学術交流支援資金「海外の大学等との共同学術活動支援」活動報告書
慶應義塾大学環境情報学部
今井むつみ
研究課題名： 動詞語意学習における文法情報の利用についての国際比較研究
メンバー
|
今井むつみ |
環境情報学部・教授 |
|
重松淳 |
総合政策学部・教授 |
|
佐治伸郎 |
政策･メディア研究科博士課程2 年 |
|
Hua Shu |
北京師範大学・教授 |
|
Guillaume Thierry |
|
|
Henrik Saalbach |
ETH |
Purpose and
Background. Argument
structure of a sentence functions as a constraint in inferring verb meanings.
English-speaking infants assume that an intransitive sentence canonically maps
to an event with nosn-caused motion and a transitive sentence maps to a caused
motion (Naigels, 1990). A recent theory proposes that children universally rely
on argument number over case-marking in making this structure-semantics mapping
(Litz, 2006). However, the argument-number constraint causes a dilemma for
children learning an argument-dropping language: when hearing a sentence with
just the subject, it is always possible that the sentence is a transitive
sentence with the object dropped; When the subject of the transitive sentence
is dropped, although the patient role should be unambiguously identified (by an
accusative case-marker in Japanese, and by word order in Chinese), it violates
the argument-number constraint. We examined how Japanese and Chinese children
deal with these problems when they infer the meaning of a novel verb. If the
argument-number hypothesis is at work, Japanese and Chinese children should map
a verb presented in a subject-only
sentence to a non-caused motion in spite of the possibility that it may be a
transitive verb. In contrast, mapping a verb appearing in the object-only frame to a caused action
should be difficult even though it is unambiguously transitive. The
developmental trajectory in the pattern of mapping is also of great theoretical
interest.
Method. Monolingual
3- and 5-year-old Chinese- and Japanese-speaking children, 30 in each age/language,
were shown two video clips side by side. In the subject-only condition, children heard a sentence containing a
novel verb only with the subject (‘usagi (rabbit) ga (Subject-marker) X-teiru (Progressive) in Japanese; ‘tuzi
(rabbit) zheng-zai (Progressive) X’ in Chinese, X beings a novel verb) while
watching two videos of (1) a character (a rabbit) doing a non-caused motion and
(2) a character (a rabbit) doing a caused motion to another character (a bear).
In the object-only condition, the
agent of the non-caused motion (rabbit) in Video 1 appeared as the patient of
the caused motion in Video 2, and they heard a sentence only with the object
(‘rabbit o (Accusative) X-teiru’ and
‘Zheng-zai X tuzi.’).
Results and
Discussion. A striking Age X Condition
interaction was found for both language groups. The 3-year-olds mapped the
subject-only sentence to the non-caused action, but failed to map the
object-only sentence to the caused action. The older children in both languages
responded at chance in the subject-only condition, but successfully mapped the
object-only sentence to the caused action involving two participants.
Thirty-two Japanese 2.5-year-olds, who were additionally tested, mapped both
the subject-only and object-only sentences to the non-caused action involving
only a single agent.
The results indicate that young
children universally start out with a simple hypothesis that a verb appeared in
a subject-only sentence is intransitive and maps it to a non-caused action, but
older children become aware of the ambiguity of this sentence frame. A
transitive sentence without the subject is difficult to map, but by 5-year-olds
of age, children overcome this problem by paying attention to a
language-specific cue (i.e., case marking and word order).
Table 1. Proportion of
mapping the subject-only sentence to
the non-caused
action
|
|
Japanese |
Chinese |
|
2-year-olds |
67.25* |
-- |
|
3-year-olds |
75.5** |
72.50** |
|
5-year-olds |
50.0
ns |
60.75ns |
*: significantly above
chance, p<.05; **: p<.01
Table 2. Proportion of
mapping the object-only sentence to
the caused action
|
|
Japanese |
Chinese |
|
2-year-olds |
28.50## |
-- |
|
3-year-olds |
63.25ns |
50.00ns |
|
5-year-olds |
75.0** |
82.5** |
##: significantly below
chance (i.e., mapping to the non-caused action), p<.01
*: significantly above
chance, p<.05; **: p<.01.