Christopher Kennedy
7
b.
Smith stepped up to the podium. His critics stepped into their offices.
c.
Jones seems unbeatable, while her opponent seems beaten.
(19)
RESEMBLANCE
relations (incoherent)
a.
?? Smith likes to play golf. Jones ate lunch.
b.
?? Smith stepped up to the podium. Jones bought a new Mercedes.
c.
?? Jones seems unbeatable. The Senate is being painted.
Kehler suggests that an elided expression that is contained in a sentence that
is part of a
CAUSE
-
EFFECT
relation does not require syntactic representation, but
an elided expression that is part of a
RESEMBLANCE
relation must have syntactic
representation. The reasoning underlying this proposal is that
RESEMBLANCE
re-
lations are, to a large extent, at least, identified on the basis of syntactic structure,
while
CAUSE
-
EFFECT
relations care only about propositional content (see Kehler
2000, pp. 540-543). The prediction of this analysis, then, is that syntactic effects
should show up only in the context of
RESEMBLANCE
relations.
This seems to be exactly right for some of the cases that are problematic for
a syntactic approach, such as examples involving syntactic non-identity (20)-(21)
and the absence of Condition A effects (22)-(23) (but see also Hestvik 1995).
(20)
a.
?? This problem was looked into by Kim, and Lee did too. (
RESEM
-
BLANCE
)
b.
? This problem was looked into by Kim, even though Lee already had.
(
CAUSE
-
EFFECT
)
(21)
a.
?? The letter from the Dean provoked a response from the Chair, and the
Provost did too. (
RESEMBLANCE
)
b.
? The Dean’s actions provoked a response from the Chair, despite the
fact that the Provost already had. (
CAUSE
-
EFFECT
)
(22)
a.
Doug
£
blamed himself
£
for the band’s collapse, and everyone else did
too. (
RESEMBLANCE
; strict reading difficult)
b.
Doug blamed himself for the band’s collapse, because everyone else
did. (
CAUSE
-
EFFECT
; strict reading possible)
(23)
a.
John wouldn’t introduce himself to everyone, but Mary did. (
RESEM
-
BLANCE
; no strict reading)
b.
Since John wouldn’t introduce himself to everyone, Mary did. (
CAUSE
-
EFFECT
; strict reading possible)
8
Ellipsis and Syntactic Representation
However, Kehler’s proposal also suffers from the same problems that afflict
a purely semantic account of ellipsis: there are contexts in which an elided VP
clearly occurs in a constituent that participates in a
CAUSE
-
EFFECT
relation, yet
also appears to be sensitive to syntactic constraints. For example, both (24a) and
(25a) manifest
CAUSE
-
EFFECT
relations, yet the elided VPs inside the relative
clauses are sensitive to the Adjunct Island Constraint (islands enclosed in brack-
ets). (24b) and (25b) demonstrate that ellipsis is possible in these examples as long
as the elided VP is not in an island.
(24)
a.
Sterling criticized every decision that Doug was upset [because Lou
did].
b.
Sterling criticized every decision that Lou did.
(25)
a.
Max refused to buy the shirt that I picked out even though it was less
expensive than the one that the salesperson complimented him [after
he did].
b.
Max refused to buy the shirt that I picked out even though it was less
expensive than the one that he did.
Similarly, (26a) and (26b) show that Condition B effects arise in
CAUSE
-
EFFECT
environments.
(26)
a.
Kim takes care of him
£
because he
£
won’t.
b.
His closest allies supported him
£
throughout this ordeal, even though
he
£
probably wouldn’t have.
In addition to these empirical problems, there is a third, more general prob-
lem with a mixed approach such as Kehler’s. If a purely semantic analysis is avail-
able in some examples, then it ought to be in principle available in all examples,
even if a syntactic analysis is preferred. In other words, an approach that allows for
the possibility of semantic recovery of VP meanings without concomitant syntactic
representation of an elided VP predicts that examples like (27a) and (28a) should be
no worse than (27b) and (28b), which violate coherence relations but do not violate
any syntactic constraints.
(27)
a.
?? This problem was looked into by Kim, and Lee did too.
b.
This problem was looked into by Kim, and Lee looked into it too.
(28)
a.
?? The letter from the Dean provoked a response from the Chair, and the
Provost did too. (
RESEMBLANCE
)
Christopher Kennedy
9
b.
The letter from the Dean provoked a response from the Chair, and the
Provost responded to it, too.
This is not an accurate characterization of the facts, however. While (27b) and
(28b) are less fluent than completely parallel structures, they are quite clearly more
acceptable than (27a) and (28a).
2 Ellipsis and Syntactic Representation
The discussion in the previous section does not provide conclusive arguments for
one answer to the representation question in ellipsis over another, but it lays out
some of the empirical ground that must be covered by any account. At the same
time, this discussion allows us to formulate very precisely a central prediction of
analyses that posit syntactic representation in the ellipsis site. Taking the deletion
analysis as the representative of such an approach (for simplicity; most of what I
will say here holds of a copying analysis as well), we can formulate the predictions
in (29).
(29)
Ellipsis and Syntactic Representation
If ellipsis involves deletion of syntactic structure, then:
a.
Elided constituents should be sensitive to syntactic constraints in gen-
eral.
b.
However, since ellipsis does not require pronunciation of the omit-
ted structure, elided constituents should be insensitive to syntactic
constraints that derive from morphophonological properties of lexi-
cal items.
In other words, the predictions of a syntactic analysis of ellipsis are more
subtle than they appear at first. For the most part, we should see the same sort
of syntactic behavior in overt and elided XPs; in particular, both elided and overt
XPs should be subject to the same set of configurational constraints. However, the
syntactic analysis does not predict that an ellipsis construction should have exactly
the same syntactic properties as a corresponding overt form. Crucially, because
ellipsis bypasses pronunciation, any constraints that make reference to the interface
between the syntax and the phonological component should be vacuously satisfied,
and therefore effectively “turned off”, in the case of ellipsis.
In the following two sections, I will show that at least one form of ellip-
sis — VP-deletion in English — behaves exactly as the predictions in (29) lead us
to expect. For the purpose of this paper, I will assume a multistratal framework
Dostları ilə paylaş: |