One striking difference between appositive and attributive adjectives concerns their size. Appositives are complex APs in the unmarked case: an adjective such as verliefdin love usually requires its PP-complement to be present, as in (194a); if its complement is not present, as in (194a'), the adjective usually gets emphatic accent. The (b)-examples, on the other hand, show that the complement is optional when verliefd is used attributively. A second difference concerns word order. Example (194c) again illustrates that an adjective such as verliefd can be preceded or followed by its complement when it is used as a complementive; cf. Sections 24.1. The examples in (194a&b) show that the same holds when verliefd is used appositively, but not when it is used attributively (cf. the head-final filter on attributive adjectives in Section 27.3, sub IB).

The primeless examples in (195) show that “heavy” APs with a complex modifier such as zo ... dat ...so ... that ... can also be used as appositives, while the primed examples again show that the appositive use of a “bare” adjective such as blijhappy yields a marked result compared to its attributive use in het blije meisjea happy girl and een vieze jasa dirty coat.

If we are dealing with comparison, the adjective can optionally be accompanied by a dan/als/van-phrase, which must follow the adjective in the predicative constructions in (196); cf. Section 26.1.3, sub V.

The head-final filter on attributive adjectives correctly predicts that such APs modified by a dan/als/van-phrase cannot be used in attributive position. However, there is a “repair” strategy that places the dan/als-phrase after the head noun; cf. Section 27.3, sub IIA. We can see this in the examples in (197).

Alternatively, the whole AP can be used as an appositive, as shown in (198). Note in passing that the appositive in (198b) can be replaced by a nominal one: Deze begeleider, de beste van de staf, zorgt voor de zwakste studentenThis supervisor, the best one of the staff, takes care of the weakest students.

We should emphasize that the dan/als/van-phrases are part of the predicatively used APs in (196), as is clear from the fact, illustrated in (199), that they can be pied-piped by topicalization of the adjective; cf. the constituency test.

This leads to the conclusion that the dan/als/van-phrases in (197) are also part of the attributive APs, and that their postnominal placement is the result of some movement operation: if we assume that the AP is base-generated in postnominal position, we can derive the attributive construction by leftward movement of the adjective across the noun, while stranding the dan/als/van-phrase (alternatively, we could assume that the AP is base-generated in prenominal position and that the dan/als/van-phrase has been moved rightward across the noun but we will not consider this option here). If so, this would make it possible to assume that the appositive constructions in (198) have the same underlying structure as the attributive ones in (197), by appealing to the head-final filter on attributive adjectives: one could then claim that the attributive construction is derived by movement of a phrase minimally containing the adjective and its complement, i.e. by stranding of the adjunct als/dan/van-phrase.

The claim that the head-final filter on attributive adjectives blocks the movement in (200a') but not in (200b') then correctly accounts for the fact that an adjective such as gekfond in (201a), which obligatorily precedes its PP-complement, cannot be used attributively; cf. (201b) and Section 27.3, sub IB. The intended meaning of (201b) must therefore be expressed by using the AP as an appositive, as in (201c).

Analyses of the kind in (200) raise non-trivial questions about the size of the moved phrase, since it is not the case that all modifiers of the adjective can be stranded. Example (202b), for instance, shows that the modifier genoeg blocks the attributive use of the modified adjective regardless of whether it is stranded or pied-piped; cf. Section 27.3, sub IIB, for a more detailed and careful discussion. Consequently, the only option is to use the complex AP appositively, as in (202c), an attested case (audio recording Paulus vangt een vos by Jean Dulieu, 1968).

The (a)-examples in (203) and (204) show that the attributive use of adjectives with a clausal complement is also blocked; cf. Section 27.3, sub IB2. This holds regardless of whether (the stranded part of) the anticipatory PP precedes or follows the adjective. The corresponding appositive constructions are acceptable; note, however, that the degraded status of (203c) suggests that the anticipatory PP cannot be split.

We will not discuss these problems with stranding/pied piping any further here, because the next subsection will show that the hypothesis that attributives and appositives have the same underlying structure encounters a number of even more troublesome problems.

There are several problems with the hypothesis that attributives and appositives have the same underlying structure. A first problem is that appositives and attributives differ in that only the former can modify pronouns and proper nouns. Some acceptable examples of appositives are given in (205). Note that examples like een bange Jan or een tevreden Marie are possible, but that the presence of the indefinite article suggests that the proper nouns are used as common nouns.

The second problem is more semantic in nature. Consider the examples in (194a&b), repeated in a slightly different form as (206a&b). Section 23.3.2 has shown that the noun phrase in (206a) refers to the intersection of the set denoted by the noun man and the set denoted by the AP op zijn vrouw verliefd. In addition, the definite determiner indicates that this intersection has one member in the given discourse domain. The noun phrase in (206b), on the other hand, refers to a known male person in the discourse domain, about whom it is said that he is in love with his wife; in this respect, the appositive behaves like the non-restrictive relative clause in (206c). The fact that (206a) and (206b) differ in interpretation in this way seems incompatible with the claim that they have the same underlying structure.

Another difference in meaning is found in the examples in (207) and (208). Example (207a) is ambiguous between a “one-set” reading, according to which the cars that sold well have the property of being both old and cheap, and a “two-set” reading, according to which both the old cars and the cheap cars sold well. The attributive construction in (207b), on the other hand, only has the “one-set” reading.

The difference between the appositive and the attributive constructions is even clearer in the examples in (208), where the antonymous adjectives oudold and nieuwnew block the “one-set” reading, because this would lead to a contradiction. As expected, the appositive construction in (207a) now only allows the “two-set” reading, according to which both the old cars and the new cars sold well, while the corresponding attributive construction in (207b) is unacceptable, because it only allows the semantically anomalous interpretation that all cars that sold well are old as well as new.

Note that the relative clauses in (209) in this case are like the attributive construction; because the relative pronoun acts as the logical subject of both conjoined predicative APs, the sentence expresses a contradiction.

The examples in (210) and (211), which contain an attributively used comparative, illustrate a third difference. In the attributive constructions in (210), the comparative can be complemented with the dan-phrase dan ik (heb)than I (have), but not with dan het mijne (is)than mine (is).

If the appositive and the attributive have a common source, we expect the same thing to be the case in the corresponding appositive construction. However, example (211) shows that this expectation is not borne out. The use of dan het mijne (is) is perfectly acceptable, while the use of dan ik heb is unacceptable. Note that example (211a) is acceptable without the verb hebbento have, but that this leads to the pragmatically strange interpretation “Jan has a house that is bigger than I am”, which is not available in (210a).

In these comparative constructions, relative clauses again behave like appositives, not attributives. This is illustrated in (212).

The general picture that seems to emerge from the discussion above is that appositive constructions are used when the attributive use of the AP is blocked for some reason. At first glance, this seems to provide support for the claim that the appositive and attributive constructions are alternative realizations of the same underlying structure. However, the last subsection has shown that there are various semantic and syntactic problems for such a proposal. We have also seen that appositives resemble non-restrictive relative clauses in several respects, which suggests that appositions are propositional in nature; an analysis emphasizing the propositional nature of appositions is given in Heringa (2012).