When an adjective with a dative complement is used attributively, the head noun must be the subject of the corresponding copular construction, while the dative acts as a complement of the adjective. This is illustrated in (81); example (81b') may be slightly marked with a non-pronominal phrase like Jan or de directeurthe manager.

While the examples in (81) show that if the dative object is optional in the copular construction, it is also optional in the attributive construction, the examples in (82) show that if the dative object is obligatory in the copular construction, it must also appear in the attributive construction. Note that, like (81b'), example (82b') is somewhat marked with a non-pronominal phrase like Jan or de directeurthe manager.

When the dative noun phrase alternates with an aan-PP in the copular construction, this is also an option when the AP is used attributively. It seems that the dative phrase is preferred when the complement is pronominal, while the aan-PP is preferred when it is non-pronominal, but the judgments seem subtle. Note that the complement of the adjective cannot occur in postnominal position; cf. *de gehoorzame/trouwe hond aan Peter/hem.

The examples in (84) provide cases in which the dative phrase is licensed by the evaluative degree element tetoo; cf. Section 25.1.3, Sub II. The primed examples show that the dative noun phrases in the primeless examples alternate with voor-PPs, which again seem to be preferred when the dative phrase is non-pronominal.

Again, the complement of the adjective does not seem comfortable in postnominal position; cf. ??een te moeilijk boek voor Peter/mij and ??een te koud zwembad voor Peter/mij. In this case, however, it is somewhat harder to judge, because voor-PPs can also occur as restrictive adverbial phrases; cf. Dit is een te moeilijk boek voor mij, which alternates with Voor mij is dit een moeilijk boekFor me this book is too difficult.

The examples in (85) show that adjectives cannot easily be used attributively when they take a genitive argument (although German is known to be more permissive in this respect). The marked status of these examples may be related to an observation in the previous subsection, namely that dative phrases also seem less acceptable when they are non-pronominal. Since genitive phrases typically express new information, they usually cannot be pronominalized, which results in lower acceptability.

Unlike what we found with dative complements, the examples in (85) cannot be saved by realizing the genitive phrase as a van-PP; this is illustrated in (86) for those examples that allow the alternation. Note that placing the van-PP in postnominal position has a greater degrading effect.

The examples in (87) demonstrate that the attributive use of fixed expressions like het spoor bijster zijnto be lost/confused, de stad meester zijnto be in command of the city, or iets kwijt zijnto have lost something is excluded.

The only way to express the intended meaning is to use the present participle of the copular verb zijnto be, which results in the rather formal or sometimes downright clumsy constructions in (88).

The examples in (89) and (90) show that, unlike most adjectives, pseudo-participles like verliefdin love and deverbal adjectives like afhankelijkdependent can either be preceded or followed by their PP-complement. The PP-complement is clearly part of the AP, since it can be pied-piped by topicalization of the AP (the constituency test), and from the fact that the preadjectival PP is located between the modifier ergvery/volkomencompletely and the adjective; cf. Sections 24.3.1, sub III, and 26.3.1 for discussion.

We conclude that in both orders the adjective and the PP are part of a single AP; consequently, we would expect that the APs in (89) and (90) can be used attributively in both orders. However, the examples in (91) show that this expectation is not fully borne out: the APs can be used attributively only when the PP precedes the adjective.

This has led to the conclusion that attributive adjectives must be immediately adjacent to the nominal projection N# that they modify; the configuration [... [AP ... A XP] N#] is excluded if the string XP between the attributive adjective and the head noun is non-null. This constraint is known as the head-final filter on attributive adjectives, since it effectively requires the adjective to be the rightmost element in its own projection. Note that we cannot simply say that the adjective must be immediately adjacent to the noun it modifies, since this would incorrectly preclude the stacking of adjectives, as in de mooie blauwe stoelthe beautiful blue chair; cf. Section 27.5 for discussion.

As can be seen in (93), the adjective gekfond must be followed by the PP op zijn vrouwof his wife. The constituency test in the primed examples shows that the string gek op zijn vrouw must again be considered a constituent, namely an AP.

Given the head-final filter on attributive adjectives in (92), we expect that this AP cannot be used attributively: the order gek op zijn vrouw violates (92), and the order op zijn vrouw gek is impossible under any circumstances. That this expectation is borne out is illustrated in (94); note also that a Google search (11/15/2022) on the strings [een/de op * gekke] did not yield any relevant result.

Adjectives such as trotsproud, tevredensatisfied and bangafraid give rise to a slightly degraded result in constructions comparable to (93b'); cf. Section 24.3.1, sub IIB. The examples in (95) show that these adjectives may lead to a slightly marked but acceptable result in attributive position when they are preceded by their PP-complement, but the result is clearly unacceptable when the PP-complement follows the adjective; this contrast follows from the filter in (92).

Since the base position of the PP is clearly to the right of the adjectives, the examples in (95) should be the result of moving the PP to an AP-external position. A first argument for this claim is that the PP must precede the degree modifier of the adjective, as shown by the examples in (96).

The examples in (97) provide further support for the claim that the PPs in (95) are external to the AP by showing that they also precede AP-external adverbial material such as altijd, voortdurend, and nog steeds; note that the markedness of these examples seems somewhat greater than in the corresponding examples in (95).

The discussion above has shown that the internal structure of attributively modified noun phrases is more complex than one would expect at first glance. We may even have to assume that noun phrases are like clauses in the sense that their internal linear order is the result of DP-internal movement (here: the PP-complement of the AP).

Section 24.1, sub II, has shown that some adjectives can take a clausal PP-complement. Consider the two primeless examples in (98), which require the anticipatory pronominal PP-complement of the adjective to be present. The primed examples show that these adjectives cannot be used attributively. Of course, the unacceptability of (98a') may be due to the head-final filter on attributive adjectives in (92), since the adjective ziek is separated from the head noun by the stranded preposition van. However, since this filter does not account for the unacceptability of (98b'), in which the stranded preposition precedes the attributively used pseudo-participle gekant, the unacceptability seems to be due to the presence of the (in this case finite) clausal complement; a Google search (July 2023) yielded three cases of the form [NP ... ertegen gekante N] without a clausal complement.

The anticipatory pronominal PP of adjectives such as boosangry and tevredensatisfied in (99) can optionally be omitted. The fact that the attributive use of these adjectives is excluded regardless of the presence or absence of the pronominal prepositional phrase shows again that the impossibility of using these adjectives attributively is apparently due to the presence of the clausal complement.

Although we have seen that there are good reasons for attributing the unacceptability of the primed examples in (98) and (99) to the presence of the clausal complement, the examples with the split anticipatory pronominal PP are also unacceptable in case the clauses are omitted; adjectives like boos and tevreden can only be used in attributive position if the split anticipatory pronominal PP is absent, as in (100a). The fact that (100b), in which the pronominal PP is moved leftward, is acceptable does not affect the matter at hand, since anticipatory pronominal PPs always have the form er + P, i.e. we are dealing with a pronominalized PP-complement of the form discussed in Subsection 1 above.