observed a timing difference with respect to the locality effects and expectations.
As we suggest, checking the compatibility of the relevant feature must take place
after the integration of the verb has been completed. Note that what previous
studies have dealt with concerning expectations was the lexical content of the
verb itself; for instance, Lev y and Keller (2013) used information on the pos-
sible argument structures of the verb, and Hus ain et al. (2014) used the col-
location of the verb and certain object NPs. Therefore, upon integrating the
verb into the structure, all the relevant checks about the lexical content of the
verb were done at that step. In our current case, the difference in timing was
observed because we dealt with complementizers that appeared as a suffi x, which
may be an independent syntactic projection in the structure.
Finally, we observed another timing difference between the High-Q condi-
tions. It seems that the emergence of TMEs was slightly faster in the local ×
High-Q condition than the distant × High-Q condition. Between the two local
conditions, TMEs were observed in region 9, and probably the effects remain
through region 10. On the other hand, TMEs in the distant × High-Q condi-
tion were clearly observed only in region 10. Although we have to wait for
some further investigation into the timing issue with respect to TMEs, we could
hypothesize that the difference was due to the dependency length between the
verb with the complementizer and the wh-phrase in the structure. Again, assum-
ing that part of TME involves the checking process for the compatibility of the
relevant features, it is reasonable to consider that it is more costly to check the
feature compatibility that spans a long distance. For instance, Spr ouse, Fukuda,
Ono, and Kluender (2011) showed that wh-interrogatives in Japanese were
judged worse when the distance between the wh-phrase and the licensing
Q-particle was long. Then, we could say that TMEs for the distant conditions
were found later than that for the local conditions because the checking process
for the feature compatibility was distance-sensitive.
7 Conclusion
We started our discussion in this chapter by raising a question about how to
make a good prediction. Focusing on the sentence-processing mechanism, we
observed that there is a certain benefi t for having more constituents. A reader
can expect a very specifi c verb to appear and can process the verb faster when
case-marked NPs narrow down the choices of verbs. At the same time, slow-
downs at the verb are observed due to the working memory cost incurred by
processing long-distance grammatical dependencies. Based on recent works by
Lev y and Keller (2013) and Hus ain et al. (2014), it is now clear that expecta-
tions and locality effects infl uence each other; when there are specifi c expecta-
tions, the locality effects are suppressed, but locality effects show up when
expectations are not strong. In this chapter, we provided data showing that
such a mutually exclusive situation arises not only where predictions are based
on thematic properties of a verb (such as the argument structure and the object-
verb collocation), but also where the reader can predict the wh-licensing
Make a good prediction 99