| |
Abstract:
Psycholinguists have long been interested in how people
process garden-path sentences such as "While the man hunted the
deer ran into the woods". Virtually all theories of parsing assume
that the comprehender prefers to take "the deer" to be the object
of "hunted", and that the structure will then have to be revised
when the verb "ran" is encountered. For the present study, we
decided to address a question that has received little attention in
the literature on sentence processing: What interpretation do
people actually obtain for sentences such as these? Our interest in
this question is motivated in part by theories of reanalysis which
assume that reanalysis will not always be complete (see Fodor &
Ferreira, 1999), and by some general theories of cognition which
assume that the representations formed for perceptual events are
not always complete and accurate (e.g., the literature on change
blindness in visual cognition).
We conducted four experiments to examine the interpretations
people obtain for garden-path sentences consisting of a preposed
adjunct clause followed by a main clause, with no comma separating
the two (as well as non-garden-pathing controls). All used visual
presentation (RSVP for Experiments 1, 3, and 4, full sentence
reading time for Experiment 2). After each sentence, the
participant was required to answer a question about it and indicate
his or her confidence in the answer. In the first two experiments,
participants received sentences such as "While the man hunted the
deer ran into the woods", and their task was to answer the question
"Did the deer run into the woods?". If participants eventually
compute the globally correct structure, then the correct answer
given the content of the sentence is "no". Ease of reanalysis was
varied by making the ambiguous phrase either short ("the deer") or
long ("the deer that was brown and furry") (Ferreira &
Henderson, 1991). Although it is possible that in the example
sentence the man was hunting a deer, there is no reason why that
inference should be stronger when the deer is modified with
adjectives. The phrase-length manipulation, then, allowed us to
separate the tendency to state that the man hunted the deer because
it is a plausible theme from the tendency to have that belief based
on the syntactic misanalysis. We found that participants were
correct less than half the time (but equally confident of their
correct and incorrect answers), and they were more likely to end up
believing that the man hunted the deer when the ambiguous phrase
was long--that is, the longer they were committed to the wrong
syntactic analysis, the more likely they were to retain the
semantic interpretation corresponding to it. The second experiment
yielded the identical pattern of results with full-sentence
self-paced reading.
For the third experiment, participants received the clauses in
either the garden-pathing order (subordinate before main) or in the
order that did not induce a garden-path ("The deer ran into the
woods while the man hunted"). In addition, they were asked to
answer both questions like "Did the man hunt the deer" and
questions like "Did the deer run into the woods?" Performance for
the question "Did the deer run into the woods" was equally good in
the garden-pathing and non-garden-pathing conditions. In contrast,
when the sentence occurred in the subordinate-main order,
participants were far less accurate to answer whether the man
hunted the deer. This experiment establishes, then, that at least a
partial reanalysis took place: Comprehenders got far enough to
locate a subject for the disambiguating verb in the main clause,
but they did not follow through on the reanalysis to the point of
revising their semantic interpretation of the preposed clause.
|
