In the literature, (8) is repeatedly quoted to
exemplify "Webelhuth's paradox", i.e. the purported ability of
scrambled elements to license parasitic gaps (characteristic of A-bar movement)
as well as to function as a binder (an A-movement effect):
(8)
..., weil
Fritz jeden Gast
[ ohne ei anzuschauen ] seinemi
because
Fritz every guestACC without
to-look-at
his
Nachbarn ti vorgestellt
hat.
neighborDAT introduced-to
has.
(example from Webelhuth: 412)
Whereas it is clear that jeden Gast is able to license the parasitic gap, it remains doubtful
whether it being in a scrambling position (i.e. probably a VP-adjoined position,
cp. Müller and Sternefeld 1994) is solely responsible for this ability. Whereas
(9) may still be interpreted as a case of vacuous scrambling by assuming a trace
t, (10) clearly shows all arguments in
their base position, but the 'parasitic gap' is still licensed:
(9)
..., weil
Fritz das Buchi (ti)
[ ohne ei anzuschauen ] weggelegt hat.
because
Fritz the bookACC without
to-look-at
put-away has.
(10) ...,
weil Fritz
dem Verkäufer das Buch
[ ohne ei anzuschauen ]
because
Fritz the salesmanDAT the bookACC without to-look-at
überreicht hat.
given-to has.
If it is maintained that 'parasitic gaps' can
only be licenced by moved elements in A-bar positions, one would have to assume
two separate vacuous scrambling movements for (10), as shown in (11), for which
one would be hard pressed to find a reasonable motivation.
(11) ...
dem Verkäuferi das Buchj ti tj [
ohne ej anzuschauen ] ...
It may be preferable to set 'parasitic gaps'
aside for now, instead of using such epiphenomena (they are not acceptable to
all native speakers to the same degree) to bolster assumptions about an issue as
central as scrambling, i.e. VP-structure.
A data set commonly mentioned to support an
A-movement analysis concerns binding:
(12)
a.
..., weil wir die Fraueni
einanderi ti
vorgestellt haben.
because we the women each
other introduced have.
b.
..., weil jedeni
seinei Mutter ti mag
because everyoneACC his
motherNOM
likes
c.
*..., weil einanderi
die Fraueni ti nicht mögen
because each otherACC
the womenNOM not
like
(examples from Corver/Riemsdijk 1994: 7)
Scrambling can create 'new' binding
relationships (12a) without causing weak crossover effects (12b), which suggest
that the landing site is an A-position. Moreover, reconstruction seems to be
blocked (12c) (cp. Corver/Riemsdijk: 7).
However, as Corver/Riemsdijk (p. 8) point out,
"there is not always consensus on the facts", and that is putting it
mildly. In contrast to cases of clear violations of (possibly universal)
syntactic principles like the ECP, analysts seem in general hesitant to declare
a given scrambling structure ungrammatical (*). Rather, the overwhelming amount
of "bad" data is marked in degrees of unacceptability (from '?*' over
'??' to '?'). This may exemplify
that scrambling structures are interdependent with so many other phenomena
(binding, definiteness / indefiniteness, specificity, intonation etc.), that
even for native speakers it is at times hard to trust their own intuitions. To
give an example, sentence (12c) becomes much better if it is supplied with the
proper intonation and, to my ears, it is perfectly acceptable if the simple
definite is changed to a demonstrative article:
(13) a.
..., weil einanderi DIE Fraueni ti NICHT
mögen.
b.
..., weil einanderi DIESE Fraueni ti NICHT
mögen.
If thus (12c) should be truly unacceptable for
some speakers (it is not to me), the reason for this may be found in the fact
that scrambling is not movement-related (cp. 5.1), but controlled by other,
weaker factors.
For the rest of this discussion, I will
restrict the data to the simplest constructions possible. Quantification,
binding, negation etc. certainly interact with scrambling structures and will
have side effects on acceptability judgments. In order to exclude such
interferences, I will use only "bare" structures without adjuncts[1],
particles, or combinations of definites and indefinites.
© Philipp Strazny 1997
[1] It is not clear how adjuncts interact with scrambling. It may be possible that adjuncts are generated in specific positions, so that arguments have to be "scrambled around" an adjunct. However, it may just be as possible that adjunction does not have any bearing on scrambling and that restrictions apply to the place of adjunct insertion. Consider the following structures:
..., daß gestern die Prinzessin dem Prinzen das Ja-Wort gegeben hat.
that yesterday the princess the prince the Yes-word given has
..., daß die Prinzessin gestern dem Prinzen das Ja-Wort gegeben hat.
..., daß die Prinzessin dem Prinzen gestern das Ja-Wort gegeben hat.
..., daß die Prinzessin dem Prinzen das Ja-Wort gestern gegeben hat.
..., daß die Prinzessin dem Prinzen das Ja-Wort gegeben hat gestern.
I doubt that the focus potential of the individual constructions (following Höhle 1982) will yield a default adjunct position with any significant degree of certainty.