Haeseryn et al. (1997:413) suggest that the -s ending functions as a nominalizing affix; consequently the adjective in the partitive genitive construction is seen as a deadjectival noun; cf. also Van Marle (1996). Deadjectival nouns ending in -s do indeed occur in Dutch. Some examples are: mooisbeautiful-s’, nieuwsnew-s’, lekkerstasty-s’, and fraaisbeautiful-s’. However, unlike the adjectives in (1a&b), these nouns have undergone idiosyncratic changes in meaning. This is quite clear in the case of nieuws and lekkers: the English translations show that the partitive genitives in (2a&b) retain their full meaning, while the corresponding nominalizations in the primed examples have undergone a specialization of meaning.

The deadjectival nouns in the primed examples in (2) exhibit various typical properties of ordinary nouns: (3a) shows that they can occur with or without a determiner, with the concomitant definite and indefinite interpretations; (3b) that they can be combined with a demonstrative; and (3c) that modification by an adjective is possible.

However, claiming that partitive genitives are nouns runs into several problems. First, we would expect that all adjectives ending in -s can be used without the preceding quantificational pronoun iets: however, the examples in (4) show that they generally require the pronoun to be present.

Second, the partitive genitives and the deadjectival nouns in (2) behave differently with respect to modification by a degree modifier: the examples in (5) show that such modification is possible with partitive genitives but is excluded with nominalizations. This strongly suggests that the partitive genitives are adjectives, while the other forms are truly nominal; cf. also the discussion of (3).

The claim that the partitive genitive is a noun would also lead to the wrong conclusion that the quantificational pronoun iets can be combined with a noun; this pattern usually yields an unacceptable result; cf. (6).

A final argument against the claim that partitive genitives are nouns would be that we would expect them to behave like the lexicalized nouns nieuwsnews and lekkerscandy in that they can be combined with a determiner such as the definite article hetthe or the demonstratives ditthis and datthat; we only discuss the neuter forms here because the lexicalized nouns are also neuter. This is generally considered impossible; examples are given in (7).

Note, however, that the examples in (7) improve considerably and are actually quite common when they are preceded by the predeterminer alall. The examples in (8) illustrate this for noun phrases with the definite article het with some examples from the internet. It seems that such examples are typically (but not always) modified by a restrictive PP-modifier or a relative clause. This does not hold for similar examples with a demonstrative pronoun like al dit/dat interessantsall these/those interesting things, al dit/dat engsall these/those scary things, and al dit/dat groensall these/those green things; a brief excursion on the internet shows that such cases can easily be used without a restrictive modifier. Finally, similar constructions do not seem to occur with possessive pronouns at all; cf. *al zijn interessants/engs/groens (lit.: all his A).

The difference between the examples in (7) and (8) may be related to the fact that the latter do not necessarily refer to a contextually determined set of entities. However, since the contrast between the two sets of examples seems to have gone unnoticed so far, we must leave it for future investigation.

Kester (1996) proposed that the partitive genitive is in fact a common attributively used adjective; it modifies a phonetically empty noun [e], the presence of which is indicated by the -s ending on the adjective. Accordingly, the structure of the partitive genitive construction is as in (9).

This proposal can be supported by the fact that partitive genitives are similar to attributively used adjectives in terms of the internal structure of the AP. Section 27.3, sub IB, has shown, for example, that an attributive adjective must be preceded by its PP-complement and (10) shows that the same thing holds for a partitive genitive; note that we have given the clause in (10b) in embedded order, to avoid the interference of PP-extraposition.

That partitive genitives are similar to attributively used adjectives is also clear from the examples in (11); if the adjective does not allow the order PP-A, the adjective is excluded both in attributive position and in the partitive genitive construction.

Note in passing that examples like (12) are only apparent counterexamples to the claim that the PP must occur preadjectivally in the partitive genitive construction: the fact that the partitive adjectives can be omitted in the primeless examples suggests that the PP is not directly related to the adjective, but functions as a modifier of the complete noun phrase.

A second reason for drawing a parallel between attributive and partitive genitive adjectives is the similarity in distribution of the -e inflection on the attributive and the -s ending on the partitive genitive: with adjectives coordinated in attributive position, the inflectional -e ending appears on all of them; likewise, with adjectives coordinated in the partitive genitive construction, the -s ending appears on all of them. This is illustrated in (13).

Of course, this does not apply to the complex adjectives in example (14). Like the attributive -e ending, the partitive genitive -s suffix is only expressed on the final adjective.

A third argument in support of Kester's analysis is based on the cases in (15), in which the element specifiekspecifically is taken to act as the modifier of the adjective christelijkChristian. Some speakers allow the modifier to bear the attributive -e ending, as in (15a'), and it has been claimed that it can also have the partitive genitive -s ending, as in (15b').

However, there may be two flaws in this argument. First, although the modifiers heel/ergvery, are accepted by most speakers with the inflectional -e ending in examples such as (16a) (cf. Section 25.1.2, sub I), they never occur with the partitive genitive -s ending in examples such as (16b).

Second, the claim that (15a') involves modification of the adjective may actually be incorrect; according to our informants who (marginally) accept it, this example has a stacked adjective reading instead of a modification reading; cf. the discussion of example (19b) in the next subsection. Actually, the fact that (16b) is unacceptable if heel is inflected can be seen as an argument against the analysis in (9), according to which the partitive genitive functions as an attributive adjective. Another problem with this proposal is that it leads to the same conclusion as the nominalization approach, namely that the quantificational pronoun iets can be combined with a noun, which is not possible in other cases; cf. (6).

The N-movement analysis, which is due to Abney (1987), is similar to the empty noun analysis in that it assumes that the partitive genitive is followed by an empty noun, but differs from it in that it assumes that the empty noun is not directly inserted into its surface position, but arrives there by movement. More specifically, the analysis assumes that the constructions in (17a) and (17b) are closely related; (17b) is derived by moving the noun iets into the determiner position occupied by the article een in (17a). The representations of (17a&b) are given in the corresponding primed examples, where DP stands for the projection of the determiner (D), and t stands for the trace left by the movement of the noun iets. The –s ending on leuk in (17b) may be a reflex of the N-movement; cf. Corver (2022:fn.55) for a concrete proposal.

A clear advantage of the N-movement analysis in (17b') over the nominalization and empty noun proposals is that it does not imply that the pronoun iets can be followed by a noun phrase in other cases, while it provides a natural account of the facts given in support of the empty noun analysis; cf. the discussion of the examples in (10), (11), and (13). Moreover, the N-movement analysis can easily account for the fact that iets can be premodified in the attributive construction in (17a), in which it functions as a regular noun preceded by the indefinite article een, but not in the partitive genitive construction in (17b), where it occupies the determiner position; we need only note that the article een in (17a) cannot be premodified by an attributive adjective. Similarly, the N-movement analysis accounts for the fact that the partitive genitive construction cannot be preceded by a determiner, since the position usually occupied by the determiner is occupied by iets itself. Note, however, that these facts are not conclusive, since the quantificational pronoun iets cannot normally be premodified or preceded by a determiner in other cases either; we will return to this issue in the discussion of the examples in (37) and (38) in Section 29.2.2, sub I. Finally, the N-movement analysis also provides a natural account of the placement of the degree modifier zoso in the (b)-examples in (18). Section 25.1.3, sub IB, has shown that the modifier zo can either precede or follow the indefinite determiner een in the (a)-examples. As shown in the (b)-examples, it can also precede or follow the noun iets in the partitive genitive construction. Since the noun iets is taken to occupy the same position as een, this similarity follows immediately; cf. Section 29.3, sub IV, for further discussion of these data.

Although we have seen that the N-movement analysis has a number of advantages, there are also some problems. First, consider the examples with coordinated adjectives in (19). Similar examples were given in (13) as evidence in favor of both the empty noun analysis and the N-movement analysis. It should be noted, however, that these data may not provide unambiguous evidence in support of these analyses. As discussed in Section 5.5, sub I, co-occurring attributives can either be coordinated or stacked, as in (19a). In the partitive genitive construction in (19b), on the other hand, the adjectives must be coordinated, as is clear from the fact that leaving out the conjunctive coordinator en/maar is impossible for the vast majority of speakers. In this respect the partitive genitive adjectives are similar to the predicative adjectives in (19c).

Second, if the partitive genitive construction is indeed derived from an attributive construction by movement, we would wrongly expect that all adjectives that can be used attributively are also possible in the partitive genitive construction, i.e. that the primed examples in (20) would be acceptable; cf. Section 29.2.3 for a more thorough discussion.

Third, we need to take into account that the pronoun cannot be the [+animate] quantificational pronoun iemandsomeone. The examples in (21) show that we can replace iets with iemand in the attributive construction in (17a), but not in the partitive genitive construction in (17b). In this respect, Dutch differs crucially from English, which allows constructions such as someone nice. The use of the percentage sign in (21b) indicates that, although all our informants reject such examples, the sequence iemand leuks is easily found on the internet; Corver et al. (2013:122) also mentions that some Dutch dialects do allow the pronoun iemand to occur. It would require further investigation in order to determine whether we are dealing with an ongoing language change or with something else.

Finally, the proposal does not easily account for the fact that the quantificational pronoun iets can be replaced by quantifier nouns such as een heleboel or quantifiers such as veel, since these elements cannot appear as the head of a common noun phrase. Note that (22a') is also a problem for the N-movement approach, because een heleboel is not a head but a phrase.

This section has shown that it is far from clear what the correct analysis of the partitive genitive construction is. We will therefore leave this problem aside, but refer to Broekhuis & Strang (1996), where an analysis is proposed based on the assumption that the noun and the adjective are in a predicative (and not an attributive) relation. Such an approach would be supported by the fact that the adjectives in partitive genitive constructions must be set-denoting, just as in predicative constructions; cf. Section 29.2.3. Hoeksema (1998) has pointed out a number of potential problems with such a proposal, while Schoorlemmer (2005) has suggested that such an approach is tenable, but only for a subset of partitive genitive constructions. We leave the proper analysis of the partitive genitive to future research and focus instead on the properties of the construction that any analysis should be able to account for.