b. kakarityoo-wa [senmu-ga dono pasokon-o
supervisor-TOPthe director-NOMwhich computer-ACC
tukatteiru-to] iimasita-ka
is.using-C said-Q
“Which computer did the supervisor say that the director is using?”In terms of expectations, TMEs are an indication that the presence of the
wh-phrase leads the parser to make a fi ne-grained prediction. More specifi cally,
when the parser encounters a non-wh NP with an accusative Case-marker, the
parser can predict there must be a transitive verb, but it cannot make any further
specifi cation of the verb: the verb may appear with the non-wh declarative
complementizer or the Q-particle ka. By contrast, if the parser encounters an
accusative wh-NP, it makes a fi ne-grained, non-thematic prediction about the
verb region; it should preferably come with ka, rather than to.
6.2 Results
Data from three participants were eliminated due to their low accuracy rates
for the comprehension questions (less than 66.6 per cent). We used reading
time data only from the trials in which the participants answered the compre-
hension question correctly. The reading-time data whose z-scores were 3 or
greater in each condition × region cell were eliminated. Because we were inter-
ested in locality effects and TMEs at the mid-verb and its spillover regions,
reading time data from the regions 8–10 were submitted to fi t linear mixed
effects models, in which the Q-position factor and the locality factor were taken
as fi xed effects, and participants and items factors as random effects. The two
fi xed effects were centered such that for the locality factor, the distance condition
was encoded as 1, and the local condition as −1; similarly, for the TME-associated
factor, the Mid-Q was encoded as 1, and the High-Q as −1. The models were
fi t using the lmer command in the lme4 package in R; the p values were com-
puted using the mixed function in the afex package. Table 6.1 lists the results
from the LMER analysis for region 8–10. Where appropriate, we compared the
conditions pairwise, the results of which are reported in the text.
In region 8 (the “mid” verb), there was a main effect of locality, showing
that the verbs in the distant conditions were read slower than those in the local
conditions. There was no main effect of the Q-position. Although there was
no interaction between the two factors, a planned pairwise comparison showed
that there was a locality effect in the High-Q conditions, indicating that the
distant × High-Q condition was read slower than the local × High-Q condition
(β = 55.05, SE = 25.91, t value = 2.12).
In region 9 (locative PP, spillover region 1), there was also a main effect of
locality as well as a main effect of the Q-position. The main effect of locality
is in fact in the opposite direction to the one in the region 8. In region 9, the
verbs in the local conditions seemed to be read slower, but it looked like this
Make a good prediction 95