|
Abstract:
Previous reading studies provide evidence for three classic
sentence processing effects: the reduced relative effect (as in 1
vs. 2 below), the effect of animacy in reduced relatives (1 vs. 3
below), and the filled gap effect (in 5 vs. 6 below). In the
research reported here, we explore whether these effects extend to
auditory presentation.
In the reduced relative sentence in (1) below, reading studies
show a garden-path effect at the by-phrase in (1) vs. (2).
(1) The defendant examined by the lawyer turned out to be
unreliable.
(2) The defendant who was examined by the lawyer turned out to be
unreliable.
In an eye tracking study, Trueswell et al. (1994) found no
garden-path effect in (3) vs. (4) where the initial NP is inanimate
and therefore not a semantically plausible agent.
(3) The evidence examined by the lawyer turned out to be
unreliable.
(4) The evidence that was examined by the lawyer turned out to be
unreliable.
In the filled gap sentence in (5) below, reading studies
typically show a "surprise" effect (longer reading times) for the
NP following the main verb.
(5) Which woman did the stranger rob the diner on First Street
with _ today?
(6) Which woman did the stranger rob _ in the diner on First
Street today?
In the current experiment, a secondary load task, the detection
of a voice changed syllable, provides an online measure of
processing difficulty (RTs and errors for detection of the word
uttered in a different voice; in the examples above, the target
words were "lawyer" and "diner").
The reduced relative effect (a significant difference between
(1) and (2) above) was found online in reaction times and errors.
There was no significant effect of the animacy of the initial NP in
reduced relatives (comparing the difference between (1) and (2) to
the difference between (3) and (4)). There was no significant
filled gap effect (difference between (5) and (6) above). Finding
the reduced relative effect shows that the voice change detection
task is sensitive to online processing effects. This makes the
failure to find the animacy effect and the filled gap effect more
meaningful.
One interpretation of these results is that in spoken language,
reduced relatives are misanalyzed just as they are in reading.
Filler-gap sentences, on the other hand, are processed differently
than in reading; the memory burden imposed by the filler may be
greater in the visual modality than in the auditory modality.
Evidence for the filled gap effect comes from self-paced reading
experiments, which differ from the experiment presented here in the
rate and control of presentation. Crain and Fodor (1985) and Stowe
(1986) both found the filled gap effect using the word-by-word
self-paced reading task. In both studies, reported reading times
for each word were above 500 milliseconds. In our experiment,
subjects had no control over the presentation rate, and the rate
was faster--approximately 250 milliseconds per word.
The failure to find the effect of animacy in reduced relatives
could be due to the auditory modality or to the presentation rate.
Trueswell et al. presented the entire sentence to subjects at once,
and subjects were in control of their reading rate at all parts of
the sentence. In our experiment, subjects heard the sentence
serially. Subjects were not in control of the presentation rate,
which was rapid (but still natural).
Future experiments in which subjects are presented with spoken
sentences at rates approximating self-paced reading times may
reveal whether processing differences are a result of the mode
alone or the rate ofpresentation.
|
