There is good reason to believe that relative elements are like wh-phrases in that they are moved into the specifier position of the complementizer (i.e. the position immediately preceding it). This cannot be shown for standard Dutch, because the phonetic content of the complementizer is obligatorily omitted in relative clauses, as indicated by the strikethrough in (300a), but it is supported by the fact that many Flemish and Frisian dialects do allow the complementizer to be expressed overtly; cf. Pauwels (1958), Dekkers (1999: §3), Barbiers et al. (2008: §1.3.1), Boef (2013: §3) and the references cited there. Example (300b) shows that the movement of the relative element in (300a) is obligatory; leaving it in situ leads to ungrammaticality.

The obligatoriness of the movement can be motivated semantically by assuming that wh-movement of the relative element creates an open proposition (i.e. a one-place predicate) that can be used to modify the head noun. In this view, a relative clause is semantically similar to an attributive modifier such as vriendelijk in de vriendelijke manthe friendly man, which also functions as a one-place predicate. This more or less classical idea is attractive, of course, because it suggests that the role of wh-movement in relativization and question formation can be unified, as the latter also creates an open predicate (to be filled in by the addressee). Although there is a debate about whether the derivation of relative clauses given in (300) is entirely correct, we will simply assume that the proposed semantic motivation for wh-movement in relative clauses is on the right track, and that any syntactic account of relativization should be able to accommodate it in order to be tenable; cf. Bianchi (1999), De Vries (2002: §4), and Salzmann (2006: §1) for reviews of the debate mentioned above.

If wh-movement in relative clauses is indeed motivated by the need to create an open proposition, we would again expect that only the relative element needs to be moved into clause-initial position. This raises the question of whether wh-movement will trigger pied piping when syntactic constraints prohibit such extraction. The examples in (301) with a prenominal possessor show that this is indeed the case: since wh-movement of the possessor would suffice to create the desired open predicate, pied piping of the larger phrase should be motivated by appealing to a syntactic constraint that prohibits extraction of the relative element from the noun phrase wiens vaderwhose father.

The constraints on extraction of relative elements are more or less the same as those on extraction of wh-elements. To avoid repeating the discussion of stranding and pied piping in Section 11.3.1.1, we will illustrate this only for PPs. The examples in (302) first show that prepositional objects such as naar wieat who require pied piping. However, if the PP has the pronominalized form waarnaar, which is easier to get for human entities in relative clauses than in wh-questions due to the presence of an antecedent with the feature [+human], stranding is possible and may even be preferred (although we have no frequency data to substantiate this).

The examples in (303) show the same for prepositional complementives; cf. Section P35.2.1.1 for a discussion of the fact that verbs of location such as zittento sit take a complementive PP. When the nominal complement of the preposition is the interrogative pronoun wie, pied piping is obligatory, while stranding seems to be the preferred option in the case of pronominal PPs.

Postpositional complementives differ from prepositional complementives in that they do not allow pied piping, but require stranding of the postposition. We illustrate this in (304) with the complementive de boom ininto the tree. Since Section P34.2.2 has shown that there are reasons to assume that pronominalization of postpositional PPs is not (easily) possible, we will not try to construct such examples here.

Finally, the (b)-examples in (305) show that circumpositional complementives such as [PP [tussen wie] door] must be split. The first member of the circumposition plus the wh-phrase (i.e. tussen wie) is preposed, while the second member door remains in situ; cf. P32.2.6 for the hypothesis that tussen wie is actually a PP-complement of the postposition door. When the circumpositional phrase is pronominalized, as in the (c)-examples, stranding of all adpositional material is clearly the preferred option.

The judgments on the examples above show that pied piping and stranding are more or less in complementary distribution; this means that wh-question formation and relativization exhibit essentially the same pattern in this respect. This suggests that we will be able to account for the examples in (302)-(305) by adopting the set of assumptions from Section 11.3.1.1, sub VI, including the “avoid pied piping” constraint; we refer the reader to that subsection for the general line of reasoning that can be readily applied to the examples in (302)-(305).

Relativization is compatible with long wh-movement: we illustrate this in example (306) for a direct object and an adverbial phrase extracted from an object clause.

As in the case of wh-question formation, long wh-movement is only possible from argument clauses; the examples in (307) show that adjunct clauses prohibit extraction of both arguments and adjuncts and should therefore be considered strong islands for wh-movement of relative elements.

Long wh-movement requires the matrix clause to contain a so-called bridge verb. Example (308b) shows that for wh-questions long wh-movement is perfectly acceptable with the verb zeggento say, but not so easy with verbs of saying that express a manner component like schreeuwento shout. Example (308c) shows that we find the same contrast with long wh-movement in relative clauses.

However, it seems that the set of bridge verbs is not identical for the two constructions. While Section 11.3.1.2 has shown that object clauses selected by factive verbs such as wetento know are weak islands for long wh-movement in wh-questions, this does not seem to hold for long wh-movement in relative clauses. In fact, a corpus of long wh-movement constructions manually collected by Jack Hoeksema shows that weten is the most common bridge verb in relative clauses derived by long wh-movement; cf. Table 5.2 in Schippers (2012). Although Schippers does not give specific examples illustrating the bridge function of weten, a Google search (July 27, 2014) on the search strings [die ik/je/hij weet dat] that I/you/he know(s) that and [die ik/je/hij eet/wist dat] that I/you/he knew that shows that this construction is indeed relatively common; the examples in (309) provide two attested examples (no longer available on the internet). Note in passing that example (309a) seems to suggest that, at least for some speakers, long wh-movement of subject pronouns in relative clauses does not lead to complementizer-trace effects; cf. also Van der Auwera (1984), Boef (2013:35), and Coppen (2013).

We conclude that long wh-movement is acceptable in relative constructions. At the same time, however, it is true that speakers seem to differ in their appreciation of relative clauses with long wh-movement. For example, Salzmann (2006:153) notes that some speakers prefer resumptive prolepsis constructions of the kind in (310) to the long wh-movement constructions in (309); note that the indices in (310) indicate co-referentiality, not movement. For corpus data, see Schippers (2012:§5) and Schippers & Hoeksema (2021).

The island sensitivity of wh-questions and relative clauses does not differ when it comes to strong islands. We will illustrate this here only for embedded questions. Example (311a) is an embedded polar yes/no question, and (311b) shows that such clauses block long wh-movement of relative elements; note that the competing resumptive prolepsis construction in (311c) does lead to an acceptable result. The examples in (312) show the same for an embedded wh-question.

This subsection briefly discusses wh-movement in so-called cleft and pseudo-cleft constructions. The cleft construction illustrated in (313a) is characterized by the fact that it contains the subject pronoun hetit, a contrastively focused complementive (here: je vriend) and a clause that closely resembles a relative clause. However, the clause does not function as a modifier of the complementive, as can be seen from the fact that it neither restricts the denotation of the head noun vriendfriend nor provides additional information about the referent of the definite noun phrase je vriendyour friend. Instead, examples like (313a) express identity statements: the person who stole the book is identified as your friend. That we are not dealing with a modifier of the complementive is also clear from the fact that the clause cannot occur adjacent to it when there is a verb in clause-final position; restrictive and non-restrictive relative clauses are normally possible in preverbal position. The number sign indicates that examples such as (313b) cannot be interpreted as an identity statement, although it can be used to refer to a certain friend who also happens to be a thief.

In the linguistic literature on Dutch, cleft constructions have received little attention, which may be related to the fact that some researchers consider it a barbarism that replaces the more regular construction using only accent, as in Je vriendje heeft het boek gestolen Your friend stole the book; cf. Paardekooper (1986:901), which suggests that French influence plays a role here. Paardekooper analyzes the clause as an extraposed subject introduced by the anticipatory pronoun het. The reason for this is that it can also be preposed, as in (314a), resulting in a construction that closely resembles the English pseudo-cleft construction. The fact that (314a) is more or less equivalent to (314b) further suggests that the clause is a free relative, and this is indeed what is suggested in Paardekooper (1986) as well as in Smits (1989: §4.2).

De Vries (2002) expresses some skepticism about claims that constructions of the above type should be identified with English cleft and pseudo-cleft constructions, because the Dutch constructions have hardly been studied in their own right (but see Daems 1974), and it is not clear whether the findings for English also apply to Dutch. Since a detailed discussion will have to wait until future research has clarified this issue, we will confine ourselves here to noting that the movement of the relative-like element die into clause-initial position has the hallmarks of wh-movement: for instance, the examples in (315) show that it is not clause-bound, but nevertheless island-sensitive in that it cannot be extracted from an embedded question or an adjunct clause.