The effect of the directionality constraint is that in many cases the only constraint imposed on word order is that the head must be final. That is, in a clause containing one or more argument-inheritance verbs, and where the finite verb occurs in final position, all complements are [DIR ]. These complements can therefore occur in any order. While the order of NP's, PP's and adverbials in the German Mittelfeld allows for a certain amount of variation, it is definitely not the case that verbal complements of argument-inheritance verbs and non-verbal complements can be scrambled. Therefore, additional constraints need to be imposed.
In Kathol's topological field approach, verbal complements of argument-inheritance verbs occupy the vc field, whereas non-verbal complements normally occupy the mf field. As mf elements must precede vc elements, this has the effect of ordering verbal complements closer to the head than non-verbal complements.
We adopt a solution that is similar to the idea of having topological fields. As we need to distinguish only between elements that are in the `verb cluster' and elements outside it, and as we do not want to make reference in our general HEAD-COMPLEMENT schema to concepts that are relevant in clausal syntax only (such as verb cluster), we introduce a slightly different terminology. The feature ZONE with values i(zone) and o(zone) distinguishes complements that appear close to the head from those that are further away:
The izone corresponds to Kathol's vc topological field.
Since we assume that the head of a phrase is always part of the izone, we also modify our HEAD-COMPLEMENT schema slightly:
The interaction of the Zone and Directionality constraints is represented schematically in (23).
One of the most important reasons for giving up the idea that there is internal structure in the verb cluster, is the fact that it allows an easy formulation of the constraint that verb clusters may not be interrupted by non-verbal constituents. In particular, in our account of Dutch, we assumed that all (obligatory) argument-inheritance verbs require their verbal complements to be of type word. Now that we have introduced a feature ZONE which distinguishes verbal complements that must be part of the `verb cluster' from other complements, we no longer need to state this constraint in the lexicon. Instead, we assume that the following constraint applies to the HEAD-COMPLEMENT schema:
Note that since the head of a HEAD-COMPLEMENT schema is izone, we can in fact remove the constraint that the head must be of type word from our revised HEAD-COMPLEMENT schema (22). More importantly, we have argued that the introduction of the Izone principle in fact leads to an improved account of partial VP fronting .