| |
Abstract:
Results from sentence-picture matching tests have indicated
that individuals with Parkinson's disease suffer from sentence
comprehension deficits (e.g., Lieberman et al., 1992; Natsopoulos
et al., 1991). However, the validity of sentence-picture matching
performance as an index of linguistic ability has sometimes been
questioned. The task makes demands on visual cognition as well as
language, and it yields measures - response times and error
percentages - that, it has been argued, may not reflect on-line
processing.
Eye-tracking technology allows us to obtain more on-line data
about sentence-picture matching. In a pilot study, we tracked the
gaze of normal and parkinsonian subjects during a sentence-picture
matching task as they listened to sentences in which semantic
constraint, voice, and grammatical structure were varied. Results
indicate that normal subjects integrated visual and linguistic
information on line. They both launched eye movements toward
pictures that could be anticipated to relate to upcoming linguistic
information, and rejected distractor pictures from consideration as
the linguistic input ruled them out. Moreover, normal subjects were
able to exploit linguistic information even before a disambiguating
word could have been completely accessed. This latter finding is
consistent with past studies in which subjects have been shown to
launch saccades based on partial lexical information
(e.g.,Tanenhaus et al., 1995). Some subjects also evinced a
"checking" stage after the sentence was finished, returning to
pictures that their earlier eye movements suggested they had
eliminated.
Parkinsonian subjects, as in past studies, made more errors and
gave slower responses than normal subjects (even with slowed motor
responses factored out) -particularly for semantically
unconstrained, passive, and grammatically more complex sentences.
On simpler, more semantically constrained sentences, they appeared
able, like normal subjects, to make eye movements that paralleled
the linguistic input. Their eyes failed to keep pace, however, with
more complex, less constrained sentences. As a result, they needed
more time after a sentence ended to determine the matching picture,
during which time they had to maintain the sentence in working
memory. Even when their eye movements kept up with the input,
parkinsonian subjects generally showed longer checking periods,
indicating greater uncertainty overall about their conclusions.
These effects were seen even when subjects reached the right
answer.
These findings are in accord with sentence processing models in
which multiple constraints are used in parallel. Moreover, they are
consistent with the notion that less typical structures or less
constrained sentences place greater demands on cognitive resources.
They also are consistent with past studies showing slowed thinking
in parkinsonian subjects.
Our results are not well accounted for by models in which
language processing occurs in isolation from nonlinguistic context.
They are more consistent with a nonmodular view in which language
is built upon circuits running between cortex and the subcortical
structures affected by parkinsonism - circuits that also subserve
motor sequencing and nonlinguistic cognitive operations (Lieberman,
in press). In particular, by affecting circuits running through
frontal regions involved in verbal working memory, parkinsonism may
lead to the resource limitations that appear to underlie the
observed problems with sentence-picture matching.
|
