Problems in the left periphery of Hungarian infinitival clauses
Krisztina Szécsényi
0 Abstract
Hungarian infinitival clauses are claimed to project a full CP with all the left peripheral
projections of a finite clause present, and the very same restrictions on the order of these
constituents. However, in certain infinitival structures the order of typical left peripheral
phrases fails to obey the expected restrictions, moreover, matrix constituents can mingle with
constituents of the infinitival clause suggesting a monoclausal structure. Adopting
Hinterhölzl’s (1999) biclausal approach to restructuring, I claim that there is a unitary
treatment for these seemingly different structures.
1 The Left Periphery in Hungarian
Hungarian is a distinctly discourse-configurational language where “both topic/comment and
focus/background divisions are reflected in surface syntax” (e.g. É. Kiss 1995; Surányi 2002).
Once present, topic and focus appear in functional projections of their own, TopP and FocP
respectively, having distinct syntactic properties: whereas the TopP projection is recursive,
FocP is not (at least not in the preverbal domain). Topics are fronted to the left with an
optional phonological stress, foci are fronted to the immediate left of the verb bearing
emphatic stress. As far as their discourse properties are concerned, topics are strictly discourse
old, while foci may be discourse-old or discourse-new.
Out of the current syntactic models where we find functional projections for topic and
focus in the C-domain (e.g. Szabolcsi 1997; É. Kiss 1998a, b; Puskás 2000) in the present
paper I adopt Szabolcsi’s (1997) approach following Beghelli & Stowell’s (1994/1995, 1997)
theory of scope, defining a strict order of topic, quantifier and focus constituents in the
preverbal field, in this order.
2 Hungarian Infinitival Constructions
As argued by Kenesei (2002) and Dalmi (2004), Hungarian nonfinite constructions in general
must be classified as clauses for the following reasons:
i. infinitival complements have TOPP and FP slots in the left periphery, where arguments of
the infinitival verb can land:
(1)
András meg-tanította a diákokat
[CP[TOPP a mondatot
Andrew pfx-taught
the students.acc
the sentence.acc
[FP csak LFG-ben
[AGRP elemezni
PRO]]]].
only LFG-iness
analyze.inf
‘Andrew taught the students to analyze sentences ONLY IN LFG.’
ii. clausal NEGP is projected within infinitival complements:
(2)
Szeretném a pénzemet
nem el-költeni
PRO a hónap végéig.
Like
the my.money.acc not pfx-spend.inf
the month end
‘I would like not to spend my money until the end of the month’
iii. the order of constituents in the preverbal domain of the infinitival clause shows the same
restrictions we find in finite clauses (Kenesei 2002):
(3)
a. Mikor vizsgálod meg [[a minden könyvet csak szemüveggel olvasó] diákokat]?
when examine pfx the every book.acc only with.glasses read
students.acc
‘When will you examine the students reading every book only with glasses?’
b. *Mikor vizsgálod meg [[a csak szemüveggel minden könyvet olvasó] diákokat]?
As the examples indicate, the C-domain is present in infinitival complements as well,
with the same restriction on constituent order as we find in finite clauses (3a,b). However, the
following sentences seem to challenge the idea that infinitival complements contain a fullfledged clausal projection-line. Whereas the grammaticality of (4a) has a lower status, with a
focused constituent present in the sentence (4b) (where the word in capital letters indicates
focus interpretation), the reverse order of the quantifier and the Topic Phrase becomes fully
legitimate. (In fact, (4b) is ambiguous, ‘tegnap’ can also be understood as referring to
‘wanting’, but the dominant interpretation is clearly the one where ‘tegnap’ modifies ‘adni’.)
Moreover, with a focused constituent present, the subject of the main clause (the underlined
constituent in (4c)) can freely scramble with constituents of the infinitival clause indicating a
monoclausal structure.
(4)
a. ?Péter oda akar [QPminden könyvet] [TopP Marinak] adni.
Peter pfx want every book.acc
Mary.dat give.inf
‘Peter wants to give every book to Mary.’
b. Péter TEGNAP
akart minden könyvet Marinak adni.
Péter YESTERDAY wanted every book.acc Mary.dat give.inf
‘Peter wanted to give every book to Mary YESTERDAY.’
c. TEGNAP
akart minden könyvet Péter oda-adni
Marinak.
YESTERDAY wanted every book.acc Peter pfx-give.inf Mary.dat
‘It was yesterday that Peter wanted to give every book to Mary.’
In the Hungarian linguistic literature we do find evidence that the presence or absence of
focus may result in reanalysis. In her analysis of verbal complex formation É. Kiss (1999)
argues that different types of complex predicates are the result of the presence or lack of
reanalysis in a structure, which, in turn, is conditioned by the presence or lack of a focused
constituent. She defines reanalysis on the following lines:
(5)
Reanalysis: In a construction in which XP is an immediate constituent of a Y
projection, and YP is an immediate constituent of a Z projection, XP is reanalyzed as
an immediate constituent of the Z projection if the phrasal boundary YP is deleted (or
is made transparent).
Sentence (6) illustrates reanalysis where the particle szét ‘apart’, originally belonging to
the verb szedni ‘take’ forming the complex verb szétszedni ‘to take apart’. The particle
obligatorily moves to the position preceding the finite verb fogja ‘will’
(6)
János szét fogja akarni kezdeni szedni a rádiót.
John apart will want.inf begin.inf take.inf the radio.acc
‘John will want to begin to take apart the radio.’
However, the properties of reanalysis occurring in embedded infinitives are different
from the reanalysis taking place in verbal complexes. In verbal complexes, reanalysis is
driven by the light verb’s need to avoid getting stressed, therefore, there is a difference
between the behavior of light verbs like fog ‘will’ and verbs like imád ‘adore’, which can be
stressed. If a focused constituent precedes a light verb, reanalysis is unnecessary, for the
focused constituent saves the verb from getting stress. This is not the case with the embedded
infinitival clauses under discussion: it is exactly the focus present in the sentence which
results in reanalysis and there is no difference in the behavior of stress-avoiding light verbs
and heavy verbs:
(7)
PÉTER imád minden könyvet Marinak (oda)-adni.
Peter love every book.acc Mary.dat (pfx)-give.inf
‘It is Peter who loves giving every book to Mary.’
3 The Scrambling Approach
In the following I point out that the word-order presented under (4c) is not the only
grammatical order, which may shed further light on what exactly is going on in this sentence.
(8)
a.
b.
c.
d.
TEGNAP akart Péter Marinak minden könyvet odaadni.
TEGNAP akart Marinak Péter minden könyvet odaadni.
?TEGNAP akart Marinak minden könyvet Péter odaadni.
TEGNAP akart Marinak minden könyvet odaadni Péter.
Based on the (sometimes slightly reduced) grammaticality of these sentences we can
conclude that the focus appears in the focus position preceding the finite verb, the order of the
rest of the constituents is essentially free. This seems to indicate that we are dealing with a
monoclausal structure here, where the focused constituent is directly followed by the verb,
with no further restriction on the order of elements in the postverbal field.
In her discussion of the nature of free word order phenomena in Hungarian É. Kiss
(2003) differentiates three relevant fields of the Hungarian sentence:
i. In the argument field word order is essentially free, influenced only by the specificity
feature of the arguments to a certain extent. (This can explain why the grammaticality of the
sentences indicated is degraded.) The postverbal free order is the consequence of a flat VP in
which the arguments mutually c-command each other.
(9)
Tegnap felhívta Pétert
egy ügyvéd.
yesterday called Peter.acc a lawyer
‘Yesterday a lawyer called up Peter.’
(10) ?Tegnap felhívta egy ügyvéd Pétert.
ii. In the operator field we find a strictly ordered hierarchy of aspectualizer, focus, and
distributive quantifier positions, where the placement of constituents determines both their
operator function and their scope.
iii. The initial topic field is different from the operator field, lacking any scopal effect due to
the [+referential], [+specific] interpretation.
Based on É. Kiss (2003) and the sentences listed in (8) it seems that these structures
should be analyzed as monoclausal.
A further argument for the monoclausal treatment of certain structures with infinitives
may come from the nature of scrambling. É. Kiss (1992) claims that scrambling is the base
generation of the arguments of a V in a random order under the V’ projection, and if it is so, it
must be local. If this requirement is observed; a constituent of V’ cannot be scrambled under a
higher V’ node. This accounts for the ungrammaticality of (12) and seems to provide evidence
for a monoclausal analysis of (14) and (15): as a result of reanalysis the arguments of the
infinitive are reanalyzed as arguments of the main verb. They appear under the same V’
projection, so they can scramble relatively freely (14), more so if there is a focused constituent
present (15):
(11)
Szeretnék a lányok, ha bemutatnám Jancsit Juliskának.
would.like the girls
if introduced John.acc Julie.dat
‘The girls would like if I introduced John to Julie.’
(12)
*Szeretnék Jancsiti a lányok [ha bemutatnám Juliskának ti]
would.like John.acc the girls
if introduced Julie.dat
(13)
Szeretnék a lányok bemutatni Jancsit Juliskának.
would.like the girls introduce.inf John.acc Julie.dat
‘The girls would like to introduce John to Julie.’
(14)
?Szeretnék Jancsit a lányok bemutatni
Juliskának.
would.like John.acc the girls introduce.inf Julie.dat
(15)
HOLNAP szeretnék Jancsit a lányok bemutatni Juliskának.
tomorrow would.like John.acc the girls introduce.inf Julie.dat
‘It is tomorrow that the girls would like to introduce John to Julie.’
An analysis in terms of reanalysis of arguments taking place followed by scrambling
could account for the problematic sentence in (4c), too.
4 Further Data to Consider
4.1 Different Types of Foci
Whereas holnap ‘tomorrow’ can be understood as focus only if it occupies the focus position
preceding the finite verb, there are other constituents that seem to appear in the focus position
of the non-finite verb. The only-phrase in (16) seems to occupy the focus position of ebédelni
‘to have lunch’:
(16)
Péter szeretne ezután
csak étteremben ebédelni.
Peter would.like from.now.on only in.restaurant have.lunch.inf
Both holnap ‘tomorrow’ and the only-phrase can appear in the sentence, and the
ungrammaticality of (18) suggests that here we are dealing with a structure where holnap
‘tomorrow’ occupies the focus position of the finite verb and csak étteremben ‘only in
restaurants’ occupies the canonical focus position directly preceding the infinitival verb.
(17)
HOLNAP szeretne Péter csak étteremben enni.
Tomorrow would.like Peter only in.restaurant eat.inf
‘Peter would like to eat only in restaurants TOMORROW.’
(18)
*HOLNAP szeretne csak étteremben Péter enni.
However, the every-phrase can directly precede the infinitive in these structures, too,
stranding the focused constituent from the verb (19) and in this case even Peter can scramble
into the clause (20). The reduced grammaticality of (21) can be explained by the violation of
the specificity requirement discussed in §4.2:
(19)
HOLNAP szeretne Péter csak a McDonaldsban minden fogást kipróbálni.
Tomorrow would.like Peter only the in.McDonalds every dish.acc try.inf
‘Peter would like to try every dish in McDonalds’ TOMORROW.’
(20)
HOLNAP szeretne csak a McDonaldsban Péter minden fogást kipróbálni.
(21)
?HOLNAP szeretne csak a McDonaldsban minden fogást Péter kipróbálni.
Sentences (20) and (21) indicate that the only-phrase does not occupy a focus position
preceding the infinitive. Further support for the claim that it is not always the case that
whenever we have two foci they belong to different verbs may come from the following
sentences:
(22)
TEGNAP kezdett el PÉTER a gyerekeknek énekelni.
yesterday started pfx Peter the children.dat sing.inf
‘PÉTER started to sing to the children YESTERDAY.’
(23)
*TEGNAP kezdett el a gyerekeknek PÉTER énekelni.
The ungrammaticality of (23), where an extra argument of the infinitive is inserted,
clearly shows that the second focus does not appear in a focus position related to the
infinitive. Rather, it can be an instance of postverbal focus, when focus occupies a focus
position following the verb. To conclude this section we can say that the presence of two foci
in this structure does not call for a biclausal analysis, either (though it has to be pointed out
that the ungrammaticality of (18) is left unexplained).1
4.3 Szabolcsi-Koopman (2000)
As many have noted in the literature (e.g. Brody 1990, 1995; É. Kiss 2002; Koopman &
Szabolcsi 2000), one important difference between Hungarian finite and infinitival clauses is
that whereas in finite clauses focus and negation cannot be separated from the verb by its
prefix or verbal modifier, in infinitival clauses it can be:
(24)
Péter nem ment haza.
Peter not went home
‘Peter did not go home.’
(25)
*Péter nem haza ment.
1
But the grammaticality of (i) seems to reject a biclausal treatment.
(i) Holnap szeretne csak a McDonaldsban ebédelni
Péter
tomorrow would.like only the in.McDonalds have.lunch.inf Peter
‘Peter would like to try every dish in McDonalds’ TOMORROW.’
Peter not home went
‘Peter did not go home.’
(the sentence is grammatical if haza ‘home’ is understood as the focus of the
sentence)
(26)
Jobb lenne
nem haza menni.
better would.be not home go.inf
‘It would be better not to go home.’
(27)
Jobb lenne
nem menni haza.
better would.be not go.inf home
(same)
(28)
Jobb lenne
csak KEDDEN haza-menni.
better would.be only on.Tuesday home-go.inf
‘It would be better to go home only on TUESDAY.’
(29)
Jobb lenne
csak KEDDEN menni haza.
better would.be only on.Tuesday go.inf home
(same)
According to Szabolcsi & Koopman (2000) “proof that nem ‘not’ and csak kedden ‘only
Tuesday’ are in the same position as in finite clauses comes from the fact that in neither case
can anything simply intervene between them and haza-menni/menni haza ‘home-go.inf/go.inf
home’”.
To account for the difference between finite and non-finite clauses Brody (1995)
proposes that in finite clauses V-to-F movement is obligatory, whereas in non-finite clauses it
is optional. This analysis supports Koopman & Szabolcsi (2000) in predicting that focused
constituents will occupy the same position in finite and non-finite structures.
While it may seem true in most of the cases, (30) shows that an every-phrase can very
easily intervene, the result is clearly grammatical:2
(30)
Jobb lenne
csak KEDDEN minden előadást
meg-nézni.
better would.be only on.Tuesday every performance.acc pfx-watch.inf
‘It would be better to watch every performance only on Tuesday.’
Moreover, when an every-phrase is inserted, prefix-verb inversion leads to
ungrammaticality, indicating unanimously that the only-phrase cannot occupy a focus position
related to the infinitive:
2
Of course, it is grammatical to say
(i) Jobb lenne minden előadást csak kedden megnézni.
‘It would be better to watch every performance only on Tuesday.’
where the quantifier is followed by the focused constituent, but there is a difference in meaning in
terms of scope. In (i) we have a certain number of performances all of which we would like to watch
on Tuesday, whereas (30) means that we could watch every performance on Tuesday, every
performance on Wednesday, etc, and we would like to watch every performance only on Tuesday, not
on other days.
(31)
*Jobb lenne
csak KEDDEN minden előadást
nézni
meg.
better would.be only on.Tuesday every performance.acc watch.inf pfx.
At this point it has to be noted that the word order attested in the infinitival clause in
(30) is never grammatical in a finite clause with focus, exactly because in that case the
focused constituent has to occupy the canonical focus position preceding the verb. Inversion
of the prefix and the verb is obligatory:
(32) a. *Csak KEDDEN minden előadást megnéztem.
b. *Csak KEDDEN minden előadást néztem meg.
c. Csak KEDDEN néztem meg minden előadást.
Proof that the focus, the quantifier, and the verb form a constituent comes from (33),
which makes it clear that in this case we are not dealing with an instance of postverbal focus
(more so because we find postverbal focus only when we also have a preverbal one):
(33)
[Csak KEDDEN minden előadást
megnézni] nem volt jó ötlet.
only on.Tuesday every performance.acc watch.inf not was good idea
‘It was not a good idea to watch every performance only on Tuesday.’
4.4 An Account of Focus Without V-to-F Movement: Kenesei (1994)
An alternative to V-to-F movement may be the assignment of the focus feature in the lexicon,
or random focus feature assignment in syntax as proposed by Kenesei (1994). The following
data support this claim:
(34)
Mari [a CIKKEKET OLVASÓ fiút]
látta.
Mari the articles.acc reading boy.acc saw
‘It’s the boy READING THE ARTICLES that Mary saw.’
(35)
Mari [a KÖNYVTÁRBAN olvasó fiút]
látta.
Mari the in.library
reading boy.acc saw
‘It’s the boy reading IN THE LIBRARY that Mary saw.’
In sentences (34) and (35) only part of the NP is focused. As Kenesei (2001) points out,
adopting Selkirk’s Focus Percolation Rule (Selkirk 1986) will easily account for how the NP
gets the focus feature even we assume the assignment of the [+focus] feature in the lexicon or
randomly in syntax:
(36)
Focus Percolation Rule: the feature [+focus] percolates onto the lexical head (and/or
the lexical head of an argument) of the XP it is assigned to. (Selkirk 1986)
If we adopt this approach an analysis in terms of V-to-F movement is unnecessary. A
further arguments provided by Kenesei (2001) is the fact that the focused phrase is not always
adjacent to V: there are at least two operator particles that may intervene between the focused
phrase and the verb stem: nem ‘not’ and emphatic is ‘too’.
This way of focus assignment should be available at least as a parametric option: e.g.
the only-phrase, which bears an inherent focus feature could have the option of not moving to
an FP directly preceding the verb.
In what follows I give a short summary of restructuring in German and have a look at
recent accounts of Hungarian infinitival clauses.
5 Restructuring in German and Hungarian
5.1 Restructuring in German
Wurmbrand (2001) differentiates restructuring vs. non-restructuring predicates in German
based on the following criteria:
(a)
(b)
(c)
(d)
(e)
the infinitival clause contains only the VP-layer, therefore it has
no embedded subject
no embedded tense
no embedded negation
no embedded structural case
Actually, in her work we do not find a binary opposition of restructuring vs. nonrestructuring predicates, the distinction is more fine-grained. Restructuring is further classified
into subcategories of lexical and functional restructuring, and the criteria above pertain to the
strictest type of restructuring only. The crucial difference between restructuring and nonrestructuring predicates is that non-restructuring predicates take CP complements, whereas
restructuring predicates take a reduced complement which can be as small as a mere VP.
5.2 Restructuring in Hungarian
Tóth (2000) claims that Hungarian and German modals involve different syntactic structures,
the former taking clausal infinitival complements, the latter only a vP-complement. If
restructuring is defined by the criteria in Wurmbrand (2001) we find only four restructuring
verbs in Hungarian: látszik ‘seem’, szokott ‘usually does’, talál ‘happen to’, fog ‘will’.
Interestingly enough, this list happens to include those verbs Kenesei (2001) identifies as the
auxiliary verbs in the Hungarian language.
However, there are other structures, too, where we find evidence for reanalysis taking
place, as pointed out by Puskás (2001):
(37) a. Nem akarok senkit
meg-hívni.
Neg want nobody.acc pfx-invite.inf
‘I don’t want to invite anybody.’
b. Senkit nem akarok meghívni.
(same)
[Tóth, 1995]
Based on the grammaticality of (37b) we have to conclude that ‘some infinitives can
undergo some sort of reanalysis in Hungarian, giving rise to the apparent counterexamples to
the clause-boundedness of n-words. Greek exhibits the same phenomena on the effect of
restructuring and infinitival domains on n-word licensing’ (Farkas & Giannakidou 1996).
The grammaticality of (37b) above clearly indicates that restructuring is not restricted to
the main predicates látszik ‘seem’, szokott ‘usually does’, talál ‘happen to’, and fog ‘will’ in
Hungarian. Returning to the sentences under discussion in this paper, it also seems obvious
that restructuring in Hungarian is not the property of certain predicates, but rather structures
with embedded infinitives where a focused constituent is present.
6 Monoclausal or Biclausal Structure?
6.1 Other Approaches to Restructuring
The approaches to restructuring presented above all assume that when restructuring takes
place the complement of a restructuring infinitive is a structure smaller than a CP, which is
motivated by the facts that restructuring infinitives involve the formation of verbal complexes
and are transparent for several types of extraction processes. Some analyses assume that it is
due to only one clause present in the structure. According to these monoclausal accounts3
verbal complexes can be base-generated and involve only one licensing domain for arguments
and adverbs (monoclausal Mittelfeld).
Biclausal accounts argue that each verb in a restructuring infinitive projects its own
clause (full or reduced) and assume special restructuring processes to account for the
monoclausal properties of restructuring infinitives, such as Verb Raising, Verb Projection
Raising to account for the formation of verbal complexes, and Long Distance Scrambling to
account for the transparency effects.
However, very often it is the case that a structure shows monoclasual and biclausal
properties at the same time, as it happens in Dutch and Hungarian. Motivated by this,
Haegeman & Riemsdijk (1986) and following them É. Kiss (1987) claim that reanalysis gives
rise to a multidimensional representation: one dimension to represent the biclausal properties,
one dimension to represent the monoclausal properties of restructuring infinitives.
While such a representation captures the essence of a restructuring construction, it
merely gives a description of the facts, it fails to give a real explanatory account.
6.1.1 Hinterhölzl (1999)
Hinterhölzl (1999) adopts a biclausal approach to reanalysis, but he claims that Verb Raising
(VR), Verb Projection Raising (VPR) and Long Distance Scrambling do not account for every
aspect of restructuring. In his analysis restructuring breaks down into movement of the
infinitival VP and the infinitival TP into designated positions in the matrix clause. These
movements are cyclic and happen for licensing purposes. Hinterhölzl shows how these
movements can be derived from a general theory of sentential complementation, assuming
that “the complementizer is essential for rendering a sentential complement (a TP) into an
argument. (...) the event denoted by an infinitive is not anchored with respect to a local Tensepredicate but is directly dependent in its interpretation on the matrix verb” (Hinterhölzl 1999:
65). It is this dependency between the matrix verb and the infinitival Tense-head that is
mediated by the complementizer. “[T]he complementizer in general serves as a place holder
for the various conditions that a verb imposes on its embedded clause. In particular, (...) the
complementizer serves as a place holder for the l-selectional features of the higher verb and
thus becomes essential for licensing the infinitival V(P) as well.” The motivation of
movement in Verb Raising and Verb projection Raising infinitives is the presence of a
defective complementizer. For this reason, the temporal dependency between the matrix verb
and the infinitive must be fixed by a movement relation.
The novelty of the approach lies in pointing out that replacing the VR and VPR-analyses
with a VO-based analysis to West-Germanic offers a more adequate account of the position of
participial, infinitival and IPP (infinitivus pro participio) complements. At the same time, the
striking similarity between the verbal complexes of West-Germanic and those of Hungarian, a
VO-language, becomes much more straightforward.4
3
For a detailed reference list of monoclausal and biclausal accounts see Wurmbrand (2001).
there have been several attempts to account for the similarity of verbal complex formation in WestGermanic and Hungarian. We also find proposals for analyzing Hungarian as an SOV language: with
regard to preverb climbing phenomena: Ackema (2004) argues that an analysis of Hungarian as an
SOV language could account for why we find so many parallels between free word order phenomena
of German and Dutch and those of Hungarian. Corver & Riemsdijk (1997) state that “scrambling is
4
In this biclausal approach monoclausaul properties of restructuring are accounted for by
T-to-T head-movement that has the effect of unifying the two TP-domains (Roberts 1997). Tincorporation in restructuring contexts creates a single Extended Projection uniting two
clauses, where arguments of the embedded clause may freely undergo A-movement into the
matrix clause.
6.1.1.1 West-Germanic
In standard Dutch certain verbs taking infinitival complements trigger what is traditionally
called Verb Raising:
(38)
dat ik Jan een liedje hoor zingen [Hinterhölzl, 1999, p. 14]
that I Jan a song hear sing
‘that I hear Jan sing a song
The rightward movement of the infinitive was first proposed by Evers (1975) to
accomodate this order with the OV-based approach to Dutch:
(39)
dat ik [Jan een liedje tR] hoor zingenR
In West Flemish we find Verb Projection Raising where we find VR in standard Dutch,
the difference between the two being that in VPR it is not only the verb that can be moved, but
also V’, or the full VP together with clitics, adverbs and sentential negation, therefore a VPRstructure must be at least as big as an IP:
(40) a. da Marie Jan nen boek tR wilt gevenR
that Marie Jan a book wants to give
‘that Marie wants to give Jan a book’
b. da Marie Jan tR wilt [nen boek geven]R
that Marie Jan wants a
book give
c. da Marie tR wilt [Jan nen boek geven]R
that Marie wants Jan a book give
d. da Valere wilt [ze morgen nen boek geven]
that Valere wants her tomorrow a book give
‘that Valere wants to give her a book tomorrow’
In what follows, due to lack of space, I will focus only on the main arguments of
Hinterhölzl (1999). The author adopts Kayne’s (1994) antisymmetry hypothesis assuming that
frequent in head-final languages, and rare or even absent in head-initial languages”. Scrambling in
head-initial languages is very often found when the head-initial classification of the given language
can be questioned easily: Hungarian as a free word-order language where it is relatively hard to define
the base word order could be mentioned as one of the examples. At this point the question arises
whether Hungarian could be an SOV language. If some evidence were found that Hungarian was
misclassified as far as its headedness-parameter is concerned, the phenomenon of scrambling available
in the language would be accounted for. On independent grounds Marácz (1989) also claims that
Hungarian is an SOV language.
For a VO-analysis of Dutch see Zwart (1993).
all phrase structure is right-branching and movement is only to the left. The approach has
welcome consequences for the problems resulting from the OV approach as well, and the
theoretical considerations which may seem stipulative at first sight can also be supported with
empirical evidence.
A problematic aspect of both OV and Vo approaches is how to account for the position
of the infinitival marker. An infinitival sequence, like zu lesen, ‘to read’ seems to be a headinitial structure. There are two ways to accomodate it with an SOV approach: assuming either
that the infinitival marker “is not an independent head but rather a verbal affix (Haider 1993)
(...) or that the infinitival verb undergoes rightward head-movement to (right)-adjoin to the
head-final infinitival marker in the IP-domain.
What is problematic for a VO approach in this respect is how to account for the leftward
movement of APs, PPs and other non-DP VP-internal constituents.5 Support for the
antisymmetry approach comes from West Flemish and Afrikaans:
(41)
West Flemish:
mee Valere te [willen [dienen boek kuopen]] een
with Valere to want that book buy
have
‘with Valere having wanted to buy that book’
(42)
Afrikaans:
Die banke moes oop gewees het, om dit gister
te [kan betaal] het
the bank should open been have
it yesterday to can buy have
‘The bank should have been open to have been able to buy it yesterday.’
The fact that material can intervene between the infinitival marker and the infinitival
verb unambiguously shows that the infinitival marker cannot be a verbal affix. Movement of
the verb cannot be head movement, either, as in (41) it contains a phrase-size constituent
(dienen boek ‘that book’). Therefore, the conclusion has to be drawn that infinitival structures
are head-initial heading a functional projection of their own which Hinterhölzl labels F1P.
Returning to the problem of leftward movement let us consider the following sentence:
(43)
dat Jan [Marie het boek morgen] wilde [geven]
that Jan Marie the book tomorrow wanted give
‘that Jan wanted to give Marie the book tomorrow’
The example shows that DPs and adverbs precede the selecting verb, whereas the
infinitive itself (and, when present, the sentential complement of the infinitive also) follow it.
Hinterhölzl (1999) argues that although the simplest way to account for this structure would
be to say ‘that the bracketed constituents preceding the matrix verb have been moved
individually via scrambling from the embedded clause into the matrix clause’, this solution is
untenable for the following reasons:
– verb particles, small clause predicates and idiomatic expressions cannot argument-scramble,
but can appear in VR constructions.6
– adverbs do not scramble. They can only undergo movement as part of a larger phrase.
5
The Case-licensing movement of DP arguments is completely justified in the minimalist framework.
These constituents can undergo focus-scrambling, however, it goes together with a special intonation
pattern which is lacking in VR structures.
6
Having disqualified the scrambling analysis Hinterhölzl bases his analysis on the
following assumptions:
i. coherent infinitives are full CP complements.
ii. a CP has the following structure:
(44)
L NPs [S-Advs [Neg [S-NPs [VP-Advs [Pred0 [F1P zu [F2P V [F3P CP [VP]]]]]]]]]]7
Now let us see how we get from an underlying structure in (a) to the sentence in (b),
where everything but the infinitive itself has been moved (and not scrambled) from the
embedded clause to the matrix clause:
(45) a. [dat vaak [VP Jan wil [CP ... [lang [F1P Ø [VP PRO lezen het boek]]]]]]
b. dat Jan het boek vaak lang wil
lezen
that Jan the book often long wants read
‘that Jan often wants to read the book for a long time’
The derivational procedure is the following:
(46) Step 1: the nominal arguments leave the VP for reasons of licensing before Spell-out
(short scrambling):
[dat vaak Jan [wil [CP [TP PRO het boeki lang [F1P [VP lezen ti]]]]]
Step 2: F1P that has been emptied up to the verb is moved into [Spec, CP] in the
embedded clause:
[dat vaak Jan wil [CP [F1P lezen ti] [TP PRO het boeki lang tF1P]]]
Step 3: the remaining TP of the infinitival is moved into [Spec, PredP] in the matrix
clause:
[dat vaak Jan [PredP [TP PRO het boek lang tF1P] wil [CP lezen tTP]]]
Step 4: the infinitival Tense-head head-adjoins to the matrix Tense-head unifying the
two TP-domains by creating a single Extended Projection (Roberts 1997).
Step 5 (optional): T-to-T head-movement opens the way for long-scrambling of the
matrix subject and the embedded direct object:
[dat Jani het boeki vaak ti [PredP [TP PRO tj lang] wil lezen]]
In German the dependent infinitive usually precedes the infinitive, one extra step is
necessary to account for this difference between German and Dutch, namely that after F1Pmovement into [Spec, CP] the infinitive overtly moves to the left of the matrix verb, leaving
behind its CP-complement. In Dutch this is assumed to be a covert operation.
7
L-NP stands for Long-Scrambled NPs (moving for reasons of specificity) and corresponds to the
traditional TP. S-NP stands for Short-Scrambled NP (nominals move here for reasons of purely formal
licensing), S-Adv for sentential adverbs. F1P is the position of the infinitival marker and moved
particle-phrases. PredP hosts Small Clauses, idiomatic expressions and directional PPs. F2P is a
functional projection dominating the licensing projection for CPs and CP complements are licensed in
F3P.
As far as West Flemish is concerned Hinterhölzl claims that “VR- and VPRconstructions are basically the same differing only in the amount of structure that can be piedpiped by F1P-movement into [Spec, CP] of the infinitival.”
To sum up, the VO-approach has several advantages over an OV-analysis, among others
that it does not have to make use of right-adjunction, it provides a relatively simple solution
for the distribution of adverbs and CP-complements in VR, and the differences between
Dutch, German and West Flemish can be easily derived. Hinterhölzl convincingly argues that
especially in bigger verb-clusters assuming TP-movement to take place is essential even in an
OV approach, so the VO approach proves to be less costly, as it can do without the
extraposition operation.
6.1.1.3 West-Germanic: Monoclausal or Biclausal?
Arguments for a biclausal analysis come from i) the positioning and interpretation of temporal
adverbials (47) and ii) binding facts (48).
(47) a. weil Peter mich shcon lange
heute besuchen wollte
since Peter me already for-a-long-time today visit
wanted
‘already for a long time has Peter wanted to visit me today’
b. *weil mich Peter schon lange
heute besucht hat
since me Peter already for-a-long-time today visited has
c. *weil das Peter schon lange
heute wollte
since that Peter already for-a-long-time today wanted
d. weil mich Peter heute schon lange
besucht hat
since me Peter today already for-a-long-time visited has
One of the core assumptions here is that a simple clause allows for only one
independent adverbial of time. (47) further illustrates that the order of time adverbials in a
restructuring construction is even different from the order that is attested in a simple clause.
The following sentences indicate that there are two distinct binding domains in coherent
infinitives of ECM-verbs:
(48) a. weil eri ihn*i/j sie nicht washen liess
since he him her not wash let
b. weil eri sie ihni/j nicht washen liess
since he her him not wash let
c. *weil eri sie sichi nicht waschen sah
since he her himself not wash
saw
d. weil eri sichi sie nicht waschen sah
since he himself her not wash
saw
e. weil er siei sichi nicht waschen sah
since he her herself not wash
saw
According to Hinterhölzl (1999) all these facts can be easily accomodated into a
biclausal approach if we assume that ECM-subjects are case-licensed in the matrix TP, objects
are Case-licensed in the embedded TP, and the principle at work as far as pronominal binding
is concerned is the following (Hinterhölzl 1999):
(49)
Pronouns and anaphors are interpreted in their Case-licensing positions (ie., they
must be reconstructed into their Case-licensing position before the Binding theory
applies at LF).
6.2 Back to Hungarian
As far as Hungarian is concerned, it is relatively easy to find infinitival constructions
involving more than one temporal adverbial with the expected interpretation. The binding
facts, however, are much more intriguing.
Interestingly enough, in a neutral sentence two independent adverbials of time sound a
bit unnatural, regardless of order:
(50) a. ?Tegnap holnap
akartam Pétert
meglátogatni.
yesterday tomorrow wanted Peter.acc visit.inf
‘Yesterday I wanted to visit Peter tomorrow.’
b. ?Tegnap
Pétert
holnap akartam meglátogatni.
yesterday Peter.acc tomorrow wanted visit.inf
However, the situation improves considerably when we also include még ‘yet, as yet’:
(51) a. Tegnap még holnap
akartam Pétert
meglátogatni.
yesterday as-yet tomorrow wanted Peter.acc visit.inf
‘Yesterday I wanted to visit Peter tomorrow.’
b. Tegnap (még) Pétert (még) holnap akartam meglátogatni.
Discussing the difference between (50) and (51), and accounting for why the sentence
becomes fully grammatical after the insertion of még would lead us very far from from the
present discussion. Howvever, if we have a look at (52), where one of the time adverbials
occupies the focus position we can see that there is no problem whatsoever with the two
independent time expressions, so in this case the biclausal approach is fully supported (this
time még ‘yet, as yet’ cannot appear in the sentence, very probably because its semantics is
incompatible with the focus interpretation):
(52)
TEGNAP akartam Pétert
holnap meglátogatni.
yesterday wanted Peter.acc tomorrow visit.inf
‘It was yesterday that I wanted to invite Peter tomorrow.’
Turning to binding we are facing a serious problem: the traditional view (e.g. É. Kiss
1987, 2002) argues for a flat VP in Hungarian based on arguments coming from Binding
Principle C:
(53) a. *Felhívta a fiúk anyja őket.
called the boys mother them
‘The boysi’ mother called themi.’
b. *Felhívták (őki) a fiúki anyját.
called
they the boys mother.acc
‘Theyi called the boysi’ mother.’
The examples show that both subject and object pronouns can c-command the rest of the
DP, hence, the VP is flat in Hungarian. Obviously, this conclusion would be devastating for
the present proposal. That it may not be the case is pointed out by Kenesei (2005), where he
presents the following data to show that there subject-object asymmetry does exist in
Hungarian, too (Kenesei 2005):
(54) a. Péter belopakodott, de a kis hülyei anyukája rögtön észrevette (őti).
Peter sneaked
but the little idiot mom
at.once saw
him
‘Peter sneaked in, but the little idioti’s mom saw himi at once.’
b. *Péter be-nézett, de (ői) nem vette észre a kis hülyei anyukáját.
Peter in-looked but he not saw pfx the little idiot mom.acc
‘*Peter looked in, but hei didn’t see the little idioti’s mom.’
These data make the Principle C argument for a flat VP in Hungarian untenable, though
further data are needed to support the biclausal analysis, which I leave for further research.
7 Conclusion
In the present paper I made an attempt to account for why certain infinitival structures fail to
project typical left peripheral constituents in the expected order in spite of the widely accepted
claim that infinitival structures are clauses, also supported by ample empirical evidence.
Based on Hinterhölzl (1999) I argue that the problematic clauses also start out as biclausal
structures, but as a result of T-to-T movement they become monoclausal. This analysis has the
welcome consequence of (1) accounting for the similarity between free word-order
phenomena of certain West Germanic languages and Hungarian and (2) giving at least the
potential for providing a uniform biclausal analysis of infinitival structures in Hungarian,
though evidence for it is weaker than in the case of German and must be supported by further
data. Another merit of the paper is pointing out an important difference between the structure
of Hungarian finite an infinitival clauses hitherto unnoticed.
References
Ackema, P. (2004). ‘Do preverbs climb?’, in K. É. Kiss & H. C. van Riemsdijk (edd.), Verb
Clusters: A study of Hungarian, German and Dutch. Amsterdam: Benjamins, 359–393.
Brody, M. (1990). ‘Remarks on the order of elements in the Hungarian focus field’, in I.
Kenesei (ed.), Approaches to Hungarian 3. Szeged: JATE, 95–121.
Brody, M. (1995). ‘Focus and checking theory’, in I. Kenesei (ed.), Approaches to Hungarian
5. Szeged: JATE, 29–43.
Corver, N. & H. C. van Riemsdijk (1997). ‘The position of the head and the domain of
scrambling’, in B. Palek (ed.), Typology: Prototypes, Item Orderings and Universals:
Proceedings of LP ’96. Prague: Charles University, 57–90.
Dalmi, G. (2004). The Role of AGRP in Non-finite Predication, Unpublished PhD dissertation.
É. Kiss, K. (1987). Configurationality in Hungarian. Dordrecht: Reidel.
É. Kiss, K. (1992). ‘Move-alpha and scrambling in Hungarian’, in I. Kenesei & Cs. Pléh
(edd.), Approaches to Hungarian IV. Szeged: JATE, 67–98.
É. Kiss, K. (1995). ‘Introduction to the volume’, in K. É. Kiss (ed.), DiscourseConfigurational Languages. Oxford: Oxford University Press.
É. Kiss, K. (1999). ‘Strategies of complex predicate formation and the Hungarian verbal
complex’, in I. Kenesei (ed.), Crossing Boundaries. Amsterdam: Benjamins, 91–114.
É. Kiss, K. (2002). The Syntax of Hungarian. Cambridge: Cambridge University Press.
É. Kiss, K. (2003). ‘Argument scrambling, operator movement, and topic movement in
Hungarian’, in S. Karimi (ed.), Word Order and Scrambling. London: Blackwell, 22–43.
Evers, A. (1975). The Transformational Cycle in Dutch and German. Unpublished PhD.
dissertation, Utrecht.
Farkas, D. & A. Giannakidou (1996). ‘How clause-bounded is the scope of universals?’,
Semantics and Linguistic Theory VI: 137-155
Haegeman, L. & H. C. van Riemsdijk (1986). ‘Verb projection raising, scope, and the
typology of rules affecting verbs’, Linguistic Inquiry 14: 417–466.
Hinterhölzl, R. (1999). Restructuring Infinitives and the Theory of Complementation, PhD
dissertation. University of Southern California.
Hinterhölzl, R. (1997). ‘A VO-based approach to verb raising’, in K. Kusumoto (ed.), NELS
27, 187–201
Kayne, R. (1994). The Antisymmetry of Syntax, Linguistic Inquiry Monograph 25. Cambridge,
MA: MIT Press.
Kenesei, I. (1994). The Syntax of Focus. Ms. Szeged: József Attila University.
Kenesei, I. (2001). ‘Criteria for auxiliaries in Hungarian’, in I. Kenesei (ed.), Argument
structure in Hungarian. Budapest: Akadémiai Kiadó, 73–106.
Kenesei, I. (2002): Arguments in Hungarian Nonfinite Constructions. Presented in Leiden.
Kenesei, I. (2005). ‘Hungarian in focus’, Journal of Linguistics 41: 409–435.
Koopman, H. & A. Szabolcsi (2000). Verbal Complexes. Cambridge, MA: The MIT Press.
Puskás, G. (2001). Word Order in Hungarian: The syntax of A-bar positions. Amsterdam:
John Benjamins.
Roberts, I. (1997). ‘Restructuring, Head Movement and Locality’, Linguistic Inquiry 48: 423–
460.
Surányi, B. (2002). Multiple operator movements in Hungarian. Utrecht: LOT.
Szabolcsi, A. (1997). ‘Strategies for Scope Taking’, in A. Szabolcsi (ed.), Ways of Scope
Taking. Dordrecht: Kluwer, 109–154.
Tóth, I. (1995). Negative polarity licensing in Hungarian. Ms. József Attila University.
Tóth, I. (2000). Inflected Infinitives in Hungarian, PhD dissertation. Tilburg: Tilburg
University.
Wurmbrand, S. (2001). Infinitives. Berlin: Mouton de Gruyter.
Zwart, J-W (1993). Dutch Syntax, doctoral dissertation. University of Groningen.