This subsection considers noun phrases in which the head noun is modified by more than one attributive adjective. Two cases can be distinguished, which will be referred to as, respectively, coordination (or symmetric co-occurrence) and stacking (or asymmetric co-occurrence) of adjectives. In the case of coordination, the adjectives are coordinated by the coordinator enand or maarbut, as in (166). In the case of stacking, the adjectives are adjacent, as in (167); no conjunctions are involved.

The presence of the coordinators en and maar clearly signals that the examples in (166) involve coordination of the adjectives; the structure of example (166b), for instance, will be approximately as given in (168a). The examples in (167), on the other hand, do not involve coordination, and the adjectives seem to be in a hierarchical relation to each other; in (167b), for example, the adjective licht modifies the noun tent, and the adjective groot modifies the nominal projection lichte tent. The structure of this example (167b) is therefore roughly the same as in (168b).

Semantically, the distinction between coordination and stacking is often a little fuzzy. For example, (166a) and (167a) seem more or less equivalent: both refer to a car that is good as well as cheap. If we abstract from the fact that the use of maar in (166b) suggests that being both big and light is unexpected for a tent, the same seems to hold for (166b) and (167b): both refer to a tent that is big and light.

We will discuss two arguments for the proposed structural difference. The first argument targets the kind of adjectives that can be combined. Coordinated adjectives usually belong to the same semantic class; for example, the coordinated adjectives in (166) all belong to the class of set-denoting adjectives. Stacked adjectives, on the other hand, can belong to different classes: for example, the adjectives kleinsmall and AmerikaansAmerican in (167c) are a set-denoting and a relational adjective, respectively. Coordinating such adjectives by maar or en, as in (169a), yields an unacceptable result, showing conclusively that adjectives belonging to different classes cannot be coordinated, but involve stacking. Example (169b) provides another illustration of the same phenomenon.

A complication is that some classes of relational adjectives have a tendency to shift their meaning towards the set-denoting adjectives; cf. Section 23.3.3. This may explain why one occasionally encounters examples such as (170), although such cases may also arise as a result of backward conjunction reduction; cf. [een [typisch [e]] en dus [relatief grote auto]] and [een [typisch Nederlandse [e]] maar toch [gulle jongen]]. We leave open which analysis is the better one, but see Subsection II for some arguments against a conjunction reduction analysis.

The second argument involves context-sensitive N-ellipsis; cf. Section 27.4, sub I. In (171) the noun phrases in the second conjunct are syntactically but not semantically reduced. For example, (171a) expresses that Peter has bought a very bad cheap car, which indicates that [e] corresponds to the complex phrase goedkope auto, and (171b) expresses that Peter has a white American car, which indicates that [e] corresponds to the complex phrase Amerikaanse autoAmerican car.

If the co-occurring adjectives in (171) were coordinated, these interpretations would be unexpected: it would imply that N-ellipsis can affect subparts of a coordinate structure. This reduction is indeed impossible if the coordinator enand is present: example (172) does not imply that Peter also has a cheap car, which shows that e corresponds only to the nominal head auto.

Cases of coordination of adjectives, as in (173a), should be distinguished from cases of coordination of noun phrases plus backward conjunction reduction, as in (173b); cf. also Section 27.4, sub I, example (137). The apparent similarity between the two constructions is due to the fact that in (173b) the head noun of the first conjunct is deleted under phonological identity with the noun of the second conjunct.

However, constructions with coordinated APs are syntactically different from conjunction reduction constructions in that there is only one determiner in the former, but two in the latter. The corresponding semantic difference is that (173a) expresses that Jan has only one car, which is both good and cheap, while (173b) expresses that Jan has two cars, one of which is good and one of which is cheap. This semantic difference can be easily demonstrated syntactically by subject-verb agreement when the NPs in (173) are used as subjects, and is further reflected by the presence or absence of the second article; cf. the examples in (174).

Unfortunately, the two syntactic tests are of no use when we are dealing with plural indefinite noun phrases. First, since the article is phonetically empty, there is structural ambiguity in examples such as (175a). Thus, the only way to distinguish the two structures in the (b)-examples is by appealing to meaning: structure (175b) expresses that we are dealing with cars that are both cheap and economical, while structure (175b') expresses that we are dealing with cars, some of which are cheap and some of which are economical.

Obviously, an appeal to subject-verb agreement is not possible either; the subject noun phrases in (176) both trigger plural agreement on the finite verb.

In such cases, however, we can sometimes appeal to various other tests. The contrast between the examples in (177) is related to our knowledge of the world: (177a) is unacceptable because a car cannot be completely black and completely white at the same time; (177b) is fully acceptable because it involves two cars.

It can also be made clear that the two constructions differ by replacing the conjunction enwith the contrastive coordinator maarbut; this is possible in (178a) with the symmetrically coordinated adjectives, but not in (178b) with the conjunction reduction construction. The reason why maar leads to an unacceptable result in (178b) is that maar cannot normally be used to coordinate noun phrases: Ik wil het meisje en/of/*maar de jongen bezoekenI want to visit the girl and/or/*but the boy; note that this would also be an argument against the backward conjunction reduction analysis proposed for the examples in (170a&b). Note that the result in (178b) seems to improve somewhat when we make the second conjunct negative, as in (178b'). However, this may have to do with the fact that this example has a parenthetic flavor and seems to require contrastive accents and additional intonation breaks; cf. Jan heeft een goede, maar geen goedkope, auto. If so, the structure in (178b') may be incorrect: we are actually dealing with a construction containing a (reduced) epenthetic clause. We leave this question for future research.

The disjunctive coordinator ofor can also be used to distinguish the two constructions. Example (179b) shows that of can be used in the conjunction reduction construction, but not in the case of coordinated adjectives. The reason why disjunction is not possible in (179a) is probably semantic or pragmatic: entities are not normally defined by a disjunction of properties.

That the coordinator of can be used with coordinated adjectives if the speaker intends to correct himself. This is illustrated in (180), which is acceptable only if the marker beter gezegd, indicating that it is a correction, is overtly expressed;

Not all attributive adjectives can be coordinated; cf. also Section 5.5, sub I, (169). The constraints on coordination are mainly semantic in nature, involving the distinction between the adjective types in (181); cf. Section 23.3.

We have seen above that set-denoting adjectives can be coordinated without any problem. However, it is not possible to coordinate a set-denoting adjective with an adjective from one of the other adjective classes. This is illustrated in (182): (182a) involves the coordination of a set-denoting and a relational (geographical) adjective, (182b) the coordination of a set-denoting and an evaluative adjective, and (182c) the coordination of a set-denoting and a modal adjective. From the unacceptability of these examples we conclude that coordinated adjectives must be of the same class.

It is not immediately clear whether it is possible to coordinate relational adjectives. An example such as (183a) seems acceptable, but still somewhat marked compared to the construction with backward conjunction reduction in (183a'). If both examples are indeed acceptable, they should differ in the number of sets involved (one in (183a), but two in (183a')), but obtaining reliable judgments seems difficult. The indefinite counterpart of these examples in (183b) is of course compatible with both analyses and so does not shed any new light on the issue.

Given the difficulty of interpreting the (a)-examples, it may be better to restrict ourselves to cases in which the noun is singular, as in (184a&b). Although examples such as (184a) can be found on the internet, we consider them to be marked compared to constructions with backward conjunction reduction, such as the one in (184b).

The contrast between the examples in (184) need not be syntactic in nature, but may have a morphological basis, since the intended reading of (184a) is more easily understood by using the adjectival compound Nederlands-Belgisch: cf. De Nederlands-Belgische afvaardiging. This is even clearer in examples such as (185): coordination, as in (185a), leads to a clearly degraded result, and the only way to express the intended interpretation “a tour through America and Europe” is by using the compound Amerikaans-Europees in (185b). Its compound status is even clearer from the fact, illustrated in (185b'), that the attributive -e ending cannot be assigned to the first member of the adjective pair.

In other cases, the impossibility of having coordinated relational adjectives may have a semantic basis: the degraded status of the examples in (186) is probably due to the fact that the coordinated adjectives are from different semantic subclasses: (186a) involves the coordination of a time and a geographical adjective, (186b) of a “movement/trend” and a geographical adjective, and (186c) of a substance and a geographical adjective.

The evaluative adjectives in (187a) and the modal adjectives in (187b) are also difficult to coordinate, which may be due to the small number of adjectives belonging to these classes. The unacceptability of (187c) is again due to the fact that it involves the coordination of adjectives belonging to different semantic classes: vermeend is a modal and drommels is an evaluative adjective.

While coordination requires the adjectives to belong to the same class, stacking of adjectives from different semantic classes is possible. The following subsections discuss the restrictions on the different combinations of the adjective classes in (181).