The use of approximative adverbial phrases is not the only way to express the notion of approximation. Dutch has a wide variety of ways of expressing this notion. The examples in (67), for instance, show that there are a number of adjectives that can serve this function; like the approximative adverbs, they precede the cardinal that they modify, which would of course follow directly if the internal structure of the bracketed phrases were as follows: [[een A cardinal] [Num [NP ... N ...]]].

The examples in (67) seem to be related to (68a), where no adjective is used and the indefinite article een (i.e. the form pronounced as [ən]) expresses the approximative meaning all by itself; the adjectives in (67) can probably be seen as modifiers that make the approximative reading more specific: the adjective klein indicates that the cardinality is slightly lower than 100, while the adjective dik expresses that the cardinality is slightly higher. The non-D-linked demonstrative zo’n can be used with a similar effect in (68b), although it seems that the adjectives in (67) cannot be easily used in such examples (although they do occur on the internet): %zo’n kleine/dikke honderd boeken.

Note also that the approximative reading of een and zo’n can be enhanced by using a juxtaposition of two cardinals, as in (69).

The suggestion that the adjectives in (67) modify the meaning expressed by the indefinite article is further supported by the fact, shown in (70), that the indefinite article cannot be replaced by a definite one, which suggests that the adjectives do not express an approximative meaning all by themselves.

However, Haeseryn et al. (1997) note that the adjectives dik and goed in (67b) are also used in speech without the article een. Although examples such as (71a) sound marked to us, especially with goed, they are easy to find on the internet. We therefore conclude that, at least for speakers who accept and produce such examples, these adjectives can express an approximative meaning in themselves. However, this does not seem to hold for klein; (71b) is certainly not acceptable.

The constructions in (68) seem to alternate with the constructions in (72), where the cardinal is expressed in a postnominal of-phrase. It is not clear to which word class of belongs; of is usually a conjunction or an interrogative complementizer, but neither of these functions seems appropriate for these examples. Nor is it possible to add an approximative modifier to such constructions, although the approximative reading can be enhanced by using cardinals in juxtaposition.

Example (73a) is another way of expressing the same meaning. The phrase een stuk of honderd in this construction seems to have more or less the same structure as the phrase in (72a) and acts as a complex cardinal modifying the noun boeken. Although we consider the use of zo’n in this construction to be marked, examples such as (73b) are easy to find on the internet, so we have marked it with “%”.

Despite the fact that the examples in (72a) and (73a) express more or less the same meaning, they exhibit quite different syntactic behavior. The first difference concerns subject-verb agreement. Since the noun preceding the cardinal in (72) is singular, it is not surprising that this noun phrase triggers singular agreement on the verb in (74a), although plural agreement seems possible for some speakers. Example (74b), on the other hand, shows that this is different with the noun phrase in (73a); the finite verb does not agree with the singular noun stuk, but with the plural noun boeken, a phenomenon common to several types of binominal constructions; cf. Section 18.1.1.

The second difference concerns the modification of the noun. The examples in (75) show that the noun boek can be modified when it is part of the structure in (73a), but not when it is part of the structure in (72a).

The third difference and concerns the fact that in (73a) the cardinal can be replaced by the modifier watsome, whereas this is impossible in (72a). The resulting structures are given in (76)

Corver (2001) suggests that the position of the of-phrases in the examples in (72) indicates the base position of cardinals in general: more specifically, cardinals are inserted in postnominal position as nominal predicates and are subsequently moved into prenominal position as a result of (obligatory) predicate inversion. This proposal may provide a better understanding of examples such as (77a), where the sequence tien of zo forms a semantic unit (meaning “approximately 10”) but is syntactically split. This would follow if we assume that the string tien of zo is base-generated in postnominal position and that the cardinal is subsequently moved into prenominal position with obligatory stranding of the of zo part, as in (77b). For the empirical motivation (partly based on cross-linguistic evidence) and the technical implementation of this analysis, see Corver (2001) and Corver & Zwarts (2005:§4).

The rejection of an idiomatic analysis in favor of the analysis in the primeless structures in (81) accounts for the syntactic productivity of the construction; many prepositions can appear in the construction. Example (82) divides the relevant prepositions into three semantic subclasses.

The examples in (83) show that the original meaning of the prepositions is easily recognized.

In addition, example (84) shows that the selection restrictions of the prepositions are also preserved: tussenbetween is followed by a coordinated phrase.

The syntactic productivity of the construction is also clear from the fact that it is sometimes possible with non-simplex prepositions.

There are also prepositions that are not possible in the approximative PP construction; examples are voorin front of and naastnext to. Corver and Zwarts (2005:§3.2) argue that this is related to the fact that the numerical dimension is simpler than the spatial dimension; the former is one-dimensional, i.e. defined only in terms of linear order and distance. Since the number scale is metaphorically related to verticality, we do not find “horizontal” prepositions like voor and naast next to “vertical” prepositions like bovenabove and onder below. In short, this restriction on the lexical choice of prepositions is not an indication of the idiomatic status of the construction, but follows from independently motivated semantic properties of the numerical system.

The examples in (86), adapted from actual examples found on the internet, show that additional modifiers can be added to the prenominal PP. Since ruim and ver modify the cardinals and not the nouns prenten/euro, such examples can be interpreted as additional evidence for an analysis along the lines of (81a&b), where the sequence P + de + cardinal is construed as a complex modifier of the noun.

We have not been able to find modifiers for sequences with prepositions that trigger the approximative meaning other than ongeveer, which leads to a more or less tautological result because it is also approximative in nature; this is clear from the fact that all examples in (87) express more or less the same meaning.

The fact that the modification of the PPs in the above examples does not seem to differ much from regular cases of modification of PPs again suggests that we are dealing with a productive syntactic pattern.

This subsection provides four arguments demonstrating that phrases such as rond de honderd euroabout a hundred euros is not PPs but noun phrases. This supports an analysis along the line in (81): [NumP [PP rond [DP de honderd]] [Num [NP euro]]].