The Semantics of Ellipsis
69
(17)
I did everything Mary did. Mary swam the English Channel and Mary
climbed Kilimanjaro, and I did too. (Fiengo and May 1994)
(18)
Whenever Max uses the fax or Oscar uses the Xerox, I can’t. (Fiengo
and May 1994)
The interpretations of these examples are tricky. (17) is the easiest. It pretty
clearly means “. . . and I swam the English Channel and climbed Kilimanjaro
too.” One wants to paraphrase (18) “Whenever Max uses the fax or Oscar uses
the Xerox I cannot use the fax or the Xerox,” but attention must be paid that
we do not analyze the sentence as meaning “. . . I cannot do either.” The correct
interpretation seems to be something like “. . . I cannot use whichever one is
being used.” Similarly, the ellipsis in (16) cannot be resolved “. . . neither of
them can do either,” but must mean something like “neither of them can do the
thing they want.” These facts are clearly beyond the ability of the common view
to capture. There are similar cases that involve NP-deletion, as pointed out in
Elbourne 2001:
(19)
John needs a hammer. Mary needs a mallet. They’re going to borrow
Bill’s.
This seems to mean “They’re going to borrow Bill’s hammer and mallet.” I will
refer to examples like these as involving split antecedents.
The fourth problem for the common view is that some cases of VP-ellipsis
and NP-deletion require no linguistic antecedents whatsoever. Many people find
the original examples of Hankamer and Sag (1976), repeated here, quite felici-
tous:
(20)
(Hankamer attempts to stuff a 9-inch ball through a 6-inch hoop.)
Sag: It’s not clear that you’ll be able to.
(21)
(Sag produces a cleaver and makes as if to hack off his left hand.)
Hankamer: Don’t be alarmed. He never actually does.
70
Paul Elbourne
And there are many examples about which there is no debate whatsoever. The
following list of VP-ellipsis cases is adapted from Merchant forthcoming, which
should be consulted for original references:
(22)
(John attempts to kiss Mary while driving.)
John, you mustn’t.
(23)
(A piece of chocolate cake is offered.)
I really shouldn’t.
(24)
(As an invitation to dance.)
Shall we?
(25)
(Mary gets John an expensive present.)
Mary, you shouldn’t have!
(26)
(Gesturing towards an empty chair.)
May I?
(27)
(Responding to the last.)
Please do.
(28)
(Seeing someone about to do a shot of Tequila.)
If you can, I can too.
(29)
(Looking at someone psyching herself up to jump across a wide gap.)
I bet she won’t.
(30)
(Seeing someone who has just died his hair green.)
You didn’t!
(31)
(Sitting next to someone doing something annoying.)
Must you?
(32)
(On witnessing someone about to do anything undesirable.)
Don’t!
Once again there are analogous cases of NP-deletion:
The Semantics of Ellipsis
71
(33)
(There are lots of barking dogs in the yard. We look at them without
speaking. I point and say:)
Harry’s is particularly noisy. (Elbourne 2001)
I will call this phenomenon ellipsis with no linguistic antecedent.
We have four problems, then, that the common view does not seem able to
deal with, those of binderless sloppy readings, ellipsis-containing antecedents,
split antecedents and ellipsis with no linguistic antecedent. In this article I lay
out a theory of ellipsis that is compatible with all these data (section 2). I com-
ment on relevant previous literature in section 3. Section 4 concludes.
It should be emphasized that I will be concentrating entirely on the seman-
tics of ellipsis, without paying attention to any syntactic constraints there may
be on when ellipsis is possible. (See Johnson 2001 for a good overview of syn-
tactic and semantic issues connected with VP-ellipsis.) And I will be concentrat-
ing on VP-ellipsis and NP-deletion, without attempting to account for sluicing,
pseudo-gapping and other kinds of ellipsis.
2 A Theory of Ellipsis
2.1
The Framework
2.1.1
Event Semantics and Little v
Following much work in event semantics and argument structure, I assume that
VPs are predicates of events (Davidson 1967, Parsons 1990, Tenny and Puste-
jovsky 2000, Pylkk¨anen 2002), and that subjects are introduced by a special
head v that takes the VP as its sister (Kratzer 1996, Pylkk¨anen 2002).
To give a flavor of the general approach, let us take the example in (34),
which will have the structure in (35). The lambda-abstractor in the syntax is
from Heim and Kratzer 1998.
(34)
Brutus stabbed Caesar.
72
Paul Elbourne
(35)
[
TP
Brutus [λ
2
[T
past
[
vP
t
2
[v [
VP
stab Caesar]]]]]]
The following lexical entries will enable vPs and VPs to be predicates of events,
with the whole sentence an assertion of the existence of certain kinds of time
intervals and events.
3
(36)
[[T
past
]] = λp
s,t
.∃t(t <
NOW
& at t : ∃e p(e) = 1)
[[v]] = λp
s,t
.λy.λe.p(e) = 1 & Agent(e, y)
[[stab]] = λz.λe.stabbing(e) & Theme(e, z)
These lexical entries give the meanings displayed in Figure 1 for different parts
of the tree. The sentence is interpreted with respect to a variable assignment g.
Figure 1 is not necessarily a serious contender for what the syntax and seman-
tics of this sentence actually look like. But it will be useful to have something
concrete to work with.
2.1.2
Pronouns and Names
I will follow Postal (1966), Stockwell, Schachter and Partee (1973), Abney
(1987), Longobardi (1994), Uriagereka (1995), Elbourne (2001, forthcoming)
and Neale (2005) in assuming that pronouns are basically determiners. In par-
ticular, third person pronouns are definite articles whose complements must be
phonologically null (Elbourne 2001, forthcoming); these complements can be
referential indices, which I take to be of type e,t , or they can be normal NPs
unpronounced because of NP-deletion. For example, the index 2 in what we
would normally write as he
2
might be interpreted, by means of a variable as-
signment mapping 2 to John, as [λx.x
= John]; since he has the same meaning
3
Some head should presumably convert the denotation of the whole sentence into a set of
possible worlds or situations, and there might also be heads that contribute illocutionary force.
I omit all these for simplicity’s sake and write as if the denotations of sentences were truth
values. I also do not properly take account of the indexical nature of tense, which presumably
must make reference to the time of utterance. I gesture towards this with the term
NOW
in the
metalanguage, which is supposed to be an indexical taking as its value the time of utterance on
each occasion of use.
Dostları ilə paylaş: |