86
Paul Elbourne
us say that one is an Agent of a plural event if and only if one is an Agent, in
the normal sense, of all the events that are atomic parts of it. We will also need
to make sense of the notion of an event (a plural event, to be sure) satisfying a
plural individual made up of VP-meanings. Let us say that for any event e and
functions f, g of type s,t , f
⊕ g(e) = 1 if and only if there exist events e and
e such that f(e ) = 1 and g(e ) = 1 and e ≤
i
e and e ≤
i
e.
Assuming the speaker is John, we finally arrive at the truth conditions in
(72) for the whole sentence:
(72)
∃t(t <
NOW
& at t : ∃e(swimming ⊕ climbing(e) = 1 &
Agent
(e, John)))
In other words, there was in the past a plural event e such that e had as its parts an
event of swimming the English Channel and an event of climbing Kilimanjaro
and John was the agent of e, in the new sense whereby he was the agent of every
atomic part of e. These truth conditions seem to be intuitively adequate.
I will shortly go on to analyze (61a) (the sentence about the globe-trotting
desires of Bob and Alice), but before doing so I should perhaps be more explicit
about the new syntax of VP-ellipsis than I have been so far. The proposal is that
a vP can be spelled out by the rules and rule-schemas in (73):
(73)
vP
→ v T
HE
P
T
HE
P
→ T
HE
A
ND
0
P
A
ND
n
P
→ A
ND
n+1
P VP
A
ND
n
P
→ A
ND
n+1
VP
T
HE
→ T
HE
RP
RP
→ R
m, st,t
RP
→ R
m, e,stt
pro
l,e
I am not aware of any cases where the RP has to contain more than one variable
of type e, so I have just listed two cases above; a more sophisticated treatment
along the lines of that given to A
ND
0
P could be devised if necessary.
The Semantics of Ellipsis
87
TP
DP
neither of them
T
λ
2
T
can
vP
t
2
v
v
T
HE
P
T
HE
T
HE
RP
R
1, e,stt
pro
2,e
A
ND
0
P
A
ND
1
P
A
ND
2
VP
sail round the world
VP
climb Kilimanjaro
Figure 4: Neither of them can. . .
Let us move on to the analysis of (61a), repeated here as (74).
(74)
Bob wants to sail round the world and Alice wants to climb Kilimanjaro,
but neither of them can, because money is too tight.
The LF for neither of them can will be that shown in Figure 4. The free variable
R
1, e,stt
will be assigned a meaning as follows:
(75)
[[R
1, e,stt
]]
g
= λx.λf
s,t
.x desires that there be an event e such that f(e) =
1 and Agent(e, x))
I will avail myself of the following simple denotations for can and neither of
them:
(76)
[[can]]
g
= λf
s,t
.it is possible that there be an event e such that f
(e) = 1
[[neither of them]]
g
= λf
e,t
.¬∃x((x = Bob ∨ x = Alice) & f(x) = 1)
88
Paul Elbourne
Given these denotations, the truth conditions for this example come out to be as
in (77). I use italicized expressions to abbreviate meanings of the VPs.
(77)
¬∃x((x = Bob ∨ x = Alice) and it is possible that there be an event e
such that σf (x desires that there be an event e such that f
(e) = 1 and
Agent(e, x) and f
≤
i
sailing ⊕ climbing)(e ) = 1 and Agent(e , x))
In other words, there does not exist an individual x such that x is Bob or Alice
and it is possible that x be the agent of an event that satisfies the unique predicate
f such that x wants to be the agent of an f-event and f is one of sailing round
the world and
climbing Kilimanjaro. This seems to be intuitively adequate.
8
(63a), repeated here as (78a), will work by the same means, as suggested by
the paraphrase in (78b).
(78)
a.
Whenever Max uses the fax or Oscar uses the Xerox, I can’t.
b.
Whenever Max uses the fax or Oscar uses the Xerox, I can’t perform
the particular action or actions out of
using the fax and
using the
Xerox that are being performed.
In other words, there will be two small VPs using the fax and using the Xerox,
and an R variable will be assigned a denotation something like “currently being
performed.” Working out an exact analysis would require us to make decisions
regarding what entities whenever quantifies over (time intervals? situations?)
and whether these are represented in the syntax. The general outlines are clear,
8
The idea of having the variable R
1, e,stt
provide extra descriptive material to modify syn-
tactically more robust material is reminiscent of the approach to quantifier domain restriction
that posits variables in the syntax, as proposed by von Fintel 1994, Stanley 2000 and Stanley
and Szab´o 2000. In particular, von Fintel (1994) sometimes has two variables in such positions,
one an individual variable bound by the subject, in order to deal with sentences like Only one
class was so bad that no student passed, where we are to understand “only one class x. . . no
student in x. . . .” The combination of a definite article plus a relation variable plus an individual
variable is also reminiscent of the LF configuration posited by Heim and Kratzer (1998) to spell
out donkey pronouns.