Partitive constructions are noun phrases referring to a subset of a presupposed set in the domain of discourse. They consist of a cardinal numeral or a quantifier expressing the cardinality or size of the subset, followed by a van-PP the nominal whose complement denotes the presupposed set. Some examples are given in (347), where the noun phrase de koekjesthe cookies refers to the presupposed set. The cardinal vier in (347a) indicates that the cardinality of the subset is 4, and the quantifier veel in (347b) expresses that the cardinality of this subset is greater than some implicitly assumed norm. Example (347c) shows that the universal quantifier alleall cannot be used, possibly because it conveys redundant information, since it refers to the same set as the noun phrase de/alle koekjesthe/all cookies. Example (347d) with the distributive quantifier elkeach, on the other hand, is acceptable: the noun phrase does not refer to the contextually determined set as a whole, but to the individual entities that make up that set.

The partitive construction is syntactically headed by the numeral/quantifier, not by the complement of van, which does not trigger number agreement on the finite verb. Example (348) shows that it is the numeral/quantifier that determines agreement, or more precisely, the phonetically empty noun that follows it; cf. the discussion of (352) below.

Nouns appearing as N1 in binominal constructions have a quantificational meaning, and so it is not so surprising that they can also appear in the partitive construction; cf. (349a). However, the examples in (349b-e) show that container, collective and, to a lesser extent, measure nouns preceded by the indefinite article eena lead to a degraded result. When such N1s are preceded by a cardinal, the result is usually more acceptable (unless they have a plural suffix as in twee kilo’s), even though most speakers then tend to interpret the N1s as more referential.

The acceptability judgments also depend on the nature of the nominal complement of the van-PP: if the noun phrase is preceded by a demonstrative pronoun instead of a definite article, the result is perfectly acceptable, and the primary reading is the quantificational one. The examples in (350) show that this applies both to expressions in which N1 is preceded by an indefinite article and to expressions in which it is preceded by a cardinal.

We have already seen in (348) that number agreement on the finite verb is triggered by the part preceding the van-phrase. The examples in (351c-e) show that this is also true for the partitive constructions in (349) and (350) with part, container and collective nouns. However, the quantifier and measure nouns in (351a-b) behave differently: they allow agreement between the verb and the nominal complement of the van-PP. This is similar to what we saw in Section 18.1.1.2, sub I: in purely quantificational constructions it is N2 that triggers agreement with the finite verb, while in the more referential constructions it is N1 that triggers agreement.

Since it is implausible that in (351a-b) the number agreement on verb is directly triggered by the complement of the van-PP, it has been suggested that the partitive construction features an empty noun following the numeral/quantifier, which has the same denotation as the nominal head of the complement of the van-PP; cf. Kranendonk (2010:§5) for a relevant discussion. This would mean that the noun phrases in (348) have the structures in (352) and that e has the same denotation as the noun student. Since the cardinal éénone must be followed by a singular noun and the cardinal vierfour by a plural noun, the agreement facts in (348) can be explained by assuming that it is the empty noun that triggers agreement on the verb.

This proposal implies that the structures of the noun phrases in (351) are as given in (353): we are dealing with regular quantificational binominal constructions in which the phonetically empty noun functions as N2. The fact that the agreement pattern of the partitive construction in (351) is identical to that of the constructions discussed in Section 18.1.1.2, sub I, now follows from the fact that they are both binominal.

The analysis in (352) and (353) turns the partitive construction into a regular noun phrase, comparable to the examples in (354), adapted from Kobele and Zimmermann ((2012:257). Since the nominal head is phonetically realized, it can have a denotation independent of the nominal part of the van-PP.

The two examples in (355) seem identical to those in (347a&b); the only difference is that the nominal complement of van is preceded not by the definite article dethe, but by the distal demonstrative diethose. Therefore, a genuine partitive reading of these examples need not surprise us. What we want to focus on here, however, is that there is a second, non-partitive reading with a meaning close to “four/many cookies of a certain kind familiar to the addressee”.

The same ambiguity arises in the examples in (356), where the van-phrase is preceded by nouns that can also appear as N1s in quantificational binominal constructions: all examples in (356) can be interpreted as either partitive or pseudo-partitive constructions. Note that constructions with the singular neuter demonstrative dat trigger the same ambiguity; cf. the examples in (356b&c).

The availability of the pseudo-partitive reading of the examples in (355) and (356) is related to the fact that the phrase van die/dat (A) + N can be used with the distribution of a DP, i.e. although it has the appearance of a PP, it can be used in positions that are normally occupied by a noun phrase; cf. Section 19.2.3.2, sub IIE. This is illustrated in (357): the translations indicate that pseudo-partitives can express that the denotation of the noun is familiar to the addressee or have an intensifying meaning comparable to English “these + Adj + Npl”.

Since PPs cannot normally function as subjects, we can conclude that the van-PPs in (355) and (356) are ambiguous: they can be interpreted either as a PP, leading to the partitive reading, or as a noun phrase, leading to the pseudo-partitive reading. According to this analysis, the partitive and pseudo-partitive readings of the examples in (356) are both binominal constructions, but they differ in that in the former case N2 has the form of an empty noun, while in the latter case it is a spurious PP that functions as N2. If this is correct, we can account for the ambiguity of the examples in (355) and (356a) by assigning to them the representations in (358).

The above analyses of partitive and pseudo-partitive noun phrases imply that van has a different status in the two constructions: in partitive noun phrases it is a regular preposition, while in pseudo-partitive noun phrases it is a spurious preposition. To substantiate this claim, we will look more closely at constructions in which the spurious van-PP is used as an argument of a verb or a preposition, and show that it behaves as a noun phrase.

The previous subsection has shown that the partitive van-phrase is headed by a true preposition, while the pseudo-partitive van-phrase is a disguised noun phrase. Something similar applies to the demonstrative. The examples in (372) and (373) show that while the distal demonstrative die/dat is part of a larger paradigm in the partitive construction, it cannot be replaced by any other determiner in the pseudo-partitive construction.

The fact that pseudo-partitive constructions can only occur with the distal demonstratives die and dat suggests that these demonstratives are used in a non-canonical way in this construction. This can also be supported by appealing to the fact that noun phrases with a distal demonstrative can normally be modified by the locational adjunct daarover there: in (374) it has a disambiguating effect on examples like vier/een aantal van die boeken and een glas van dat bier; the presence of daar emphasizes the partitive readings found in (372) by blocking the pseudo-partitive reading found in (373).

The prosodic properties of the demonstrative also suggest that it is a defective form in the pseudo-partitive construction. Demonstratives are typically used in contrastive contexts and can therefore easily be assigned contrastive accent: niet dit maar dat boek not this but that book. However, the demonstrative in the pseudo-partitive construction resists accent, as can be seen from the fact that the examples in (375) can only be interpreted as true partitive constructions.

Finally, (376b) shows that the demonstrative in the spurious van-PP cannot be followed by a cardinal or quantifier, again suggesting that the defective demonstrative is not a regular determiner.

Since the partitive construction refers to a subset of a presupposed set, the complement of van must be definite. This predicts that the indefinite demonstrative determiner zulk(e)such cannot occur in the partitive construction. As shown in (377), this prediction is correct; the noun phrases following van have only a type reading and in this sense resemble the pseudo-partitive reading.