive here jeden is in the

socalled Mittelfeld Since binding is p ossible it follows on the present analysis that there is no

empty category of the longdistancescrambled element rather this must b e basegenerated in its



cstructure p osition

References

Alsina A Predicate Composition A Theory of Syntactic Function Alternations Do ctoral

dissertation Stanford University

Bayer JKornlt J Against scrambling as an instance of Movealpha In N CorverH

Riemsdijk eds Studies on Scrambling Mouton de Gruyter Berlin

Bech G Studien uber das deutsche verbum innitum Munksgaard Kop enhagen

Berman J Topicalization vs Left Dislo cation of Sentential Arguments in German In Pro

ceedings of the LFG Conference in Grenoble

Bresnan J a Linear Order Syntactic Rank and Empty Categories On Weak Crossover In

M Dalrymple et al eds Formal Issues in LexicalFunctional Grammar CSLI Stanford

Bresnan J b Morphology Comp etes with Explaining Typ ological Variation in Weak

Crossover Eects Pap er presentedattheMITWorkshop on OptimalityinSyntax May

Bresnan Jc LexicalFunctional Syntax Barcelona Version Ms Stanford University

Bresnan JZaenen A Deep Unaccusativity in LFG In K Dziwirek et al eds Gram

matical Relations ACrossTheoretical Perspective CSLI Stanford

Choi HW Weak Crossover in Scrambling Languages Precedence Rank and Discourse

Pap er presented at the Annual Meeting of the Linguistic So ciety of America New Orleans

Dalrymple M The Syntax of Anaphoric Binding CSLI Stanford

Dowty D Thematic Protoroles and Argument Selection Language

Frank RLee YSRambow O Scrambling as nonop erator movement and the sp ecial



This was also prop osed byBayerKornlt in the framework of GB

LFG JBerman Empty Categories in LFG

status of sub jects Proceedings of the Third Leiden Conference for Junior Linguistsedby S Bar

biers M den Dikken and C Levelt

FreyW Syntaktische Bedingungen fur die semantische Interpretation Studia grammatica

XXXV Berlin

Haider H Deutsche Syntax generativ Parameter der deutschen Syntax Habilitations

schrift Teil I I Universitat Wien

Haider H Fakultativ koharente Innitkonstruktionen im DeutschenArbeitspapiere des SFB

Nr Universitat Stuttgart

Haider H Deutsche Syntax Generativ Vorstudien zur Theorie einer projektiven Gram

matik Tubingen

Jackendo R On Larsons Treatment of Double Ob ject Construction Linguistic Inquiry

Kaplan RZaenen A Longdistance Dep endencies Constituent Structure and Functional

Uncertainty In M Dalrymple et al eds Formal Issues in LexicalFunctional Grammar CSLI

Stanford

King T Conguring Topic and in Russian CSLI Stanford

Maha jan A The AABAR Distinction and Movement Theory Do ctoral dissertation MIT

Reis M Wer glaubst du hat recht On Socalled Extractions from VerbSecond Clauses

and VerbFirst Parenthetical Constructions in German In Sprache und Pragmatik Germani

stisches Institut der Universitat Lund Lund

Web elhuth G Principles and Parameters of Syntactic Saturation Oxford University Press Oxford