the parser can make a strong prediction, the facilitation effects based on
expectation are more pronounced, and in another situation where properties
of the verb are not highly predictable, locality effects do appear more strongly.
As we discussed above, Husain et al. (2014) further arg ue that such comple-
mentarity between expectations and locality arises because, when an item is
highly predictable, the parser can integrate the item into the structure before
the parser actually encounters that item. In the Hindi case they discussed,
the crucial cue for the parser to make a prediction is information about the
collocation of the object and the verb. On the other hand, the crucial cue
for the prediction in the German cases of Levy and Keller (2013) and
Konie czny and Döring (2003) is t he presence of NPs in the preverbal posi-
tion that helps the parser predict the argument structure of the verb to
follow.
Now we would like to investigate further other types of prediction by the
parser that may lead to the modulation of the locality effects. If Husain
et al.’s (2014) argument is applicable to more general cases, then it seems
that a wide range of information about the verb, including highly grammatical
cues, may impact the locality effects. On the other hand, it could be that
their argument may be applicable only to certain types of expectation, such
as information about the argument structure of the verb. In the following
experiment, we used wh-interrogatives in Japanese as target sentences. The
use of wh-interrogatives provides a particularly interesting testing ground for
further investigating the interaction between the locality effects and expec-
tation-based facilitations, because they involve two types of grammatical
dependencies at the same time: thematic dependencies and wh-Q dependen-
cies. The presence of a wh-phrase strongly invokes an expectation for a
Q-particle to come, which may or may not be attached to its theta-assigning
verb. If the prediction for the Q-particle suppresses the locality effects, then
the suggestion by Husain et al. (2014) is shown to be applicable quite gener-
ally, in that information other than the argument structure of the verb has
an influence on the locality effects.
6 Experiment
We conducted a self-paced reading experiment, examining whether informa-
tion other than the argument structure of the verb would influence the
emergence of locality effects. As briefly discussed above, we set up a paradigm
of expectation triggered by wh-phrase, which obligatorily forms a dependency
relationship with clause-final Q-particle ka in Japanese. The Q-particle is a
scope-marking verbal suffix, and it may appear with the verb assigning a
thematic role to the wh-phrase. This is the earliest possible position where
the required Q-particle may be placed. Alternatively, the Q-particle, being a
scope marker, may be attached to a verb in a clause higher than the one
containing the wh-phrase.
Make a good prediction 91