The Theory of Complex Predicates and the få

advertisement
76
Working Papers in Scandinavian Syntax 57, 1996, 76-91.
The Theory of Complex Predicates and the
Norwegian Verb få 'get'
1
Helge Lødrup
University of Oslo
1. It has often been observed that Norwegian sentences with the verb få 'get'
can be ambiguous. Cf. (1). (Most translations are literal.)
(1) Per fikk reparert bilen
'Per got repaired the car'
(1) has an "active" interpretation, where the subject of få is the agent of the
repairing, as well as a "passive" interpretation, where the subject of få is a
benefactive, and the agent of the repairing is not specified2. (It can be
specified in an adjunct, for example av kona si 'by his wife'.) I assume that this
semantic difference reflects different syntactic constructions. The få of these
constructions will be referred to as the active få and the non-active få,
respectively3. This paper focuses on the construction with the non-active få.
The non-active få is often assumed to be an auxiliary in a passive
construction. I will argue that it is not, and propose that the non-active få and
the participle that it governs constitute a complex predicate. I will also give
an explanation of the curious fact that the non-active få cannot govern a verb
without an object. Finally, I will discuss the relation between sentences like
(1), whose object follows the participle, and sentences like (2), whose object
precedes the participle, and argue that they are more different than is usually
assumed.
(2) Per fikk bilen reparert
'Per got the car repaired'
My discussion is limited to Norwegian. The relevant facts are roughly the
same in Danish, while the Swedish picture is more complex. (See Hedlund
1992, chapter 2.) Many languages have constructions that are similar to the
non-active få construction. (See e.g. Askedal 1984, Haider 1984, Haegeman
1985.) Comparisons would be interesting, but are not included in the present
paper.
1
I would like to thank several people who have discussed the topic of this paper with
me, including Janne Bondi Johannessen, Arild Hestvik and Leiv-Otto Marstrander. I
would especially like to thank Kirsti Koch Christensen, who has spent hours reading
and commenting preliminary versions of this paper.
2
(1) also seems to have a causative or resultative interpretation. In section 2, I argue
that this interpretation is not distinct from the "passive" interpretation.
3
It is possible to get the same ambiguity with the verb ha 'have' in certain modal
contexts, cf. Jeg vil ha renset dette teppet 'I will have cleaned this carpet'. (See Hansen
1972.)
77
2. The basic meaning of a sentence of the form subject - non-active få - VP is
that the event denoted by the VP takes place to the advantage or
disadvantage of the subject. The malefactive interpretation, which is less
common, is illustrated by the authentic examples (3)-(4).
(3) forbrukere har fått misbrukt stjålne eller tapte plastkort
'consumers have got misused stolen or lost plastic cards'
(4) befolkningen var slaver som fikk skåret av seg tungen
'the population was slaves who got cut off them the tongue'
This description of the meaning of the non-active få is not without its
problems, however. Consider (5).
(5) Per fikk reparert bilen
'Per got repaired the car'
(5) means that the car was repaired, and that this happened to the advantage
(or disadvantage) of Per. Per´s involvement is not specified. For example, the
car may have been repaired by Per's wife without Per knowing, or Per may
have hired someone to repair it. The last case could be argued to represent a
distinct interpretation in which få is a causative or resultative verb. This
causative or resultative få would have no grammatical properties of its own,
however. When it comes to the grammatical properties discussed below, it
always follows the non-active få. I would prefer to say that even if the subject
of få does something to make the event denoted by the embedded VP come
about, this does not represent a distinct interpretation. Cf. (6) with the main
verb få.
(6) Per har fått ny jobb
'Per has got new job'
Again (6) does not specify Per´s involvement. He may have struggled for
years to get this job, or he may have found an offer in his mail box.
We could think of the event denoted by the embedded VP in the nonactive få construction as a kind of abstract object that the subject "gets". There
are many ways in which we can "get" objects. The conceptualization
represented by få focuses upon the fact that the object ends up with us, not
how this happened. Therefore, it does not seem to be linguistically relevant
what Per might have done to bring about the result described in (5) or (6).
For these reasons, I prefer to say that there is no distinct causative or
resultative interpretation of få in a sentence like (5), and that its subject is
always benefactive (excluding, of course, the active få)4.
3. The non-active få is traditionally assumed to be an auxiliary in a passive
construction. This analysis is implicit in traditional Norwegian and Danish
grammars (for example, Diderichsen 1962:135, Hansen 1967:157, Lie 1979:71).
When få takes a participle, it is treated as an auxiliary. It is pointed out that
4
There is a causative or resultative verb få in Norwegian, for example in a sentence like
Per fikk gulvet rent 'Per got the floor clean'. Even if the subject of this verb must be
agentive, there seems to be something benefactive about it as well.
78
the agent of få can be the same as the agent of the main verb, or different
from the agent of the main verb. From a somewhat different point of view,
Askedal 1984 and Ryen 1990 also assume få to be a an auxiliary in a passive
construction.
Challenging the traditional view, three questions could be asked:
a) Is it true that the active få is the same verb as the non-active få?
b) Is it true that the non-active få is an auxiliary in a passive construction?
c) Is it true that the participle that is governed by the non-active få is passive?
In this section, I will answer "no" to a) and b), and "yes" to c).
a) Is the active få the same verb as the non-active få?
Diderichsen 1962:135, Hansen 1967:157 and Lie 1979:71 identify the active få
and the non-active få. This identification can also be found in two recent
papers. Christensen and Taraldsen 1989 propose that both the active få and
the non-active få take a participle phrase with a PRO subject; PRO is
coindexed with the subject of the active få, while it is a PROarb with nonactive få. Taraldsen 1995 argues that both the active få and the non-active få
take a subject that is moved from the participle phrase; it does not have to be
the subject of the participle phrase, however.
There are important semantic and grammatical differences between the
active få and the non-active få, which I will now discuss.
The active få takes an agentive subject. The basic meaning of a sentence
of the form subject - active få - VP is that the subject manages to carry out the
action denoted by the VP. The embedded participle takes a PRO subject that
is coindexed with the subject of få, so the active få is a control verb5. The nonactive få is different. It takes a benefactive (or malefactive) subject. The
embedded participle does not have a subject that is coindexed with the
subject of the non-active få. In taking a benefactive (or malefactive) subject,
the non-active få is like the main verb få and the root modal få in sentences
like (7)-(8).
(7) Per fikk en kake
'Per got a cake'
(8) Får jeg komme inn?
'Get I come in?' [i.e. 'May I come in?']
There are two interesting differences between the active få and the non-active
få that are related to their subject roles. These differences concern controller
selection and imperative formation. (See Faarlund 1985.) A verbal
complement headed by the active få follows the main rule for control,
selecting the object (if there is one) as its controller. The non-active få shows
5
As an alternative, the active få could be assumed to be a raising verb (as in Taraldsen
1995). However, få differs from ha 'have' and other raising verbs in not allowing a
formal subject. Cf. Det har/*får snedd 'It has/gets snowed', Det har/*får kommet en mann
'It has/gets come a man'. Participles do not usually take PRO subjects, so the
embedded participle with the active få is exceptional in this respect.
79
deviant control behavior, however, allowing control from a subject, even
with an object accessible. Cf. (9)-(10).
(9) Per bad Pål om å få PRO utbetalt pengene [active få]
'Per asked Pål to get PRO paid out money'
[object control]
(10) Per bad Pål om å få utbetalt pengene [non-active få]
'Per asked Pål to get paid out money'
[subject control]
Also with imperative formation, the active få follows the main rule, while the
non-active få shows deviant behavior. The active få has ordinary second
person imperatives. Imperatives of the non-active få, on the other hand, are
first person. Cf. (11)-(12).
(11) Få PRO utbetalt pengene! [active få]
'[You] get paid out the money!
(12) Få utbetalt pengene! [non-active få]
'[I] get paid out the money!
The deviant behavior of the non-active få with respect to controller selection
and imperative formation is related to its subject role. (See Faarlund 1985.)
The main verb få and the root modal få behave like the non-active få. The
active få, on the other hand, behaves like the resultative verb få that appears
in sentences like Per fikk gulvet rent 'Per got the floor clean'6.
There is another important grammatical difference between the active få
and the non-active få. The active få can govern a verb without an object, while
the non-active få cannot. (13) is impossible with the non-active få.
(13) Per fikk arbeidet
'Per got worked'
(13) can only mean that Per managed to work, not that somebody worked for
the advantage (or disadvantage) of Per. (More in section 5.)
The active få cannot be passivized, while the non-active få can7. Cf. (14)(15).
(14) *Det fås arbeidet
'It get-PASSIVE worked'
(15) Det fås utbetalt penger i kassen
'It get-PASSIVE paid out money at the cash register'
6
The data are not completely clear. There are imperatives that allow an interpretation
that is both second person and benefactive, in addition to the expected first person and
benefactive, cf. (Gå til kassen og) få utbetalt pengene! '(Go to the cash register and) get
paid out the money!'
7
få does not have a passive participle; it only allows the morphological passive with
the suffix -s. This kind of passivization is typical of a small group of Norwegian verbs
that includes ha 'have', behøve 'need', trenge 'need'. What these verbs have in common is
that they assign a benefactive role to their subject.
80
(15) can only mean that somebody pays out money at the cash register to the
advantage (or disadvantage) of someone, not that somebody manages to pay
out money at the cash register.
We see, then, that there are several arguments that the non-active få and
the active få should not be identified as the same verb.
b) Is the non-active få an auxiliary in a passive construction?
Even if it is true that the non-active få and the active få are not the same verb,
it could still be the case that the non-active få is an auxiliary in a passive
construction. In the following, I will argue that there is no passive auxiliary
få8. I will accept the traditional view that the active få is an auxiliary (but see
Lødrup 1996).
First, consider the important fact that the non-active få can be passivized.
(Cf. 15 above.) Needless to say, auxiliaries can in general not be passivized,
and a sentence with a passive main verb cannot be passivized once more.
Another important fact is that the non-active få assigns a thematic role to
its subject. The uncontroversial passive auxiliaries, være 'be' and bli 'become'
on the other hand, have no thematic role to assign to their subject; they are
raising verbs. This explains the contrasts in (16)-(18).
(16) En medalje blir/*får overrakt Per
'A medal is/gets presented Per'
(17) Per blir/*får tenkt på
'Per is/gets thought of'
(18) Det blir/*får overrakt Per en medalje
'It is/gets presented Per a medal'
Another argument that the non-active få takes a subject that is thematic, and
not raised from the participle phrase, is the following: Verbs like bebreide
'reproach', nekte 'refuse' and pålegge 'impose on' require their benefactive role
to be syntactically realized, cf. (19)-(20).
(19) *Vi påla nye plikter
'We imposed new duties'
(20) *Nye plikter ble pålagt
'New duties were imposed'
If the subject of få is raised from the embedded participle phrase, it should be
possible for the benefactive of these verbs to be realized as the subject of få, in
the same way as it can be realized as the subject of a passive auxiliary. This
means that (21) should be possible, just like (22), if the subject of få is raised
from the embedded participle phrase.
(21) *Han fikk stadig pålagt nye plikter [from Askedal 1984]
'He got always imposed new duties'
(22) Han ble stadig pålagt nye plikter
'He was always imposed new duties'
8
Cf. also Haider 1984, who argues that the corresponding construction in German,
known as the Rezipientenpassiv, does not involve a passive auxiliary.
81
However, (21) is ungrammatical, so the subject of få is not raised9.
Could få be a passive auxiliary even if it is not a raising verb? The central
auxiliaries are raising verbs. However, most linguists would say that the root
modals (and the active få) are auxiliaries, even if they are control verbs. If we
for a moment ignore the difference between raising and control, we could
say that the subject of an auxiliary must at least be a suitable subject for the
verb that the auxiliary governs. But even this modest requirement is not
satisfied by the non-active få. The subject of få is not a suitable subject for the
embedded participle, cf. for example (23).
(23) Per fikk reparert bilen
'Per got repaired the car'
We see, then, that there are important arguments against the auxiliary status
of the non-active få.
c) Is the participle that is governed by the non-active få passive?
The correct part of the traditional analysis of the non-active få is the
assumption that the embedded participle is passive. A couple of arguments
for this view are the following.
First, the set of verbs that cannot occur with the non-active få seem to be
exactly the set of verbs that lack passive participles. Cf. for example (24)-(25).
(24) *Per fikk tilhørt laget - *Laget ble tilhørt (av Per)
'Per got belonged the team' - 'The team was belonged (by Per)'
(25) *Vi fikk hatt bilen i deres garasje - *Bilen ble hatt i deres garasje (av oss)
'We got had the car in their garage' - 'The car was had in their garage (by us)'
Second, passive participles that have no active counterpart, like tilsendt 'sentto' are possible with the non-active få, cf. (26)-(28).
(26) Han ble tilsendt et brev
'He was sent-to a letter'
(27) *Jeg har tilsendt ham et brev
'I have sent-to him a letter'
(28) Han fikk tilsendt et brev
'He got sent-to a letter'
4. When the non-active få is not an auxiliary, what is it? An important
property of the non-active få construction is that the embedded VP seems to
have no subject. The subject of få is not the subject of the embedded VP, as
shown above (see 23). The DP following the embedded participle cannot be
its subject, since it is in the object position. One could stipulate a PROarb
subject, as in Christensen and Taraldsen 1989. I will explore another
possibility, and assume that the embedded VP does not have a subject in any
sense of the term, at any level of representation. (It does, of course, have a
9
Taraldsen 1995, who assumes that the subject of få is raised from the embedded
participle, makes a parallel argument, but with the opposite conclusion. His crucial
example involves the verb tilsende 'send to' which to me only exists as a passive
participle. He acknowledges the existence of cases like (21), which he claims to be
exceptions.
82
highest thematic role, which is not assigned to a DP, because the verb is
passive.) This means that the non-active få construction is monoclausal, even
if there are two verbs involved.
There is one construction in Norwegian that has important properties in
common with the non-active få construction, namely the la 'let' construction.
(On similarities and differences, see Christensen 1993.) The relevant use of la
is represented by the rather bookish (29).
(29) Vi lot løslate fangene
'We let [someone] release the prisoners'
la is also used in sentences like Vi lot Per løslate fangene 'We let Per release the
prisoners', which is an ordinary "raising to object" construction.
The similarity between (the relevant uses of) la and få is striking. With la,
as with få, there is an alternative word order; the object can precede the
embedded verb, as in (30).
(30) Vi lot fangene løslate
'We let the prisoners release'
With la, as with få, the embedded verb seems to have no subject specified on
any level10.
The most important difference between the constructions is that with la,
the embedded verb must be an active infinitive. Other differences will not be
discussed here (see Christensen 1993).
Taraldsen 1983, 1984, 1991 proposes that the VP above la is a projection
of both la and the embedded verb, so la and the embedded verb constitute a
complex predicate. My analysis of the non-active få construction is based
upon the same idea, even if my implementation is somewhat different from
Taraldsen´s.
Several researchers have established that two verbs together can form a
complex predicate, which takes one subject, one object, etc. The verb that
governs the other verb in phrase structure is most often a causative or
causative-like verb, or a "light verb"11. (See for example Rosen 1990, Di Sciullo
and Rosen 1990, Ritter and Rosen 1993, Taraldsen 1991, Alsina 1992, 1993,
Butt 1995. The theory of Alsina is especially important for this paper.)
I assume that a complex predicate is derived through an operation on the
argument structures of the verbs involved. The argument structure of the
non-active få is underspecified, as in (31). Its non-subject argument is not a
thematic role, but a position that combines with the argument structure of an
embedded participle, shown in (32). The result is a complex predicate with
an argument structure like (33) (to be revised below).
10
With the verb høre 'hear', a similar construction is possible, as in Jeg har hørt si at han
kommer 'I have heard say that he comes'. This construction is somewhat marginal in
modern Norwegian.
11
It has been proposed that an auxiliary (or auxiliary-like) verb and a main verb should
be analyzed as a complex predicate in some languages. This seems to be well
motivated for cases like Italian volere 'want'. (See Rosen 1990, chapter 4.) This kind of
analysis is unmotivated for Norwegian, however, where auxiliaries behave like main
verbs in most respects.
83
(31) få <benefactive < >>
(32) reparert 'repaired' <agent theme>
(33) få-reparert <benefactive <agent theme>>
The embedded agent is unlinked (or realized as an "agent phrase") because
the embedded verb is passive.
The complex predicate can passivize. Then the benefactive is unlinked,
and the theme is linked to the subject or the object, giving sentences like (34)(35).
(34) Penger fås utbetalt i kassen
'Money get-PASSIVE paid out at the cash register'
(35) Det fås utbetalt penger i kassen
'It get-PASSIVE paid out money at the cash register'
Within Lexical Functional Grammar, a complex predicate construction is
monoclausal in functional structure. At this level of representation, the
complex predicate is one unit that takes one subject, one object etc. For
example, the (simplified) functional structure of Per fikk reparert bilen 'Per got
repaired the car' is as in (36).
(36)
| SUBJ Per
|
| PRED få-reparert |
| OBJ bilen
|
The two verbs forming a complex predicate do not have to be a syntactic unit
in constituent structure. This is well known from the Romance causative
construction, which has been a central topic in the literature on complex
predicates (Rosen 1990, Di Sciullo and Rosen 1990, Alsina 1993). It is also the
case with få and the embedded participle, cf. (37)-(38), in which the two verbs
are not adjacent.
(37) Da fikk Per reparert bilen
'Then got Per repaired the car'
(38) Når fås bilen reparert?
'When get-PASSIVE the car repaired?'
This means that the complex predicate formation cannot take place in the
lexicon, it must take place in syntax. Alsina 1993:293-305 shows how this can
be done within Lexical Functional Grammar. He assumes that a VP can have
two functional heads when one of them is an incomplete predicate with an
underspecified argument structure, and the other is an ordinary predicate.
The predicates compose to specify this argument structure of the complex
predicate. (This is rather similar to the analysis of la in Taraldsen 1983, 1984,
1991.)
84
In a tree like (39), both få and the VP above the embedded participle are
functional heads that combine12.
(39)
VP
V
VP
|
få
'get'
V
DP
|
|
reparert
'repaired'
bilen
'the car'
An argument for the complex predicate analysis is the fact that a pronoun
that cliticizes on få can realize an argument of the embedded participle. Cf.
(40).
(40) Han fikk-seg pålagt nye oppgaver
'He got himself imposed on new duties' [i.e. 'He had new duties imposed on him']
In (40), seg 'himself' is an obligatory indirect object of the embedded
participle, which cliticizes on få. If seg is replaced by an ordinary DP, it
cannot precede the participle. In (41), both the indirect and the direct object
of the embedded participle cliticize on få13.
(41) Han fikk-seg-det forklart mange ganger
'He got himself it explained many times' [i.e. 'He had it explained to him many times']
It is well known from the Romance languages that complex predicate
constructions allow a pronoun that realizes an argument of the embedded
verb to cliticize on the higher verb. The phenomenon is called clitic climbing.
(See e.g. Manning 1992 and references there.) An example is the Italian (42).
(42) Maria lo fara leggere a Gianni [Di Sciullo and Rosen1990:117]
'Maria it made read to Gianni'
Clitic climbing seems to be more restricted in Norwegian than in Romance.
(For example, sentences like (40)-(41) become less natural if the reflexive
pronoun is replaced by a non-reflexive pronoun.) Even so, sentences like
(40)-(41) give an important argument for the complex predicate analysis.
Askedal 1984 points out that sentences like (40)-(41) do not allow the
interpretation with the active få. The reason is that clitic climbing is only
12
A prediction of this analysis is that the embedded VP cannot be topicalized. The
theory of unbounded dependencies in LFG (Kaplan and Zaenen 1989) only allows
topicalization of a constituent that has a grammatical relation, and the embedded VP
does not have one. This prediction is not correct; a sentence like the following is not
really unacceptable:
?Reparert bilen fikk han ikke 'Repaired the car he did not get'
13
Sentences like Jeg fikk den reparert 'I got it repaired' are probably not relevant to the
discussion here. The reason is that they can get the same analysis as sentences like (2)
above, in which the DP precedes the embedded participle.
85
allowed in complex predicate constructions, and the active få construction
does not contain a complex predicate. (Askedal 1984 interprets the facts in a
different way.)
5. Accepting the complex predicate analysis, there are still two important
properties of the non-active få construction to be accounted for:
a) få has no object of its own in the non-active få construction. This is
remarkable, because få, apart from its auxiliary uses, never occurs without an
object.
b) The non-active få never appears with an intransitive verb. A sentence like
(43) is only grammatical with the active få. It can only mean that Per
managed to work, not that someone worked to the advantage (or
disadvantage) of Per.
(43) Per fikk arbeidet
'Per got worked'
The fact that the embedded participle is passive cannot be the reason that it
cannot be intransitive, since unergative verbs have passive participles in
Norwegian.
The contrast is especially striking with verbs that allow their object to be
unexpressed. Cf. (44)-(45).
(44) Per fikk bakt en kake
'Per got baked a cake'
(45) Per fikk bakt
'Per got baked'
(44) is ambiguous between the active and the passive interpretation, while
(45) can only have the active interpretation
The point is that the non-active få must have an object. Clausal
complements can be objects in Norwegian (cf. Lødrup 1991), and sentences
like (46)-(47) allow the passive interpretation.
(46) Per fikk opplyst hvordan maskinen virker
'Per got informed how the machine works'
(47) Per fikk anbefalt å unngå fuktighet
'Per got recommended to avoid moisture'
It is possible to construct sentences in which an implicit object is enough to
get the passive interpretation. Cf. (48).
(48) Når får vi servert?
'When do we get served?'
(48) can get the passive interpretation if the implicit object has a strong
degree of presence, for example if it is said in a restaurant. This possibility
86
seems to be somewhat marginal, however, and it is disregarded in the
following discussion.
The properties a) and b) are also shared by la. In the relevant
construction, la lacks an object of its own. And the embedded verb must take
an object, cf. (49).
(49) *Vi lot arbeide
'We let work'
The complex predicate analysis does not in itself say anything about the
lacking object of få or the obligatory transitivity of the embedded participle.
To account for this I will apply the theory of causatives in Alsina and Joshi
1993. They propose that causative predicates in some languages are threeplace predicates, taking a theme argument in addition to the causer and the
caused event. This theme argument is semantically identified with an
argument of the embedded predicate; this is called fusion. This means that
one syntactic argument must realize both the theme role of the causative
predicate and one of the thematic roles of the embedded predicate. This role
can be the logical subject or the logical object; this is a parameter of variation.
Chichewa allows both possibilities, cf. (50)-(51) (from Alsina and Joshi
1993:8).
(50) Nu˘ngu i-na-phík-ítsa kádzidzi maûngu
porcupine SUBJ-PAST-cook-CAUS owl pumpkins'
'The porcupine made the owl cook the pumpkins'
(51) Nu˘ngu i-na-phík-ítsa maûngu kwá kádzı¯dzi
porcupine SUBJ-PAST-cook-CAUS pumpkins by owl
'The porcupine had the pumpkins cooked by the owl'
In (50), the object 'owl' realizes both the theme of the causative predicate and
the subject role of the embedded predicate 'cook'. In (51), the object
'pumpkins' realizes both the theme of the causative predicate and the object
role of the embedded predicate. (See also Alsina 1992.)
Even if the non-active få is not a causative, I assume that it takes three
semantic arguments. These arguments are a benefactive, a theme and one
that combines with the argument structure of the embedded participle. The
logical subject of the embedded participle does not link, since the verb is
passive. (It can license an "agent phrase", however.) The logical object of the
embedded participle is identified with the theme of få. Cf. (52)-(53).
(52) få <benefactive theme <
>>
|_____|
(53) få-reparert 'get repaired' <benefactive theme <agent theme>>
|_____________|
This means that the object of the embedded participle is also the object of få.
This gives us an explanation for the obligatory transitivity of the embedded
participle, få needs an object. To be more exact, the embedded participle must
have a thematic object; with a non-thematic object there can be no
87
fusion. This is the reason that (54), which contains an idiom with a nonthematic object, is impossible with the non-active få.
(54) Per fikk drevet dank
'Per got driven dank'
The idiom drive dank means 'idle about'. (54) can only mean that Per managed
to idle about, it cannot mean that someone idled about to the advantage or
disadvantage of Per.
There are also independent arguments for the analysis in (52), which will
be discussed below.
First, note that the object of (55) shows the definiteness effect, while the
object of (56) does not.
(55) Det fås utbetalt penger/*pengene i kassen
'It get-PASSIVE paid out money/the money at the cash register'
(56) Per får utbetalt penger/pengene i kassen
'Per gets paid out money/the money at the cash register'
In the functional structures of (55) and (56), penger 'money' is not the object of
utbetalt 'paid out', it is the object of the complex predicate. In (56), this
complex predicate has a thematic subject, while in (55) it does not. The object
of (55) is therefore predicted to show the definiteness effect, just like the
object of (57).
(57) Det fås penger/*pengene i kassen
'It get-PASSIVE money/the money at the cash register'
My second argument for the analysis in (52) has to do with locality. The
complex predicate is derived by fusing the argument structures of få and the
embedded participle. It has been proposed that operations on argument
structure are always local. Alsina 1993:619 maintains that they cannot go
"down" more than one level of embedding. This means that the complex
predicate can only consist of få and the topmost embedded participle, which
furthermore must have an object role immediately accessible. This locality
restriction predicts that the relevant object cannot be the object of a
preposition. And sentences like (58) are ungrammatical.
(58) *Per fikk passet på barna av naboen
'Per got looked after the kids by the neighbor'
The locality restriction also predicts that the verb with the accessible object
role cannot be further embedded in any way. For example, it cannot be
embedded under a raising verb. In (59)-(60), the verb with the relevant object
role (opphevet 'abolished', fjernet 'removed') is embedded under a passive
raising verb (forsøkt 'attempted', anbefalt 'recommended'). And (59)-(60) are
ungrammatical.
(59) *Norge fikk forsøkt opphevet boikotten
'Norway got attempted abolished the boycott'
88
(60) *Per fikk anbefalt fjernet mandlene
'Per got recommended removed the tonsils'
6. In section 1, it was pointed out that there are two possible word orders
with få, as in (61)-(62).
(61) Per fikk reparert bilen
'Per got repaired the car'
(62) Per fikk bilen reparert
'Per got the car repaired'
(62) seems to have the DP in the object position of få. I will call the word
order of (61) V DP, and the word order of (62) DP V.
In the literature V DP and DP V are seen as instances of the same
construction. This view is implicit or explicit in Diderichsen 1962:135,
Hansen 1967:157 and Ryen 1990. It has also been proposed that DP V
sentences are derived from DP V sentences, or the other way around14. There
are, however, differences between V DP sentences and DP V sentences that
indicate that they should be treated separately.
The first difference concerns dialects with participle agreement. In these
dialects, the V agrees with the DP in DP V sentences, but not in V DP
sentences. The same is the case in the written "Nynorsk" standard of
Norwegian, cf. Christensen 1989. (And the same is true of Swedish, cf.
Hedlund 1992:59.) Cf. (63)-(64).
(63) Han fekk blomane tilsende
'He got the flowers sent-PLURAL'
(64) Han fekk tilsendt blomane
'He got sent-SUPINE the flowers'
The agreement shows that the DP in DP V sentences should be a subject for
the participle.
The second difference concerns locality. It was pointed out above that the
complex predicate can only consist of få and the topmost embedded
participle, which furthermore must have an object role immediately
accessible. However, this locality restriction seems not to be operative in the
DP V sentences (even if it must be admitted that the contrasts are subtle). Cf.
(65)-(67).
(65)?Per fikk barna passet på
'Per got the kids looked after'
(66) Norge fikk boikotten forsøkt opphevet
'Norway got the boycott attempted abolished'
(i.e. 'Someone attempted to abolish the boycott for Norway')
(67) Per fikk mandlene anbefalt fjernet
'Per got the tonsils recommended removed'
(i.e. 'Someone recommended to remove the tonsils for Per')
14
Taraldsen 1995 proposes that DP V is derived from V DP by moving the DP. Another
possibility is to see DP V as basic and V DP as derived. The head of the embedded VP
is then moved to the position after få. This is essentially the analysis proposed by
Hedlund 1992:75-76 for Swedish dialects that allow V DP.
89
The third difference concerns the range of possible verbs. Experiencer verbs
are acceptable in DP V sentences, but not in V DP sentences. Cf. (68)-(71).
(68) Kokken får dessertene sine likt av alle
'The cook gets his deserts liked by everybody'
(69) ??Kokken får likt dessertene sine av alle
'The cook gets liked his deserts by everybody'
(70) Den pensjonerte læreren får timene sine savnet av alle
'The retired teacher gets his classes missed by everybody'
(71) ??Den pensjonerte læreren får savnet timene sine av alle
'The retired teacher gets missed his classes by everybody'
This difference might seem to be strange, but it is a part of a more general
phenomenon. Romance languages do not allow experiencer verbs in the
causative faire-par-construction, cf. Bordelois 1988, Alsina 1993:203-4. So there
is independent evidence that experiencer verbs can be prohibited in complex
predicate constructions (even if this fact is difficult to explain)15.
These three differences indicate (with varying strength) that DP V
sentences do not involve a complex predicate. Instead, they represent an
ordinary "raising to object" construction. The DP is the "raised" object, and
the VP is an object predicate (called XCOMP in Lexical Functional Grammar).
This means that the DP V sentences get the same syntactic analysis as the
sentences (72)-(73).
(72) Per fikk katten inn
'Per got the cat in'
(73) Per fikk hendene skitne
'Per got the hands dirty'
The DP is the object of få. At the same time it is the subject of the embedded
predicate in the functional structure. The (simplified) functional structure of
Per fikk bilen reparert 'Per got the car repaired' is as in (74), where the dotted
line indicates that the object of få is also the subject of the XCOMP16.
15
There is also a problematic difference between V DP and DP V concerning double
objects. Herslund 1986 points out that DP V sentences do not allow double objects in
Danish. This seems to apply to Norwegian as well, cf. *Hun fikk blomstene sendt
generalen 'She got the flowers sent the general'. V DP sentences allow double objects, cf.
Hun fikk sendt generalen blomstene 'She got sent the general the flowers'. This difference
between DP V sentences and V DP sentences also concerns sentences with la (as
Herslund 1986 points out). It is not obvious if it is relevant here, or how it should be
accounted for.
16
It is not obvious if V DP and DP V sentences with la are as different as they are with
få. Taraldsen 1991:225-27 claims that there is an important structural difference
between them; in DP V sentences the DP is in a subject position, while in V DP
sentences it is in an object position. His argument is that the DP in DP V sentences can
antecede a reflexive. I am not sure that reflexives give an argument in this context,
since it is not too difficult to make a non-subject antecede a reflexive. And I find no real
difference in acceptability between (i) and (ii), where the reflexive is the possessive sine
'their'.
(i) Vi lot dem løslate sammen med foreldrene sine 'We let them release with their
parents'
(ii) Vi lot løslate dem sammen med foreldrene sine 'We let release them with their
parents'
90
(74)
|
|
|
|
|
|
SUBJ Per
PRED få
OBJ bilen
XCOMP
|
|
|
|
| SUBJ
| |
| PRED reparert | |
7. I have tried to show how the properties of the non-active få construction
follow naturally from the theory of complex predicates (here in the version of
Alsina 1993) when we apply the theory of causatives in Alsina and Joshi
1993.
LITERATURE
Alsina, A. 1992 On the argument structure of causatives. Linguistic Inquiry 23, 517556.
Alsina, A. 1993 Predicate Composition: A Theory of Syntactic Function Alternations.
Dissertation. Department of Linguistics, Stanford University.
Alsina, A. and S. Joshi 1993 Parameters in causative constructions. In Papers from the
27th Regional Meeting of the Chicago Linguistic Society 1991, 1-15. Chicago.
Askedal,
J.O.
1984
Zum
kontrastiven
Vergleich
des
sogenannten
"bekommen/erhalten/kriegen-Passivs" im Deutschen und entsprechender
norwegischer Fügungen aus få und dem Partizip Perfekt. Norsk lingvistisk
tidsskrift 2, 133-166. (Oslo: Novus.)
Bordelois, I. 1988 Causatives: From lexicon to syntax. Natural Language and Linguistic
Theory 6, 57-93.
Butt, M. 1995 The Structure of Complex Predicates in Urdu. Stanford: CSLI.
Christensen, K.K. 1989 Partisippkongruens, A´-argumenter og hv-flytting. In E.H.
Jahr and O. Lorentz (eds.) Syntaks/Syntax. Oslo: Novus. 363-385.
Christensen, K.K. 1993 Conditions on complementation in Norwegian la-and fåcausatives. Talk given at Scandinavian Syntax Workshop, The XIVth
Scandinavian Conference of Linguistics and The VIIIth Conference of Nordic and
General Linguistics.
Christensen, K.K. and K.T. Taraldsen 1989 Expletive chain formation and past
participle agreement in Scandinavian dialects. In P. Benincà (ed.) Dialect
Variation and the Theory of Grammar. Dordrecht: Foris. Pp. 53-83.
Di Sciullo, A.M. and S.T. Rosen 1990 Light and semi-light verb constructions. In K.
Dziwirek et al. (eds.) Grammatical Relations: A Cross-Theoretical Perspective.
Stanford: CSLI. Pp. 109-125.
Diderichsen, P. 1962 Elementær Dansk Grammatik. København: Gyldendal.
Faarlund, J.T. 1985 Imperatives and control. First person imperatives in Norwegian.
Nordic Journal of Linguistics 8, 149-60.
Haegeman, L. 1985 The get-passive and Burzio's generalization. Lingua 66, 53-77.
Haider, H. 1984 Mona Lisa lächelt stumm - Über das sogenannte deutsche
´Rezipientenpassiv´. Linguistische Berichte 89, 32-42.
Hansen, A. 1967 Moderne dansk III. København: Grafisk forlag.
Hansen, E. 1972 Modal interessens: Nu bør det (komme) frem. Danske studier 67, 536.
Hedlund, C. 1992 On Participles. Stockholm: Department of Linguistics, Stockholm
University.
91
Herslund, M. 1986 Causatives, double objects and the ergativity hypothesis. In Ö.
Dahl (ed.) Papers from the Ninth Scandinavian Conference of Linguistics. Institute
of Linguistics, University of Stockholm. Pp. 142-153.
Kaplan, R.M. and A. Zaenen. 1989. Long-distance dependencies, constituent
structure and functional uncertainty. In M.R. Baltin and A.S. Kroch (eds.)
Alternative Conceptions of Phrase Structure. Chicago: The University of
Chicago Press. Pp. 17-42.
Lie, S. 1979 Innføring i norsk syntaks. Oslo: Universitetsforlaget.
Lødrup, H. 1991 Clausal complements in English and Norwegian. Norsk lingvistisk
tidsskrift. 105-136. (Oslo: Novus.)
Lødrup, H. 1994 "Surface proforms" in Norwegian and the definiteness effect. In M.
Gonzalez (ed.) Proceedings of the North East Linguistics Society 24. Amherst:
GLSA, Department of Linguistics, University of Massachusetts. Pp. 303-315.
Lødrup, H. 1996 Properties of Norwegian auxiliaries. In Proceedings of The IXth
Conference Of Nordic And General Linguistics, Department of Scandinavian
Studies and Comparative Literature, University of Oslo.
Manning, C.D. 1992 Romance is so complex. Technical Report CSLI-92-168, Center for
the Study of Language and Information, Stanford University.
Ritter, E. and S.T. Rosen 1993 Deriving causation. Natural Language and Linguistic
Theory 11, 519-555.
Ryen, E. 1990 Få - en liten studie av et mangesidig hjelpeverb. NOA 12, 1-13
(Department of Linguistics, University of Oslo).
Rosen, S.T. 1990 Argument Structure and Complex Predicates. New York: Garland
Publishing.
Taraldsen, K.T. 1983 Parametric Variation in Phrase Structure: A Case Study.
Dissertation, Univerity of Tromsø.
Taraldsen, K.T. 1984 Some phrase structure dependent differences between Swedish
and Norwegian. In Working Papers in Scandinavian Syntax 9.
Taraldsen, K.T. 1991 A directionality parameter for subject-object linking. In R.
Freidin (ed.) Principles and Parameters in Comparative Grammar. Cambridge;
MIT Press. Pp. 219-268.
Taraldsen, K.T. 1995 Participle-based small clause complements of få 'get' in
Norwegian. In A. Cardinaletti and M.T. Guasti (eds.) Syntax and Semantics 28:
Small Clauses. San Diego: Academic Press. Pp. 207-233.
Helge Lødrup
Department of
Linguistics
University of Oslo
Pb. 1102, Blindern
N-0317 Oslo, Norway
helge.lodrup@ilf.uio.no
Download