(8)a. Early, clause-mate Q-particle:
dare-ga kita-ka sitteiru.
who-NOM came-Q know
“(I) know who came.”
b. Late, higher Q-particle
dare-ga kita-to omoimasu-ka?
who-NOM came-C think-Q
“Who do (you) think came?”
In the experiment, we varied the position of Q-particle ka, crossed with the
manipulation of the distance between the wh-phrase and the verb. Note that
the presence of wh-phrase made the test materials just like the Hindi high-
expectation conditions. In Hindi, the specific lexical content of the verb is
predicted by the object, while in our Japanese sentences, what is predicted by
the wh-phrase is the encounter with the Q-particle at the earliest possible posi-
tion. Thus, it is worth testing whether the locality effects would be incurred
by this type of grammatical expectation. Additionally, in the conditions in which
the Q-particle is not attached to the theta-assigning verb, a surprisal effect might
be incurred, because the parser would prefer to resolve incomplete dependencies
as soon as possible (Miyamoto and Takahashi, 2002; mor e on this below). In
a sense, these conditions can be considered counterparts of Hindi low-expectation
conditions. We would like to see if stronger locality effects are observed in these
“less expected” conditions than in the “expected” conditions in which the
Q-particle is attached to the clause-mate, theta-assigning verb.
Another important point we should pay attention to with respect to the
findings in Levy and Keller (2013) and Husain et al. (2014) is that they observed
locality effects when the structure being processed requires a lot of working
memory resources. In fact, in an experiment by Grodner and Gibson (2005),
the lo cality effects were not so pronounced when the structure was simple; their
locality effects clearly appeared when the relevant part of the sentence was
embedded inside of relative clauses. In Hindi experiments by Husain et al.
(2014), the sentenc es were rather long, which obviously consumes much work-
ing memory. It has been shown in various measurements that sentences with
wh-dependency require more working memory resources to process in general
(King and Kutas, 1995, Kluender an d Kutas, 1993, Kaan et al. 2000, Phillips,
Kazan ina and Abada, 2005). We examine the locality effect using wh-interrogative
sentences in Japanese.
6.1 Method
Thirty undergraduate students participated in a self-paced reading experiment
run by Linger software (written by Douglas Rohde). Twenty-four sets of
target sentences were prepared and distributed into four lists according to
92 Hajime Ono et al.