effect emerged due to the slow reading time in the local × High-Q condition.
In order to examine the slow reading time in the local × High-Q conditions,
further planned comparisons were conducted. We found that there was a simple
main effect of the Q-position in the local conditions (β −75.08, SE = 27.73,
t value = −2.71), showing that the local × High-Q condition was read slower
than the local × Mid-Q condition. On the other hand, such an effect of the
Q-position was absent in the distant conditions.
Finally, in region 10 (a nominative NP, a spillover region 2), there was a
main effect of the Q-position, showing that the High-Q conditions were read
slower than the Mid-Q conditions.
6.3 Discussion
One of the main interests in this experiment was to examine whether and how
the locality effects and expectations influence each other. Given the findings in
Le vy and Keller (2013) and Hu sain et al. (2014), we were specifically interested
in testing whether one would suppress the other, and whether the locality effects
would be observed only when there was no strong expectation. Below, we will
briefly summarize the findings and discuss implications we can draw for the
connection between the locality effects and expectations.
In region 8 (critical region), there was a main effect of locality, indicating that
the critical verb in the distant conditions was read slower than that in the local
conditions. Although it is unfortunate that the interaction of the two fixed effects
was not significant, a further planned pairwise comparison suggested that the
slow reading time of the distant × High-Q condition was driving the main effects
of locality. The reading time data showed that there was no locality effect in the
Mid-Q conditions where the Q-particle appeared with the critical verb. Further-
more, the comparison between the two High-Q conditions where the Q-particle
was missing from the verb that was in the same clause as the wh-phrase showed
that there were actually locality effects observed between the two.
As discussed above, if we take the Mid-Q conditions as equivalent to Husain,
et al’s Hindi high expectation conditions, and the High-Q conditions as the
Hindi low expectation conditions, the current pattern of results in this region
turned out quite similar to what has been observed in Hus ain et al. (2014):
locality effects showed up only when expectations were not strong. In other
words, when expectations were in fact strong (our Mid-Q conditions), locality
effects did not show up. Such a correspondence between Hindi data and Japa-
nese data leads us to claim that the types of expectations that may suppress
locality effects are not restricted to information about argument structure of
the verb, but are more generally found. In the current experiment, what was
predicted by the wh-phrase was the presence of the Q-particle appearing at the
critical verb as a suffix, and obviously this wh-licensing morpheme had no direct
relation to the argument structure of the verb. Recall that Hus ain et al. (2014)
argue that locality effects are not observed in the high expectation conditions
because the reader can expect a specific verb to appear in the high expectation
conditions, and the reader can integrate the verb into the structure before
Make a good prediction 97