|
Abstract:
Julie Van Dyke, University of Pittsburgh,
jvandyke+@pitt.edu
A number of researchers have appealed to working memory to
explain limitations on recovery from temporarily ambiguous
sentences (Gibson, 1991; King and Just, 1991; Lewis, 1996). This
work continues that tradition with a focus on understanding the
precise nature of the limitation. Ferreira and Henderson (1991)
manipulated length and syntactic complexity of the ambiguous
region and found no complexity effect. Rather, they attributed
difficulties in recovery to the position of the head of the
ambiguous phrase relative to the disambiguating material.
In this work we directly pit syntactic complexity against
length, controlling for position of the head. We use
subject/object ambiguities as in:
(a) John knew the man was causing trouble.
and show that the easy recovery from the ambiguous object in
(a) becomes difficult with a particular intervening structure. In
particular we show an effect of what we call syntactically
interfering structures (Lewis, 1996) as in:
(b) John knew the man who thinks the university knew that he
was cold was causing trouble.
This is in contrast to syntactically non-interfering
structures as in:
(c) John knew the man he saw travelling on the train with a
suitcase was causing trouble.
We define interfering structures in the context of a parsing
model (NL-Soar) which has a limited indexing structure for
partial constituents. Thus, in sentence (b) the embedded clauses
interfere with the main verb so that when the object ambiguity is
resolved, the matrix verb is no longer available for repair. In
sentence (c) there is no interference with the matrix verb, so we
expect to find successful repair.
We are presently evaluating this hypothesis using a self-paced
moving window paradigm in which subjects are asked to give
grammaticality judgements. We expect to show independent effects
of length and interference and an interaction with
Minimal/Non-Minimal attachment sentences. This interaction
indicates that the interference effect can not be explained
simply by the syntactic complexity of the ambiguous region. We
argue that a more satisfying account comes in terms of
interfering structures in working memory blocking reanalysis.
Additionally, we suggest that the reason Ferreira and Henderson
(1991) found no effect of structure is because all of the
structures they used in their stimuli were non-interfering.
Ferreira, F. Henderson, J. M. (1991). "Recovery from Misanalysis
of Garden-Path Sentences."
Journal of Memory and Language,
30, 725-745.
Gibson, E. A. (1991). A computational theory of human
linguistic processing: Memory limitations and processing
breakdown. Doctoral dissertation, Carnegie Mellon.
King, J. Just, M. A. (1991). "Individual Differences in
Syntactic Processing: The Role of Working Memory."
Journal of Memory and Language,
30, 580-602.
Lewis, R. L. (1996). "Interference in Short-Term Memory: The
Magical Number Two (or Three) in Sentence Processing."
Journal of Psycholinguistic Research,
25:1, 93-115.
|
