| |
Abstract:
Full and reduced relative clauses are frequently used in
studies of sentence comprehension and ambiguity resolution.
However, little is known about how these constructions are used
in natural text, and one criticism of ambiguity studies that use
reduced relatives is that the stimuli are unnatural and may not
be revealing about normal sentence comprehension processes.
This study examined a sample of 378 naturally occuring
relative clauses extracted from the Brown corpus. The relative
clauses were extracted from the set of all participle uses of 40
regular verbs selected from four different ranges of participle
preference (a measure of a verb's tendency to be used in
participial rather than past tense form). Non-restrictive
relative clauses were included in the sample, but a separate
analysis in which they were excluded showed almost identical
results.
Reduced relatives, despite a long-standing reputation for
difficulty, were actually much more common than full relatives:
330 or 87% of the relatives in the sample were reduced, only 48
or 13% were full, F(1,44)=3D22.68, p < .001. The apparent
conflict between reduced relatives' presumed difficulty and their
overwhelmingly greater frequency is resolved in part by the fact
that the naturally occurring reduced relatives were likely to
occur with at least two factors that have been shown to ease
their processing difficulty. Trueswell (1996) found that reduced
relatives were easier to comprehend when they contained verbs
with a high participle preference. Trueswell, Tanenhaus and
Garnsey (1994) likewise found that reduced relatives modifying
inanimate nouns were easier to process than those modifying
animate nouns. The naturally occurring relatives in the sample
showed significant tendencies to have both of these properties
(participle preference: F(3,44)=3D3.74, p<.05; noun animacy:
F(1,44)=3D9.83, p<.01). In all, 54% of the relatives in the
sample had both of these properties, and 97% had one or the
other.
These findings have implications for ambiguity studies that
use full and reduced relatives because they indicate that the
naturalness of the stimuli may be an important issue. Unlike the
constructed stimuli that are used in some experiments, naturally
occurring reduced relatives may be easier than full relatives
(they are certainly more common) and they almost always occur
with one or more factors that can make them easy to process.
Trueswell, J. (1996). "The role of lexical frequency in
syntactic ambiguity resolution."
Journal of Memory and Language,
35, 566-585.
Trueswell, J., Tanenhaus, M., and Garnsey, S. (1994). "Semantic
influences on parsing: use of thematic role information in
syntactic ambiguity resolution."
Journal of Memory and Language,
33, 285-315.
|
