Main clauses

Main clauses with a finite verb in initial position (as in yes/no-questions) are of type ques. Main clauses in which the finite verb appears in second position (as in declarative sentences or WH-questions) are of type root. The attributes associated with these types are:

$$\small\begin{minipage}[t]{.9\textwidth} \begin{avm} \begin{d... ...{\it root}\\ sem & Qlf \end{displaymath}\end{avm}\end{minipage}$$ (24)

Dutch main clauses differ from subordinate clauses in that the finite verb in main clauses appears in first or second position. There is a tradition, both in transformational and non-transformational grammar, to account for this fact by postulating a dependency between the finite verb and the position where finite verbs occur in subordinate clauses. The advantage of postulating such a dependency is that the grammar rules used for subordinate clauses are also applicable in main clauses. In transformational grammar, a dependency of this type can be established by means of a head-movement operation which moves the verb from its final position to a position at the beginning of the sentence.

Within the framework of HPSG [31,21] we can obtain a similar dependency by postulating a verbal trace, i.e. a verbal sign without phonological content, at the end of the clause. Using this verbal trace as the head, we can use the VP rules discussed above to build up a VP as usual.

The rule for introducing such a verbal trace is given in figure 9. Note that the sign for verbal traces differs from that of an ordinary verb in that its subcategorisation list in not instantiated, but made reentrant with VSLASH:VSC. Similarly, the semantics of the verbal gap is reentrant with VSLASH:VSEM. Furthermore, a verbal gap is a basic (i.e. non-complex) lexical verb, with no phonological content (i.e. [NULL NULL]). We can also safely assume that verbal traces are finite, as main clauses are always headed by a finite verb. The value of SUBJ is the empty list, as VPs headed by a verbal trace never combine with a subject directly (as will be shown below). Finally, SLASH also can be assumed to be empty.³

Figure 9: Verbal Gap

There are two rules which combine a finite verb with a VP containing a verbal trace, and which also introduce a subject (figure 10). Both rules are highly similar (they are therefore both instances of a MAIN-CLAUSE-STRUCT). The only difference is the category of the mother, and the order of the daughters. The vfirst-rule introduces phrases of the type ques, i.e. instances of verb-first clauses, in which the subject follows the main verb. The subject-first-rule introduces phrases of type root, in which the subject is first, and the main verb follows the subject. The constraints imply, among others, that the VP must contain a verbal trace, that the SC-information of the main verb is reentrant with VSLASH:VSC of the VP (and thus, indirectly, with the SC-value of the verbal trace), and that the semantics of main verb is shared with the value of VSLASH:VSEM on the VP (and thus, indirectly, with the semantics of the verbal trace). Note also that the VP acts as semantic head of the construction. This is necessary in order to ensure that the effect of verbal modifiers within the VP is properly taken into account. An example derivation of a subject first main clause is given in figure 11.

Figure 10: Rules for verb-first and subject first main clauses (y/n questions and simple declarative sentences)

Figure 11: Rob koopt geen kaartje (Rob does not buy a ticket)