| |
Abstract:
Sentence and discourse processing both depend on the
integration of information from different and often non-adjacent
constituents and thus may seem very similar in their working memory
demands. For example, in English, number information is crucial for
both the processing of intra-sentential subject-verb agreement and
inter-sentential pronoun-antecedent agreement. In both cases, the
dependent elements (the subject and the verb, or the antecedent and
the pronoun) could be separated by intervening text. Despite this
similarity between sentence and discourse processing, theorists
disagree on whether the two kinds of processing draw on a shared
pool of resources or on separate resource pools (e.g., Caplan &
Waters, 1999; Just, Carpenter, & Keller, 1996). This
disagreement lies at the heart of current debates concerning the
autonomy of sentence processing.
The research reported here attempted to resolve this
disagreement by comparing the effect of intervening text on the
on-line processing of sentences and discourse. Participants were
recruited from two populations: healthy normal adults and
Alzheimer's patients. The performance of the Alzheimers patients
was of special interest because of their well-known working memory
impairment. We reasoned that if sentence and discourse processing
draw on shared resources then the processing of long distance
dependencies in sentences and in discourse should be similarly
affected in the patients.
We conducted two cross-modal naming experiments. Experiment 1
tested the processing of subject-verb number agreement within
sentences, and Experiment 2 investigated the processing of
pronoun-antecedent number agreement across sentences. Both
experiments included a length manipulation such that processing was
tested with varying amounts of intervening material between the
agreeing constituents (see the sample items at the end of this
abstract.) Both experiments tested patients with Alzheimer's
disease and age-matched normal participants. Verbal working memory
was assessed for all participants.
The results of these two experiments revealed a 3-way
dissociation between processing number agreement in sentences and
in discourse:
1. Intervening material only hindered agreement processing in
sentences (Experiment 1) but not in discourse (Experiment 2).
2. The Alzheimer's patients were only impaired in discourse
processing (Experiment 2) but not in sentence processing (Exper.
1).
3. Working memory performance correlated with discourse processing
performance (Experiment 2) but not with sentence processing
performance
(Experiment 1). These results are not compatible with a simple
shared resource model in which both sentence and discourse
processing draw on the same working memory resources in a similar
fashion (e.g., Just, Carpenter, & Keller, 1996). While these
results could be interpreted as indicating separate resources for
sentence and discourse processing, they could also be explained on
the basis of differences in the frequency and predictability of the
agreeing elements: Almost all sentences in English include a verb
that agrees in number with the grammatical subject, but fewer
sentences include a pronoun. We discuss our findings in terms of a
new theory of verbal working memory that assigns special importance
to the predictability and frequency of both sentence and discourse
dependencies.
References:
Caplan, D., & Waters, G. (1999). Verbal working memory and
sentence comprehension. Behavioral and Brain Sciences, 22(1),
77-126.
Just, M. A., Carpenter, P. A., & Keller, T. A. (1996). The
capacity theory of comprehension: new frontiers of evidence and
arguments. Psychological Review, 103(4), 773-780.
|
