|
Abstract:
Working memory capacity is the ability to store and retrieve
information while performing other mental computations. In the
Listening Span and Reading Span tests (Daneman & Carpenter,
1980), subjects hear or read sets of sentences, then recall their
final words. W.m. span is the largest set size with accurate
recall.
These measures correlate reliably with general language ability
(Daneman & Merikle, 1996), but their ability to explain
individual differences in parsing is a topic of heated debate
(Waters & Caplan, 1996; Just, Carpenter & Keller, 1996).
For example, to understand 'the dog [that the cat chased _ ]', one
has to retrieve 'the dog' after processing 'that the cat chased',
and Capacity Theory (Just & Carpenter, 1992) states this is
constrained by w.m. Waters & Caplan, however, argue that the
w.m. tests are inappropriate because they involve conscious
processing and, unlike gap-filling, there is no relationship
between the stored words and subsequent interpretation. They
dispute the findings which support Capacity Theory, and argue that
psycholinguistic processing involves a separate working memory
system, while Just and Carpenter reject that view.We argue that the
w.m. tests may be correlated with automatic parsing simply because
both tap general language processing abilities (Daneman &
Merikle, 1996). In Experiment 1, 90 children (4;7 to 10;11)
completed a Listening Span task and a gap filling task: Smyth &
Flowers (1998) report that filled gap (2) and nested dependency (3)
sentences are harder to process than sentences like 1:
1. Which can is easy to put _ in this jar?
2. Which can is easy to put this jar in _ ?
3. Which can is this jar easy to put _ in _ ?
Gap filling scores were correlated with Listening Span, but the
correlation with Age was even higher. After Age was partialled out
in a hierarchical regression, w.m.scores did not account for any
further variance.
In Experiment 2, 80 adults completed Reading Span, the
Nelson-Denny Reading Test, and two parsing tasks. In the first
(Smyth 1998), subjects judged sentences like 4 and 5:
4. I don't know which jar the can would be easy to put these
coins in _ .
5. I don't know which jar these coins would be easy to put _ in
the can.
Both have two fillers but only one gap; 4 is harder to reject
because a filled gap effect disrupts processing before any gaps
have been discharged. In the other task (McElree 1999), subjects
judged the acceptability of sentences with 0, 1, or 2 clauses
between the gap and the filler:
6. This is the track that the runners [...] admired _ / amused _
.
The greater the distance, the more likely it is that subjects
fail to discriminate between acceptable and unacceptable
sentences.
We will report the results of a hierarchical regression with
Listening Span and the Nelson Denny scores on each of these parsing
tasks. If Listening Span does predict parsing ability, then it
should explain the variance in the parsing tasks even after reading
scores are partialled out. Results will be discussed in terms of a
distinction between lexical and syntactic interpretations of the
Active Filler hypothesis.
|
