| |
Abstract:
In recent years, it has been demonstrated that lexical
information is rapidly integrated with contextual information in
order to predict and narrow down the domain of reference. More
specifically, and based on the experimental technique developed by
Tanenhaus et al. (1995), several eye-movements studies have shown
that people fixate on a certain object in a visual scene as soon as
they encounter auditory lexical items which distinguish that object
from other potential referents in the visual scene (such
distinguishing information includes adjectives in NPs (Sedivy et
al., 1999), selectional restrictions of verbs in VPs (Altmann &
Kamide, 1999), and the semantic relationship between a preposition
and a subsequent NP in a PP (Chambers et al., 1998 )). For
instance, Altmann & Kamide (1999) demonstrated that
eye-movements were driven to a picture of a cake more frequently
after subjects heard "The boy will eat _(II_(J" than after they
heard "The boy will move_(II_(J", when the cake was the only edible
object amongst several moveable objects in the same scene (a ball,
a toy car, a toy train set, as well as the cake). However, it is
crucial to note that the majority of those studies looked at the
effect of heads in a head-initial language (English _(IQ_(J e.g., V
in a VP; Altmann & Kamide, 1999; P in a PP; Chambers et al.,
1998). An important question to be addressed here is whether
similar patterns of predictive eye-movements will be found during
the processing of pre-head constituents in head-final languages.
Strict variants of head-driven parsing (e.g., Pritchett, 1991,
1992) argue that structural analyses of the sentence occur only
after the head has been encountered. Within this framework,
processing of head-final structures should result in no predictive
eye-movements of the kind revealed in the previous English
studies.
We are currently conducting an eye-tracking experiment in which
native speakers of Japanese are presented with pairs of a visual
picture and an auditory sentence. There are two sentences (e.g., 1a
& 1b) for each picture (e.g., 2):
(1a) Josee-ga fooku-de yukkurito supagettii-o taberu.
woman-nom fork-with slowly spaghetti-acc eat.
"The woman will eat the spaghetti with the fork slowly."
(1b) Josee-ga fooku-no-yoko-ni yukkurito naifu-o idoosuru.
woman-nom fork-next to slowly knife-acc move.
"The woman will move the knife next to the fork slowly."
(2) A visual scene consisting of "a woman", "a fork", "a plate
of spaghetti", "a knife" and "a napkin"
The pre-head-driven parsing account (e.g., Inoue, 1991; Bader
& Lasser, 1994; Koh, 1997; Kim, 1999; Kamide & Mitchell,
1999) predicts that people will fixate on the spaghetti after
hearing "fork-with" even before hearing "spaghetti-acc" in (1a),
since, in the visual scene, the spaghetti is the most plausible
object for an action involving a fork as instrument. However, such
eye-movements (to the spaghetti before "spaghetti" is heard) could
be due to a simple lexical association between "fork" and
"spaghetti". To rule this out, (1b) serves as the control condition
for (1a). Anticipatory eye-movements to the spaghetti will be
interpreted as support for the pre-head-driven parsing account if
hearing "fork-next to" does not guide eye-movements to "spaghetti"
in (1b) as frequently as "fork-with" does in (1a). In our
presentation, we shall discuss the results of the experiment in the
light of (pre-) head-driven parsing theories as well as the recent
empirical studies in the same experimental paradigm using a
typologically different language.
Altmann, G.T.M., & Kamide, Y. (1999). Incremental
interpretation at verbs: Restricting the domain of subsequent
reference. Cognition, 73(3), 247_(IP_(J264.
|
