Professional Documents
Culture Documents
Light Verbs in Korean and Hindi The Locality Probl PDF
Light Verbs in Korean and Hindi The Locality Probl PDF
net/publication/228731450
CITATIONS READS
0 2,513
5 authors, including:
Some of the authors of this publication are also working on these related projects:
All content following this page was uploaded by James Pustejovsky on 30 May 2014.
Jongsup Jun
(Hankuk University of Foreign Studies)
Jongsup Jun. 2006. Light Verbs in Korean and Hindi: The Locality Problem as an
Epiphenomenon of the Surface Case Resolution. Language and Linguistics 37, 209-232.
A well-known property of the Korean light verb construction (=LVC)
is the locality mismatch between theta-marking and case-marking on the
accusative-marked theme; e.g. Kim paksa-ka mullihak-ul yenkwu-lul
hayssta 'Dr. Kim studied physics'. The accusative theme receives its
theta-role inside the verbal noun phrase (=VNP), but seems to be case-
marked by a verbal element outside the VNP. Interestingly, the locality
problem never occurs in the Hindi LVC. In this paper, I develop a case
theory under the framework of conceptual semantics (Jackendoff 1990,
1997, 2002) to explain why the locality problem occurs in one language,
and not in another. In this theory, case is assigned in the syntactic and
semantic modules independently; case alternation results from the
mismatch between syntactic case and semantic case. The theme argument
receives syntactic genitive inside the VNP, and semantic accusative
because it is lower in rank in the thematic hierarchy than the actor argument.
Korean allows case alternation when syntactic case conflicts with semantic
case. We set up the parameter in the opposite direction for Hindi; i.e. Hindi
realizes only the syntactic case when syntactic case does not match its
semantic counterpart. This simple parameter shows that the locality
problem is merely an epiphenomenon of the surface case resolution.
1. Introduction
study-ACC did
'Doctor Kim studied the origins of the earthquake'
The acceptability of the genitive theme in the true LVC varies from close
to ungrammatical to fully acceptable under relevant discourse situations.
The study of linguistic data becomes particularly difficult when the data
do not usually occur in everyday conversations. The use of elicited unnatural
data is analogous with laboratory experiments in biology or chemistry, where
the experiments are carefully designed to elicit certain data that are not
readily available in natural situations. This is why Jackendoff (1994) draws
parallelism between the elicited linguistic data and the experimental data
in laboratory situations. This is also why the linguistic science deals with
intuitively unclear and hence controversial data to explain speakers'
knowledge of a language (Schütze 1996). The genitive theme in (1a) is
not immediately acceptable to many speakers, perhaps because there are
other syntactic and pragmatic reasons to disfavor the use of the genitive
theme in the true LVC. Nevertheless, the fact that we can readily create
situations where the genitive theme can be used in the true LVC shows
there lies something behind the curtain.
2) The VN also licenses the agent role to the external argument Kim paksa.
In many earlier studies, the external argument is also the argument of the
light verb. Theories differ depending upon whether there is argument sharing/
fusion between the light verb and the VN (O'Grady 1995; J. Yoon 1991;
Butt 1995, 1997; T. Mohanan 1997; J-R Kim 1993; Kageyama 1991; J.
K. Lee 2001 Rosen 1989), whether the external argument of the light verb
controls that of the VN (Isoda 1991; Matsumoto 1992a, b; Sato 1993);
or whether the Korean HATA or the Japanese SURU is in fact a heavy
verb (Y-H Kim 1981; Hasegawa 1991; Kajihara 1991; Terada 1990;
Uchida & Nakayama 1993). Because of this complex situation, the internal
argument of the VN provides a better ground for explaining the locality
mismatch between case and theta-roles than the external argument.
212 언어와 언어학 제37집
In (2), mullihak gets both the theme role and the genitive case
inside the VNP. Both theta-marking and case-marking are local here.
When mullihak is marked accusative, however, it is outside the
VNP, as shown in (3).
Without further ado, I assume that the external argument Kim paksa
belongs to HATA in (1a) as far as the surface syntax is concerned; i.e.
there might be argument sharing, fusion, control, or whatever in the
deep syntax, but, on the surface, the external argument behaves as the
external argument of HATA.
3) The nominative-marked Kim paksa is always outside the VNP, and brings
about non-local theta-marking. Since Kim paksa gets both case and the
theta-role directly by HATA in some theories, the discussion of the external
argument tends to be theory-specific, or unnecessarily complicated (cf.
Footnote 3). For my thorough treatment of the external argument based
Light Verbs in Korean and Hindi▪Jongsup Jun 213
does not match the theta-marking domain in (3). This is the locality
mismatch between case-marking and theta-marking in the Korean
(and also Japanese) LVC.4)
The LVC is important in south Asian languages like Hindi. T.
Mohanan (1994, 1997) discusses two types of the Noun-Verb complex
predication in Hindi.
light verb (ki-) agrees with them in gender (i.e. masculine kiyaa vs.
feminine kii). Notice that mohan gets NP-internal genitive when the
VNs agree with the light verb in gender.
In (5), however, the nominative-marked VNs (pasand 'liking' and
ksamaa 'forgiveness') do not agree with the light verb. The verb is
inflected for the default masculine singular -aa, even though its
nominative-marked arguments are feminine. When the VNs do not
agree with the light verb in gender, the internal argument mohan gets
verbal case.
After studying behavioral differences in agreement, passivization,
negation, and gapping between (4) and (5), Mohanan (1994, 1997)
concludes that the host VNs in (5) are inherent part of the VN-LV
complex predicates, whereas the host VNs in (4) are syntactic arguments
of the verb. The bracketing in (6) and (7) shows the difference.
Since mohan cannot get verbal case in (8) and (9), no locality mismatch
between theta-marking and case-marking arises in the true LVC in
Hindi.5)
The difference between the Korean LVC and the Hindi LVC immediately
raises one important theoretical and empirical issue. Why does the
locality mismatch occur in one language (i.e. Korean), and never in
5) It is interesting that Hindi scholars have not shown interest in the locality
problem for the light verb construction (Miriam Butt, p.c.). As far as Hindi
is concerned, the locality mismatch between theta-marking and case-marking
never arises, and hence never asks for special attention from scholars. When
we look at Hindi from a cross-linguistic perspective, however, the locality
problem arises. That is, we are struck by the fact that the Korean/ Japanese
LVC displays the locality mismatch whereas the Hindi counterpart does not.
216 언어와 언어학 제37집
(10) a. VN (x, y, z)
b. LV ( ) <ACC>
c. VN (z) + LV (x, y) <ACC>
Light Verbs in Korean and Hindi▪Jongsup Jun 217
3.2 Morphosyntax
VNP LV
VN LV
I assume that the two minimal structures for the LVC in (14) are
constructions in the sense of construction grammar (Goldberg 1995),
and that these constructions are lexical entries like regular words.
The lexicon also has a number of VNs, light verbs like HATA. (15)
shows part of such a lexicon.
(15) a. VP-construction
VP
VNP LV
b. V-construction
V
VN LV
c. HATA LV
d. pinan 'blame' VN
e. yenkwu 'study' VN
This will yield the true LVC yenkwu-lul hata. We do not need any
esoteric pre-syntactic lexical operation or syntactic movement.
8) RRG (=Role and Reference Grammar) also develops its semantic case theory
that has nothing to do with meaning per se (Van Valin & LaPolla 1997).
In RRG, case assignment is semantic, but it refers to a thematic hierarchy,
224 언어와 언어학 제37집
One may wonder why we need this mechanism for semantic case
when we already have a mechanism for syntactic case. The reason
lies in the architecture of language in conceptual semantics. Since
syntax and semantics are two autonomous levels of grammar, we need
correspondence rules between syntax and semantics. In order to define
correspondence rules, we first have to identify syntactic and semantic
entities that correspond to each other. In other words, we need an
interface component between syntax and semantics. J. S. Jun (2003)
argues that case plays this interface role between syntax and semantics.
To illustrate the point, we have three NPs in syntax and three THINGs
in semantics. We mark one of the three NPs with nominative, and
mark one of the three THINGs with nominative. Correspondence rules
refer to this case marking in both syntax and semantics, and map the
NP with nominative marking in syntax onto the THING with nominative
marking in semantics. For this reason, syntactic and semantic markings of
case are required by the foundational architecture of conceptual semantics.
The crucial aspect of this theory is the surface case resolution. In
most cases, syntactic case matches its corresponding semantic case.
That is, an NP gets nominative in syntax, because it is the subject.
Its corresponding THING argument gets nominative in semantics,
because it is the actor. Therefore, the NP gets nominative because
its syntax and semantics collaborate to mark it nominative.
A problem occurs when syntactic case conflicts with the semantic
case. When syntactic case does not match the semantic case, languages
differ in choosing syntactic case over semantic case, semantic case over
syntactic case, either syntactic or semantic case, both syntactic and
semantic cases, etc. The LVC provides one such example. In the LVC,
the NPOBJ of a VN remains inside the VNP, as shown in (18). Following
the standard assumption, I assume that the adnominal genitive is strictly
(18) VP
VNP LV
NPOBJ VN
GEN
(18) and (19) show that there is mismatch between the syntactic
genitive and the semantic accusative.
I assume that Korean and Japanese choose either syntactic or
semantic case when the two cases do not match.9) This explains why
we have case alternation between genitive and accusative in the true
LVC in Korean. Because the NPOBJ is an internal argument of the
VN, it is generated inside the VNP, and gets adnominal genitive in
syntax. On the other hand, the corresponding THING argument of
the NPOBJ in the conceptual structure gets accusative. Because Korean
allows either syntactic or semantic case when they conflict, either
genitive or accusative surfaces, as shown in (1a).
What I propose for the Hindi LVC is that it chooses only the syntactic
case when it conflicts with the semantic case. When the syntactic genitive
9) For detailed arguments, see the discussion of the aspectual nominal construction
in Korean and Japanese in J. S. Jun (2003).
226 언어와 언어학 제37집
conflicts with the semantic accusative as in (18) and (19), only the
syntactic genitive surfaces. This explains why case-alternation never
occurs on the NPOBJ in the Hindi LVC.
(22) *VP
NPTheme VNP LV
NPGoal VN HATA
(23) VP
NPGoal VNP LV
NPTheme VN HATA
4. Conclusion
Japanese LVC in the literature, the locality mismatch does not occur
in Hindi. That is, the locality mismatch per se must be parameterized
to occur or not to occur in any theory. In this paper, I have tried to
parameterize the locality mismatch in terms of the surface case resolution
between the syntactic and the semantic case in my conceptual semantic
case theory. Within this view, case alternation in the Korean LVC is
the result of the mismatch between the syntactic genitive and the
semantic accusative. One empirical gain of the theory is that it provides a
natural account for the hierarchical argument transfer. Now that we
have gained insights into case as well as the light verb construction
in Korean and Hindi, it is time to pursue our inquiry into case based
on more cross-linguistic data.
참고문헌