Both verbs and verb phrases are of type v:
The features LEX, NULL, and VFORM are specific for v. The feature VFORM is used to distinguish finite, infinitive, te-infinitive and past participle verbs (and verb phrases headed by such verbs). The feature LEX is used to distinguish lexical verbs ( ylex) from verbal phrases that are not lexical ( nlex). The feature ylex subsumes two further subtypes basic complex, to distinguish basic and complex lexical verbs. The latter are combinations of a verb and a separable prefix ( aan+komen, arrive) or combinations of a modal verb and a main verb ( wil vertrekken, want to leave). The feature NULL is used to distinguish verbal traces (i.e. verbal signs without phonological content) from other verbal signs. The features SUBJ, SLASH, and VSLASH and NULL are discussed in the section below on sentential syntax.
There are a number of similar rules for combining a verb or a verbal
projection with one of its complements. One rule combines a noun
phrase complement with a verbal head ( een kaartje kopen, buy a ticket):
A verbal modifier can be either an adverb, a PP, or a temporal NP. There are unary rules rewriting signs
of type modifier into each of these categories. One such rule is the
At the moment, we allow all modifiers to precede or follow the verb ( ik moet morgen in Assen zijn/ in Assen zijn morgen/ morgen zijn in Assen, I must be in Assen tomorrow, ik moet tien uur in Assen zijn/?in Assen zijn tien uur, I must be in Assen at ten o'clock). Therefore, there are two similar rules, vp_v_mod and vp_mod_v, in which a verb combines with a modifier. The first is illustrated here:
A special type modifier (with SEM and MOD as only attributes) in combination with three unary rules is used to introduce the various types of verbal modifier. A sample derivation is given in figure 6 (the value of the features SLASH and VSLASH is not shown, but is on all verbal signs in this derivation).
Finally, there are two VP-rules that give rise to `complex'
lexical expressions, instead of phrases. Firstly, consider the
This allows us to derive phrases such as (dat ik) een kaartje wil kopen (that I want to buy a ticket) where the finite modal verb combines with the infinitival verb before combining with the object of kopen (figure 7). Note that it is essential that the modal verb selects a [LEX YLEX] argument in this case, as this excludes the derivation of ungrammatical expressions such as (dat ik) wil een kaartje kopen. The result of combining a modal with an infinitival verb is [LEX COMPLEX] (i.e. subsumed by [LEX YLEX]). This implies that such combinations can be selected by another modal verb (i.e. (dat ik) een kaartje zou willen kopen, that I would like to buy a ticket).
Next, consider the v_part_v rule:
It should be obvious that these two rules, and the limited form of argument inheritance we allow (i.e. structure sharing of SC-lists only, and no concatenation of SC-lists), is not sufficient to account for the full range of verb clustering data in Dutch. For one thing, the grammar as it stands cannot handle `inverted' word orders ( (dat ik de trein) halen moet, that I must catch the train), where the infinitive precedes the modal verb. It is rather straightforward to include rules for inverted word orders. A potentially more problematic omission is the fact that perception verbs ( horen, zien) and causative laten, which also introduce verb clusters ( (dat ik) Rob een kaartje laat kopen, that I let Rob buy a ticket), cannot be accounted for. The analysis of this construction in van Noord and Bouma  is based on the notion `argument inheritance'. This presupposes the possibility of recursive constraints in syntax (to concatenate SC-lists) as well as rules with an indefinite number of daughters. Both are excluded within the present formalism.