In multiple questions, wh-movement only applies to a single wh-phrase; the second (third, etc.) is left in situ. This may seem surprising at first, given the hypothesis discussed in Section 11.3.1.1, sub II, that wh-movement is needed to create operator-variable chains. For this reason, it has been argued that examples such as (271a) involve covert (invisible) movement of the second wh-phrase; cf. May (1985) and Lasnik & Saito (1992). However, it may also be the case that the second wh-phrase remains in situ because it does not independently take scope, as the formula ?x?y (x has read y) does not properly express the meaning of example (271a). Instead, multiple questions have a so-called pair-list reading, which is given in (271b). An appropriate answer is therefore a list of ordered pairs <x,y>: Marie has read Max Havelaar by Multatuli, Jan has read De Kapellekensbaan by Louis-Paul Boon, Els has read De zondvloed by Jeroen Brouwers, and so on.

We refer the reader to Van Riemsdijk & Williams (1986: §13), Dayal (2006/2017), and Bayer (2006) for reviews of proposals that are able to derive the pair-list reading without movement of the second wh-phrase. To avoid confusion, it should be pointed out that the notion of ordered pair used above of course refers to the specific case of only two wh-phrases. The notion of n-tuple would have been more appropriate to include cases with three or more wh-phrases, such as (272a), but we will follow the general practice of simply using the notion of pair-list reading.

Example (273a) shows that multiple questions need not be main clauses, but can also be embedded. An informal semantic representation of this example is given in (273b): John wondered for which ordered pairs <x,y> it is true that x has read y.

The wh-phrases in (271a), (272a) and (273a) are clausemates, but this is not necessary: example (274a) shows that the second wh-phrase can also be more deeply embedded. This example again has a pair-list reading, which is given in (274b). An appropriate answer should provide a list of pairs <x,y> such that it is true that x says that Peter is reading y: Marie says that Peter is reading Max Havelaar, Jan says that Els is reading De Kapellekensbaan, etc.

It is important to note that pair-list readings do not arise when the second wh-phrase occupies a scope position itself. This is illustrated in (275), where the second wh-phrase watwhat is wh-moved into the initial position of the embedded clause (as indicated by the trace; we have not indicated the trace of the matrix subject who for the sake of simplicity). Such examples can only be interpreted as in (275b); appropriate answers to such questions identify the agent of the matrix verb, but not the theme of the embedded verb: Marie (vroeg zich af wat Peter leest) Marie (wondered what Peter is reading). This shows that in multiple questions it is not only possible, but actually required that at least one wh-phrase remains in situ.

Multiple wh-questions have not only a special meaning, but also a characteristic intonation pattern: both wh-phrases must be accented, as indicated by the small caps in the (a)-examples in (271)-(274). This will help us to distinguish multiple wh-questions from regular wh-questions like the primeless examples in (276), where the unaccented pronoun wat is interpreted existentially, with the meaning “something”. This leads to the informal semantic representations given in the primeless examples.

The examples in (276a&b) also show that it is possible to include the expletive erthere in regular questions, which is consistent with the fact that the non-D-linked subject pronoun wiewho is compatible with it; cf. Wie komt er?Who is coming?. Although the judgments are subtle, it seems clear to us that adding the expletive to multiple questions like (271a)/(273a) is more difficult. If the resulting examples in (277) are indeed degraded, this would suggest that wh-phrases in multiple questions are (to some extent) D-linked. Of course, this would fit well with the pair-list reading of such questions, since they seem to involve entities from the domain of discourse.

The wh-phrases given in Subsection I are all arguments. The examples in (278) show again that the subject can form a pair with the direct object, a pair with the indirect object, or a triple with both objects.

The examples in (279) show that the subject need not be involved; the pair may also involve two objects; the two examples in (279) illustrate this for constructions with respectively a nominal and a prepositional indirect object.

The fact illustrated above that in situ wh-phrases can be embedded in prepositional indirect objects raises the expectation that they can also be embedded in prepositional objects. The examples in (280) show that this prediction is borne out. Note that the acceptability of example (280b) is special in that the sequence op watfor what is usually replaced by the pronominal PP waaropfor what in (280b'), but a Google search (June 11, 2024) shows that both sentences occur on the internet; the number of results, which have been manually checked, is given in square brackets. The acceptability of (280b') may be related to the pronoun being heavily accented.

Given the special nature of the (b)-examples in (280), we give another example of this alternation with the phrasal verb recht hebben (op)to be entitled to in (281). Both forms are common on the internet; the raw results of our Google search (June 11, 2024) are again given in square brackets. Example (281b) is interesting in its own right, as it shows that the R-pronoun waar is preferably moved leftward (a search for the unsplit pattern returns only 8 hit). This shows that the earlier claim that the second wh-phrase remains in situ is only true in so far as it cannot undergo wh-movement.

Multiple wh-questions are not affected by the position of the prepositional object (Koster 1987:213); the primeless examples in (282) show that the object op wie/wat can occur before or after the main verb in clause-final position. However, example (282b') shows that in the case of the pronominal PP waarop, placement before the main verb seems to give a better result.

Multiple wh-questions are also possible with wh-adjuncts. This is especially true for spatial waarwhere and temporal wanneerwhen, but it is also possible for adjuncts like waaromwhy and hoehow, which are impossible in English according to Lasnik & Saito (1992: ch1); cf. Haider (2010: §3.4), which claims that this is a more general difference between VO and OV-languages. To give an indication of the relative frequency of such cases, we give the results of our Google search (June 11 2024) on the string [wie heeft waar/wanneer/waarom/hoe] in square brackets. We manually checked the results, but stopped counting for waarom and hoe as soon as we found 20 relevant cases.

Haider (2010: §3.4) has shown that the same pattern is found in German, and also notes that adverbs like waarwhere and wanneerwhen can co-occur in multiple questions, while adverbs like waaromwhy and hoehow cannot (regardless of their order). This is also true for Dutch, as shown in (284).

Since the (b)-examples are perfectly acceptable when the second wh-phrase is omitted, we must be dealing with a co-occurrence restriction on waarom and hoe; cf. Haider (2010:119ff) for the claim that this restriction is universal and should be related to the semantic type of these adverbial phrases.

Subsection I mentioned that the fact that the second (third, etc.) wh-phrase is left in situ has led to the claim that it undergoes covert (invisible) movement. A serious problem for this claim is that the second wh-phrase can occur in positions where traces of wh-phrases cannot normally occur. We will illustrate this here for a number of islands that are strong in Dutch; cf. Section 11.3.1.3. In order not to complicate the discussion unnecessarily, we limit ourselves to wh-phrases functioning as arguments.

Multiple questions with interrogative proforms like wiewho, watwhat and waarwhere seem to adhere to fairly strict order restrictions in the sense that the canonical word order is not affected by wh-movement. The examples in (292) show that in transitive constructions the subject usually precedes the direct object, just as in declarative clauses such as dat <*dat boek> Jan <dat boek> gekocht heeft (which we have given in its embedded form to eliminate the interference of topicalization).

It is worth noting that examples like (292b) are claimed to be acceptable in German (Haider 2010:115), which may be due to the fact that the order of subjects and objects is less strict in German than in Dutch.

For ditransitive constructions the tendency to preserve the unmarked order in multiple wh-questions means that the order of the nominal arguments will be: subject > indirect object > direct object. We illustrate this in (293) for multiple wh-questions based on the reference sentence dat Jan/Hij Marie/haar een boek wil geventhat Jan/he wants to give Marie/her a book. The asterisk in (293b') indicates, of course, that the intended interpretation is not available.

Subjects and direct objects tend to precede prepositional indirect objects in multiple wh-questions, although speakers seem to be less rigid in this case. We illustrate this in (294) for questions based on the reference sentence dat Jan een boek aan Marie wil geventhat Jan wants to give a book to Marie. The relatively acceptable status of (294b') may be related to the fact that the prepositional indirect object can precede direct objects in focus constructions; cf. dat Jan aan Marie een boek wil geven.

Nom-dative verbs usually allow the theme-subject and the indirect object to occur in both the S-IO and the IO-S order (cf. Section 2.1.3, sub IIF), and this also holds for multiple wh-questions with these verbs. We show this in (295) for questions based on the reference sentence dat <dat boek> Peter <dat boek> goed is bevallenthat Jan likes that book a lot. This correspondence shows that examples of this kind cannot be used to argue that Dutch is like German in that it does not impose ordering restrictions on the subject and object in multiple wh-questions.

Subjects and direct objects usually precede prepositional objects, and (296) shows that this order is maintained in multiple wh-questions. The (a)-examples are based on the reference sentence dat Jan op zijn vader wachtthat Jan is waiting for his father and the (b)-examples on the reference sentence dat de rechter Peter tot het betalen van een boete veroordeeldethat the judge sentenced Peter to pay a fine.

Subjects also usually precede spatial/temporal adverbial phrases. Although there may be a slight preference for objects to precede such adjuncts, both orders seem to be acceptable in multiple wh-questions, which is consistent with the fact that the order of objects and spatial/temporal adverbial phrases also varies in the middle field of the clause: dat hij <de man> gisteren/in Amsterdam <de man> heeft ontmoetthat he met the man yesterday/in Amsterdam.

The generalization that seems to cover all the cases above is that the wh-phrase whose canonical position is closest to the clause-initial position will be the one that undergoes wh-movement. This generalization may follow from some version of Chomsky’s (1973) superiority condition (in which superiority refers to asymmetric c-command), if we take the view that linear order is ultimately derived from the structural, hierarchical relation between phrases; cf. Kayne (1994). We will not explore this option here, but simply use the notion of superiority condition as a convenient label for the generalization mentioned above.

Although the superiority condition provides a relatively adequate description of the order of the interrogative proforms in the earlier examples, it seems to conflict with cases involving more complex wh-phrases. This can be illustrated quite easily by examples with a complex wh-subject and a complex wh-object, since many speakers allow both orders in (298); cf. Dayal (2006: §2) for a review of similar facts from English.