92
Paul Elbourne
(92)
Theory the Second
(i)
VP and NP nodes may be bare.
(ii)
vPs may be spelled out as in (73).
(iii)
DPs may be spelled out as in (81).
(iv)
VPs and NPs in A
ND
0
Ps and AND
0
Ps must be bare.
(v)
A bare VP or NP node must be replaced at LF by a copy of a
phrase of the same syntactic category drawn from the linguistic
environment.
In other words we view the trees in Figures 3–5 not as being base-generated
but as deriving from structures that originally had bare VPs and NP nodes in
their A
ND
0
Ps and AND
0
Ps. All else proceeds as previously described, and we
still retain the option of handling the ellipsis-containing antecedent cases and
binderless sloppy readings with the simpler structures posited earlier.
10
Before we leave these data, we should note that it is also possible to concoct
labored but not ungrammatical examples that combine the traits of the various
species that we have been examining. (93), for example, is a combination of a
split antecedent case and a case of an ellipsis-containing antecedent:
10
An alternative way of unifying Theory the First with the structures posited for split an-
tecedent cases would maintain that VP-ellipsis and NP-deletion always involve definite de-
scription structures of the type posited in (73) and (81). In cases without split antecedents we
would just have one VP or NP as part of the structure, and the operator A
ND
1
or AND
1
. If we
suppose that some trivial property is generally available for the denotation of RP and SP when
these phrases are redundant, it turns out that a definite description structure with just one NP or
VP is semantically equivalent to just having the NP or VP there by itself. I marginally prefer
the option given in the text because of the complexity of the structures that result in cases of
ellipsis-containing antecedents if we suppose that we always have definite descriptions in ellip-
sis. But the issue is a subtle one, and the theory described in this note has a certain kind of unity
that cannot be claimed by the one in the text.
The Semantics of Ellipsis
93
(93)
When Bob had to sail round the world and Mary had to climb Kili-
manjaro, they didn’t want to; and when Bob had to swim the English
Channel and Mary had to climb K2 they didn’t, either.
It is simple to derive this example in the current theory: the overt want to is
followed by a little v and a silent T
HE
P containing two bare VP nodes, and
before these VP nodes are filled in this structure is copied and used to resolve
the VP-ellipsis in the second sentence. There are then two separate processes of
resolving split antecedent ellipsis, one in each sentence.
2.5
Ellipsis with No Linguistic Antecedent
Recall the cases of ellipsis with no linguistic antecedent in (22)–(33), some of
which are repeated in (94)–(98).
(94)
(John attempts to kiss Mary while driving.)
John, you mustn’t.
(95)
(A piece of chocolate cake is offered.)
I really shouldn’t.
(96)
(As an invitation to dance.)
Shall we?
(97)
(Mary gets John an expensive present.)
Mary, you shouldn’t have!
(98)
(There are lots of barking dogs in the yard. We look at them without
speaking. I point and say:)
Harry’s is particularly noisy.
The question is how to integrate these cases into the framework developed in
section 2.4.
Roughly speaking, what the examples of ellipsis with no linguistic antecedent
have in common is that there is some obvious sensory (in these cases visual)
94
Paul Elbourne
clue to the property conveyed by the unpronounced phrase. The clue need not
always be an instance of the relevant action or entity before the very eyes of the
speaker and hearer: there is kissing in the scenario of (94) and a dog in that of
(98), but not necessarily any eating in that of (95), and certainly not any eating
of the piece of cake being offered; and there may or may not be dancing actually
taking place when (96) is uttered, provided that it is clear to speaker and hearer
that they are at a dance. But in the cases of VP-ellipsis there must at least be
an obvious result of the action in question or a stimulus towards performing it.
It is hard to be more precise, and I will leave the matter here for now, pending
further research.
I propose that in these cases too we have full syntactic VPs and NPs at LF.
11
So in (94), for example, we might have something like [
VP
kiss me now]. We
can now emend our theory to the following, which is the final version:
(99)
Theory the Third
(i)
VP and NP nodes may be bare.
(ii)
vPs may be spelled out as in (73).
(iii)
DPs may be spelled out as in (81).
(iv)
VPs and NPs in A
ND
0
Ps and AND
0
Ps must be bare.
(v)
A bare VP or NP node must be filled in at LF by a VP or NP that
is highly salient.
(v)
A VP or NP is highly salient if and only if:
11
One sometimes gets the impression that some theorists think that verb-phrase meanings
that are merely contextually salient or able to be worked out, as opposed to occurring as the
value of some constituent in the linguistic environment, should not come to be represented as
syntactically fully-fledged VPs. It is unclear what the grounds for this view could be, however.
If it ever happens that we think of things and then put our thoughts into words, which is not
implausible, we are extraordinarily adept at moving from non-linguistic to linguistic modes of
representation.
Dostları ilə paylaş: |