Van-PPs are probably the most common PPs in a noun phrase, and can function either as complements or as adjuncts. The van-PPs in (56) clearly function as complements. First, they express the theme arguments of the deverbal nouns kopenbuying and makermaker; the nouns and the van-PPs thus stand in a similar semantic relation as the input verbs and their (nominal) complements. Second, the presence of the van-PP is required by the semantics of the derived nominal head (unless its semantic contribution is somehow contextually implied), just like the presence of the complements of the input verb. Third, the preposition van has no lexical content but merely expresses the relation between the head and its complement. It is often claimed that the use of van is motivated by the fact that verbs and nouns differ in that only the former can assign (abstract) case to its nominal complement; the use of van is necessary to formally license the theme argument by assigning case to it.

In (57), the van-PPs function as adjuncts; the information provided by the PPs is needed to identify the paper or book referred to but there is nothing in the semantics of these nouns that requires their presence. This is clear from the fact that it is perfectly possible to speak simply about een fietsa bike or een kranta newspaper. These indefinite noun phrases thus differ crucially from een makera maker, which inevitably evokes the idea of an object that has been created; the use of this indefinite noun phrase will be distinctly odd if the context does not provide any information about this object. Moreover, the preposition van in (57) differs from van in (56) in that it is lexical: this is clearest in example (57a), where it expresses a possession relation; in (57b) the relation can be regarded as one of time.

The above examples might suggest that a van-PP only functions as a complement of the head noun if the latter is derived and inherits the arguments of the base. This is indeed the normal rule, although there are two classes of exceptions: the first class is formed by the relational nouns, and the second by the so-called picture/story nouns, which in a sense could be said to have an agent and a theme argument; cf. Section 16.1, sub III. Some examples are given in (58) and (59); recall that the theme argument of story nouns is usually expressed by an over-PP.

PP constituents in the noun phrase can also be introduced by other prepositions. The PPs in the primeless examples in (60) clearly function as adjuncts, given that the nouns in question can also occur without them, as shown in the primed examples. Moreover, all the head nouns in (60) are non-derived, so there is no question of inherited arguments. Adjunct PPs like these can have a variety of semantic roles (location, direction, means, property, etc.).

Some researchers have argued that PP-adjuncts are easy to recognize: whenever the PP is headed by a preposition other than van, the PP is not a complement but an adjunct (cf. Booij & Van Haaften 1987; Hoekstra 1986): they maintain that the PPs aan pleinvreesfrom agoraphobia and naar Amsterdamto Amsterdam in the primeless sentences in (61) are adjuncts of the derived nouns lijdersufferer and reizigertraveler, despite the fact that the preposition is identical to that selected by the input verb lijdento suffer and reizento travel. Others claim that the PPs are inherited from the input verb. One reason for taking the second position is that the PPs in (61a&b) differ in the same way as the PPs in the corresponding verbal constructions in (61a'&b'): the selected preposition in the (a)-examples is functional in the sense that it has no lexical content but simply serves to express the relation between the head and its theme argument, whereas the preposition in the (b)-examples is lexical in the sense that it has retained its original directional meaning and introduces a predicative complement.

The most common case in which the complement need not be expressed syntactically is when its referent is recoverable from the context. In (64a) the necessary information is provided by the extra-linguistic context, and in (64b) by the linguistic context.

With relational nouns denoting body parts, the latter option is even grammaticalized: not mentioning the internal argument in the noun phrase leads to a default interpretation in which some other argument in the clause is interpreted as the possessor; in (65a&b) the required information is provided by the subject Marie, and in (65c) by an indirect object haarher. In these inalienable possession constructions, the article can also be replaced by a possessive pronoun explicitly expressing the possessor argument; cf. Section V3.3.1.4 for a more detailed discussion of these constructions.

Note that the choice between an indefinite or definite article in (65) depends on whether the body part in question is unique to each individual or not. If an indefinite article is used with a unique body part, the inalienable possession reading will not be available: an example such as (66a) will be interpreted to mean that Jan broke someone else’s nose. A similar effect occurs when a definite article is used with a non-unique body part: (66b) will be interpreted to mean that Jan broke a bone or, less favorably, someone’s leg. Note that replacing been with linkerbeen makes the referent unique again and the example acceptable, cf. (66c).

When the possessor is an indirect object, as in (66b), using an indefinite noun phrase with a unique body part actually makes the sentence infelicitous. The same is true, to a lesser extent, when we use a definite noun phrase with a non-unique body part.

The examples in (68) show that the internal arguments of a noun cannot be expressed in generic contexts. Example (68c) shows that these contexts also allow the use of an indefinite noun phrase for inalienably possessed unique body parts, which is impossible in the case of specific reference (cf. (66a)).

Replacing the indefinite noun phrases in (68) with specific ones leads to unacceptable results. However, they may become more acceptable, if the noun is modified by adjectives like goedegood or idealeideal. Note that, under the intended generic reading of (69b), the PP indicates that we are dealing with an ideal of Marie; most likely she is not even married.

The predicatively used noun phrases in (70) denote characteristic traits of the subject of the clause and exhibit a similar behavior to the generic noun phrases in (68) and (69): the complement of the noun cannot be expressed. As in (69b), the PP in (70c) again indicates that we are dealing with an ideal of Marie; this sentence does not imply that Peter is Marie’s husband.

Nouns derived from a pseudo-intransitive verb with a habitual reading inherit the property that mention of the complement is not required; the (a)-examples in (71) illustrate the normal, non-habitual use of the verb rokento smoke and the derived noun rokersmoker; the (b)-examples illustrate their habitual use.

Complements can be left unexpressed when a noun is quantified, modified or negated. This is illustrated in example (72a) for the quantifier iedereevery and in (72b) for the negator geenno. Such constructions are only fully acceptable if the sentence can be given a generic interpretation, as in (72), or if the implied argument refers to a contextually determined group of individuals, such as the children of a particular school class, as in (73).

In contexts where the focus is on the existence of the referent, or on establishing a relation between a noun and some other entity, this other entity typically does not appear in the form of a PP either. For instance, in example (74a) a relation is established between the noun phrases een koningina queen and dit landthis country. In such a context, the noun koningin, which normally requires a complement, can appear as an indefinite noun phrase without a complement. The same is true of the noun phrase een dampkringan atmosphere in (74b).

On the other hand, when the noun occurs in a definite noun phrase, a related argument is always implied, and the relationship between the noun and the implied entity is presupposed (e.g. “the queen of this country” and “the atmosphere of the earth” in examples (75a&b)).

Incorporation of the theme argument of a deverbal noun as the initial part of a compound is quite common in Dutch, especially with er and ing-nominalizations. Examples of incorporation with er-nouns can be found in (76). These examples show that incorporation results in adicity reduction of the derived noun, since the argument slot originally held by the incorporated argument is no longer available. This means that the er-noun usually requires the expression of a certain argument, but that this is no longer possible if the argument has been incorporated.

Adicity reduction is not restricted to cases where an argument is incorporated. In many cases the incorporation of some other element (e.g. an adjunct) can also block the expression of a theme argument. This is illustrated in (77) for an er-noun with an incorporated purpose adjunct (cf. voor je brood werkento work for a living) and an incorporated instrument adjunct; cf. Section 16.2.3.1, sub IIA5, for further discussion.

In the case of ing-nominalizations, theme incorporation also seems to lead to a reduction of adicity, although the effects may not be as strong as in the case of er-nominalization. Section 15.3.1.3, sub III, has already shown that incorporation is possible with both NP and PP-themes of the input verb.

Example (79a) shows that, after incorporation of the theme argument of an ing-nominalization derived from a ditransitive verb, the resulting compound noun is preferably used without an argument: the expression of the recipient is possible but seems somewhat marked. Example (79b) further shows that, after the incorporation of a theme argument, the agent cannot be expressed in the form of an agentive door-PP; the use of a van-PP is possible, but this may be due to the fact that this PP actually expresses a possessive relation.

The examples in (78) and (79) suggest that compound ing-nouns are lexicalized, which is why they do not accept arguments. However, this may not be a general rule, since there are some ing-nominalizations with an incorporated adjunct that do seem to allow the expression of arguments. In example (80), for instance, the incorporated adjunct leeftijdage, which provides the ground on which discrimination takes place, does not block the presence of the theme argument van oudere werknemersof older employees.

There seem to be different forms of extraction from the noun phrase: topicalization, relativization and wh-movement, PP-over-V and scrambling. Since they all have their own peculiarities, they will be discussed separately. We will mainly use van-PPs for illustration here, as these provide the clearest results; the behavior of PPs headed by prepositions other than van will be discussed in Sections 16.2.2 to 16.2.5.

The discussion above seems to suggest that PP-complements can be easily extracted from the noun phrase, but that this is not possible with PP-adjuncts. However, under certain specific conditions it seems possible to extract PP-adjuncts. This will be discussed in the following subsections.

Subsection B discussed a number of exceptions to the general rule that adjunct PPs cannot be extracted from a noun phrase. Since these exceptions all involve van-PPs, we should be careful not to jump to the conclusion that adjunct extraction is possible under certain circumstances, but rather to first see if some alternative analysis is possible. We suggest that a large number of these exceptions can be explained in principle by assuming that the PPs in question are generated as restrictive adverbial phrases outside the noun phrase, whose main function is to restrict the domain of discourse; cf. Kaan (1992:§5) for a somewhat different proposal.

That this is a realistic possibility is clear from the examples in (119). In (119a) the van-PP can easily be interpreted as a source argument selected by the verb horento hear. However, this is less likely in the case of (119b): the information that Peter will come is not necessarily provided by Jan himself, but could have been acquired in some other way. The analysis of the van-PP in example (119c&d) as a source argument of the verb wetento know is even less likely: in (119c) the source of the evaluative assertion expressed by the embedded clause is most likely the speaker himself, and in (119d) the source cannot be the referent of the noun phrase embedded in the van-PP, since it does not refer to a human entity. From this we conclude that, at least in (119b-d), the van-PP need not or cannot be construed as the complement of the verb.

That this is a realistic possibility is clear from the examples in (119). In (119a) the van-PP can easily be interpreted as a source argument selected by the verb horento hear. However, this is less likely in the case of (119b): the information that Peter will come is not necessarily provided by Jan himself, but could have been acquired in some other way. The analysis of the van-PP in example (119c&d) as a source argument of the verb wetento know is even less likely: in (119c) the source of the evaluative assertion expressed by the embedded clause is most likely the speaker himself, and in (119d) the source cannot be the referent of the noun phrase embedded in the van-PP, since it does not refer to a human entity. From this we conclude that, at least in (119b-d), the van-PP need not or cannot be construed as the complement of the verb.

That the supposedly extracted van-PPs are actually adverbial phrases could also be argued on the basis of example (121a), which contains the focus particles alleenonly and ookalso: under the standard assumption that the string preceding the finite verb constitutes a single constituent, the extraction analysis would incorrectly predict that (121b) should also be acceptable. Recall, however, that in the case of PP-over-V the particle and the noun phrase must also be split, so there may be some interfering factor at play; cf. Barbiers (1995). The examples in (122) provide similar cases with constituent negation.

More evidence for the analysis proposed here can be found in (123): with a van-PP in sentence-initial position, the sentence is acceptable with both the definite article hetthe and the possessive pronoun zijnhis (though the latter may be slightly marked). With the PP immediately after the head, on the other hand, only the definite article can be used. This suggests that, at least in the case of the construction with the possessive determiner, but more likely in both cases, the van-PP in (123a) is not extracted from the noun phrase zijn werkhis work but is generated as an independent adverbial constituent.

Examples of a similar kind are given in (124). Again, the van-PP cannot have been extracted from the subject, since it cannot appear in its original position, following the head; cf. the contrast in acceptability between the two orders in (124).

The discussion above seems to lead to the conclusion that many cases of alleged extraction of PP-adjunct from noun phrases are merely apparent, and should be reanalyzed as involving independent adverbial phrases. This, of course, has serious consequences for the extraction test as a whole, since it may be that the alleged cases of extraction of PP-complements from noun phrases also involve independent adverbial phrases; cf. Broekhuis (2005/2016) for a relevant discussion.