| |
Abstract:
A primary goal of language comprehension theories is to
explain why some sentences are harder to understand than others.
Toward that goal, experimental evidence has been collected about
which sentences are hard and which are easy. The usual approach is
to use sentences containing temporary ambiguities, which either
meet or violate comprehenders' expectations. Sentences like the
now-classic example below (Mitchell & Holmes, 1985)
1. The historian read the manuscript had been destroyed.
are difficult, presumably because they violate an expectation
that a noun following a verb is its direct object. In an attempt to
distinguish among competing comprehension models, several studies
have examined factors that ease the difficulty of such sentences,
with much of the attention focused on properties of the
ambiguity-introducing verb ("read" in the example; e.g., Ferreira
& Henderson, 1990; Trueswell, Tanenhaus, & Kello, 1993;
Garnsey, Pearlmutter, Myers, & Lotocky, 1997). Most studies
have found that comprehenders' expectations about the continuation
of a sentence is influenced by how often its particular verb is
used with a particular construction (called "verb bias").
Specifically, readers have difficulty when a "DO-bias" verb is
followed by an embedded clause, but little or no difficulty when a
"clause-bias" verb is followed by a clause. Thus, the difficult
embedded clause becomes easy when it is consistent with the verb
bias. This result has been taken as one source of support for
constraint-satisfaction models of comprehension over modular
two-stage models.
However, Frazier (1995) has argued that demonstrating that hard
sentences can be made easy isn't the crucial test for arbitrating
among models, since it can be explained as ease of recovery from a
misanalysis. More telling would be the opposite scenario, in which
a simple DO-continuation is made hard. Such an effect is predicted
by constraint-based theories, but not by modular two-stage models
since such sentences should not lead to any misanalysis in the
first place.
In our study, subjects read sentences like those below, with
verb bias crossed with sentence continuation type, in a self-paced
moving-window paradigm.
Consistent with previous results, the disambiguating region
(indicated above with caps) in (2a) was read more slowly than in
(3a). More importantly, the reverse was true for the direct object
versions: the disambiguating region in (3b) was read more slowly
than in (2b), and the size of the verb bias effect was equivalent
to that in (a) versions. Thus, verb bias had an equally strong
influence in simple and complex sentences, which is consistent with
constraint-satisfaction models but not with modular two-stage
models.
References
Ferreira, F., & Henderson, J. M. (1990). Use of verb
information in syntactic parsing: Evidence from eye movements and
word-by-word self-paced reading.
Journal of Experimental Psychology: Learning, Memory, and
Cognition,
16, 555-568.
Frazier, L. (1995). Constraint satisfaction as a theory of
sentence processing.
Journal of Psycholinguistic Research,
24(6), 437-468.
Garnsey, S. M., Pearlmutter, N. J., Myers, E., & Lotocky, M.
A. (1997). The contributions of verb bias and plausibility to the
comprehension of temporarily ambiguous sentences.
Journal of Memory and Language,
37(1), 58-93.
Mitchell, D. C., & Holmes, V. M. (1985). The role of specific
information about the verb in parsing sentences with local
structural ambiguity.
Journal of Memory and Language,
24, 542-559.
Trueswell, J. C., Tanenhaus, M. K., & Kello, C. (1993).
Verb-specific constraints in sentence processing: Separating
effects of lexical preference from garden-paths.
Journal of Experimental Psychology: Learning, Memory, and
Cognition,
19, 528-553.
|
