Next: Comparison and Conclusions Up: Word Order Constraints on Previous: Word order within the

Word order in Dutch verb clusters

In this section we revise and extend the analysis of Dutch presented in [18].

As in German, Dutch subordinate clauses are verb-final. If the clause is headed by a modal, an auxiliary, or a verb such as horen (to hear), proberen (to try), helpen (to help) or laten (to let) (these are so-called `verb-raising' verbs), the head of its non-finite VP-complement must occur right of the head of the main clause. This is illustrated in (51-c)a,b. As the head of the non-finite VP can be a verb-raising verb itself, the construction can (in principle) lead to an arbitrary number of crossing dependencies between pre-verbal complements and verbs subcategorizing for these complements. This is illustrated in (51-c)c, where subscripts are used to make the dependencies explicit.

$\ex. \ag. dat Jan het boek {\em wil} lezen\\ that John the book wants read\\ {... ... the book wants let read\\ {\em that John wants to let Mary read the book} \par$

Following our proposals for German in the prevous section, we assume that Dutch `verb-raising' verbs select for a list of complements consisting of a verb and the complements selected by that verb. The word order constraints for complement-inheritance verbs make reference to GVOR. Furthermore, complement-inheritance verbs select an i-zone complement. Some lexical elements are given below.

$\ex. \a. \begin{avm} {\rm willen} ({\em to want}) $\mapsto$ \avmjprolog[ head ... ...w$\ \avmjpostlog] \> \\ arg-s & \<\@2,\@3,\@4\> \avmjpostlog] \end{avm}\par$
Note that directionality only orders complements relative to the head; the order of complements with respect to each other is left open. In previous work on Dutch [18], we have assumed that obliqueness provides an ordering of complements: if C₁ is more oblique than C₂ (i.e. the SYNSEM-value of) C₁ follows C₂ on COMPS), than C₁ appears closer to the head than C₂. For German, we have been satisfied with the assumption that the order of non-verbal o-zone complements is not (only) determined by obliqueness. For Dutch, this appears to be much less likely. Consider the example in (53), for instance. The perception verb zien is an complement-inheritance verb selecting for a verbal complement and an NP object (Marie). Both the NP Marie and the PP naar huis are o-zone complements of the head ziet. The order of both is strictly determined by the order on COMPS, however.

$\ex. \ag. dat Jan Marie naar huis ziet lopen\\ that Jan Marie towards home sees... ...{\em that Jan sees Mary walk home} \b. *dat Jan naar huis Marie ziet lopen \par$
This suggests that the Obliqueness word order constraint must apply for o-zone elements. This constraint, applicable to Dutch, is given in (54).

$\ex. {\bf Obliqueness Constraint (for Dutch):} \a. \begin{avm} \@1 \avmjprolog[... ...vm} precedes \begin{avm}\@1 \end{avm} on the {\sc comps} list of the head. \par$
A derivation for the VP of example (51-c)c can now be given as follows:

$\ex. {\sc \begin{tabular}[t]{ccccc} \multicolumn{5}{c}{ \node{vp1-vp}{vp$\lang... ...1}{vp1-wil} \nodeconnect{vp1-v2}{vp1-laten} \nodeconnect{vp1-v3}{vp1-lezen} \par$
The word order given in this example is the only grammatical word order, and the only word order allowed by the analysis. This is so, because the NPs are ordered to the left of the verbs because of directionality; the order of the NPs is determined by obliqueness; and the order of the verbs is determined by the Governor Constraints.

In comparison with the analysis presented in [18], the introduction of governor word order constraints has no consequences for the word orders that can or cannot be derived in the case of ordinary verb raising verbs. ¹⁶

It is more interesting for cases where governors and governees do not appear in strict left-to-right order, as in participle inversion constructions and in constructions involving separable prefixes.

Separable Prefixes

Certain verbs in Dutch subcategorize for a so-called `separable prefix'. These particle-like elements appear as part of the verb in subordinate clauses, but appear in clause-final position if their governor heads a main clause:

$\ex. \ag. dat Jan Marie {\em aan} spreekt\\ that John Mary {\sc prefix} speaks\\ {\em that John speaks to Mary} \b. Jan spreekt Marie aan \par$
In complex verb-sequences, the prefix can appear not only as part of its governor, but also in positions further to the left:

$\ex. \ag. dat Jan Marie zou hebben {\em aan} gesproken\\ that John Mary would h... ...u hebben gesproken \d. $^*$\ dat Jan Marie zou hebben gesproken {\em aan} \par$
These examples can now be treated straightforwardly. A verb selecting a separable prefix requires this prefix to be [GVOR $\rightarrow$ , i-zone]. This predicts correctly that this prefix can occur anywhere to the left of the governing verb within the verb cluster. The lexical entry for spreken and an example derivation are given below.

$\ex. \begin{avm} {\rm spreken} ({\em to speak (to)}) $\mapsto$ \avmjprolog[ hea... ... inner \\ gvor & $\rightarrow $ \avmjpostlog] \> \avmjpostlog] \end{avm}\par$

$\ex. \begin{tabular}[t]{ccccc} \multicolumn{4}{c}{ \node{nlpart-vp}{vp$\langle\... ...connect{nlpart-v1}{nlpart-hebben} \nodeconnect{nlpart-v2}{nlpart-gesproken} \par$
Note that apart from the word order given in (59), the orders aan zou hebben gesproken and zou hebben aan gesproken are derivable as well.

The verb-cluster sometimes also contains an adjective. The following examples are similar to a large set of examples given in [2]:

$\ex. \ag. dat de Sovjetunie zich niet zal laten {\em bang} maken door cruiserak... ...m he had made it clear to the pupils that the exam would not be difficult} \par$
The adjective can be placed at any position left of its governor, just like separable prefixes. The difference with compound verbs which require a separable prefix is not always easy to make. Such examples are not treated as compound verbs in [2].

Interestingly, such verbs can often take a full adjectival phrase instead of a single adjective. In those cases, however, the ADJP cannot be part of the verb cluster:

$\exi. \a. dat de Sovjetunie zich niet [ADJP erg bang] zal laten maken door cruis... ...had [ADJP heel erg duidelijk] gemaakt dat het examen niet moeilijk zou zijn \par$

These examples can be analysed straightforwardly by assuming that certain verbs subcategorize for an adjectival phrase. Those verbs do not constrain the ZONE feature of this adjectival phrase. This implies that if such an ADJP is selected in the I-ZONE, it must be a word; if it is selected as part of the O-ZONE it can be either a word or a full phrase.

Participle Inversion

The tense auxiliaries hebben (have) en zijn (be) are verb raisers which take a past participle as complement. These auxiliaries are special, however, in that they may take their complement either to the left or to the right:

$\ex. \ag. dat Jan het boek heeft gelezen\\ that John the book has read\\ {\em that John has read the book} \b. dat Jan het boek gelezen heeft \par$
In more complex examples, the pattern is that the participle is either to the left of the finite verb, or the last verb of the cluster:

$\ex. \ag. dat Jan het boek moet hebben gelezen\\ that John the book must have ... ...that John must have read the book} \b. dat Jan het boek gelezen moet hebben \par$
Examples such as:

$\ex. ?* dat Jan het boek moet gelezen hebben \par$
are acceptable only to a minority of speakers.

The analysis of these examples proceeds as follows. For those speakers that accept examples such as (64), we assume that the lexical entry for hebben leaves the GVOR value of its complement unspecified:

$\ex. \begin{avm} {\rm hebben }({\em to want}) $\mapsto$ \avmjprolog[ head & ve... ...\ comps & \@1 \\ zone & inner \avmjpostlog] \> \avmjpostlog] \end{avm}\par$
This results in a situation in which the participle may occur anywhere left of the auxiliary within the cluster: both (63)a, (63)b and (64) are derivable.

For the standard dialect we assume that the participle may only precede the governing auxiliary if it also precedes the head. Thus, the features for directionality and governor must agree. A modified lexical entry which implements this constraint is given below:

$\ex. \begin{avm} {\rm hebben} ({\em to want}) $\mapsto$ \avmjprolog[ head & ve... ...er \\ gvor & \@2 \\ dir & \@2 \avmjpostlog] \> \avmjpostlog] \end{avm}\par$
Note that the reentrancy between GVOR and DIR avoids the introduction of two separate lexical entries (i.e. one for the situation where the participle follows the auxiliary, and one for the situation where it precedes it). In (64), the participle gelezen precedes its governor, but follows the head of the phrase (i.e. zou), and thus no derivation on the basis of the modified lexical entry for hebben is possible.

Summarizing, this analysis correctly predicts the following set of facts, involving both a separable prefix and a participle:

$\ex. \ag. dat Jan Marie zou hebben aan gesproken\\ that John Mary would have ... ...arie zou gesproken aan hebben \f. *dat Jan Marie zou gesproken hebben aan \par$
The most interesting cases are the pairs (67)e,f and (67)i,j. Example (e) is allowed, because the prefix aan precedes its governor gesproken; in (f) this is not the case, hence this example is ruled out. The examples (i) and (j) are allowed only in those dialects which allow (64).

Next: Comparison and Conclusions Up: Word Order Constraints on Previous: Word order within the

Noord G.J.M. van
1998-09-28