The Semantics of Ellipsis
89
however.
9
There are also examples of split antecedents involving NP-deletion, as we
saw in (19), repeated here as (79). I will analyze the variant in (80), which is
more revealing of structure since the quantifier each actually seems to bind into
the NP-deletion site.
(79)
John needs a hammer. Mary needs a mallet. They’re going to borrow
Bill’s.
(80)
John needed a hammer. Mary needed a mallet. Each borrowed Bill’s.
We will need a set of rules for spelling out silent NPs parallel to the ones we
saw for VPs in (73). The rules and rule-schemas in (81) will suffice.
(81)
DP
→ D THEP
THEP
→ THE AND
0
P
AND
n
P
→ AND
n+1
P NP
AND
n
P
→ AND
n+1
NP
THE
→ THE SP
SP
→ S
m, et,t
SP
→ S
m, e,ett
pro
l,e
Translating the proposal just explored with respect to VP-ellipsis into the NP
domain, we arrive, then, at the slightly simplified LF in Figure 5 for the last
sentence of (80). I ignore any complexity there may be behind the surface forms
each and
Bill’s. The new operators THE and AND
2
will receive the interpreta-
tions in (82) and (83), parallel to the interpretations of T
HE
and A
ND
2
.
(82)
[[THE]]
g
= λF
et,t
.λG
et,t
.σf(F (f) = 1 & G(f) = 1)
9
The present apparatus can also be put into service to analyze cases of overt conjunction
of VPs, as in
John walked and sang. All that is needed is for the overt
and in this position to
mean λf
s,t
.λg
s,t
.f ⊕ g. There are arguably conceptual advantages to having and produce a
sum of entities when it appears between VPs, just as it does when it conjoins expressions of
type e. See Krifka 1990 and Lasersohn 1995 for detailed proposals concerning the non-boolean
conjunction of VPs, and Winter 2001 for discussion.
90
Paul Elbourne
TP
DP
each
T
λ
2
T
T
past
vP
t
2
v
v
VP
borrow
DP
D
Bill’s
THEP
THE
THE
SP
S
1, e,ett
pro
2,e
AND
0
P
AND
1
P
AND
2
NP
hammer
NP
mallet
Figure 5: Each borrowed Bill’s
(83)
[[AND
2
]]
g
= λf
e,t
.λg
e,t
.λh
e,t
.h ≤
i
f ⊕ g
The nouns
hammer and
mallet receive the denotations one might expect, and the
free variable S
1, e,ett
will receive the following interpretation from the variable
assignment g:
(84)
[[S
1, e,ett
]]
g
= λx.λf
e,t
. x needs an f
(85)
[[hammer]]
g
= λx.x is a hammer
(86)
[[mallet]]
g
= λx.x is a mallet
The Semantics of Ellipsis
91
I abstract away from the complexities inherent in the analysis of transitive in-
tensional verbs like
need. Allowing ourselves the convenient lexical entries in
(87), (88) and (89) for each, Bill’s and borrow, we arrive at the truth conditions
in (90) for the last sentence of (80). I use italicized words to abbreviate the
meanings of hammer and mallet.
(87)
[[each]]
g
= λf
e,t
.∀x((x = John ∨ x = Mary) → f(x) = 1)
(88)
[[Bill’s]]
g
= λf
e,t
.ιx(x is Bill’s & f(x) = 1)
(89)
[[borrow]]
g
= λx.λe.borrowing
(e) & Theme(e, x)
(90)
∀x((x = John ∨ x = Mary) → ∃t(t <
NOW
& at t : ∃e(borrowing(e)
& Agent(e, x) & Theme(e, ιy(y is Bill’s & σf(x needs an f & f ≤
i
hammer ⊕ mallet)(y) = 1)))))
The claim, then, is that the last sentence of (80) is true if and only if, for all x
such that x is Mary or John, x was the Agent of a borrowing event whose Theme
was the unique item of Bill’s that satisfied the unique predicate f such that x
needed an f and f was one of hammer and mallet. This seems to be accurate.
It is time to consider how to integrate the model that we have built up for
split antecedent cases with the theory that we developed in previous sections for
binderless sloppy readings and ellipsis-containing antecedents. Recall Theory
the First in (39), repeated here as (91):
(91)
Theory the First
VP-ellipsis and NP-deletion consist in the generation of bare VP and
NP nodes, respectively. These structures are sent to PF. There is an LF
process of resolving the ellipsis, whereby the bare nodes are replaced
with a copy of a phrase of the same syntactic category drawn from the
linguistic environment.
We can combine this theory with the procedures we have posited to deal with
the split antecedent cases by adopting the following statement: