In their self-paced reading experiment, they observed a main effect of expec-
tations at the crucial verb region; the high expectation conditions were in general
read faster than the low expectation conditions. Also, they observed a significant
interaction between the expectation factor and the distance factor. Specifically,
while the presence of an extra adverbial did not increase or decrease the reading
time of the verb in the low expectation conditions, a speedup was observed in
the long × high expectation condition, compared to the short × high expecta-
tion condition. In other words, the verb was read faster when the distance
between the object and the verb was long. One thing that should be paid
attention to in their results is that they increased the length between the object
and the verb with an adverbial phrase. Recall from the German experiments of
Konieczny and Döring (2003) and Levy and Keller (2013) that addi ng an
adjunct phrase in the structure did not facilitate the processing of the verb. It
seems unlikely that the adverbial greatly contributes to the prediction of the
verb because, in the high expectation conditions in this experiment, the verb
was already predicted. It could be that such a specific sentential context somehow
boosts the impact of adding an adverbial, which is usually not strong enough
to facilitate the processing of the verb. At this point, it is unclear what exactly
causes the difference between German and Hindi cases, but further investigation
is necessary about the role of an adverbial phrase with respect to the locality
effects in those languages.
When we return to the main findings of their experiment, the results indicate
that the impact of locality (i.e., the distance between elements to be integrated)
is highly modulated by expectations. The Hindi case shows that the locality
effects (slowdowns) appear only when there is no strong expectation at work.
Husain et al. (2014) argued that such complementarity arises because the
parser can build a detailed structural representation when the verb is highly
expected given the preverbal constituents. In the high expectation conditions
above, the parser not only builds a detailed phrase structure for VP, but also
integrates the expected verb before the parser actually encounters the verb.
Note that the locality effects are assumed to be, at least in a substantial part,
a cost of integrating the verb and the dependent elements in the structure
(Gibson 1998 among others); there fore, no locality effects should be found
at the verb region when the verb is fully predicated in advance, evoking no
need for integration at the point of the actual encounter. Husain et al. found
a distance-based expectation facilitation (i.e., the more intervening elements,
the easier) only in the high expectation conditions because the counteracting
locality effects (the more intervening elements, the harder) were absent.
5 An interim summary and forward
The findings in Levy and Keller (2013) and Husai n et al. (2014) show that
t he facilitation effects based on expectations and the processing costs based
on locality are two forces that counteract each other. In one situation where
90 Hajime Ono et al.