The use of a pronominal PP often depends on whether the complement of the preposition is [+human] or [-human]. This is especially clear in the case of referential personal pronouns. Resumptive pronoun constructions of the type in (13) make it possible to test the restrictions on the co-occurrence of certain types of antecedents and the R-pronoun; cf. the discussion of (12).

Note that the resumptive pronoun er must appear in its phonetically strong form daar in clause-initial position, as in (14).

This may be due to the fact that (with the exception of subject pronouns and the expletive er) phonetically weak elements cannot be placed in clause-initial position; the difference between (13b) and (14) is therefore similar in nature to the difference between the two examples in (15). We refer the reader to Section N18.2.1.1, sub VB, for a more detailed discussion.

We discuss the behavior of [-human] and [+human] referential pronouns with respect to the resumptive pronoun test in separate subsections.

The demonstrative R-pronouns daarthere and hierhere from Table 1 are also preferably interpreted as referring to [-human] antecedents. Examples such as (27a) are perfectly natural if the speaker refers to an object, but are decidedly odd when used to refer to a person. Example (27b) again shows that pronominal PPs are acceptable in generic constructions with [+human] bare plural antecedents.

Although demonstrative pronouns are perfectly possible in the subject or the object position of the clause as an independent argument of the verb (i.e. without an accompanying noun), they are marked when used as an independent argument in the complement position of a preposition.

The examples in (29) show that relative pronouns with a [+human] antecedent can be realized both as the regular relative personal pronoun wiewho and as the relative R-pronoun waar. Although normative grammars are often opposed to (29b), it is this form that is usually found in colloquial speech; cf. Section N16.3.2.2, sub III, for discussion. The contrast between the (a) and (b)-examples shows that the preposition must be pied-piped by the [-R] pronoun, whereas preposition stranding is possible and even slightly better in the case of a [+R] pronoun. This again confirms our previous finding that preposition stranding is only possible with R-pronouns; cf. example (3).

The examples in (30) show that the formation of a pronominal PP is obligatory if the antecedent of the relative pronoun is [-human]. This implies that the relative pronouns die and dat cannot occur as the complement of a preposition.

Interrogative pronominal PPs can only be used if the preposition has a [-human] complement. A speaker who knows that Jan is waiting for a person will not use the construction in (31b); this construction can only be used if the speaker expects the answer to involve a [-human] entity, or if he has no expectation at all. The primed examples show that preposition stranding is excluded with the [-R] pronoun wiewho, but acceptable and even preferred with the corresponding [+R] pronoun waar. Again, we can conclude that preposition stranding is only possible with R-pronouns. The number sign in (31a') indicates that this sentence is possible when read as including the particle verb opwachtento await/waylay, which is not relevant here.

The examples in (32) show that the formation of a pronominal PP is strongly preferred when the speaker expects the answer to involve a [-human] entity; examples such as (32a) are only acceptable as echo-questions.

Existentially quantified R-pronouns also refer strictly to [-human] entities. A speaker who uses (33b) is expressing that the thing he is waiting for is not a [+human] entity. Something similar holds for the negative counterpart of this R-pronoun in (33b').

It seems that if the existentially quantified pronoun refers to a [-human] entity, the formation of the pronominal PP is more or less optional. This is shown in (34).

If the complement of the preposition expresses sentence negation, it must be moved to a certain position in the middle field of the clause; cf. Section V13.3.2. This is clear from the fact that the PP-complement of the adjective in (35a) cannot occupy its regular postadjectival position but must precede the adjective; cf. Sections V13.3.1 and A24.3.1, sub IIB, for further discussion. The contrast between (35a') and (35b') shows that preposition stranding is possible (and obligatory) only with the R-pronoun; the pronoun niemandnobody also moves, but must pied-pipe the preposition. This again confirms our earlier claim that preposition stranding is restricted to R-pronouns.

In the case of universally quantified pronouns, pronominal PPs also strictly refer to [-human] entities. A speaker who uses (36b) is expressing that the things the doctor will look at are not [+human] entities. So, (36a) can be used to express that the doctor will examine all patients, while (36b) expresses that the doctor will examine the patient(s) thoroughly.

As in the case of the existentially quantified pronouns, the formation of the pronominal PP is more or less optional if the universally quantified pronoun refers to a [-human] entity, as shown in (37).

The previous subsections have shown that [+human] pronouns usually do not easily allow R-pronominalization, although two exceptions have been attested. First, although non-neuter referential personal pronouns only marginally allow it, R-pronominalization of neuter pronouns is easily possible and even obligatory if the neuter pronoun is singular. Second, R-pronominalization of [+human] relative pronouns seems to be the preferred option in colloquial speech. R-pronominalization of [-human] pronouns, on the other hand, is usually obligatory; only the (existentially and universally) quantified pronouns behave differently in this respect. The discussion from the previous subsections is summarized in Table 2, where P stands for the preposition in question.