Beruflich Dokumente
Kultur Dokumente
0. Introduction...........................................................................................................................................
Acknowledgements...............................................................................................................................
1. On the syntax of focus in Hungarian...................................................................................................
1.1 É.Kiss (1992): A VPinternal focus................................................................................................
1.1.1 Problems with a VPinternal focus analysis..........................................................................
1.2 Bródy(1995): Focus Criterion; FP or +Ffeature on T....................................................................
1.2.1 Focus Phrase..........................................................................................................................
1.2.2 Bródy (1990) and (1995) on Vfocusing.................................................................................
2. On the syntax of negation in Hungarian.............................................................................................
2.1 Puskás (1994): Neg Criterion; NegP and obligatory FP.................................................................
2.2 Critique of Puskás’ Analysis...........................................................................................................
2.2.1 Negcliticization on V.............................................................................................................
2.2.2 FP projection in neutral sentences.........................................................................................
2.2.3 Representational HMC...........................................................................................................
2.3 Negation in Hungarian: Two Possible Analyses.............................................................................
2.3.1 Analysis A: nem is [Spec, NegP]............................................................................................
2.3.2 Analysis B: nem is head of NegP............................................................................................
2.4 Cooccurrence of Focus and Negation............................................................................................
2.5 Proposed Analysis of Negation in Hungarian.................................................................................
2.6 Location of negative adverbials and Nwords................................................................................
3. Syntax of the subjunctive in Hungarian.............................................................................................
3.1 É.Kiss (1991) on the syntax of HKMs............................................................................................
3.2 É.Kiss (1995): TP projection...........................................................................................................
3.3 The proposed structure of HKMs....................................................................................................
3.4 Manzini’s Theory of the Subjunctive..............................................................................................
3.5 Adaptation of Manzini’s account to Hungarian..............................................................................
3.6 Tests of the Relevance of Manzini’s Account for the Subjunctive in Hungarian..........................
3.6.1 The Neg, Q and If Operators..................................................................................................
3.6.2 Islandeffects...........................................................................................................................
3.6.3 Parasitic GapLike Effects......................................................................................................
3.6.4 Interaction between the subjunctive and polarity items.........................................................
3.6.5 The Conditional Triggered by the IfOperator......................................................................
3.7 A decisive argument in favour of Manzini’s account: Long Operator Movement.........................
3.7.1 Interactions between the subjunctive and long operator movement: five characteristics.....
1
Contents
3.7.2 Explanation of I......................................................................................................................
3.7.3 Explanation of V.....................................................................................................................
3.7.4 Explanation of III and IV........................................................................................................
3.7.5 Explanation of II.....................................................................................................................
2
3.8 Summary of the Results, Further Support and Remaining Problems.............................................
3.8.1 Summary of the Results...........................................................................................................
3.8.2 Further Relevant Observations..............................................................................................
3.9 Word order predictions for long operator movement.....................................................................
3.9.1 Cooccurrence of focus and the subjunctive..........................................................................
3.9.2 Cooccurrence of negation and the subjunctive.....................................................................
4. Syntax of the imperative in Hungarian...............................................................................................
4.1 Rivero and Terzi (1995)..................................................................................................................
4.2 Zanuttini (1994)...............................................................................................................................
4.3 Rivero (1994)...................................................................................................................................
4.4 The proposed analysis of the imperative.........................................................................................
4.5 Cooccurrence of the imperative and nonneutral elements...........................................................
4.6 A special word order variation of the imperative...........................................................................
5. Conclusion..............................................................................................................................................
Appendix....................................................................................................................................................
An alternative analysis : no separate Negprojection............................................................................
Interaction of negation and focus..........................................................................................................
Interaction of negation and the subjunctive..........................................................................................
Interaction of negation and the imperative............................................................................................
A problem: position of Nwords...........................................................................................................
References..................................................................................................................................................
Figure 1........................................................................................................................................................
Figure 2........................................................................................................................................................
Figure 3........................................................................................................................................................
Figure 4........................................................................................................................................................
Figure 5........................................................................................................................................................
Figure 6........................................................................................................................................................
Figure 7........................................................................................................................................................
Figure 8........................................................................................................................................................
Figure 9........................................................................................................................................................
Contents
Figure 10......................................................................................................................................................
Figure 11......................................................................................................................................................
Figure A1.....................................................................................................................................................
4
Szendroi Kriszta
5
The syntax of focus in Hungarian
There are various syntactic accounts of contrastive focus in Hungarian. I shall
describe two of them in detail. The first, É.Kiss (1992), assumes that the focus
position is in [Spec, VP], and that this position is not unique to a focused constituent,
but it holds many other elements, such as whwords, and verbal modifiers (VM)
(Section 1.1). Section 1.2 describes the second, Bródy (1995). He argues for a
different analysis following May (1985). Section 1.2.2 is devoted to a consideration
subsequent chapters.
1.1 ƒ.Kiss (1992): A VPinternal focus
ƒ.Kiss (1992) assigns the following structure to the Hungarian sentence:
3
S
T VP
Imrei
F VÕ
Erzsitj
V XP XP
ismeri 0i 0j
Imrenom Elisabethacc know
‘Imre knows Elisabeth.’
Figure 1
3
ƒ.Kiss (1995) presents a slightly different analysis of the Hungarian sentence, one which contains a
Tense Projection. In that analysis, she maintains VPinternal focus. Since we are not yet concerned
with the status of the TP projection in Hungarian, let us postpone the description of ƒ.Kiss (1995)
until Section 3.2.
(Figure from ƒ.Kiss, 1992:99)
She claims that the Hungarian sentence consists of an optional Topic position
under S, and a VP. The [Spec, VP] is a Focus position which hosts foci, negated
constituents and whwords. Quantifiers adjoin to VP, and sentential negation is under
VÕ, immediately in front of the Verb. Arguments are base generated under VÕ and
are subsequently moved to operator positions or to the Topicposition. At this level,
the subject is not hierarchically different from the internal arguments. The main
organizational principle that holds in Hungarian syntax is that operators ccommand
and (if they can) precede their scope.
1.1.1 Problems with a VPinternal focus analysis
According to Bródy (1990:206), the problem with É.Kiss (1987) (and Horváth
(1986)) is that they provide no natural way to account for the cases where the V is
itself focused.
He also claims that É.Kiss (1987) and Horváth (1986) differ in that they
postulate different positions for the focused element. ÔBut they agree in that both
assumed that the focused category occupies the same position as the verbal modifier
(VM) since these elements are usually in complementary distribution.Õ He adds that
ÔAs observed by Kenesei (1986) in the case of infinitives, the VM can but does not
1990:210)
1.2 Br—dy(1995): Focus Criterion; FP or +Ffeature on T
The syntax of focus in Hungarian
There are some Weak Crossover configurations that are acceptable in Hungarian, and
unacceptable in English. Br—dy suggests, in line with MahajanÕs account for Hindi,
that there are additional Apositions that can be postulated between the whelement
and the bound pronoun. He claims that AgrO is at least one of these additional
positions. This entails that [Spec, VP] cannot host foci or whelements.
1.2.1 Focus Phrase
which hosts the focused element (unless it is the V that is focused, in which case it
overtly adjoins to F). Following May (1985), he posits a Focus Criterion: the focused
element has a +f feature that it has to check, at least at LF, and that the checking
domain of an XP that has a +F feature must contain a +f phrase at S and at LF. In
other words, the Nfeatures of F are strong in Hungarian, the focused phrase moves
overtly to [Spec, FP] (Br—dy 1995:31).
In addition to the above, he also claims that in a tensed sentence that contains
a focused element, the V moves overtly to F, the supporting data being sentences that
preverbal; it is immediately in front of the V. In sentences which have a focused
element, the VM follows the V. This, he claims, is an indication of Vmovement.
4
‘VM stands for verbal modifier. Hungarian has a large number of verbal modifiers that form a
lexical unit with the verb [...]. For example leszól contains le (down) and szól (calls) – ‘speaks ill of
somebody’ or kiharcol contains ki (out) and harcol (fights), – ‘obtains by fighting’. In such cases the
V and the VM clearly form one lexical unit: the meaning of whole is not compositional. Furthermore
the VM can modify the thematic and Case frame of the V. E.g. szól takes a dative goal while leszól
takes a direct object accusative theme. We should mention here also that the VM position may be
filled also by elements that clearly do not belong there lexically: embedded small clause and
infinitival predicates, [...], adverbials and indefinite expressions being central cases.’ (Br—dy
1990:202)
(1) (a) MARI h’vta fel PŽtert.
MARY rang VM Peter.
(b) *MARI felh’vta PŽtert.
MARY VMrang Peter.
‘Mary rang up Peter.’
In infinitival sentences Verbraising seems to be optional:
It would be better toVMring/toring VM PETER.
Pollock (1989) observed similar behaviour in French and English infinitival
constructions. In line with his and EmondsÕ analyses, Br—dy claims that in a
sentence which contains a focused element, the Vfeatures of a +Tns T are invariably
strong, while those of a Tns T are optionally strong. If the sentence (either tensed or
infinitival) does not contain a focused element or a negation or any other kind of
special element, there is no overt Vmovement. To capture the interaction between
the strength of T and the presence of F, there are two possibilities. First, Br—dy
argues that F does not project a full phrase, +F is located on the T head, but this idea
is seriously problematic (Br—dy, 1995:39), since it is incompatible with the well
motivated assumption that T is universally involved in nominative Case assignment
(or checking). Assuming that a head has only a single Spec, T can only check both
focus and nominative with the help of forming a chain with a higher functional head,
The syntax of focus in Hungarian
movementÕ, [Spec, AgrSP] being higher than [Spec, TP]. So, Br—dy abandons this
view.
Second, Br—dy suggests that the +F feature reappears on T, which it needs to
check against F. Thus, ÔIf the Vfeature (+F) of the F head is strong, as it appears to
be in Hungarian, T must be adjoined to F overtly. If T is itself strong, (as in the case
of +F +Tns T and optionally in the case of +F Tns T) then the verb must be adjoined
to T overtly, and since T must be adjoined to F, the whole complex will be located in
FPÕ (Br—dy, 1995:40).
1.2.2 Bródy (1990) and (1995) on Vfocusing
Bródy (1990) does not accept É.Kiss (1987), because he claims that it does
not provide a natural analysis of focused verbs. Let us take a closer look at how his
analysis extends to this case.
(3) J—zsi ESZI a levest(, nem issza).
Joseph EATS the soup (not drinks pro).
(4) LEjött a lepcsön? Nem, FELSZALADT.
Has he come DOWN the stairs? No, he ran UP.
In this case, according to Br—dy (1990:212), no other constituent can be
focused. See example (5):
(5) *Nem JçNOSSAL VITTEM le a szemetet (hanem...)
Not JOHNWITH TOOKI down the rubbish (but...)
‘I didnÕt TAKE down the rubbish with JOHN (but...)’
Example from Br—dy 1990:212.
Br—dy (1990) says that the ungrammaticality of (5) Ôis a strong evidence for
the claim that it is in fact the V that assigns +f to the category in spec of FPÕ (Br—dy
1990:212). He posits a kind of Ôdoubly filled COMP filterÕ, saying that it is either
the Spec of FP that is filled or the VM+V complex is under F, but not both.
(6) ?J—zsi [FP A LEVEST ESZI](, es nem a kaka—t issza.)
Joseph the SOUPACC EATS ( and not the cocoaacc drinks)
The syntax of focus in Hungarian
‘Joseph EATS the SOUP (and does not drinks the cocoa.)’
Example (6) suggests that although the ‘doublyfilledCOMP’filter of Bródy
(1990) may be argued to hold, it is clearly not the V that assigns +f to the XP in
[Spec, FP]. In (Br—dy 1995:33), he also abandons the view that it is the V that has
the +f feature in the case of XPfocusing, and posits a +Ffeature on T and F . 5
I would rather analyze (4) (repeated here as (7)) differently, saying that in this
case it is the VM that is focused, the V appears in F only because focusing in general
triggers VtoF.
(7) [FPFELszaladt.]
UP ran
‘He ran up.’
The fact that it is not the V that assigns +f to the focused XP is crucial for
analyses of negative and imperative constructions put forward later.
(8) J—zsi [FP A ZŠLDSƑGET nem ]?szereti]?
Joseph the VEGETABLESACC not likes
‘Joseph does not like VEGETABLES.’
(9) J—zsi [FP A SZEMETET vidd ]le!
Joseph the RUBBISHACC takeimp VM!
‘Joseph, take down the RUBBISH!’
5
Horváth (1995) also claims that the +F feature is not on the V, but rather on a functional head.
I think the arguments of Bródy (1995) against a VPinternal focus are
sufficiently grounded. His analysis, although ÔtranslatableÕ to Minimalist feature
driven movement, has one slight weakness: it is not ÔminimalÕ, in a vague sense of
the word, to posit two identical features on two functional heads (+F on T and F) in
analysis.
The syntax of negation in Hungarian
In this chapter I attempt to give a structural analysis of another nonneutral
element of the Hungarian sentence: sentential negation. In section 2.1 I describe the
only detailed analysis in the literature, that of Puskás (1994). Section 2.2 contains a
critique of this analysis and is concluded with its rejection. In Section 2.3 I describe
two possible analyses of sentential negation and enumerate the advantages and
drawbacks of both of them. I reject Analysis A and provide further empirical support
for Analysis B in Section 2.4. Section 2.5 contains the proposed structure (Figure 7),
and a brief summary. Section 2.6 provides some further data.
2.1 Pusk‡s (1994): Neg Criterion; NegP and obligatory FP
Pusk‡s (1994) gives a somewhat different analysis of the Ôfocus fieldÕ in
follows. Similarly to the WhCriterion and the now familiar Focus Criterion, she
adopts from Haegeman & Zanuttini (1991) the NegCriterion: a NEG operator must
0 0
be in a SpecHead relation with an X +NEG; an X +NEG must be in a SpecHead
relation with a NEG operator. Haegeman (1995) argues that the NEGCriterion may
apply at LF or at S. Pusk‡s (1994) argues that in Hungarian, it applies at S.
Pusk‡s has a similar view on FP as Br—dy (1995), but a slightly different
0
one. According to her, the V moves and adjoins to F takes up the +FOCUS feature
there. If a constituent is preposed, it moves to [Spec, FP] and receives a +FOCUS
feature from the V. If nothing is preposed, the +FOCUSfeature is Ôrealised as a
stress on the verbÕ (Pusk‡s, 1994:68). Thus even in neutral sentences she assumes V
toF.
She posits an Agr and a T projection, and following Belletti (1990), she relies
on the morphological structure of the inflected verb to determine the order of the
functional projections. She assumes the following order of functional projections for
the Hungarian sentence: [FP [AgrP [TP [VP ]]]] . 6
According to her, in a negative sentence, the NegP is lower than AgrP, but its
Ôexact position within IP is irrelevantÕ (Pusk‡s, 1994:69). She assigns the following
structure to the Hungarian negative sentence:
FP
FÕ
0
F AGRP
AGRÕ
0
AGR NEGP
NEGÕ
0
NEG TP
0
T VP
Figure 2
(Figure from Pusk‡s, 1994:70)
6
Note that Br—dy assigns the structure: [AgrP [FP [TP [VP ]]]].
The syntax of negation in Hungarian
not saw1s Johnacc
‘I havenÕt seen John.’
(b) J‡nost nem l‡ttam
Johnacc not saw1s
(c) *nem J‡nost l‡ttam
not Johnacc saw1s
(Examples from Pusk‡s, 1994:70)
Examples in (10) suggest, that since the V and nem are always adjacent, nem
may have the same status as pas in French, and that it occupies [Spec, FP]. But we
0
thus posits that nem is in Neg .
whowith not talked2s
‘Who did you not talk to?’
(b) PƒTERT nem l‡ttam.
PETER not saw1s
‘I didnÕt see PETER.’
We can also see from (11), that nem appears between the focused element in
0
[Spec, FP] and the V in F . Therefore, she proposes that nem cliticizes on the V. She
adopts the same idea as Belletti (1990): the +NEG feature is on the head of NegP.
0 0 0 0 0
The V moves to F via T and Agr . The Neg cliticizes on the V in F . This latter
movement seems to be a violation of the ECP, but representationally it is not, because
0
the complex head in F carries both the indices i and j, thus the traces i and j are
properly governed. (See Figure 3.) Pusk‡s also claims that it is in FP that the Neg
Criterion is satisfied.
7
Note that it is generally assumed that whwords move to [Spec, FP] in Hungarian.
The syntax of negation in Hungarian
FP
(J‡nost) FÕ
F AGRP
nemi l‡ttamj
+neg AGRÕ
AGR NEGP
tj
NEGÕ
NEG TP
ti tj É
I didnÕt see John/pro.
Figure 3
(Figure from Pusk‡s, 1994:73)
2.2 Critique of PuskásÕ Analysis
PuskásÕ analysis has three weaknesses. First, she assumes that nem in
Hungarian is a syntactic clitic, cliticizing on the V in F. Second, she assumes VtoF
movement in neutral sentences, and third, she relies on the representational nature of
the HMC. Let us consider them in turn.
2.2.1 Negcliticization on V
In my view, the fact that nem forms one phonological word with the V is not
sufficient for assuming syntactic cliticization. One could just as well claim that want
in a wannaconstruction cliticizes on to. Such a claim would need syntactic support.
For example, in English not can move to C together with the auxiliary has in a
sentence like HasnÕt John seen his father? . 8
not make a difference that this clause can be embedded with a hogy complementizer,
i.e. the V is clearly not in C, what is important here is that the V surfaces above
surface together with the V in this position, which is not the case in Hungarian.
(12) *Ne add MARINAK oda a könyvet!
not give TO MARY VM the book
‘Don’t give the book TO MARY!’
This is, however, not sufficient to prove that Neg does not cliticize on the V.
As pointed out by Bródy, the ungrammaticality of (12) may also be due to the fact
that the prefocus position does not allow negation. This, indeed, seems to be true. As
it can be seen from (14), the prefocus position does not allow a negated constituent
either.
(14) *Nem Józsi csak MARIT szereti.
not Joseph only MARY loves
8
Example from (Ackema, Neeleman & Weerman 1993:28).
The syntax of negation in Hungarian
‘Not Joseph only loves MARY.’
Another strange property of (12) is worth noting here. It is not only
incompatible with a preverbal negation, but also with a negation in lower positions.
See examples (15) and (16).
(15) *Add ne MARINAK oda a könyvet!
give not TO MARY VM the book
‘Don’t give the book TO MARY!’
(16) *Add MARINAK ne oda a könyvet!
give TO MARY not VM the book
‘Don’t give the book TO MARY!’
These examples are compatible with a Negcliticization analysis. Following
Puskás (1994) one might claim that (156) are not grammatical because the V cannot
excorporate from the Neg+V complex head. But they are also compatible with an
analysis that claims that the V cannot move to the Cfield because of an intervening
Neghead, as Rivero and Terzi (1995) argue for the incompatibility of certain
imperatives with negation . 9
2.2.2 FP projection in neutral sentences
Pusk‡sÕ analysis has a second weakness. It assumes that even in sentences
0
where there is nothing in [Spec, FP], FP projects and the V moves to F . This analysis
cannot capture the empirical fact that in neutral sentences the V appears after the VM,
while in nonneutral ones it appears before it. Also, there are no obvious triggers for
appears in front of the VM, but stays in situ otherwise.
9
See Section 4.1 for discussion.
The syntax of negation in Hungarian
2.2.3 Representational HMC
Third, Puskás admits that her analysis relies on the representational nature of
the HMC. In fact, she makes use of this property twice. The Neg head moves to F
skipping Agr, and the V moves from T to Agr, skipping Neg and from then on to F.
In my opinion, PuskásÕ analysis reveals a crucial property of Belletti’s (1990)
analysis which is the following. Belletti (1990) derives the order of functional
projections from inflectional morphology. Let us assume that in a particular language
we arrive at the order of functional projections illustrated in Figure 4. If – following
Belletti (1990) – we allow for a representational HMC, we can derive a full range of
surface orders of heads quite easily, the only restriction being that the lowest two
heads have to end up in the same position, thus have a common head for their chains . 10
F1P
F1 F2P
F2 F3P
F3 F4P
F4 F5P
F5 F6P
F6 VP
Figure 4
F6+F4+F2toF1 we can derive F6+F4+F2+V+F5+F3+F1 complexhead situated in
F1. To give one more example, by VtoF5, V+F5toF2, V+F5+F2toF1 and F6to
10
A similar, but not identical proposal was put forward in Bobaljik and Jonas (1996). I consider their
analysis much more convincing.
F4, F6+F4toF3, F6+F4+F3toF1 we can derive F6+F4+F3+V+F5+F2+F1 situated
in F1.
I think it is not theoretically preferable to claim that the order of functional
projections is morphology driven but at the same time allow for such irrestrictive
movement of functional heads . 11
2.3 Negation in Hungarian: Two Possible Analyses
constrained head movement. I accept that movement respects the Minimal Link
Condition, to be precise MLC is not a filter that filters out improper movements but it
is part of the definition of movement.
Minimal Link Condition:
A can raise to target K if there is no B such that raising of B to K is
legitimate under Last Resort and B is closer to K than A is (where
closeness is defined in terms of ccommand).
(Manzini, lecture notes)
Last Resort:
Move F raises F to target K only if F enters into a checking relation
with a sublabel of K.
(Chomsky 1995:280)
Let us try to account for sentential negation in Hungarian. Consider the
following two analyses.
11
Pollock (1993) argues that Belletti (1990) is not necessarily compatible with a feature checking
approach, where heads are basegenerated with their inflection. On the contrary, one would expect an
order where inflections are Ôpeeled offÕ one by one starting with the outmost one, thus
[[[Root parl] erTense/Mood] ons Agrs] can be checked first by Agrs, then by Tense/Mood thus maintaining
the order derived in Pollock (1989). (Haegeman,1995:33)
The syntax of negation in Hungarian
2.3.1 Analysis A: nem is [Spec, NegP]
If we assume that nem is in a separate functional projection, NegP, it would
seem natural to say that the +NEGfeature is checked in NegP, and that the V only
moves further to FP, if there is a focused constituent in the sentence, otherwise FP is
not projected at all. If we assume that NegP is above TP, as Puskás did, we can say
that in finite, negative sentences the V moves to T and the V+T complex moves to
Neg. This analysis would also be theoretically preferable, since it does not rely on the
analysis cannot be the head of NegP, otherwise leftadjunction of V+TtoNeg would
12
derive Ôtensed VnemÕ order contrary to empirical data.
NegP], and that the head of NegP is occupied by an empty head.
better wouldbe not VMring Peter
‘It would be better not to ring up Peter.’
(b) ?Jobb lenne nem h’vni fel PŽtert.
better wouldbe not ring VM Peter
‘It would be better not to ring up Peter.’
In nonfinite sentences, Vmovement is optional. Thus, for (17a), we say that
the V moves to Neg only after Spellout. Technically speaking we can arrive at the
above configuration, along the lines of Br—dy (1995), if we say that the +NEG
0
feature reappears on T, and the Vfeatures of Neg are strong, thus T always moves
to Neg. If the sentence is +Tns (and optionally if it is Tns), the V overtly moves to T
12
We can rescue this analysis if we dismiss KayneÕs prohibition on rightadjunction, but I do not
consider this option here.
and V+T moves to Neg. Thus, in our analysis, the position of NegP within the IP is
not irrelevant; it has to be above TP, which is coincidentally the same as in Belletti
(1990) and Pollock (1989).
Unfortunately, it is easy to see that this analysis is wrong: it wrongly predicts
a Ôfocused NP, tensed V, nemÕ order in sentences such as (4b), and also (18). If the
Vfeatures of F are strong, we would expect VtoF to occur.
Joseph VEGETABLES not like3s
‘Joseph does not like VEGETABLES.’
(b) *Józsi A ZÖLDSÉGET szereti nem.
Joseph VEGETABLES like1s not
2.3.2 Analysis B: nem is head of NegP
Let us hypothesize that nem is the head of NegP. But, crucially, we take it to
be a free morpheme, which does not allow adjunction. In other words, the V feature
of Neg is weak.
The advantage of this analysis over Analysis A is that it explains naturally
negative sentences with a focused element. The surface order found in these
constructions is Ôfocused NP, nem, VÕ (See Example (18a)). As the Vfeature of F is
strong, we would expect VtoF movement to occur, but the intervening Neghead
blocks this movement. It is arguable that the Vfeature on F is satisfied by NegtoF
movement.
There is a technical problem that arises: we cannot account for the fact that in
tensed sentences the V obligatorily precedes VM if the sentence contains a negation
in the same way as Bródy (1995). We cannot posit a reappearance of +Neg on T,
The syntax of negation in Hungarian
and assume that the Vfeature of a +Neg, +Tns T are invariably strong and those of a
+Neg, Tns are optionally strong, because we assumed that the Vfeature of Neg is
weak. Thus, we need a different explanation for the claim that the tensed V surfaces
in T in a Hungarian negative sentence.
Also, we have to account for the adjacency between nem and the finite V. We
have to stipulate that in Hungarian [Spec, TP] is not projected and that adjunction to
13
TP is not allowed either.
13
We might accept the claim of Bobaljik and Jonas (1996) that [Spec, TP] is only projected in
languages that have Transitive Expletive Constructions.
2.4 Cooccurrence of Focus and Negation
The cooccurrence of negation and focus provides further data that we have to
account for.
We know from É.Kiss (1987:67) that in the case of sentential negation nem is
always adjacent to the V.
Johnacc not seen
‘I didn’t see John.’
(b) Nem láttam Jánost.
(c) *Nem Jánost láttam.
take scope over the focused constituent, as we would expect, since we know that
Hungarian operators ccommand and precede their scope (É.Kiss, 1987:55). See
Examples (20) and (21).
(20) Nem JÁNOST láttam.
‘I didn’t see JOHN.’
shown in (22).
14
Note that the negative marker with subjunctive and also imperative and conditional sentences is not
nem, but ne. They are in complementary distribution; thus I claim that they occupy the same position.
In my opinion the difference between the two is not a syntactic issue. They can either be distinct
lexical items or be selected by the V or by T.
The syntax of negation in Hungarian
(21)
(a) Nem MÁRIÁT várta meg János este a mozi elött (, hanem Évát.)
not MARYacc waited1s VM John evening the cinema infrontof…
‘John waited not for Mary in front of the cinema in the evening, but for
Eve.’
(É.Kiss 1995:55)
(b) Ne JÓZSIT küldd oda(, hanem Mártát!)
not JOSEPHacc sendimp there…
‘Don’t send JOSEPH there, but MARTHA!’
(22)
(a) Nem MÁRIÁT várta meg János este a mozi elött(, hanem csak sétált.)
not Maryacc waited1s VM John evening the cinema infrontof…
‘John was not waiting for MARY in front of the cinema in that evening,
but was only taking a walk.’
(É.Kiss 1995:55)
(b) Ne JÓZSIT küldd oda(, hanem magad menj!)
not JOSEPHacc sendimp there…
‘Don’t send JOSEPH there, but go yourself!’
Also, as it is shown in Example (24), it can cooccur with the postfocus nem
only if it takes scope over the focused constituent, but not the whole sentence. É.Kiss
calls this ‘contrastive negation’, as opposed to ‘constituent negation’.
(23)
(a) Nem MÁRIÁT nem várta meg János este a mozi elött (, hanem Évát.)
not MARY not waited1s VM John evening the cinema infrontof…
‘It was not Mary that John didn’t wait for in front of the cinema in the
evening(, but Eve).’
(Modified example from É.Kiss 95:54)
The syntax of negation in Hungarian
(b) Ne JÓZSIT ne küldd oda, hanem Mártát!
not JOSEPHacc not send there, but MARTHA!
‘It is not Joseph who you shouldn’t send there, but Martha.’
(24)
(a) *Ne JÓZSIT ne küldd oda, hanem magad ne menj!
not JOSEPHacc not send there, but yourself not go!
‘It is not Joseph who you shouldn’t send there, but do not go yourself!’
(b) *Nem MÁRIÁT nem várta meg János este a mozi elött, hanem nem volt
kedve arra sétálni.
not MARYacc not waited VM John evening the cinema infrontof, but not
was feltlike there towalk
‘It was not MARY that John didn’t wait for in front of the cinema in the
evening, but he didn’t feel like walking that way.’
(c) ??Ne EZT a könyvet ne vegyük meg, hanem ne költsük a pénzt.
not THIS book not buyimp1pl VM, but not spendimp1pl the moneyacc
‘Let’s not buy not THIS book, but let’s not spend money.’
(d) ??Nem MARINAK nem adom oda az új albumot, hanem nem adok
kölcsön semmit.
not TOMARY not give1s VM the new albumacc, but not giveloan1s
nothing
‘I don’t give the new album not TO MARY, but I don’t lend anyone
anything.’
Examples (20)(23) argue for another ÔnegpositionÕ above the focused XP . 15
Manzini and Savoia (1997) argue that negation is also possible in some Italian
contexts where VtoC occurs. They argue for the existence of another Neg position
15
Note that the focused constituents (in caps) can always have a contrastive meaning, so I assume that
they are located in the Cfield, i.e. in F.
in the Cfield, above Focus and under Topic. As far as I can tell, this analysis is
compatible with the Hungarian data. However, there is one crucial thing to account
for, namely, that the upper Neg and the Focus is strictly adjacent. This has to be a
above. 16
The fact that this uppernegation can take scope either over [Spec, FP] or over
the whole clause follows since it ccommands and precedes both [Spec, FP] and the
whole clause. The phenomenon illustrated in (24) remains to be accounted for.
2.5 Proposed Analysis of Negation in Hungarian
On the basis of the above I adopt the following structure for a Hungarian
sentence containing sentential negation.
NegP
Neg’
Neg FP
F’
F NegP
Neg’
Neg TP
T VP
V
Figure 5
16
In the preferred case this would follow from the theory. Thus this analysis is consistent with a
theory that does not allow adjunction to XPs, for example on the grounds that Spec positions and XP
adjoined positions are indistinguishable.
The syntax of negation in Hungarian
The structure in Figure 5 adopts Manzini & Savoia’s (1997) idea of an extra
Negposition in the Cfield and Analysis B. It is consistent with recent syntactic
developments. As it was noted earlier, a technical problem arises, namely that we
cannot posit a reappearance of +Neg on T in order to derive ‘nemfinite VVM’
order in sentences without a focused element. An admittedly ‘vulnerable’ solution
might be to claim that even in this case we posit a +Ffeature on T. I do not go into
this technical problem any further here. (See the Appendix for an idea of a completely
different analysis.)
2.6 Location of negative adverbials and Nwords
É.Kiss (1995:221) claims that negative adverbials are in [Spec, FP].
Evein everybody only little trusts PREV
‘Everybody trusts Eve only a little.’
(b) *Évában mindenki CSAK KISSÉ megbízik.
(c) *Évában CSAK KISSÉ mindenki megbízik.
(d) *CSAK KISSÉ Évában mindenki megbízik.
(Examples from É.Kiss, 1995:221)
‘nothing’. She claims that ‘ki (who) and semmit (nothingacc) compete for the same
position’. 17
nothingacc whonom not saw3s
‘Who saw NOTHING?’
(c) Ki nem látott semmit?
whonom not saw3s nothingacc
‘Who didn’t see anything?’
(Modified examples from Puskás, 1994:74)
The syntax of negation in Hungarian
However, as Bródy pointed out, this is not necessarily the case since the
following sentences are grammatical:
nothingacc not atall TO MARY gave3p VM
‘They did not sell anything TO MARY.’
(b) Semmit se MARINAK nem adtak el.
nothingacc not at all TO MARY not gave3p VM
‘There wasn’t anything that they did not sell TO MARY.’
The above word order can be accounted for if we assume that semmit in (27)
complex head formed by is+nem) is under Neg above FP, see Figure 6 and (28). We
can further claim that (26a) is ungrammatical because nem has moved to F and thus
the lower [Spec, NegP] is unavailable for checking +Negfeatures of semmit ‘nothing
acc’. If this analysis is correct, it shows that the upper negation is indeed a full
projected NegP as the lower one; it also has a Spec position that is a possible landing
site for checking +Negfeatures.
(b) [NegP Semmiti se [FP MARINAK k neml [NegP tl [TP adtakj [VP tj el ti
tk ]]]
NegP
semmiti
Neg’
Neg FP
se
MARINAKk F’
F NegP
neml
Neg TP
tl T VP
adtakj tj el ti tk
Figure 6
The syntax of the subjunctive in Hungarian
The use of the subjunctive in Hungarian appears to be strongly restricted both
semantically and structurally. According to Farkas (1992), the indicative is the mood
of assertion, whereas the matrix subjunctive is the mood of imperative utterances, and
together with the conditionaloptative it is used to express wishes (e.g. Ott essem el
Žn... ÔLet me fall there...Õ). Embedded subjunctives are lexically governed, and the
class of verbs/ heads subcategorising the subjunctive can be described in semantic
terms: directives like mond ÔtellÕ, javasol Ôsuggest, desideratives like akar
sz’nŸ Ônot likelyÕ. She also claims that the realis/ irrealis distinction, which was
traditionally used to distinguish subjunctives from declaratives, is clearly on the right
track, though insufficient.
As I am interested in the syntactic rather than the semantic description of the
(1994) offers a syntactic account for the subjunctive in Italian. In what follows, I will
argue that it applies to Hungarian data as well.
The subjunctive in Hungarian – leaving aside for the moment the imperative
forms and certain matrix forms of wishes (see above) – only appears embedded in the
specific clausetype called hogy kštšsz—s mellŽkmondat (HKM), clause introduced
by the complementizer hogy Ôthat’. A structural description of HKMs can be found
in ƒ.Kiss (1991).
In Sections 3.13.3 I will recall the analysis of É.Kiss (1991) and point out
18
that it is incompatible with the results of Chapters 1 and 2. Then I propose a structure
in Figure 9, which is in line with both É.Kiss (1991) and the first two chapters.
Sections 3.4 contains the description of ManziniÕs theory of the subjunctive,
which I adapt to Hungarian in Section 3.5. I will show that it is indeed sensitive to
islandeffects and parasitic gap like effects (Section 3.6). A language particular
account in Section 3.7. I summarize the results in Section 3.8.
Finally, in Section 3.9 I show that the analyses given in Chapters 1,2, and 3
can account for various word order variations found in Hungarian sentences with a
focused XP.
3.1 ƒ.Kiss (1991) on the syntax of HKMs
ƒ.Kiss (1991) says that nouncomplement clauses in English are barriers to
movement, i.e. they are strong islands, because the CP in them is structurally not a
19
complement, but it is adjoined to the noun head. On the other hand, verbcomplement
pronoun bears case corresponding to the thetarole of the clause. It is obligatorily
spelt out except when in the NOM or ACC, in which case it agrees with the V and
communication function of the clause. The clause either follows it immediately, or
stands at the end of the matrix sentence.
18
I will not deal with noun heads that trigger the subjunctive, and I will therefore omit the structural
description of these from ƒ.Kiss (1991).
19
To keep the notation consistent, I use CP where ƒ.Kiss uses S’’.
The syntax of the subjunctive in Hungarian
On the basis of the above, she concludes that the pronoun is the structural
head of the complement clause from which the clause can, or if it is focused, must be
extraposed. Thus she claims that nouncomplement clauses in English (29b) , some 20
clauses (29c) are structurally the same.
20
(29b) is also the translation of (29a).
(29)
(a) Hallottam [NP [NP a h’rt,] [CP hogy meg’gŽrt Marinak kŽt dolgot]]
(c) Hallottam [NP [NP (azt), ][CP hogy meg’gŽrt Marinak kŽt dolgot]]
I heard [that he had promised two things to Mary]
(Modified examples from ƒ.Kiss 1987: 138)
She assigns to them the structure in Figure 7. She claims that owing to
Visibility, sentential arguments are assigned case, but CPs cannot bear case. That is
why they are subordinated to an NP. The complex NP functions as the argument of
the V: the pronoun gets case and since it is semantically empty, the CP gets theta
role. Thus the CP, being an adjunct to the N, will be a Blocking Category, the NP
dominating it inherits barrierhood, so no extraction is possible from out of the CP.
Extraction is only possible if the pronoun is phonologically empty. In this case, she
claims that the NP projection is removed, the CP is directly governed by the V. In this
latter case Visibility is satisfied by the percolation of the case assigned to CP on the
specifier of CP.
SÕÕ
J‡nos SÕ
S
V DP
b’zik DP SÕÕ
abban hogy SÕ
hozz‡ S
V NP
megy Mari
John believehe thatin that tohim goshe Mari
The syntax of the subjunctive in Hungarian
‘J‡nos believes that Mary will marry him.’
Figure 7
3.2 É.Kiss (1995): TP projection
I consider ƒ.KissÕ claim that the pronoun forms one constituent with the CP
sufficiently grounded. I cannot however simply accept her account, because it is not
consistent with results in Chapter 1 and 2 in at least two ways. First, it contains a VP
internal focus; and second, it has no TP projection. In Chapter 1, I gave sufficient
arguments in favour of an FP projection. Concerning the TP, it was argued that it
plays a role in nominative Case checking and that it is universally projected.
earlier work I assumed that in the Hungarian sentence the topic and the VP are
dominated by an S node; i.e., the sentence is not the projection of a category but is an
instantiation of a predication relation [...]. No empirical evidence has been presented
against this view. Nevertheless, in order to keep in line with current theoretical
considerations, I will hypothesize here that the Hungarian sentence is also the
projection of a functional category: that of tense and mood. That is the Hungarian
sentence is a TP or MP.Õ (É.Kiss, 1995:209)
She goes on to say, ÔThe claim that the Hungarian sentence is the projection
of tense can also be given weak empirical support. Namely, long operator movement 21
is easier from infinitives and from conditional and subjunctive clauses than from
finite present or past clauses,...Õ (É.Kiss, 1995:209).
However, the analysis in É.Kiss (1995) is still incompatible with findings in
Chapter 1 and 2 in that it assumes a VPinternal focus.
3.3 The proposed structure of HKMs
21
See Section 3.7 for a detailed analysis of long operator movement constructions.
The syntax of the subjunctive in Hungarian
Let us assign the structure in Figure 8 to the Hungarian sentence. The V and
its arguments are basegenerated under VP. I assume a flat structure to the VP as
ƒ.Kiss, but (apart from a weak empirical support based on Example (36)) I do not
give any proof for this assumption. If T is +Tns, +F/(+Neg), the V moves to T
overtly. At the left edge of the sentence we find CP and FP projections. The [Spec,
22
FP] position is available as a landingsite for phrases to check their +F feature. For 23
simplicityÕs sake I omitted any Agrtype functional heads.
CP
CÕ
C FP
FÕ
F TP
TÕ
T VP
V NP NP
Figure 8
The structure assigned to HKMs can be seen in Figure 9. If the pronoun is not
spelt out, the DP projection is missing; the CP is sister to the V.
22
The properties of CP are not addressed here. I simply assume its existence and that the Chead
hosts complementizers, such as hogy.
23
Topics are possibly in [Spec, TopicP], which projects between FP and CP. Topics do not have
operator features, therefore they are not members of the Q/ If/ Neg/ F etc. class. A clause can contain
more than one topic. Topic positions are strictly above FP in Hungarian. Rizzi (1995: Fn17) claims
that the unavailability of the postfocus topic position(s) is due to the fact that there is obligatory Vto
F movement in Hungarian.
CP
CÕ
C FP
FÕ
F TP
TÕ
T VP
V DP
DP CP
CÕ
C FP
FÕ
F TP
TÕ
T VP
V DP*
Figure 9
3.4 ManziniÕs Theory of the Subjunctive
dependency, it is sensitive to islands and parasiticgaplike effects. Lexical predicates
that trigger the subjunctive embed intentional operators. Verbs of the wanttype, for
claims that in a case like this, the verb embeds an operator in the matrix C position
dependency includes the matrix C, T and V positions and the embedded C and T
positions to respect HMC/ MLC. (See Figure 9) The fact that the matrix C is linked
to the embedded C follows because, for example in Basque, some operators actually
The syntax of the subjunctive in Hungarian
select the embedded complementizer, and also because in Italian the complementizer
che appears to be inevitable in embedded subjunctives. The embedded Cembedded T
dependency can be explained in terms of Tenseanchoring.
3.5 Adaptation of ManziniÕs account to Hungarian
appears in HKMs. If I am to accept ManziniÕs account, I want to claim that the
matrix V embeds an intentional operator in C, which creates a dependency CTV
dependency slightly, and also include the matrix or the embedded F, if either is
projected. We will have to show that this is indeed a syntactic dependency, i.e. it is
sensitive to island and parasiticgaplike effects. The fact that the appearance of hogy
is inevitable in all cases supports ManziniÕs account in the same way as she uses the
inevitable appearance of Italian che as a supporting argument.
CP
CÕ
C FP
Op FÕ
F TP
TÕ
T VP
ind V DP
akar DP CP
CÕ
C FP
hogy FÕ
F TP
TÕ
T VP
subj V DP
egyen fagyit
Akarom, hogy egyen fagyit.
want1s that eatS3s ice creamacc
‘I want him to eat ice cream.’
Figure 10
The syntax of the subjunctive in Hungarian
polarity items, reflexives and disjoint reference follows from ManziniÕs model. I will
partly examine the interaction of these with subjunctives in Hungarian, but most of
what can be done will be left for future investigation.
3.6 Tests of the Relevance of ManziniÕs Account for the Subjunctive in
Hungarian
3.6.1 The Neg, Q and If Operators
According to Manzini, the subjunctive in Italian can be triggered by operators
such as the Neg, Q and If, even in sentences where the predicate would not select the
subjunctive. This is not as straightforward in Hungarian:
(30)
(a) Non sa che io sono/ sia andato
(aÕ) Nem tudja, hogy elmegyek/ *elmenjek
He doesnÕt know that I have/ haveS gone
(b) Sai che lui e/ sia andato
(bÕ) Tudod, hogy elmegy/ *elmenjen
Do you know that he has/ hasS gone
(c) Se sai che lui e/ sia andato, ...
(cÕ) Ha tudod, hogy elmegy/ *elmenjen...
If you know that he has/ hasS gone, ...
(The Italian examples are from Manzini)
But in some cases the Neg operator has a clear role in the appearance of the
subjunctive:
The syntax of the subjunctive in Hungarian
(31)
(a) Val—színŸ, hogy moziba megyek /*menjek ma este
Likely that cinemato goI/ *goIS tonight
‘It is likely that I go to the cinema tonight.’
(b) Nem val—színŸ, hogy moziba megyek / menjek ma este
Not likely that cinemato goI/ goIS tonight
‘It is not likely that I go to the cinema tonight.’
(32)
(a) Lehet, hogy moziba megyek/ *menjek ma este
(b) Lehetetlen, hogy moziba megyek/ menjek ma este
‘It is possible/ impossible that I go/ goS to the cinema tonight’
(33)
(a) Kiz‡rt, hogy moziba megyek/ menjek ma este
(b) Nincs kiz‡rva, hogy moziba megyek/ menjek ma este
‘It is/isnÕt out of the question that I go/ goS to the cinema tonight’
Farkas (1992) gives a semantic explanation and claims that the subjunctive
becomes possible because nem val—sz’nŸ Ônot likelyÕ and lehetetlen ÔimpossibleÕ
are epistemics expressing remote possibility.
3.6.2 Islandeffects
As for islandeffects, Hungarian displays adjunctisland effects, as in (34), and
also complex NPislandeffects (35), though some of these are problematic and are
not considered here in detail (see Fn 16).
(34) SzeretnŽm, mert fontos 0/ *legyen nekem
‘I would like to have it, because it is / *isS important for me’
(35)
(a) J—zsinak kell [a tudat [,hogy a gyerekek biztons‡gban vannak/ ?
legyenek]]
‘J—zsi needs the knowledge that the children are/ ?areS safe’
(b) J—zsinak kell, [hogy a gyerekek biztons‡gban *vannak/ legyenek]
‘J—zsi needs that the children *are/ areS safe’
The syntax of the subjunctive in Hungarian
(c) Fontos [a hír [, hogy Mari férjhez megy/ *menjen]]
‘The rumour/news that Mari marriesind/ *S someone is important’
(d) Fontos, [hogy Mari férjhez megy/ menjen]
‘It is important that Mari marries ind/ S someone’
subject is topicalized. On the other hand, in Hungarian subjectislandeffects do not
focusing do not change the mood.
(36)
(a) Fontos, hogy M‡ria boldog legyen It is important that Mary beS happy
(b) Lehetetlen, hogy... It is impossible that...
(c) KŽtsŽges, hogy... It is doubtful that...
(d) K’v‡natos, hogy... It is desirable that...
(e) ElkerŸlhetš, hogy... It is avoidable that...
(f) ElkerŸlhetetlen, hogy... It is unavoidable that...
(g) SzŸksŽges, hogy... It is needed that...
(h) SzŸksŽgtelen, hogy... It is needless that...
(I) Elengedhetetlen, hogy... It is inevitable that...
(j) Kiz‡rt, hogy... It is out of the question that...
(k) Kell, hogy... It is needed that...
(l) Szabad, hogy... It is allowed that...
(m) ElŽg, hogy... It is enough that...
(n) Tœl+Adj, hogy... It is too+Adj that...
(o) Adjbb ann‡l, hogy... It is more+Adj than...
(p) Fš, hogy... It is of prime importance that...
(q) FeltŽtel, hogy... It is a condition that...
It is almost always possible to construct sentences corresponding to these,
which have the HKM associated with the pronoun in such a way that the pronoun is
not the subject ( (37a) = (36a)) but the object of the sentence (37b).
The syntax of the subjunctive in Hungarian
It is important that Mary beS happy
(b) (Azt) Fontosnak találom, hogy Mária boldog legyen
I find it important that Mary beS happy
The structure in Figure 9 describes sentences like (37b), but (37a) seems to be
adjoined to TP, we would have serious problems explaining why this structure fails to
display subjectisland effects. ƒ.Kiss (1987) claims that we do not find subjectobject
asymmetry in whmovement (and in long operator movement) because the subject
and the object are not assigned hierarchically different positions in the Dstructure of
the sentence. Under a nonconfigurational VP approach we can also account for the
phenomenon that the subjunctive in Hungarian does not seem to show subjectisland
effects. So, the structure in Figure 9 applies to sentences like (37a) as well. If there is
no pronoun, let us take the CP as sister to the head. If the pronoun shows up, the head
will be sister to a DP, which will have two daughters: the DP containing the pronoun
and the CP (See Figure 9).
3.6.3 Parasitic GapLike Effects
As we have seen, subjunctive constructions in Hungarian show the relevant
islandeffects. Let us see if they show parasiticgaplike effects. Recall that I assume,
following Manzini (1994), that the subjunctive forms a dependency: CFTCT. Thus
a matrix C licences the subjunctive in a complement clause. As we have seen in (34),
the subjunctive cannot be licensed in an adjunct clause. The interesting phenomenon
Manzini notes in Italian is that the subjunctive can be licensed in an adjunct clause if
it is adjunct to a complement clause provided the complement clause is subjunctive.
The adjunct CP and the complement TP are taken to be sisters by adjunction. This
construction is independently argued to licence parasitic gaps elsewhere by Manzini.
In Manzini (1994) she argues that the fact that this construction licences multiple
subjunctives is in fact a ‘parasiticgaplike’ phenomenon in so far as it involves a
branching dependency, see (38).
(38) C F T V C T
C T
(39)
(a) SzeretnŽm, hogy felvegyenek, mert ’gy kaphatok/ *kaphassak diplom‡t.
WouldlikeI that accepttheymeS because thisway cangetI/ *cangetIS
degreeacc
‘I would like to be accepted, because I could get a degree this way.’
(b) Lehetetlen, hogy azŽrt jšjjšn ide, mert franci‡t tanul/ *tanuljon
Impossible that for come3sS hereto because Frenchacc learnhe/ *learn
heS
‘ItÕs impossible that he comes here because he is learning French.’
We can see from (39) that this branching dependency does not apply in
Hungarian; multiple subjunctives are not allowed.
3.6.4 Interaction between the subjunctive and polarity items
Neg, Q and If – the operators that licence the subjunctive in Italian are known
to form a natural class: they can licence negative polarity items (Manzini, 1994:45).
If both negative polarity items and the subjunctive is assumed to be ‘the spellout of
indefinites bound by certain operators’ (Manzini, 1994:9) then we expect them to co
The syntax of the subjunctive in Hungarian
occur. It was pointed out to me by É.Kiss that bármit is ‘anythingacc’ is a negative
24
polarity item in Hungarian. Let us see if in Hungarian the subjunctive can cooccur
cannot necessarily bind a subjunctive in a relative clause, even if the head of the
relative clause is a polarity item, but (40ab) are grammatical.
(40)
(a) KŽtlem, hogy bármit is l‡t/ lásson.
doubt1s that anything seeind3s/seeS3s
‘I doubt that he sees/seesS anything.’
(b) Lehetetlen, hogy bármit is ért/ értsen.
impossible that anything understandind3s/S3s
‘It is impossible that he understands/understandS anything.’
(c) Nem l‡tok egy t‡rgyat sem, ami tetszik/ *tessen
not see1s an object even that like/ *likeS
‘I donÕt see any object that I like.’
3.6.5 The Conditional Triggered by the IfOperator
We get a clearer picture if we extend our data: let me include sentences with
the conditional. I claim that these are indeed sensitive to parasitic gap like effects, as
in (41).
(41)
24
The polarity item bármit is ‘anythingacc’ can be licensed by a Neg in a superordinate clause (i), or
by a clausemate If (ii), but it cannot be licensed by a clausemate Neg (iii).
(i) Nem hiszem el, hogy Józsi bármit is ért.
‘I don’t believe that Joseph understands anything.’
(ii) Ha bármit is megtudsz, értesíts!
‘If you learn anything, tell me!’
(iii) *Józsi nem ért bármit is.
‘Joseph doesn’t understand anything.’
(a) Kiz‡rt, hogy azŽrt jšn ide, mert nincs / *ne lenne m‡s dolga.
It is out of the question, that forthat comeheInd here, because nothashe
/ *not hasC other business
‘It is out of question that he comes here because he has/ *would have
nothing else to do.’
(b) Kiz‡rt, hogy azŽrt jšnne ide, mert nincs / ne lenne m‡s dolga.
It is out of the question, that forthat comeheC here, because nothashe /
not hasC other business
‘It is out of question that he would come here because he has/ would have
nothing else to do.’
(c) Kiz‡rt, hogy azŽrt jšn ide, mert unatkozik / *unatkozna.
It is out of the question, that forthat comehe here, because isboredhe /
*is boredheC
‘It is out of question that he comes here because he is/ *would be bored.’
The syntax of the subjunctive in Hungarian
(d) Kiz‡rt, hogy azŽrt jšnne ide, mert unatkozik / unatkozna.
It is out of the question, that forthat comeheC here,because isboredhe/
isboredheC
‘It is out of question that he would come here because he is/ would be
bored.’
The conditional also licences polarity items, or vice versa, can be licensed by
tárgyat sem is a polarity item licensed by a clausemate Neg. In (42c) bármit is is not
licensed, and in (42d) it is licensed by a Neg in the superordinate clause. We can see
that wherever the polarity item is possible (42b,d), the conditional is also possible.
seeI an objectacc that likeIInd/*S/*C
‘I see an object that I like.’
(b) Nem l‡tok egy t‡rgyat sem, ami tetszik/ *tessen/ tetszene.
Not seeI an object neither that likeI / *likeIS / likeIC
‘I donÕt see any objects that I likeC.’
(c) L‡tom hogy bármi is *tetszik/ *tessen/ *tetszene.
see1s that anything like2s*ind/*S/*C
‘I see that you like anything.’
(d) Nem l‡tom hogy bármi is ?tetszik/ *tessen/ tetszene.
not see1s that anything like2s?ind/*S/ C
‘I don’t see that you like anything.’
This extension of the data can be justified by the fact that in sentences with
the Ifoperator, Italian uses subjunctive and Hungarian uses the conditional. Along
the lines of ManziniÕs argument I can say that the Ifoperator is embedded under the
matrix C, and it triggers subjunctive on the embedded T in Italian, and conditional in
Hungarian.
The syntax of the subjunctive in Hungarian
(43) Ha gondolod, hogy elmegy/elmenne
if you think that he goind/gocond
‘If you think that he goes ...’
Se sai che lui è/sia andato
‘If you know that he hasind/ hassubj gone’
3.7 A decisive argument in favour of ManziniÕs account:
Long Operator Movement
It is argued in ManziniÕs article that long distance binding of the Icelandic
reflexive sig is only possible if it is across a subjunctive. Hungarian subjunctives have
a characteristic which is similar in nature. This phenomenon can be best explained by
the following examples:
(Examples (78) É.Kiss, 1995:210)
These sentences are examples of long operator movement, the verbs they
contain are called bridgeverbs. It is assumed here that most Hungarian speakers
would either accept all three, or would only accept (44b) and (44c). In either case the
sentence with the subjunctive (or the conditional) is generally considered better than
the one with the indicative.
Recall from Section 3.2 É.Kiss’ (1995) argument, that this construction is an
empirical argument in favour of the existence of a TP projection in Hungarian. Her
explanation is as follows: ‘... [this] fact can be easily described if the VP is governed
by T, and a [independent] T is an Lmarker, while a [+independent] T is not, turning
25
the VP into a barrier.Õ Her analysis cannot be accepted under a nonVPinternal
example, in the case of English, assuming that T Lmarks V results in overgeneration,
because this would allow for direct VtoC movement in a sentence like *Havei he
where the pronoun shows up.
3.7.1 Interactions between the subjunctive and long operator movement:
five characteristics
If I am to say, that long operator movement is not excluded in other contexts
but is at least better with the subjunctive, and I want to account for the structure of the
subjunctive in terms of ManziniÕs syntactic dependencyanalysis, I have to show
how does the subjunctive facilitate long operator movement and I also have to show
that ManziniÕs model explains ƒ.KissÕ other observations as well. É.Kiss (1987)
says that sentence intertwining, that is long operator movement, has three interrelated
characteristics:
25
ƒ.Kiss adopts a proposal of Antal (1977), who assumes that tense and mood are in complementary
distribution in the Hungarian sentence. ƒ.Kiss attributes T the following possible values:
T Ð> [+independent]/ [independent]
[+independent] Ð> present/past
[independent] Ð> [+infinitival]/ [+modal]
[+modal] Ð> imperative/ subjunctive/ real conditional/ irreal conditionalÕ
This is in line with the idea behind ManziniÕs account that in fact subjunctive mood is realized on the
Tense projection (at least in some languages).
The syntax of the subjunctive in Hungarian
No 1 There is no pronoun associated with the embedded clause.
No 2 Clause contains hogy or other complementizer
No 3 Matrix V agrees with the object raised from the embedded sentence.
If we want to account for the interaction of long operator movement with the
subjunctive, we face a more complicated picture. At first sight, we predict that since
both the subjunctive and the long operator movement are syntactic dependencies, they
will appear in similar contexts. On the basis of No 1 we might jump to the conclusion
that the subjunctive is also blocked if the pronoun is spelt out. This bournes out to be
false. It is easy to find examples where the subjunctive is better without the pronoun,
but is still possible with it. There are also a few examples where the sentence is worse
without the pronoun, as in (45).
fearIyouacc thatfrom/0 that gothereyou
‘I am worried that you go there. (It is a dangerous place.)’
Let us now set up a new set of observations on the interaction of long operator
movement and the subjunctive.
I The subjunctive appears both with a pronoun associated with the HKM
and without it.
II Long operator movement is better from subjunctive embedded clauses
than from indicative ones.
III If there is long operator movement, the pronoun cannot be spelt out.
(cf. No 1)
IV The clause contains the complementizer hogy. (cf. No 2)
V The matrix V agrees with the preposed object. (cf. No 3)
The syntax of the subjunctive in Hungarian
3.7.2 Explanation of I
Let us begin by explaining observation I.
The fact that the subjunctive is possible if the pronoun is spelt out might seem
a little strange at first sight, since we would expect the DP to act as a barrier, and
block the CFTVCT dependency. We might want to claim that the DP is not a
parasitic gaps. However, if this was the case, it would be rather difficult to account
for the fact that long operator movement is still impossible in these contexts.
On the other hand, looking at the MLC described in Section 2.3 we can claim
that movement of the embedded C to the matrix V is not blocked by the intervening
D, because D does not really ‘intervene’ in the dependency. The following condition
applies: the embedded C moves to D (and not to the matrix V) if and only if D c
commands b iff a does not dominate b and every category dominating a dominates
b.’ (Roberts, 1997:27) We can now see that the above condition is not satisfied here
since the first conjunct does not apply: the category DP dominates D (a), but does not
dominate C (b), because not every segment of it dominates it.
3.7.3 Explanation of V
possible from argument clauses, where the clause is coindexed with the AGR marker
of the V, hence V follows. She claims that if the pronoun does not show up, the CP
can bear case which appears at [Spec, CP]. The XP on the way up has to enter this
position, which would facilitate its ACCcase marking by the matrix predicate.
In a structure like the one in Figure 9, the preposed object cannot get ACC
case in the lower [Spec, CP]. However, we can claim that if the pronoun is not spelt
checked until something moves up to the position where it can be checked, i.e. to
some [Spec, AgrO] in the matrix sentence.
3.7.4 Explanation of III and IV
Let us now move on to III. Long operator movement, if the pronoun appears,
will cross one barrier, thus constitutes a weak Subjacency violation. It also violates
the ECP, if a nonargument is moved. See Figure 11 for illustration.
CP
CÕ
C FP
Op FAGYIAcc FÕ
F TP
TÕ
T VP
ind ti''' VP
V DP
akar
DP CP
azt ti'' CÕ
C FP
hogy FÕ
F TP
TÕ
The syntax of the subjunctive in Hungarian
T VP
subj ti' VP
V DP
egyen ti
*FAGYIT akarom azt, hogy egyen.
ICE CREAM want1s that that eatS3s
‘I want him to eat ICE CREAM.’
Figure 11
crossing the lower TP, a Blocking Category, but let us assume, not a barrier. Moving
out of this intermediate position is however blocked. One way to explain this is to
claim that the lower CP, being not Lmarked is a BC and a barrier. This is however a
little problematic, if we want to claim that it is the clause that gets the thetarole from
the V, the pronoun being semantically empty. A supporting argument could lie in the
definition of Lmarking in Chomsky (1986:14): Lmarking can be identified as direct
thetamarking. Direct thetamarking, in turn, can only be done if the marker and the
marked elements are sisters. Since the CP is not sister to the V head it cannot be L
marked by it.
The DP, interestingly, will neither be a barrier inherently, because it is L
marked, nor will it inherit barrierhood from the lower CP, since it does not dominate
it, because only one segment of it dominates it. Movement out of the lower CP is thus
a weak Subjacency violation. 26
26
As far as I can tell, this is a problem of the Barriersframework (Chomsky, 1986). This
configuration is precisely an adjunctisland. We would expect movement to be blocked, and the
Barriersapproach only accounts for a weak Subjacency violation because an XPadjunction structure
is used as an ‘escapehatch’ in other contexts (e.g. movement out of the VP.)
There is an alternative way to achieve the same result. We can posit an “extra
barrier” in exactly the same way as Chomsky (1986:367) proposes for Noun
islands, thus either the NP or the CP, which is complement to the N, is a barrier. The
transmit barrierhood to the NP) because the complementizer cannot be deleted in such
sentences. This implies that the Cposition is not properly governed, thus, that it is in
undeletable.
The syntax of the subjunctive in Hungarian
Chomsky (1986:37) also talks about a parametric variation between English
and Italian, namely that in English the most embedded tensed IP, in Italian tensed CP
is an “extra barrier”. He also mentions that it may be the case that the parametric
variation involves not the distinction tense vs. infinitive but the distinction indicative
vs. infinitivesubjunctive.
If the pronoun is spelt out, movement out of the CP does not seem to show
accept the first explanation to the weak Subjacency violation observed in this case,
namely that the CP is not Lmarked.
government of the intermediate trace in [Spec, CP] from outside the clause. Thus we
have an ECP violation if a nonargument is moved out.
3.7.5 Explanation of II
The only remaining observation to be accounted for is II. Based on the fact
that the CP is indeed sister to the V head, I claim that the CP is Lmarked by the V.
However, movement is blocked by a Subjacency effect, a ‘typically weak’ one as
Chomsky (1986:36) calls it, if the lower clause is indicative. I would like to claim that
an “extra barrier” occurs in this case, but movement freely occurs if the lower clause
is in the subjunctive. In my opinion there is a correlation between this “extra barrier”
and the T of the clause being [+independent]. This correlation is very natural in the
sense that a clause with a [+independent] T forms a ‘unit’; in some sense this is the
idea behind the concept of an island. One might claim that an indicative rootclause is
some sort of an island (perhaps one might call it ‘Rootisland’). The CFTVCT
dependency can be argued to account for the absence of this “extra barrier”, to the
extent that this dependency relates the matrix C and the embedded C (or T) and
therefore it ‘breaks the unitnature’ of the embedded sentence. In a metaphoric way, it
is not an ‘island’ any longer, it is a ‘peninsula’. This is of course only speculation, a
formal account remains to be given. I conclude without a formal explanation of the
grammatical judgements for (44) and thus of Observation II.
3.8 Summary of the Results, Further Support and Remaining Problems
3.8.1 Summary of the Results
Using the structure in Figure 9 we can account for the following observations:
HKM and without it.
II Long operator movement is better from subjunctive embedded clauses.
III If there is long operator movement, the pronoun cannot be spelt out.
IV The clause contains the complementizer hogy.
V The matrix V agrees with the preposed object.
3.8.2 Further Relevant Observations
Transylvanian dialect of Hungarian. In terms of ManziniÕs syntactic dependency, I
have to argue that this dialect has an empty operator in this position.
The syntax of the subjunctive in Hungarian
She also observes that long operator movement can cross several sentence
boundaries, but the constituent does not have to go to the highest possible position.
Several constituents can be preposed at the same time even to landing sites in
approach. I do not have an answer to the question why the constituent gets preposed
to a higher clause Focus position, if it can also surface in an embedded FP.
ƒ.Kiss observes that the long movement of the PreV, the perfectivity operator,
is only possible in the presence of a matrix predicate of modal meaning. In the case of
nonmodal Vs, which themselves can be marked for perfectivity, this possibility is
excluded. I think this observation is not only true for the perfectivity operator, but
also for other verbal prefixes.
John VM wants that eat1s the supperhisacc
‘John wants me to eat UP his supper.’
(b) *[J‡nos [megi [mondta, [hogy [[egyem ei a vacsor‡j‡t]]]]]]
John VM said that eatI the supperhisacc
‘John told me to eat UP his supper.’
(ƒ.Kiss 1987: 126)
(47) (a) [J‡nosnak [kii [kell, [hogy [[hœzzam ei a fog‡t]]]]]]
Johndat VM must that pullIS the toothhisacc
‘I have to pull out John’s tooth.’
(b) *[J‡nos [kii [mondja, [hogy [[hœzza ei a fogamat]]]]]]
John VM says that pullhe the toothmyacc
‘John says that he pulls out my tooth.’
I think it is possible to argue that these verbal prefixes are moved to the matrix
[Spec,FP], as I argued for Example (7) in Section 1.2.2. The subjunctivedependency,
then, could easily account for the fact that in this case only bridgeverbs triggering the
subjunctive appear.
In addition to the above she observes that only Vs not subcategorized for an
interrogative complement are bridgeverbs. I have no explanation for this but I would
suggest that it might be related to semantic characteristics of bridgeverbs, thus no
syntactic explanation is needed.
The syntax of the subjunctive in Hungarian
3.9 Word order predictions for long operator movement
3.9.1 Cooccurrence of focus and the subjunctive
I have included Br—dy’s (1995) Focusaccount in the structure in Figure 9. In
a sentence containing long operator movement the matrix clause has a FP projection,
above TP. This clause is tensed, thus T is +Tns, +F: it requires overt VtoT, which in
turn moves to F. Thus, we predict that if Hungarian had a bridgeverb that has a VM,
the V would precede VM. The lower clause is also +Tns, but let us assume it is F.
Thus Verbraising is predicted not to appear.
PETERacc wantI that looksS at/atlooksS my work
‘I want PETER to look at my work.’
essentially two ways to account for this phenomenon. First, using the dependency
described in Section 3.4 that accounts for the subjunctive on the lower T, and second,
by claiming that the XP is not preposed from its baseposition, but from a lower
clause Focusposition. Let us start by the latter: if indeed the lower clause projects a
FP, and XP moves to its Spec, it would check its +f feature. Therefore, for it to move
further, we need to posit at least one of the following two assumptions: the XP either
has another feature to check in the matrix [Spec, FP], or it moves to matrix FP in
order to check the +Ffeature of the matrix F position. The second idea would violate
the principle of Greed, to say the least, and the first also leads to serious problems
because it entails that an XP can have features that a lower FP cannot check, while a
matrix FP can.
Let us try then, to account for the marginal grammaticality of (48b) using the
dependency described in Section 3.4. Following Bródy (1995), the matrix F has a +F
feature, which reappears on the matrix T. Let us now assume that this +F feature is
available at the foot of the chain as well, namely on the lower T. If it is, then it sounds
plausible to claim that it may trigger overt VtoT in the lower clause.
On the basis of the above, we can account for an even stranger phenomenon.
Let us see sentences where a matrix constituent is focused:
The examples in (49) contain sentences where the matrix verb does not trigger
the subjunctive on the lower verb, whereas in (50) it does. The a) and b) examples
contain no focus while the c) and d) examples have a focused element in the matrix
clause. The b) and d) examples show Vmovement.
The syntax of the subjunctive in Hungarian
As we can see in (50d), Vraising is possible in a subjunctive lower clause if
the matrix clause contains a focused element, even if this XP is not preposed from the
lower clause. The CFTVCT dependency can account for this data. Let me note
that the idea, that Vraising is possible in (48b) because the focused element is moved
from the lower [Spec,FP], could not be generalized to account for the grammaticality
of (50d).
3.9.2 Cooccurrence of negation and the subjunctive
Long operator movement sentences that contain sentential negation have the
following word order variations:
Joseph wanted3s that not upgiveS1s/ giveS1s up.
‘Joseph wanted me not to give up.’
(c) J—zsi nem akarta, hogy feladjam.
(d) *J—zsi nem akarta, hogy adjam fel.
Joseph not wanted3s, that upgiveS1s/ giveS1s up
‘Joseph didn’t want me to give up.’
Examples (51a) and (51b) suggest that Vmovement is obligatory in negative
embedded subjunctive clauses. This is not surprising, since this is the word order
found in simple sentences. However, (51d) is ungrammatical. Whatever triggered V
raising in the simple clause and the embedded clause cannot do so via the
subjunctivechain.
I assume that this is possible because operators that are claimed to belong to
one semantic class do not necessarily behave exactly alike in syntax. Much the same
way as Q and If do not license the subjunctive but at least partly licence negative
polarity items in Hungarian, whereas they do both in Italian, we can say that the +F
feature is available at the foot of the subjunctive chain in Hungarian, while the feature
responsible for VtoT in negative sentences is not. 27
27
The solution to this question would have to go hand in hand with the solution to the problem of
triggering VtoT in negative nonfocus sentences.
The syntax of the imperative in Hungarian
On the basis of the previous discussion, I attempt to find a structural analysis
of the imperative mood in Hungarian. After a brief literature review (Section 4.14.3)
I propose an analysis in Section 4.4. Sections 4.5 and 4.6 contain empirical data,
accounted for by the proposed analysis.
4.1 Rivero and Terzi (1995)
Rivero and Terzi (1995:301) divide languages into two classes:
Class I: Imperative Vs have a distinctive syntax.
Class II: Imperative Vs lack a distinctive syntax.
Other Class I characteristics are that imperatives cannot be negated, and that
imperative Vforms are the only finite forms that precede clitic pronouns. Rivero and
Terzi offer the following explanation for these characteristics: Ôin Class I a strong V
feature in C encodes the logical mood of imperatives, and the V inflected with
Imperative morphology must raise overtly to check itÕ.
The fact that these imperatives are incompatible with negation follows
course, assume that the Neg cannot incorporate V. They can account for the second
characteristic of Class I imperatives by making the assumption that clitics adjoin to
the empty functional head which takes IP as a complement. Thus, VtoC yields the
28
Zanuttini, in one of her previous works (1991), has put forward a different explanation. Her view
rests on two core notions. The first is the claim that true imperatives lack a Tense projection
altogether, the second is what she explains as follows: Ôthe occurrence of certain functional elements
is parasitic on that of other functional projectionÕ, i.e. preverbal negation, Neg1P, is parasitic on the
presence of TP (Zanuttini 1994:125). (She abandons this view in Zanuttini (1994) on the basis that the
parasitic nature of Neg1P on TP is an extra stipulation.)
required VCl order. They explain the fact that this functional head does not act as a
barrier to VtoC by claiming that unlike Neg and C, this head does not have operator
features, or any kind of Abar characteristics, reducing the problem to Relativized
Minimality.
Class II imperatives, on the other hand, have no distinctive syntax. They can
be negated freely. They only precede the clitics, if otherwise the clitics were
sentenceinitial. In this case, any finite verb would precede the clitics, not only
imperatives. With respect to negation, they differentiate languages, such as Ancient
Greek, that do not allow Neg to incorporate V, from other languages, such as Serbo
Croat, that allow V incorporation to Neg . In Ancient Greek, negative imperatives are
29
possible, and they have a structure in which Neg projects a phrase immediately above
IP, and the V remains in IP. In SerboCroat, negative imperatives are also possible.
Here, Neg projects a phrase and allows the incorporation of V, thus the complex head
NegV can either stay there or raise further, for example to C to support a sentence
initial clitic (see Rivero and Terzi (1995:311 Ex (16)). For all Class II languages the
imperative features are licensed in IP.
Imperatives in Hungarian use the same inflectional endings that subjunctive
verbs. Contrary to subjunctives, they can appear in matrix clauses.
29
ÔIncorporationÕ seems to allow for the fact that the linear order is Neg+V.
The syntax of the imperative in Hungarian
As we can see from the examples in (52), the V precedes the VM, so it
imperative Vs. This is a characteristic of Class I languages.
(53) J—zsi, ne menj ki!
Joseph no goimp out
‘Joseph, donÕt go out!’
Example (53), on the other hand, suggests that Hungarian is not a Class I
language, since it freely allows negative imperatives. So, it seems that (53) argues
against a Vmovement analysis of imperatives in Hungarian . 30
4.2 Zanuttini (1994)
Let us consider Zanuttini (1994) for a second idea. She distinguishes ÔtrueÕ
ÔborrowedÕ from another paradigm, e.g. Hungarian imperative forms are the same
as the subjunctive ones. She claims that it is only the former class that is incompatible
with preverbal negation. According to Zanuttini (1994:122), it follows from RiveroÕs
account that (53) is an argument for a short Vmovement to a landing site below
NegP, for example T.
Zanuttini (1994) adopts the idea of Laka (1990) that true imperatives and
negative markers in Spanish are in complementary distribution because they occur in
the same position.
She assumes the following structure: [CP [PolP [FP1 [FP2 [FP3 [ ]]]]]]; The features
of the Polhead can be both positive or negative. ‘If a negative marker is present in
30
Let me briefly mention here, that (50) is also compatible with V incorporation into Neg, as Rivero
and Terzi (1995) argue for SerboCroat. But following the discussion of adjunction to Neg in Section
2.2.1. I reject this analysis.
the sentence and moves to PolP, either at Sstructure or at LF, it will match the
emphatic affirmation moves to PolP, the sentence will express emphatic affirmation.
If neither an overt negative element, nor a marker of emphatic affirmation are present,
the verb will move to PolP (at LF) and the sentence will be affirmative (without
special emphasis on affirmation).’ (Zanuttini, 1994:132).
She further assumes that in languages with preverbal negation, such as Italian
assumptions: FP1 is missing with true imperatives; in finite clauses with the
indicative or the subjunctive the V is leftadjoined to the head of FP3; pronominal
clitics are leftadjoined to the head of FP2. She claims that this gives the observed
word order cliticfinite verb. In languages with an infinitiveclitic order she assumes
that the infinitival V (or the gerund) has moved one step further and leftadjoined to
the head of FP1. This is also the case for those suppletive imperative verb forms that
missing, they leftadjoin to the head of PolP. Since in this case, and only in this case,
a negative marker cannot leftadjoin to the head of PolP, she derives the observed
data, namely that only true imperatives are incompatible with preverbal negation.
(Zanuttini, 1994:1337)
There is however a crucial problem with this analysis. She assumes that the
features of PolP are strong and does not check this feature before Spellout in
sentences that contain neither a negative element nor a true imperative V. Infinitives
are claimed to be adjoined to F1, indicative or subjunctive Vs to F3 or if they are
suppletive imperative forms to FP1, but nothing is claimed to adjoin to Pol. If the
The syntax of the imperative in Hungarian
features of Pol are strong the V should move to Pol before Spellout. However, in this
case it would not be possible to account for the following observation: ‘infinitives,
gerunds and certain suppletive imperatives exhibit the same word order as true
imperatives in the presence of pronominal clitics, but contrast with true imperatives in
their behaviour in the presence of a preverbal negative marker.’ (Zanuttini,1994:127)
4.3 Rivero (1994)
Rivero (1994) gives further arguments in favour of the claim that in the
Balkan languages, ÔtrueÕ imperatives move to C, while ÔsuppletiveÕ imperatives
move to T/Agr. She also claims that ‘verbs and auxiliaries raise to the category that
contains tense, person, number (and aspect) in all finite Balkan clauses.’ (Rivero,
1994:78).
embedded contexts. According to Rivero (1994) this argues for the fact that true
imperative Vs are in C. This argument is clearly not tenable in Hungarian where the
Cfield is much more articulated. There are no ÔtrueÕ imperatives in Hungarian, but
nevertheless Vmovement is possible in embedded contexts.
Second, she claims that the relative position of the V to clitics is symptomatic
of VtoC movement. She assumes that clitics are immediately above T/AgrP. In true
imperative constructions, clitics are obligatorily postverbal. In suppletive forms they
appear preverbally. In Hungarian, this argument can only be implemented if we find
some adverbs that are clearly above TP and under FP.
4.4 The proposed analysis of the imperative
Let us summarize the above. Rivero and Terzi (1995) claim that languages
that show Vraising with imperatives exhibit VtoC movement. In this case
incorporation to the Neghead on the way up is not allowed. Languages that do not
show a distinct syntax of imperatives have negative imperative constructions, either
by short Vmovement or by incorporation to Neg. In neither case does the imperative
V behave differently from nonimperative ones. Zanuttini (1994) tried to account for
the fact that only true imperatives are incompatible with preverbal negation. I
showed that her analysis is not satisfactory. Rivero (1994) argues that certain
imperative forms do not show VtoC, but only move to T, i.e. these forms do not
have a distinct syntax from other finite verbforms. She enumerates three arguments
in favour of this analysis none of which are easily adaptable for Hungarian.
Hungarian imperatives show a distinct syntactic behaviour (cf. (52)). I assume
sentential negation (cf. (53)) in which case the Neghead intervenes on the VtoF
path, thus V performs short movement to T, and Neg moves to F. The +Ffeature, that
triggers VtoF, reappears on T, triggering VtoT. Thus the V surfaces in F if the
sentence is not negated, otherwise it is in T.
The syntax of the imperative in Hungarian
4.5 Cooccurrence of the imperative and nonneutral elements
Let us see how the above analysis accounts for the word order in nonneutral
imperative sentences.
The following examples show that the imperative verb is indeed adjacent to a
focused constituent, as a VtoF analysis predicts.
JOSEPHacc sendimp
‘Send JOSEPH!’
(b) *JÓZSIT holnap küldd!
JOSEPHacc tomorrow sendimp
‘Send JOSEPH tomorrow!’
(c) *JÓZSIT gyorsan küldd!
JOSEPHacc quickly sendimp
‘Send JOSEPH quickly!’
Negation with imperatives is possible in three ways:
Joseph not goimp VM
‘Joseph, don’t go there!’
(b) JÓZSIT ne küldd oda!
JOSEPHacc not sendimp VM
‘Don’t send JOSEPH there!’
(56) (a) Ne JÓZSIT küldd oda(, hanem Mártát!)
not JOSEPHacc sendimp VM (but Marthaacc)
‘Don’t send JOSEPH there(, but Martha)!’
(b) Ne JÓZSIT küldd oda(, hanem magad menj!)
not JOSEPHacc sendimp VM (but yourself go)
‘Don’t send JOSEPH there, but go yourself!’
not JOSEPHacc not sendimp VM (but Marthaacc)
‘Don’t send not JOSEPH there, but Martha!’
(b) *Ne JÓZSIT ne küldd oda (, hanem magad (ne) menj!)
not JOSEPHacc not sendimp VM (but yourself (not) go)
‘Don’t send not JOSEPH there, but (don’t) go yourself!’
Following the discussion of similar nonimperative constructions in Section
2.4, and assuming the structure of Figure 5, I assume that the intervening Neghead
blocks VtoF, and the Vfeatures of F are satisfied by NegtoF movement. In this
case the V surfaces in T, as argued by Zanuttini (1994) and Rivero (1994). Prefocus
negation found in (56) and (57) is located in the NegP above FP. The strict adjacency
of the negative element and the V is due to the unavailability of [Spec, TP] and
adjunction to TP.
4.6 A special word order variation of the imperative
Since the T in the imperative sentence is +Tns, +F, we would expect VtoF to
be obligatory. However, there are several sentences in Hungarian which contradict
this analysis. Vmovement seems to be optional in the case of matrix imperatives. See
Examples (58a) and (58b).
The syntax of the imperative in Hungarian
the soupacc VMeatimp
‘Eat up the soup!’
(b) A könyvet elolvasd!
the bookacc VMreadimp
‘Read the book!’
We can try to account for the optionality in (58a) and (58b) along the lines of
Br—dyÕs analysis for infinitive constructions (see Examples (2a) and (2b)), by
saying that imperative utterances pattern with infinitives, in that their T is optionally
strong. However, the word order found in (58) appears to be different. Intonation
suggests that the V itself is in focus, i.e. adjoined to F.
There are essentially two ways to account for this phenomenon: first, by
claiming that the VM+V form a complex V, by incorporation or by other means. This
solution would, first of all, have to account for the fact that in all the other examples
of Vmovement mentioned so far, the VM did not raise. Also, this way we would
for the ungrammaticality of *A LEVEST megedd! (Eat up the SOUP!).
Let us turn to the second idea. If we allow the VM to undergo focusing, which
coincidence that VM and V are adjacent, they are in SpecHeadrelation because both
of them had to move to the checking domain of the FP.
The following examples get the same explanation with the slight difference
that because of the intervening Neghead, the V performs VtoT (and Neg moves to
T).
31
See Br—dy (1990:213) for the ‘doublyfilled Comp filter’: it is either the case that [Spec, FP] is
filled, and the bare V moves to F, or that VM+V, i.e. V+ moves to F, and [Spec, FP] must stay empty.
The syntax of the imperative in Hungarian
that sure that Joseph VM not holds that snake
‘It is sure, that Joseph won’t touch that snake.’
(b) Meg ne fogd!
VM not hold!
‘Don’t touch!’
that sevensacraments that Joseph VM not forgets the message
‘It is sure, that Joseph won’t forget the message.
(b) El ne felejtsd!
VM not forget!
‘Don’t forget!’
The only problem with this analysis is that nem cannot surface above the VM,
or if it can, it can only negate the VM. This is probably due to the fact, that some
VMs (e.g. meg) are semantically empty.
(61) *Ne ELfelejtsd!
not VMforgetimp
‘Don’t forget!’
(62) *Ne MEGfogd!
not VMholdimp
‘Don’t touch!’
(63) Ne FELmenj(, hanem le!)
not upgoimp (but down)
‘Don’t go UP(, but down!)
construction:
The syntax of the imperative in Hungarian
(64) Nem megmondtam, hogy bontott csirkét hozzál haza!
Not VM told , that...
‘Haven’t I told you , that...’
Szendroi Kriszta
5. Conclusion
My aim was to give a syntactic account of the subjunctive, the imperative and
partly the conditional mood in Hungarian. The word order found in such sentences is
to a large extent influenced by the presence or absence of nonneutral elements such
as focus and negation. Therefore, in the first two chapters I tried to give an account of
nonneutral indicative sentences. I adopted Bródy’s (1995) analysis for the focus and
structure (cf. Figure 5): the Hungarian sentence contains a Focusprojection, and two
Negationprojections, one above FP, the other above TP. The Vfeatures of F are
strong and the +F feature reappears on T, so the V surfaces in F in a finite sentence
that contains a focused XP, unless the lower NegP is projected. In the case of a post
focus negation, the V surfaces in T and the Neghead moves to F.
operator (Neg, Q, If, etc.) in the matrix C via a CTVCT dependency. Since this is
a syntactic dependency, it is sensitive to islands and shows parasiticgaplike effects. I
showed that the same dependency licences the Hungarian subjunctive mood as well. I
language, gives further support for Manzini’s theory.
distinct syntax, i.e. the V does not surface in the same position where it surfaces in
neutral sentences.
The V in the neutral Hungarian sentence does not move out of the VP. If a
VM is present, it strictly precedes the V. In nonneutral sentences, as well as in
imperatives, the V is claimed to move to F (or to T). This claim crucially relies on the
following assumption: the fact that the V precedes the VM in these sentences is
indicative of Vmovement (cf. Section 1.2.1).
Section 2.4), and on the interaction of these with the subjunctive (cf. 3.9) and the
imperative (cf. 4.5). I tried to account for a wide range of data including marginally
grammatical or ‘rare’ constructions (e.g. data in 2.6 and 4.6).
The main problem that remains to be explained is the technical problem that
because this feature is assumed to be weak. In my opinion, this technical problem
leads to deeper theoretical problems, such as the adequacy of the Minimal Link
discussed extensively in the current literature, see Bródy (1997) for a radically
different solution.
Szendroi Kriszta
Appendix
which is not discussed in the thesis. It is possible to account for a wide range of data
using the structure outlined below. However, there are problematic cases. Although
more work is needed on this analysis, it still seems to be worth mentioning it here.
An alternative analysis : no separate Negprojection
The analysis relies on the following statement by Haegeman (1995:127):ÔThe
NEGcriterion requires a Spechead relation between a negative operator and a head
with the feature [NEG]. This does not commit us to saying that each negative
sentence must contain a NegP. In the case of the WHcriterion, Rizzi does not
postulate that the Spechead relation between the WHoperator and the whhead be
realized on a specific whprojection. Rather, the WHfeature is hosted by I or by C. It
When a language has an overt negative head, Neg 0 projects into a functional
projection NegP.Õ
Let us assume that the +Negfeature is on F, and that it is ‘realized’ in the FP
in different positions if different parts of the FP are filled. Let us further assume that
the fact that the V appears in F in Hungarian is due to the fact that the Vfeatures of F
are strong, and following Bródy (1995) let us posit that the +F or +Negfeature re
appears on T.
FP
nem FP
Spec FÕ
Appendix
nem FÕ
F TP
Figure A1
90
Szendroi Kriszta
The structure in Figure A1 is developed on the basis of Laka’s (1990) SP . In 32
preferable, because it contains adjunction to an XP and an X’. However, it is worth
considering, because it captures the fact that the sentential operators Q, Neg, F, which
are known to form a natural class, can occupy the same syntactic position in the
sentence. Also, the technical problem of the strength of the Negfeature that arose in
Analysis B is solved: under this analysis, it is not contradictory to assume that this
feature is strong.
Let us see how Figure A1 accounts for basic data.
which it has to check against the +Negfeature on F. Thus, it moves to [Spec, FP] and
nem appears adjoined to FP. In this case, nem takes scope over [Spec, FP].
If no XP is focused, sentential negation triggers VtoF, because of a +Neg on
F and the reappearence of +Neg on T, and nem surfaces adjoined to FÕ. Most
probably [Spec, FP] is not projected.
(A1) Józsi nem szereti a zöldséget.
Joseph not likes the vegetableacc
‘Joseph does not like vegatables.’
Interaction of negation and focus
If an XP is preposed, sentential negation is still realized as a +Neg on F and
on T. However in this case, [Spec, FP] is projected, and nem either appears adjoined
32
A very similar analysis can also be found in Szabolcsi (1996).
Appendix
to FP (see (A3)), in which case it takes the focused constiuent in its scope because it
ccommands and precedes it, or appears adjoined to F’ (see (A2)), in which case the
focused constituent is outside the scope of negation.
92
Szendroi Kriszta
(A2) Józsi A ZÖLDSÉGET nem szereti.
Joseph VEGETABLES not likes
‘Joseph does not like VEGETABLES.’
(A3) Nem JÁNOST láttam.
not JOHNacc saw1s
‘I haven’t seen JOHN.’
If both constituent negation and sentential negation appear, two +Neg features
are checked, two nem surface: one adjoined to FP, the other to FÕ and some sort of
filter prevents the upper nem to take scope over the whole clause.
(A4) *Ne JÓZSIT ne küldd oda, hanem magad ne menj!
not JOSEPHacc not send there, but yourself not go!
‘It is not Joseph who you shouldn’t send there, but do not go yourself!’
Interaction of negation and the subjunctive
Recall from Section 3.9.2, that long operator movement sentences that contain
sentential negation have the following word order variations:
Joseph wants that I donÕtS upgive/ give up.
(c) J—zsi nem akarta, hogy feladjam.
(d) *J—zsi nem akarta, hogy adjam fel.
Joseph doesnÕt want that I upgive/ give up.
Appendix
Examples (A5a) and (A5b) suggest that Vmovement is obligatory in negative
embedded subjunctive clauses. This is not surprising, since this is the word order
found in simple sentences. However, (A5d) is ungrammatical. I assume that this is
possible because operators that are claimed to belong to one semantic class do not
necessarily behave exactly alike in syntax. Much the same way as Q and If do not
license the subjunctive or negative polarity items in Hungarian, but they do in Italian,
we can say that the +F feature is available at the foot of the subjunctive chain in
Hungarian, while the +NEG feature is not.
Interaction of negation and the imperative
The interaction of negation and the imperative provides a stong argument in
favour of the proposed structure in Figure A1 (cf. Section 4.5).
Negation with imperatives is possible in three ways:
(A6) Józsi, ne menj oda!
‘Joseph, don’t go there!’
JÓZSIT ne küldd oda!
‘Don’t send JOSEPH there!’
(A7) Ne JÓZSIT küldd oda(, hanem Mártát!)
‘Don’t send JOSEPH there, but Martha!’
Ne JÓZSIT küldd oda(, hanem magad menj!)
‘Don’t send JOSEPH there, but go yourself!’
(A8) Ne JÓZSIT ne küldd oda (, hanem Mártát!)
‘Don’t send not JOSEPH there, but Martha!’
*Ne JÓZSIT ne küldd oda (, hanem magad (ne) menj!)
‘Don’t send not JOSEPH there, but (don’t) go yourself!’
94
Szendroi Kriszta
I assume that there is VtoF movement in imperative sentences even if they
are negated. The possibility of negated imperatives in a long Vmovement analysis
can be explained under the structure proposed in Figure A1, because there is no
intervening Neghead; Neg is not projected, it is a feature on F.
A problem: position of Nwords
Recall from Section 2.6 É.Kiss’ claim that negative adverbials are in [Spec,
FP] and Puskás’ claim that Nwords are also in [Spec, FP]. As Bródy pointed out, this
is not necessarily the case since the following sentences are grammatical:
nothingacc at all TO MARY gave3p VM
‘They did not sell anything TO MARY.’
(b) Semmit se MARINAK nem adtak el.
nothingacc at all TO MARY not gave3p VM
‘There wasn’t anything that they did not sell TO MARY.’
Under the structure proposed in Figure A1, these sentences are difficult to
explain. The only possibility is to claim that semmit is in fact in Topicposition, but
this analysis is not tenable because of semantic reasons. The topic is the logical
subject of the sentence, and an Nword cannot act as one. Nwords cannot be
topicalized.
Szendroi Kriszta
Szendrši Kriszta
97
References
Manzini, M. Rita (1994), Sentential Complementation: the Subjunctive, Ms, University College
London.
Manzini, M. Rita, and Leonardo M. Savoia (1998), Negation parameters and their interaction in Italian
dialects, Ms, University of Florence.
May, Robert (1985), Logical Form, its Structure and Derivation, MIT Press, Cambridge, Mass.
Pollock, JeanYves (1989), Verb Movement, Universal Grammar, and the Structure of IP, Linguistic
Inquiry 20, 365424.
Pollock, JeanYves (1993), Note on Clause Structure, Ms, Amiens.
Pusk‡s, Genoveva (1994), Sentential negation in Hungarian, Rivista di Linguistica 6.1, 538.
Rivero, Mar’a Luisa (1994), Clause Structure and Vmovement in the Languages of the Balkans,
Natural Language and Linguistic Theory 12, Kluwer, 63120.
Rivero, Mar’a Luisa and Arhonto Terzi, (1995), Imperatives, Vmovement and logical mood, Journal
of Linguistics 30, CUP, 301332.
Rizzi, Luigi, (1995), The Fine Structure of the Left Periphery, Ms, University of Geneva, October
1995.
Roberts, Ian, (1997), Comparative Syntax, Arnold, London.
Zanuttini, Raffaella (1994), Speculations on negative imperatives, Rivista di Linguistica 6.1, 6789.
Mood in Hungarian
Igemód a magyarban
SZAKDOLGOZAT
Írta Szendroi Krisztina
ELTE BTK Elméleti Nyelvészet szak
Témavezeto: É.Kiss Katalin tudományos tanácsadó
1998
SZAKDOLGOZAT
Írta Szendroi Krisztina
ELTE BTK Elméleti Nyelvészet szak
1998