| |
Abstract:
Theories of reanalysis that make predictions about processing
difficulty on dispreferred structures must take into account two
factors: (1) the cost associated with recovering dispreferred
material (repair costs, diagnosis costs, re-activation costs,
suppression of preferred interpretation, etc.) and (2) the cost
inherent in making the correct attachment, independent of the
effects of ambiguity. Ferreira & Henderson (1991; 1998) account
for reanalysis difficulty in terms of the decay of the lexical
material required for reanalysis, predicting that longer distances
make reanalysis more difficult. The current paper suggests that
this theory offers an account of reanalysis which is incomplete in
that it does not clearly distinguish what affects the cost of
recovering the dispreferred interpretation vs. what affects the
basic attachment processes. Ferriera & Henderson apparently
intend the decay theory to cover both factors, but we will show
that their experimental materials did not include the relevant
unambiguous baselines to test this.
The cost of the correct attachment can be estimated by comparing
unambiguous non-minimal attachment sentences. In order to
distinguish pure decay (distance) from effects of interfering
structure, we have used sentences which differ with respect to the
noun phrase that serves as the subject of the embedded sentential
complement. The distance effect is the comparison of short and
long, non-interfering sentences, while the interference effect is
the comparison of non-interfering and interfering sentences.
Examples are given below:
Short: The hotdog vendor noticed that the boy was crying on the
curb.
Long, Non-interfering: The hotdog vendor noticed that the boy
who was annoying the other customers was crying on the curb.
Long, Interfering: The hotdog vendor noticed that the boy who
thought the fries were burnt was crying on the curb.
Counter to Ferreira & Henderson's predictions,
grammaticality judgments show a large effect of structural
interference (p < .001) and no effect of distance.
The cost of recovery can be estimated by comparing these
unambiguous sentences with their ambiguous counterparts (i.e.
without the post-verbal 'that'). Here, we find a significant effect
of distance (p < .05) but not of interference on the size of the
garden path effect.
These data are consistent with a model where decay/distance--
and not interference--play a role in the *recovery* of the
dispreferred structure. Conversely, they suggest that
interference--and not decay--is the most important factor in making
the attachment itself. The Ferreira & Henderson account can be
made consistent with this data by assuming that decay affects
recovery only, not attachment. We discuss the implications of this
interesting cross-over interaction for other models of garden path
recovery, including both serial (e.g., Fodor & Inoue, 1998;
Lewis, 1998) and parallel (e.g., Gibson, 1991; MacDonald,
Pearlmutter, & Seidenberg, 1994) models.
Ferreira, F., & Henderson, J. M. (1991). Recovery from
misanalysis of garden- path sentences. Journal of Memory and
Language, 30, 725-745.
Ferreira, F., & Henderson, J. M. (1998). Syntactic reanalysis,
thematic processing, and sentence comprehension. In J. D. Fodor
& F. Ferreira (Eds.), Reanalysis in Sentence Processing .
Boston: Kluwer Academic.
Fodor, J. D., & Inoue, A. (1998). Attach Anyway. In J. D.
Fodor & F. Ferreira (Eds.), Reanalysis in Sentence Processing .
Boston: Kluwer Academic.
Gibson, E. A. (1991). A computational theory of human linguistic
processing: memory limitations and processing breakdown.
Unpublished Ph.D Dissertation, Carnegie Mellon University,
Pittsburgh.
Lewis, R. L. (1998). Leaping off the garden path: Reanalysis and
limited repair parsing. In J. D. Fodor & F. Ferreira (Eds.),
Reanalysis in Sentence Processing . Boston: Kluwer Academic.
|
