|
Abstract:
On the basis of two whole-sentence anomaly judgment
experiments, we propose a two-stage storage model along the lines
of Berwick & Weinberg (1984) and Caplan & Waters (1999).
The first stage is involved in storing incomplete sentence
material, whereas the second stage involves propositional storage.
We will make some specific proposals about how these stores are
used during the processing of various syntactic structures and
about the different costs relating to the use of the two stores. In
Experiment 1, embedding (right-embedded relative clauses vs.
center-embedded relative clauses; cf. 1 and 2) and articulatory
suppression (no counting vs. counting out loud from one to six)
were orthogonally varied.
1. The man kisses the child who sees the girls. (literal
translation)
2. The man who sees the girls kisses the child. (literal
translation)
This experiment showed an interaction of embedding by
articulatory suppression. In the no-counting condition, there was
no effect of embedding. However, in the counting condition, RTs
were longer for center-embedded clauses than for right-embedded
clauses.
Experiment 2 also orthogonally varied embedding and counting,
but using adverbial clauses (cf. 3 and 4) instead of relative
clauses.
3. The journalist reported the crime, although the dangerous
criminal threatened him. (literal translation)
4. The journalist reported, although the dangerous criminal
threatened him, the crime. (literal translation)
Here we also found an interaction between embedding and
counting. However, this interaction was entirely opposite to that
found in Experiment 1. Effects of embedding were absent in the
counting condition, whereas in the no-counting condition, RTs were
slower for right-embedded clauses than for center-embedded
clauses.
These experiments show that adverbial and relative clauses are
processed differently when center-embedded. In the counting
condition of Experiment 1, center embeddings yield longer RTs than
right embeddings. In the center-embedded sentences, the main-clause
subject can only be integrated with the rest of that clause once
the embedding has been processed and must be kept active until this
occurs. The fact that this pattern only emerges in the counting
condition suggests that storage of the main-clause subject depends
on articulatory rehearsal.
The finding that in the no-counting condition of Experiment 2
right-embedded adverbial clauses are read more slowly than
center-embedded ones suggests that a different store is used for
these structures: readers transfer completed propositions to a
propositional store. Center-embedded structures, which do not allow
transfer, are read faster than right-embedded structures; this
seems to indicate that the cost of keeping information active in
working memory is less than the cost of transfer to and retrieval
from the propositional store. The lack of such an effect in the
counting condition indicates that counting out loud prevents
readers from transferring information to the propositional
store.
In summary, we would like to propose the following:
A. Articulatory suppression impedes articulatory rehearsal of
unintegrated verbal material;
B. Articulatory suppression prevents readers from transferring
propositions to the second-stage store; and
C. Keeping propositional content active in the first-stage store
is less costly than transfer to and retrieval from the second-stage
store.
Berwick, R., & Weinberg, A. (1984). The Grammatical Basis of
Linguistic
Performance: Language Use and Acquisition. Cambridge, MA.: MIT
Press.
Caplan, D., & Waters, G. (1999). Verbal working memory and
sentencecomprehension. Brain and Behavioral Science, 22(1),
77-126
|
