|
Abstract:
This study examines children's ability to use referential
context when making syntactic choices in production and
comprehension. Recently, it has been shown that 4-5 year-olds
sometimes fail to take into account referential information when
resolving PP attachment ambiguities, in both their initial parse
and their final interpretation of the ambiguous phrase, but perform
competently for unambiguous variants of such phrases (Trueswell et
al. 1999). When adults interpret sentences such as (1), their
initial parse is highly dependent on the referential context: They
assume that "on the napkin" refers to a destination in one-referent
(i.e. one frog) contexts and that it refers to a modifier in
two-referent (i.e. two frog) contexts.
1. Put the frog on the napkin into the cup.
2. Trueswell et al.(1999) found that children misparse these
sentences even in the two-referent context. Specifically, children
erred in their actions in a manner consistent with a destination
interpretation, i.e. they were at chance when selecting between the
two frogs and frequently moved one frog to an empty napkin (a
"false destination").
We examined why children fail to consider the referential
principle (Crain & Steedman 1985) in this situation. One
possibility is that children of this age lack the skill to evaluate
contrastive information in the scene. If so, we would expect them
to also to fail to consider the referential principle in
production. Another possibility is that the presence of two
potential referents (2 frogs) was not salient to the children in
the original Trueswell et al. study. In this case, we would expect
childen's performance to improve if the presence of the two
potential referents is emphasized.
We designed a study to test these two hypotheses. Children
(N=16, mean age 4yr 8mo) were told a story involving two animals
(e.g. two frogs) each doing different actions. At the end of the
story, each frog rests on a platform (e.g. a napkin and a towel).
The child is then asked either a Contrastive question (CQ) about
one of the frogs (e.g. "Which frog baked cookies?") or a
Noncontrastive question (NQ) ("What did the animals do at Mrs.
Squid's house?"). In order to answer the CQ correctly, the child
must produce a modifying PP. After answering the question, the
child is told to "touch the happy face" and then given a command
like (1). On 72% of the trials, children spontaneously
distinguished between the referents by using a modifying PP in the
CQ condition, e.g. replying "the one on the napkin." However, these
same children continued to misparse the "put" sentences, showing
correct actions on only 22% of the CQ trials. Similar insensitivity
was present on NQ trials (19% correct). As in Trueswell et al.,
errors were consistent with a destination interpretation: Children
were at chance when selecting between frogs and typically moved a
frog to the false destination.
The results thus reveal a strong asymmetry between production
and comprehension. This suggests that children may be able to
better integrate linguistic and contextual information when
speaking than when listening, and that comprehension mistakes do
not derive from lack of grammatical competence or pragmatic
knowledge. We suggest that the different demands of production and
comprehension drive such a developmental pattern: If comprehension
truly is an automatized guessing game, one might expect the child
comprehender to depend on (highly reliable) syntactic and lexical
cues to resolve ambiguity, even when relevant referential cues were
used immediately before for production.
|
