While early transformational accounts of the Dutch cross-serial dependency construction assumed that the clause-initial complements of the verbs present in the verb-cluster are sisters (after a ``pruning'' transformation had eliminated VP-nodes whose head had been V-raised to a clause-final position ()), non-transformational accounts have assumed that the initial part of the clause, preceding the verb cluster, has considerable internal structure. However, there is little theory-independent motivation for postulating the existence of several verbal projections dominating the non-verbal arguments in the clause.
For German, it has been argued that the phenomenon of ``partial fronting'' (6) requires the existence of partial (or ``contoured'') VP's in the ``Mittelfeld''.
Given the un-controversial assumption that the fronted material in (6a,b) is a constituent, it seems at first sight that at least two different ways of analyzing (6c) must exist, even-though it is semantically unambiguous. Accounts of partial VP-fronting using contoured VP's ([20,12]) have indeed made this assumption. The disadvantage of such an analysis is not only that it introduces spurious ambiguity, but also that it makes crucial use of this fact to account for the data.
An interesting alternative is provided in , who argues that the fact that partial VP's can be fronted not necessarily implies that such constituents exist within the ``Mittelfeld''. His analysis assumes that the normal clause structure contains a ``flat'' VP and that partial VP's can only be derived in fronted position. This eliminates the spurious ambiguity problem and furthermore has the advantage that one does not have to motivate the existence of partial VP's in the Mittelfeld.