| |
Abstract:
Consider the following sentence.
(1) Barney attacked the husband of the woman who was reading a
newspaper.
Who was reading a newspaper?
When native English speakers read syntactically ambiguous
sentences such as (1), the preferred response is
the woman.
The Construal hypothesis for sentence processing predicts the Late
Closure response (Frazier & Clifton, 1996). In contrast, if
native Japanese speakers and native Spanish speakers read sentences
such as (1) in Japanese or Spanish, the preferred response is
the husband.
The Tuning hypothesis predicts the Early Closure response for
native Japanese and native Spanish speakers (e.g., Kamide &
Mitchell, 1997; Mitchell & Cuetos, 1991). A
language-independent model of sentence processing suggests that
bilinguals use a consistent parsing strategy in both languages;
i.e., they use the preferred parsing strategy in their native
language and in their second language (Bates & MacWhinney,
1981; MacWhinney, 1997). A language-dependent model, on the other
hand, suggests that bilinguals use language-specific parsing
strategies; i.e., bilinguals switch parsing strategies when they
switch languages (Fernández, 1998). Previous cross-lingual
research has paid no attention to the language of the instructions.
For example, readers tested on Spanish materials are instructed
typically in Spanish; readers tested on Japanese materials are
instructed typically in Japanese. The present research was designed
to examine the impact of the language of instruction when the
participants are instructed in their native language versus their
second language.
Twelve Japanese-English bilinguals read Early Closure, Late
Closure, and Ambiguous sentences. (See Examples.) Participants read
instructions printed on an index card; half the participants read
instructions printed in English, the other half read the
instructions printed in Japanese. All texts were presented in
English. A two-choice comprehension question followed each
sentence.
When native Japanese speakers read Early Closure and Late
Closure sentences printed in English, they demonstrated the typical
native-English Late Closure preference strategy. An interesting
pattern emerged, however, for readers instructed in English versus
Japanese. When instructed in English, readers were more likely to
make Late Closure responses to Late Closure sentences than when
instructed in Japanese. In contrast, when instructed in Japanese,
readers were more likely to make Early Closure responses to Early
Closure sentences than when instructed in English. The instruction
impacted ambiguous sentences differentially. Specifically, when
instructed in English, readers tended to respond to ambiguous
sentences with Late Closure choices; when instructed in Japanese,
they tended to respond to ambiguous sentences with Early Closure
choices. These results suggest that when discourse clues in the
written text are absent, the context of the environment (i.e., the
language of instruction) plus its accompanying parsing preference
guide sentence processing. A cross-linguistic model that
incorporates both language-dependency and language-independency
principles will be discussed.
|
