|
Abstract:
Results will be presented of two eye-movement studies
investigating the Repeated Name Penalty (RNP) of Centering Theory
(CT) in an intra-sentential context.
Findings in the framework of CT (Gordon et al., 1993, Gordon and
Chan, 1995 ) have shown that the syntactic prominence and the form
of referring expressions are important variables in local discourse
coherence; for example, entities introduced into a discourse by
expressions in syntactically prominent positions, in particular the
subject position, resist subsequent reference by full referring
expressions (i.e. a name). A processing disadvantage, termed the
RNP, has been observed when a name rather than a pronoun is used in
subsequent discourse to refer to such prominent entities.
While psycholinguistic work in CT has focused mainly on
inter-sentential anaphora (but see Gordon and Hendrick, 1997), the
present studies seek to test the predictions of CT in an
intra-sentential context. The experimental items used in the first
study consist of a preposed adjunct clause introducing 3 entities
(subject, direct object and indirect object), followed by a main
clause, such as
(1) When Sue showed off the kids to Ron last night, she/Sue
looked radiant.
The referents in the subject and indirect object position are
introduced by a name and the direct object by a definite NP. There
are two experimental manipulations : the subject of the main clause
(a) unambiguously co-refers with one of the three entities
introduced in the adjunct (SUBJ, OBJ, and INDIRECT OBJ. conditions)
and (b) is expressed by a pronoun or by the same referring
expression appearing in the adjunct (NAME and PRONOUN
conditions).
Our predictions are based on two assumptions : (a) following
experimental findings by Gordon and Hendrick (1997) we assume that
similar principles govern intra-sentential and inter-sentential
co-reference and (b) we adopt Keenan and Comries (1977)
NP-Accessibility Hierarchy, namely Subject>Object>Indirect
Object. We therefore predict the size of the RNP to be positively
correlated with the accessibility /prominence of the antecedent,
being greatest with a subject antecedent and decreasing
progressively as we go down the hierarchy.
The results of the first study did not confirm the predictions:
we observed a significant RNP (name > pronoun) for both subj and
indirect object antecedents, but no interaction. Most surprisingly,
when the referent in object position was the antecedent, the
opposite effect was observed, i.e. a significant repeated pronoun
penalty (pronoun > name) occurred.
This result suggests that the subject position is as prominent
as the indirect object position, as far as our experimental items
are concerned. It also shows that the object position is less
prominent than the indirect object position, contrary to our
predictions. However this latter result may be due to the fact that
different types of NPs (name vs definite description) were used to
introduce referents in these two positions. Findings by Garrod and
Sanford (90) suggest that entities introduced by name are more
prominent than ones introduced by a definite description. A second
study was carried out to test whether the name vs definite
description contrast was responsible for the result of expt 1. The
experimental items of the first experiment were modified so that
the definite description now appeared in the indirect object
position and the name in the object position, as in
(2) When Joe showed little Amy to the guys, she/Amy started
screaming.
Only the object and indirect object position were compared. The
only significant result was a marginal RNP for the object position
at first pass. Because we observed neither a RPP nor a RNP for the
indirect object position, we conclude that both position and NP
type play a role in determining a referents prominence status.
|
