The most important innovation of our analysis is the fact that we assume that only one rule is needed to derive subordinate clauses. The rule is an instance of the head-complement-schema given in (12).
This schema inherits from the rule schema given in (8). Therefore, the head-feature principle, subcategorization principle, nonlocal feature principle, and the directionality principle hold for this rule. The subcategorization principle together with the information in (12) implies that the non-head daughters that are selected by the head in this rule form a suffix of the SUBCAT list of the head. The remaining SUBCAT elements of the head form the subcategorization requirements of the mother node. Thus, this schema can be used to form both partial and saturated verb-phrases. In the first case only a subset of the SUBCAT elements are selected (and hence the SUBCAT list of the mother is non-empty). In the latter case all SUBCAT elements are selected (and hence the SUBCAT list of the mother is empty).
The HEAD-COMPLEMENT schema assigns a ``flat'' clausal structure to phrases containing a clause-final verb sequence. As an example we give a derivation for (13) in (14).
This example can best be understood by looking at the verb lezen
first. This ordinary transitive verb subcategorizes for an
(accusative) noun phrase NP3. The verb laten is a typical
verb-raiser. It selects a verb, all subcat elements of this verb (in
this case NP3) and an accusative noun phrase of its own, NP2. This verb in turn is selected by the finite modal
wil. This modal subcategorizes for laten, all subcat elements of
laten and a nominative subject NP1. Because wil is the
head of the rule it is the subcat list of wil that will effectively
be used by the rule.
The HEAD-COMPLEMENT schema does not require that all complements must be selected. No ``spurious ambiguity'' results from this liberal formulation, as (12) requires that its head must be lexical (i.e., of type word), whereas the mother is non-lexical (i.e., of type phrase). This prevents a complex phrase licensed by (12) to be the head of another phrase licensed by (12), and thus ``contoured'' VP's cannot be derived.
This should be compared with analyses in which a verb cluster is
assumed, such as the analysis of [8]. As
explained in the introduction, in such analyses it is difficult to
account for the fact that raising is obligatory; those accounts need
ad-hoc feature-passing mechanisms.
By not imposing any constraint on the length of SUBCAT of the mother in (12), we do allow unsaturated, partial, VP's. Such phrases do play a role in the discussion below.
Since we assume that subjects are part of the SUBCAT list of finite verbs, there is no need for a separate schema to combine a head with a subject.
