| |
Abstract:
Coordinate structures present one of the classic puzzles in
the theory of human linguistic ability. Despite many intriguing
proposals, there is as yet nothing like concensus on how syntactic
principles and processing systems construct the relationship
between each individual conjunct and other constituents outside the
coordinate structure.
This paper explores an unusual approach. It questions whether
these relations are constructed and represented within the same
hierarchical framework that's applied in the absence of
coordination. Thus, in "The water samples need testing" extending
the subject NP via incorporation of a prepositional phrase or
relative clause readily establishes the hierarchical nature of the
subject/verb agreement relation. However, there are coordinate
structures where a prescriptively irrelevant NP near the agreeing
verb can confuse writers. Thus, in "Jill's ability and her desire
to help has led to a career in medicine" it appears that the second
conjunct has somehow overcome whatever control mechanism links the
coordinate structure as a whole to the agreement marking on the
verb. Temporarily neglecting possible syntactic accounts, facts of
this sort might be explained by holding that relationships between
individual conjuncts and constituents outside the coordinate phrase
are constructed within a nonhierarchical memory representation in
which linear proximity significantly affects the relationships
constructed. Such a theory might hold that though each conjunct has
internal syntactic structure, links between conjuncts and other
constituents are not syntactically constructed, that interpretive
processes are primarily responsible for establishing links between
conjuncts and other constituents. An approach of this sort has some
attractions: it correctly predicts that there will be no c-command
relations among conjuncts; it offers an account of the seemingly
extrasyntactic character of the 'coordination of likes' principle
by making interpretive processes responsible for coordinate
integration; it entails the Coordinate Structure Constraint because
it provides for none of the syntactic paraphernalia in the highest
node in the coordinate structure that would be needed to implement
extraction.
This paper will review a number of experimental results relevant
to this proposal, including recent evidence on Binding Theory
violations (p<.001) that seems to follow from this proposal. It
will also report new evidence of sensitivity to the amount or
nature of the material separating conjuncts from an agreeing verb,
especially where the presence or absence of coordination exerts a
demonstrable effect. For example, in "The water samples in/and the
material from the pond/ponds needs testing" the plurality of the
distractor "pond/ponds" is irrelevant to the agreement marking on
"needs", whether the subject NP is coordinate or not. Indeed, in
the two variants of this sentence with a non-coordinate subject NP
the alternate forms of the distractor have no discernable effect on
judged acceptability. However, the distractor does produce a
reliable effect (p=.05) when the subject NP is coordinate.
Apparently the boundary surrounding the coordinate structure is
permeable, allowing "needs" to sometimes spuriously 'agree' with
the singular "pond" in a way that does not occur with a
non-coordinate subject NP.
The paper argues that these and other effects motivate further
exploration of the possibility that linear, non-hierarchical memory
structures are involved in the analysis of coordinate
structures.
|
