The unacceptable examples in (142a-c) and (143a-c) become acceptable when an associate for the quantifier is added in the form of a noun phrase or a pronoun.

As with the other floating quantifiers, allemaal must follow its associate, unless it is placed in sentence-initial position and its associate is a weak pronoun; the examples in (148) and (149) are all acceptable with the weak pronoun zethey but degraded with the noun phrase die mensen/boekenthese people/books.

That we are dealing with floating quantifiers in (148) and (149) is clear from the fact that the quantifier and its associate are not adjacent. Actually, there is not much chance of mistaking the floating quantifier allemaal for a modifier: this use is not possible for allemaal if the noun phrase contains a determiner (cf. Table 9). This is demonstrated again by the examples in (150): since the noun phrase and the quantifier cannot occupy the clause-initial position together, we can safely conclude that they do not form a constituent. Incidentally, this again suggests that the term quantifier float should not be taken too literally; cf. Section 21.1.4, sub III, for discussion.

However, recall from the discussion of the examples in (134) that the data for strong pronouns is not so clear. Although marked, example (151a) suggests that it is at least marginally possible for allemaal to function as a modifier of a strong pronoun: allemaal and the pronoun can be placed together in the clause-initial position (the constituency test). The examples in (151b&c) illustrate the same point as (151a), given that floating quantifiers cannot normally be scrambled across the adverb gisteren; cf. the discussion of the examples in (115) and (116).

The quantifier allemaal is not able to form a constituent with a weak pronoun; as in the case of alle in Section 21.1.2.2, sub IIB4, this requires the pronoun to be strong.

As in the case of all other al-quantifiers, the associate of the floating quantifier allemaal can be independently quantified, although some people may object to the primed examples on pragmatic grounds.

In this case, this fact cannot be used to query the “quantifier-float” approach, because the examples in (155) are unacceptable for independent reasons; cf. the discussion of (150). Note in passing that our judgment of (155b) differs from that in Coppen (1991:133), where it is assigned a mere question mark; our informants, however, categorically rejected the two examples in (155).

As shown in (156), the floating quantifier allemaal seems freer than the other types of al-quantifiers in that it can take the neuter pronoun hetit, the singular demonstratives dit/datthis/that, and the interrogative pronoun watwhat as its associates.

Unlike the personal pronouns in (151), these pronouns certainly cannot form a constituent with the modifier allemaal: *Het allemaal is verkocht; *Wat allemaal heb je gelezen? At first glance, this seems to be another argument against the “quantifier-float” approach, although it must be pointed out that there are reasons to assume that the function of allemaal in these examples is different from its function in the examples discussed earlier. First, it seems that allemaal can be used with an adverbial function: example (157a) seems to favor a reading in which het/dit refers to a set of entities that have all been read, but it can also refer to a single entity that has been read from cover to cover, a reading that is the only one possible in example (157b) with the adverbial helemaal.

Second, the semantic contribution made by allemaal in the wh-constructions in (156b) differs from ordinary universal quantification, and it can be replaced with preservation of meaning by zoal(what) for instance/among other things (lit.: so-all). This would suggest that we can classify the use of allemaal in (156) as adverbial. If so, allemaal is the only al form that can be construed both as an adverb and as a floating quantifier.

The (d)-examples in (142) and (143), repeated here as (158a&b), show that allemaal cannot easily be used as a complement of a PP: it is marginally possible when it refers to a human discourse topic, and impossible when the discourse topic is inanimate (although some speakers are more lenient in this respect). Example (158a') further shows that adding a pronominal associate to allemaal in (158a) leads to a perfectly acceptable result. Adding a pronominal associate to allemaal in (158b) does not improve the result, which is to be expected because strong pronouns cannot usually be used to refer to inanimate entities. Note that using a weak pronoun would make the result ungrammatical for the independent reason that weak pronouns cannot form a constituent with the quantifier; cf. the discussion of (152).

The primed examples in (158) thus show that [+human] and [-human] pronominal associates of allemaal differ in that only the former can form a constituent with allemaal within the PP. A similar conclusion was drawn for al in (101b&b'), which was supported by the fact that scrambling and topicalization of the PP requires pied piping of the quantifier. Example (158a') requires the same, as shown in (159).

However, the situation is even more complex: in (160a) the scrambled PP is not adjacent to the quantifier, and in (160b) the PP can be topicalized without the quantifier allemaal, provided that the latter occupies the same position as in (160a).

The cases in (159) and (160) thus seem to show that allemaal can be generated either as part of the PP or independently to the left of the base position of the PP containing its associate; in the latter case, the PP must be moved to some position to the left of the quantifier. If so, we expect R-pronominalization to yield an acceptable result also with a split PP: the R-pronoun (i.e. the actual associate of allemaal) must be moved to a position to the left of the independent quantifier allemaal, while the preposition can be stranded to the right of this quantifier; the split PP is indicated by italics.

The results in (161) seem indeed acceptable; the fact that they sound a bit marked may be related to the fact that we are dealing with indirect-object PPs; the examples in (162) involving PP-complements are impeccable.

The above examples all involve a pronominal associate of allemaal. The judgments for full noun phrases such as zijn vriendenhis friends seem to be different from what we found in (159) and (160) for the strong pronoun; the primed examples in (163) show that neither scrambling nor topicalization of the PP can pied-pipe the quantifier. This suggests that allemaal can only be construed as part of the PP when the latter takes a pronominal complement.

In examples such as (164a), taken from Perridon (1997:185), it is not immediately clear whether allemaal is a floating quantifier belonging to the subject wat je over mij vertelt or a modifier belonging to the predicate nominal leugens.

Semantically, (164) is not on a par with (165a), where the quantifier clearly modifies the noun phrase and allemaal is be rendered as “all sorts of”. Instead, it patterns with (165b), where allemaal has a very high-degree reading; it is nothing but lies that you have told about me.

However, as discussed in Section 21.1.5.1, sub IIC, it is not clear that allemaal as a quantifier of very high degree is part of the noun phrase. Evidence that allemaal in (164) and (165b) can be construed as a floating quantifier with the pronominal subject as its associate is given in (166), where the nominal predicate is replaced by an adjectival one, which of course cannot be modified by allemaal; at least in these examples allemaal must be construed as a floating quantifier.