Section 36.2.1, sub III, has shown that PP-complements and PPs introducing a specific semantic role (e.g. instrument) allow R-pronominalization. When these PPs function as clausal constituents, they typically allow R-extraction as well, and this subsection discusses the distribution of the resulting stranded prepositions in the clause. We will start with PP-complements of verbs. Example (111a) shows that such PP-complements can occur in various positions within the clause. Their unmarked position seems to be left-adjacent to the verb(s) in clause-final position, but they can also occur more to the left (especially if the nominal part of the PP is contrastively accented) or in extraposed position. This is supported by the fact, illustrated by (111b), that stranded prepositions of PP-complements are also usually left-adjacent to the verb(s) in clause-final position: an adverbial phrase (here: the negation niet ‘not’) between the stranded preposition and the clause-final verb(s) leads to unacceptability, and the same is true if the stranded preposition is postverbal.

An example such as (112b) is only an apparent exception to this general rule since it does not involve R-extraction. Example (112a) first shows that R-extraction of the demonstrative R-word daar is not obligatory. Consequently, (112b) can be derived by leftward movement of the complete pronominalized PP daarover ‘about that’.

The R-pronoun daar can undergo R-extraction from the PP-complement of spreken ‘to talk’, but only when the pronominalized PP is left-adjacent to the clause-final verb(s), as in the perfectly acceptable example (113a), which is derived from (112a). Example (113b) is derived from (112b) and is completely unacceptable due to a freezing effect: the R-pronoun daar is extracted from a PP that has been A'-scrambled (cf. Section V13.3), thus forming an island for extraction; cf. Koster (1978a: §2.6.4.4), Corver (2006b/2017) and Ruys (2008).

The examples in (114) are counterparts of those in (112) and (113), with the weak R-pronoun er. The (a)-examples show that unsplit patterns are consistently marked in such cases. This is particularly true in (114a'), in which the PP is A'-scrambled; this is probably because weak pronouns cannot be contrastively accented. As expected, the (b)-examples show that the split pattern is perfectly acceptable when the stranded preposition is left-adjacent to the clause-final verb(s), but not when it precedes the negation.

Example (115a) shows that pronominalized PP-complements are usually degraded in extraposed position, particularly with the weak R-pronoun er. Therefore, it is not surprising that R-extraction is also excluded in (115b). According to the traditional analysis, in which the main verb selects a complement-PP to its left, this shows that extraposition of pronominalized PPs should be excluded. More recent proposals, in which the main verb selects a PP-complement to its right, must have a mandatory rule that places the PP to the left of the verb (after which R-extraction can apply); for a detailed review of the relevant literature, see Zwart (2011:Part III), although there are alternative proposals, such as in Barbiers (1995a: §4), which are not discussed there.

Note that daarover in (115a) may actually be right-dislocated; if so, it would be a main-clause external phrase, and its marginal acceptability in the indicated position would be irrelevant to the present discussion; this could lead to the conclusion that extraposition of pronominalized PPs is categorically blocked.

Section 36.2.1, sub IIIA, has shown that prepositions that introduce a specific semantic role allow R-pronominalization; the examples given there show that PPs functioning as clausal constituents allow R-extraction as well. Example (116) illustrates this again with instrumental and comitative met-PPs. The primeless examples in (116) show that met-PP with a lexical noun phrase or proper name can precede or follow the direct object of the verb. However, when R-extraction occurs, the stranded preposition must follow the direct object, as in the primed examples.

This restriction on the placement of stranded prepositions has led to the hypothesis that stranded prepositions must be left-adjacent to the verb(s) in clause-final position, although there are also systematic counterexamples: cf. Hoekstra (1979) for an early discussion. We will discuss these in the following subsection.

This restriction on the placement of stranded prepositions has led to the hypothesis that stranded prepositions must be left-adjacent to the verb(s) in clause-final position, although there are also systematic counterexamples: cf. Hoekstra (1979) for an early discussion. One problematic case is illustrated by the examples in (117), which paraphrase the (a)-examples in (116); however, they differ in that the transitive verb openen ‘to open’ is replaced by the periphrastic causative open maken ‘to make open’, in which open functions as an adjectival complementive. Example (117a) shows that the instrumental met-PP must precede the adjective open in conformity with the fact that complementives are usually left-adjacent to the verbs in clause-final position. Therefore, in the (b)-examples, the complementive open and the stranded preposition met compete for the surface position left-adjacent to the clause-final verbs. This results in word-order variation: in (117b) the preverbal position is taken by the complementive, and in (117b') by the stranded preposition mee. The fact that the stranded preposition mee and the clause-final verb maakte can be separated by the complementive open refutes the above-mentioned hypothesis in its strongest form.

That the (b)-examples in (117) are not isolated case is evident from example (118), where similar cases with the prepositional complementive in de muur ‘into the wall’ are found: in (118b) the position left-adjacent to the clause final verbs is taken by the prepositional complementive in de muur and in (118b') by the stranded preposition mee.

However, there are several factors that may force adjectival predicates to precede stranded prepositions; cf. Section A28.2.2, sub III, and Ruys (2005/2008) for more discussion. Consider the examples in (119) with the gradable adjectival complementive bang ‘afraid’, which should be compared with those in (117) and (118). Example (119a) again shows that non-pronominal instrumental met-PPs must precede the adjective. The (b)-examples in (119) demonstrate that the position of the stranded preposition can be affected by the presence of a degree modifier; adjectival complementives without the degree adverb erg ‘very’ can precede or follow the stranded preposition mee, while those with the degree adverb must precede it.

It is more difficult to demonstrate the same for prepositional complementives, given that they usually do not allow degree modification. This is only possible in special cases such as in de war ‘confused’ in (120), which are not locational in nature but denote properties. The (b)-examples show that when the degree modifier erg is present, the pattern is the same as that found in (119): the stranded preposition must follow the complementive, while this is not necessary when the modifier is absent.

In locational prepositional complementives, the modifier is usually a measure phrase such as the adjective diep ‘deep’ in (121). Adding the measure phrase diep to the prepositional complementive in (118a) makes no difference in word order; the judgment on the two word orders in (121a) remains similar. Moreover, as expected on the basis of (119) and in (120), the stranded preposition in the (b)-examples must follow the modified prepositional complementive diep in de muur. This means that we again find the same pattern as with adjectival complementives.

Similar results are expected for nominal complementives, but examples are difficult to construct, making this difficult to establish. Ruys (2008:577) provides an example suggesting that stranded prepositions behave differently in that they must follow the nominal predicate; our own attempt in (122) seems to show the same thing. We will leave this deviant behavior of nominal complementives for future research and refer to the discussion in Ruys (2008) for an initial attempt at an explanation.

There is yet another deviant case, but now in the opposite direction, i.e. one in which the stranded preposition cannot be left-adjacent to the clause-final verb(s). Section 32.3.1.5, sub II, has argued that verbal particles also act as a kind of complementive. This correctly predicts that they can intervene between the stranded preposition and the clause-final verbs (if present). However, unlike the complementive phrases in (117) and (118), they cannot precede stranded prepositions.

The main conclusion from the previous subsection is that predicative complements, which are located left-adjacent to the verb(s) in clause-final position in the unmarked case, compete with stranded prepositions for the same position. Examples (117) and (118) show that there are typically two possible word orders, with either the predicate or the stranded preposition left-adjacent to the verb. However, there are factors (such as the cases of modification discussed here) which may favor one of the two orders. The discussion has further shown that the hypothesis mentioned below (116) that stranded prepositions of PP-complements of verbs must be left-adjacent to the verbs in clause-final position cannot be maintained in full. This is also evident from the fact that there can be more than one stranded preposition in a single clause, as shown in (124a), with two occurrences of R-extraction, one from the instrumental met-PP and one from the prepositional complementive in de muur. This means that the two stranded prepositions compete for the position left-adjacent to the verb(s) in clause-final position, with one inevitable loser: the stranded preposition mee of the instrumental PP cannot follow the stranded preposition of the complementive.

The well-formed order of the stranded preposition generally reflects the unmarked order of the two full prepositional phrases. This is easiest to see when one of the stranded prepositions heads a predicatively used PP, as in (124a), but this also holds for other cases; cf. Ruys (2005/2008). Note that the R-word er in (124b) is interpreted as the pronominal R-word associated wit both mee and in. This conflation of syntactic functions will be discussed in more detail in Section 36.5.3.

The data discussed in this section present a problem for the OV-approaches to Dutch. To get the stranded propositions left-adjacent to the verbs in clause-final position, the direct object or complementive must move to the left across the stranded preposition. However, as Ruys (2008) also points out, these movements do not exhibit the properties established for A- and A'-scrambling in Section V13.2 and V13.3. For example, complementives to the left of stranded prepositions do not seem to need contrastive accent, as is usually the case for phrases undergoing A'-scrambling, nominal objects occurring to the left of stranded prepositions are not necessarily part of the presupposition of the clause. The latter is illustrated by the examples in ( 125 ) with the object pronoun wat: such pronouns never undergo regular A-scrambling but they still must precede stranded preposition.

Things may be somewhat easier for approaches assuming a universal underlying VO-order, as they independently require leftward movement of nominal objects and complementives; cf. Zwart (2011:Part III). This means that only their landing site, to the left or the right of the stranded preposition, must be sensitive to the factors determining surface order. It is important to mention that Ruys (2008) convincingly argues that the constraints regulating the word-order variation described earlier may be of an extra-syntactic nature, which suggests that the ultimate analysis may depend on the interface (or bare output) conditions postulated in the current minimalist program. A more or less ready-made proposal for such an analysis can be found in Broekhuis (2008: §2/§5.3), where the leftward movement of nominal objects and complementives is analyzed as an instance of so-called short A-scrambling (related to valuation of agreement features), and the phonological and semantic interface conditions imposed on the syntactic output are formulated in terms of an optimality-theoretic evaluation; we leave this as a topic for future research.

In support of Ruys’ proposal, we conclude with a brief justification for assuming non-syntactic constraints on the placement of stranded prepositions. Ruys’ main reason is that the distribution of stranded prepositions in the middle field of the clause is quite similar to that of “weak” PPs, i.e. PPs with a weak pronominal complement. This is illustrated with the examples in (126): the (b)-examples show that the stranded preposition mee and the weak PP met ze ‘with them’ differ from the PP met de poppen ‘with the dolls’ in that they can follow the adjective bang afraid.

The (b)-examples show that the weak PP met ze ‘with them’ behaves like the stranded preposition mee in that it can occur between the complementive and the postverbal verb. This is typically mandatory when the complementive contains a modifier (but see the discussion below for an adjustment of this claim). Ruys convincingly argues that this implies that the variation in word order cannot be regulated by appealing to purely syntactic principles, such as the popular proposal from the 1980s that the instrumental PP must be in a position where the trace of the extracted R-word can be formally licensed by the verb under government. Instead, it is proposed that there is a prosodic constraint at play that appeals to the notion of “weak” PP, i.e. PPs with phonetically weak or empty nominal complement (a weak pronoun or trace). There may also be a constraint that appeals to the weight (or complexity) of the complementive phrase. To see this, let us return to the examples in (119) to (121), which show that the stranded preposition of an instrumental met-PP must follow adverbially modified complementive phrases. Our discussion focuses on the examples in (121), repeated as (127), assuming it applies to the other cases as well.

What we did not mention earlier is that there is an alternative placement of the stranded preposition mee, as in (128a), where it occurs between the modifier of the complementive diep and the complementive in de muur itself. The acceptability of (128a) suggests that the stranded preposition is (or can be) part of the complementive phrase, but there is reason to reject such an analysis. First, note that (128b) shows that (gradable) measure phrases can be extracted from prepositional complementives by wh-movement; cf. Section 34.1.2, sub II. This opens the possibility to derive the order in (128a) by A'-scrambling of the measure phrase into a more leftward position in the middle field of the clause (e.g. the contrastive focus position). This leads to parallel analyses for the two examples in (128), sketched in the primed examples.

It is difficult to provide a syntactic explanation for the difference in the judgments on examples (127b) and (128a) in terms of e.g. the aforementioned syntactic relation between the verb and the trace tj of the extracted R-word in the instrumental PP. This is because movement of the modifier of the complementive does not seem to affect this relationship. An account in terms of the weight of the complementive, on the other hand, seems quite natural, as the weight of the remnant of the complementive PPs in (128) is reduced to that of the unmodified complementive in (118).