Edson T. Miyamoto1 & Shoichi Takahashi2
1 Nara Institute of Science and Technology, 2 Massachusetts Institute of Technology
It has been argued that the processing of in-situ wh-phrases in Japanese is similar to the processing of fronted wh-phrases in English ([1]). For the latter construction readers posit a gap as early as possible, whereas for the Japanese construction it has been proposed that readers posit a question particle (QP) at the first complementizer position available as in the following.
(1) EmbeddedSubject-nom WhObject-acc V-that MatrixSubject-nom MatrixV-QP?
In (1), the embedded object wh-phrase requires a QP whose earliest possible occurrence would be in the embedded complementizer position. However, because this position is filled with the affirmative complementizer that, a slow-down (compared to a non-wh condition) has been shown to occur due to a typing mismatch effect (TME; [1]), in a parallel to the filled-gap effect (FGE).
The symmetry between the TME and the FGE if correct may lead to a better understanding of wh-phrase processing across languages. However, there are alternative explanations for the slow-down in (1), which do not require assuming a grammatical dependency between whs and QPs. If the parallel between the TME and the FGE is to hold, the TME should only occur at grammatically relevant positions (see [2] for the absence of FGE within islands).
We conducted a self-paced reading experiment in which a wh-phrase can only be grammatically licensed by a QP in the matrix complementizer position as in (2).
(2) WhMatrixSubject-nom EmbeddedSubject-Nom Object-acc V-that MatrixV-QP?
A QP can occur in the embedded complementizer position, but it cannot license the wh subject in the matrix clause (i.e., the wh would still require a QP in the matrix clause). The TME predicts no slow-down at that because the wh creates no expectation for a QP in this position. The prediction was confirmed: (2) is numerically faster than a non-wh condition, whereas there is a slow-down in (3) below (2x2 interaction: F1(1,30)=4.62, p<0.05; F2(1,23)=4.28, p=0.05).
(3) WhEmbeddedSubject-nom Object-acc V-that MatrixSubject-Nom MatrixV-QP?
Although (3) was read more slowly at V-that than in a non-wh condition, the difference was not statistically significant. Thus, another experiment was conducted including the scrambled sentence (3) and its canonical counterpart, (4).
(4) MatrixSubject-Nom WhEmbeddedSubject-nom Object-acc V-that MatrixV-QP?
There was no interaction between (3), (4) and their respective wh controls (with embedded QPs instead of embedded that). But there was a main effect as (3,4) were significantly slower than their controls (F1(1,37)=6.14, p<0.05; F2(1,23)=3.68, p= 0.067). The result suggests that a TME occurs for wh-subjects, although its magnitude seems smaller than for wh-objects as in (1). Such a subject/object asymmetry in the TME may shed light on the equivalent asymmetry for the FGE ([2]) and is currently being investigated in new experiments.
References
[1] Miyamoto, E. T., & Takahashi, S. (2001). The processing of wh-phrases and interrogative complementizers in Japanese. In N. Akatsuka, S. Strauss, & B. Comrie (Eds.), Japanese/Korean Linguistics, 10. Stanford: CSLI.
[2] Stowe, L. A. (1986). Parsing WH-constructions: Evidence for on-line gap location. Language and Cognitive Processes, 1, 227-245.