The data presented under 4.2 shows that
objects, but not subjects, can be fronted along with the verb, which is true for
the vast majority of German verbs. Nevertheless, there are verbs that exhibit
opposite behaviour, as noted in Haider (1993: 197) and shown in (22):
(22)
a.
..., daß den Mann der
Schlag getroffen hat.
that
the manACC the strokeNOM hit
has.
"The man had a stroke".
b.
?..., daß der Schlag den Mann getroffen hat.
c.
Den Mann hat der Schlag getroffen.
d.
Der Schlag hat den Mann getroffen.
e.
*Den Mann getroffen hat der Schlag.
f.
Der Schlag getroffen hat den Mann.
Here, the subject can be fronted with the verb
(f), but the object cannot (e). Moreover, there seems to be a strong adjacency
requirement for subject and verb, which even works against scrambling inside an
embedded clause (b). (f) directly falsifies the hypothesis that only objects can
be fronted along with the verb. However, the ungrammaticality of (e) suggests
that one may still be able to draw a generalization, if formulated as "for
each verb, the argument that appears leftmost in the default word order cannot
undergo V2 in conjunction with the verb".
The question now is how the generalization
above may be captured in formal terms. One could assume that a verb arbitrarily
assigns a "don't-touch-me"-feature to one of its arguments. This
feature would then block movement of the so marked argument along with the verb.
Obviously, this strategy is far from ideal, since, ad hoc assumptions would be
needed to restrict movement to maximal projections that do not
contain an argument carrying such a feature. Since it is commonly assumed
that syntactic mechanisms are blind as to the internal structure of the phrases
they apply to, this would clearly bring an unwanted turn to the theory.
Furthermore, methodological considerations make it preferable to base
restrictions on structural operations like movement on structural conditions.
It is hard to see how the VP-structure
proposed by Frey/Tappe (1992) and Haider (1993) would by itself allow to make a structural
distinction among the different arguments: after all, all arguments are seen as
adjuncts to VP that do not take the verbal projection to a higher level. Under
5.2, I will thus attempt to embellish the theory with a proposal made by
Bayer/Kornfilt (1994) to account for the grammaticality differences in (22).
© Philipp Strazny 1997