Section 18.1.1.3, sub IIE, has shown that N1 can only be modified by a limited set of attributive adjectives, namely those with a quantificational meaning or those indicating size. In other cases, attributive adjectives preceding N1 actually modify N2 (see Section 18.1.1.2, sub IIB, for details), despite the fact that in these cases gender and number agreement is always with N1, not with N2.

In (109) the singular neuter noun glas/stuk requires that the attributive -e ending be absent, whereas agreement between the adjective and the non-neuter substance noun N2 would have required the presence of the -e ending. This is clear from the fact, illustrated in (110), that the -e ending must be present if the adjective follows N1. This again shows that if N2 functions as the semantic head of the QBC, this does not imply that it also functions as the syntactic head.

The attributive inflection on the adjective lekker in (109) is sensitive to the number and definiteness feature of the full binominal phrase; if the singular N1 is replaced by a plural one, or if the indefinite article een is replaced by the definite article het, the adjective must have the -e ending. This is shown for (109a) in (111): note that we have replaced the non-neuter N2 wijn with the neuter N2 bier to block interference of the gender feature of this noun.

When the adjective immediately precedes N2, the adjective is not sensitive to the number and definiteness feature of the full binominal phrase. This is shown in (112), where the adjective agrees with the neuter substance noun in all cases, although it should be noted that for some speakers examples (112b&c) are somewhat marked.

Attributive adjectives modifying N2 can only precede N1 if they are set-denoting adjectives, i.e. adjectives that can usually also occur as the predicate in a copular construction. Placing a non-set-denoting adjective modifying N2 before N1 usually leads to a degraded result.

Furthermore, the attributively used set-denoting adjectives must denote a property of N2; in cases such as (114), where the adjective has a classifying function instead, the adjective cannot precede N1 either.

Finally, it should not be possible to construe the attributively used adjective with N1: in examples such as (115a), the construal of the adjective with N2 is blocked by the fact that it can also express a property of N1; in order to modify N2, the adjective must occur after N1, as in (115b).

Modifying PPs never intervene between N1 and N2, regardless of whether they modify N1 or N2. Consider first the examples in (116): the PPs met een dekselwith a lid and met statiegeldwith a deposit clearly belong to the container nouns doos and krat (which is also clear from the fact that N2 can be omitted), but nevertheless they follow N2. The fact that the PP cannot be placed between N1 and N2 suggests that the PP is actually modifying a phrase containing both N1 and N2, not just N1. If this is indeed true, then the structure of these noun phrases is as indicated in the primed examples.

In the examples in (116), the referential meaning of the N1s is emphasized at the expense of their quantificational force; (116a), for example, does not refer to a quantity of cigars, but simply to a box with cigars; the construction is more or less synonymous with een doos met sigarena box with cigars. Consequently, it is N1 and not N2 that functions as the semantic head of the examples in (116). This is also clear from the fact that examples such as (117), where the verb forces a reading in which N2 acts as the semantic head of the QBC, are semantically anomalous when a PP-modifier of N1 is present.

Since modification of N1 by a PP suppresses the quantificational meaning of N1, we expect that purely quantificational nouns cannot be modified by a PP: that this proves to be true is clear from the fact that the examples in (118) only allow an interpretation in which uit die pot/fles modifies N2, which is clear from the fact that N2 cannot be omitted. However, since we have seen that the PP can also modify the complete QBC, one might want to argue that these examples can be ambiguous between the structures in the primed and doubly-primed example.

Whatever the conclusion about the structure of the examples in (118a&b), it seems that the analysis suggested in the doubly-primed examples is not available when N1 is referential. This is made clear by the examples in (119). Despite its complexity, example (119a) seems acceptable: the PP zonder pitten must be interpreted as a modifier of N2, and met een deksel as a modifier of N1. Changing the order of the two PPs, as in (119a'), makes the construction completely unacceptable, which would follow immediately if we assume that the PP modifying N2 is embedded in the noun phrase headed by N2, as in structure (119b), but not if we assume that it is external to a phrase containing both N1 and N2.

Like PP-modifiers, relative clauses never intervene between N1 and N2, regardless of whether it is N1 or N2 that is modified. Some examples are given in (120): the relative clauses in these examples can only be construed with the container nouns doos and krat, which is clear from the fact that N1 triggers singular agreement on the finite verb of the relative clause, and from the fact that N2 can be omitted. Nevertheless, the relative clauses must follow N2. The fact that the relative clause cannot be placed between N1 and N2 suggests that it modifies a phrase containing both N1 and N2, not just N1. If this is correct, the structure of these noun phrases is as indicated in the primed examples.

In (120), the referential meaning of the N1s is emphasized at the expense of their quantificational force. This explains why examples such as (121), where the verb forces a reading in which N2 acts as the semantic head, are semantically anomalous when the relative clause is present.

Since modification of N1 by a relative clause suppresses the quantificational meaning of N1, it is to be expected that purely quantificational nouns cannot be modified: that this is true is shown by the fact that the examples in (122) only allow an interpretation in which the relative clause modifies N2. This is clear not only from the semantic interpretation, but also from the fact, illustrated in (122a), that it is N2 that triggers number agreement on the finite verb in the relative clause, and that it is N2 that triggers gender agreement on the relative pronoun in (122b). Note that example (122b) with the relative pronoun die improves when the indefinite article is replaced by the definite article de, which is of course due to the fact that N1 is then construed as a referring expression.

Given that the relative clause can in principle modify the complete QBC, one might want to claim that the examples in (122) are ambiguous, and can be associated with either the structures in the primeless or the primed examples in (123).

There is reason to believe that both structures are indeed available. First, recall from Section 18.1.1.3, sub IIA, that purely quantificational N1s cannot normally be preceded by a definite article, but that this becomes possible when the QBC is modified by a relative clause; this is illustrated again in (124).

We also showed in that section that this is a more general phenomenon: proper nouns like Amsterdam, which normally do not license a definite article, can be preceded by one when they are modified by a relative clause; cf. het Amsterdam *(dat ik ken uit mijn jeugd)the Amsterdam *(that I know from my childhood). The crucial point is that the definite article is licensed in the antecedent of the relative pronoun, and this suggests that in (124b) it is the full QBC that acts as the antecedent of the relative pronoun: the definite article precedes N1, not N2. This suggests that the structures in the primed examples in (123) are possible alongside the primeless ones.

It seems, however, that the primed structures are not available when N1 is referential. This can be seen from the examples in (125). Despite its complexity, example (125a) seems acceptable: the first relative clause must be construed with the N2 sinaasappelen and the second with the N1 kistje, which is clear from the fact that they agree with the respective relative pronouns in number/gender. Changing the order of the two relative clauses, as in (125a'), leads to unacceptability, which would follow immediately if we assume that the relative clause modifying N2 is embedded in the nominal projection headed by N2, as indicated in (125b), but not if we assume that it is external to a phrase containing both N1 and N2.

For completeness’ sake, note that the same order restriction seems to hold when the modifiers are a PP and a relative clause, respectively. The examples show that the modifier of N2 always precedes the modifier of N1; example (126b') is of course possible but not under the intended reading that the oranges are from Spain.

This section has shown that both N1 and N2 can be modified. When N1 is modified, it appears that the entire QBC is within the scope of the modifier. When N2 is modified, either the entire QBC or the projection of N2 can be in the scope of the modifier, depending on the status of N1: if N1 is purely quantificational, both structures seem to be available; if it is referential, the scope of the modifier seems to be limited to the projection of N2.