Example (98a) shows that there are different ways to express a person's age. The sentence in (98a) is perfectly acceptable, but feels a bit laborious; the sentences in (98b) and (98c) are the more colloquial ways of expressing Marie’s age.

On the assumption that the meaning of a sentence is compositionally constructed, the fact that the three examples in (98) are semantically equivalent suggests that we are dealing with syntactically similar constructions, and the easiest way to express this is to assume that the examples differ in the phonetic realization of their constituting parts: some elements may be silent, as indicated by small caps in (99). This means that in all cases we are dealing with a copular construction with an adjectival predicate modified by a nominal measure phrase.

Similar constructions are given in (100), although not all variants are as felicitous as their structural counterparts in (99); in particular, the use of the silent measure noun meter seems impossible in this case.

There are additional restrictions on the use of silent nouns, the most important being that the adjectival predicate must be predicated of noun phrases referring to one or more persons (or animals); inanimate noun phrases are usually excluded when a silent element is present.

The singly-primed examples in (101) are only possible in special contexts (see below); a similar restriction does not apply to their structural counterparts in (99) and (100), which would be correctly interpreted as referring to Marie’s age or height, even if they were uttered unexpectedly. The reason for this may be that we tend to classify people according to their age and body size: the latter perhaps to a lesser extent, which may explain why Marie is (bijna) twee can only be interpreted as referring to age. Although size is important in the classification of both people and furniture, the former are usually considered in terms of height, while furniture is usually considered in all three dimensions. Not surprisingly, in contexts where the height of mountains (in mountaineering) or the age of a car (e.g. in economic life) is particularly relevant, the examples in (102) may be passable without an overt adjectival predicate.

The doubly-primed examples in (101) with the silent measure nouns jaar and meter are unacceptable, even in contexts that leave no doubt about the intended interpretation.

The relevant studies collected in Kayne (2005/2010/2019) have shown that the assumption of silent nouns can explain several unexpected properties of constructions with cardinals (and quantifiers). We will illustrate this for the Dutch counterpart of John will easily make it to a hundred, which Kayne analyzes as John will make it to the age of a hundred years, in which the elements in small caps are silent. Now consider the two semantically equivalent alternants in (103). A problematic feature of (103b&c), pronounced as Jan haalt de honderd (jaar) gemakkelijk, has always been the obligatory presence of the definite article de before the cardinal honderd; the article de can now easily be explained as belonging to the silent noun leeftijd (although it does raise the question of why Dutch differs from English in not allowing a silent article de, which may be related to the obligatory expression of the indefinite article a in a hundred).

The above discussion suggests that there are both semantic and syntactic reasons to think that syntactic representations can contain silent material. Although this raises many as yet unanswered questions about the distribution and licensing of such elements, some of which will be briefly addressed later, we will see in the following subsections that such silent elements can help resolve several long-standing problems related to “unexpected” morphological and syntactic material in certain constructions with cardinal numerals, such as the definite article de in (103b) above.

Section 20.1.1.5 has argued that there are reasons to assume that cardinal numerals are nominal in nature, and that we therefore expect that they can be pluralized. Section 20.1.1.2, example (25), has further shown that there are a number of cases for which this is uncontroversial: cardinals denoting “round” figures like honderdhundred, duizendthousand, miljoenmillion, etc., can all be pluralized by -en. However, this is not true for all cardinals: replacing honderdenhundreds in (104a) with tienen is impossible (although the intended meaning can be expressed by the undisputed noun tientallentens). However, example (104b) shows that there are contexts in which forms such as tienen (as well as tweeën, etc.) can easily be used.

Although the cardinals in (104) are all suffixed with -en, the cases in (104a) and (104b) are semantically different: the suffix -en attached to the cardinal honderda hundred expresses the multiplicity of the number, i.e. n x 100 students; the suffix -en attached to tien/twee, on the other hand, does not express the multiplicity of the number, because the subject as a whole simply refers to a group of ten/two persons (including the speaker). The absence of a plurality meaning makes it highly doubtful that we are dealing with pluralization of the cardinal. This doubt is strengthened by the fact that the suffix -en found in constructions like (104b) does not always coincide with the plural marker normally found with the cardinal in question: in example (105a) the suffix -en appears despite the fact that, according to the phonological constraints on pluralization, the plural of the cardinal zevenseven is usually formed with the suffix –s; cf. (105b).

Furthermore, cardinals in contexts such as (104b) are restricted to the lower number range; it seems that cardinals up to six lead to perfectly acceptable results, that cases based on numbers between seven and twelve are possible but occur with decreasing frequency, and that higher numbers lead to questionable results. This is illustrated in (106); we expect the acceptability judgments on (106c&d) to vary somewhat from speaker to speaker.

Booij (2005) concludes from the above examples that the use of pluralized numerals is idiosyncratic in nature and limited to very specific syntactic contexts, which would militate against the idea that we are dealing with syntactically productive constructions. However, Subsection III will follow Corver & Kranendonk (2008) in arguing that constructions such as wij zessen with a seemingly pluralized cardinal do involve productive syntactic patterns.

Booij (2005) lists a number of constructions in which “plural” cardinals like drieën can be used. Three of his examples are given in (107).

The following subsections will discuss the three cases in (107) in turn. There is yet another kind of collective adverbial phrase, with the invariant pronominal form zijn, which will be discussed immediately after our discussion of the first kind represented by (107c).

Cardinals can be used in names of, e.g., royalties, machines, and products. Some examples are given in (144); such cases are characterized by the fact that the proper noun precedes the cardinal and that the cardinal does not have its normal cardinality reading.

The fact that the proper noun precedes the numeral is not surprising in the light of Longobardi’s (1994) claim that proper nouns are heads of regular noun phrases, but differ from lexical nouns in that they can move into the D-position: for the proper noun Jan, this leads to the representation [DP [[D JanN] [NP tJan ]]]. The assumption of N-to-D movement also accounts for the fact that definite articles cannot be used in the expressions in (145), where the lexical noun assumes name-like properties.

The loss of the cardinality reading by the cardinal numerals is evident in the fact that the noun phrases in (144) and (145) all refer to singular entities. In fact, it is not even possible to say that they have taken on an ordinal guise: room 214, for example, can refer to a particular room on the second floor and does not necessarily imply that there is also a room 167 on the first floor.

De Belder (2007) has argued that the date expression in (146a) also involves naming, but differs from the previous cases in that it involves a silent noun, day, as loosely indicated in (146b); for a more detailed discussion of the representation of (146), we refer to De Belder’s article.

That the silent noun in (146a) functions as a name is clear from the fact that it does not allow a definite pronoun; cf. *de drie februari. In this respect, examples such as (146a) differ crucially from their counterparts with an ordinal numeral such as de derde februarithe third of February, which, as we will argue, includes the silent noun day; such cases are discussed in Section 20.1.2, sub IV.

The previous subsections addressed some peculiar syntactic and morphological properties of cardinal numerals and suggested that they could be explained by postulating silent nouns like leeftijd age, jaar year stuk piece, persoon person, or dag day. We also followed Kayne (2005; 2010) in assuming that silent nouns are not the result of deleting the phonological features of the nouns, but that they are stored in the lexicon without phonological features. In fact, Corver & Kranendonk (2008) claims that these nouns have little descriptive content and should therefore be characterized as grammatical nouns in the sense of Emonds (1985; 2000). Empirical evidence for these claims can be found in Kayne’s (2003; 2007) discussion of the quantification constructions in (147).

The fact that the quantifier few can bear the comparative and superlative morphemes -er/-est provides conclusive evidence for analyzing it as an adjective. However, if few were indeed an adjective modifying the noun books, the grammaticality contrasts between the primeless and primed examples in (148) would be surprising; these cases clearly show that a/every should not be interpreted as a determiner/quantifier of the nouns books and days. However, since determiners and quantifiers cannot normally be combined with adjectives either (cf. *a green/every green), it is also unlikely that their presence in (148) is directly licensed by the adjective few.

Kayne explains these facts by proposing that the determiner/quantifier is licensed by a silent noun number, which is both premodified and content-licensed by the quantificational adjective few. Assuming that the resulting noun phrase [a few number] has the same modifying function as the cardinal numeral in (149a), the primeless examples in (148) can be assigned the structural analyses (149b&b'), in which a/every few has a modifying role similar to that of nearly in (149a).

Because the silent noun number in the (b)-examples has no overt counterpart, the conclusion that it is listed in the lexicon without a phonological matrix seems inescapable.

Another reason to think that we are dealing with silent nouns, listed as such in the lexicon, is that they often have a very strict interpretation. Consider the examples in (150), repeated with a different cardinal numeral from Subsection I.

The crucial example is (150c), where the silent noun would normally be interpreted in the same way as the overt noun jaar in (150a&b). Note that this would be the case even in a baby health center, where the age of babies is more likely to be expressed in terms of days, weeks, or months. This would be difficult to explain if we were dealing with phonetic reduction of the noun triggered by the linguistic context, while it follows naturally when we are dealing with a silent grammatical noun expressing salient cognitive notions in everyday experience: the age of a person is usually counted in years, not in days, weeks or months.

Obviously, the assumption of silent grammatical nouns in the lexicon raises many (as yet unanswered) questions. Consider, for example, the primeless examples in (151). The structural analysis of rond vier uur and rond vieren in the primed examples simply follows the proposal in Section 20.1.1.5, sub III, and is not at issue here; we focus on the properties of the silent noun uur hour, which is in fact part of the (incomplete) inventory of silent nouns provided by Kayne.

The assumption of the silent noun uur in the primed examples of (151) seems justified in view of the fact that the examples are synonymous, but it also raises at least two additional questions. The first concerns the content-licensing of the silent noun. One possibility is to say that the (a)-examples refer to an eventuality (i.e. the arriving of Jan) that is inherently time-bound, and that this is sufficient to content-license the silent noun. But even if this were true, it would leave us with the (b)–examples, which do not refer to an eventuality, but simply express the local time in the speaker’s here-and-now. The second question concerns the suffix -en in vieren; it is unlikely that this is a plural affix attached to the silent noun uur, since it cannot appear on its overt counterpart uur. This shows that the assumption of silent nouns does not solve all cases of “unexpected” morphological affixes on cardinals; that such affixes can also be present for purely syntactic reasons is suggested by Corver & Kranendonk’s (2008:§5.1) discussion of the licensing of contextually triggered N-ellipsis.