next up previous
Next: Partial VP fronting Up: Partial VPs Previous: Partial VPs

Subsections


Extraposition and partial extraposition

Extraposition

Some verbs subcategorize for a saturated verbal complement (sometimes marked with the infinitival complementizer om), as the following examples illustrate.

\eenumsentence{
\item \shortex{7}
{dat & Jan & heeft & geprobeerd & [(om) & het...
...pl}) & the book
& to read]}
{\em that Jan has forced Piet to read the book}
}

It is not entirely clear whether we should analyse such examples as involving extraposition (the standard transformational point of view), or rather as straightforward selection to the right ([9]). Here we will assume that such verbal constituents to the right of the verb-cluster are simply analysed as arguments that need to be selected to the right, using the feature DIR introduced earlier.

Verbs that only allow extraposition of saturated verb-phrases (such as the verbs beloven (to promise) en dreigen (to threaten) in their non-modal reading) will not inherit the arguments of their verbal argument, but simply require that the verbal argument is saturated:

\enumsentence{
beloven $\mapsto$\begin{avm}
\begin{displaymath}head & verb\\
...
...subcat & \<\> \\
dir & right \end{displaymath}\>
\end{displaymath}\end{avm}}
The word om is analysed as a verbal element (with VFORM om) that subcategorizes for a saturated verb-phrase with VFORM te:

\enumsentence{
om $\mapsto$\begin{avm}
\begin{displaymath}
head & verb$[$om$]$\\...
...bcat & \<\> \\
dir & right \end{displaymath} \>
\end{displaymath}
\end{avm}}

The disjunctive specification of the VFORM feature of the verbal argument thus accounts for the optionality of om. Verbs which do subcategorize for a VP, but do not allow this complement to be headed by om (such as menen (to believe/suppose)), simply select a complement whose head is verb[te].

The main difference between extraposition and raising constructions is that extraposition verbs require that their verbal argument is saturated. A verb raiser, on the other hand, requires that its verbal argument is lexical--and inherits the arguments of this verbal argument.

Another difference between the verb raising construction and the extraposition construction is the infinitivus pro participio (IPP) effect. The verb hebben ordinarily requires that its verbal argument is headed by a participle. However, if this verbal argument is a verb-raiser then an infinitive is required:

\eenumsentence{
\item
\hspace{8pt}\shortex{6}
{ dat & Jan & de vrouwen & heeft &...
...he women & has & seen ({\sc prt}) & sleep}
{\em that Jan saw the women sleep}
}

The IPP effect can be used to distinguish between the verb-raising and the extraposition construction.

\eenumsentence{
\item
\shortex{6}
{dat & Jan & Piet & heeft & zien & slapen}
{t...
...et & has & forced ({\sc prt}) & to sleep}
{\em that Jan forced Piet to sleep}
}

Partial Extraposition

The interesting thing about extraposition in relation with verb clusters is the occurrence of partial verb-phrases that seem to mix the properties of ordinary extraposition on the one hand and verb-raising on the other hand ([6]). Thus we get sentences such as:

\eenumsentence{
\item
\shortex{3}
{ dat & Jan {\em de trainer van zijn gelijk} ...
...m dat Jan {\em de trainer} heeft gemeend {\em van zijn
gelijk te overtuigen}
}

Verbs that allow partial extraposition mix properties of ordinary extraposition verbs with the properties of raising verbs. On the one hand, partial extraposition verbs also inherit the remaining arguments on the SUBCAT list of a verbal argument that is selected to the right. But unlike raising verbs this argument need not be lexical. Furthermore the VFORM om (cf. below) is never allowed in case arguments are inherited. Finally the IPP effect does not occur in partial extraposition constructions.

In our analysis partial extraposition verbs do not specify a lexicality requirement on their verbal argument. Furthermore, unexpressed elements from the SUBCAT list are inherited by the head. Thus partial extraposition verbs are similar to verb-raising verbs in the sense that they inherit subcategorization information from their verbal argument. For example, the participle gemeend will have the lexical specification given in (22). The derivation for  (21c) is given in (23).

\enumsentence{
gemeend $\mapsto$\begin{avm}
\begin{displaymath}head & verb$[$p...
...bcat & \@1 \\
dir & right
\end{displaymath}} \>
\end{displaymath}\end{avm}}

\enumsentence{
}
\begin{center}
\leavevmode
\unitlength1pt
%% picture(310.16,139...
...(281.68,10.06){\hbox{}}
\put(0.00,120.00){\hbox{}}
\par\endpicture
\end{center}

Our analysis correctly predicts that om can never occur with partial extraposition. As indicated above, the complementizer om only combines with a saturated verb-phrase:

\eenumsentence{
\item \hspace{8pt}\shortex{6}
{ dat & Jan & de vrouwen & heeft &...
...te kussen
\item
\hspace{8pt}dat Jan heeft geprobeerd om de vrouwen te kussen
}

Note that the following sentence is ungrammatical:

\enumsentence{
$^*$\ dat Jan {\em van zijn gelijk} heeft gemeend {\em de trainer te overtuigen}
}

The contrast between this sentence and (21c) is explained by our assumption that only a suffix of the subcategorization elements can be selected by Rule 1 (12). Thus the VP de trainer te overtuigen cannot be built (with the relevant reading of overtuigen).

A difference between partial extraposition and verb-raising is the IPP effect: this effect only occurs with verb-raising constructions but not in (partial) extraposition constructions. There is a class of verbs (including proberen (to try)) that allow both verb-raising and (partial) extraposition. Our analysis predicts that these verbs can occur both in their infinitival form and in their participle form, when they are selected by the auxiliary hebben. This prediction is correct:

\eenumsentence{
\item \hspace{8pt}\shortex{6}
{ dat & Jan & de vrouwen & heeft &...
...({\sc prt}) & the women & to kiss}
{\em that Jan has tried to kiss the women}
}

The first example is the verb-raising variant of probeer, the second example is the partial extraposition variant of probeer in which the extraposed argument happens to be lexical too. The third example shows that verb-raising is obligatory if the IPP effect is present. The subcategorization requirements of hebben are discussed in section 6.


next up previous
Next: Partial VP fronting Up: Partial VPs Previous: Partial VPs
Noord G.J.M. van
1998-09-29