10
Ellipsis and Syntactic Representation
in which the syntactic component generates a pair of representations, one that in-
terfaces with the phonological component (PF), and one that interfaces with the
semantic component (LF) (as in Chomsky 1995, etc.). I will show that VP-ellipsis
constructions are sensitive to configurational constraints on (LF) representations,
but not to morphophonological constraints governing the interpretation of PF rep-
resentations, in line with the predictions in (29). I should note that I am adopting a
multistratal framework for convenience. The predictions in (29) hold of any theory
that includes both configurational and morphophonologically defined “interpretive”
constraints on syntactic well-formedness: ellipsis constructions should be sensitive
to the former but not the latter.
3 Ellipsis and Parasitic Gaps
3.1 “Non-parasitic” Gaps
As originally observed by Kim and Lyle (1996), apparent parasitic gap chains do
not show island effects when the expected position of the parasitic gap is contained
in a deleted VP (see also Lappin 1992).
(30)
Wh-islands
a.
Which article
£
did you summarize
¥¦£
after Jim asked [who had read
PG
£
]?
b.
Which article
£
did you summarize
¥¦£
after Jim asked [who would be
willing to]?
(31)
Adjuncts
a.
Which movie
£
did you see
¥¦£
because Polly was so excited [after going
to
PG
£
]?
b.
Which movie
£
did you see
¥¦£
because Polly was so excited [after she
did]?
(32)
Complex NPs
a.
Mayor Daley
£
, whom everyone met
¥¦£
except [the people who didn’t
know that there would be an opportunity to see
PG
£
], discussed Chicago
politics.
b.
Mayor Daley
£
, whom everyone met
¥¦£
except [the people who didn’t
know that there would be an opportunity to], discussed Chicago pol-
itics.
Christopher Kennedy
11
(33)
Coordinate structures
a.
Which books
£
did you read
¥¦£
after learning that Erik had bought
PG
£
and enjoyed them
£
?
b.
Which books
£
did you read
¥¦£
after learning that Erik had and enjoyed
them
£
?
Kennedy (1997) argues that these facts have a fairly straightforward and simple
explanation: the putative parasitic gap chains in (30b)-(33b) do not show island
effects because, contrary to what might be initially assumed, these sentences do not
actually contain parasitic gaps. Instead, the “gap” in the deleted VP is actually a
pronoun. In other words, the structural descriptions of (30b)-(33b) are as in (34a)-
(34d), where the struck-through text corresponds to the deleted VP.
(34)
a.
Which article
£
did you summarize
¥¦£
after Jim asked who would be
willing to [
VP
summarize it
£
]
b.
Which movie
£
did you see
¥¦£
because Polly was so excited after she
did [
VP
saw it
£
]?
c.
Mayor Daley
£
, who everyone met
¥¦£
except the people who didn’t
know that there would be an opportunity to [
VP
meet him
£
], discussed
Chicago politics.
d.
Which books
£
did you read
¥¦£
after learning that Erik did [
VP
read them
£
]
and enjoyed them
£
?
The reason that VP-deletion is possible here, despite the apparent non-identity be-
tween deleted and antecedent VPs, is that these examples are just instances of what
Fiengo and May (1994) call “vehicle change”: the observation that in many con-
texts, pronouns and other expressions (in particular, A-traces) “count as” identical
for the purpose of licensing deletion.
1
One of the arguments that Kennedy presents in favor of the claim that the
elided VPs in (30b)-(33b) contain pronouns, not parasitic gaps, comes from crossover
effects. As is well known, parasitic gaps show strong crossover effects (i.e., they
are subject to Condition C; see Postal 1993; Cinque 1990). This is illustrated by the
contrast in (35).
(35)
a.
Who
£
were they investigating
¥¦£
before he
£
knew they suspected
PG
£
?
b.
Who
£
were they investigating
¥¦£
before you knew they suspected
PG
£
?
1
See Merchant 2001 for a semantic licensing condition on ellipsis, based on Schwarzchild’s
(1999) theory of focus and deaccenting, that derives vehicle change.
12
Ellipsis and Syntactic Representation
Sentences like (30b)-(33b), however, do not show strong crossover effects:
(36)
a.
Who
£
did Maureen vote for because he
£
asked her to?
b.
Which students
£
did Otis report without them
£
knowing he had?
c.
Who
£
did you call before she
£
asked you to?
If the elided VPs in these examples contain pronouns, rather than parasitic
gaps, then the facts in (36) follow: these sentences have structures that are com-
pletely parallel to examples like the ones in (37), which also do not show crossover
effects.
(37)
a.
Who
£
did Maureen vote for
¥¦£
because he
£
asked her to vote for him
£
b.
Which students
£
did Otis report
¥¦£
without them
£
knowing he had re-
ported them
£
c.
Who
£
did you call
¥¦£
before she
£
asked you to call her
£
If, however, the elided VPs in these examples contained parasitic gaps, then they
should be as ungrammatical as corresponding examples with overt parasitic gaps,
such as those in (38).
(38)
a.
Who
£
did Maureen vote for
¥¦£
because *he
£
/Charles asked her to sup-
port
PG
£
b.
Which students
£
did Otis report
¥¦£
without *them
£
/you knowing he
suspected
PG
£
c.
Who
£
did you call
¥¦£
before *she
£
/Marcus asked you to visit
PG
£
While the sentences under consideration do not show crossover (Condition
C) effects, they
do show Condition B effects:
(39)
a.
Who
£
did Maureen recommend because he
£
/Louis wouldn’t?
b.
Which students
£
did Otis report even after telling them
£
/you to?
c.
Who
£
did you try to serve before seeing that she
£
/I already had?
Again, this follows if the “gaps” in the ellipsis sites are pronouns:
(40)
a.
Who
£
did Maureen recommend
¥¦£
because he
£
wouldn’t recommend
him
£
b.
Which students
£
did Otis report
¥¦£
even after telling them
£
to report
them
£