position so that what remained could be elided as a phrase. The difficulty
here comes from mismatches between the kinds of leftward movements seen
in nonelliptical structures and the ones needed to make this account work. The
asymmetry can be seen in the oddity of the following examples compared to
their putatively elliptical descendants in (40) and (41).
(48)
a. On the stoop it is!
b. Moving pretty fast that is!
c. From Spain it is!
d. Driven exactly 10,000km it’s been.
e. Recommended for ages 6 and older it is.
f. One of Anthony Carroll’s best men he is.
(49)
a. Barbara Partee that is.
b. Rob’s mom it is.
c. Nova Scotia it is.
d. The ngo-gyin, the song of mourning it is.
Stainton correctly points out that these sound like ‘Yoda-speak’ (p. 141).
One response to this objection would simply be to drop the requirement that
only constituents be elided: all we’d need is a different implementation of the
syntax-semantics interface requirements for ellipsis. (Though Stainton rightly
objects that this would make the putative ellipsis less like better understood
elliptical phenomena; the question is of course how serious we would like to
take this nonparallelism—one can point to other ‘elliptical’ phenomena such
as gapping and possibly Right Node Raising which have very different prop-
erties from sluicing and VP-ellipsis.) A second response would be to deny the
relevance of the status of the above moved examples to (40) and (41) at all:
after all, the argument could run, there are lots of differences in acceptability
between apparently optional versions of the same structures (#There are some
men tall
vs. Some men are tall; I saw the book vs. ?The book was seen by me,
etc.). In the area of ellipsis, in fact, it has been claimed that some movements
necessarily feed ellipsis. This claim is best known applied to pseudogapping
(see Lasnik 1999, Merchant 2008a), where the remnant movement must co-
occur with VP-ellipsis, but it has also been claimed for the obligatory pres-
ence of VP-ellipsis in subject-aux-inverted comparatives (Merchant 2003, but
see Culicover and Winkler 2007 for a more refined picture).
9
3.2.3.
Final problems with syntactic ellipsis
Even if all the above is correct, and one wishes to accept that there is syn-
tactic ellipsis for the above cases, there remains a subset of examples that
are problematic (I set aside the ‘ordering’ examples until section 5). The first
problematic example is the following, adapted from Stainton 2006:107:
Hans and Franz are playing a boring game one day in which each person
takes turns naming an object which reminds him of a particular person. Their
conversation consists of sentences such as
(50)
Die
the
Lampe
lamp.
NOM
erinnert
reminds
mich
me
an
on
meinen
my.
ACC
Onkel
uncle
Wolfram.
Wolfram
‘The lamp reminds me of my uncle Wolfram.’
They go their separate ways and a few days later, Hans is sitting in a bar when
Franz walks in the door. Hans points at a nearby beer-stained old wooden
table and says,
(51)
Mein
my.
NOM
Vater!
father
(‘My father!’)
In the same context, it would be odd to say either (52a) or (52b):
(52)
a. Das
that
ist
is
mein
my.
NOM
Vater!
father
‘That is my father!’
9
Stainton 2006:140 also mentions the utterance of ‘Moronic jerk!’ at a passing motorist as
unassimilable to *
moronic jerk, given the lack of an article. I think that the key to
understanding such examples is to realize that they occur in a ‘calling’ function (as Stainton
mentions in his fn. 17 on p. 140), which requires the vocative case in many languages. As
usual, Greek is particularly illuminating, since it always requires a definite article with names
used as arguments, but disallows an article when the name is used in the vocative (and it
shows a morphological difference): to call to Alexandros, one says, Alexandre! (vocative), not
o Alexandros
(nominative); to call someone a jerk, one says, Vre iliTie! (vre = vocative particle
indicating impatience) where iliTie is in the vocative, not the nominative iliTios.
b. Meinen
my.
ACC
Vater!
father
‘My father!’
Stainton raises this example as a failure of connectivity, since in German,
the object of the preposition an which is required by the verb erinnern (‘re-
mind’) must appear in the accusative case. Nonetheless, (52b) is impossible
in this situation: instead, we find the nominative as in (51). Stainton posits
that what can be asserted by (51) in this context is that same as what (53)
would assert in this context (or more strictly speaking, the speaker asserts
THAT REMINDS ME OF MY FATHER):
(53)
Das
that.
NOM
erinnert
reminds
mich
me
an
on
meinen
my.
ACC
Vater.
father
‘That reminds me of my father.’
Stainton points out that if the asserted content is generated by German
words corresponding to those appearing in (53), we would expect (52b) to
be fine, and (51) to be odd. I agree that this example is challenging for the
syntactic ellipsis account, but not for the reason stated. The limited syntac-
tic ellipsis account does not suppose that the asserted content derives from
German words—syntactic connectivity effects are predicted only in short an-
swers and the like, where the ‘fragment’ is based on a structure with an acces-
sible linguistic antecedent. It is reasonable to suppose that such accessibility
to linguistic structure in the above game has eroded over the intervening days
(in fact, linguistic cues erode much more quickly than that, generally within
a clause or two, as the psycholinguistic literature explores; see Arregui et
al. 2006). So the puzzle is not that the accusative is unacceptable here (which
would also violate the ban on P-stranding in German); the puzzle is why the
nominative can be used, given the oddity of (52a).
10
The problem is that in German, the demonstrative das appears to be less
useable for abstract properties of individuals in copular sentences than that
10
My own investigations with German speakers has led me to believe that the empirical
situation is somewhat more complicated than this: for many speakers, (51) and (52a) have
approximately the same degree of felicity in the given situation (some report low felicity,
others higher, but with no intraspeaker variation). Obviously, for such speakers, there is no
problem to be addressed. But for the sake of the discussion, I concentrate on speakers that
share the judgments Stainton reports.