There is another set of examples in which partial VP's occur. Examples of Dutch partial VP topicalization are (sometimes only marginally) acceptable:
Our analysis of such partial VP topicalizations is based on the analysis of [21] for German, but improves upon that analysis because we do not need a special rule to build partial verb-phrases in topic position.
Following Nerbonne's analysis, we assume a lexical complement extraction rule for verbs, which moves a complement from SUBCAT to SLASH. The rule differs from the standard complement extraction rule in that the extracted complement and the element added on SLASH only structure-share the value of their features (HEAD and SUBCAT in particular):
![\enumsentence{
\par$
\begin{array}[t]{c}
\begin{avm}
\begin{displaymath}head & v...
...end{displaymath}} \>
$~\bigoplus$ \@5 \end{displaymath}\end{avm}\end{array}$}](img28.png)
Thus, if a verb subcategorizes for a complement of type word, this requirement will not be present once the element has been moved to SLASH_INHER. Consequently, a verb-raiser which ordinarily requires that its argument is lexical (i.e., of type word) may allow for a non-lexical verbal projection (of type phrase) in topic position. Everything else remains the same: as before, unexpressed SUBCAT elements are inherited by the verb-raiser.
Apart from this lexical rule [21] needs a special mechanism to construct partial VP's. In order to prevent spurious ambiguities (of the kind discussed in [22]) this rule is supposed to build such VP's only in topic position. For this reason the binary feature FOCUS is introduced. In the current analysis this gymnastics is not needed at all: our Rule 1 introduced in (12) also creates partial VP's. On the other hand, we do not get spurious ambiguities because of the lexicality specifications on the head and the mother of Rule 1. Consider for example the relevant lexical specification of the verb heeft in (27a) (leaving the adjunct out of the discussion -- for an analysis of adjuncts that is compatible with the analysis given here, see [29]). This specification (30) is output of the lexical rule given above and is produced on the basis of (29).
The derivation of (27a) is illustrated in (31). It is assumed here that verb-fronting is analysed by a rule that is also an instance of the head-complement-schema (12), in which word-order is the same as in rule 1 except that the finite verb precedes everything else. Furthermore the topic is selected by a filler-gap rule of the kind that is standard in HPSG.
This section illustrated how the ``flat'' analysis is compatible with the occurrence of partial VP's. We have shown that the fact that partial VP's can be topicalized and extraposed need not imply that such partial VP's occur in the verb-cluster. Furthermore such partial VP's do not need to give rise to the spurious ambiguities of [22], and neither require the ad-hoc rule of [21].
![\eenumsentence{
\item[a.]
\shortex{5}
{ [$_{pvp}$\ bezocht] & heeft & Jan & dat...
... that conference & never}
{\em Jan has never wanted to visit that conference}
}](img27.png)
![\enumsentence{
heeft $\mapsto$\begin{avm}
\begin{displaymath}
head & verb$[$fin$...
...erb$[$prt$]$\\
subcat & \@3 \end{displaymath}} \>
\end{displaymath}\end{avm}}](img30.png)
