| |
Abstract:
In recent years, findings from error-elicitation studies have
indicated that there are production processes such as subject- verb
agreement (SVA) that are sensitive to manipulations of
morpho-syntactic variables but not semantic or phonological
variables (e.g., Bock & Miller, 1991, Bock & Eberhard,
1993). The research reported here extends the syntactic line of
inquiry. Here we demonstrate, using the 'preamble-repetition,
sentence completion' task, that the SVA process is sensitive to the
order of constituents within a complex subject NP.
It has been found repeatedly that SVA errors arise when the head
of the subject is modified by a clause or phrase containing a
number-mismatching noun ("the interloper"). In research reported at
earlier CUNY conferences, we found (1) more errors following a
preamble such as "The manager of the huge multinational banks" (9%)
than after a syllable-matched preamble such as "The manager of the
loan department at the banks" (3%) (Nicol & Barker, 1997) and
(2) more errors following "The statue which stood in the gardens
near the mansion" (9%) than "The statue which stood in the garden
near the mansions" (3%) (Nicol, Forster & Vigliocco, 1994).
These findings suggest that it is NOT the proximity of the
interloper to the VERB that matters, but rather the proximity to
the HEAD NP. However, there are other possible explanations. The
first finding could be explained in terms of "competition": the
potential effect of a plural nonhead is reduced if there is another
singular nonhead. The second finding could be explained by appeal
to greater complexity associated with alternating number. The
alternating-number explanation is supported by the fact that there
are also more errors following "The statues which stood in the
garden near the mansions" than "The statues which stood in the
gardens near the mansion". Typically, one does not see interference
from singular interlopers. However, the plural head cases elicited
an unusual number of other types of errors as well, including
responses in which the sentence ending had the intermediate NP as
the logical subject (e.g., "The statues which stood in the garden
near the mansion is watered regularly"). Hence, it seemed possible
that the plural head cases were generally more complex and the
excessive errors observed for these conditions was obscuring the
proximity effects. It would clarify matters if we could eliminate
at least the head-misselection errors from the set of
responses.
Results indicate that the appearance of an interloper in a
position closer to the head NP has a significantly larger effect
when the head is singular than when it is plural. Hence, even if
alternating number elicits errors due to increased complexity,
there is an ADDITIONAL effect in the singular-head condition. We
argue that this is the standard interloper effect: A plural nonhead
NP produces greater interference when it is close to the (singular)
head than when it is more distant from the head.
This result suggests that subject-verb agreement in English is
computed over a representation in which constituents are linearly
ordered. In addition, this result is compatible with two views of
agreement computation: (1) Verb specification occurs prior to the
complete assembly of the complex NP, if the NP contains multiple
phrases/clauses; interference may arise if a number- mismatching
nonhead is co-active with the head when the verb is specified. (2)
Verb specification is achieved via "feature percolation" from the
head NP to the verb; interference is greater from NPs that are
hierarchically closer to the head NP because the "syntactic path"
is shorter. Further research is needed to distinguish these.
|
