An important property of the constructions in (513a&b) and (514a) is that they are impersonal, in the sense that the subject pronoun hetit is non-referential. That this is the case can be gathered from the fact that this pronoun cannot be replaced by any referential element (while retaining the intended meaning). This is illustrated in (515) for the demonstrative pronouns ditthis and datthat. The number sign indicates that (515a) is possible if the adjective gezelligcozy is predicated of “having/doing this in the room is cozy”, but cannot be used to say anything about the coziness of the room.

At first glance, the claim that the subject pronoun het is non-referential is problematic for the copular constructions in (513a), because the adjectival complementives warmwarm and gezelligcozy must be predicated of some entity. This problem could perhaps be solved for the adjective warm, in that it takes a quasi-referential subject by positing that it resembles weather verbs such as vriezento freeze, but this seems less likely for adjectives such as gezelligcozy. To solve this problem, it has been proposed that the pronoun het actually functions as an anticipatory pronoun, coindexed with the locational PP, which acts as the logical subject of the adjective. If we extend this proposal to impersonal constructions like (513b) and (514), we arrive at the representations in (516).

These representations not only resolve the question as to what the adjective/verbs in (513a&b) and (514a) are predicated of, but may also make intuitive sense in light of the fact that the nominal parts of the locational PPs appear as the subject of the alternate constructions in the primed examples in (513) and in (514b). However, let us not jump to conclusions, because the two alternants are not semantically equivalent. This is clear from the examples in (517), taken from Janssen (1976:69): while (517a) unambiguously refers to the space inside the car, (517a') can also be used to refer to the car itself (its engine may need fine-tuning, for example); similarly, while the PP in (517b) can refer to a get-together organized by the Janssen family, the subject in (517b') must refer to the people themselves.

The claim that the locational PPs in (516) function as logical subjects of the clauses not only provides an answer to the question pertaining to the semantic properties discussed in the previous subsection, but is also supported by their syntactic behavior. Let us first rule out two possible alternative analyses. First, the locational PP in de tuin in (514a) cannot be analyzed as a complementive, because (518a) shows that it can be placed after the clause-final verb, and (518b) shows that it need not be left-adjacent to the clause-final verbs but can easily be separated from them by other phrases (here vaak) in the middle field of the clause.

The examples in (519) further show that the locational PP differs from unsuspected PP-complementives in that it does not allow R-pronominalization; the PP can only be pronominalized using locational proforms like hierhere and daarthere.

A second possibility would be that the locational PP functions as an adverbial phrase. However, this seems to contradict the fact that it cannot be omitted; example (520a) cannot be interpreted as an impersonal construction, but is only acceptable if the neuter pronoun het is referential, i.e. functions as the pronominalized counterpart of an example such as (520b).

The conjecture in (516) that the locational PP functions semantically as the logical subject of the clause is compatible with these facts. A possible problem for this conjecture is that the (a)-examples in (521) show that the PP cannot be placed in the regular subject position of the clause; it can only be placed in the clause-initial position if it is topicalized, in which case the non-referential pronoun het must appear in the subject position right-adjacent to the finite verb in second position. However, this is compatible with the proposed analysis if we assume that the regular subject position can only be occupied by a noun phrase and that this is precisely the reason why the anticipatory pronoun is used in this construction.

In fact, this also explains why het is not needed in the alternants of (521) in (522); since the reference objects in these constructions are syntactically realized as noun phrases, they can of course be placed in regular subject position, so that the insertion of the anticipatory pronoun het is unnecessary (hence blocked).

The syntactic status of the van-PP is not immediately clear. A first observation is that this PP seems to prefer a position after the clause-final verb, which immediately rules out an analysis according to which the PP functions as a complementive.

The examples in (524) show that R-pronominalization of the van-PP is possible; this favors an analysis according to which the PP functions as a complement of the verb; however, it is not conclusive because certain adverbial phrases also allow R-pronominalization.

Another argument for assuming that the van-PP is a PP-complement and not an adverbial phrase is that omitting this PP leads to a severely degraded result: such an effect is expected for PP-complements, but not for adverbial phrases.

Note that the examples in (525) are semantically incoherent; the verb is taken in its literal sense as a verb denoting undirected motion, while the (logical) subject does not seem to be able to satisfy the selection restrictions imposed by this verb. Adding the van-PP apparently lifts the selection restriction imposed by the verb on its subject. This is consistent with the proposal in Vandeweghe (2020: §3.1.3) that the nominal complement is a kind of demoted subject of the verb comparable with the nominal complement of agentive door-PPs in passives: cf. Mieren krioelen in de tuin ‘Ants are swarming in the garden’. This may even be clearer in the examples in (526).

Although plausible at first glance, there are serious problems with proposals of this kind, which will be discussed in the next subjection.

The fact that the (logical) subject need not satisfy the selection restriction that a verb like krioelento crawl imposes on its agentive argument may suggest that the meaning of the constructions in (514) is non-compositional. One way to avoid this conclusion is to assume that the predicative relations in the clause are expressed in a non-canonical way. We will consider one option here, which we will show to be untenable in the light of a wider set of data.

First, consider the examples in (527), which show that the verb krioelen requires its agentive subject to be plural or to be headed by a noun denoting a collection of entities; using a singular noun phrase such as de mierthe ant leads to an unacceptable result.

The examples in (528) show that the verb krioelen imposes similar restrictions on the nominal part of the van-PP similar to those on the subject in (527); the nominal part of the PP must be plural or refer to a collection of entities. Note in passing that, despite the definiteness of the noun phrase, it does not refer to a contextually determined set of entities, but has a high-degree reading in the sense that it expresses that there are many ants/a lot of vermin; cf. Hoeksema (2009) and Vandeweghe (2020).

This may suggest that the van-PP functions semantically as the logical subject of the verb. If so, this means that we are dealing with a rather complex set of predication relations, which are schematized in the figures in (529). The two constructions are identical in that the verb is predicated of the nominal part of the van-PP. The complex verbal phrase krioelen van de mieren functions as a predicate which is subsequently predicated of the reference object de tuinthe garden, directly if the latter is realized as the subject of the clause, or via the anticipatory pronoun het if it is realized as a locational PP.

529

a.

to be rejected

b.

to be rejected

There are several possible problems with such analyses. The first is that the predication relation I between the verb and its internal argument is usually not syntactically encoded by the preposition vanof in standard Dutch (although this has been different in earlier stages of the language, as pointed out in Vandeweghe 2020). This does not a priori mean that such an analysis would be untenable, since it is argued in Section N17.4.2.1 that this preposition can establish such a relation in metaphorical N-van-een-N constructions such as een schat van een kata treasure of a cat, in which the noun schat is predicated of the second noun; cf. Die kat is een schatthat cat is a treasure. A second semantic problem is that establishing predication relation I in the structures in (529) should produce a proposition; since propositions are saturated predicates, they cannot normally be predicated of another argument, and this means that we have to make additional stipulations in order to make predication relation II possible. The third and probably most problematic aspect of the analyses in (529) is that it predicts that in constructions of this type the verb is always predicated of the nominal part of the van-PP. However, the examples in (530) show that this need not be the case.

The verbs barstento burst, stikkento suffocate and stervento die are clearly not predicated of the noun phrase de toeristen. Instead, the original meaning of the verbs has bleached and the construction as a whole simply takes on a quantitative aspect of meaning; there is an extremely high number of tourists in town. It may further be noted that the syntactic properties of the verbs barsten, stikken and sterven in the constructions in (530) also differ considerably from their properties in their more regular use. This is shown in (531) and (532) for the verb stikken. Example (531) shows that this verb, being a telic unaccusative verb, forms its perfect tense with the auxiliary zijnto be and cannot be combined with durative adverbial phrases like een uur langfor an hour.

The constructions in (532), on the other hand, exhibit properties of atelic predicates:

they form their perfect tense with the auxiliary hebbento have and can be combined with durative adverbial phrases such as de hele zomerall summer.

To summarize the discussion, we can conclude that the meaning of the constructions under discussion cannot be determined in a compositional way. The verbs in this construction also have the property that their meaning has bleached; they do not denote the same state of affairs as in their more regular use, a semantic change that is also reflected in their syntactic behavior.

An essential aspect of the meaning of the two constructions under discussion seems to be that there is a high concentration of entities at a particular location. It has further been claimed that the two alternants differ in the spread of these entities. Constructions with a nominative subject like (514b) and (530b) are given a holistic interpretation: e.g. (530b) expresses that wherever you go in town, there will be many tourists. Impersonal constructions like (514a) and (530a), on the other hand, have been claimed to be consistent with a partial interpretation: for instance, (530b) may be true if there is a high concentration of tourists in certain limited areas of town.

The nominative/PP alternation under discussion seems to be very productive, and many verb types can enter the construction. Example (513) has already shown that the alternation can occur in copular constructions. The examples in (511) and (512) in Subsection I have further shown that positional verbs with the complementive adjective volfull also enter this alternation; one example is repeated here as (533).

The examples in (534) provide a number of other possible cases with an adjectival complementive, although these are somewhat harder to judge because they have a more or less idiomatic flavor. The examples in (534) are similar to those in (514) with the verb krioelen and (530) with the verbs barstento burst, stikkento suffocate and stervento die: they contain an obligatory van-PP and express that there is a high concentration of entities denoted by the nominal part of the van-PP at the location denoted by the reference object die krantthat newspaper/de stadthe city.

Another set that allows the alternation consists of verbs denoting light and sound emission. Note that the van-PP in (535a') is optional, but this may be due to the fact that schitterento glitter can also be used as a monadic verb: De diamant schitterdeThe diamond sparkled. These constructions again express that there is a high concentration of entities denoted by the nominal part of the van-PP at the location denoted by the reference object de luchtthe sky/de tuinthe garden.

Finally, the examples in (536) provide a number of examples of bodily sensation/function that seem to particularly favor the construction in which the reference object is realized as the subject of the clause.