argument structure in hungarian existential sentences

19

Upload: thangdaotao

Post on 08-Nov-2014

17 views

Category:

Documents


4 download

TRANSCRIPT

Page 1: Argument Structure in Hungarian Existential Sentences

Argument Structure in Hungarian Existential Sentences

�Agnes Bende-Farkas

e-mail: [email protected]

August 18, 1999

1 Abstract

In Hungarian the De�niteness E�ect is triggered by lexical properties of certain verb classes. This

paper discusses two properties of these classes: (i) their argument structure, in in particular, the

relationship between such a verb and its internal argument. (ii) The way these verbs describe change.

The framework I will use is DRT with �-abstraction (Muskens [37], Kamp{van Eijck [48]). Some

familiarity will be assumed with Hungarian sentence structure and the semantics of (syntactic) Topic

and Focus (�E.Kiss[16]).

2 Introduction: The Data and the Eighties

In English (and in Germanic languages in general see Freeze[19]) the the source of the De�niteness

E�ect ((1), Milsark[35]) is syntactic, being tied to a particular construction.

a: There are some/many/few cats on the roof

b: �There are all/most/the cats on the roof

c: �There is Macavity/every cat/the cat/it on the roof

d: [Expletive][be][NP(Subj)][Coda]

(1)

In there-insertion contexts, the preverbal subject position is �lled by the expletive there, and the

`logical subject' NP is postverbal. It is usually followed by a coda, which can be a small clause

predicate or a locative PP. Beside the copula be, occasionally unaccusative motion verbs, like arrive

or run (into/out of) can also appear in there-insertion contexts. There are a number of well-

studied meaning di�erences between this construction and \standard" subject{predicate sentences

(Milsark[35], Reuland{ter Meulen[41]).

These well-known facts have been reiterated here in order to mark the contrast with Hungarian. In

this language, the De�niteness E�ect has di�erent sources, while it exhibits by and large the same

semantic restrictions as the English De�niteness E�ect.

As described in Anna Szabolcsi's pioneering work (Szabolcsi [43]), the De�niteness E�ect in Hun-

garian is triggered by lexical factors, and word order or information structure is partly determined

by these. 1

�This paper has bene�ted substantially from discussions with Dorit Abusch, Greta Dalmi, Hans Kamp, Peter

Krause, Barbara Partee, Mats Rooth, Antje Ro�deutscher, Gabriella T�oth, Ede Zimmermann and the participants

of the 1999 Klausurtagung of the Stuttgart PhD Programme in Linguistics. I am grateful to all of them and wish to

take responsibility for remaining errors.1There is another aspect of the De�niteness E�ect/information structure relation in Hungarian. As noted by

Szabolcsi and several authors after her, the presence of Focus overrides or neutralizes the De�niteness E�ect:

1

Page 2: Argument Structure in Hungarian Existential Sentences

(2) and (3) show the De�niteness E�ect with the verbs van `be', kap `receive', ��r `write' and vesz

`buy'.

a: Van k�onyv/k�et k�onyv/(n�emi) tej

Is book/two book/(some) milk

\There is a book/two books/some milk"

b: �Van a k�onyv/Mari k�onyve/minden k�onyv

Is the book/Mari book-Poss3Sg/every book

\There is the book/Mary's book/every book"

(2)

(Ex. (7) Szabolcsi [43]:324)

a: Kaptam/��rtam/vettem k�onyvet/k�et k�onyvet

Got-1Sg/wrote-1Sg/bought-1Sg book-Acc/two book-Acc

\I received/wrote/bought a book/two books"

b: �Kaptam/��rtam/vettem minden k�onyvet/a k�onyvet

Got-1Sg/wrote-1Sg/bought-1Sg every book-Acc/the book-Acc

\I received/wrote/bought every book/the book"

(3)

(Based on ex. (8) from Szabolcsi [43]:324-325)

� In Hungarian entire verb classes show the De�niteness E�ect. The common trait of these classes

is existence/availability (as seen from the list in the Appendix). This semantic criterion is

mixed in a non-trivial way with morphological/syntactic considerations, however.2 The point

is, Hungarian De�niteness E�ect verbs are morphologically simple, unpre�xed verbs. Also,

they are \light" or incorporating verbs, in that they need to be preceded by some nonmaximal

projection, usually a secondary predicate, with which they form one V (Koml�osy[30]). (And

then they become the \complex" verbs the next paragraph is about.) It is important to

note for this paper however, that this distinguished preverbal slot can also be �lled by a(n

incorporated) nominal (as in (9a) in Part 3.1), or by a covert perfectivity operator. In fact,

most De�niteness E�ect examples in this paper are of the latter kind, with an empty operator

in place of an overt secondary predicate.

Complex verbs (see e.g. Ackerman[3], Koml�osy[30], and also the sentences in (4)) do not trigger

the De�niteness E�ect, even if they do have the relevant semantic properties. In the literature

on Hungarian, a \complex" verb usually means a De�niteness E�ect light verb combined with

a secondary predicate: a pre�x, a resultative predicate, a locative complement a.s.o.:

a: J�anos minden falat [�(pirosra) festett]|resultative

John evey wall-Acc [(red-onto) painted]

\John painted every wall (red)"

b: J�anos minden sz�eket [�(be-)hozott]|pre�x

John every chair-Acc [(in-)brought

\John brought (in) every chair"

c: J�anos minden vend�eget [�(a k�onyvt�arba) vezetett]|locative

John every guest-Acc [(the library-into) led]

\John led every guest (into the library)"

(4)

(ii) a: �J�anosTop hozott minden sz�eket

JohnTop brought every chair-Acc

\John brought every chair"

b: J �ANOSFoc hozott minden sz�eket

JohnFoc brought every chair-Acc

\For every chair, it is John who brought it"

2This point has emerged in discussions with Mats Rooth.

2

Page 3: Argument Structure in Hungarian Existential Sentences

� The argument subjected to the De�niteness E�ect is invariably the internal argument: the

object of transitives, and the subject of unaccusatives. 3

It was said earlier that in Hungarian De�niteness E�ect verbs partly determine the information

structure of the sentence they appear in. This amounts to the following: (i) if the verb is in the

Rheme part of the sentence (Rheme meaning loosely non-Topic) and (ii) there is no Focus in the

sentence, then the internal argument of the verb can only occur in the Rheme part, in postverbal

position. Basically, this amounts to a ban on Topicalisation for the internal argument (whereas other

arguments of the verb are not restricted in this way). This is shown in the contrast between (5a-b)

and (5c).4

a: J�anosTop tavalyTop [V P��rt/vett egy k�onyvet (Marinak)]

John last-year wrote/bought one book-Acc (Mary-Dat)

\Last year John wrote/bought a book (for Mary)"

b: MarinakTop [V P��rt/vett egy k�onyvet J�anos]

MaryTop last-yearTop wrote/bought one book-Acc John

|same as the a sentence|

c: ???Egy k�onyvetTop (J�anosTop) (MarinakTop) tavalyTop [V P��rt/vett

One book-AccTop (JohnTop) (Mary-DatTop) last-yearTop wrote/bought

(J�anos) (Marinak)]

(John) (Mary-Dat)

|same as above|

(5)

Szabolcsi's classi�cation needs some correction. This is apparent, for instance, from her (much

later) typology of Hungarian NPs (Szabolcsi[44]). The point is, the De�niteness E�ect proper in

Hungarian seems to be con�ned to MON " NPs only. This is because MON # and non-monotone

NPs are di�erent from the more usual MON " weak NPs, having to appear in preverbal Focus

position.5 (On Hungarian word order and discourse functions see �E.Kiss[16].) This in turn means

that the information structure of sentences with such NPs is not of the existential/presentational

type seen for instance in (3a). The presence of these NPs in Focus position overwrites the initial

existential/presentational information structure of a typical De�niteness E�ect sentence like (3a).

(It could be seen in footnote 1 that syntactic Focus in general overrides the De�niteness E�ect.)

3Word order can be relevant in sentences with the verb van `is'. Even in these cases, though, word order di�erences

are clearly related to di�erences in information structure. Also, note that van `be' is analysed as forming a complex

verb `be-on-the-roof' in the grammatical b example below. That is, if a verb from the relevant class is part of a

complex predicate, it no longer triggers the De�niteness E�ect|just as one could see from (4).

(iii) a: [V P [V 0Van egy macska a tet}on]]

Is a cat the roof-on

\There is a cat on the roof"

b: [Egy macskaTop][V P [V 0 [V a tet}on van]]]

A cat the roof-on is

\One of the cats is on the roof"

4It could be seen from (5) and the preceding footnotes that in Hungarian lexical � novelty requirements and

information structure can interact in a non-trivial way. A proper account of this is well beyond the scope of this

contribution.5A puzzle not noted in Szabolcsi[44] is that sok `many/much', or rengeteg `a great many', `a lot' cannot appear

in the postverbal position associated with the De�niteness E�ect. Rather, they seem to be con�ned to the preverbal

positions of Topic, Quanti�er or Focus, all of which are marked from a discourse point of view:

(iv) a: �J�anos hozott sok sz�eket

John brought much chair-Acc

\John has brought many chairs"

b: J�anos [SOK sz�eketFoc] hozott

John MUCH chair-AccFoc brought

\John has brought MANY chairs"

3

Page 4: Argument Structure in Hungarian Existential Sentences

� Non-monotone and non-negative MON # NPs are shown in (6):

J�anos legfeljebb/pontosan K�ET k�onyvet kapott/��rt

John at-most/exactly two book-Acc received/wrote

\John received/wrote at most/exactly TWO books"

(6)

(6) is an instance of amount quanti�cation (Kroch [32], Szabolcsi [44]). The fact that in (6) the object

NP is in syntactic Focus position suggests that such determiners involve a presupposition/Focus

partitioning of the sentences in which they appear. (6) presupposes that there is a set of books John

wrote/received, and it asserts that the cardinality of this set is at most/exactly two.

As seen in (7), it is indeed awkward to coordinate postverbal MON " and MON # NPs in

Hungarian. I take it as a sign of semantic incompatibility.

a: ???J�anos kapott/��rt egy reg�enyt �es legfeljebb h�arom verset

John received/wrote a novel-Acc and at-most three poem-Acc

\John received/wrote a novel and at most three poems"

b: ?J�anos kapott/��rt egy reg�enyt �es pontosan h�arom verset

John received/wrote a novel-Acc and exactly three poem-Acc

\John received/wrote a novel and exactly three poems"

(7)

� Negative quanti�ers like semmi(t) `nothing(-Acc)', or egy k�onyvet sem `not a single book(-Acc)'

have to occur either in preverbal Focus position as in (8a), or postverbally, as in (8b).

a: J�anos semmit/egy k�onyvet sem kapott/��rt

John nothing-Acc/one book SEM received/wrote

\John has written/received nothing/no books"

b: J�anos nem kapott/��rt semmit sem/egy k�onyvet sem

John SEM received/wrote nothing-Acc SEM/one book SEM

\John didn't receive/write anything/any books"

(8)

Although negative quanti�ers have to occur in Focus position in Hungarian (at least partly), they

do not involve a presupposition/assertion division of the sentence.

This discussion on non-MON " NPs has shown that in Hungarian these do not behave as the

weak NPs of other languages. Apart from some suggestions, this paper does not o�er a detailed

account of their semantics. From now on, by \weak NP" I will usually mean MON " NPs.

After this excursus on NP types in Hungarian, a summary of Szabolcsi's �ndings is in order. So,

what one has in Hungarian is that

� strong NPs cannot be the internal arguments of De�niteness E�ect verbs (in sentences lacking

overt syntactic Focus).

� De�niteness E�ect-verbs have an EXIST component in their lexical entry. With statives this is

their referential argument itself. With process or accomplishment verbs it should be associated

with their consequent state: the state of something being available or known to be at some

location, or with a possessor.

� Corollary: semantic explanations of the English De�niteness E�ect carry over to Hungarian.

For Szabolcsi[43], this meant the tautology/ contradiction account in Barwise{Cooper [5]. As

far as my own work (especially work for the future...) is concerned, I take the De�niteness E�ect

to be basically a novelty constraint (Prince [38], Blutner[7],[6], McNally[34]). In Hungarian,

it is a (lexical) novelty constraint on events and discourse referents that stand for Themes of

these events.

4

Page 5: Argument Structure in Hungarian Existential Sentences

3 What Depends on What, and How

In this part I present two proposals as regards the formalisation of the De�niteness E�ect in Hun-

garian. One proposal, outlined in Part 3.1, resorts to Semantic Incorporation (van Geenhoven[49]).

The other proposal, discussed in 3.2, involves the composition the DE-verb with the determiner of

the object NP. The two approaches will be \tested" against coordination in 3.3. 3.3 also o�ers an

improved, almost-�nal method to combine De�niteness E�ect verbs and object NPs.

3.1 Semantic Incorporation

In this part the relationship between a DE-verb and its internal argument is de�ned as Semantic

Incorporation (van Geenhoven [49], van der Does{de Hoop [46]). The proposal that Hungarian full

NP internal arguments are (also) incorporated into DE-verbs (just like their preverbal bare nominal

counterparts) goes back in fact to Szabolcsi [43].

So, in this part, internal argument NPs are assumed to be incorporated into De�niteness E�ect

verbs. In e�ect this amounts to saying that the internal argument discourse referent is introduced

by the verb itself. It follows then that (weak) NPs are said to contribute a property, and that weak

determiners are analysed as modi�ers/adjectives. Weak determiners have often been analysed as

adjectival, as in Higginbotham[20]. For MON # NPs this strategy has disadvantages, as pointed

out e.g. in Blutner[6], but here it is harmless, because discussion is con�ned to MON " NPs.6

The motivation for treating full NP objects as incorporated is that the internal argument position

of the verb can also be �lled by bare nominals, which are said to be incorporated into their host verb

at some level of linguistic representation. This has in fact to do with the \light" verb property of

De�niteness E�ect verbs. That is, these verbs need to be preceded by some nonmaximal projection,

with which they form a V ; their structure can be schematised as follows: [V V ] (this was mentioned

brie y in Part 1). This \distinguished" preverbal slot can host bare common nouns (usu. internal

arguments, or sometimes obliques). Hungarian linguists usually agree that this is certainly semantic

incorporation, even if the question of incorporation at other levels of linguistic structure is far from

being settled (Kiefer[29], �E.Kiss[16], Koml�osy[30], and papers presented at the CLITE'98 conference

in Szeged). Before presenting the proposal for full NP objects proper, here are the main cases with

incorporated bare nominal objects.

(9) shows a De�niteness E�ect verb with a preverbal, incorporated bare nominal object. Such

nominals are not suitable antencedents for pronominal anaphora:

J�anos [sz�eket hozott]. ??K�enyelmes volt

John [chair-Acc brought]. Comfortable was

\There was a chair-bringing event by John. ?It was comfortable"

\John was chair-bringing (just then).: : : "

(9)

Bare nominals can also occur postverbally, as in (10a). Such sentences usually have an experiential

Perfect reading, as in the mini-discourse (10a,b). This is often, but not always, associated with

Verum Focus. With many (but not all) verbs a more usual Perfect reading is also available, as in

6A natural assumption about this kind of incorporation and Hungarian is that the entire NP gets incorporated into

the verb, and not the head noun only. This latter type of incorporation is found, for instance, in West Greenlandic

(van Geenhoven[49]). But then West Greenlandic has syntactic incorporation, whereas this particular Hungarian

phenomenon is assumed to involve semantic incorporation only.

5

Page 6: Argument Structure in Hungarian Existential Sentences

the mini-discourse (10a,c):

a: J�anos hozott (m�ar) sz�eket

John brought (already) chair-Acc

\John has (indeed/already) brought some chair(s)"

b: : : :Teh�at nem is annyira lusta

: : : So not too that-much lazy

: : : \So he is not that lazy (after all)"

c: : : :Teh�at le-�ul-het-�unk

: : : So pfx-sit-may-1Pl

: : : \So we can sit down now"

(10)

With (10) I meant to show that bare nominals are not always con�ned to the preverbal position

reserved for incorporation or complex predicate formation. In sentences without Focus they occur

postverbally precisely when the sentence has perfect aspect.7 This is relevant for the following reason.

If both bare nominal and full NP objects are assumed to be incorporated, one may nevertheless wish

to properly distinguish between these two kinds of incorporation. Even though bare nominals can

appear preverbally, and full NPs cannot, word order alone is not a su�cient distinctive criterion

(at least not in the syntax). If one wishes to take it into account, this should also include the

aspectual changes that accompany word order variation in Hungarian (seen e.g. in the contrast

between durative (9) and perfective (10a)).

After this (syntactic) digression, (11b) introduces the basic incorporating representation of hoz

`bring' (temporal information will be ignored for the time being). This is the representation that

expects a bare nominal internal argument. It is in fact (i) a translation of the static schema from

van Geenhoven[49] into DRT, with (ii) a language-speci�c modi�cation motivated by the inaprop-

priateness of pronominal links to the nominal (seen in (9)). In (11b) the internal argument discourse

referent is introduced in the DRS that corresponds to the event description. It follows from the

de�nition of subordinate DRS-es (Kamp[23], Kamp{Reyle[24]), that it is inaccessible for further

reference.

a: hoz (bring);

b: �P:�x:

e

e:

x bring � to �

P �(�)

(11)

The nominal itself contributes only a property to the representation of the verb. � is a familiar

location discourse referent whose introduction is motivated in part 4. With verbs like kap `receive'

or lop `steal' it can be understood as the Bene�ciary.

a: sz�eket hoz (chair-bring);

b: �x:

e

e:

x bring � to �

chair�(�)

(12)

Now (11) is inappropriate for (13a), which contains a full NP internal argument. The reason for this

is that (i) full NP internal arguments are accessible antecedents, which means that their discourse

7Then it is not unreasonable to say that in such cases the distinguished preverbal position is �lled by a covert

perfectivity operator (Ede Zimmermann (p.c.)).

6

Page 7: Argument Structure in Hungarian Existential Sentences

referent is to be introduced at an accessible DRS-level. (ii) Their descriptive content is to be

introduced at the same level as their discourse referent (and not internally to the event description-

DRS).

J�anos hozott egy sz�eket. Mari k�enyelmesnek tal�alta

John brought a chair-Acc. Mary comfortable-Dat found+Def3Sg

\John brought a chair. Mary found it comfortable"

(13)

There is a way to turn (11) into an incorporating representation that allows pronominal anaphora.

This is to apply existential disclosure (Dekker[12]) to the event-internal discourse referent �, by (i)

introducing a discourse referent in the main DRS, (ii) equating it with � in the event description-

DRS, and (iii) abstracting a property from the resulting representation. This can be achieved

by means of an \excorporation" or \extraction" operator E (I in (14a) is a variable of the type

((ed); (ed))8of incorporating transitive verbs):

a: E : = �I:�P:�x: [

P �( ); I(��:[� = ])(x)]

b: E(hoz ) = �P:�x: [

e

P �( )

e:

x bring � to �

� =

]

(14)

The problem with (14) is that it can combine with postverbal bare nominals, and does not block

anaphoric links with them. And, conversely, the basic incorporating representation (9) can now com-

bine with full NP arguments, incorporating them at the wrong DRS-level, and disallowing anaphoric

links.

This can be prevented by category assignment: the basic representation (9) can be assigned

category CN jV P , and the \derived" entry (14b) can be assigned category V P=NPpred. The category

NPpred would be assigned the same (e; d) type as bare nominals, with di�erences in the semantics

proper. Nevertheless, this is (i) stipulative, and (ii) does nothing but push semantic information into

the syntactic component. Presumably, the operator E would also need to have a syntactic correlate.9

To conclude, here is a summary of the pros and cons of the Semantic Incorporation approach to

Hungarian verbs and full NP objects.

This approach provides (i) a simple and straightforward account of the dependency of the internal

argument NP on the De�niteness E�ect verb. (ii) Also, it o�ers a simple, and natural-looking way

to express the novelty property of these NPs, while remaining faithful to the original DRT treatment

of inde�nites. But, (i) it is hard to keep \genuine" incorporated N 0-s and full NPs apart, except by

stipulation. (ii) Coordination is problematic (Barbara Partee (p.c.)). The other proposal (presented

in Part 3.2) also su�ers from this problem; coordination will therefore be discussed in a separate

subsection.

3.2 There Were Three Arrivals

This part contains the second proposal on combining Hungarian De�niteness E�ect verbs with their

internal arguments. According to this proposal, the De�niteness E�ect verb is to have access to

8; is Muskens' and Kamp{van Eijck's asymmetric merge. When K1 and K2 are merged, one forms ConK1[ConK2

,

and UK1[ U

0

K2, where U 0

K2contains those elements of UK2

that do not occur in the conditions of K1.9Alternatively, it could be stipulated that every De�niteness E�ect verb is ambiguous, i.e. it has two incorporating

entries, one which expects a nominal, and one which expects a full (weak) NP. To complicate matters further, one

could assign these verbs a third, non-incorporating entry, to be used in sentences with Focus. The unattractiveness

of this alternative is obvious.

7

Page 8: Argument Structure in Hungarian Existential Sentences

determiner information from the internal argument NP. In e�ect, the determiner is said to combine

with the verb �rst, yielding a generalized quanti�er that says e.g. that there were three x such that

x arrived. My motivations for doing so are the following:

� Weak determiners are symmetric (Higginbotham [20], Keenan [27], Zucchi [54]): [[Det N 0 V ]]

is equivalent to [[Det V N 0]]; that is, saying that three arrivals were students is the same as

saying that three students arrived. (See also Doetjes{Honcoop[13].)

� In Hungarian, [VDE Det N 0] form the core part of the Rheme, thus being one unit in informa-

tion structure; accordingly, some freedom may be allowed within that unit.

� The pre�xed variants of De�niteness E�ect verbs also show some kind of connection to the

internal argument NP, by making its determiner partitive-speci�c, as shown in (29b) from

Section 4. It is reasonable then to assume some access to the determiner of the internal

argument NP.

A related proposal from the literature is de Swart's universal/existential insertion mechanism.

(de Swart [10],[11],[9]). The motivation for these schemata is to restore generalized quanti�er types

for predicative NPs, while preserving the intended predicative interpretation. Universal as opposed

to existential insertion is needed to capture the intended scope relations and readings with MON #

NPs.

Existential insertion, for MON " NPs:

a: 9I(BE(�x:N 0(x)))

b: 9I : �P:�Q:9 :[Q( ) ^ P ( )]

(15)

Universal insertion, for MON # NPs:

a: 8I(BE(�x:N 0(x)))

b: 8I : �P:�Q:8 :[Q( )! P ( )]

(16)

The proposal presented here is similar in spirit to these schemata, in that quanti�er (or determiner)

information takes scope, so to speak, over the verb and the N 0, and in that order.

According to another proposal (Bende-Farkas[4]), in English there-insertion contexts the deter-

miner of the postcopular NP is combined with material contributed by there be. This accounts for

the unusual scope properties of these NPs and goes some way toward accounting for the \episodic"

or stage-level poperties of English existential sentences. Ultimately, Hungarian existential sentences

exhibit similar dependencies between the verb and the internal argument.

Preliminary: I take the basic translation of a DE-verb as incorporating, as in (11), repeated here as

(17):

hoz (bring); �P:�x:

e �

e:

x bring� � to �

P �(�)

(17)

If VDE is the DRS for a Hungarian transitive DE-verb, as the DRS from (17), one can have the

following ( D is a determiner type ((ed; (edd))) variable):

�D:�x:[D(��:[VDE (x)(� :[ = �])])](18)

(18) basically does two things to an incorporating verb: (i) the incorporated internal argument is

existentially disclosed (to await binding by the determiner), and (ii) the verb is turned into a function

that expects a determiner as argument (returning a generalised quanti�er).

8

Page 9: Argument Structure in Hungarian Existential Sentences

Given (18), hozott h�arom : : : (`brought three : : : ') will be represented as in (19c):

a: (��:[�P:�u:

e �

e:

u bring� �to �

P �(�)

(x)(� :[ = �])])] =red

� : [

e �

e:

x bring� � to �

� =

(x): = A(x))

b: �x:�Q: [X

jX j = 3; A(x)(X) ; Q�(X) ] =red

c: �x:�Q: [X

jX j = 3;

e �

e:

e

: : :

X = �

(x);Q�(X)]

(19)

It is to be noted that on this approach, NPs correspond uniformly to generalized quanti�ers of type

(edd), and determiners are assigned the type (ed; (edd)).

(18) is like taking de Swart's existential insertion operation and Dekker's existential disclosure

(Dekker [12]) as one step. (18) could in principle be factored into these two distinct steps, but I am

not sure I would like to have unconstrained existential disclosure for DE-verbs.

Comments:

� The motivation behind (18) is that full NP objects are dependent on De�niteness E�ect verbs,

only, this dependency is not incorporation.

� (18) yields the intended interpretation only with weak NPs. D can in principle be a strong

determiner, but since the verb contributes the restrictor and the N 0 the nuclear scope, it

cannot yield what we want. Proper names and pronouns contain no determiners, so they

cannot combine with DE-verbs as de�ned in (18).

� Due to the locality of the Det(V ) combination, the object NP cannot be introduced at a DRS-

level where it would get wide scope relative to other quanti�ers or operators; nor can it be

independent from the event argument of the De�niteness E�ect verb (although this is but the

�rst step toward a fuller account of this dependency).

� (18) is problematic when the object NP is a coordinate structure. This will be discussed in

Part 3.3.

� (18) also presents a technical problem, because the abstracted variable x for the subject is

in the wrong place. In (19) the order of abstracted variables is �x:�Q: : : :, and it should be

�Q:�x: : : : This can be overcome by relying on the subject pro-drop property of Hungarian

and allowing the subject su�x to contribute a pronominal element x. If there is an overt

subject elsewhere in the sentence (in existential sentences it is usually in Topic position), then

x is either bound to it, as presuppositions are bound to their antecedents (van der Sandt[47]),

or it can be re-abstracted once the Rheme part of the sentence ([V Det N 0

Obj ]) is put together.

9

Page 10: Argument Structure in Hungarian Existential Sentences

The coverage of (18) is restricted to monotone increasing NPs. Monotone decreasing and non-

monotone NPs are either presuppositional, or their negation component has to take wide scope.

� At most/exactly n N type NPs are assumed to involve a presupposition and an assertion part.

For instance, legfeljebb h�arom sz�eket (`at most three chairs-Acc') presupposes that there is a

group of chairs, and there is an eventuality in which these chairs are involved; the assertion is

that the cardinality of this group is at most three.

� Negative quanti�ers: I take nem: : : egy N' SEM (`not a single N") to be one unit in the

semantics. The reason for this is that the Focus position in Hungarian often contains kataphoric

\pointers" to constituents that cannot appear in Focus proper for syntactic reasons. In this case

the pointer and the constituent it is linked to are said to share the Focus feature and acquire

exhaustive interpretation. For instance, akkor (`then', `if') in (20a) is in Focus position, and

its construal is `if and only if ', and not simply `if'.

a: J�anos AKKOR f�el, ha egeret l�at

John THEN is-afraid, if mouse-Acc see-3Sg

\John is afraid i�[sic!] he sees mice"

b: J�anos NEM hozott egy sz�eket ?(SEM)

John not brought a chair-Acc (SEM)

without sem: \There is a chair John hasn't brought in"

with sem: \Not a single chair did John bring"

(20)

Then (nem): : : egy : : : sem (`not a single : : : ") is represented as:

�P:�Q: [:

x

P �(x)

Q�(x)

](21)

3.3 Conjunction

Both Semantic Incorporation and the `determiner incorporation' technique fare rather badly with

respect to conjunction. (Barbara Partee (p.c.))

The main reason why conjunction is problematic for Semantic Incorporation is incorporation

itself, viz the assumption that the object NP does not introduce a discourse referent of its own. If

this is so, the discourse referent introduced by the verb cannot \know" that it is to take the i-sum

of two (or more) discourse referents.

a: J�anos hozott h�arom macsk�at �es k�et kuty�at

John brought three cat-Acc and two dog-Acc

\John brought three cats and two dogs"

b: �P: [

x � e

P �( )

e:

x bring � to �

� =

] ( ��: [ cat�(�)

j�j = 3

] ; ��: [ dog�(�)

j�j = 2

] )

(22)

In (22b) the object discourse referent introduced by the verb is to be the sum of two collective

discourse referents. The problem is, there is no source for these referents, because NPs are said not

10

Page 11: Argument Structure in Hungarian Existential Sentences

to introducce discourse referents. Moreover, it can be read o� of (22b) that the sequencing operation

; yields absurd results. If one is to amend the Semantic Incorporation solution, full object NPs are to

be allowed to introduce their own discourse referents, which would get uni�ed with, or bound to, the

discourse referent introduced by the verb. And, of course, then this is not Semantic Incorporation any

more. (But, curiously enough, it converges with the revised version of the determiner incorporation

technique.)

The determiner incorporation alternative does not yield good results with conjunction, either.

The reason is, if one has a structure [V [Det1 N 0

1 and Det2 N 0

2]], the verb has access only to

the leftmost determiner, i.e. Det1. And then this violates the Coordinate Structure Constraint.

The solution proposed by Barbara Partee (p.c.) is to separate cardinality information from the

introduction of a discourse referent. This implies the following:

� discourse referents are introduced separately from determiners;

� (weak) determiners are regarded as adjectives/modi�ers (as part of N 0);

� if two weak object NPs are conjoined in a Hungarian existential sentence, one is entitled to

take their non-Boolean conjunction (Link[33], Kamp{Reyle[24], Krifka[31], Winter[53]); this in

e�ect means the introduction of a new discourse referent, the i-sum of the conjunct discourse

referents.

In order to implement this proposal, one has to overcome certain technical di�culties, mostly related

to type assignment.

As a preliminary, here is the representation of hozott h�arom macsk�at (`(she) brought two cats');

A stands for the existentially disclosed entry of the verb:

a: �P:�Q: [

X

P �(X)

Q�(X)

] (A)(�Z: jZj = 3

cat�(Z)

) =red

b: �P: [

e x � X

P �(X)

e:

x bring � to �

� = X

] (�Z: jZj = 3

cat�(Z)

) =red

c:

e x � X

e:

x bring� � to �

� = X

jX j = 3

cat�(X)

(23)

If one has a conjoined NP, like three cats and two dogs, the following steps are necessary: (i) conjoin

(=merge) the representations of the conjuncts and (ii) introduce a new discourse referent that is the

i-sum of the discourse referents introduced by the head nouns of the conjuncts.

11

Page 12: Argument Structure in Hungarian Existential Sentences

The representation of (the Hungarian equivalent of) three cats and two dogs will then be (24).

� is non-Boolean conjunction, or i-sum.

a: �P: [

Y

jY j = 3

cat�(Y )

P �(Y )

] � �P: [

Z

jZj = 2

dog�(Z)

P �(Z)

] = : : : =

b: �P: [

X Y Z

X = Y � Z

cat�(Y )

jY j = 2

dog�(Z)

jZj = 3

P �(X)

]

(24)

From a comparison of (23) and (24) one can see that (23) cannot readily be extended to conjunction

cases, and this is because of the Coordinate Structure Constraint. Also, when composing material

that yields (24), the NP does not provide the higher type that needs to be combined with the verb.

This is at odds with the determiner composition technique, where the verb expects a determiner of

type ((ed; (edd)).) That is, (24) has neither the \loose" structure, nor the type that yields itself to

the determiner incorporation technique.

One possible way to break up (24), and provide separate access to the referent X , is to say that

X corresponds to a \determiner" of type (ed; (edd)) (let this complex object be labelled as Q). Then

(24) can be taken as the output of applying Q to the two conjoined NPs.

a: (24)

b: �P:�Q: [

X

P �(X)

Q�(X)

] (��: �P:[

Y Z

� = Y � Z

P �(�) : : :])

c: (: = Q(C))

(25)

What one has in (25b) is that the conjoined NP is broken up into a \quanti�er", or numberless

determiner Q and a property C. Now there is a type mismatch in (25b), because C, is of type

(e; (edd)), whereas Q expects a property (ed) type argument. (The initial type of C is edd, from

which an e type argument has been abstracted (�� : : :), when the quanti�er Q has been removed). A

second, more pervasive problem is that it does not make much sense (apart from reasons of typing)

to abstract a variable (�), when it is clearly known what it is, viz the i-sum of the discourse referents

Z and Y . The �rst problem can be corrected by means of a minor adjustment. The second one is

more persistent, and can be solved (I think) if one returns to a more \traditional" variant of DRT.

(But this is work for the future.)

So, the problem of typing can be overcome if the unsaturated (and redundant-looking) property

argument place in C is �lled with the combinator I(= �f:f). This reduces the type of C to the

12

Page 13: Argument Structure in Hungarian Existential Sentences

desired property type ed. Then Q can now combine with the verb �rst, and then with C.

a: hozott h�arom macsk�at �es k�et kuty�at:

((s/he) brought three cats and two dogs)

b: �P:�Q: [

X

P �(X)

Q�(X)

] (A)(��:

Y Z

� = Y � Z

: : :

) =red

c:

e x � X Y Z

e:

x bring� � to �

� = X

X = Y � Z

: : :

(26)

(26c) is the �nal rendering of hozott h�arom kuty�at �es k�et macsk�at (`(s/he) brought three cats and

two dogs'). As before, the \quanti�er" Q combines with the verb (abbreviated as A) �rst, and only

after this with the (defective) coordinate structure.

To end this section, here is a brief comparison of (26) and Semantic Incorporation. (26) yields

the same output as the excorporation operator E (cf. (14b) from the Part 3.1). Only, in (26) the

discourse referent is supplied by the \quanti�er" Q and not by the verb or by the operator E .

Coordination has shown that the object NP has to be allowed to come with its discourse referent.

On the other hand, Q is seen to mediate between the verb and the object NP.

A potential drawback of the Semantic Incorporation strategy (as with other strategies aimed at

keeping scope local) is the possibility to excorporate (see Bende-Farkas[4] for discussion). The

determiner incorporation technique on the other hand does not allow the object NP to be introduced

at a higher level. More precisely, the `object NP' in this case can be either the entire NP, complete

with the i-sum discourse referent X , or the \property" NP ��:K : : :. Neither of these options will

work, because at the original site the De�niteness E�ect verb will not receive the arguments it

expects.

4 Ways of Describing Change

This part is about the way Hungarian De�niteness E�ect verb describe change. In fact, English type

verbs can also describe change in this way, only (i) this fact tends to be overlooked by existing formal

treatments. Also, (ii) the Hungarian case is clearer and more convenient to study, because there

is a tight correspondence between verb \complexity" and transition types in this language. Unlike

English verbs, Hungarian De�niteness E�ect verbs can only describe change in one way. Pre�xed or

resultative counterparts of these verbs on the other hand can only describe change in the \English"

way (with some additional familiarity/speci�city e�ects).

4.1 What Remains Constant

The internal argument of a Hungarian De�niteness E�ect verb is dependent on the event argument

of that verb. One way to explicate this dependency is to posit a familiar (quasi-)indexical variable

in the lexical entry of the verb. This is nothing but the discourse referent � , �rst introduced in

(11), in 3.1. With verbs like hoz `bring' it stands for the Goal of the event, and it is assumed to be

13

Page 14: Argument Structure in Hungarian Existential Sentences

familiar from context.10 (See also Erteschik-Shir [17], Borschev-Partee [8], Abusch{Rooth [2], only,

in these proposals, a variable for spatiotemporal location is introduced in the Topic/Rheme part of

a representation; for Hungarian, this variable is to be part of a lexical entry too).

The aim of the discussion in this subsection is to motivate the introduction of this distinguished

variable in the lexical entries of DE-verbs. The main observation is that Hungarian DE-verbs describe

change in a \non-standard" way, by focusing on what is new at some distinguished location/among

the possessions of a familiar possessor.

The \typical", or default, way English-type verbs describe change can be seen in (27b):

a: Ten cats were lost

b: John found one of the tabbies

(27)

Informal description: take a discourse referent, and observe its \transition" from a Source to a Goal

(Jackendo� [21], Zwarts{Verkuyl [55]). (N.b. cases when the Source is familiar vs cases when only

some unspeci�ed Source (6=Goal) is in your representation).

Formal semantics since Dowty[14] has followed the \English" pattern:

� take a generalised quanti�er Qx as the translation of the internal argument;

� take a state that holds of that variable: �(x) at t;

� take a state that is in some sense the opposite of the previous state, s.t. it holds at t0 (t < t0,

and the interval (t; t0) is in some sense minimal);

:�(x) at t0;

� conjoining the two formulae amounts to having a transition operator BECOME;

� quanti�ers are to be assigned wide scope over BECOME for semantic considerations. So,

what we have is Qx:BECOME(�(x)).

a: A door opened;

b: 9x:[door(x) ^ BECOME(open(x))]

(28)

Hungarian pre�xed verbs show the `English' pattern, plus additional familiarity properties that lead

to coherent mini-discourses like (29):

a: Az el}ot�erben �allt h�arom sz�ek �es k�et fotel.

The lounge-in stood three chair and two armchair

\There were three chairs and two armchairs in the lounge"

b: Azt�an J�anos be-hozott k�et sz�eket

Then John in-brought two chair-Acc

\Then John brought in two of the chairs"

(29)

The Source of the event described in (29b) is the lounge mentioned in the �rst sentence. Note that

the inde�nite k�et sz�eket (`two chairs') in the second sentence has a partitive-speci�c interpretation.

There is consensus in the Hungarian linguistics community that this is due to the presence of the

verbal pre�x (see for instance several papers in Kenesei[28]).

In a context like (29a), Hungarian De�niteness E�ect verbs usually contribute to incoherent, or

disconnected, pieces of discourse:

a: Az el}ot�erben �allt p�ar b�utordarab

The lounge-in stood a-few piece-of-furniture

\There was some furniture in the lounge"

b: #J�anos hozott egy sz�eket

John brought a chair-Acc

\John brought a chair"

(30)

10This proposal concerns verbs that describe a state or change of location or possession (Hungarian equivalents of

`buy', `bring', `keep', `steal' a.s.o.). It does not necessarily carry over to verbs of creation like `paint', or verbs of

change of state like `cut' or `dry'

14

Page 15: Argument Structure in Hungarian Existential Sentences

The point is, the chair mentioned in (30b) is not understood as a subset of the group mentioned in

(30a). Nor is the lounge mentioned in (30a) is understood as the Source of the event described

in (30b). (Unless this is made explicit, e.g. by adding the anaphoric adverbial onnan `from-

there' to (30b).) That is, (30a) is not understood as providing the precondition state (Kamp{

Ro�deutscher [26],[25]) of the event in (30b).

The kind of link that is allowed is to a previously introduced location that serves as the Goal

for the De�niteness E�ect verb:

a: Ebben a szob�aban sok�aig nem volt b�utor,

This-in the room-in for-a-long-time not was furniture,

\For a long time there was no furniture in this room,

b: am��g J�anos nem hozott egy sz�eket

until John not brought a chair-Acc

(this was the case/that is,) until John brought a chair"

(31)

(31) shows is that the adverbial in the �rst sentence provides the Goal of the event described in the

second sentence. Indeed, the state in (31a) provides the precondition state of the event in (31b).

This way of describing change can be schematised as follows:

� take a location (or a Possessor/Bene�ciary), and

� describe what is new there in virtue of some event.

Given a Goal (location, possessor,: : : ) and t0 < t1,

a: s0; t0 X0 = fa 2 Dejs0: [[AT ]]t0(a; )g

b: s1; t1 X1 = fa 2 Dejs1: [[AT ]]t1(a; )g

(32)

the individuals denoted by the Theme argument of a DE verb are said to be in X1 n X0. The

eventuality described by the DE-verb is a transition from s0 to s1. The intuition is that the Theme

is new at .

4.2 New Relative to What

(32) in the previous subsection seems to imply that it is su�cient for the internal argument discourse

referent to be new only relative to the location . A similar proposal has been made for English in

Zucchi[54]:

There-sentences are felicitous only in contexts which entail neither that the intersection

of the set denoted by the N 0 of the postverbal NP with the set denoted by the XP is

empty nor that it is nonempty.

Zucchi [54]:51)

This however is not su�cient: the internal argument has to be new relative to the preceding context,

as shown by the unacceptability of (33b), where the discourse referent for the man is new relative

to the coda of the there-sentence:

a: John met a man yesterday

b: �There's the man John met in the garden

(33)

The same is valid for Hungarian, as seen in (30b). That is, it is not su�cient to have novelty only

relative to, or at, the Goal. Accordingly, (32) from the previous part is not su�cient to capture the

kind of novelty involved in the De�niteness E�ect (in English or Hungarian).

An additional reason for being wary to locate the novelty constraint on Themes in the seman-

tics (as in (32)) is that two new (and distinct) discourse referents may as well correspond to the

15

Page 16: Argument Structure in Hungarian Existential Sentences

same individual. In (34a-b) one needs to have two distinct (and new) discourse referents that may

nevertheless correspond to the same individual in the model:

a: J�anos tal�alt egy macsk�at

John found a cat-Acc

\John found a cat"

b: K�et h�et m�ulva ism�et tal�alt egy macsk�at

Two week later again found a cat-Acc

\Two weeks later he found a cat again"

(34)

4.3 Event Structure

The following list sums up the relevant properties of the event structure (Moens{Steedman[36],

Pustejovsky[39], Kamp{Ro�deutscher[26],[25]) of Hungarian De�niteness E�ect verbs:

� the precondition states of these verbs can be supplied only by descriptions that involve Goals

(as in (31)); that is, links that involve the Theme (or a discourse referent that contains it,

or may be bridged to it) are not licensed (as in (30));11 preceding context can only provide

precondition states that involve the Goal of the event described;

� Dowty's BECOME is not used in decomposition;

� most of these verbs have a consequent state just in case their object is a full NP, with quantized

referential properties (this was implicit in the brief discussion on bare nominal vs full NP

objects in Part 3.1). Then it is sensible to link the introduction of a consequent state to the

presence of the object discourse referent (whereas with pre�xed verbs the pre�x seems to be

responsible for the introduction of the consequent state). It remains to be seen however to

what extent this kind of consequent state yields itself to modi�cation by adverbs such as again

(Dowty[14], Kamp{Ro�deutscher[26],[25], von Stechow[51],[52]).

The following is the �rst version of a meaning postulate that links events described by Hungarian

De�niteness E�ect verbs to their consequent states:

� e x

e:

x V � to �

= �

)

s

s = RES(e)

s: R�(�; )

(35)

Appendix

Classes of Hungarian De�niteness E�ect Verbs:

EXIST: van `be', akad `happen, chance to be', `occur', tart `keep';

BECOME AVAILABLE IN A PARTICULAR FASHION: �erkezik `arrrive', t�ort�enik `hap-

pen', ker�ul `become available',: : :

11This kind of precondition state can be provided after a De�niteness E�ect verb has been introduced:

(iv): a: J�anos tal�alt egy macsk�at

John found one cat-Acc

\John has found a cat"

b: K�es}obb kider�ult, hogy a szomsz�ed�e

Later was-found-out, that the neighbour's

\It was found out later to belong to the neighbour"

16

Page 17: Argument Structure in Hungarian Existential Sentences

COME INTO (NON-)EXISTENCE IN A PARTICULAR FASHION: sz�uletik `be born',

�ep�ul `become/get built', f}o `cook' (intrans.),: : :

CAUSE TO BECOME AVAILABLE IN A PARTICULAR FASHION: kap `get', tal�al `�nd',

szerez `obtain', hoz `bring', visz `take', lop `steal', vesz `buy', ad `give', ker��t `make available',

`acquire', : : :

CAUSE TO BECOME (NON-)EXISTENT, IN A PARTICULAR FASHION: rajzol `draw',

fest `paint', f}oz `cook', eszik `eat', �ep��t `build', : : :

(Based on Szabolcsi [43]:323-324)

References

[1] Werner Abraham and Sjaak de Me�y, editors. Topic, Focus and Con�gurationality. John Ben-

jamins, Amsterdam, 1986.

[2] Dorit Abusch and Mats Rooth. Pragmatic Evaluation of Inde�nites , 1998. Paper presented at

the 6th Symposium on Logic and Language, Budapest.

[3] Farrell Ackerman. Complex Predicates and Morpholexical Relatedness: Locative Alternation

in Hungarian. In Sag and Szabolcsi [42].

[4] �Agnes Bende-Farkas. Quanti�cation in Existential Sentences, 1999. To appear in the Proceed-

ings of the 1999 Formal Grammar Conference; Utrecht.

[5] Jon Barwise and Robin Cooper. Generalized Quanti�ers and Natural Language. Linguistics

and Philosophy, 4:159{219, 1981.

[6] Reihnard Blutner. Dynamic Generalized Quanti�ers and Existential Sentences in Natural Lan-

guages. Journal of Semantics, 10(1):33{64, 1993.

[7] Reinhard Blutner. Type Shifts in Dynamic Logic and the De�niteness E�ect. In K�alm�an and

P�olos [22], pages 49{66.

[8] Vladimir Borschev and Barbara H. Partee. Formal and Lexical Semantics and the Genitive of

Negated Existential Sentences in Russian. In �Zeljko Boskovi�c et al., editor, Formal Approaches

to Slavic Linguistics: The Connecticut Meeting, Ann Arbor, 1997. Michigan Slavic Publications.

[9] Henriette de Swart. Inde�nites between Predication and Reference. Paper presented at SALT'99,

Santa Cruz.

[10] Henriette de Swart. Scope Ambiguitites with Negative Quanti�ers. In von Heusinger and Egli

[50], pages 145{164.

[11] Henriette de Swart. Inde�nites in a Type-Shi�ng Perspective. In Proceedings of the 11th

Amsterdam Colloquium, ILLC:Amsterdam, 1997.

[12] Paul Dekker. Transsentential Meditations. PhD thesis, University of Amsterdam, Amsterdam,

1993.

[13] Jenny Doetjes and Martin Honcoop. The Semantics of Event-Related Readings: A Case for

Pair{Quanti�cation. In Szabolcsi [45], pages 263{310.

[14] David Dowty. Word Meaning in Montague Grammar, volume 7 of Studies in Linguistics and

Philosophy. Reidel/Kluwer, Dordrecht, 2 edition, 1979/1991.

[15] Urs Egli, Peter E. Pause, Christoph Schwarze, Arnim von Stechow, and G�otz Wienold, editors.

Lexical Knowledge in the Organization of Language, volume 114 of CILT. John Benjamins,

Amsterdam and Philadelphia, 1995.

17

Page 18: Argument Structure in Hungarian Existential Sentences

[16] Katalin �E.Kiss. Sentence Structure and Word Order. In Ferenc Kiefer and Katalin �E.Kiss [18].

[17] Nomi Erteschik-Shir. The Dynamics of Focus Structure. Cambridge University Press, Cam-

bridge, 1997.

[18] Ferenc Kiefer and Katalin �E.Kiss, editor. The Syntactic Structure of Hungarian, volume 27 of

Syntax and Semantics. Academic Press, New York, 1994.

[19] R. Freeze. Existentials and Other Locatives. Language, 68:553{595, 1992.

[20] James Higginbotham. Inde�niteness and Predication. In Reuland and ter Meulen [41], pages

43{70.

[21] Ray Jackendo�. Semantic Structures. M.I.T. Press, Cambridge, Mass., 1990/91.

[22] L�aszl�o K�alm�an and L�aszl�o P�olos, editors. Papers from the Second Symposium on Logic and

Language. Akad�emiai, Budapest, 1990.

[23] Hans Kamp. A Theory of Truth and Semantic Representation. In Jeroen Groenendijk et al.,

editor, Formal Methods in the Study of Language. Mathematisch Centrum, Amsterdam, 1981.

[24] Hans Kamp and Uwe Reyle. From Discourse to Logic. Kluwer, Dordrecht, 1993.

[25] Hans Kamp and Antje Ro�deutscher. Drs-Construction and Lexically Driven Inference. The-

oretical Linguistics, 20, 1994.

[26] Hans Kamp and Antje Ro�deutscher. Remarks on Lexical Structure and Drs-Construction.

Theoretical Linguistics, 20, 1994.

[27] Edward L. Keenan. A Semantic De�nition of \Inde�nite NP". In Reuland and ter Meulen [41],

pages 286{317.

[28] Istv�an Kenesei, editor. Approaches to Hungarian, volume 5. Attila J�ozsef University, Szeged,

1995.

[29] Ferenc Kiefer. Noun Incorporation in Hungarian. Acta Linguistica Hungarica, 40(1{2):149{177,

1990{91.

[30] Andr�as Koml�osy. Complements and Adjuncts. In Ferenc Kiefer and Katalin �E.Kiss [18].

[31] Manfred Krifka. Boolean and Non-Boolean `And'. In K�alm�an and P�olos [22], pages 161{189.

[32] Anthony Kroch. Amount Quanti�cation, Referentiality, and Long Wh-Movement, 1989. Ms,

University of Pennsylvania.

[33] Godehard Link. The Logical Analysis of Plurals and Mass Terms: A Lattice-Theoretical Ap-

proach. In Rainer B�auerle and Christoph Schwarze and Arnim von Stechow [40], pages 303{323.

[34] Louise McNally. Existential Sentences without Existential Quanti�cation. Linguistics and

Philosophy, pages 353{392, 1998.

[35] Gary Milsark. Towards an Explanation of Certain Peculiarities of the Existential Construction

in English. Linguistic Analysis, 5(3):1{29, 1977.

[36] Marc Moens and Mark Steedman. Temporal Ontology and Temporal Reference. Computational

Linguistics, 1988.

[37] Reinhard Muskens. Combining Montague Semantics and Discourse Representation Theory.

Linguistics and Philosophy, 19(2):143{186, 1996.

[38] Ellen F. Prince. Toward a Taxonomy of Given-New Information. In Peter Cole, editor, Radical

Pragmatics. Academic Press, New York, 1981.

18

Page 19: Argument Structure in Hungarian Existential Sentences

[39] James Pustejovsky. The Syntax of Event Structure. Cognition, 41:47{81, 1991.

[40] Rainer B�auerle and Christoph Schwarze and Arnim von Stechow, editor. Meaning, Use and

Interpretation of Language. De Gruyter, Berlin/New York, 1983.

[41] Eric J. Reuland and Alice ter Meulen, editors. The Representation of (In)de�niteness. M.I.T.

Press, Cambridge, Mass., 1987.

[42] Ivan Sag and Anna Szabolcsi, editors. Lexical Matters, volume 24 of CSLI Lecture Notes. CSLI,

Stanford, 1992.

[43] Anna Szabolcsi. From the De�niteness E�ect to Lexical Integrity. In Abraham and de Me�y [1].

[44] Anna Szabolcsi. Strategies for Scope Taking. [45].

[45] Anna Szabolcsi, editor. Ways of Scope Taking, volume 65 of Studies in Linguistics and Philos-

ophy. Kluwer, Dordrecht, 1997.

[46] Jaap van der Does and Helen de Hoop. Type-Shifting and Scrambled Inde�nites. Journal of

Semantics, 1998.

[47] Rob van der Sandt. Presupposition Projection as Anaphora Resolution. Journal of Semantics,

9:333{377, 1992.

[48] Jan van Eijck and Hans Kamp. Representing Discourse in Context. In Johan van Benthem and

Alice ter Meulen, editors, Handbook of Logic for Linguistics. Elsevier, 1997.

[49] Veerle van Geenhoven. Semantic Incorporation and Inde�nite Descriptions. PhD thesis,

T�ubingen University, T�ubingen, 1996.

[50] Klaus von Heusinger and Urs Egli, editors. Proceedings of the Konstanz Workshop \Reference

and Anaphoric Relations", number 79 and 84 in Working Papers in Linguistics, Konstanz, 1996.

[51] Arnim von Stechow. Lexical Decomposition in Syntax. In Egli et al. [15], pages 81{117.

[52] Arnim von Stechow. The Di�erent Readings of Wieder `Again': A Structural Account. Journal

of Semantics, 13:87{138, 1996.

[53] Yoad Winter. Flexible Boolean Semantics. PhD thesis, UIL-OTS Utrecht, 1998.

[54] Alessandro Zucchi. The Ingredients of De�niteness and the De�niteness E�ect. Natural Lan-

guage Semantics, 3:33{78, 1995.

[55] Joost Zwarts and Henk J. Verkuyl. An Algebra of Conceptual Structure: An Investigation into

Jackendo�'s Conceptual Semantics. Linguistics and Philosophy, 1994.

19