The question as to whether postpositional phrases like de boom ininto the tree in (68a) can undergo R-pronominalization is not easy to answer. Let us first recall some basic facts. The defining property of postpositional phrases is that the adposition follows its nominal complement and that they always have a directional meaning (i.e. express the notion of a path). Furthermore, postpositions differ from prepositions in that they do not have to be adjacent to their nominal complement, which can be shifted to the left across clausal constituents, such as the adverbial phrase gisteren in (68b). Since the nominal complement can be extracted from the postpositional phrase, we correctly predict that moving the relative pronoun diewhich into the clause-initial position in (68c) also leads to a perfectly acceptable result.

The prepositional counterpart of the postposition in in (69) does not have a directional meaning, but a locational one: in (69a) the PP in de boom expresses that Marie has undergone a change of location. We have further seen that leftward movement of the nominal complement of a preposition is not possible, i.e. there is no preposition stranding in Dutch. This means that the structures in (69b&c) with a change-of-location meaning are ungrammatical; this is indicated here by the use of an asterisk.

The preposition in does not take a [-human] pronominal complement, but obligatorily undergoes R-pronominalization, as shown in (70a). Like the nominal complement of the postposition in in (68), the R-pronoun can be shifted to the left across clausal constituents, and it is therefore not surprising that we can also have the relative construction in (70c).

The discussion so far has tacitly assumed that the examples in (70b&c) have only a prepositional source. We have given no evidence for this so far, but there is some a priori reason to adopt such a hypothesis. The fact that the nominal complement de boom in (68b) can be easily pronominalized by the (weak) object pronoun ʼmhim shows that R-pronominalization is unnecessary for postpositional phrases (and hence blocked); the difference between postpositions and prepositions would then be visible not only in their position relative to their non-pronominal complement, but also in the form of their pronominal complement. A promising syntactic test that might help to determine whether the pronominal PP in examples like (70b&c) is prepositional or postpositional in nature can be based on the examples in (71); they show that postpositions are like verbal particles in that they can permeate clause-final verb clusters, whereas (stranded) prepositions cannot.

Unfortunately, the permeation test is problematic for several reasons. First, permeation of the verbal cluster by a stranded preposition is perfectly acceptable in certain southern varieties of Dutch; cf. Sections V4.3 and V4.5. This means that the test will only yield reliable results if we restrict ourselves to speakers of the northern part of the Netherlands. A second problem is that even the speakers passing the judgments in (71a') are often not very sure about them in the relevant relative constructions, which may also vary from case to case. Consider the examples in (72), which, according to the permeation test, involve postpositions.

As expected, most of the northern speakers we consulted strongly prefer the use of the regular relative pronoun diewhich in (72a&b); the use of the relative R-pronoun waar is usually considered to be marked. Surprisingly, however, die and waar in (72c) are generally judged to be equally acceptable; some speakers even prefer the use of the relative R-pronoun waar. If the permeation test is trustworthy, the contrast between (72a) and (72b) should lead to the conclusion that postpositional phrases with in can sometimes, but not always, undergo R-pronominalization; the fact that we are dealing with the same form in, which precludes an appeal to the lexical properties of the postposition, leaves this remarkable fact a mystery.

We could solve this problem by simply rejecting the permeation test. This step might be justified by the fact that the examples in (73) show that the judgments on the use of the relative R-pronoun waar in (72) are more or less identical to those on the use of prepositional phrases in the examples in (73); this suggests that the primed examples in (72) are, after all, pronominalized prepositional phrases.

Maintaining the generalization that postpositional phrases cannot undergo R-pronominalization would also allow us to account for the minimal pair in (74) with the complex postposition achternaafter and the circumposition achter ... aanafter. The primed examples show that the nominal complement of the postposition achterna can only be pronominalized by a [-R] pronoun, whereas in the case of the circumposition achter ... aan it must be pronominalized by a [+R] pronoun.

The discussion so far seems to show that the permeation test is not completely reliable and that, despite the fact that the stranded adpositions permeate the clause-final verb cluster, we are dealing with pronominalized prepositional phrases in the examples of (72). There are, however, other problems for the claim that postpositional phrases do not allow R-pronominalization. First, consider the resultative construction in (75). The two (a)-examples show two things: (i) the postpositional phrase de haven ininto the harbor cannot be replaced by the prepositional phrase in de haven; (ii) pronominalization of the nominal complement of the postposition, de haven, by the weak pronoun ʼmhim does not seem to be easily possible. However, the judgments on the relative construction in (75b) show that the relative pronoun can at least marginally take the form of an R-word: the use of waar gives rise to a marked but reasonably acceptable result, whereas the use of the regular pronoun die is (surprisingly) rejected by most speakers.

A further potential problem is illustrated by example (76b). Since the adposition af cannot be used as a preposition, this example may involve a pronominalized postpositional phrase. In this case, however, there is also an alternative analysis, according to which the R-word waar is not a pronoun corresponding to the noun phrase de berg in (76a) but a pro-form of the adpositional phrase van de berg. This means that we can probably ignore this case.

It will be clear from the discussion above that we are not yet in a position to give a definitive answer to the question as to whether or not postpositional phrases can undergo R-pronominalization. If it is truly impossible for stranded prepositions to permeate verb clusters, we must conclude that R-pronominalization of postpositional phrases is possible (which, incidentally, leaves us without an account of the relative grammaticality judgments in (72)). If, however, it turns out that stranded prepositions can sometimes permeate the verb cluster, the answer depends on the grammaticality status of (75b). If this example with the R-word waar is declared grammatical, then R-pronominalization of postpositional phrases should be considered possible, but if it is declared ungrammatical, then we can claim R-pronominalization of postpositional phrases to be impossible (and thus account for the grammaticality judgments on the examples in (72) by referring to the similar judgments on the examples in (73)). Since we are not able to shed more light on this issue at this point, we have to leave it to future research and simply conclude that, despite the tricky cases discussed above, postpositional phrases usually do not allow R-pronominalization.