Postulating a verb-complex requires ad-hoc methods for characterizing what distinguishes a verb-complex from other verbal projections. Postulating partial VP's to account for partial fronting of VP's introduces the spurious ambiguity problem. The alternative we propose below assumes that the constituent structure for the cross-serial dependency construction does not contain a verb-cluster and does not contain partial VP's. Instead, all verbs and all non-verbal complements are directly dominated by S. We demonstrate that this not only accounts for cross-serial word order, but also that it is compatible with the analysis of a number of constructions that appear to be problematic for a flat analysis.
Note that the ``flat'' analysis that we propose is different from analyses based on sequence union ([25,26]) or similar non-concatenative operations ([14]). In those analyses only the derived structure is flat. In contrast, our analysis assumes that both the derived structure and the derivation structure is flat.
After briefly introducing our version of HPSG in the next section, we present the basic rule of the grammar in section 3 and demonstrate how it accounts for cross-serial word order. In section 4 we discuss two constructions in Dutch involving partial VP's. In section 5, we give an account of anaphoric binding that is largely independent of clause structure. In section 6, we demonstrate how verb-sequences with inverted word order and verb-sequences containing particles can be accounted for.