Nominal arguments precede the clause-final verbs. This is true for both subjects and direct objects alike, regardless of whether they are indefinite or definite.

This restriction is especially clear in the case of indirect objects: while prepositional indirect objects can easily be extraposed, their nominal counterparts cannot. To eliminate possible interference from the presence of a direct object, the examples in (22) illustrate this using a regular passive construction; example (22b) is only possible when the preposition aan is present.

One ostensible exception to the general rule that nominal arguments cannot be extraposed has already been discussed in Section 12.1, sub IV: afterthoughts and backgrounded noun phrases can be placed postverbally. However, we have seen that these should not be considered as extraposed phrases, but as right-dislocated, parenthetical constituents. VP-topicalization can be used to support this view. The examples in (23) first show that a direct object must be pied-piped under VP-topicalization if it is in its base position; under neutral intonation (i.e. without contrastive accent) the direct object can only be stranded if it has previously left the VP by leftward scrambling across the adverb graaggladly.

Example (24b) shows that right-dislocated noun phrases can easily be stranded in postverbal position, while the (c)-examples show that pied piping is only possible in the case of afterthoughts, in which case we have to use quite distinct intonation breaks, and even then some speakers tend to reject it.

Enumerations, as in example (25a), are another possible exception to the general rule that nominal arguments must precede the clause-final verbs; cf. Haeseryn et al. (1997:1376). Such enumerations are preceded by an intonation break and cannot be pied-piped under VP-topicalization, which again suggests that they are parenthetical in nature: such examples are therefore special in that the “true” direct object need not be pronounced.

Haeseryn et al. also notes that in formal contexts, nominal arguments may occasionally occur postverbally. This order, called “expressive”, is quite obsolete: it is used especially when the postverbal noun phrase represents newsworthy information: example (26) might be used as “breaking news” in a news broadcast, but not in a biography as a neutral way of expressing where and when the singer André Hazes died. Such cases are clearly part of the periphery of the language and should therefore be ignored in a synchronic syntactic description of the core grammar.

Finally, note that free relatives (i.e. relative clauses without an overtly realized antecedent) can easily be found in postverbal position, just as relative clauses with an overt antecedent can. If free relatives were noun phrases, this would be a counterexample to the claim that nominal arguments cannot be extraposed, but the examples in (27) show that the two cases can be unified if we assume that the antecedents of free relatives are syntactically present but lack phonetic content. We will return to the extraposition of relative clauses in Section 12.4.

Clausal complements occupy the postverbal position, as in (28a). Normally, it is not possible for complement clauses to precede the postverbal verbs: example (28b) is only acceptable as a direct speech construction, i.e. if the speaker intends to express that Jan literally uttered the phrase “dat het hem spijt”; cf. Section 5.1.2.4, sub II, for a discussion of such cases.

Factive clauses, such as the bracketed phrase in (29), are another apparent exception to the general rule, but Section 5.1.2.3 has shown that it is plausible that the preverbal clause in (29b) is actually nominal in nature; cf. that section for a detailed discussion.

Example (30b) shows that the clausal complement in (28a) can be pied-piped under VP-topicalization; we have added some adverbial material to the sentence to make the resulting structure more balanced. The fact that pied piping is possible strongly suggests that the complement clause is part of the verbal projection. This conclusion may be supported by the fact that stranding of the complement clause is definitely marked compared to pied piping.

The (b)-examples in (31) show that the results are quite different when the clause is introduced by the anticipatory pronoun hetit. The fact that the clause must be stranded in this case suggests that it has a different position than the argument clause in (30), which is not introduced by hetit; it is not extraposed (i.e. clause-internal) but right-dislocated (i.e. main-clause external); cf. Section C37.3 for a discussion of right dislocation.

This conclusion is also supported by the fact that argument clauses that are not introduced by het show a different behavior with respect to wh-extraction than the corresponding clauses that are introduced by het; Section 11.3.1.1, sub III, has shown that wh-extraction is allowed only in the absence of this anticipatory pronoun. If the anticipatory pronoun functions as the true direct object, while the clause is merely an apposition, this follows from the claim that wh-extraction is possible only from complement clauses; cf. the discussion in Subsection A.

That the anticipatory pronoun functions as the true object is supported by the fact, illustrated in (33), that its associate clause is optional: direct objects are normally obligatory, and it is clear that the pronoun must be present if the clause is omitted. Note in passing that the number sign indicates that the string without the pronoun is used in academic circles as a translation of Latin dixiI have spoken meaning “I have said all I had to say”; this is clearly not part of Dutch core grammar and can thus be ignored in our syntactic description.

This subsection has shown that argument clauses are obligatorily extraposed. This was shown only for finite clauses, but the same holds for opaque and semi-transparent infinitival argument clauses, while transparent infinitival argument clauses undergo a process of verb-cluster formation. Since discussing this would simply repeat much of the discussion in Section 5.2, we will not digress here.

The examples in (34) show that PP-complements typically occur before or after the verbs in clause-final position, where we also find (main-clause external) right-dislocated phrases, such as afterthoughts and backgrounded phrases; cf. Section C37.3.

Extraposed PP-complements are fairly easy to distinguish from afterthoughts and backgrounded phrases: because arguments are usually obligatory, afterthoughts and backgrounded phrases require some anchor in the “true” argument position. This can be easily illustrated with the verb houdento like, which obligatorily takes a van-PP as complement: cf. the unacceptability of (35a). The examples in (35b&c) show that the presence of a pronominal PP such as daarvanof that is only possible, and then in fact obligatory, if the postverbal PP is preceded by an intonation break.

However, some verbs such as wachtento wait in (36a) optionally take a PP-complement. In such cases the resulting pattern is different. The (b)-examples in (36) first show that postverbal PPs must be preceded by an intonation break when a pronominal PP such as daaropfor that is present; in this respect constructions with an optional PP-complement behave just like constructions with an obligatory PP-complement. Recognizing afterthoughts is not very difficult, since the PP is preceded by an intonation break and given a contrastive accent, but distinguishing between extraposed and backgrounded PPs is more difficult, since it relies mainly on the intonation break, which may not be very prominent in speech. The main point for our present purposes, however, is that the intonation break in the (c)-examples is optional in slow, careful speech; we can therefore conclude that both extraposition and backgrounding right-dislocation are available.

There are at least two reasons to assume that extraposed PPs are part of the clause. The first reason is phonological in nature and concerns the fact that (non-contrastive) sentence accent can easily be placed on the extraposed PP; in fact, it is the neutral placement of this accent. In the case of right dislocation, on the other hand, the sentence accent must precede the right-dislocated PP. This is shown in (37), where the sentence accent is indicated in italics.

The fact that sentence accent can occur on extraposed PPs conclusively shows that extraposed PPs are clause-internal. A second reason for assuming this is that they can be pied-piped under VP-topicalization, as shown in (38a), although it should be noted that some speakers prefer the order in (38a'), in which the PP-complement is preverbal; this may be due to the fact that there is no information-structural reason for extraposition, given that the clause-initial VP as a whole functions as a focus of topic. Example (38b) shows that stranding the PP-complement leads to a degraded result when the sentence is pronounced with a neutral (non-contrastive) intonation pattern.

Because the contrast between the two primeless examples in (38) is not as sharp as one would like, we illustrate the same again in (39) with the verb rekenen, which takes a PP-complement headed by opon when used as a PO-verb meaning “to count/bank (on)”.

Backgrounded PPs are not easily pied-piped by VP-topicalization, as can be seen from the fact that at least some speakers allow them only in postverbal position, as indicated by the judgments on the (a)-examples in (40). Example (40b') shows that afterthoughts are marginally possible after topicalized VPs with an anticipatory pronominal PP, but only if they are preceded and followed by very distinct intonation breaks. Example (40b) shows that afterthoughts can also occur in postverbal position. We illustrate the same again in (41) with the verb rekenen (op)to count (on).

The discussion above suggests that extraposed and right-dislocated PPs occupy different positions. Since extraposed PPs are like extraposed clauses in that they are obligatorily pied-piped under VP-topicalization, the simplest theory would be that they occupy the same structural position in the clause. If this were true, we would expect extraposed PPs to behave like extraposed clauses in that they allow wh-extraction. However, this expectation is not borne out as extraposed PP-complements are islands for wh-extraction; the examples in (42) show that wh-extraction is possible only if the stranded preposition immediately precedes the clause-final verbs.

The previous subsections have shown that nominal, clausal, and prepositional arguments exhibit different extraposition behavior, as indicated in Table 1.

type of argument	extraposition option	islandhood of extraposed phrase
nominal	impossible	N/A
clausal	obligatory	extraction possible
prepositional	optional	extraction impossible

In early generative grammar it was generally assumed that Dutch has an underlying OV-structure: objects are uniformly base-generated to the left of the clause-final verbs. This implies that constructions with extraposed objects are derived by rightward movement. De Haan (1979) pointed out that this movement analysis of extraposed object clauses is problematic in view of the fact that they allow wh-extraction in bridge-verb contexts; this is inconsistent with the movement analysis because movement creates syntactic islands (the so-called freezing effect). De Haan therefore concluded that argument clauses are base-generated to the right of the clause-final verb position.

If nominal and clausal direct objects have the same underlying base position, there is only one option left: they are base-generated in the surface position of the clause, and the nominal phrase undergoes an obligatory leftward movement to a position preceding the clause-final verb position. Although this raises the question why extraction from nominal arguments is sometimes possible (as is clear from so-called wat-voor splits), this position seems to be currently taken by many (but not all) generative linguists; cf. Zwart (1997/2011: §9) and Broekhuis (2008: §2).

The fact that extraposed PP-complements allow wh-extraction only in preverbal position strongly suggests that they differ from extraposed argument clauses in a non-trivial way. More specifically, they differ from extraposed clauses in that they cannot be base-generated in postverbal position. In principle, there are two ways to account for extraposed complement PPs: either the PP is moved rightward across the verb into the postverbal position, as was commonly assumed in early generative grammar, or some verbal projection is moved leftward to a position to the left of the PP; we refer the reader to Barbiers (1995a) and Kayne (2000: Part III) for the latter option.

What is most important for our present discussion is that we can conclude that extraposition should not be regarded as a uniform phenomenon that can be explained by a single (movement) rule. The ramifications of the pattern given in Table 1 are currently still under investigation; an overview of a number of theoretical options is given in Section 9.4, sub I, to which we refer the reader for further discussion and suggestions for further reading.