The differences in the placement of nominal and clausal complements relative to the clause-final verbs shown in (48a&b) have been a focus of attention since the rise of early generative grammar. The assumption that direct objects are inserted in the complement position of the verb inevitably led to the conclusion that the alternate placement of direct objects in the sentence is the result of some movement transformation. Thus, the question arose as to which is the base position of the direct object: that of the nominal complement in (48a) or that of the clausal complement in (48b)? The consensus on this question in the mid-1970s seemed to be that Dutch is underlyingly an OV-language, and that objects must therefore uniformly be base-generated in preverbal position; examples such as (48b) are thus derived by an obligatory extraposition rule, which moves the clause from the preverbal object position to some postverbal position; cf. Koster (1973/1974/1975).

Although the extraposition approach remained dominant until the mid-1990s, it was clear from the beginning that it was not without problems; cf. De Haan (1979). The most conspicuous problem had to do with freezing: since extraposition is movement, and movement usually produces a freezing effect, the extraposition approach predicts that clausal complements are islands for extraction; however, the fact that wh-movement takes place from the postverbal embedded clause in example (49) shows that this prediction is false. Recall from Section 9.3, sub I, that the intermediate trace t'i is due to the fact that long wh-movement takes place via an escape hatch in CP.

One way to save the assumption that Dutch is underlyingly an OV-language and thus requires the direct object to be base-generated in preverbal position would be to assume that the postverbal clause is actually not the true object of the verb, but that it is a dependent of a phonetically empty anticipatory pronoun comparable to het in (50a), where the indices indicate the relation between the pronoun and the clause. The (empty) pronoun would then be the true object of the verb; cf. Koster (1999). However, this analysis is generally rejected because (50b) shows that the presence of an overt anticipatory pronoun normally blocks wh-movement from the embedded clause; cf. Hoekstra (1983), Bennis (1986), and many others. Note that we have included the particle nog in these examples because some speakers seem to prefer some material between the anticipatory pronoun and the clause-final verb.

If we want to maintain that nominal and clausal objects are base-generated in the same position, the obvious alternative to explore is to assume that they are both base-generated in postverbal position, and that the nominal object is moved to some preverbal position. This approach has become popular since Kayne (1994), where it was argued that rightward movement is excluded on general grounds, so that movement is uniformly to the left. One virtue of this approach is that we know independently that noun phrases can be moved to higher/more leftward positions; for example, it is commonly assumed that the subject of a passive clause is (or at least can be) moved from the position occupied by the direct object of the corresponding active clause into the regular subject position of the clause, as in (51b), in order to obtain nominative case and/or to establish agreement with the finite verb.

In line with this tack, we might assume that the nominal object also moves from its underlying postverbal position to some higher position where it can be assigned accusative case or establish abstract (i.e. phonetically invisible) object-verb agreement (which is morphologically expressed in many other languages). A possible problem with this proposal is that it incorrectly predicts freezing of the nominal direct object; example (52a) shows that the phrase wat voor een boekwhat kind of book functions as a single nominal phrase, which strongly suggests that (52b) is derived by extracting the element wat from this complex phrase, and thus that nominal objects are not islands for wh-extraction.

The discussion above shows that we can only maintain the assumption that nominal and clausal complements are base-generated in the same position if we assume that certain obligatory movement operations do not result in freezing; cf. Broekhuis (2008) for a proposal to this effect. However, it may also be the case that the presupposition that nominal and clausal complements are base-generated in the same position is incorrect, and that they are simply base-generated in a preverbal or postverbal position, respectively, as proposed in De Haan (1979:44) and Barbiers (2000) for Dutch, and in Haider (2000) for German. A possible problem with this solution is that the verb and the postverbal clause form a base-generated constituent, which leads to the false prediction that postverbal clauses must precede extraposed PPs: this means that extraposition of the PP tegen Peterto Peter in (53a) should lead to the order in (53b), whereas the order in (53b') is actually much preferred.

This subsection has briefly discussed three approaches to the placement of nominal and clausal arguments: two movement approaches (one involving rightward movement of clausal and one involving leftward movement of nominal arguments) and one base-generation approach. We have seen that they all run into various possible problems for which special provisions should be made.

Subsection A has shown that nominal and clausal complements are strictly ordered with respect to the clause-final verbs. This does not hold for PP-complements, which can normally occur either to the left or to the right of these verbs.

In such cases, it seems easy to determine the base position of the PP: assuming that the two positions are related by movement, we predict that the PP in the derived position will exhibit a freezing effect. The fact, illustrated by the examples in (55), that R-extraction is possible from the preverbal but not from the postverbal PP thus leads to the conclusion that the preverbal position is the more basic one; cf. Koster (1978: §2.6.4.4), Corver (2006b/2017) and Ruys (2008). This can be taken as support for an underlying OV-order in Dutch, assuming that PP-complements are base-generated in the complement position of the verb.

The conclusion that the postverbal placement of PP-complements is the result of an extraposition operation, which has become known as PP-over-V, seems almost inescapable if one assumes that movement invariably produces a freezing effect. However, there are also serious problems with the claim that the stranded prepositions in (55) occupy the complement position of the verb. First consider example (56a), which shows that so-called VP-topicalization involves movement of a larger verb phrase that may include at least the direct object, i.e. the complement position of the main verb. The earlier conclusion that stranded prepositions must occupy the base position of the PP-complement thus implies that they are VP-internal and consequently must be pied-piped by VP-topicalization. However, the (b)-examples in (56) show that the stranded must be stranded; pied piping leads to a severely degraded result; cf. Den Besten & Webelhuth (1990).

If we accept the freezing effect as a diagnostic for movement, the acceptability of (56b) suggests that PP-complements are not base-generated as a complement of the verb at all, but VP-externally. Analyses of this kind have indeed been proposed on independent grounds and amount to saying that extraposition of PPs does not result from rightward movement of the PP but from an (optional) leftward movement of the VP into a position left-adjacent of the PP, as in (57). An early proposal of this kind can be found in Barbiers (1995a), where it is claimed that the landing site of the VP is the specifier of the PP, and that this turns the PP into an island for extraction; this account for the fact, illustrated in (55), that R-extraction is only possible when the PP precedes the verb.

A possible problem for (57) is that PP-complements are not generated within the lexical projection of the verb, but this can be solved if we follow Kayne (2004), which claims that PP-complements of verbs are not inserted as a unit but derived in the course of the derivation; the preposition is inserted as a functional head, which attracts a nominal complement of the verb. After this, the VP can be moved into the specifier of the PP. The (simplified) derivation is given in (58).

The proposals in (57) and (58) have given rise to a rich research program; cf. Den Dikken (1995) for an attempt to describe the extraposition of clauses discussed above in terms of leftward VP-movement, and Kayne (2000: Part III) and later work for similar attempts to explain a wide range of seemingly rightward movement phenomena in these terms. Because of the highly technical nature of these proposals, we will not discuss them, but mention them here as evidence for the claim that it is not a priori obvious whether PP-complements are base-generated to the left or to the right of clause-final verbs and, even more surprisingly, that it is not even obvious that they are base-generated in the complement position of the verb.

The early extraposition approach considers the clause-final verbs as the pivot around which a number of syntactic processes take place. Complements are inserted in preverbal position and various category-specific movement rules lead to a reordering of the verb and its complements. Such rearrangements are excluded for nominal complements, obligatory for clausal complements, and optional for PP-complements. The central role ascribed to the verb is aptly captured by the term PP-over-V in the case of extraposition of PPs. Later research has shown, however, that the pivotal role of verbs may be an accidental property of Dutch and German (i.e. the Germanic OV-languages). This can be illustrated by the English examples in (59).

Despite the fact that nominal, clausal and prepositional complements all follow the main verb in English, it is clear that they exhibit a distributional difference similar to the corresponding elements in Dutch. The fact that clausal complements must follow time adverbs such as yesterday, whereas nominal complements usually precede such adverbs, shows that these complements occupy different positions. The fact that the PP-complement can either precede or follow the adverbial phrase a long time reflects the distributional behavior of the Dutch PP. The correspondence between the Dutch and English examples shows that it is not so much the position of the complements relative to the verb that is at stake here, but rather their absolute position; in both Dutch and English, the three types of complements simply occupy different positions in the clause. The hypothesis would therefore be that Dutch and English behave identically when it comes to the placement of the complements of the verb, but differently when it comes to the placement of the verb itself. One implementation, which seems to be widely accepted by the current generation of generative grammarians, is the claim that the lexical domain of the clause is not just a simple projection of the verb V, as suggested by the representation in (10), repeated here as (60a), but consists of at least two projections: one headed by a root element, usually (somewhat misleadingly) represented by V, and another headed by a so-called light verb v, as indicated in (60b); cf. Chomsky (1995). Recall that X in this structure stands for an unspecified number of functional heads that may be needed to provide a complete description of the structure of the clause.

The basic intuition behind the structure in (60b) is that all verbs are in fact derived from some nonverbal root by affixation with the verbal morpheme v. Although the Dutch light verb v is usually phonetically empty, this hypothesis is empirically supported by Latinate verbs such as irriterento irritate: this verb can be taken to be derived from a nonverbal root irrit-, which can also be used as input for the adjective irritant or the noun irritatie. The Dutch light verb v can thus be seen as a zero morpheme comparable to -eren in (61a).

The correspondences between Dutch and English can now be explained by assuming that in these languages nominal, clausal and prepositional complements occupy the same surface positions in the clause, while the differences can be explained by assuming that the root V moves to (merges with) the light verb v in English but not in Dutch embedded clauses. This is shown for nominal and clausal complements in (62). The presumed difference in V-to-v movement between English and Dutch can in fact be held responsible for the fact that English emerges as a VO-language, while Dutch emerges as an OV-language; cf. Broekhuis (2008/2011/2023) and Barbiers (2000) for more detailed discussion.

62

Note that the schematic representation in (62) is not intended to make any claim about the base positions of nominal and clausal complements; it may well be that VP is actually a larger constituent within which the nominal or clausal complement has moved to its surface position; cf. Johnson (1991), Koizumi (1993), and Broekhuis (2008) for leftward movement of nominal objects within this VP-domain.

This subsection has briefly discussed the distribution of postverbal arguments: nominal and clausal arguments occur pre and postverbally, respectively, while PP-complements can occur on either side of the clause-final verbs. Starting from the claim that complements are all base-generated in the complement position of the verb, generative grammar has attempted to account for the different placement options by means of specific rearrangements in the clause. Early proposals included obligatory extraposition of clausal arguments and optional PP-over-V. Since the mid-1990s, proposals have been developed that involve leftward movement of nominal complements and verbal projections. And there are also proposals that simply reject the claim that nominal and clausal arguments are base-generated in the same position. The debate over the derivation of extant surface orders is ongoing and far from settled, and this subsection has reviewed only a small number of empirical facts that have played a crucial role in motivating/testing the various proposals. A more detailed description of the data can be found in Section 12.1.