<< >> TOC
The data shown in (19) under 4.2 suggest that
the traditional distinction of 'external' and 'internal' arguments may play a
role in the description of scrambling phenomena. An external argument is usually
considered to be marked as such in the lexicon and syntactically licensed by a
functional head (F in Haider's model, I
in Williams 1994). Grimshaw (1990) proposes that instead of arbitrary marking
(e.g. as 'designated' in Haider's framework), prominence relations within the
A-structure determine the external argument, i.e. the argument that is both
highest in the thematic hierarchy and leftmost
in the logical representation (e.g. the causative agent) will be considered
Whereas in English, external arguments are
always saturated 'externally', i.e. outside of the VP, the same is not the case
for German, i.e. there is no evidence that a typical German subject is not
inside VP. The relation towards VP may thus not be a sufficient measure to
pick out a specific argument as 'external'.
As (22) shows, case (nominative) does not in itself provide a diagnostic
for externality either.
Bayer/Kornfilt propose a mechanism of
"complex category formation" that may provide such a diagnostic. Their
analysis rests on the assumption that there is no principle which would demand
that feature clusters have to be saturated in all languages in the same way.
English may require that certain features (e.g. tense, agreement etc.) enter the
syntactic derivation independently from the verbal features, but German may
allow/require that the feature clusters expressed in V and I in English are
saturated en bloc. BK capture this by
positing a "complex category formation" process, which amounts to
merging V and I at the beginning of the derivation to form a complex category [ V
/ I ]. It may now be possible that in spite of being part of the same
complex category, I and V nevertheless project independently in such a way that
all internal arguments are simple adjuncts to VP, whereas the external argument
additionally projects I' to IP.
Now, the A-structure of the idiomatic
expressions den MannACC
der SchlagNOM treffen
(the object is exchangable) would differ from e.g. die
PrinzessinNOM den PrinzenACC
heiratet in the following way:
[x TREFFEN y]
[ x HEIRATEN y ]
According to Grimshaw's (1990) prominence
theory, x (der Schlag) in (31a) should be the external argument, just like x
(die Prinzessin) in (31b): both
may be considered AGENTs and are most prominent in the logical form.
Syntactically, however, derSchlag differs from die
Prinzessin, insofar as it cannot be fronted with the verb. The only
difference in (31) between der Schlag
and die Prinzessin is that l-conversion
would apply to die Prinzessin first
and to der Schlag last. It could thus
be that the prominence effect is simply a side effect of a more basic mechanism:
the position of the l-operator
determines externality. Since l-position
coincides with prominence for most verbs, prominence theory would thus usually
make correct predictions, although on the basis of a slightly wrong assumption.
For (31b), x is the external argument projecting the I-system, whereas y
is the external argument for (31a). The sentences (22a) and (16a), repeated here
as (32a) and (b) would thus have the structure in (32c) and (d) respectively:
daß den Mann der Schlag getroffen hat.
..., daß die Prinzessin den Prinzen geheiratet hat.
Since both den Mann and die Prinzessin
are only licensed as occupants of the unique [Spec, IP]-position, they cannot be
fronted as members of VP. This explanation works under the assumption that IP is
needed in situ to enter a bidirectional licensing relationship with F, which
means that IP cannot be fronted. V2-fronting thus involves 'disentangling' the
VP/IP-complex, i.e. 'extraction' of VP.
B/K's formulation of complex category
formation is an attempt to make the proposal conform to standard 2-dimensional
notation. To better visualize the processes involved, it may be preferable to
think of complex categories in 3-dimensional terms, as sketched in (33):
Such a 3-dimensional representation would
allow for structural Gestal switches: with respect to its formal properties, I
would be the head of the clause, whereas V functions as a semantic head.Or,
in other conceptual terms, sentence representations have several tiers,
including but not limited to: one formal (I-system), one semantic (V-system),
and one morphophonological tier (the overt lexical expressions associated with
each terminal node). Such a tier analysis
would have one advantage over B/K's model: extraction of VP does not infringe on
the integrity of the I-system, i.e. no complex process of 'disentangling' is
<< >> TOC
© Philipp Strazny 1997
 Bayer/Kornfilt attribute the first formulation of such a view Abney (1986). Proposals arguing for syntactic Gestalt switches include Chomsky (1995a) arguing that heads are both minimal and maximal, and Pesetsky (1994), who proposes that double object constructions in English are organized in both layered (tertiary branching) and cascade (binary branching) structures.
 This is not to argue that morphological features would have to be realized on the same tier as phonological features.