is in English, apparently. (Though das certainly has such property or situ-
ational uses in general, as the subject of verbs meaning ‘bother’, ‘surprise’
etc. and in similar nonsubject contexts as well.) Instead, what one would like
is something more along the lines of the following:
(54)
Derjenige,
the.one
der
who
in
in
der
the
gegebenen
given
Beziehung
relationship
zu
to
dem
the
gerade
just
angedeuteten
indicated
Tisch
table
steht,
stands
ist
is
mein
my
Vater.
father
‘The one who stands in the given relationship to the just indicated
table is my father.’
This example has the desired syntactic property of having Mein Vater in
the nominative case, here as the subject in an inverted specificational copular
clause. There is however no hope for a theory that would allow such a syn-
tactic object to be deleted on the basis of the context, linguistic or otherwise.
The example therefore stands as a datum that cannot be accounted for under
the limited ellipsis account discussed above.
The second problematic example has the same basic difficulty: it simply
fails to be equivalent in the given context to any possible reading of That is
X
.
(55)
A father is worried that his daughter will spill her chocolate milk. The
glass is very full, and she is quite young, and prone to accident. He
says, ‘Both hands!’
In this context, one cannot say, ‘That’s both hands!’ or ‘Those are both
hands!’ and expect the child to understand this as a command to use both
hands to hold her glass. Nor, as Stainton points out, does it help to think that
there is an elided verb ‘use’ here—doing so is equivalent, he shows, to aban-
doning the limited syntactic ellipsis account for one that is unconstrained and
ultimately untenable. For Stainton, the property needed here (namely USE)
is given by the context, and both hands supplies its argument, but there is no
syntactic or semantic representation of USE present.
11
In sum, there are simply some examples that the syntactic ellipsis analysis
cannot accommodate.
11
I merely note that in a language like Russian that has instrumental case, such an example
appears necessarily in the instrumental, which is also the case assigned by the Russian verb
pol’zovat’sja
(‘use’):
3.2.4.
Summary
The above discussion has shown that for almost every kind of example, some
more or less plausible syntactic ellipsis story can be told. But at the end of the
day, I feel like the boy with his thumb in the dike: the dike is going to keep
springing leaks, and while I may not run out of theoretical thumbs, one can’t
help but feel tired trying to plug all the leaks. Theorists with more syntactic
leanings than I have may feel this strategy is worth pursuing to more extremes
than I do. I’m willing to concede that syntactic ellipsis is required only when
connectivity effects are observed, and that this holds only in two subcases:
first, when there is a linguistic antecedent as in short answers, and second,
when there is a syntactic slot to be filled, as discussed in section 5 below. For
the rest, including the many ‘bare nominative’ examples, a syntactic solution
seems to me to be less attractive on the whole than the alternatives.
This means that the syntactic conclusions of Shopen 1972 and Barton
1990 cannot be escaped: ‘bare’ phrases must be generable on their own, with
no local syntactic embedding of any kind. Once we accept this conclusion,
we must begin to explore its consequences for the models of grammatical
competence we construct (see the papers in Progovac et al. 2006 for several
relevant proposals). The urgent task becomes what to make of the mecha-
nisms for handling what otherwise look like syntactic dependencies, in form
(case, number, gender, person, anaphoricity, aspectual marking, etc.). Only
some, not all, of the cases of interest can even possibly be handled under a
syntactic ellipsis analysis.
4.
Ellipsis
semantic
as ‘slot-filling’
The question, then, given a divisa et impera strategy, is whether the remain-
ing cases can be handled with a semantic ellipsis analysis. In this section, I
concentrate in particular on the three main subcases: phrases that pick out
individuals, properties, and quantifiers. There is a way of construing the se-
(i)
dvumja
rukami!
two.
INSTR
hands.
INSTR
(ii)
*dve
ruki!
two.
NOM
/
ACC
hand.
GEN
This may be a case where the case itself contributes some semantic restriction on the kinds
of predicates it can be combined with; such a strategy fails in general for structural cases,
however, and so can’t be used to account for accusatives, as discussed in section 5.
mantics of such expressions which I believe is fully consonant with Stain-
ton’s points about their interpretations in context, but which makes use solely
of commonly assumed, independently posited, semantic combinatorics. The
basic idea is to let the semantic value of these expressions (what they ‘say’)
include a variable over the relevant kind of object, and to let this variable
receive its value in the usual way, namely by an assignment function (or its
equivalent in variable-free treatments) whose content is of course itself en-
tirely determined by context in the Gricean pragmatic way.
The core of the debate seems to come down to whether Stainton is right
when he writes that “what is asserted ... is fully propositional; but what
is metaphysically determined by slot-filling and disambiguated expression-
meaning is something less than propositional” (2006:228) and that ordinary
words and phrases used in isolation “don’t have ‘slots’ that yield something
propositional when they are used in context” (2006:158). Here I think that
there is a reasonable reading of slot-filling under which the large majority of
examples adduced can be handled as propositional after all.
By ‘slot-filling’, Stainton means the contextually determined values of
items like indexicals and pronouns, as well as other elements whose semantic
value is generally taken to be a variable. As usual, the paradigm case is that
of a pronoun. An example like (56a) has the semantic value in (56b) (setting
aside the number and gender contributions of he), which, under the variable
assignment in (56c), has the truth conditions in (56d).
(56)
a. He
2
left.
b. leave(x
2
)
c. g = [ x
2
→ sam ]
d.
leave
(x
2
)
g
= 1 iff Sam left
The case of pronouns is the simplest one, especially when these pick out
individuals, type . On widespread conceptions of the semantic compo-
nent of the grammar, such variables are put to a variety of uses. Consider the
following expressions:
(57)
expression
type
a.
sick
< e,t >
b.
sick
( john)
< t >
c.
sick
(x
3
)
< t >
d.
λ x
3
[sick(x
3
)]
< e,t >
Dostları ilə paylaş: |