Christopher Kennedy
13
c.
Who
£
did you try to serve
¥¦£
before seeing that she
£
already had served
her
£
The conclusion to draw from these facts is that the gaps in the elided VPs
in these examples are pronouns. It follows that (30b)-(33b) can be assigned the
syntactic representations in (34a)-(34d), explaining the absence of island effects.
More generally, if the principles of the Binding Theory (in particular, Condition B)
apply to syntactic representations, as is standardly assumed, then it must be the case
that elided VPs are syntactically represented.
Before moving on, we should first consider an alternative view of the Bind-
ing Theory, which views constraints on coreference as constraints on the morpho-
syntactic expression of particular types of meanings, rather than as structure-based
constraints on the possible interpretations. In such a model, the facts discussed here
would not necessarily provide evidence for syntactic representation in ellipsis, as
pointed out in Hardt 1999. An example of this type of approach to the Binding
Theory is developed in Reinhart and Reuland 1993, in which Condition B is stated
as in (41).
(41)
A reflexive predicate is reflexive-marked.
In essence, this constraint requires any predicate (at least) two of which’s argu-
ments are co-valued to be morphologically (or lexically) marked as reflexive. (41)
thus correctly predicts that an example like (42a) is ungrammatical on the reading
indicated by the coindexing even if there is no internal structure to the VP.
(42)
a.
Who
£
did Otis nominate
¥¦£
because she
£
couldn’t?
b.
ab1cPedgfh¥i¦p5%XEGqr&%@6s¥t9uP
BECAUSE
vxwyd
P%XEGqr&%@6s¥P3
The predicate is reflexive (in Reinhart and Reuland’s sense), as indicated by the log-
ical representation in (42b), but it is clearly not reflexive-marked. More precisely,
if there is no structure to the elided VP, it could never be reflexive-marked.
2
The problem with this analysis is that it is too strong. In particular, without
some weakening, it rules out reflexive interpretations across the board: the second
2
This is presumably exactly the right analysis of (i), in which the presence of the VP-anaphor it
indicates an absence of internal structure.
(i)
Who
did Otis nominate
because she
couldn’t do it?
If this predicate is reflexive-marked, then the sentence is perfectly acceptable, as expected:
(ii)
Who
did Otis nominate
because she
couldn’t do it herself
?
14
Ellipsis and Syntactic Representation
conjunct in a simple example like (43a) should also violate (41), because the reflex-
ive predicate in the second conjunct is not reflexive-marked. If the elided VP has
internal structure, however, as indicated in (43b), then it is reflexive-marked, even
if this is not apparent in the surface string.
(43)
Otis served himself, and Alex did too.
(44)
Otis served himself, and Alex did [
VP
serve himself] too
3.2 “Missing” Parasitic Gaps
The analysis of the facts in the previous section builds on the idea that the deleted
VP can be given the “non-parasitic” gap analysis in (45a). However, as pointed
out by Shimada (1999) and Postal (2001), the alternative “missing” parasitic gap
structure in (45b) is a possible analysis of the elided VP, but it is not chosen since it
would result in an ill-formed structure.
(45)
a.
[
VP
... pro ...]
the non-parasitic gap structure
b.
[
VP
...
PG
...]
the missing parasitic gap structure
If we could find contexts in which (45b) had to be the actual analysis, and if
such contexts behaved syntactically just like p-gap constructions (obeyed islands,
showed strong crossover effects, etc.), then we would have even more evidence that
elided VPs have syntactic structure. The only difference between (45a) and (45b)
is a structural/syntactic one (semantically, both pro and
PG
are interpreted as bound
variables), therefore if we can show that some instances of VP-deletion must have
the former structure and some the latter, we will have provided evidence for syn-
tactic representation in the ellipsis site. Postal (2001) makes essentially this point
in the conclusion of his paper; here I am simply fleshing out the argument in more
detail.
We begin with Engdahl’s (1985, p. 41, fn. 19) observation that VP-deletion
improves some parasitic gaps outside the deletion site:
(46)
a.
Otis is a person who
£
I admire
¥¦£
because close friends of
PG
£
became
famous.
b.
Otis is a person who
£
I admire
¥¦£
because close friends of
PG
£
seem
to.
(47)
a.
Which film
£
did you see
¥¦£
because a critic of
PG
£
was excited?
b.
Which film
£
did you see
¥¦£
because a critic of
PG
£
advised you to?