b. * Hans.
Hans
This is because German, unlike English, does not permit preposition-
stranding under leftwards movement:
(21)
a. * Wem
whom.
DAT
hast
have
du
you
mit
with
gesprochen?
spoken
(lit.) ‘Who did you speak with?’
b. * Hans
Hans
habe
have
ich
I
mit
with
gesprochen.
spoken
(lit.) ‘Hans, I spoke with.’
The correlation between the obligatory presence of the preposition in (19b)
and the necessity of pied-piping the preposition in overt movements such as
in (21) is directly captured on the ellipsis account in Merchant 2004a, where
it’s proposed that the short answer consists of a fragment which had under-
gone a kind of movement to a clause-peripheral position with concomitant
ellipsis of the clausal node under the landing site of the fragment. The result-
ing sentence
syntactic
is given in (22), where F is a functional head which hosts
the E(llipsis) feature which triggers nonpronunciation of its complement, here
the clausal node TP.
(22)
FP
¨
¨
¨
¨
r
r
r
r
Hans
1
¨
¨
¨
r
r
r
F[E]
I spoke with t
1
For such cases, Stainton tends to agree (though he hedges a bit) that ellip-
sis is involved, writing ‘some of [this] data reinforce my standing view that
direct answers to immediately prior interrogatives may well involve genuine
syntactic ellipsis’ (2006:137, with similar remarks on p. 144 and in Stainton
1997).
It’s important to note that to date, no-one has even hinted at how to account
for these facts without using a theory of preposition-stranding, and no-one
has ever proposed a theory of preposition-stranding that distinguishes Ger-
man from English on anything but morphosyntactic grounds. In other words,
whether the grammar of a language makes available preposition-stranding is
an irreducibly syntactic fact about the language, not a semantic one, or a prag-
matic one. (Whether a speaker in a given context will choose to make use of
P-stranding of course is dependent on nonsyntactic factors; but even factors
that favor P-stranding will be powerless in a language like German.) There is
a language-internal effect of connectivity between the grammatical form of
the short answer and some aspect of the form of the question.
Another striking connectivity effect in short answers involves voice: what-
ever voice is used by the questioner must underlie the answer, determining the
form of the short answer. In German, this can be seen in both directions (pas-
sive voice in the question, active in the answer and vice versa); in English, in
only one direction.
(23)
Voice connectivity in short answers
a. English
Q: Who is sending you to Iraq? A: *By Bush.
b. German
i. Q: Wer
who.
NOM
hat
has
den
the
Jungen
boy
untersucht?
examined?
A: * Von
by
einer
a
Psychologin.
psychologist
Q: ‘Who examined the boy? A: [intended:] ‘(He was exam-
ined) by a psychologist.’
ii. Q: Von
by
wem
who.
DAT
wurde
was
der
the
Junge
boy
untersucht?
examined
A: * Eine
a
Psychologin.
psychologist.
NOM
Q: ‘Who was the boy examined by?’ A: [intended:] ‘A psy-
chologist (examined him).’
It is crucial to note that these effects emerge only when ellipsis is involved,
and are not due to more general conditions on felicitous answers or discourse
coherence, as the following control cases demonstrate.
(24)
No voice connectivity in nonelliptical answers
a. English
Q: Who is sending you to Iraq? A: I’m being sent by Bush.
b. German
i. Q: Wer
who.
NOM
hat
has
den
the
Jungen
boy
untersucht?
examined?
A: Er
he
wurde
was
von
by
einer
a
Psychologin
psychologist
untersucht.
examined
Q: ‘Who examined the boy?’ A: ‘He was examined by a psy-
chologist.’
ii. Q: Von
by
wem
who.
DAT
wurde
was
der
the
Junge
boy
untersucht?
examined
A: Eine
a
Psychologin
psychologist.
NOM
hat
examined
ihn
him
untersucht.
Q: ‘Who was the boy examined by?’ A: ‘A psychologist ex-
amined him.’
Such connectivity effects form the best argument that there is syntactic
ellipsis involved in fragment answers, and that the unpronounced syntax must
be identical in some way to the syntax in the question asked. These effects
tell strongly against approaches like Ginzburg and Sag’s 2000 and Culicover
and Jackendoff’s 2005, which posit no syntax internal ellipsis sites at all (see
Merchant 2008b for discussion). While this conclusion continues to appear
to me to be inescapable, there remain certain issues with fragment answers
that have to be addressed on this analysis as well.
The first and in my view most serious issue concerns the lack of island
effects in certain contexts. Apparent movement sensitivities to islands are
variable under ellipsis. So while (25) shows a standard island effect (namely
illicit wh-extraction out of a relative clause), and this effect persists when
VP-ellipsis is applied to the higher VP as in (27), it is famously absent in
an equivalent sluicing case like (26) (see Merchant 2001 for discussion and
references). In Merchant 2008a I propose to capture this distinction by mak-
ing reference to the variable amount of structure elided in the two cases: in
sluicing, more, in VP-ellipsis, less. If the grammatical mechanisms that trig-
ger island deviancies are encoded along the path of extraction (for example,
through the use of illicit intermediate traces, here marked *t), then ellipsis can
variably eliminate these from representation that is pronounced. If a higher
node is elided (as in sluicing), so will all *ts be; while if a lower node is