The most striking property of the N-van-een-N construction is that the two nouns usually agree in number: if N1 is singular, N2 must also be singular; if N1 is plural, N2 must also be plural. This is illustrated in (172).

As shown in (173), the N-van-een-N construction resembles the copular construction in this respect, where number agreement between the subject and the predicative noun phrase is also generally obligatory; cf. Section 22.2, sub IV, for some exceptions. This supports the idea that the two nouns in the N-van-een-N construction stand in a predicative relation.

Occasionally, however, predicative singular nouns can be predicated of a plural subject, as in (174a), and this is even the norm when the predicate is a mass noun, as in (174b). Bennis et al. (1998) claim that for some people the corresponding N-van-een-N constructions in the primed examples are also acceptable (to varying degrees), and cases like (174b') can indeed be found on the internet using the search string [dat tuig van een]. This emphasizes the similarity between the N-van-een-N construction and the copular construction, and thus supports the idea that N1 and N2 are in a predicative relation in the N-van-een-N construction.

It has also been claimed that examples such as (175b), in which N2 is a mass noun, are at least marginally possible for some speakers (we found one example on the internet). However, the corresponding copular construction is absolutely unacceptable. Note, however, that in English, singular mass nouns that trigger plural agreement on the finite verb (like the police in The police are coming) can occur as the subject in a copular construction with a plural nominal predicate: The police are idiots. When a mass noun triggers singular agreement on the finite verb, however, this is impossible: %The government is/are idiots. The unacceptability of Dutch examples such as (175a) may therefore be due to the fact that all Dutch mass nouns trigger singular agreement on the finite verb.

A hotly debated issue regarding the N-van-een-N construction is whether N1 or N2 is the semantic head of the construction. The fact that in constructions such as (176a), the N-van-een-N construction can be replaced by a noun phrase headed by either N1 or by N2 has led to the idea that the construction is ambiguous and that either of the two nouns can function as the semantic head of the construction.

This conclusion, however, seems to be wrong, because the acceptability of (176b) is only due to the fact that the noun phrase twee schatten is used as a (metaphorical) predicate, just as in the N-van-een-N construction. If the binominal phrase is used as an argument, as in (177), the direct object twee schatten in (177b) cannot be construed metaphorically, but must refer to entities that are part of the regular denotation set of the noun schattreasure; consequently, (177b) refers to a different state of affairs than (177c). On the other hand, the fact that (177c) can be used to refer to the same state of affairs as (177a) clearly shows that it is N2 that acts as the semantic head of the N-van-een-N construction.

Since the two nouns in the N-van-een-N construction generally agree in number, it is difficult to say which of the two nouns triggers agreement on the finite verb. To determine this, we must appeal to the more exceptional and perhaps controversial cases in (174b) and (175b). Our own judgments suggest that non-linguistic factors may be decisive when we are dealing with a singular N1: in (178a) singular agreement seems to be preferred, whereas in (178a') it is plural agreement that is preferred. In examples such as (178b), where N1 is plural, we always seem to have plural agreement. The data in (178) shows that the fact that N2 is the semantic head of the construction does not necessarily imply that it is also the syntactic head of the construction; cf. the discussion in 18.1.1.2, where we reached a similar conclusion for quantificational binominal constructions.

Another way to determine the syntactic head of the construction is to consider which determiner the N-van-een-N construction takes. If the definite determiner agrees in gender with N1, we conclude that N1 is the syntactic head of the construction, and if it agrees with N2, we conclude that N2 is the syntactic head. Unfortunately, we cannot show this on the basis of the definite articles de and het, since we have already seen in (170a) that definite articles cannot be used in the N-van-een-N construction. However, gender agreement can also be illustrated by demonstrative pronouns: the demonstrative diethat/those is non-neuter and/or plural, whereas the demonstrative datthat is singular neuter.

The examples in (179a&b) and (180) show that the demonstratives can easily be used when the two nouns select the same demonstrative, i.e. when they both select die or dat. According to some speakers, the mixed singular examples are excluded, while other speakers accept at least some of these examples. In the case of (179c&c'), the judgments seem to differ among speakers: some prefer the primeless example, in which the demonstrative agrees with N1 but not with N2, while others prefer the primed example, in which the demonstrative agrees with N2. The judgments on the (d)-examples in (179), on the other hand, seem clearer: the primed example is generally rejected, while the primeless example is accepted by at least some speakers. Similar judgments were collected for the possessive pronouns onze[-neuter,+sg] our and ons[+neuter,+sg] our in (181). For a detailed discussion, we refer the reader to Everaert (1992).

Table 6 summarizes the results above. Examples in which the determiner agrees in gender with the two nouns are always possible. If the two nouns differ in gender, agreement between the determiner and N1 is obligatory for at least one group of speakers. For another group of speakers, the gender of N1 affects the agreement pattern: if N1 is [-neuter], agreement between the determiner and N2 is preferred, but if N1 is [+neuter], agreement between the determiner and N2 is also blocked for them. Obviously, the cases in which there is no agreement at all produce the most degraded results, which is not reflected in the judgments in the table.

To conclude this discussion of agreement, we would like to point out that the set of determiners preceding the N-van-een-N construction is rather limited. For example, the demonstratives die/dat in (179) and (180) above do not really have a deictic function, but rather seem to express some kind of affective meaning, as can also be found in e.g. die Jan toch!, which is said about Jan when he does/says something special and the speaker wants to express his approval or (mild) disapproval of what Jan does/says. As shown in (182a), the use of demonstratives in their deictic function generally leads to a bad result, as does the use of the definite article does. Only in very special contexts are “true” demonstratives possible. For instance, an interrogative example such as (182b) seems possible if the speaker hugs the cat in question and thus shows that he himself likes it, but not if he just points at it. In other words, (182b) is only possible as a kind of rhetorical question.

The fact that the N-van-een-N construction conveys a strong personal evaluation of the referent of the whole construction may also account for the fact that first-person possessive pronouns are more common in the construction than second or third-person ones. When acceptable, the use of second-person and third-person possessive pronouns generally conveys an ironic message; the speaker of (183b), for example, confronts the hearer with a fact that is inconsistent with the description of the cats as being “schatten”.

Due to restrictions like these, the set of determiners preceding the N-van-een-N construction is largely limited to the cases discussed above and the indefinite articles een/∅ and derivatives of them like zo’nsuch a, geenno and wat eenwhat a. Some examples with these indefinite determiners are given in (184).

The modification of nouns in the construction is subject to several restrictions. For example, it is impossible to insert an attributive adjective immediately before N2; the only exceptions are classifying adjectives in collocations like Cyperse kattabby in (185c).

It is possible to use an attributive adjective to modify N1, but generally these modifiers are amplifying or affective in nature and do not attribute a property to N1, which is of course not surprising given that N1 is not referential in nature.

According to some, an attributive adjective preceding N1 can also be used to modify N2, which would be consistent with the fact that it is N2 that acts as the semantic head of the construction. Some examples, taken from Den Dikken (1995b), are given in (187).

Insofar as the examples in (187) are acceptable, it is clear that roodharig must modify N2. Note that the data in (187) is in line with the results regarding gender agreement in Table 6: the (a)-examples show that when N1 is neuter and N2 is non-neuter, the adjective must agree with N1, whereas the (b)-examples show that when N1 is non-neuter and N2 is neuter, speakers seem to vary with respect to the noun that triggers agreement — for some speakers it is N1, as in (187b), while for others it is N2, as in (187b'). So again, we have to conclude that the feature [+neuter] on N1 blocks gender agreement with N2 for all speakers.

Although attributively used adjectives can precede N1, postnominal modifiers cannot immediately follow it, as shown in (188a). Probably the impossibility of modifying N1 is again due to the fact that N1 is not referential in nature. Example (188b) shows that postmodifiers following N2 are possible, but in these cases we cannot immediately decide whether the PP modifies N2 or the whole N-van-een-N construction.

In order to find out whether the modifier in (188b) is modifies N2 or the complete N-van-een-N construction, we can consider relative clauses such as those given in (189). The fact that the relative pronoun must agree in gender with N2 shows that it is this noun that is modified, and not the complete N-van-een-N construction.

The indefinite article preceding N2 cannot be replaced by other kinds of determiners. The indefinite article seems to be sensitive to the number of N2: if N2 is singular the indefinite article is eena, and if it is plural, the article has the null form.

The “%” before een in (190b) expresses that while Bennis et al. (1998) claim that examples such as schatten van een katten are possible, other speakers consider the result to be at best highly marked. Nevertheless, the Npl van een Npl construction is perfectly acceptable to all speakers in exclamative contexts like (191a), where both nouns are preceded by the indefinite article een (pronounced with schwa: [ən]). As shown in (191b), such a combination of een and a plural noun is not restricted to N-van-een-N constructions of this kind, but is typical for this kind of exclamative constructions; cf. Section 19.1.4.2, sub I.

This suggests that the indefinite article and N2 can simply be analyzed as a noun phrase. However, there is a problem with this conclusion; een is also possible with N2s that normally cannot be preceded by an indefinite article. The most striking case concerns proper nouns: normally a proper noun like Marie is not preceded by an indefinite article (*een Marie), but in (192) it is obligatory.

The same thing can perhaps be shown on the basis of substance nouns, which normally cannot be preceded by an indefinite article either: een pracht van een wijn/kaas (lit.: a beauty of a wine/cheese). A caveat is in order, however, because speakers tend not to construe the N2s in such cases as substance nouns; instead, the noun wijn is interpreted as referring to a certain kind of wine, and the N2 kaas can refer to an actual object (i.e. the concrete result of the production process).

There are also proper nouns that can be preceded by a definite but not an indefinite article, for example de/*een Westerkerk or het/*een paleis op de Dam. Again, these proper nouns must be preceded by een in the N-van-een-N construction; note that the definite article, which is normally present, cannot be used in these binominal constructions.

The facts in (192) and (193) have led to the suggestion that een is not actually part of the noun phrase headed by N2, but is present to perform some other function; cf. Bennis et al. (1998) for discussion.

Since the preposition van cannot be replaced by any other preposition, it has been suggested that it is a spurious preposition. Alexiadou et al. (2007:246) suggest that this is also supported by the fact that, unlike true van-PPs, the sequence van + noun phrase cannot undergo R-pronominalization; cf. (194b). Another fact that might point in this direction is that this sequence cannot be separated from the sequence preceding it; this is illustrated for PP-over-V in (194c).

Bennis et al. (1998) also adopt the claim that van is a spurious preposition, and they have suggested that its syntactic function is to signal the (metaphorical) predicative relation between N1 and N2; cf. die kerel is een boomthat fellow is (like) a tree. It is claimed that van is in a sense comparable to the copula zijnto be in a copular construction; cf. Den Dikken (2006:§5) for a detailed technical implementation of this idea.

The N-van-een-N construction can be used in all regular NP-positions, i.e. both as an argument and as a nominal predicate. In (195) we give examples in which the construction functions as subject, direct object, indirect object, complement of a preposition, and predicate in a copular construction.