actually seeing it. Then, when the reader sees the actual verb, the fully elaborated
phrase structure has already been built, and there is not much to do in terms
of integration. In our current results, the Q-particle was fully predicted by
the presence of the wh-phrase. The reader can build the structure with the
Q- particle that is yet to be seen. When the reader actually sees a verb with
the Q-particle, the representation of the Q-particle has already been integrated
into the structure, presumably without the lexical content of the verb by itself.
Turning to the High-Q conditions, where the verb was followed by a declara-
tive complementizer to, we observed locality effects: the distant × High-Q
condition was read slower than the local × High-Q condition. The fact that
locality effects showed up in this comparison also fits well with the findings
from Hus ain et al. (2014). In the target sentences, the verb without the Q-particle
was not expected, given the existence of the wh-phrase in the clause. We could
interpret the finding along the line of the analysis elaborated in Hus ain et al.
(2014): because the verb with the declarative complementizer to was not expected
at all, the representation for the declarative complementizer could not be sup-
posed or built in before the reader actually saw the declarative complementizer.
Because no benefit was available from the expectation-based preprocessing, the
verb and the complementizer in the High-Q conditions must be processed upon
the reader seeing the verbal complex. It has been assumed that the integration
of the verb requires the reactivation of the arguments, and it is quite costly to
reactivate linearly distant elements because the activation of such an element
has already decayed (Lew is 1996, Gib son 1998). Therefore, the distant × High-
Q condition was read slowly, compared with the local × High-Q condition.
Note that in addition to the locality effect in region 8, we also observed
slowdowns of High-Q conditions in region 9 and 10, though the detailed pat-
terns are slightly different. Let us first look at the effects we observed in region 9.
In this region, the local × High-Q condition was read slower, compared with
the local × Mid-Q condition, and it is plausible that this is a TME (Miy amoto
and Takahashi 2002; we will come back to the lack of TMEs in the distant
conditions shortly). This finding is quite crucial because as far as we are aware
of, no timing difference has ever been observed with respect to the emergence of
locality effects and expectation-related effects, such as TMEs in our case (see
Ono and Nakatani 2010, 201 4 for some related discussions). Also, the current
results showed that locality effects were observed earlier in the sentence than
the expectation-related effects. We believe that there is a principled reason why
locality effects must be observed earlier than TMEs. Recall that locality effects
are claimed to reflect the processing costs associated with the integration of the
verb into the structure (Gib son 1998). TMEs, on the other hand, could be a
slowdown effect involving the checking process with respect to the feature
compatibility between the wh-phrase and the verb plus the particle. If such a
checking process can take place only after the verb and the relevant morpheme
are actually integrated into the structure, then TMEs must be observed after
the locality effect, which is a cost associated with the integration of the verb.
The above interpretation can also account for why no other previous work has
98 Hajime Ono et al.