had no major impact on the reading time of the verb, suggesting that the lexical
or grammatical information that lowers the processing cost of the verb is selec-
tive. Having more materials diminishes the reading time only if the presence of
those materials can reduce the number of possible outcomes and contribute to
a better prediction, possibly by providing more specifi c information about the
argument structure of the verb (see also Konieczny, 2000 ). It is certainly at issue
exactly what kind of lexical or grammatical (and presumably many other) infor-
mation is connected to expectations discussed so far (Vasishth and Le wis, 2006,
Levy, 2008, Nak atani and Gi bson, 2010, Staub and Clift on, 2006).
3 Locality effects
Although it has been reported in the literature that more information leads to
processing facilitation, there are opposite cases in which adding more materials
in the structure slows down the parser. For instance, Grodner and Gib son (2005)
observed reading-time slowdowns, called locality effects, when the length of the
dependency between the two linguistic elements becomes larger. Dependency
Locality Theory (DLT, Gibson, 2000) a nd Similarity-Based Interference (Gordon,
Hendric k and Johnson, 2001, 2004) are repre sentative accounts in the literature
for these distance-based “locality” effects (Warren and Gibs on, 2002).
Grodner and Gib son (2005) compared the following set of materials, in which
the complexity of the embedded subject in the relative clause is manipulated.
In (3a), the subject is a simple NP, the nurse; in (3b), a PP, from the clinic, is
attached to the subject; in (3c), the subject is followed by a relative clause.
(3) a. The administrator who [the nurse] supervised scolded the medic while...
b. The administrator who [the nurse from the clinic] supervised scolded
the medic while...
c. The administrator who [the nurse who was from the clinic] supervised
scolded the medic while...
They measured the reading times of the embedded verb supervised and observed
that the verb in (3c) was read the slowest, and there was a gradual increase in the
reading times from (3a) to (3c). According to DLT, the reading time of the verb
supervised is partially determined by the number of discourse referents and events
intervening the verb and its subject. Compared with (3a), (3b) and (3c) had a dis-
course referent clinic between the nurse and supervised, the presence of which incurs
a cost for integration at the verb. In (3c), the tensed verb was, representing an event,
incurs another cost, thus the reading time for the verb in (3c) was the greatest.
The locality effects are not limited to the dependency between the subject and
its predicate, but also observed in various long-distance dependency relations. For
instance, Phillips, Kazan ina, and Abada (2005) compared the following two sen-
tences with different dependency lengths. The wh-phrase which accomplice origi-
nates as the object of the verb recognize (noted with the line) in both sentences.
In (4a), the wh-phrase is at the edge of the embedded clause where it originates,
and in (4b), the distance between the wh-phrase and the verb recognize is larger.
86 Hajime Ono et al.