|
Abstract:
The comprehension of definite reference requires establishing
a relevant domain within which a referent can be uniquely
identified. However, relatively little is known about the
constraints relevant for establishing domains and how they are
used in real-time comprehension. One problem is that domains take
into account specific contextual information that is difficult to
manipulate using standard methodologies.
We used a head-mounted eye-tracker to record participants'
eye-movements as they heard instructions to manipulate objects in
a display. Previous work using instructions such as "Pick up the
red block" has established that listeners assign interpretations
incrementally, selecting the referent as soon as it becomes
unique given the potential referents in the display. However,
these results are consistent with at least three different
construals of the domain in which the referent is understood to
be unique, namely the set of entities within the denotation of
either: (a) the noun phrase; (b) the NP constrained by the
semantics of the clause in which it is contained; (c) the NP
constrained by both semantics and nonlinguistic factors such as
possible actions.
In Experiment 1, participants were given instructions such as
"Pick up the cube. Now put it inside/below the can." When several
different containers were present, identification of the referent
was delayed until after the noun (e.g. 'can') for both 'beside'
and 'inside' instructions. However, identification occurred after
the definite article when 'inside' instructions were used with
single container displays. Thus, the semantic relation evoked by
the preposition 'inside' immediately restricted the referential
domain.
Experiment 2 introduced a nonlinguistic constraint. Displays
contained two instances of the described container (e.g. two
cans) of two different sizes. The moved object's size was varied
such that it fit in either both or only one of the cans.
Instructions used either a definite or indefinite article (e.g.
"put the cube inside the/a can"). Eye-movement latencies and
proportions of looks to other objects in the definite condition
indicated the size of the moved object restricted the domain of
reference. Moreover, latencies were comparable to those in
control trials with only one can present. Crucially, the opposite
pattern occurred with indefinite articles (longer latencies
occurred when the object fit into only one can) ruling out an
explanation based on a simple problem-solving strategy.
In sum, domains of interpretation for referential expressions
are constructed and updated dynamically as an utterance unfolds
in time, taking into account both linguistic and nonlinguistic
factors.
|
