Consider the examples in (142). Since (142a&b) are impersonal passives, the first occurrence must be the expletive er; the same thing holds for (142c) since the associate noun phrase [twee [e]] of quantitative er is an indefinite noun phrase. These examples show that the expletive er cannot co-occur with the other uses of er.

The pattern in (142) has given rise to the idea that er must be placed in a unique, designated [+R]-position in the middle field of the clause, i.e. the [+R]-position in (139). The assumption that this position is unique accounts for the fact that only one occurrence of er is allowed; [+R] can be occupied by only one R-word, so that the other occurrence(s) would violate Axiom I; cf. Van Riemsdijk (1978:ch.5).

Axiom I also gives the correct predictions for the examples in (144): the examples in (144a&b) show that pronominal er cannot co-occur with locational or quantitative er, and (144c) shows that the co-occurrence of quantitative and locational er is also correctly ruled out. Since the grammaticality judgments on the examples in (144) remain the same when we reverse the interpretations (e.g. when we interpret the first occurrence of er in (144c) as locational and the second as quantitative), we have discussed all possible combinations.

We conclude with a brief remark on axiom I. This axiom is overruled for occurrences of er that must be adjacent to the preposition, such as er in the temporal phrase er voor/nabefore/after it; cf. (145). This follows from the fact, discussed in Section 36.3.1, that these temporal PPs do not allow R-extraction; since er remains PP-internal, it is correctly predicted that the clause can contain two occurrences of er in such cases.

The pattern of grammaticality judgments is slightly different for main clauses. If the expletive er occupies the regular subject position after the finite verb, as in (146), we find a pattern that is essentially identical to the one in (142).

However, if the expletive er occupies the clause-initial position, as in (147), it can co-occur with quantitative er, but not with pronominal or locational er. While the pattern in (146) follows directly from axiom I, the pattern in (147) is somewhat puzzling; we will not provide an explanation for this anomalous pattern here; cf. Odijk (1993: §6.5/1994) for a proposal.

Similar problems do not seem to arise with the other types of er, since they never occur in sentence-initial position; hence, they invariably exhibit the pattern found in the main clauses in (144). This conclusion may not be uncontroversial, however. Neeleman & Van de Koot (2017:4386) appear to disagree with the claim that locational and prepositional er cannot be placed in sentence-initial position: basing themselves on the sentences in (148), they assume that the first occurrence of er is prepositional in (148a) and locational in (148b). The implication would be that these R-words can also co-occur with quantitative er. We have reasons to reject this: in Section 36.5.3, we provide compelling reasons for assuming that the first occurrence of er is expletive, while the second occurrence of er has both a quantitative and a prepositional function in (148a) and (perhaps) a quantitative and locational function in (148b). This analysis of the examples in (148), which is the standard one in the literature and can also be found in Haeseryn et al. (1997), is fully consistent with the claim that sentence-initial er must be expletive.

There is a second difference between the description given so far and the one in Neeleman & Van de Koot (2017: §2.2): they claim that the co-occurrence restriction illustrated in (144) can always be overridden if (i) the two occurrences of er are not adjacent and (ii) one of the occurrences of er is quantitative. Two examples purporting to show this are given as the primed examples in (149). However, the primed examples with a single occurrence of er expressing both functions are much better than the primeless examples (as well as the other examples that violate the co-occurrence restriction presented in that study).

Although we agree that the primeless examples in (149) sound a bit better than those in (142)/(144), we leave it to future research to investigate whether they should be considered part of the colloquial languages (e.g. because they occur in informal speech).

We conclude from the discussion in the previous subsections that Axiom I provides a largely adequate description of the attested facts; leaving aside the data dispute that might arise from some of the judgments presented in Neeleman & Van de Koot (2017: §2.2), the only problem seems to be the grammaticality judgment on example (147c). It should be noted, however, that the unacceptability of the examples in (142), (144), (146), and (147a&b) does not necessarily mean that the intended meanings cannot be expressed. Some of them can, due to the fact that er is able to perform more than one function at the same time. This is discussed in detail in Section 36.5.3.

Wh-movement of an R-word is sometimes blocked by the presence of another R-word. Consider example (160a), which is ungrammatical under the interpretation that waar is part of the pronominal PP. This can be accounted for by assuming that the wh-word waar cannot be moved into the clause-initial position in one fell swoop, but must be moved via one of the two [+R]-positions in (159); since we have seen that the weak locational R-word er is moved via the strong [+R]-position into the weak one, neither of these positions is available for the wh-word, so movement of waar into the clause-initial position is blocked. When the locational R-word is strong, as in (160b), it does not move into the weak [+R]-position; if we assume that this weak position is accessible to the wh-phrase, then wh-movement of waar is predicted to be possible.

The examples in (161) show that the same contrast can be found in the case of topicalization of demonstrative R-words.

The judgments on the examples in (160) change under the reverse interpretation, i.e. with the preposed interrogative R-word as locational and the second R-word as pronominal. In (162a), pronominal er can be placed in the weak [+R]-position, and the locational R-word waar can be moved via the strong [+R]-position into the clause-initial position. In (162b), however, the locational R-word must be moved via the strong R-position into clause-initial-position, so there is no landing position for the demonstrative R-word hier (the empty weak [+R]-position is not accessible since hier is not a weak R-word).

The examples in (163) show that the same thing seems to hold for demonstrative pronominal R-words.

The judgments on the examples in (162b) and (163b), which are those of Huybregts (1991), are disputed by Hans Bennis (p.c.), who considers (162b) and (163b) fully acceptable and suggests that the relevant reading can be forced in relative clauses such as (164). We agree that a locational interpretation of the relative pro-form waar is more or less acceptable, but its pronominal reading is still possible and seems in fact seems to us to be the more prominent one, which would be in line with Huybregts’ judgments on (162b) and (163b).

The proposed analysis correctly predicts that expletive and quantitative er never block wh-movement of a locational or pronominal R-word, as the latter can be moved into the clause-initial position via the strong [+R]-position. The examples in (165) and (166), which are the wh-movement counterparts of (156) and (157), show that this prediction comes true. Note in passing that, like (156b) and (157b), (165b) and (166b) also allow a reading in which waarwhere acts as a locational pro-form; the weak R-word er then simultaneously performs the functions of expletive and pronominal R-word; cf. Section 36.5.3, sub I.

However, topicalization seems to differ from wh-movement; examples such as (167) with topicalized demonstrative daar differ sharply from the examples in (165) with interrogative waar; the use of the percentage mark indicates that some speakers seem to accept such cases and that some cases can be found on the internet.

The markedness of the examples in (167) is probably not because the expletive blocks topicalization of the R-word daar, but rather due to the fact that the presence of the expletive er depends not only on whether a (definite) subject is present, but also on whether the clause contains presuppositional material; cf. Section N21.1.2, Bennis (1986: §3) and Koeneman (2000: §4) for more discussions of this restriction on the expletive er. For instance, example (168a), in which both the subject and the direct object are (non-specific) indefinite, must contain the expletive, whereas (168b), which contains a definite direct object, is at best marginal if the expletive er is present; if the direct object is a pronoun, as in (168c), er must definitely be absent. The contrast between (165) and (167) can therefore be attributed to the fact that interrogative elements like waar are by definition indefinite, while demonstrative elements like daar are definite.

Replacement of waar by daar does not affect the grammaticality judgments when we are dealing with quantitative er, as shown in the examples in (169); the judgments are similar to those in (166).

That the judgments on (166) and (169) are similar is consistent with the fact, illustrated in (170), that quantitative er must be realized, regardless of the presence of presuppositional material.

A problem for the hypothesis developed above are multiple wh-questions like those in (171), where the first [+wh] R-word must be interpreted as locational and the second as pronominal. If, as has been assumed so far, the locational wh-phrase must be wh-moved via the strong R-position, the landing site of the pronominal wh-phrase should be occupied by a trace, so that we wrongly predict the structure in (171a) to be ungrammatical. Structure (171b), on the other hand, is then wrongly predicted to be possible, since the pronominal R-word can in principle be moved via the weak R-position. We leave this problem to future research.

Note that (171b) improves significantly when an emphatic accent is assigned to the second occurrence of waar. This suggests that focused locational pro-forms are not moved into the strong [+R]-position but into a designated focus position when they are contrastively focused; cf. Section V13.3.2.

Wh-movement of [-R] phrases are also sometimes blocked by R-words. Subsections 1-3 will show that this is especially true for predicative locational phrases: the blocking effect can be easily established for pronominal R-words (Subsection 1), but not for locational R-words, because locational phrases block the movements anyway (Subsection 2); expletive and quantitative er do not induce this blocking effect (Subsection 3). Finally, Subsection 4 will show that non-predicative phrases can freely cross all types of R-words.

So far, we have only discussed the co-occurrence of two R-words with different functions. However, it is also possible to have more than one pronominal PP. Consider the examples in (178), where (178a) is the reference sentence. The examples in (178b&c) first show that the PP voor dat boek and the circumpositional phrase naar de bibliotheek toe both allow pronominalization.

In addition, example (179a) shows that it is not possible to have two occurrences of er, which follows from the claim expressed by axiom I in (159a) that there is only one weak [+R]-position available. However, if the second occurrence of er is replaced by a strong demonstrative form, as in (179b), the result is perfectly acceptable; this would follow from Axiom II in (159b), according to which demonstrative R-words can be placed in the strong [+R]-position. In fact, (179b) seems to be ambiguous in that the pronoun er can be interpreted either as the complement of the preposition voor or as the complement or the circumposition naar ... toe; judgments are subtle, however, and speakers of Dutch may prefer one of the two readings, possibly also depending on the intonation pattern of the example.

It is important to note in passing that (179a) becomes perfectly acceptable on the intended reading if one of the two occurrences is omitted: Jan is er gisteren voor naar toe gegaanJan went there (i.e. to it) for it yesterday. This is due to the fact that one occurrence of er can perform more than one function, i.e. as the complement of both voor and naar ... toe; cf. Section 36.5.3 for discussion.

Given the assumptions made so far, we predict that wh-movement is possible in the case of two pronominal PPs and that the resulting example in (180a) will be ambiguous, in the way indicated in the two (b)-examples. Although it is not easy to judge from the data, it seems to us that this prediction is indeed correct.

That example (180a) is indeed ambiguous is further supported by the fact that the two relative constructions in (181) are both acceptable.

The judgments on (180a) do not seem to change if we replace er by a strong form, as in (182a). The resulting example remains ambiguous, as shown again in the two (b)-examples.

Note that the ambiguity in (182) is not found if we replace the strong R-word daar with the strong R-word hierhere, as in (183a); the reading expressed by the structure in (183b') does not seem to be readily available.

Replacing the verb gaan with komen, as in (184a), yields an unambiguous result again, but now the structure in (184b) is the marked one.

The difference is due to the fact that the verbs komento come and gaanto go differ in that only the former licenses a reading in which the locational pronominal PP hier ... naar toe refers to the deictic center (i.e. the “here”) of the speaker: Jan komt/#gaat hier naar toe Jan comes/goes here (with the number sign indicating that the structure is perfectly acceptable when hier naar toe refers to some other location). This account for the fact that the choice of verb tends to disambiguate example (183/184a), and thus provides additional evidence for the claim that the ambiguity reported for (180a) and (182a) is real.

Example (185a) shows that clauses can contain two locational phrases and (185b) shows that the two locational phrases can both be replaced by a locational pro-form. Since the two pro-forms in (185b) seem to require emphatic accent, it is not surprising that the pro-forms must both be strong; (185c) shows that using a weak pro-form gives rise to a severely degraded result.

The unacceptability of (185c) is expected, given our earlier conclusion that the weak locational pro-form er is moved into the weak [+R]-position via the strong [+R]-position, so that the latter is no longer accessible to the demonstrative locational pro-form. The acceptability of (185b) is surprising though, since we have seen that only one strong [+R]-position is available. Perhaps this shows that emphatic focus exempts the strong locational pro-form from moving to the strong [+R]-position, which would be consistent with the fact that the placement of the second strong R-word in front of the adverb altijd gives rise to a marked result: ??Jan slaapt hier daar altijd. See also the discussion of (171b) in Subsection IVA.

In view of the unacceptability of (185c) it is not surprising that (186a) is also unacceptable: wh-movement must proceed through the [+R]-positions, but these are occupied by the weak locational pro-form er and its trace. Although a bit marked, example (186b) seems acceptable, which would follow if hier occupies the strong [+R]-position and waar is moved into sentence-initial position via the weak [+R]-position.

Quantitative er can be used to license an interpretive gap in both the direct object and the indirect object of the clause, as shown in (187). Since quantitative er must be moved into the weak [+R] position, we expect it to precede both objects, regardless of the position of the interpretive gap e.

At first sight, the acceptability of the examples in (187b&c) might lead us to expect that the clause can contain more than one occurrence of quantitative er. However, this would be incompatible with axiom I in (159a), according to which the clause can contain only one weak [+R]-position, so that at most one occurrence of the weak pronoun er can be licensed: so, example (188) is predicted to be unacceptable, regardless the position of the second occurrence of er.

Although we believe that this prediction is indeed correct, we have used the percentage sign to indicate that cases such as (188), in which the two occurrences of er separated by phonetic material, are given as fully acceptable in Neeleman & Van de Koot (2017: §2.2). We believe that this is not justified: the intended reading can only be expressed by using a singular occurrence of er as a licenser of both gaps, as in the fully acceptable construction Jan heeft er [NP twee e] [NP drie e] gegeven; cf. Section 36.5.3. We leave it to future research to show if our rejection of the structure marked with % is correct or not.