22
Ellipsis and Syntactic Representation
(71)
The Cubs need a left-handed hitter more than they do...
VP
VP
V
need
FP
DegP
left-handed
£
F’
F
DP
¥
DP
D
a
NP
DegP
¥¦£
NP
pitcher
Crucially, only a syntactic analysis of the sort given in (66) (together with the
assumption that pseudogapping involves VP-deletion plus extraction of the remnant
XP) can account for the (c) readings in (68)-(70). An alternative explanation of the
facts would be one that posited some kind of “attributive modifier ellipsis”, in which
the structure assigned to e.g. (69a) would be (72).
(72)
The Cubs need a left-handed hitter more than they do [
VP
need
¥¦£
] [
DP
a [
NP
[
DegP
left-handed] [
NP
pitcher]]]
£
The problem with this sort of analysis is that it also predicts that the (b) sentences
in (68)-(70) should have (c) readings, which is false. That is, such an analysis
cannot capture the fact that ellipsis of the attributive modifier in these examples
is “parasitic” on pseudogapping, whereas this follows directly from the structural
analysis proposed in K&M.
5
Only an analysis that assumes movement of DegP to
SpecFP plus deletion (i.e., syntax on the PF side of the derivation) allows for the
construction of a syntactic constituent of the type needed to get the (c) reading only
when the verb (phrase) is also deleted.
The third argument presented by K&M for the FP-structure in (66) comes
from particular uses of the verb make. This verb has an “evaluative” use that re-
quires an attributive modifier, as shown in (73).
(73)
a.
Peaches make delicious tarts.
5
This discussion is stated in terms of a deletion analysis of ellipsis, but the same arguments hold
of semantic analyses. A purely interpretive analysis such as the one developed in Dalrymple et al.
1991 runs into exactly the same problems of overgeneration outlined above, since it must posit an
operation that recovers “missing” attributive modifier meanings. If such an operation is possible in
the (a) sentences in (68)-(70), then it should also be possible in the (b) sentences, contrary to fact.
Christopher Kennedy
23
b.
# Peaches make tarts.
This restriction appears to be relaxed in pseudogapping contexts, however:
(74)
a.
Peaches make delicious pies more often than they do tarts.
b.
# Peaches make delicious pies more often than they make tarts.
The acceptability of (74a) follows directly given the FP structure in (66) and the
analysis of the pseudogapping sentences discussed above: (74a) can be assigned
the structure in (75), which satisfies the requirement that the complement of make
have an attributive modifier.
6
(75)
Peaches make delicious pies more often than they do [
VP
make [
FP
[
DegP
delicious]
£
F
¥
]]
[
DP
¥¦£
tarts]
Taken together, these facts provide strong support for the hypothesis that
attributive modifiers can, and sometimes must (see (65) above), raise from their
base position at the NP level to the specifier of a functional head within the nominal
projection but above the determiner. Since this movement is clearly licensed, it
must be the case that the locus of LBC violations is movement out of SpecFP, not
movement from the base position of the attributive modifier. That is, the problem
must be at the FP level, not the DP level.
This conclusion, together with the fact that ellipsis eliminates LBC effects
in attributive comparative deletion, lead K&M to propose a formulation of the LBC
in terms of the morphophonological expression of syntactic feature combinations.
Specifically, K&M claim that extraction of left branch modifiers in English (and re-
lated languages) is regulated by the principle of Full Interpretation (Chomsky 1981,
1986, 1995), which requires that every element in a particular interface represen-
tation have an interpretation at that interface. In the case of the syntax-phonology
interface, this means that all terminal nodes — structured bundles of syntactic fea-
tures — must have a phonological value. Following Halle and Marantz 1993, K&M
assume that a syntactic object “has a phonological value” if and only if it can be
paired with a corresponding morphophonological matrix from the lexicon. This
leaves open the possibility that the syntactic component can derive representations
that are well formed in all respects except that they contain objects without mor-
phophonological instantiations.
According to K&M, this is exactly what happens in LBC contexts. Like
the inverted DegPs, an attributive wh-operator must move through SpecFP (cf. how
6
I set aside here the question of what is responsible for this constraint. One possibility, given the
well-formedness of sentences like These peaches would make a hell of a pie is that evaluative make
subcategorizes for an FP.
24
Ellipsis and Syntactic Representation
tall a man vs. * a how tall man), with the result that the head of FP is assigned a
[+wh] feature by spec-head agreement. K&M’s proposal is that left branch effects
in English arise because the lexicon lacks a F
Q
head. On this view, examples
like those in (76)-(77) are ungrammatical not because of a constraint on movement
per se, but because the syntactic representations are unpronounceable: they violate
Full Interpretation at the PF interface, because F
Q
has no morphophonological
instantiation.
(76)
The Cubs start a more talented infield than [wh
£
the Sox start [
FP
¥¦£
F
[
DP
an
¥¦£
outfield]]]
(77)
How talented
£
do the Sox start [
FP
¥¦£
F
Q
[
DP
an
¥¦£
outfield]]?
In order to generate a well-formed structure, the [+wh] feature on F
must be
eliminated. This can happen in two ways. The first option is to pied-pipe the entire
FP with the wh-operator, in which case the [+wh] feature on F
can be checked
in the normal way. This is the strategy taken in questions such as (78), but it this
option is unavailable in null operator constructions (see Grosu 1994).
7
The second option is to delete a constituent containing the offending F
object. This is what happens in the well-formed examples of attributive comparative
deletion, in which a constituent containing the gap is deleted. (78) illustrates the
case where a VP has been deleted.
(78)
The Cubs start a more talented infield than [ wh
£
the Sox (do)
[
VP
start [
FP
¥¦£
F
Q
[
DP
an
¥¦£
infield]]]]
We now have an answer to our puzzle: if the LBC is a constraint on the mor-
phophonological instantiation of syntactic representations, and if ellipsis involves
deletion of syntactic representations (or, alternatively, an instruction to “bypass”
morphophonological instantiation, `a la Wasow 1972), then the fact that ellipsis of
7
Caterina Donati points out that we also don’t see this kind of pied-piping in comparatives in
which the operator is overt, however (in Italian, Bulgarian, and Romanian), so this may not be a
complete answer.
Movement of the entire FP to SpecCP may be exactly what is going on in examples of attributive
CD in which only an argument is missing, such as (i), however.
(i)
Jones produced as successful a film as [[
FP
wh
F
d
e
fhghij
[
DP
a
film]]
k
Smith produced
lk
]
This sort of analysis cannot be ruled out, if movement and deletion in SpecCP is part of the grammar
of comparatives (see Kennedy to appear for arguments to this effect). Moreover, given that English
does not allow deletion of argument DPs except in movement constructions (assuming a ‘copy and
delete’ theory of movement), this may be the only plausible analysis for such comparatives.
Dostları ilə paylaş: |