The primeless examples in (123) show that N-ellipsis can be licensed by a syntactically realized noun in a preceding sentence, while the primed examples show that N-ellipsis is excluded when [e] precedes the overtly realized noun. The unacceptability of the singly-primed examples is due not to the empty element but to the overtly realized one, as will be clear from the fact that applying N-ellipsis to the latter, as in the doubly-primed examples, will lead to a fully acceptable result if the discourse provides a suitable antecedent for both empty nouns.

The restrictions on the distribution of [e] are similar to those on the distribution of referential personal pronouns. This becomes clear when we compare the examples in (123) with those in (124), where coreference is indicated in italics. Just like [e] in the primeless examples in (123), the pronoun hij in (124a) depends for its interpretation on the direct object in the preceding sentence, and just like the empty noun in the singly-primed examples in (123), the pronoun in (124b) cannot precede its antecedent. Coreference becomes possible, however, when the antecedent is itself a pronoun, as in (124c); cf. the doubly-primed examples in (123).

Note that N-ellipsis is also possible with a syntactically realized neuter antecedent, as opposed to the case in which the antecedent is determined by the non-linguistic context. This will become clear by comparing the examples in (125) with the one in (122b).

With the sentences in (123) coordinated by enand, the judgments remain the same, which shows that N-ellipsis can occur in coordinated CPs. We will see below that N-ellipsis can also occur in coordinated verbal projections that are smaller than complete clauses: (126) involves the coordination of verbal projections with a subject (i.e. IPs), and (127) involves the coordination of verbal projections without a subject (i.e. VPs). We refer the reader to Chapters V10 to V13 for a detailed discussion of the internal structure of clauses, and to Section C39.4.1 for a discussion of clauses of the kind in (127), which are sometimes (incorrectly) referred to as forward conjunction reduction constructions.

In (128) we see that N-ellipsis can also apply to verbal projections phonetically reduced by gapping (indicated by strikethrough).

The examples in (129) show that N-ellipsis does not lead to a fully acceptable result in coordinated noun phrases; we illustrate this here for coordinate structures functioning as subject and object.

The percentage sign is used to indicate that the examples in (129) are rejected by most speakers, while speakers who do accept them still prefer backward conjunction reduction, resulting in structures in which the elided noun precedes the overtly realized noun; cf. [NP [de grote [e]] en [de lichte tent]] and [NP [het sterke [e]] en [het lieve paard]]. N-ellipsis as a result of backward conjunction reduction is discussed in Subsection IC.

Context-sensitive N-ellipsis does not only occur in successive sentences and coordinate structures, it can also occur in subordinate structures. This is shown by the primeless examples in (130). Although some speakers accept the primed examples under the intended interpretation, for most speakers the overt noun must precede the empty noun.

The interpretation of the empty noun is again similar to the interpretation of a personal pronoun in this respect; cf. the discussion of (124). This is illustrated in (131), where coreference is indicated by italics.

The ordering restriction on overt and phonetically empty nouns is not a surface phenomenon. For example, the primeless examples in (132) show that topicalization of the complement/adjunct clauses in (130) does not block N-ellipsis in the subordinate clause. However, topicalization of the subordinate clauses also makes N-ellipsis in the main clause fully acceptable; this is illustrated by the primed examples in (132), which should be compared with the primed examples in (130).

The examples in (133) show that the interpretation of [e] is again similar to the interpretation of the referential personal pronouns in this respect: we refer the reader to N22.1 for more discussion of the conditions on the interpretation of the personal pronouns.

Context-sensitive N-ellipsis can also occur in simple clauses. This is illustrated by the two primeless examples in (134), in which the subject triggers N-ellipsis in a direct/prepositional object and a prepositional complementive, respectively. Unlike what is the case in the complex sentences in (130), however, it seems occasionally possible for the empty noun to precede the overtly realized one in simple clauses; this is illustrated by the primed examples.

The cases of N-ellipsis in the primeless and primed examples of (134) seem to behave differently in several respects. For example, the primed examples require a special intonation contour; the contrastive accent (indicated by small capitals) must be placed on the attributive adjectives. Another striking difference is that the overt and empty nouns in the primeless examples need not have the same number, whereas this seems to be required in the primed examples; this is demonstrated in (135) for cases with the neuter noun paardhorse, which takes the determiner het in the singular and the determiner de in the plural; this means that we are dealing here with one old horse and two or more young horses.

Traditionally, the cases in (143) are considered to be cases of nominalization. There are at least four facts that support this position. First, the examples in (149) show that pluralization of the de-group is possible, which is typically a nominal and not an adjectival property. Note that this argument is mainly based on orthographic convention, since the plural -n is not pronounced in standard Dutch, and that the het-group does not provide the same evidence, since it consists only of non-count noun phrases.

Second, the -e affix also appears on adjectives that do not allow attributive inflection; examples are the adjectives in (150) that orthographically end in -en. This ending is normally pronounced as schwa, whereas in the primeless examples it is realized as [ən]; the primed examples show that the form with –e cannot be used in attributive position.

Note that if the adjective orthographically ends in -e (schwa), the -e affix is absent; the /n/ in (151) is the plural ending, which should not be confused with the /n/ of besproken in the examples in (150).

Third, the examples in (152) were given as independent evidence for the fact that the -e ending can be used as a nominalizing affix. The alternative of treating it as an attributive inflection runs counter to the fact that possessive pronouns normally do not inflect. A nominalization approach would also take into account the fact that the resulting forms can be preceded by a determiner. Note that the form jullie cannot be nominalized, which may be related to the observation in Subsection B that this prohibition also applies to adjectives ending in /i/.

		singular	plural
1st person		de/het mijne	de/het onze
2nd person	colloquial	de/het jouwe	*de/het jullie(-e)
	polite	de/het uwe	de/het uwe
3rd person	masculine	de/het zijne	?de/het hunne
	feminine	de/het hare
	neuter	de/het zijne

The examples in (152) are perhaps not entirely comparable to the cases of context-insensitive N-ellipsis discussed here, since their interpretation is generally context-dependent; cf. (153a). This is not the case, however, for their plural counterparts in (153b), which are typically used to refer to a particular set of people.

Finally, the examples in (154) pose a serious problem for the N-ellipsis analysis to be discussed below, since the postulated empty nominal element [e] cannot be replaced by a phonetically realized one. Since the nominalization approach does not postulate such an empty element, these facts indirectly support this approach.

The arguments given in the previous subsection provide important evidence for a nominalization approach to the constructions in (143). However, there are also arguments against such an approach and for an N-ellipsis approach, according to which we are dealing with genuine attributively used adjectives in these cases.

A first argument concerns pluralization and could be interpreted as a counterargument to the first argument for nominalization. Although nouns ending in a schwa can often have either a plural -s or a plural -n morpheme, the former seems to be more common than the latter (perhaps because the latter is not pronounced in speech). However, the primeless examples in (155) show that the cases under discussion categorically resist a plural in -s; the primed examples in (155) illustrate the regular form of pluralization for comparison.

This argument against nominalization is probably not very strong, since a noun like waardevalue does not allow the plural -s either: waarden versus *waardes. The same is true for a group of nouns ending in the phonetic sequence [idə], which also do not allow the plural -s for many speakers: druïdendruids versus %druïdes. Therefore, it may be that, like these nouns, the putative nominalizing affix -e simply has the idiosyncratic property of selecting the plural -n morpheme.

Another morphological difference between the constructions in (143b) and simple nouns ending in schwa is that the former resist diminutive formation, whereas the latter generally allow it; cf. the contrast between the primeless and primed examples in (156). This follows naturally from the N-ellipsis analysis, since adjectives cannot be input for diminutive formation.

There are also syntactic arguments for the N-ellipsis approach to the constructions in (143) and against a nominalization approach. First, the allegedly nominalized adjectives in (157) can be modified by degree modifiers like onweerstaanbaar (157a) and erg/zeervery (as is clear from the absence of the attributive -e inflection in these forms); since degree modification of nouns is usually not possible, as shown in *de erg/zeer kapitalisten (lit.: the very capitalists), we should conclude that we are dealing with adjectives in (157).

The fact that modification by an attributive adjective is also possible is not relevant in this context; this is expected in either approach, since we may just be dealing with stacked adjectives. Compare:

A second syntactic argument for N-ellipsis concerns PP-complements of deverbal adjectives such as afhankelijkdependent. Section 27.3, sub IB1, has shown that such complements must precede the adjective in attributive position: it is impossible to place the PP-complement between the adjective and the noun or after the noun. If we assume an N-ellipsis analysis, we immediately account for the fact that the PP van een uitkering usually precedes afhankelijken in (159b); it is unexpected for the PP-complement of a (deadjectival) noun. Note that the percentage sign indicates that we found some cases of this order on the internet.

As expected, when the adjective is associated with a phrase that can follow the noun, it can also follow the (supposedly) nominalized adjective. This is demonstrated in (160) with adjectives in their equative, comparative and superlative use, which can be combined with postnominal als, dan and van-phrases, respectively.

More cases of a slightly different nature are given in the primeless examples in (161), which should be compared with the primed examples.

Finally, examples such as (162a) have also been used to refute the nominalization analysis: since the adjectival participle geplaatste is combined with an argument (a locational PP), nominalization would have to take a phrase as input. However, since (162b) shows that inf-nomalizations such as plaatsen clearly retain their selection restrictions, the same may be true for the nominalized participle in (162a).

The facts in (157) to (161) follow immediately if we assume N-ellipsis, while it is not clear how they could be accounted for under the nominalization approach.

Corver & Van Koppen (2011) claims that the –e on the adjective is not morphological in nature, i.e. it is neither a nominalizing affix nor the attributive –e inflection. Instead, it is taken to involve pronominalization: e (as well as its plural counterpart en) is a pronominal clitic corresponding to English one in the rich ones, which is phonetically supported by the preceding adjective. So far, this analysis has rested mainly on comparative data from Dutch dialects but there is at least one striking standard Dutch fact that supports this proposal. Consider again the examples in (150), repeated here as (163), which were given earlier in support of the nominalization approach.

Since besproken cannot be inflected in attributive position, the –e must have some other function. Both the nominalization and the pronominalization approach readily account for this; the former analyzes –e as a nominalizing affix, while the latter analyzes it as a pronominal clitic: cf. the representations in (164). The N-ellipsis approach does not seem to have an adequate answer to this challenge.

Several of the arguments used above in support of the nominalization and N-ellipsis approaches can also be used to support the pronominalization approach. For example, Corver & Van Koppen (2010) has already shown that the analysis in (164b) can also be used to account for inflected possessive pronouns such as mijnemine, which is analyzed as [mijn [N epron]] in specific Dutch dialects (although, surprisingly, the possibility of applying this analysis to the standard language is not considered); cf. also the introduction to this section. Furthermore, the pronominalization approach would also account for the fact, already illustrated in (157) and (159), that the putative nominalized adjective can be modified by a degree adverb in de erg/zeer rijkenthe very rich and preceded by a PP-complement in de van een beurs afhankelijken, since this approach assumes that we are dealing with regular attributive constructions; cf. the representations in (165).

This shows that there are good reasons to believe that the pronominalization approach might combine the favorable aspects of the two previous approaches. Since this is not the place to develop this approach further, we must leave it to future research.