Document 11401139

advertisement
3
Working Papers in Scandinavian Syntax 45 1990 pp. 3-12.
Helge Lødrup:
VP-TOPICALIZATION AND THE VERB GJØRE IN NORWEGIAN1
1. It is well known that VP-topicalization in Norwegian (as in Swedish and
Danish) can be accompanied by the verb gjøre ´do´. gjøre can be
obligatory, as in 1-2, or optional, as in 3-4:
1 Spille/Spiller golf gjør jeg aldri
´Play-INF/Play-PRES golf I never do´
2 Spille/Spiller golf tror jeg at jeg aldri gjør
´Play-INF/Play-PRES golf I think I never do´
3 Spille/Spilt golf har jeg aldri gjort
´Play-INF/Play-PART golf I have never done´
4 Spille golf vil jeg aldri gjøre
´Play-INF golf I will never do´
It has often been observed that the obligatory gjøre in 1-2 satisfies the
requirement for a finite verb.
An interesting feature of the gjøre-construction is the variation in the form
of the verb in 1-3. For ease of exposition, I will overlook the forms to the
right of the slash for a while, and return to them in section 3.
To be exact, gjøre is not limited to topicalizations, it is used whenever a
VP enters into an unbounded dependency. Cf. the cleft sentences 5-6:
5 Det var ikke akkurat spille golf jeg gjorde
´It was not exactly play golf I did´
6 Det var ikke akkurat spille golf jeg ville (gjøre)
´It was not exactly play golf I would (do)´
From a transformational point of view, the gjøre-construction has been
viewed as particular and problematic. (See e.g. Faarlund 1978: 2.2.5,
Lorentz 1979:119) The topicalization is, of course, derived by
This paper has benefited from discussion at the ´Workshop on functional
elements and their projections´ (Oslo 1989), and ´Grammatik i fokus´
(Lund 1990), especially Anders Holmberg has been very helpful. I am also
grateful to Jan Anward for his comments.
1
4
movement, the problems are represented by gjøre. First, there is the
question where gjøre "comes from". Somehow gjøre must be inserted in
the right context (if it is not present in deep structure and deleted in other
contexts, which is a crude paraphrase of the proposal in Källgren 1972).
Second, there is the question of the status of gjøre and the relation
between gjøre and the topicalized VP.
The latest answer to these questions is given by Källgren and Prince 1989.
(Their analysis is based on Swedish, with finite VPs in sentences like 1-2.)
They say that gjøre is a pro-VP, and derived as the lexical realization of
the trace of the topicalized VP.
This idea of gjøre as a pro-element is rather well established. Several
Scandinavian grammarians have called gjøre a pro-verb or a pronominal
verb, e.g. Hansen 1967:69, Thorell 1977:75-76, Lorentz 1979:119 and Lie
1979:43 (in passing).
This kind of analysis suffers from both theoretical and empirical problems.
First, it involves the concept "lexically realized trace", which seems to be
problematic in linguistic theory. Second, it assumes that verbs may be proelements. But there are no clear instances of verbs being pro-elements in
Norwegian. And VP already has a pro-element, namely the pronoun det
´it/that´. (This is pointed out by Fretheim 1974 and Faarlund 1978:2.2.5.)
Cf. 7-9:
7 (Har du spist middag?) Ja, jeg har det
´(Have you eaten dinner?) Yes, I have that´
8 (Skal du spise middag?) Ja, jeg skal det
´Shall you eat dinner?) Yes, I shall that´
9 (Liker du jordbær?) Ja, jeg gjør det
´(Do you like strawberries?) Yes, I do that´
A sentence with VP-topicalization and gjøre can contain det, cf. 10-11
(pointed out to me by Jan Anward):
10 Spille golf, det har jeg aldri gjort
´Play golf, that I have never done´
11 Spille golf, har du gjort det?
´Play golf, have you done that?´
The distribution of det is as with NP-topicalization:
12 Golf, det har jeg aldri spilt
´Golf, that I have never played´
13 Golf, har du spilt det?
´Golf, have you played that?´
5
In 10-13, det is a proform for the topicalized constituent. In 10 and 12,
there must be a trace of det after gjort and spilt, and in 11 and 13 there can
be no such trace. These facts are hardly compatible with the view that
gjøre is a proform and/or a lexically realized trace. (Källgren and Prince
1989 claim that gjøre with det is another gjøre, but there seems to be little
or no evidence for this, at least in Norwegian.)
2. In this paper, I will present an alternative to the pro-VP analysis. The
paper is based on Lexical Functional Grammar (LFG), but I believe that
the main points can be transferred to other frameworks as well.
In LFG, movement and traces do not exist. (Cf. Kaplan et al 1987, note
that their treatment of unbounded dependencies differ from Kaplan and
Bresnan 1982:4.7.) Unbounded dependencies are accounted for in the
functional structure (i.e. the representation with information about
grammatical relations and morphosyntactic properties). A constituent that
enters into an unbounded dependency is generated directly in its surface
position by the phrase structure rules, and is supplied with an equation
saying that it has some grammatical relation to a predicator somewhere
"down" in the sentence. Overgeneration is avoided by well-formedness
conditions on functional structures, which demand that every predicator
has exactly the number and kind of arguments it needs (Kaplan and
Bresnan 1982:4.5).
The question of where gjøre "comes from" does not exist in LFG. gjøre
can only come from the lexicon, and it can only be generated directly in its
surface position. The question of the status of gjøre and the relation
between gjøre and the topicalized VP is more interesting.
I will argue that gjøre is an auxiliary verb, not unlike English do in
sentences like 14-15:
14 Play golf, I never do
15 Play golf, I have never done
This is not a new idea, it was proposed by Fretheim 1974 and Faarlund
1978:2.2.5, within the framework of generative semantics2
There are, of course, other possible analyses of the gjøre construction.
One possibility is deriving it from the gjøre det construction (as in 10)
with deletion of det. This analysis is not compatible with the framework
of LFG, and will not be examined further.
2
6
Being an auxiliary verb primarily means being a verb. Auxiliary verbs are
raising verbs, in LFG this means that they take a non-thematic subject and
a verbal complement with obligatory control3. The LFG treatment of
auxiliaries is discussed in Falk 1984, Netter 1988.
If gjøre can be an auxiliary, it is necessary to establish that the auxiliary
gjøre is distinct from the main verb gjøre. An important difference is that
the auxiliary gjøre is semantically empty. There are no semantic
constraints on the complement of gjøre. Note especially that the notion of
"doing" something is not involved, cf. 16:
16 Like jordbær kjenner jeg ingen som gjør
´Like strawberries I know nobody who does´
(This is pointed out by Fretheim 1974) A stative verb like like ´like´ does
not satisfy our notion of "doing" something. If somebody asks Hva gjør
du ´What do you do´, it is impossible to answer Jeg liker jordbær ´I like
strawberries´.
Even passive verbs may be used in the gjøre-construction, cf. 17:
17 Misforstås gjør de alltid
´Misunderstand-PASS they do always´
The auxiliary analysis accounts for important features of the gjøre
construction. It is well known that some VPs cannot be the verbal
complement of gjøre. Cf. 18-21:
18 *Ha spilt golf gjør jeg ikke
´Have played golf I do not´
19 *Ville spille golf gjør jeg ikke
´Want-to go play golf I do not´
20 *Ha penger gjør han ikke
´Have money he does not´
21 *Ville hjem gjør han ikke
´Want-to (go) home he does not´
Auxiliaries differ from ordinary raising verbs in that they can
pronominalize their verbal complement: (Har du spist?) Ja, jeg har det
´(Have you eaten?) Yes, I have that´ (Hagen 1979). It is possible that
modal verbs in their optative use should be regarded as equi verbs, as
originally proposed by Ross.
3
7
VPs that cannot be the verbal complement of gjøre, are headed by an
auxiliary verb, or to be more precise, a verb that may be used as an
auxiliary 4.
Under the pro-VP analysis, it is rather strange that these VPs cannot be
"pronominalized". Under the auxiliary analysis, this is not unnatural. As is
well known, auxiliaries impose grammatical restrictions on their verbal
complements. These restrictions may concern grammatical form, e.g. the
choice between a participle and an infinitive. They may also concern the
kind of head allowed, e.g. få does not allow a modal to head its verbal
complement. (Cf. *Trikken får ville bli forsinket ´The tram gets-to begoing-to be late´.) gjøre restricts its verbal complement in basically the
same way, prohibiting auxiliaries from heading it.
Treating gjøre as an auxiliary verb also gives a natural account of sentences
with gjøre det, cf.:
22 (Liker du jordbær?) Ja, jeg gjør det
´(Do you like strawberries?) Yes, I do that´
23 Spille golf, det har jeg aldri gjort
´Play golf, that I have never done´
24 Spille golf, har du gjort det?
´Play golf, have you done that?´
It is a general property of auxiliary verbs in Norwegian that they can
pronominalize their verbal complement with the pronoun det , cf.
25 (Har du spist middag?) Ja, jeg har det
´(Have you eaten dinner?) Yes, I have that´
26 (Skal du spise middag?) Ja, jeg skal det
´(Shall you eat dinner?) Yes, I shall that´
27 Spille golf, det har jeg aldri kunnet
´Play golf, that I have never been-able-to´
28 Spille golf, har du kunnet det?
´Play golf, have you been-able-to that?´
A potential problem for the auxiliary analysis is represented by sentences
like 29:
29 (Har du spist middag?) *Ja, jeg gjør det
´(Have you eaten dinner?) Yes, I do that´
This is the main rule. få ´get´ and bli ´become´ do not sound so bad when
they are used as main verbs. Cf. (?)Få penger gjør han aldri ´Get money
he never does´, (?)Bli syk gjør han aldri ´Become ill he never does´.
4
8
What prevents the pronoun det from referring to the VP headed by har
´have´? This must be connected with the restriction that the verbal
complement of gjøre cannot be headed by a verb that may be used as an
auxiliary. The pronoun cannot be interpreted in a way that would violate
this restriction. A parallel case is 30:
30 (Vil trikken bli forsinket?) *Ja, den får det
´(Will the tram be late?) Yes, it gets-to that´
Again, the pronoun cannot be interpreted in a way that would violate the
restriction on the verbal complement of the auxiliary.
3. A potential problem for the auxiliary analysis is the possibility of finite
VPs in sentences like 31:
31 Spille/Spiller golf gjør jeg aldri
´Play-INF/Play-PRES golf I never do
There is some variation in Norwegian, Swedish and Danish concerning the
finiteness of topicalized VPs. Consulting standard grammars, we find that
all three languages allow both possibilities, although one gets the
impression that finite VPs are the preferred possibility in Swedish, while
infinite VPs are more common in Danish and Norwegian. (See e.g. Thorell
1977:227 on Swedish and Hansen 1967:69 on Danish.). This does not, of
course, imply that every speaker allows both possibilities.
What should we do with this variation? One possibility is to regard the
constructions with finite and infinite VPs as basically different, and
describe them as independent of each other. Another possibility is to
choose one as basic, and describe the other as derived by some low level
rule. Choosing this approach, I think there are good reasons for the infinite
form to be considered as basic:
First, tense is a category that in all other cases is realized only once5.
Looking at the other Germanic languages, we find the infinite form as the
only possibility in the parallel constructions of English and German.
Second, I know no general process that could give us the infinite form from
the finite form. On the other hand, deriving the finite form from the infinite
form can be viewed as part of a more general phenomenon: Under certain
conditions, a verb may copy the morphosyntactic features of its governing
verb, instead of taking on
Källgren and Prince 1989 mention that tense may be realized twice with
left dislocation, but this is a construction that is related to topicalization.
5
9
the form that the governing verb normally requires. Cf. the somewhat
substandard 32-34
32 Jeg har prøvd å gjort det (Vinje 1987:320)
´I have tried to done it´
33 Jeg hadde villet gjort det
´I had wanted-to done it´
34 Slutt å skrik!
´Stop-IMP to cry-IMP
Infinitive is expected, but the verb has picked up the participle/ imperative
form from the governing verb. Whether this feature copying is part of a
more general phenomenon is a question not to be considered here. Vinje
1987:320 implies that it is an instance of "attraction". Attraction is an
established notion in traditional grammar. But it is supposed to account
for a diversity of unexpected morphosyntactic forms, and it is not clear
that it is more than a cover term6. An interesting line of research is
represented by Anward 1988, who develops a theory of verb-verb
agreement.
A crucial argument that feature copying is involved in the gjøreconstruction comes from the variation in 35:
35 Spille/Spilt golf har jeg aldri gjort
´Play-INF/Play-PART golf I have never done´
It is quite obvious that the participle form must be due to feature copying
from gjort ´done´. A participle is not possible with infinitive gjøre:
36 Spille/*Spilt golf vil jeg aldri gjøre
´Play-INF/Play-PART golf I will never do´
4. So far, we have a rather unproblematic analysis of the gjøre
construction, using traditional and well motivated categories. There is,
however, one problem with the auxiliary-analysis that is not easy to
handle: the impossibility of gjøre when VP is in the sentence "where it
belongs". Cf. 37:
37 *Jeg gjør aldri spille golf
´I never do play golf´
For example, Vinje 1987:154 claims that attraction accounts for the
agreement in the following sentence: For formannen var det ideene, ikke
saken som var viktig ´To the chairman, it was the ideas, not the matter,
that was important-SG´. It is not obvious that this example and 32-34
have any interesting grammatical properties in common.
6
10
It might be pointed out that some dialects do allow sentences like 37 (as
does English). Faarlund 1978:73 provides examples like the following:
38 Du gjer berre dreg det ut (From the author Tarjei Vesaas)
´You do only postpone it´
39 I sovore vêr gjorde ein ikkje ottast at nokon uroa dei (From the author
Jens Tvedt)
´In such weather one did not fear that anybody disturbed them´
But in most dialects, the verbal complement of gjøre must enter into an
unbounded dependency. Let us call this restriction R. (Note that this
restriction does not affect the proform det.) R seems to be a strange
restriction, as an argument which enters into an unbounded dependency
usually can appear "where it belongs" as well.
The necessity of the restriction R might be taken as an argument against
the auxiliary analysis. On the other hand, it will be needed by any
alternative to the pro VP-analysis which shares the assumption that the
topicalized VP is a complement of gjøre. And there is some evidence that
this kind of restriction is necessary in the description of natural languages.
In a theory like LFG, where there is no movement or traces, there is no a
priori reason that an argument which enters into an unbounded
dependency must be able to appear "where it belongs" as well. This is
pointed out by Kaplan et al 1987, who give the examples 40-41:
40 That he might be wrong he didn´t think of
41 *He didn´t think of that he might be wrong
A similar case from Norwegian is 42-43 (pointed out to me by Kirsti Koch
Christensen):
42 Penger har jeg ikke mange av
´Money I have not many of´
43 *Jeg har ikke mange av penger
´I have not many of money´
There is, then, some evidence that certain arguments of certain predicators
must enter into unbounded dependencies.
An independent argument for treating the gjøre-construction as a
construction with an obligatory unbounded dependency comes from
sentences like 44:
11
44 Spise gjør de ikke stort av (authentic)
´Eat they do not much of´
(Note that stort in 44 is used as a quantifier, just like mange in 42.) It is
impossible to imagine the corresponding sentence without an unbounded
dependency.
Accepting that certain arguments of certain predicators must enter into
unbounded dependencies, it would not be unnatural to expect the opposite
case: that certain arguments of certain predicators must not enter into
unbounded dependencies. And such cases seem to exist. Many speakers of
English do not allow indirect objects to enter into unbounded
dependencies, cf. 45 (from Ziv and Sheintuch 1979):
45 *The girl that I gave flowers is here
Another case is represented by the Swedish 46-477:
46 Jag är glad att du kom
´I am happy that you came´
47 *Att du kom är jag glad
´That you came I am happy´
The fact that certain arguments of certain predicators must or must not
enter into unbounded dependencies represents a challenge to our current
understanding of unbounded dependencies. The problem is, however, a
general one that does not only involve the gjøre construction. While we are
still waiting for a better understanding of unbounded dependencies, I think
the auxiliary analysis should be given a chance.
Another case is, curiously enough, represented by the non-auxiliary
raising verbs. Cf.
(i) *Å spille golf sa jeg at han forekommer meg
´To play golf I said he seems to me´
(ii) *Spille golf tror jeg aldri jeg har sett ham
´Play golf I think I have never seen him´
In LFG, both forekomme ´seem´ and se ´see´ are raising verbs. gjøre and
the non-auxiliary raising verbs seem to have opposite demands concerning
their verbal complement: It must, or must not, enter into an unbounded
dependency. (Note that the impossibility of (i) - (ii) would follow if,
contrary to LFG assumptions, their topics are taken to be non-maximal
projections.)
7
12
LITERATURE
Anward, J. 1988 Verb-verb agreement in Swedish. McGill Working
Papers in Linguistics. Special Issue on Comparative Germanic Syntax.
Faarlund, J.T. 1978 Verb og predikat. Ein studie i norsk
verbalsemantikk. Oslo.
Falk, Y.N. 1984 The English auxiliary system. Language 60.
Fretheim, T. 1974 The double object constraint on pronominalization.
University of Oslo Working Papers in Linguistics No. 5.
Hagen, J.E. 1979 Counting objects. A dubious constraint on Norwegian
pronominalization. Nordic Journal of Linguistics 2.
Hansen, A. 1967 Moderne dansk III. København.
Kaplan, R.M. and J. Bresnan 1982 Lexical-functional grammar: A formal
system for grammatical representation. In J. Bresnan (ed) The mental
representation of grammatical relations. Cambridge, Mass.
Kaplan, R.M. et al 1987 Functional uncertainty. The CSLI Monthly,
January. Center for the Study of Language and Information, Stanford
University.
Källgren, G. 1972 Restriktioner på verbfrasframflytting. PILUS 13
(Papers from the Institute of Linguistics, University of Stockholm.)
Källgren, G. and E.F. Prince 1989 Swedish VP-topicalization and
Yiddish Verb-topicalization. Nordic Journal of Linguistics 12.
Lie, S. 1979 Innføring i norsk syntaks. Oslo.
Lorentz, O. 1979 Norsk setningsform. Oslo.
Netter, K. 1988 Non-local dependencies and infinitival constructions in
German. In U. Reyle and C. Rohrer (eds) Natural language parsing and
grammatical theory. Dordrecht.
Thorell, O. 1977 Svensk grammatik. Stockholm.
Vinje, F.-E. 1987 Moderne norsk. Oslo.
Ziv, Y. and G. Sheintuch 1979 Indirect objects - reconsidered. In Papers
from the fifteenth regional meeting. Chicago Linguistic Society. Chicago.
Helge Lødrup
Department of linguistics and philosophy
Pb 1102, Blindern
N-0315 Oslo 3
NORWAY
Download