| |
Abstract:
Abstract: Accounts of sentence processing frequently appeal
to retrieval from working memory. We will discuss three
whole-sentence anomaly judgment experiments suggesting that the
sort of retrieval taking place is a major determinant of processing
load. In experiment 1, we found that relative-clause complexity
(subject-relative vs. object-relative clauses), propositional
complexity (one-action vs. two-action sentences), and articulatory
suppression were independent. Experiment 2 replicated the
independent effects of the subject-object relative complexity and
articulatory suppression, but demonstrated that articulatory
suppression interacted with complexity of embedding (right vs.
center). On the basis of those results, we would like to extend the
model proposed by Caplan & Waters (1999, BBS). Our extension
focuses on the type of retrieval needed at a particular point in
the sentence rather than on the linguistic levels involved in
processing. The present data suggest three types of retrieval. 1)
With object-relative clauses, readers have to retrieve and revise
an already built up structure. 2) In cases of propositional
complexity, propositional content has to be retrieved. 3) And
finally, in a center-embedded sentence, readers have to retrieve
the main-clause subject after they have processed the embedding.
The interaction with articulatory suppression suggests that this
subject is retrieved using an articulatory format. Currently, we
are conducting an experiment which orthogonally varies complexity
of embedding, adjectival load (one vs. three consecutive
adjectives), and articulatory suppression to see whether adjectival
load also requires articulatory retrieval.
|
