Rayner, 1982, Crain a nd Steedman, 1985, Altmann and Steedman, 1988,
Trueswe ll, Tanenhaus, and Garnsey, 1994, Tanenha us, Spivey-Knowlton, Eber-
hard, and Sedivy, 1995, Bader, Meng, and Bayer, 2000, Miyamot o, 2002, among
many others). In this chapter, however, we focus on a very restricted set of
“useful” information, so the kinds of information that will be discussed here
are by no means comprehensive. We review some relevant work in the following
section.
Assuming that having more information is beneficial, we can still easily imagine
the cases in which having many constituents before seeing a verb would not
help us predict the identity of the verb. One such case is the one in which the
intervening constituents are not arguments or adjuncts directly associated with
the verb. In fact, having too many constituents seems to cause problems for
processing a verb. It is known in the sentence-processing literature that there
is an increase of the memory load, called locality effects, when the parser has
to process a dependency that spans a long distance (Yngve, 1960, Chomsky
and Miller, 1963, King an d Just, 1991, Gibson, 1998). The parser may slow
down due to having too many constituents in such a case.
In this chapter, we are interested in the relation between expectations and
locality effects that seem to be counteracting each other. Having a lot of infor-
mation helps the reader make a good prediction about the verb, but a large
amount of information involved in an incomplete dependency may impose a
lot of memory load. Previous work on expectations shows that not all types of
information lead to the facilitation of the processing of the verb. We review
Koniecz ny and Döring (2003) in the next section, who reported that a preverbal
constituent contributed to the facilitation only when it was an argument of the
verb. In section 3, we review relevant observations about locality effects. After
discussing expectations and locality effects, we will discuss some recent empirical
findings that locality effects are not independent of expectations, but are influ-
enced by expectations. Manipulating the phrase structure in German embedded
sentences, Levy an d Keller (2013) found that locality effects sometimes emerged
with the facilitation by expectations suppressed, and locality effects sometimes
disappeared with the facilitation by expectations observed. Furthermore, Husain,
Vasishth, and Srinivasan (2014) found that the locality ef fects wer e obser ved
only when expectations were weak.
Then, one question arises: what kind of information leading to expectations
influences the emergence of the locality effects? We will introduce results from
our experiment in which the locality and expectations are manipulated in wh-
interrogative sentences in Japanese. The results show that the locality effects
disappeared when there was a specific expectation at work, which is quite similar
to the results in Husain et al. (2014). Also, we will see that an expectation-
related slowdown occurred later than the locality effects. This finding is particu-
larly noteworthy because no previous work reported that locality effects and
the expectation-related effects occurred at the same time in a single sentence.
We argue that the way expectations relate to the locality ef fects is dif ferent
depending on the type of prediction involved.
84 Hajime Ono et al.