Ono, H. 1 , Yoshida, M. 2 & Fujii, T. 3
1 Tsuda College
2 Northwestern University
3 Yokohama National University
This study shows that the parser predicts the upcoming syntactic structures and utilizes the predicted structure to resolve the wh-verb dependency. A self-paced reading experiment on Japanese multiple wh-questions (1) reveals that, the matrix Wh-subject leads to the prediction of the upcoming matrix interrogative-verb (V-int(errogative)), and the parser links the embedded Wh-object to this predicted but not-yet seen matrix Verb-int, resolving wh-dependency.
2x2 factorial design manipulated the matrix Subject type (Wh/NP-subject) and Embedded Verb type (V-int/dec(larative)).
(1) {Wh/NP-subject} [NP-subject Wh-object {V-int/dec}] V-int
The RT difference of the embedded verb reached a significant interaction (t=-2.22, p<.05): Pairwise comparisons revealed V-int > V-dec in the Wh-subject conditions (t=-1.70, p<.09). Normally the parser links the embedded Wh-object to closest embedded V-int: A surprise effect emerges, if it reveals as V-dec (Miyamoto & Takahashi 2001). In this experiment, however, the V-int, the closest to the embedded Wh-object, caused a slowdown. This result indicates the following: The parser attempts to link the Wh-object to matrix V-int, which is predicted by the presence of wh-subject. Because the Wh-object had already been linked to the V-int, another V-int in the embedded clause was not expected.