|
Abstract:
Agreement features participate in several grammatical
dependencies. This poster will explore how features are represented
by the processor during the comprehension of two of these:
filler-gap and subject-verb agreement dependencies. Evidence will
be presented for a structure-dependent passing of features from a
filler to the position of its gap, giving rise to difficulty in
checking an intervening subject-verb agreement relation.
Kimball and Aissen (1971) report data from a dialect of English
in which a local subject-verb mismatch as in (1a), 'manager think',
is acceptable if it is on the extraction path between a plural
relative pronoun and its trace. The mere presence of a c-commanding
NP does not suffice, as illustrated by the ill-formedness of
(1c).
1a. (*)Lucine dislikes the people
i
who
i
[the manager think] t
i
know the answers.
1b. Lucine dislikes the people
i
who
i
[the managers think] t
i
know the answers.
1c. *Lucine dislikes the people
i
who
i
t
i
think [the manager know] the answers.
1d. Lucine dislikes the people
i
who
i
t
i
think [the managers know] the answers.
For speakers of most dialects of English, (1a) is ungrammatical.
Nevertheless, intuitions suggest (1a) is more acceptable than (1c)
even for these speakers. We hypothesized that this perceived
difference in acceptability has its basis in processing: in (1a),
the plural feature of the NP which is head of the relative clause
('people') is passed through the tree from the relative pronoun
('who') to the trace it binds. When the intervening agreement
relation is checked, it may appear that this 'passing' plural
feature is available in the local tree structure to license the
plural feature of 'think'.
This hypothesis is supported by two word-by-word incremental
acceptability judgment studies. For the paradigm in (1), rejection
rates showed the predicted interaction between position of the gap
and grammaticality: there was less of a difference between (1a) and
(1b) than between (1c) and (1d). A second study tested the late gap
(1a,b) sentences only, manipulating the number of the filler. The
difference in rejection rates for the ungrammatical vs. grammatical
sentences was smaller when the plural feature was available from
the filler to license the plural verb (as in 1a,b) than when it was
not (i.e., when the filler was singular ('person')).
Unlike most existing studies of agreement features in
processing, the current results cannot be modeled in terms of local
effects on a verb from an adjacent feature-marked noun in the
subject NP (though see Nicol, Forster and Veres (1997) where
features move within the subject NP). Instead, these results
provide preliminary evidence that features are passed through the
phrase marker.
Kimball, J. Aissen, J. (1971) "I think, you think, he think."
Linguistic Inquiry,
2, 241-246.
Nicol, J. L., K.I. Forster, and C. Veres. (1997) "Subject-verb
agreement processes in comprehension."
Journal of Memory and Language,
36, 569-587.
|
