The pronouns die and dat can be used with antecedents referring to both human and non-human referents. Which of the two pronouns is used depends on the gender and number of the antecedent: die is used with [-neuter, singular] or [±neuter, plural] antecedents, whereas dat can only be used with [+neuter, singular] antecedents. In other words, dat can only be used with antecedents that take the neuter definite article het, and die is used in all other cases. This is illustrated in Table 4. Note that we gloss die/dat as who when the antecedent is [+human] and as that when the antecedent is [-human], in accordance with the preferred English rendering of the pronoun.

		singular	plural
[-neuter]	[+human]	de man die daar loopt the man who there walks ‘the man who is walking there’	de mannen die daar lopen the men who there walk ‘the men who are walking there’
	[-human]	de bal die daar ligt the ball that there lies ‘the ball that is lying there’	de ballen die daar liggen the balls that there lie ‘the balls that are lying there’
[+neuter]	[+human]	het kind dat daar speelt the child who there plays ‘the child who is playing there’	de kinderen die daar spelen the children who there play ‘the children who are playing there’
	[-human]	het boek dat daar ligt the book that there lies ‘the book that is lying there’	de boeken die daar liggen the books that there lie ‘the books that are lying there’

Note, however, that in informal language the pronoun die is increasingly used with [+neuter, singular] antecedents when the antecedent has a [+human] or [+animate] referent. Thus, instead of the expressions in (92a&b), we may find the corresponding primed examples.

It is not entirely clear what the scope of this usage is. Haeseryn et al. (1997:330) give some examples of non-restrictive relative clauses with nouns referring to a person, including an example with the diminutive of a proper noun; cf. taaladvies.net/taal/advies/vraag/208.

However, it is not the case that this use is limited to non-restrictive relative clauses: it is easy to find examples of restrictive relative clauses on the internet. Two clear cases are given in (94): the first is the title of a story on You tube, and the second is part of a review of a theatrical performance.

It has been suggested that examples such as (92a) are common because the neuter gender of the nominal head clashes with the sex of the referent of the noun phrase; cf. onzetaal.nl/advies/diedat.php. This may indeed be relevant, but it cannot be the whole story, since examples such as (92b') can be found in contexts that give no indication of the dog’s sex: in this case, the grammatical gender of the stem of the diminutive form seems to be the determining factor. It seems clear that more research is needed before we can say anything definitive about this phenomenon.

The relative pronouns die and dat can be combined with different types of quantified antecedents. This is shown in (95) for existentially quantified noun phrases, and in (96) for universally quantified noun phrases.

In archaic and literary (poetic) Dutch, the relative pronoun die can also take the quantified pronoun alall as its antecedent, leading to constructions such as Al die dit leest is gekall who read this are mad. In (formal) Dutch, however, the relative pronoun wie is preferred in this context; cf. Subsection B2, below.

The relative pronoun die can also be used in so-called semi-free relative constructions, i.e. restrictive relative clauses with an antecedent that has little semantic content and no independent reference. In these constructions, die is used with the antecedents degene(n) and diegene(n)the one(s), both of which are used only for [+human] referents.

The relative pronoun dat does not seem to be favored in these constructions: the [-human] antecedent dat must be followed by the relative pronoun wat, which is probably motivated by the fact that using dat would lead to haplology. But also with datgene the use of wat seems to be much preferred, although there are many examples with dat on the internet.

Since the antecedent in semi-free relative constructions has no independent reference, relative clauses of this type are always restrictive. For completeness’ sake, example (99) shows that neither die nor dat can be used in free relative constructions, i.e. these relative pronouns always require an overtly realized antecedent.

The examples in (100) show that the relative pronouns die and dat can have the same syntactic functions as a regular noun phrase, viz. as subject or object of the relative clause.

The examples in (101) also show that the relative pronouns die and dat cannot function as the complement of a preposition, regardless of whether the PP is an argument, as in (101a&b), or an adverbial phrase, as in (101c&d). In this respect, these pronouns behave like personal pronouns referring to inanimate entities; cf. Section P37.1.

Note that stranding of the preposition, as in (102), does not improve matters, which is of course consistent with the fact that Dutch does not allow preposition stranding by extracting a noun phrase; cf. Section P37. Note in passing that mee in (102c) is the stranded form of the preposition met.

This means that Dutch must resort to other means to express the intended meanings. Subsection B below will show that, in the case of [+human] antecedents, this can be done by replacing die/dat with the pronoun wie. An alternative, which is also available when the antecedent is [-human] and will be discussed in Subsection III, is to use a (split) pronominal PP waar ... P.

The examples in (103) and (104) show that die/dat can also be used in restrictive relative clauses with an antecedent that functions as a complementive. This is possible regardless of whether the relative pronoun functions as an argument or a predicate in the relative clause. This is shown in the (a) and (b)-examples, respectively. Relative clauses of this kind are discussed in more detail in Section 17.3.2.3.3, sub IC.

In some varieties of spoken Dutch, the relative pronoun die is sometimes used in possessive constructions such as (105a), where it is followed by the reduced possessive pronouns z’nhis or d’rher, which can also be found in possessive constructions like Jan z’n boekJanʼs book and Marie d’r boekMarieʼs book. Example (105a') shows that the resulting construction, which is considered substandard and which is not acceptable to all speakers of Dutch, is restricted to the singular, which may be related to the fact that the third-person plural possessive pronoun huntheir has no reduced form; cf. the discussion in Section 19.2.2.4, sub I. Example (105b) shows that the relative pronoun dat differs markedly from die in that it can never be used in this way.

The relative pronoun wie is restricted to [+human] referents and can be used regardless of the gender, number, or definiteness of its antecedent. This is illustrated in examples (107).

Example (108) shows that it is not easy for wie to take an existentially or universally quantified antecedent: the pronoun die is generally the strongly preferred option in this case. This is shown by (108).

However, the examples in (109a&b) show that the pronoun wie can be used to modify the universally quantified pronoun alall when it is used to refer to persons, although this particular use is characterized by a high degree of formality. Note that both the antecedent and the pronoun can be interpreted as either singular or (perhaps somewhat markedly) plural, as the form of the finite verbs of the main and relative clauses shows.

Example (110a) shows that the universally quantified pronoun al cannot be used without the relative clause introduced by wie. This also explains why it cannot be used with a non-restrictive relative clause, as shown in (110b). For more details on the quantifier al, see Section 21.1.2.1.

Relative clauses introduced by the pronoun wie can easily be used as free relatives, i.e. without a phonetically expressed antecedent. Some examples are given in (111). Free relatives occur most often as the subject of a generic matrix clause like (111a&b), but (111c) shows that it is certainly not impossible to use a free relative subject to refer to a specific person.

In (111) there is matching in syntactic function between the free relative in the main clause and the relative pronoun in the relative clause, but it is also possible to have a mismatch between these two functions. In (112a), for example, the free relative functions as the subject of the main clause, while the relative pronoun functions as the direct object of the relative clause. The acceptability of (112a) contrasts sharply with the unacceptability of the German example in (112b), taken from Van Riemsdijk (2006/2017:§2), which is generally attributed to the fact that German, unlike Dutch, has morphological case: the relative pronoun in the German example must be accusative in order to play its role in the relative clause, but nominative in order for the free relative clause to play its role in the main clause: this morphological conflict (absent in Dutch) causes the unacceptability of (112b).

The examples in (113) show that free relatives can also function as direct objects; again, the free relative can have a generic or a specific interpretation. These examples again show that Dutch is not subject to a matching restriction: the free relatives in (113) function as direct objects of the main clauses, while the relative pronouns function as subjects of the relative clauses. In German, examples such as (113) are considered ungrammatical or archaic; cf. Van Riemsdijk (2006/2017:§4.2.2) for discussion.

Example (114a), taken from the internet, shows that a free relative can also be used as the complement of a preposition, although it should be noted that using a free relative as part of an indirect object introduced by aanto, as in (113b), seems less favored than using a nominal indirect object.

A free relative is usually analyzed as a noun phrase headed by a phonetically empty antecedent for the relative pronoun wie, and not as a clause (Van Riemsdijk 2006/2017). The evidence for this claim is that the free relatives with wie have the syntactic distribution of noun phrases, and not that of clauses: they must precede the verbs in clause-final position, even when an anticipatory pronoun is present. This is illustrated in (115) by free relatives acting as subject and object of the clause.

The examples in (116) show that the semi-free relative constructions are marked, both in comparison with the corresponding free relatives in (111) and in comparison with semi-free relatives like (97b&c) with the relative pronoun die.

Although the more or less idiomatic constructions in (117a&b) resemble free relative constructions, they are different in several respects. First, the particle ook is obligatory and adds a concessive meaning to the construction (“no matter who you ask/see...”), making the construction as a whole more or less equivalent to constructions with a universal quantifier: (117a) could be paraphrased as “Everyone gives the same answer” and (117b) as “Everyone has a cell phone”. Second, the primed examples show that it is completely impossible for the construction to take an overt antecedent or to appear in the form of a semi-free relative construction.

Moreover, the wie-constructions in the primeless examples do not function as arguments, but rather as clausal adjuncts. Since Dutch is a verb-second language, the finite verb in declarative main clauses is preceded by a single constituent, and since the subject in the primeless examples in (117) occupies this position, the wie-phrase must be clause-external. This conclusion is also supported by the fact, illustrated by (118), that the wie-phrase, unlike regular constituents of the clause, cannot occupy the first position of the clause itself.

For the sake of completeness, note that, according to Haeseryn et al. (1997:361), modifiers like onverschilligindifferent and om het evenirrespective can perform the same function as the particle ook in (117). We cite one example in (119); to our ear, the use of onverschillig sounds rather formal and somewhat forced.

Another highly idiomatic type of expression, typically found in proverbs, is illustrated by the constructions in (120); cf. Stoett (1923-5: #2503). Originally, the second part of the construction functioned syntactically as a relative clause introduced by the relative pronouns die and dat. Today, however, it is more common to use the pronouns wie and wat, respectively, as evidenced by the results of a Google search (January 12, 2015). In (121) we give similar examples with wat/dat.

The examples in (122) show that the pronoun wie cannot function as the subject or direct object of a relative clause in standard Dutch, although it should be noted that in some varieties of Dutch (such as the dialect spoken in Amsterdam) the relative pronoun wie can also be used as the direct object of a relative clause, hence the % mark in (122b&c).

The relative pronoun wie normally functions instead as the object of a PP, where the PP as a whole can function either as a complement or as an adjunct of the relative clause: the examples in (123a&b) illustrate the former, involving an indirect object and a PP-complement, respectively; example (123c) illustrates the latter.

Since the pronoun wie functions as a regular noun phrase, the preposition is obligatorily pied-piped. This means that stranding the preposition, as in (124), leads to unacceptability; in such constructions the relative pronoun must take the form of the R-pronoun waar, as will be shown in Subsection III.

Although the examples in (122) have shown that the relative pronoun wie cannot function as a subject or a direct object, it can function as the nominal indirect object of a ditransitive relative clause. This means that the examples in (107) alternate with the examples in (125). The examples in (125) also show that the bare pronoun wie (i.e. not wie in a PP) can be replaced by the relative pronoun die/dat; cf. Subsection A.

It seems that the relative pronoun wie can also be used for other types of dative noun phrases, although the judgments are perhaps less clear. In (126a&b) we are dealing with nom-dat verbs, i.e. unaccusative verbs with a dative argument (cf. V2.1.3), and it seems possible to relativize the dative argument with either wie or die. Both options are certainly available for examples such as (126c), where the dative is not an argument of the copular verb zijnto be, but of the adjective trouwloyal; cf. Section A25.2.

In some varieties of spoken Dutch, the relative pronoun wie can be used in possessive constructions like (127a&b), where it is followed by the reduced possessive pronouns z’nhis or dʼrher, which can also be found in possessive constructions like Jan z’n boekJanʼs book and Marie dʼr boekMarieʼs book. Its use is restricted to the singular, which may be related to the fact that the third-person plural possessive pronoun huntheir does not have a reduced form; cf. the discussion in Section 19.2.2.4, sub I.

The constructions in (127a&b) are often considered substandard and not acceptable to all speakers of Dutch. Speakers who do not accept these examples usually use the genitive form wienswhose in possessive constructions like these; cf. Subsection II.

Describing the use of the relative pronoun wat is complicated by the fact that a process of language change seems to be underway in which the use of wat is on the rise (Van der Horst 1988, Schoonenboom 1997/2002); many of the uses of wat in the examples below seem to be relatively recent innovations in the language and therefore meet with normative resistance.

The nominal antecedents of the relative pronoun wat are [+neuter] and [+singular]. That the antecedent must be neuter can be seen in (129a&b): despite the fact that the two nouns voorstelproposition and suggestieproposition are close synonyms, only the neuter noun voorstel can enter the construction. That the antecedent must be singular can be seen in (129c). Note that all examples in (129) are acceptable with the relative pronoun dat or die.

Haeseryn et al. (1997:338) claim that wat can only be used with an indefinite antecedent, but they immediately add that in informal speech wat can be used with any singular neuter noun and is actually preferred to dat in large parts of the language area; a cursory look on the internet shows that this is indeed the case. A Google search (April 17, 2008) for the strings het aanbod wat/dat er is gave the results in (130). However, it seems that wat is more widely accepted with an indefinite antecedent than with a definite antecedent, hence the % mark in (130b).

The examples so far have all been with abstract nouns. Most speakers of Dutch seem to be less willing to accept examples with concrete nouns, although examples with both indefinite and definite antecedents can easily be found on the Internet; two examples are given in (131). Note that (131b) again shows that wat can also occur with definite antecedents.

The examples in (131) are restrictive relative clauses. The judgments are different for non-restrictive relative clauses, where the use of wat seems to be more generally accepted and even preferred by some speakers with indefinite antecedents. This is illustrated by the constructed examples in (132). First, example (132a) is a restrictive clause, and dat is preferred to wat by most speakers of the standard variety, regardless of the definiteness of the antecedent. Example (132b) is a non-restrictive relative clause with an indefinite antecedent: all speakers accept the relative pronoun wat, and some (but not all) speakers even prefer it to the relative pronoun dat. Finally, example (132c) is a non-restrictive relative clause with a definite antecedent, and most speakers accept both wat and dat (although speakers’ preferences seem to vary from person to person). In short, while the precise status of the examples in (132b&c) may be somewhat unclear, it seems safe to conclude that most speakers accept wat in non-restrictive clauses.

The examples in (132) show that wat can also be used when the antecedent is animate or human, although many speakers will object to these examples even more than to those with inanimate nouns in (131). The examples with wat in (133) are taken from the internet.

The antecedent of non-restrictive relative clauses with the relative pronoun wat need not function as an argument, but can also function as a nominal predicate, and in this respect wat is crucially different from the relative pronouns die and dat, which cannot normally be used in this context. This is illustrated in (134). Note that the relative pronoun wat is not sensitive to the gender and number specification of the predicative noun phrase; this is related to the fact, discussed in Subsection 5 below, that wat can also take an AP predicate as its antecedent.

Finally, we can point to a typical use of the pronoun wat in constructions with a nominal antecedent in the form of an elliptical superlative expression followed by a restrictive relative clause. Examples are given in (135a&b), in which wat can be seen as coreferential with the phonetically empty noun [e] modified by the superlatives mooistemost beautiful and meest belachelijkemost ridiculous; example (135c) has a more or less idiomatic flavor. Note that dat can also be used in these cases.

As shown in (136), a similar construction occurs with expressions like enigeonly and eerste/laatstefirst/last, which are similar to superlative phrases in that they are usually used in definite (i.e. uniquely referring) noun phrases.

The relative pronoun wat is also used in combination with quantified pronominal antecedents. Some examples are given in (137). Although it is possible to use either wat or dat, in some cases one of the two pronouns is clearly preferred. This is particularly clear in the examples in (137a&c): a cursory look on the internet shows that the string [alles wat] occurs about twenty times as often as [alles dat], while [zoveel dat] occurs about ten times as often as [zoveel wat]; the string [iets wat] in (137b) occurs about twice as often as [iets dat].

When the antecedent is niets, as in (138), both wat and dat are acceptable, although the former seems to be more popular: a Google search (January 12, 2015) for the string [niets wat/dat ik] yields 294 hits for wat and 359 hits for dat, many of which contain the irrelevant construction Het is niet voor niets dat ik ...It is for a good reason that I .... Nevertheless, there seem to be several interfering factors which, to our knowledge, have not yet been investigated so far. While in example (138a) wat is about six times as frequent as dat, our Google search also showed that in (138b) dat and wat occur with about the same frequency. The search strings are given in square brackets.

Finally, example (139a) shows that in some cases both the use of wat and dat seem to lead to a degraded result. The preferred way to express the intended thought is as in (139b). In all likelihood, we are dealing here with modification by a PP headed by the preposition van, which takes a free relative as its complement; cf. niets van dat allesnothing of that all, where dat alles is also the complement of van.

The examples in (140) show that the relative pronoun wat can be used in so-called semi-free relative constructions, in which case it is coreferential with the antecedent dat(gene). Replacing wat with the relative pronoun dat in these examples leads to an unacceptable result, which is probably motivated by the fact that this would lead to haplology. But with datgene, the use of wat also seems to be much preferred, although there are numerous examples with dat on the internet.

Wat can also be used in free relative constructions, i.e. without an overt antecedent. Examples with a subject are given in (141). These examples show that the result is best when the free relative is in clause-initial or extraposed position; example (141c), in which the subject occupies the regular subject position immediately after the finite verb in the second position of the clause, is marked. This suggests that, despite the fact that free relatives are usually claimed to be part of a noun phrase with a phonetically empty antecedent, the free relative in (141) has more or less the distribution of a subject clause. In fact, the similarity with a regular clause goes even deeper, as can be seen from the fact that example (141b) must contain an anticipatory pronoun hetit, just as it would if we were dealing with a regular subject clause (De Vries 2002:281; Van Riemsdijk 2006/2017:§2). Note that while example (141a) seems compatible with both a generic and a specific interpretation of the free relative, example (141b) clearly favors a specific interpretation.

The examples in (142) show more or less the same thing for embedded clauses: since topicalization is excluded in Dutch embedded clauses, the subject is preferably placed in extraposed position with the anticipatory pronoun het in subject position.

A free relative functioning as a direct object also exhibits the syntactic behavior of a clause: (143a&b) shows that the free relative is preferentially placed in clause-initial or extraposed position. In the latter case, the anticipatory pronoun hetit is optional, as it would be in a regular object clause; example (143c), in which the free relative occupies the regular object position, is marked again.

It seems that the examples in (143) do not easily allow for a generic interpretation, but that such an interpretation is not completely impossible is clear from the constructions in (144a); example (144b) is a similar, more idiomatic expression.

Example (145) provides the embedded clauses corresponding to (143): since topicalization is excluded in Dutch embedded clauses, the object is preferably placed in extraposed position with an optional anticipatory pronoun het in object position.

That free relatives with wat behave like regular clauses is also clear from the examples in (146), where the free relative is part of a PP-complement of the verb: the PP can be in extraposed position as a whole, as in (146a), or the free relative can be extraposed in isolation, in which case the clause must contain the anticipatory PP (i.e. er + P). What seems impossible, however, is to place the complete PP in a position preceding the clause-final verb. This is exactly the pattern we also find with complement PPs containing a regular finite clause; cf. Section P34.4.

The distribution of free relatives with wat is very different from that of free relatives with wie, discussed in Subsection B3 above, which have the distribution of noun phrases, not clauses. It is not entirely clear what the correct analysis of the examples in (141) to (146) is. De Vries (2002:281) suggests that we may be dealing with a kind of heavy NP shift or right dislocation, with het or a phonetically empty pronoun pro occupying the original position of the free relative, as indicated in (147a). But there are three problems with this proposal. First, it leaves unexplained why we could not have the empty pronoun pro with heavy DPs like het beeldje in de etalage in examples such as (147b): it forces us to assume that pro is only possible with DPs that have the form of free relatives. Second, it leaves unexplained why free relatives with wie in examples such as (147c) cannot undergo the same kind of movement as free relatives with wat. In short, it forces us to adopt the ad hoc solution that pro is only possible with DPs that have the form of a free relative with wat. Finally, it leaves unexplained why (147a) does not require the obligatory intonation break (indicated by an em-dash) that we find in (147b).

Another possibility suggested by De Vries (2002:281) is that we are dealing with an apposition. The main reason for assuming this is that sometimes an apposition marker like en welnamely can be used, which he illustrates with example (148a). Again, there are several problems with this suggestion. The first problem, which De Vries himself points out, is that the marker en wel requires the pronoun het to be present; moreover, it is very difficult to pronounce (148b) with the intonation contour typically associated with appositions, i.e. with an intonation break before the free relative, when this marker is not present.

Second, the addition of the marker to the earlier examples with postverbal free relatives seems to be completely ruled out, which is illustrated in the primeless examples in (149) for the examples in (141b) and (143b). The primed examples show that the free relatives can optionally be pronounced with the intonation contour associated with appositions, which shows that an apposition reading is possible. However, this intonation contour requires that the pronoun hetit be present; this can only be illustrated when the free relative is an object, as in (149b''), since the pronoun is always obligatory with free relative subjects in extraposed position.

The third and final problem with the proposal that we are dealing with appositions is that it predicts that free relatives with wie can be used in the same way, and thus can occur in postverbal position; we have already seen in (147) that this expectation is not borne out. Given these problems with the two proposals discussed above, we conclude that the examples in (141) to (146) present an as yet unsolved problem.

We now continue the discussion with a special kind of free relative construction, which can act as the predicate of the clause. Some examples can be found in (150). The predicative free relatives are sometimes called transparent free relatives, since they themselves contain a predicate (in italics), which is semantically the most prominent part of the construction; cf. Schelfhout et al. (2004). For instance, the examples in (150) are more or less equivalent to the examples in (151); the free relative simply adds the information (at least partly due to the presence of noemento call) that we are dealing with an assessment by the speaker, which leaves open the possibility that other people might have a different opinion; cf. Van de Velde (2009:§8.3.1.4).

Transparent free relatives are always introduced by wat, which is remarkable given that wat seems to function as the logical subject of the embedded predicate and thus might be expected to be sensitive to the [+human] feature of the subject of the clause. Nevertheless, replacing wat by wie in (150b) leads to an unacceptable result, as shown in (152); cf. Schelfhout et al. (2004). This seems to have to do with the fact that transparent free relatives are evaluative (i.e. property denoting) expressions; this may be incompatible with the use of [+human] wie as the head of the free relative, because such free relatives always refer to persons (not properties).

That the free relatives in (150) are transparent in the sense that it is actually the embedded predicative phrase that is active in the main clause is especially clear in example (150a), in which the embedded predicate is adjectival: like the regular set-denoting adjectives, the predicative free relative construction can be used in attributive prenominal position, as illustrated by (153). There are at least two things remarkable about the structure in (153a). First, the adjective duurexpensive is inflected with the attributive -e ending, which we also find with the regular attributively used adjective in (153b), and not with the uninflected form duur, which we find with the predicatively used adjective in (151b). Second, the adjective in (153) follows the clause-final verb noemen, which would never be possible in other cases: dat ik deze auto <duur> noem <*duur>.

Transparent free relatives also occur in argument positions. This is most natural when the free relative is the complement of a preposition, as in (154a), which is a slightly adapted example from the newspaper Het Parool (“Kraamkamer vol ideeën”; March 29, 2008), or a direct object; when the free relative functions as a subject, the result seems somewhat marked. Note that the primeless examples are again more or less equivalent to the primed examples: the free relative construction in the primeless examples only adds the information (through to the use of the verb blijkento turn out) that at the time of the sale it was not known that the shares were worthless.

The markedness of example (154c) may be caused by the fact that the transparent free relative precedes the verb in clause-final position, since example (154b) also becomes somewhat marked when it is embedded; cf. (155a). Note, however, that extraposition of the transparent free relative does not improve matters; on the contrary, it worsens the result, as can be seen from the fact that (155a') is only possible with an intonation break before the free relative. That extraposition of the transparent free relative worsens the result is also shown by the unacceptability of the extraposition counterpart of (154c) in (155b).

The fact that transparent free relatives cannot be in extraposed position shows that they have the distribution of noun phrases; they are thus really different from regular free relatives with wat, which, as we have seen, have the distribution of clauses; cf. the discussion of (141) to (146). Another difference concerns the fact, illustrated in (154c) above, that transparent free relatives can be used in an expletive construction, which shows that they are indefinites; regular free relatives, on the other hand, are definite or generic and therefore never occur in an expletive construction; Van de Velde (2009:§8.3.1.4). This is clear from the fact that example (141a) does not have an expletive counterpart, i.e. [wat hij doet] is lovenswaardigwhat(ever) he does is praiseworthy does not alternate with *er is [wat hij doet] lovenswaardig.

The fact that transparent free relatives behave like indefinite noun phrases might lead to the idea that they contain some phonetically empty quantificational pronoun. Such an idea might be supported by the fact that all transparent free relative constructions in (154) alternate with semi-free relative constructions headed by the pronoun ietssomething: iets wat later waardeloze aandelen bleken te zijnsomething that turned out to be worthless shares. Nevertheless, it does not seem possible to assume that the examples in (154) contain a silent iets, since the semi-free relative construction with iets triggers singular agreement on the finite verb, and not plural agreement, as is the case with the transparent free relative in (154c): Er werd/*werden ons daar iets [wat later waardeloze aandelen bleken te zijn] verkocht.

We conclude with the more or less idiomatic constructions in examples (156a&b), which resemble free relative constructions but are different in a number of ways. First, it is not possible for the construction to take an overt antecedent or to appear in the form of a semi-free relative construction; cf. (156a'&b'). Second, the particle ook (or a modifier like onverschilligindifferent and om het evenirrespective) is obligatory, and adds a concessive meaning to the construction (no matter what you say/do...).

Third, the wat-constructions in the primeless examples do not function as arguments, but rather as clausal adjuncts; since Dutch is a verb-second language, the finite verb is preceded by a single constituent in declarative main clauses, and since the subject occupies this position in the primeless examples in (156), the wat-phrase must be clause-external. This conclusion is also supported by the fact, illustrated in (157), that the wat-phrase, unlike regular constituents of the clause, cannot occupy the first position of the clause itself; cf. also the discussion of (118).

Finally, note that similar constructions are possible with wh-phrases in sentence-initial position. Because the examples in (158) are clearly not free relatives, and given that wat can also be used as an interrogative pronoun, we conclude that we are not dealing with free relatives in (156) either; cf. Van Riemsdijk (2006/2017:§5.2) for further arguments.