New nouns can also be formed by compounding, i.e. by combining an existing noun with another free morpheme. The rightmost element of a compound determines the syntactic category of the whole, a generalization known as the right-hand head rule; cf. Williams (1981b). This means that in nominal compounds the second element is always a noun. The first element, on the other hand, can be a noun, a verb, an adjective, a preposition, or a numeral. Examples are given in (241). Note that in the examples in (241c'), the first element is an adverbially used adjective. For example, a zwartwerkermoonlighter is not a worker who is black, but one who works without paying income tax.

There are three types of nominal compounds. The vast majority of compound nouns is endocentric, i.e. with the second element acting as the semantic head. In this type of compound, the compound denotes a subset of the set denoted by the second noun (AB ⊂ B): huisdeurfront door denotes a particular type of door. Only a few compound nouns are exocentric, i.e. with no element acting as the semantic head. In such cases, the compound cannot be paraphrased as a particular type of the entity denoted by the second (or first) element: wijsneus does not denote a nose but a particular type of person (know-it-all). Another small subset is that of copulative compounds, where both members are nominal. The denotation of compounds in this category is determined by the denotation of both members (i.e. AB = A ∩ B): a kind-sterchild-star is both a child and a star.

In some cases, there may be some doubt as to whether we are dealing with an endocentric or exocentric compound. Such cases are formations like neppistoolfake gun or speelgoedpistooltoy gun. It is clear that in these cases we are not dealing with entities that are prototypical members of the set denoted by the noun pistool, but on the other hand it is not obvious that these entities do not belong to this set, given that examples such as Dit pistool is niet echtThis gun is not real/a fake (Friesch Dagblad, December 18. 2005) sound perfectly acceptable. Obviously, the author of this sentence considers fake guns to be a subset of the set denoted by pistoolgun. From this, we conclude that language users simply treat the formations under discussion as endocentric compounds.

Although there are certain tendencies, the semantic (syntactic) relationship between the two elements of an endocentric compound noun is largely unpredictable. This unpredictability is nicely illustrated by the pair beendermeelbone-meal and kindermeelinfant cereal; the former denotes meal made from bones, while the latter usually denotes flour used to make porridge for children. The list in (242) shows that virtually every conceivable semantic relationship can be found in Dutch compounds.

Many compounds are lexicalized in the sense that the meaning of the compound cannot be fully inferred from the meaning of its constituent parts. Not only exocentric compounds, for which this is to be expected, but also endocentric and copulative compounds can be lexicalized. Examples are given in (243).

Many, but not all, N + N compounds contain a linking element between the two members. In many cases, the form of the linking element depends on the first member of the compound and closely resembles the plural ending of that member: -e(n)-, -s- or -er-. Some examples of compound nouns with such a linking element are given in (244). Note that the presence of the linking element does not imply any notion of plurality; cf. hondenkophead of a dog. The choice between the linking elements -en- and -e- is subject to complicated and still controversial orthographic rules (Woordenlijst der Nederlandse Taal, edition 1995, §5.1.1/2). Since the pronunciation of the two linking elements is identical, these rules are a political rather than a linguistic issue; cf. Booij (1996) for a detailed discussion and historical background.

De Haas & Trommelen (1993) claims that in addition to the linking element -er-, there is a linking element -eren-. Some examples from the Van Dale dictionary are given in (245). It is not entirely clear whether we can conclude from these examples that there is indeed a linking element -eren-. First, the first member goederen in compounds like (245a) can be seen as a plurale tantum (the singular noun goed only occurs in a number of fixed expressions), so there is no reason to assume that we are dealing with a linking element. The examples in (245b) potentially involve a linking element -eren-, but these formations seem to alternate with the compound volkenmoord and Volkenbond with the linking element -en-. Furthermore, since the formations in (245b) do not belong to the colloquial register, it seems doubtful that we can conclude from these examples that there is indeed a linking element -eren-.

Another example given in De Haas & Trommelen is kinderengejoelshouting of children. This seems to be a relatively recent coinage, although the form kindergejoel would still normally be used in speech. This is especially the case when we want to express that the screaming is coming from a single child, in which case (246a) would be downright unacceptable and we would have to use kindergejoel. This observation may also be relevant for evaluating the examples in (246b-d), which can be found on the internet, and in which the first member is also necessarily interpreted as referring to a non-singleton set of entities: -eren- cannot be used if the first member is interpreted as singular; -er-, on the other hand, is common if the first member is interpreted as plural, as in hoenderhokchicken coop. Furthermore, a Google search (2023/3/13) revealed that all forms in (246b-d) alternate with the expected form with the linking element -er-. We found instances of both beenderenkuil (18) and beenderkuil (26), the latter of which seems to be archaeological jargon, and of both eierenaanvoer (9) and eieraanvoer (30), and that both kalverenmarkt and kalvermarkt occur relatively frequently (>100 hits). For this reason, we think it is better to leave it open for the moment whether -eren- should be considered a linking element of the relevant kind.

The examples in (247) clearly show that the linking element -s need not be related to the plural suffix -s: the primed examples show that the first members of these compounds do not take the suffix -s in the plural.

The examples in (248), adapted from De Haas & Trommelen (1993), suggest that it is completely unpredictable whether or not a linking element will appear, and if so, what form it will take: we cannot appeal to the first member of the compound to tell us whether a linking element will appear (examples a-c), nor to the second member of the compound. Nevertheless, there seem to be certain tendencies, but we refer the reader to De Haas & Trommelen (1993: §2.9.1) for a discussion of these.

Finally, note that in some cases the linking element -s is optional and subject to individual variation: objectpositieobject position, for example, seems to alternate freely with objectspositie.