Long wh-movement does not seem to differ from local wh-movement in terms of the syntactic functions of the moved elements. The examples in (213) show that it can affect clausal constituents of all types: argument, complementive, and adjunct. As in the case of local wh-movement, the only requirement seems to be that an interrogative form is available. Recall that we will omit the intermediate trace in the clause-initial position of the embedded clause if it is not immediately relevant to our discussion. Note that the wh-phrase wanneerwhen in (213d) can also be construed as a modifier of the matrix clause, but this is of course not the reading intended here.

The examples in (214) further show that long wh-movement is not restricted to clausal constituents, but can also be applied to wh-elements embedded in clausal constituents (provided that local wh-movement also allows stranding). We illustrate this in (214) with an interrogative modifier of an adjectival complementive and a split wat-voor phrase, respectively.

The examples discussed in the previous subsection suggest that long wh-movement does not impose any special conditions on the syntactic function of the moved element. Nevertheless, it is necessary to say more about long wh-movement of subjects, since it triggers special effects in many languages. This is illustrated for English by the examples in (215), which show that long wh-movement of subjects, but not of objects, requires the omission of the complementizer that. Chomsky & Lasnik (1977) introduce the so-called that-trace filter, which explicitly forbids the configuration [... [C that] ti ...], but since it is possible to find similar facts in languages other than English, we will use the more general term complementizer-trace filter.

Long wh-movement of subjects also triggers a special effect in French. The translation of (215a) in (216a) shows that the subject trace cannot occur when the declarative complementizer appears in its regular form, que, but requires it to appear as qui; cf. Kayne (1976). Example (216b) further shows that this que/qui alternation does not apply in the case of long wh-movement of e.g. an object.

The Dutch translations of the examples in (215)/(216) in the primeless examples in (217) suggest that the subject-object asymmetry found in English and French does not occur in standard Dutch, since both are perfectly acceptable; cf. Dekkers (1999). It seems that for at least some speakers the question of whether the subject-object asymmetry occurs depends on the type of interrogative noun phrase: while non-D-linked subject pronouns such as wiewho in (217a) easily allow long wh-movement, D-linked subjects such as welke jongenwhich boy in (217a') are marked (but certainly passible) for such speakers.

A possible reason for the difference in acceptability of the two (a)-examples may be that, despite appearances, the traces of the two wh-phrases do not occupy the same position in the clause. We will first illustrate this difference in the position of the subject traces with the intransitive examples in (218) and (219) (i.e. without a definite object). The examples in (218) show that long wh-movement of wie requires the presence of the expletive erthere; long wh-movement of welke jongen is passible for at least some speakers, provided that the expletive is present.

What is crucial for our argument is not so much the (admittedly subtle) effect of D-linking on the acceptability of an overt expletive in the two examples in (218), but the contrast between the examples in (218) and those in (219); whereas omitting the expletive is completely impossible in both examples in (218), it is at least marginally possible in (219a), and even preferred in (219a).

The acceptability contrasts between (218) and (219) can be explained by appealing to the complementizer-trace filter. First, consider the two (a)-examples. Since the expletive er is right-adjacent to the declarative complementizer dat or the finite verb in non-subject-initial clauses such as Gisteren heeft er iemand gelogenYesterday someone lied, it can be assumed to occupy the regular subject position. If we further assume that the absence of the expletive indicates that the subject has been moved into the regular subject position (a marked option for non-D-linked wie), the difference in acceptability between the two (a)-examples follows from the complementizer-trace filter: if the expletive er is not present, the C-position is immediately followed by a wh-trace, which is forbidden by the complementizer-trace filter when the C-position is filled by the complementizer dat, but allowed when it is filled by the finite verb. The contrast is even clearer in the case of the two (b)-examples, since the markedness of er signals that D-linked wh-phrases are preferably wh-moved via the regular subject position. But again, we see that the presence of er makes long wh-movement possible.

Let us return to the contrast between (217a) and (217a'). At first glance, the above proposal does not seem to do much to account for this, as these examples do not contain the expletive er. If this indicates, as suggested above, that the subject has been wh-moved via the regular subject position, we would predict these examples to be both unacceptable, contrary to fact. However, the fact that the expletive is not realized is not due to the position of the subject, but to another factor, namely that the realization of expletives is not only sensitive to the (in)definiteness of the subject, but also depends on the presence of presuppositional material in the clause; cf. Section N21.1.2 for details. Consider the examples in (220), where the subjects are all interpreted as non-specific indefinites, and where er should not be interpreted as a locational proform, but as a pure expletive.

The contrast between the two examples in (220a&b) shows that the definiteness of the object may affect the distribution of the expletive er. This is even clearer in (220c), where the referential personal pronoun het blocks realization of the expletive. Consequently, in order to show that the acceptability of long wh-movement of the subject depends on D-linking, we also have to control for the definiteness of the object. This has been done in the examples in (221), which show that with an indefinite object omission of the expletive again has a severely degrading effect in the case of long but not in the case of local wh-movement. The contrast between the primeless and primed examples (221) thus shows again that wh-movement of subjects is sensitive to the complementizer-trace filter.

It is important to note that the complementizer-trace filter crucially involves a phonetically realized complementizer. This is clear from the examples in (222), which show that local wh-movement into the clause-initial position of the embedded clause does not require the presence of the expletive er; in short, the empty complementizer Ø does not trigger the complementizer-trace effect. The primed examples in (221) have already shown that the complementizer-trace filter crucially involves a phonetically realized complementizer, not just a phonetically filled C-position, since finite verbs in second position do not evoke this effect.

We can summarize the above discussion by saying that, in accordance with the complementizer-trace filter, long wh-movement of subjects is possible only if they skip the canonical subject position right-adjacent to the complementizer, which is signaled by the presence of the expletive er (unless its overt realization is ruled out for independent reasons). Having said this, we must add a caveat related to the fact that Maling & Zaenen (1978) have suggested that there are regional varieties of Dutch in which the expletive er can be freely omitted. Although this claim is controversial, there may indeed be variation in speakers’ judgments when it comes to eliminating the expletive in the examples discussed in this subsection. For a more detailed discussion, see Bennis (1986: §3.6.1) and Schippers & Hoeksema (2021:§4).

The acceptability of long wh-movement depends on the properties of the embedded clause from which the wh-phrase is extracted. The examples in (223) show that the embedded verb must be an argument of its matrix clause; long wh-movement from complementive or adverbial clauses is prohibited.

The examples in (224) show that long wh-movement is also impossible from argument clauses introduced by the anticipatory pronoun hetit. This would follow immediately from the observation above, if we assume that the anticipatory pronoun is the “true” argument of the verb, while the clause is an adjunct or a right-dislocated (i.e. clause-external) element.

Long wh-movement is possible not only from object clauses but also from subject clauses, as can be seen from the fact that impersonal passivization of (223a') leads to a perfectly acceptable result; this is shown in (225a'). The use of an anticipatory pronoun again blocks long wh-movement, as shown by the fact that the passivized counterpart of example (224b) in (225b') is unacceptable. For completeness, the primeless examples show that the corresponding cases without long wh-movement are both acceptable.

It should be noted, however, that at least some speakers perceive an argument-adjunct asymmetry in the case of subject clauses. Thus, while all speakers accept argument extraction from both object and subject clauses, as in the (a)-examples in (226), some speakers consider adjunct extraction from subject clauses to yield a worse result than from object clauses, as in the (b)-examples. This suggests that subject clauses differ from object clauses in that they are weak islands for wh-movement.

The acceptability of the passive example in (225a') raises the expectation that long wh-movement is also possible from subject clauses in unaccusative constructions. This seems to be confirmed by the fact that the modal verb blijkento turn out licenses long wh-movement provided that the anticipatory pronoun hetit is not present; cf. Bennis (1986: §2). Even speakers who consider wh-extraction in example (227b) marked with the expletive er will agree that there is a sharp contrast in acceptability with the version with the anticipatory pronoun het.

However, long wh-movement from subject clauses is quite rare, because subject clauses are usually obligatorily introduced by the anticipatory pronoun het. For example, the modal verb schijnento seem differs from blijken in that it does not allow the impersonal construction with the expletive erthere, so that long wh-movement is categorically excluded.

The primeless examples in (229) show that the anticipatory pronoun het cannot occur when the subject clause is in main-clause initial position, while the primed examples show that long wh-movement is nevertheless impossible. This shows that long wh-movement is only possible from subject clauses in clause-final position, although it is not clear whether this should be considered a restriction on wh-movement, as subject clauses are never possible in the middle field of the clause: cf. Koster (1978b).

We conclude with a brief digression on matrix verbs that normally select a prepositional object, such as klagen (over)to complain about. Although Section 2.3.1, sub VI, has shown that many of these verbs allow the omission of the anticipatory pronominal PP when the prepositional object is clausal, long wh-movement is usually excluded.

The verb hopen (op)to hope for seems to be an exception. Example (231a) first shows that this verb selects a prepositional object; the use of a nominal object (without op) leads to an unacceptable result. Example (231b) shows that the anticipatory pronominal PP erop can easily be omitted when the object is clausal; in fact, it is the preferred option. Finally, example (231c) shows that long wh-movement is acceptable when the pronominal PP is not present.

The examples in (231) therefore suggest that verbs selecting a prepositional object may license long wh-extraction after all. But things are not so simple, given that pronominalization of the embedded clause in (231c) may result in het: cf. De directeur hoopt het The manager hopes [for] it. In fact, het can also be used as an anticipatory pronoun with hopen, as in De directeur hoopt het [dat hij een grote bonus krijgt] The manager hopes [it] that he will get a big bonus. This shows that hopen can actually be a transitive verb if it selects a clausal complement. From this, we conclude that the acceptability of (231c) does not count as a counterexample to the claim that wh-extraction is not possible from prepositional object clauses.

Subsection III has shown that long wh-movement is only possible if the embedded clause has the syntactic function of subject or direct object. This does not mean, however, that long wh-movement is possible from any subject or direct object clause, as this may also depend on properties of the matrix predicate: while certain matrix verbs can function as so-called bridge verbs, others cannot. An important semantic factor involved is factivity: bridge verbs are non-factive in the sense that their use does not imply that the speaker presupposes the truth of the proposition expressed by their complement clause. This explains the contrast in acceptability between the two (b)-examples in (232); while the use of wetento know in (232a) implies that the speaker presupposes the proposition “Peter bought an Ipad” to be true, the use of denkento think does not.

There are several other factors that determine whether a particular verb allows long wh-movement. For example, although verbs of saying are typically non-factive, they do not allow long wh-movement if they also express a manner component: while the “neutral” verb zeggento say easily allows long wh-movement, the verb fluisterento whisper does not, as its meaning includes the additional manner component “without vibration of the vocal cords”.

The above discussion suffices to illustrate that long wh-movement requires not only that the embedded clause is an argument of the verb, but also that the matrix verb must meet certain criteria in order to function as a bridge verb. For further discussion, we refer the reader to Section 5.1.6, where the distinction between bridge and non-bridge predicates is discussed in more detail. Further restrictions on long wh-movement will be discussed in Section 11.3.1.3, where we will focus on so-called islands for wh-movement.

Long wh-movement is obligatory in standard Dutch in order to form a question in which a constituent of an embedded clause takes scope over a matrix clause; if long wh-movement is excluded for some reason, such a question simply cannot be formed. Since adverbial clauses do not allow long wh-movement (cf. Subsection III), it is impossible to question the object een vette bonusa big bonus in (234a), as is clear from the fact that the two (b)-examples in (234) are both unacceptable as regular wh-questions. The number sign indicates that with the right intonation pattern the utterance in (234b) can be interpreted as an echo-question or with an existential interpretation of watsomething, but we can ignore this here.

That long wh-movement is obligatory to derive questions in which a constituent of an embedded clause has matrix scope is also clear from examples such as (235), in which the object clause, unlike the adjunct clause in (234), does allow long wh-movement: the contrast between the two (b)-examples shows that leaving the wh-phrase in situ blocks the question interpretation. We have added the intermediate trace t'i to the representation in (235b') because its presence will become relevant in the discussion below.

The obligatoriness of long wh-movement is expected on the hypothesis (discussed in Section 11.3.1.1, sub II) that wh-movement is needed to create an operator-variable chain. However, it leaves unexplained that standard Dutch differs markedly from some of its dialects (as well as from German) in that it does not allow so-called partial wh-movement and/or wh-doubling. Partial wh-movement is illustrated in (236a) by an example taken from Barbiers, Koeneman & Lekakou (2010); it is characterized by the fact that the actual scope position of the wh-phrase (here: wie) is marked by some placeholder (here: the wh-element wat); the wh-phrase cannot remain in its clause-internal base position, but must at least move into the initial position of its own clause.

Wh-doubling is illustrated in example (237a), and is characterized by the fact that the wh-phrase is found not only in its scope position, but also in the initial position of the embedded clause; cf. Boef (2013) for a discussion of a similar phenomenon in relative clauses. Again, this construction is not part of the standard language.

According to Barbiers, Koeneman & Lekakou, the two (a)-examples in (236) and (237) can be seen as the result of successive cyclic movement if we adopt Chomsky’s (1995a: §3) copy theory of movement, according to which movement is a two-step operation: first, the content of the “moved” phrase is copied and then inserted at some higher position. The difference between long wh-movement and wh-doubling is simply that in the former case only the highest copy is spelled out phonetically, whereas in the latter case all copies in a clause-initial position are spelled out; this is indicated in (238), where the strikethrough indicates that the copy is not spelled out.

Partial wh-movement is analyzed in essentially the same way as wh-doubling, with the difference that watwhat is considered a partial copy of wiewho; these pronouns are the spell-out of virtually the same set of features, with the exception of [+human], which is missing in wat; cf. Barbiers, Koeneman & Lekakou (2010) for details. If the proposed analysis is on the right track, this would provide evidence for the successive cyclic movement approach to long wh-movement. Note, however, that the proposal is controversial; we refer to Schippers (2012: §4) and Pankau (2014) for extensive reviews of various proposals and further discussion.

Section 5.2 has shown that there are three formally different types of infinitival clauses: om + te-infinitivals, te-infinitivals, and bare infinitivals. Some typical examples are given in (239).

It seems that long wh-movement from om + te-infinitival clauses yields a more degraded result than long wh-movement from te-infinitival clauses. This can be easily demonstrated with the verb proberento try, since this verb is possible with both clause types; although some speakers object to both primed examples in (240), our informants consider (240a') much worse than (240b'). Note that we give the examples in the perfect tense to show that both examples involve extraposed clauses. The labels CP/TP indicate that the two types of infinitival clause differ in size; cf. Section 5.2.2 for a detailed discussion.

The degraded status of examples such as (240a') suggests that om + te-infinitivals differ from finite declarative clauses in that they do not accommodate long wh-movement via the specifier of their CP, which may in fact be consistent with the finding in Section 11.3.1.1, sub IV, that embedded infinitival wh-questions are not common in colloquial speech. If this is true, it implies that long wh-movement from te-infinitivals in examples such as (240b') differs from long wh-movement from finite declaratives in that it must be applied in one fell swoop; this is of course also suggested by the fact that TPs do not contain the position normally associated with wh-movement, the specifier of CP. That wh-movement is possible in one fell swoop in (240b') is not surprising in view of the fact, discussed in Section 5.2.2.3, that extraposed te-infinitivals are semi-transparent in the sense that they allow the infinitival clause to be split, as illustrated in (241). If this split is the result of leftward scrambling of the object het boek, there is no obvious reason to assume that leftward wh-movement of the interrogative pronoun watwhat would be impossible in (240b').

Section 5.2.2.3 has also shown that there are two types of te-infinitivals. The semi-transparent type, already illustrated in the (b)-examples in (240), is characterized by the fact that the matrix verb occurs as a perfect participle in the perfect tense and that splitting the infinitival clause is considered marked by at least some speakers. The transparent type is characterized by the fact that the matrix verb appears as an infinitive in the perfect tense and that splitting of the infinitival clause is obligatory in the northern variety of standard Dutch as a result of verb clustering. This type can again be illustrated by the matrix verb proberento try, since this verb can also take transparent te-infinitivals as its object. Examples like (242a) exhibit monoclausal behavior and it is therefore not surprising that wh-movement of the object of the infinitival verb sturento send is perfectly acceptable for all speakers.

Bare infinitival complements always exhibit monoclausal behavior; the examples in (243) show that, as expected, bare infinitivals freely allow wh-movement of the complement of the infinitival verb.

The above discussion suggests that wh-movement via a local escape hatch does not apply in the case of infinitival complement clauses, and that wh-extraction from such clauses must therefore apply in one fell swoop. Note, however, that the literature has not paid much attention to wh-extraction from om + te and te-infinitivals so far; it might be useful to investigate our claim in more depth, as judgments are not always very clear (perhaps due to the interference of constructions with infinitival goal clauses, which are also introduced by om: cf. Wat doet u [om af te vallen]?What do you do to lose weight?).