The R-word er can perform four different functions. It can be used as an expletive in an existential/presentational construction (190a), as the pronominal R-part of a pronominal PP (190b), as a locational pro-form (190c), and as an indicator of the nominal gap in quantitative constructions (190d).

Sometimes the R-word er can be used to express more than one of these functions at the same time, which can be seen in the examples in (191). It is easy to prove that er can simultaneously perform the function of the noun phrase de sigarenkistthe cigar box and the noun sigarencigars in example (191a); in (191b) er must be part of the pronominal PP er ... in and is obligatory; in (191c) we have quantitative er, which is likewise obligatory. Since both pronominal and quantitative er are obligatory, er must perform both functions in (191d).

However, it does not seem to be the case that all functions can be conflated. First, it seems that locational er cannot be conflated with pronominal er; in example (192b), er is simply interpreted as part of the pronominal PP er ... over; pronominalization of the locational PP op het bal in (192a) requires that a strong locational R-word like daarthere be used.

Second, the same thing holds for quantitative and locational er; according to us example (193b), er can only be interpreted as the licenser of the nominal gap [e]; pronominalization of the locational PP in Amsterdam in (193a) requires that a strong locational R-word like daarthere or hierhere be used.

The restrictions may be related to the fact that locational er generally has an adverbial function: because adverbial phrases are usually optional, there is no syntactic reason for the hearer to assume that er expresses a locational meaning in (192b) and (193b), so that the most obvious interpretations of these examples are purely pronominal and purely quantitative, respectively.

In claiming that the locational and quantitative functions of er cannot be conflated in (193b), we go against a long tradition (starting with Bech 1952) which claims otherwise. The examples given in support of the assumption that these two functions can be conflated usually take the form in (194), where the context, provided in (194a), restricts the assertion of (194b) to students who are present in the classroom. The question we have to ask, however, is whether the fact that the students referred to in (194b) are in the classroom is expressed by the element er, or whether it is an inference from the context. To answer this question we can look at example (194c), where er can only be understood as the locational pro-form: this example sounds rather forced with er present (due to its redundancy), which suggests that er does not express the locational meaning in (194b), i.e. that we are dealing with an inference from the context.

The claim that quantitative and locational er cannot be conflated also contradicts the claim in Bennis (1986: 180) that the primeless examples in (195) are fully acceptable: since wonento live and doorbrengento spend time often require the presence of a locational phrase, these examples would provide evidence in favor of the claim that the quantitative and locational functions of er can be conflated. However, we believe that in these cases the presence of a strong locational R-word, as in the primed examples, is preferred; the primeless examples are marginal at best.

We leave it to the reader to decide whether our arguments against the traditional view are conclusive, but in the following we will assume that they are. Thus, leaving aside for the moment the possibility that er performs the same function more than once (which is discussed in Subsection II for cases in which er is associated with two pronominal PPs or two quantitative noun phrases at the same time), we expect the following conflations of functions to be possible: all other combinations are excluded by the two observational generalizations in (192) and (193).

The possibilities in (196a) are of course trivial: no conflation has taken place, as in the examples in (190). In the following two subsections, we will therefore focus on the options in (196b&c).

This subsection shows that the predicted conflations of functions in (196b&c) do indeed occur. The dual functions expletive + pronominal and expletive + quantitative are illustrated in (197b&c), and in (197d) the only possibility of combining three functions is illustrated.

The dual function quantitative + pronominal has already been demonstrated in (191); and we will simply repeat the examples here.

Example (199c) demonstrates the last option in (196b). It is clear from the examples in (199a&b) that we are dealing here with a conflation of the expletive and locational functions of er: (199a) shows that (in the absence of any other locational phrase or qualifying adjectival phrase) locational er is obligatory in this construction, while (199b) shows that an indefinite noun phrase requires the presence of expletive er. Consequently, we cannot but conclude that er performs both functions in (199c).

Section 36.5.1 has shown that the behavior of the weak R-word er is similar to that of weak object pronouns in that it usually cannot occupy the clause-initial position in main clauses. The only exception to this generalization is the expletive er, which behaves like weak subject pronouns in that it can occur in this position. The exceptional status of expletive er raises the question as to whether the placement of expletive er affects the conflation of the functions of er. To find out, we examine the main clauses corresponding to (197) and (199).

The examples in (200) show that expletive er in the regular subject position immediately after the finite verb in second position induces exactly similar judgments as in the embedded clauses in (197).

Example (201a) only illustrates that expletive er can also occupy the first position in the clause. The two (b)-examples bear out that sentence-initial placement of the expletive does not affect the conflation of the expletive and the pronominal functions of er, while the expression of the pronominal function by a separate occurrence of er, as in (201b'), leads to unacceptability.

Things are different, however, in the case of the conflation of expletive and quantitative er. The (b)-examples in (202) show that the conflation of these two functions is not possible: the quantitative function must be expressed by a separate occurrence of er in the regular position of non-expletive er (i.e. after the finite verb), as in (202b').

In dealing with three functions, as in (203), we require an additional occurrence of er. Given the data in (201) and (202), it seems reasonable to assume that the first occurrence of er expresses the expletive and the pronominal functions, while the second one only expresses the quantitative function, but it should be noted that we have no independent evidence to support this claim.

The (b)-examples in (204), finally, provide the main clause counterpart of (199c) and show that the locational pro-form behaves like pronominal er in that conflation with the expletive is also required if er occupies the sentence-initial position.

Example (205a) contains two PPs, both of which allow R-extraction; cf. (205b&c). The (d)-examples show that two stranded prepositions must occur with only a single occurrence of pronominal er, which shows that two occurrences of pronominal er are obligatorily conflated. Similar examples can be found in (206).

However, constructions in which er is understood as associated to two stranded adpositions are not always available. For instance, examples like (207d) and (208d) seem ungrammatical, despite the fact that the (b) and (c)-examples show that the two PPs both allow R-extraction by themselves.

The difference between (205) and (206), on the one hand, and (207) and (208), on the other, has not been discussed in the literature, so that we can only speculate about what causes the difference in the judgments on the two sets of examples. The most striking difference is that the former set involves complementive PPs (viz. naar de bibliotheek toe and uit het slot), while the latter two examples do not. Our conjecture, therefore, is that this type of conflation is only possible when one of the two pronominal PPs is a predicatively used locational/directional phrase. Future research will have to show whether this conjecture is correct.

As R-words other than er can also strand prepositions, the question arises as to whether the possibility of conflation is a typical property of the weak R-word er or a more general property of R-words. The data in (209) suggest the former; note that we cannot give examples with relative pronouns, as their reference is determined by their (unique) antecedent.

In order to have two stranded prepositions, there must generally be two R-words, as in (210). The examples are perhaps difficult to process, but it seems that the first R-word in each of the examples in (210) must be interpreted as the pronominalized object of the locational preposition uitout of; the reading in which the first R-word is construed as the pronominal object of the stranded preposition meewith seems to be excluded. Somewhat surprisingly, this is also true for the interrogative R-word in (210d), in which waar (and not er) is construed as the pronominal object of uit. This means that these examples all have a similar structure featuring a nested dependency: ... R-wordi ... R-wordj .. Pj .. Pi, where the indices indicate which R-word is construed with which adposition.

To conclude this subsection, we will discuss another case, taken from Haeseryn et al. (1997:488). As in (205) and (206), the examples in (211) involve a complementive PP, viz. in de krant, so that they come under the earlier hypothesis that one of the two PPs involved must be a complementive in order for them to be licensed by a single occurrence of er. What is special about this example, however, is that the over-PP seems to function as a modifier of a subject noun phrase. Since the subject noun phrase is indefinite, er in (211a) is an expletive. In (211b&c), er functions simultaneously as an expletive and as a pronominal R-word. In (211d), er is construed with two stranded prepositions. Example (211e) illustrates how complex such constructions can become: in addition to the three functions it already has in (211d), er also functions as quantitative er in this example.

Example (212a) shows that it is also possible to conflate two instances of quantitative er. That we are dealing with conflation here is clear from the examples in (212b&c): when the direct object is a full noun phrase, as in (212b), er is obligatorily present to indicate the nominal gap in the subject; when the subject is a full noun phrase, er is present to indicate the nominal gap in the direct object: consequently, er in (212a) must perform both functions simultaneously. Note in passing that a subject with a nominal gap may precede quantitative er; cf. Section N19.3 for further discussion.