| |
Abstract:
Many experimental studies have documented the rapid effects
of discourse/pragmatic referential constraints on on-line language
processing (e.g. Altmann & Steedman, 1988; Sedivy, Tanenhaus,
Chambers & Carlson, 1999). However, in contrast to adults'
efficient use of referential information, there is reason to expect
that children would not show the same sensitivity. Children often
produce ambiguous references, and frequently fail to establish
antecedents of pronouns and definite noun phrases, suggesting that
they either lack the pragmatic principles governing felicitous
reference, or cannot adapt their speech for effective
communication. In addition, Trueswell et al. (in press) found that
5 year-olds demonstrate a lack of sensitivity to referential
factors in resolving PP attachment ambiguities.
Two aspects of 5-6 year-old children's referential communication
were explored in this study: 1) their ability to identify a unique
referent for a definite noun phrase in production and processing
and 2) their ability to establish reference with respect to
mutually established common ground.
An elicited production task tested whether children would use
modifiers (adjectival, prepositional phrases, or relative clauses)
to uniquely identify a target referent when two referents
corresponding to the the target's head noun were present in a
visual array. The results showed that children reliably used
modification more often when referential disambiguation was
necessary, suggesting that they clearly understand the pragmatic
principles requiring definite noun phrases to refer to a unique
entity in discourse. There were also interesting differences in the
type of distinguishing property: Scalar adjectives were used
primarily in situations where referential disambiguation was
necessary, while color adjectives were used freely, regardless of
whether the target referent needed modification. Spatial relations
typically encoded by locative phrases (e.g. the orange on the
plate) were least frequently produced, even in situations that
required modification for disambiguation. This suggests inherent
difficulties in the use of locative phrases as modifiers, a factor
relevant for explaining the lack of referential effects in the
Trueswell et al. study, which used PPs as modifiers in all of the
stimuli.
A second pragmatic constraint on referential communication
requires the coordination of interlocutors' discourse
representations in order to determine what information is shared in
common ground. Results from processing studies suggest that even
adults show evidence of difficulty integrating common ground
information, prompting hypotheses of a two-stage model in which
common ground information is used late (Keysar, Barr & Balin,
1998). An elicited production task investigated children's reliance
on common ground information. Subjects viewed a vertical display
containing four objects, one of which was hidden from an
experimental confederate's view. Three conditions were compared:
one in which a single noun was sufficient to identify the target
object (e.g. the cup); one in which a scalar property distinguished
between the target and some other competitor object visible to both
participants (e.g. the big cup), and one in which the competitor
object was visible to the child, but not to the confederate, making
modification redundant from the confederate's perspective. Children
used modification significantly more frequently in the second
condition, when both the target and competitor object were visible
to both parties, than in either of the other two conditions,
indicating use of common ground in their production. An on-line
comprehension task using eye movement monitoring showed
particularly striking use of common ground information. Eye
movement data showed no evidence of interference of the competitor
object when it was hidden from the confederate's view, even from
the very earliest moments of processing a target description.
|
