Impersonal middles are like adjunct middles, but differ from regular middles in that they can easily take an intransitive verb as input. Again, this may be because the subjects of impersonal middles do not correspond to the direct objects of the input verbs.

The examples in (271) show that adjunct middles behave like regular middles, but differ from passives in that the subject of the input verb cannot be expressed by an agentive door-PP. Still, the notion of agent seems to be implied; the evaluative modifier provides an assessment of some property of the subject in relation to the activity denoted by the verb, and thus indirectly introduces the notion of agentivity.

Transitive verbs can be used as input for impersonal middle formation only if they can be used as pseudo-intransitives; overt realization of the objects in the middle constructions in the primed examples in (272) leads to unacceptability.

For the same reason, ditransitive verbs cannot normally be the input verb for impersonal middle formation; this is at best marginally possible if the direct object can be omitted, as in the (b)-examples of (273).

Impersonal middle formation is normally not possible on the basis of unaccusative verbs, although again we may have to make an exception for unaccusative verbs such as vallento fall, which allow a controlled stage-context reading; cf. Section 3.2.2.3, sub I.

The examples in (275) show that impersonal middles take the auxiliary hebbento have in the perfect tense; this is also true for middle verbs derived from unaccusative verbs such as vallen, which normally take zijn.

The discussion in the previous subsection has shown that impersonal and adjunct middles are similar in that they both take (pseudo-)intransitive verbs as input; cf. Section 3.2.2.3, sub I. Furthermore, we have seen that the two constructions are also very similar semantically. Let us therefore assume for the moment that impersonal middles are the impersonal counterparts of adjunct middles: the non-referential pronoun het is an anticipatory pronoun associated with the adverbial PP which acts as the logical subject of the clause. The impersonal middles in (270) and (272) would then be assigned the structures in (276), in which coindexing is used to express the proposed relation between the anticipatory pronoun het and the adverbial phrase. We also refer the reader to Section A28.6 for a discussion of comparable copular constructions: Heti is warm [in de tropen]i It is hot in the tropics.

The syntactic and semantic similarities between adjunct and impersonal middles can now be explained by assuming that subjects of adjunct middles correspond to adjuncts coindexed with the anticipatory subject pronoun het in impersonal middle constructions. If so, we expect impersonal middles to be subject to similar restrictions as adjunct middles. For example, Section 3.2.2.3, sub VC, argued that the unacceptability of examples such as (277b) shows that subjects of adjunct middles cannot correspond to the nominal part of a PP-complement. We therefore correctly predict the unacceptability of (277c), in which the anticipatory pronoun is coindexed with a PP-complement. Note in passing that this example improves considerably when we add a locational adverbial phrase, as in (277c'), which is expected because the anticipatory pronoun can then be construed with the adverbial phrase.

Section 3.2.2.3, sub VC, also argued that the subject of an adjunct middle cannot correspond to the nominal part of a predicative PP. This was illustrated by the examples in (266), repeated here as (278), which show that the adverbial and predicative PPs can be distinguished by their meaning: an adverbial PP simply indicates where the activity denoted by the verb takes place, whereas the predicative PP refers to the new location that the subject of the clause obtains by performing the activity denoted by the verb.

The subject in the adjunct middle construction in (267), repeated here as (279a), clearly corresponds to the adverbial PP: it is the jumping on the trampoline that is said to be nice, not the jumping onto the trampoline. Example (279b) shows that the same holds for the corresponding impersonal middle.

We conclude from the discussion above that anticipatory pronouns in impersonal middles can only be coindexed with adverbial phrases. This may have interesting consequences for cases in which the status of a particular PP is unclear, such as the locational PP that co-occurs with the verb wonento live/reside in example (280a). Since this PP is obligatory, it is sometimes suggested that it is selected (e.g. as a PP-complement or a complementive) by the verb; cf. Klooster (2001a:154-6). However, the fact that this example has an impersonal middle counterpart can be seen as a conclusive argument for analyzing it as an adverbial phrase.

In relation to the hypothesis that the non-referential pronoun het in the impersonal middle functions as an anticipatory pronoun associated with an adverbial PP acting as the logical subject of the clause, it may be useful to note that impersonal middles are special in that their nominalizations can easily be used as nominal predicates. The resulting copular constructions in (281) are also impersonal in nature; the subject pronoun het cannot be replaced by e.g. a deictic pronoun and seems to function as an anticipatory pronoun co-indexed with the adverbial PP; we refer the reader again to Section A28.6 for a discussion of comparable constructions.

Finally, note that impersonal middles have no corresponding constructions in which their participles are used as attributive modifiers or secondary predicates. This is not due to the verb, but to the fact that the subject pronoun het is non-referential and can therefore not be modified or function as an argument of a predicate.

The previous subsection suggested that adjunct middles always have an impersonal middle counterpart. It does not seem to be the case, however, that the reverse is also true: impersonal middles seem to be possible with a somewhat wider range of adjunct types than adjunct middles. This can be seen by comparing the examples in (282); the non-middle construction in (282a) has an impersonal but not an adjunct middle counterpart.

The examples in (283) suggest that the contrast in acceptability between impersonal and adjunct middles can, at least sometimes, be related to meaning. The nominal part of the adverbial met-PP in (283a) can easily be used as the subject of an adjunct middle, whereas the nominal part of the zonder-PP in (283b) cannot. This is clearly related to the fact that the implicit negation expressed by zonderwithout is irrecoverably lost in the adjunct middle in (283b''); this example is only possible with the meaning “It is nicer to ride with a helmet”. Note that Zonder helm rijdt lekkerder is more or less acceptable; however, this is not a middle construction, but a construction with a PP subject.

In other cases it is less clear what determines the contrast, although the examples in (284) suggest that the nominal part of the adjunct must refer to an entity/entities that is/are in some conventional relation to the activity denoted by the verb, in order to be able to appear as the subject of an adjunct middle.

Ackema & Schoorlemmer (1994, 2006/2017: §6.2) suggests that the contrast may also be related to the ability of the adjunct-PP to undergo preposition stranding; cf. the contrast between the primeless and primed examples in (285).

Ackema & Schoorlemmer also suggests that subjects of adjunct middles correspond to argument-PPs, not to adjuncts, which would make the adjunct middle a kind of regular middle. We do not follow this suggestion, for two reasons. First, analyzing the PPs in question as arguments is rather dubious; second, this proposal contradicts the findings of Section 3.2.2.2, sub IA, that subjects of regular middles never correspond to the nominal part of uncontested PP-complements of the input verb.

Impersonal middles usually require the presence of an evaluative modifier listed in (179), such as gemakkelijkeasily. However, as in the case of adjunct middles, impersonal middles sometimes also occur with adjectives like snelfast and vlotsmoothly, which do not select an experiencer voor-PP: compare the impersonal construction in (286b) with example (259b), repeated here as (286a).

The examples in (287) show that the experiencer selected by evaluative adjectives such as gemakkelijkeasily cannot normally be realized overtly (with the same caveat made in Section 3.2.2.2, sub IC, for the regular middle).

As in the case of adjunct middles, the evaluative modifier can be (at least marginally) omitted if the negative adverb nietnot is present or the verb is emphatically accented, as in (288a&b), or if the evaluation is expressed in some other way, e.g. by using the metaphorical expression like a train ‘very well, smoothly …’ in (288c).