outside Madison merely indicates that the so-labeled road goes to Minneapo-
lis/St. Paul, of course).
Is this enough to establish Stainton’s desired point? It all depends on
whether we take the utterance of (43a) to assert that THIS CHAIR IS FOR
[an editor of Linguistics and Philosophy], or whether this content is merely
implicated (perhaps conventionally) as part of the meaning of the labeling-
function (assuming the latter is in play here as well). Such considerations
apply also to most of the examples in (41). Is the act of labeling something
an assertion in the desired sense? If it is, then the debate is over, for me, at
least—I have no reason to think that every label has a TP structure subject
to elision. (Though even if this debate ends here, the hard work of sorting
out the linguistic facts continues). If not, it’s not clear just what the debate is
about: if I label something conventionally (such as by putting a title on my
paper, or wearing a nametag, or pointing at myself and saying ‘Tarzan’), but
knowingly mislabel, have I lied? What exactly is the difference between lying
and misleading when applied to such cases? Imagine a criminal, Sam, who
wants to rob a train and whose plan is to make the train engineer stop the train
in a field by mounting a sign at the side of the railroad tracks:
(44)
End of track
Imagine that in fact the track does not end near where Sam has placed the
sign. Does the sign ‘lie’? (Intuition says no: lies need intentional agents.) Has
Sam thereby ‘lied’? Certainly his intent was to mislead, and he used appropri-
ate linguistic means to do this (posting the sign in Greek in Kansas would do
little for his schemes). While we have clear intuitions that someone who says
or writes an intentional falsehood has lied (committed libel or perjury, etc.),
it’s less clear that ‘lying’ applies to (44). This is one of the reasons that laws
recognize degrees, and that there are different laws in different jurisdictions.
3.2.2.
Other syntactic questions
Other technical questions arise for a syntactic ellipsis analysis. The first in-
volves the inability of ‘fragments’ to be embedded (unlike for example VP-
ellipsis and sluicing, which are not limited to matrix clauses):
(45)
Even though it is obviously true,
a. no-one noticed that *(it’s) on the stoop!
b. Jack didn’t exclaim that *(it was) moving pretty fast!
c. few mail carriers recognized that *(it is) from Spain!
d. she said that *(he is) one of Anthony Carroll’s best men.
(46)
a. Jenny told us that *(that is) Barbara Partee.
b. Anita hinted that *(it was) Rob’s mom.
c. Katerina complained that *(it was) Nova Scotia.
On a syntactic account, this fact would have to be captured by positing a
feature dependence between the E trigger for ellipsis and the matrix clause,
such as [Clause-type:Matrix], using the notation of Adger 2002. The ques-
tion is whether it is independently plausible to believe that certain operations,
rules, or items are restricted in just this way: as Stainton 2006:116 puts it,
‘if these posited expressions really do have sentential structures, they should
embed in all sentential frames’. In fact there are many phenomena across lan-
guages that are limited in exactly this way, a fact that can be captured in a
variety of ways (such as with a lexical featural mechanism, as Adger pro-
poses), but whose existence is in no doubt. The asymmetry between matrix
(or ‘root’) clauses and embedded ones traditionally goes under the rubric of
the ‘Penthouse Principle’ (so named in Ross 1973; see de Haan 2001 for a
recent discussion), whose generalization we can state informally but mem-
orably as ‘the rules are different if you live on the top floor’. Examples of
matrix-only phenomena are legion; I list here just a few.
(47)
Phenomena occurring only in matrix clauses
a. German and Dutch verb second (Vikner 1995):
Das
the
Buch
book
hat
has
er
he
gelesen.
read
‘He read the book.’
b. Hidatsa declarative clause marker -c (Boyle 2007):
puušíhke-š
cat-
DET
.
DEF
mašúka-š
dog-
DET
.
DEF
éekaa-c
see-
DECL
‘The cat sees the dog.’
c. Imperatives (in many, perhaps all, languages; Postdam 1998):
Ánikse
open.
IMP
.2s
tin
the
porta!
door
(Greek)
‘Open the door!’
d. Subject-auxiliary inversion in English questions (McCloskey 2006):
How many presents did he get? vs. *How many presents he got?
We were surprised at {*how many presents did he get. | how many
presents he got.}
e. Albanian polar interrogative particle a:
A
Q
je
are
ti
you
të
AGR
lodhur?
tired
‘Are you tired?’
f. Greek dubitative interrogative particle araje (Giannakidou to ap-
pear):
Tha
FUT
perasi
pass.3s
araje
PRT
tin
the
eksetasi?
exam
‘Will he pass the exam, I wonder?’
g. English question-modifying ‘so:’
So who came?
I wonder (*so) who came?
h. English tag-questions:
He’ll pass, won’t he?
I wonder if he’ll pass (*won’t he).
It is at best unclear whether the idiosyncratic syntactic restrictions that
the analysis of such phenomena require are the kind that would be naturally
extended to capturing the matrix-only nature of fragments. Since there is to
my knowledge no overall understanding of what, if any, commonalities such
phenomena have, there is no good way to know whether the ‘fragment’ re-
striction falls into a natural class with them.
8
The second major technical question involves the movement involved.
Recall that the movement was entirely motivated by the theory-internal de-
cision to elide only constituents (on a par with VP-ellipsis and sluicing)—
maintaining this claim necessitated moving the remnant to a clause-external
8
Particularly interesting in this regard is the case of imperatives, which have no good se-
mantic reason for not being embeddable under predicates like ‘command’ or ‘order’ (and in
some languages—Spanish, Greek, etc.—imperatives can’t even be embedded under negation).
Most analyses of imperatives take the problem to be one of the morphosyntax; an approach
that might link the ‘fragments’ restriction with that on imperatives might seek an answer in
their illocutionary force instead, in particular in its syntactic realization (see for example Speas
and Tenny 2003).