There is a broad concensus among researchers working within the paradigm of HPSG that complement inheritance of the kind proposed in Hinrichs and Nakazawa [3,4] is an essential operation in the analysis of the verb cluster in German and Dutch. Both languages have a class of verbs (including auxiliaries and modals) that subcategorize for a (possibly) unsaturated verbal complement, and for all the complements on the COMPS-list of this verbal complement. Most analyses of German have assumed that these complement inheritance verbs combine with their verbal complements to form a phrase consisting of (lexical) verbs only. This phrase is usually referred to as the verbal complex.
In this paper, we argue that the word order of German as well as Dutch verb clusters can be accounted for without introducing a verbal complex.1Our analysis rests on the assumption that a single HEAD-COMPLEMENT schema exists, which licences phrases consisting of a lexical head and an arbitrary number of its complements. This schema allows a complement inheritance verb to combine with its verbal complement, as well as the complements of this complement, in one step. A consequence of this analysis is that there is no room within the verb phrase for partial VPs or a verbal complex. The advantage of such an account is that there is no need to distinguish between a rule schema for verbal complexes and for (partial) VPs. Furthermore, a `flat VP' implies that phrase structure does not impose any constraints on word order. Therefore, the full range of word order possibilities found in German and Dutch verb clusters is captured by a single schema. Of course, the main challenge for this `flat VP' analysis is to demonstrate that it can do so without leading to vast overgeneration. This is the main topic of the current paper.
In the next section, we introduce the German data and discuss the analysis of Hinrichs and Nakazawa [3,4] as well as a number of related approaches. Next, we present our analysis of German. It uses a general HEAD-COMPLEMENT schema in conjunction with linear precedence statements. We demonstrate that the proposed set of LP-statements accounts for all ordering possibilities encountered in the German verb cluster. Furthermore, we argue that our analysis leads to an improved account of partial VP fronting .
The account to word order adopted here, is considerably more sophisticated than the proposal in , in which an account of the Dutch verb cluster was presented which relied primarily on ordering in terms of obliqueness. In section 4, we demonstrate that the improvements that were necessary in order to account for German also lead to a smoother account of some of the more problematic Dutch data.