The fact that wh-moved phrases consisting of a single word, such as watwhat in (145a), move into clause-initial position is expected on the hypothesis, discussed in Subsection II, that wh-movement derives an operator-variable chain in the sense of predicate calculus. However, this is not true for the fact that there are also cases of wh-movement in which wh-movement applies to phrases containing non-interrogative material, like welke autowhich car, wiens autowhose car, and wiens vaders autowhose fathers car’ in (145b-d); the non-interrogative parts of the wh-phrases are italicized.

The hypothesis that wh-movement derives an operator-variable chain requires only the movement of the interrogative pronouns; the movement of the non-interrogative material in these examples is therefore semantically superfluous. Consequently, there must be some other reason why wh-movement of the interrogative demonstrative and possessive pronouns in (145b-d) pied-pipes the non-interrogative parts of these noun phrases. The reason is syntactic: it is simply impossible to extract determiners from noun phrases in Dutch. The examples in (146) show that while it is possible to wh-move a complete direct object, it is impossible to extract an interrogative demonstrative pronoun from it.

The examples in (147) show essentially the same thing for possessive pronouns: while it is possible to wh-move a complete direct object, it is impossible to extract (a subpart of) a possessive determiner from it. The (a)-examples provide cases with the formal, genitive form wiens, while the (b)-examples provide cases with the more colloquial sequence wie z’n; in both cases pied piping is obligatory.

The examples in (148) are added to show that the wh-element need not be the determiner of the wh-moved noun phrase itself, but can also be more deeply embedded: the wh-element wiens is the determiner of the noun phrase wiens vader, which in turn is the determiner of the wh-moved noun phrase wiens vaders boek.

We conclude from the above discussion that pied piping is a repair strategy that is employed when wh-movement of the wh-element itself is blocked for syntactic reasons. Since we will limit ourselves in the next subsections to providing an empirical description of the contexts that disfavor wh-extraction and thus favor pied piping, we refer the reader to Corver (1990: §7-9) for a more theoretical discussion of the syntactic restrictions on wh-extraction (as well as a cross-linguistic examination of the relevant data).

Example (149a) shows again that pied piping of noun phrases can be triggered by interrogative demonstrative and possessive determiners like welkewhich and wienswhose; Subsection A has already shown that this may be related to the fact that it is not possible to extract determiners from noun phrases.

Interrogative determiners are not only capable of pied-piping head nouns, but also various other NP-internal constituents. This is particularly striking in the case of postnominal modifiers: while the primeless examples in (150) show that such modifiers can occur in extraposed position, the primed examples show that they must be pied-piped under wh-movement; cf. Guéron (1980). Note that the questions in the primed examples are special in that the use of the modifiers presupposes that the speaker has information that allows him to narrow down the set of possible answers: since we can assume that the addressee has more knowledge of the situation than the speaker and thus also has this information, the explicit mention of the modifier may feel somewhat forced. A more detailed discussion of the extraposition and pied-piping behavior of relative clauses can be found in Section N16.3.2.3.

Subsection A has shown that the wh-element need not be an immediate constituent of the pied-piped noun phrase, but can also be more deeply embedded: example (151a) illustrates this again for a possessive pronoun embedded in the determiner of a pied-piped noun phrase. Note in passing that constructions such as (151b) are sometimes judged less felicitous because the noun phrases tend to become difficult to compute; this is also true for non-interrogative noun phrases such as [[Peters moeders] auto]Peterʼs motherʼs car.

It also seems possible to embed the wh-element hoehow in a quantifier phrase such as veelmuch/many, although this fact may be obscured by the orthographic convention of writing the formation hoe + veel as a single word. The fact that hoeveel in (152b) corresponds to heel veelvery many in (152a) strongly suggests that we are actually dealing with two separate words.

Example (152b) also shows that the interrogative quantifier hoeveel cannot be extracted from its noun phrase, thus supporting the hypothesis put forward in Subsection A that pied piping can be regarded as a repair strategy. Note that the extraction of hoe is also impossible: *Hoei heeft Marie [ti veel boeken] gelezen?

Speakers occasionally seem to have different judgments on pied piping triggered by a more deeply embedded wh-element. This is well illustrated by examples such as (153b), in which the wh-element hoehow corresponds to the degree adverb ergvery in (153a). Judgments on such examples vary: Broekhuis (2013: §5.2.1) gave similar examples as perfectly acceptable, while Corver (2003:292) claimed that such examples can only be interpreted as echo-questions; yet another judgment can be found in Vos (1994:130), where examples such as (153b) are assigned a question mark. Of course, it is difficult to decide whether the markedness of (153b) results from a syntactic constraint or from the computational complexity of the structure, which speakers can easily avoid by using the more or less synonymous but computationally simpler question Hoe groot is Els haar autoHow big is Els car?’.

For completeness’ sake, note that it is also possible to find examples such as (154) on the internet. Although Vos (1994) claims that such examples are acceptable in colloquial speech, we doubt that the construction should be considered part of the standard language, as many speakers reject it; cf. Corver (2003) for a more detailed discussion of this construction.

The examples discussed so far show that wh-elements located to the left of a nominal head can pied-pipe the non-interrogative part of the noun phrase. Pied piping seems to be more difficult when the wh-element is to the right of the nominal head. This contrast can be seen in the constructions with a prenominal genitive possessor and a postnominal van-PP in the (a) and (b)-examples of (155); cf. Vos (1994:130).

At first glance, example (156a) seems to show that the contrast between the two primed examples in (155) can be derived from the hypothesis that pied piping is a repair strategy: the acceptability of example (156a) suggests that the postnominal van-PP can be extracted from the noun phrase. A closer look, however, shows that the van-PP need not be interpreted as the possessor of the direct object. First, (156b) shows that the interrogative van-PP can also be used when the possessor is expressed by a possessive pronoun, which makes it very unlikely that the van-PP also functions as a possessor: cf. *[zijn auto van Peter] (lit. his car of Peter). Second, (156c) shows that the direct object can be pronominalized without affecting the van-PP, whereas pronominalization usually affects the whole noun phrase. The examples in (156b&c) thus show that the van-PP must have some other syntactic function, probably that of indirect object with the thematic role of source; cf. Section 3.3.1.3.

The above discussion shows that we should be careful not to jump to the conclusion that wh-extraction of PPs from noun phrases is possible, but that each case should be examined on its own merits. For example, it is striking that most of the examples analyzed in the literature as cases of wh-extraction of PPs from noun phrases involve PPs headed by van or voor. However, Section N15.2.1.5, sub III, has shown that van and voor-PPs can also be used as restrictive adverbial phrases. It is therefore imperative to examine cases with other prepositions in order to establish conclusively that wh-extraction of postnominal PPs is possible, but the primed examples in (157) show that such cases usually do not allow wh-extraction; cf. Vos (1994:139-40) and Broekhuis (2016).

Now let us return to the hypothesis that pied piping is a repair strategy. Since stranding is excluded or at least quite marginal in the primed examples in (157), this hypothesis predicts that the pied piping examples in (158) are acceptable, but this is not borne out: these examples are impossible as wh-questions.

It follows that the hypothesis that pied piping is a repair strategy should not be interpreted in such a way that pied piping of the remainder of the noun phrase can be used to form the desired question whenever wh-extraction is excluded; it may be the case that certain semantically plausible questions simply cannot be formulated for reasons yet to be determined; cf. De Vries (2002: §8.5) for a concrete proposal.

The examples in (159) show that prepositional phrases with an interrogative pronominal complement require pied piping; this is illustrated for such PPs in various syntactic functions. Since stranding of the preposition would invariably lead to unacceptability, these examples are fully consistent with the hypothesis advanced in Subsection A that pied piping can be regarded as a repair strategy. We will ignore the stranding data in this subsection, but return to it in Subsection VI, where the stranding behavior of post and circumpositional phrases is discussed. Note in passing that all the examples in (159) involve the [+animate] pronoun wiewho; we will see in Subsection VI that the [-animate] pronoun watwhat is also not possible in this context, but we will ignore that for now.

The wh-element triggering pied piping need not be the complement of the pied-piped PP, but can be more deeply embedded. The examples in (160) illustrate this for a prepositional object, a complementive, and an adverbial phrase in which the wh-element functions as a determiner of a nominal complement of the pied-piped PP.

The examples in (161) show that the degree of embedding can be even greater. Example (161a) shows that the degree modifier hoehow of a quantifier of a nominal complement of a PP will pied-pipe the entire PP, and (161b) shows that the same is true for the degree modifier hoehow of an attributive modifier of a nominal complement of a PP. A Google search (June 5, 2024) on the string [met een hoe grote] has shown that such examples can be easily found on the internet in both main and embedded clauses, despite being syntactically more complex than the disputed example %[Een [hoe grote] auto] heeft Els?How big a car does Els have in example (153b) from Subsection B.

In the examples discussed so far, the wh-element is in the prenominal position of a nominal complement of the PP. At first glance, the result seems to be infelicitous if the wh-element is in postnominal position, as shown by the primed examples in (162). The percentage signs indicate that such examples are often claimed to allow only an echo-interpretation (Vos 1994:127), while there are also speakers who allow them as regular wh-questions and attribute their markedness to computational complexity (De Vries 2002: §8.5).

Vos (1994) has further shown that pied piping is perfectly acceptable in wh-questions such as (163a). She suggests that this is only possible if the postnominal PP functions as a modifier, but this would incorrectly predict that examples such as (163b) would be unacceptable, since the relational noun centrum clearly selects the PP van welke stad (cf. Section N14.2.3); to our ears, this example has more or less the same status as (163a).

An alternative explanation for the acceptability contrast between the wh-examples in (162) and (163) could be that the complex noun phrases in the primed examples in (162) alternate with the structures with a possessive pronoun ([op/achter [wiens broer]]), whereas such alternants are not available for the noun phrases in (163). The markedness of the primed examples in (162) could then be attributed to syntactic blocking, in the sense that the structures with a possessive pronoun are simply preferred; this would be consistent with De Vries’ claim that the markedness of the primed examples in (162) is due to their computational complexity.

There may be other factors that affect acceptability judgments. Vos (1994) claims that an example such as (164a) is unacceptable, even though it seems to involve the same degree of syntactic complexity as the examples in (163). De Vries (2002) considers similar cases to be acceptable but difficult to understand, suggesting that the infelicity of this example may again be non-syntactic in nature. What we want to suggest here is that the cause of the awkwardness lies in the nature of the nouns involved: the nouns broerbrother and vriendfriend in (164) are both relational nouns and can therefore only receive a correct interpretation if the relational argument is known to the addressee. Out of context, this condition is not fulfilled for the noun vriend in (164a), and it may be that this is the reason for the degraded status of this example. That this line of thinking may be on the right track is supported by the fact that example (164a) improves if we replace the relational noun vriend with a non-relational noun such as meisjegirl, as in (164b), which is as acceptable as (163b).

The discussion above suggests that wh-elements in postnominal position are capable of triggering pied piping of PPs, unless they are part of a postnominal van-PP that has an alternative expression as a prenominal possessive pronoun. The discussion of the examples in (164) has also shown that in some cases there may be non-syntactic factors at play that obscure the proper syntactic generalization; since these factors have not been fully investigated, we must leave them to future research.

Pied piping of APs is quite limited and usually involves the interrogative degree adverb hoehow. This is illustrated in (165) for a complementive and an adverbially used AP. The fact that the adjectives cannot be stranded shows that such cases are consistent with the hypothesis put forward in Subsection A that pied piping is a repair strategy.

The degree adverb hoe can also be embedded more deeply as part of a gradable degree modifier, such as drukbusy in (166). The (b)-examples show that pied piping sometimes leads to a marked result in such cases; the preferred option seems to be wh-extraction of the complete adverbial modifier, although preferences seem to vary from case to case and speaker to speaker, which is why Corver (1990: §8) marked both (b)-examples as grammatical.

The contrast in the stranding behavior of the simple degree modifier hoe and the complex modifier hoe A is again illustrated in the examples in (167): while the complex wh-phrase hoe goed in (167a) must be interpreted as a degree modifier of bereikbaar, the simplex wh-phrase hoe in (167b) cannot; it can only be interpreted as a manner adverbial.

The preference for stranding (if real) only holds in cases such as (166), where the AP is a complementive. In other functions, such as supplementive, pied piping is the only option. This contrast is illustrated in (168).

For completeness’ sake, the examples in (169) show that adjectives cannot be pied-piped by their complement: (169a) shows that wh-movement of the PP op wie must strand the adjective boos and (169b) shows that wh-movement of the noun phrase welke opera must strand the adjective zat. We refer the reader to Section A25.1.2.4 for more data and a more detailed discussion.

We conclude by noting that pied piping of APs by other wh-elements to the right of the adjectival head does not seem to occur; Subsection VI will show that such wh-elements normally strand (part of) the AP.

Wh-movement does not pied-pipe verbal projections. For instance, example (170) shows that wh-movement of a direct object cannot pied-pipe the VP of which it is a part, but must be extracted from it. Note that in contrast, Section 11.3.3, sub VIC, will show that VP-topicalization is possible: [Een boek gelezen]i heeft Jan ti.

Similarly, a wh-phrase that is part of an embedded object clause cannot pied-pipe the clause (although topicalization of clauses is possible), but must be extracted from it.

The patterns in (170) and (171) are expected on the basis of the hypothesis put forward in Subsection A that pied piping is a repair strategy. However, pied piping of an embedded clause is also impossible if wh-extraction is blocked. The examples in (172) show this for a case in which the wh-element is part of an adverbial clause; (172b&c) show that pied piping and stranding both lead to an unacceptable result, and (172d) shows that refraining from wh-movement is not an option either. As a consequence, it is simply impossible to formulate the desired question. Note that the linear strings in (172c&d) are acceptable as declaratives if wat is interpreted as the existential quantifier “something”, but this is of course not intended here.

The impossibility to formulate certain questions is not as exceptional as it may seem at first glance, as this is typically the case when a wh-element occurs in a so-called syntactic island; we will return to this issue in Section 11.3.1.3.

This subsection has shown that pied piping is possible when the wh-element is embedded in a noun phrase, an AP or a PP, but impossible when it is embedded in an (extended) verbal projection. For the cases discussed here it seems observationally adequate to say that pied piping is possible whenever stranding (wh-extraction) is excluded. Subsection VI on stranding will discuss more cases that are also covered by this generalization, while Subsection VII will show that there are also cases that violate it. The discussion in this subsection was somewhat complicated by the fact that the judgments in the literature are sometimes contradictory; we argued that this may be due to the interference of a number of non-syntactic factors, which should be further investigated in the future. More extensive data sets on pied piping are given in Corver (1990: §7-10), Vos (1994), and De Vries (2002: §8.5).

Subsection VC has shown that wh-movement of the nominal complement of a prepositional phrase usually pied-pipes the whole PP. However, this subsection will show that there are also cases in which wh-movement of the nominal complement strands the adposition; this holds for the pronominal PP waar+PP it, as well as for postpositional and circumpositional PPs. Our discussion of these cases is followed by an attempt at analysis. We conclude with a discussion of stranding by wh-movement of the modifier of a PP.