The first test focuses on the fact that the finite verb agrees in number with the subject of the clause. Since the two nouns in a QBC can differ in number, we can determine the syntactic head of the construction by looking at the number specification of the finite verb: the noun with which the verb agrees is the syntactic head. Example (8) illustrates this for the quantifier noun boela lot and the collective noun groepgroup. In (8a), the number specification on the finite verb clearly shows that we should consider the plural N2 studentenstudents as the syntactic head of the construction and not the singular N1 boel. In (8b), on the other hand, the singular agreement on the verb clearly shows that it is the singular N1 groep that acts as the syntactic head.

Since we have seen in Section 18.1.1.1 that the noun aantal is ambiguous between a quantifier and a collective reading, it is expected that QBCs with this noun will show mixed behavior with respect to subject-verb agreement. The examples in (9) show that this expectation is indeed borne out. Note, however, that the two examples seem to differ in their preferential agreement pattern: a search in the Corpus Gesproken Nederlands by Van Eerten (2007) revealed that in examples such as (9a) the majority of cases (76%) exhibit plural agreement, whereas in examples such as (9b) there is a clear preference for singular agreement (70%). This may be related to the fact that the QBC in (9) is indefinite and that placement of indefinite phrases into the clause-initial position triggers a partitive reading, which may be more readily available in the referential reading of N1.

If N1 is a measure noun, there are also two possibilities: in (10) the verb can exhibit singular agreement, in which case it agrees with the singular N1 kilokilo, or plural agreement, in which case it agrees with the plural N2 appelsapples. To our ear, the primeless examples are equally good, while the primed examples with the QBC in the clause-initial position, which is always somewhat marked, clearly prefer singular agreement.

This suggests that in this case we are also dealing with an ambiguity between a purely quantificational and a more referential reading of the noun. This seems to be supported by the fact, illustrated in the primeless examples in (11), that measure nouns exhibit different behavior with respect to pluralization in the two constructions: N1 is marked plural when it agrees with the verb, but not when the verb agrees with N2, the substance noun melkmilk. From this we should conclude that when the measure noun liter is not the syntactic head of the QBC, it loses its ability to trigger a plural, which could be interpreted as an indication that it has lost its referential status as a count noun.

This is also supported by a semantic difference between the examples. In (11a) we perceive the milk as two quantificational units of one liter each; we will call this the package-unit reading, since it implies that the milk was purchased in containers that each contain one liter of milk. In (11b), on the other hand, we do not perceive the milk as available in specific units: there may be a single container containing two liters of milk, or there may be more, as long as the total quantity is (approximately) two liters. The primed examples show again that the QBC can only occur in the clause-initial position if N1 has a referential reading.

The ambiguity described above can only arise if the descriptive content of N1 is weak: it is difficult to determine what the denotation set of nouns like boela lot, aantalnumber, kilokilo and literliter is. When N1 has a clear descriptive content, like the collective and container nouns in (12), agreement with this noun is strongly preferred.

Part nouns like reepbar in (13) also seem to have descriptive content, and we therefore expect them to trigger agreement on the verb. This is indeed the case, although we cannot show it by appealing to the agreement facts alone, because part nouns are always used in combination with a substance noun, which invariably triggers singular agreement on the verb: the fact that the verb in (13a) is singular does not tell us much. The plural agreement in (13b), of course, conclusively shows that N1 can act as the syntactic head of the construction, but since we have seen in (11a) that the verb must agree with plural N1s, this is still not enough to exclude the possibility that N2 can act as the syntactic head in (13a). However, the fact that the part-noun counterpart of (11b), given in (13c), is unacceptable seems sufficient to conclude that the part nouns must act as the syntactic head of a QBC: if N2 could act as the syntactic head of the construction, this example should be acceptable.

The examples in (8) to (13) have shown that the number features that trigger number agreement on the finite verb can be located either on N2 or on N1. The actual choice seems to depend on whether N1 is referential or purely quantificational. A purely quantificational noun like boellot in (8) apparently does not have the necessary features to trigger agreement on the verb, whereas referential nouns like the collective noun kuddeherd in (12a) or the part noun reepbar in (13) do have these features. Other nouns, such as the measure noun liter, seem to have an intermediate status, and the question of whether they trigger agreement on the verb or not depends on whether they have a purely quantificational or a more referential function.

That both N1 and N2 can act as the syntactic head of the construction can also be shown by demonstrative pronouns. Demonstratives agree with the head noun in gender and number: if the head noun is [+neuter, singular], the proximate and distal demonstratives are ditthis and datthat, whereas in all other cases they are dezethis/these and diethat/those; cf. Section 19.2.3.1. The examples in (14), which contain a neuter singular N1, show that the distal demonstrative can indeed agree with both nouns. In the primeless examples the proximate demonstrative agrees with the neuter N1, whereas in the primed examples it is the plural/non-neuter N2 that triggers agreement. Some people object to the primed examples, but the pattern is very common, especially with the noun paar, which can be easily found on the internet. Examples such as (14b') seem less numerous, but they are also found on the internet.

The same thing can be shown for the proximate demonstratives: our search for the string [deze paar] resulted in 551 hits, and 46 out of the first 50 instantiated the construction; our search for the string [deze pond] resulted in just a single instance of the desired construction.

As expected, the two options in (14) and (15) have different interpretations. This is clearest in the (a)-examples with the noun paar: in the primeless examples, the QBC refers to two ducks that belong together and form a couple; in the primed examples, on the other hand, the noun paar has a purely quantificational meaning and refers only to a small number of ducks. Something similar holds for the (b)-examples: in the primeless examples, the QBC refers to a single piece of cheese, whereas no such implication holds for the primed examples. This again suggests that agreement with N1 is only possible if it is referential: if it is purely quantificational, it is N2 that enters the agreement relation. This conclusion seems to be supported by the fact, illustrated in (16), that diminutive formation is blocked when N2 agrees with the demonstrative: this suggests that N1 has lost its referential status in this case. We will return to this in Section 18.1.1.3, sub I.

Again, the ambiguity only occurs with nouns that have little descriptive content. It does not occur with container, collective, and part nouns. In (17a&b), the container noun fles and the collective noun kudde are non-neuter, whereas the non-count nouns bier and vee are neuter, and only the non-neuter demonstratives can be used. In (17c) the part noun stuk is neuter, while the N2 kaas is non-neuter, and only the neuter demonstrative yields an acceptable result.

The two subsections above have shown that subject-verb agreement as well as gender marking on demonstrative pronouns can be determined by either N1 or N2, depending on the type of noun we are dealing with: if N1 is purely quantificational, it is always N2 that triggers agreement; if N1 has descriptive content (i.e. if it is a container, collective or part noun), it is N1 that triggers agreement; measure nouns are special in that they allow both patterns. Some nouns are ambiguous and can be used either as a quantifier noun or as a noun of another type.

Verbs can impose several semantic selection restrictions on their arguments. Verbs like zich verspreidento disperse and omsingelento surround, for example, generally require a plural noun phrase as their subject: in (18a'), for example, the use of the singular noun phrase de studentthe student leads to a semantically anomalous result. That the restriction is semantic and not syntactic is clear from the fact that the use of singular noun phrases referring to collections of entities, such as politiepolice, results in an acceptable construction. The symbol “$” is used to indicate semantic incompatibility.

That satisfaction of the semantic restriction is not a syntactic matter is made even clearer by the QBC constructions in (19): in (19a) the semantic restriction is satisfied by the syntactic head of the construction, but in (19b) the noun that triggers agreement and the noun that satisfies the semantic restriction are different.

Example (19b) shows conclusively that there is no a priori reason to assume that N2 can only act as the semantic head of the QBC if N1 has a purely quantificational meaning. And the examples in (20) show that there is indeed no such restriction. Example (20a) shows that a verb like verzamelencollect requires the direct object to refer to a set of separable entities like stamps or pieces of furniture. The unacceptability of (20b) shows that a noun phrase headed by a container noun like doosbox does not satisfy this selection restriction. The acceptability of (20c) therefore shows that in QBCs with a container noun it is N2 that satisfies the semantic restrictions.

The same can be shown by appealing to other kinds of semantic restrictions. For example, a verb like rokento smoke selects a direct object that refers either to a substance, like tobacco, that can be smoked, or to an entity made of that substance, like a cigar; cf. in (21a). Example (21b) is infelicitous, since a noun phrase like een doosa box does not satisfy this selection restriction. Consequently, the fact that (21c) is acceptable shows that the selection restrictions of the verb can be satisfied by N2 despite the fact, discussed in Section 18.1.1.2, sub I, that N1 is always the syntactic head of the construction.

Note that we have put aside the fact that example (21b) is acceptable under a generic/habitual interpretation: Hij rookt een doos per dagHe smokes a box per day. In such cases we are dealing with an elliptical QBC construction: Jan does not smoke the box, but its contents. Such constructions are only acceptable if information about the contents of the box is available to the addressee.

It is important to note that the descriptive content of the container noun in the QBC een doos sigaren in (21c) has been backgrounded in favor of the package-unit reading: the QBC does not normally refer to a box with a certain content, but to a certain number of cigars. However, this does not mean that this is always the case. Consider the examples in (22), where the verb sluitento close is substituted for the verb rokento smoke in (21). The examples in (22a&b) show that the noun phrase sigaren cannot satisfy the semantic selection restrictions of this verb, whereas the noun phrase een doos can. From the fact that (22c) is acceptable, we conclude that N1 functions as the semantic head of the QBC, which implies that it has retained its descriptive content: we are still referring to a box with certain contents, not to a number of cigars. The contrast between (21) and (22) thus shows that QBCs headed by a container noun are ambiguous.

It seems that measure, collective, and part nouns behave like container nouns. We will therefore limit our discussion to these types by showing in (23) that in QBCs headed by these nouns, N2 can also satisfy the semantic selection restrictions imposed by the verb.

This subsection has shown that most QBCs are ambiguous depending on whether N1 receives a more referential or a more quantificational interpretation: in the former case it is N1 that functions as the semantic head of the construction, in the latter case it is N2. The question of which head functions as the semantic head is independent of the question of which head functions as the syntactic head: the two functions may or may not be performed by the same noun. The quantifier nouns differ from the other nouns in that they never function as the semantic head of the construction, which is related to the fact that they do not have much descriptive content to begin with.

That N2 can be the semantic head of the construction is also clear from the fact that the QBC as a whole can be modified by attributive modifiers that belong to N2 rather than N1. Some examples are given in (24). The primeless and primed examples are more or less synonymous, suggesting that in both cases the attributive adjective modifies N2.

That it is not N1 that is modified is particularly clear from the examples in (25a&b): here the adjectives can only modify the noun glas, so that the primeless and primed examples are no longer synonymous. Furthermore, example (25b) receives an anomalous interpretation (which seems to be marginally accepted by some speakers).

The unacceptability of (26) points in the same direction: since viesunsavory and lekkertasty are antonyms, the structure leads to a contradiction (although it should be noted that such examples are sometimes used as puns).

The fact that the adjective can modify N2 does not imply that it also agrees with this noun in number/gender. This is illustrated in (27): example (27a) shows that the non-neuter substance noun wijn requires the use of the inflected form of the adjective; in (27b), on the other hand, the -e ending is missing because the adjective agrees with the singular neuter noun glas.

There seem to be certain restrictions on the availability of the intended reading that are not entirely clear. For example, although the QBC in (28a) can be found on the internet (2 hits), we have the impression that the order in (28a') is much preferred. Furthermore, example (28b) shows that if the adjective and N2 form a fixed collocation, such as witte wijnwhite wine, the adjective must immediately precede N1: the primeless example can only refer to a white bottle.

Finally, if the attributive adjective can also be used to modify N1, the reading in which the adjective preceding N1 modifies N2 is excluded: the two (a)-examples in (29) are not synonymous, and example (29b) does not lead to a contradiction.

So far we have only considered container nouns, but the (a) and (b)-examples in (30) show that similar facts can be found with collective and part nouns, respectively. That we are dealing here with a modifier of N2 and not a modifier of N1 is supported by the fact that N1 can only be modified by a very small class of attributively used adjectives; cf. Section 18.1.1.3, sub IIE, for discussion and examples.

However, this does not mean that the modifier can always precede N1: the examples in (31) show that quantifier and measure nouns do not license this kind of modification; the modifier of N2 must follow N1.

That N2 can be the semantic head of a QBC can also be shown by the interpretation of the reciprocal pronoun elkaareach other, which must have a c-commanding plural antecedent; cf. Chapter 23. For our present purpose it suffices to say that a reciprocal pronoun functioning as a (PP-)object of the verb can be interpreted as coreferential with the subject of the clause, but not with a noun phrase embedded in the subject of the clause. In (32a), for example, elkaar can be bound by the subject de ouders van Jan en MarieJan and Marieʼs parents, but crucially not by the noun phrase Jan and Marie. The same holds for (32b), where the noun phrase hun ouderstheir parents can be coreferential with elkaar, whereas the possessive pronoun huntheir embedded in the subject cannot.

The examples in (33) show that N2 cannot be regarded as embedded in the subject in the same way as, for instance, the possessive pronoun huntheir in (32b). Regardless of the type of N1, N2 is able to bind the reciprocal elkaareach other. Note that we do not include examples of a QBC with a part noun, because these nouns can only be combined with non-count nouns, which cannot act as antecedents of a reciprocal.

The fact that N2 can act as the antecedent of the reciprocal pronoun shows that it can act as the semantic head of the QBC. Note that the agreement on the verb shows that N2 need not be the syntactic head of the construction; this is the case when N1 is a quantifier noun, as in (33a), but not in the other cases.