An overtly expressed subject of the bare infinitive makes it very unlikely that we are dealing with a bare-inf nominalization. The reason is that in nominalizations the subject of the input verb is never expressed by a prenominal noun phrase: it is either left implicit or expressed by a postnominal van or door-PP. We illustrate this in (691) using nominalizations of an intransitive, a transitive, and an unaccusative verb. Note that we used det-inf nominalizations in the primeless examples because bare-inf nominalizations strongly favor their nominal argument in prenominal position; we refer the reader to Section N14.3.1.2, for a detailed discussion of the position and form of the arguments of the two types of nominalization.

The crucial thing for our present purpose is that the primed examples in (691) are unacceptable regardless of whether the determiner hetthe is present or not, whereas the AcI-constructions in (692) are perfectly acceptable.

The fact that the subject of the input verbs of the nominalizations in (691) cannot be expressed by a noun phrase in prenominal position makes it very unlikely that the bracketed phrases in (692) are bare-inf nominalizations; we can safely conclude that we are dealing with bare infinitival complement clauses. That this is the correct analysis is also clear from the fact that the bare infinitival clauses allow splitting: the bare infinitives preferably follow the perception verbs in clause-final position and are thus normally separated from their nominal arguments, which must precede the clause-final verb cluster as a whole (test I and II in Table (689)).

More support for the assumption that we are dealing with bare infinitival complement clauses is provided by the fact that the presence of the bare infinitive triggers the IPP-effect (test III); in perfect-tense constructions such as (694), the perception verbs cannot occur as past participles, but must instead occur in their infinitival form. Furthermore, the fact that the bare infinitives cannot precede the perception verb shows that they cannot be the head of bare-inf nominalizations (test II); cf. Jan heeft die film gezienJan has seen that movie.

Although these facts already firmly establish that the bracketed phrases in the AcI-constructions in (692) cannot be bare-inf nominalizations, we will still apply the remaining tests for the sake of completeness. First, the primeless examples in (695) show that the bare infinitives can follow the negative adverb niet, but cannot be preceded by the negative article geenno; tests V and VI thus confirm that we are dealing with bare infinitivals.

Second, the examples in (696) show that even though the bare infinitives can precede the perception verbs in clause-final position, they cannot be moved further to the left by focus movement, despite the fact that the intended meaning of these examples is perfectly plausible: “that Jan liked to see .....”; test IV thus again confirms that the bracketed phrases are not nominal but verbal in nature.

In short, we have ample evidence to conclude that the presence of a noun phrase corresponding to the subject of the bare infinitival is incompatible with analyses according to which the perception verb zien in (692) is complemented by a bare-inf nominalization; instead, we are dealing with bare infinitival object clauses.

The subject of the embedded bare infinitival clause can be left implicit under certain conditions. Examples such as (697) suggest that the bare infinitive must be transitive; omitting the subject of monadic (intransitive and unaccusative) verbs usually produces a marked result.

The degraded status of the examples in (697b&c) supports the conclusion from the previous subsection that perception verbs are incompatible with nominalizations, because bare-inf nominalizations are not sensitive to the adicity of the input verb: the examples in (698) are all perfectly acceptable.

However, the examples in (699) show that there is at least one exception to the general rule that the subject of monadic verbs cannot be omitted in AcI-constructions; verbs expressing sound emission usually yield perfectly acceptable results.

Since we cannot a priori rule out an analysis according to which the primed examples involve bare-inf nominalizations, we will examine these cases in more detail. Before doing so, it should be noted that the infinitival clause in example (697a), which is the one typically discussed in the descriptive and theoretical literature, also involves sound emission. This example may therefore belong to the same type as the examples in (699), which seems to be supported by the fact that its pseudo-intransitive counterpart, Ik hoorde zingen (lit. I heard sing), is acceptable; cf. Vanden Wyngaerd (1994: §3). For this reason, we will include example (697a) in our examination below.

Again, the tests from Table (689) should be helpful in determining whether the bare infinitive is nominal or verbal in nature, or whether it can be both. Let us first consider whether the infinitive can be verbal. If so, it should be part of the verbal complex and thus be able to appear last in the clause-final verb cluster (test I and II). The examples in (700) show that this is indeed the case, regardless of whether the subject of the infinitive is overtly expressed or implied. The fact illustrated in (700b) that the constituent headed by the bare infinitive can be split by the verb horen also shows that the bare infinitive is part of the verb cluster.

Test III applies to the examples in (701). That the bare infinitive can be verbal is shown by the fact, illustrated by the primeless examples, that it triggers the IPP-effect, again regardless of whether the subject of the infinitive is overtly expressed or implied. However, the primed examples show that the IPP-effect does not occur when the bare infinitive precedes the verb sequence. This, together with the fact that the subject must be omitted in this case, shows that the bare infinitive can also be nominal.

That we are dealing with a nominalization is also clear from the fact, illustrated in (702), that the phrase headed by the bare infinitive can undergo focus movement (test IV); as expected, this again requires the subject to be implicit.

That the bare infinitives can be verbal or nominal is also supported by the fact that the bare infinitive in example (703) can be preceded by either the negative adverb nietnot or the negative article geenno (tests V and VI). In the former case, the bare infinitive must be interpreted as verbal, as is also clear from the fact that its subject can be overtly expressed, whereas in the latter case it must be parsed as nominal, as is also clear from the fact that its subject must be left implicit.

The above discussion has established that the perception verb horento hear can take a bare infinitival complement clause with an implicit subject. It is doubtful whether verbs like ziento see or vindento consider also have this possibility. The examples in (704), for instance, show that leaving the subject of the infinitival complement of zien implicit always leads to a degraded result. This suggests that leaving the subject implicit is only possible when the bare infinitive is a verb of sound emission; cf. Petter (1998:145).

The previous two subsections have established that perception verbs can take bare infinitival clauses as their complement. In fact, this seems to be the only viable analysis for constructions in which the subject is expressed by an accusative noun phrase; subjects of bare-inf nominalizations are never expressed by nominal phrases. On the other hand, if the subject is left implicit, it would in principle be possible to analyze the projection of the bare infinitive as a bare-inf nominalization. We examined such constructions with the tests from Table (689) and found that such structures are ambiguous. However, we also saw that the ambiguity arises only with the verb horento hear, i.e. when the bare infinitive is a verb of sound emission.

Subsection I claimed that the noun phrases de zonthe sun and de deurthe door in the AcI-constructions in the primed examples in (706) are subjects of the infinitival verbs. A first argument for this claim is that these noun phrases also function as subjects of the verbs opkomento rise and klapperento rattle in the finite complement clauses in the primeless examples.

Another argument is that the finite and infinitival clauses behave in a similar way under pronominalization: the pronoun dat can be used to pronominalize not only the finite clauses in the primeless examples, but also the bracketed phrase in the primed examples. The fact that the noun phrases de zon and de deur are part of the pronominalized phrases clearly shows that they are part of the infinitival clause and therefore are not introduced as internal arguments of the perception verbs. From this we can conclude that they function as subjects of the bare infinitival clauses.

That the accusative noun phrases in AcI-constructions are not internal arguments of the perception verbs is also clear from the fact that they can be pronominalized by the weak anaphor zich. The primeless examples in (708) first show that this is never possible when the accusative object is an internal argument of the verb; this is consistent with the observational generalization from Section N22.4 that the simplex reflexive zich cannot be bound by a co-argument. The coindexing in (708) indicates the intended binding relation.

The examples in (709) show that the simplex reflexive zich is possible in AcI-constructions with perception verbs. Note that the simplex reflexive can often be replaced by the complex reflexive zichzelf without a clear difference in meaning.

Constructions like (709a&b) do not seem very common in speech and may feel somewhat forced, for the simple reason that people tend not to register their own actions by visual or auditory means. Examples such as (709c), on the other hand, are quite common, and the same goes for more special infinitival constructions with ziento see such as (710), which express an illusory/epistemic reading. Note in passing that in examples like (709c) and (710) weak reflexives seem to be preferred to complex reflexives.

That the simplex reflexive is possible in examples like (709) and (710) is consistent with the generalization that the simplex reflexive zich cannot be bound by a co-argument, provided that the reflexive functions as the subject of the bare infinitival; it would be surprising, however, if it were an internal argument of the perception verb.

Although it is generally assumed that the subject of the infinitival clause is assigned accusative case by the perception verb, this is not easy to establish in Dutch. Of course, it is clear that we are not dealing with nominative case: the form of the pronoun in example (711a) shows that the subject of the infinitival clause is assigned objective case. And that we are dealing with accusative (and not dative) case may be supported by the fact that this case is indeed morphologically expressed in the German counterpart of this example in (711b); cf. Drosdowski (1995:739).

However, there is little independent evidence for the claim that the accusative case is assigned by the perception verb. One way to establish this would be to appeal to passivization: the fact that the accusative subject of the infinitival clause in the English example in (712a) becomes the subject of the matrix clause in the passive construction in (712b) can be taken as evidence for “exceptional case marking” of the subject of the infinitival clause by the matrix verb to expect.

Unfortunately, this kind of evidence is not available for Dutch AcI-constructions: passivization of such examples is always impossible. The (a)-examples in (713) show this for a construction in which the infinitive is monadic (i.e. intransitive or unaccusative), and the (b)-examples for a construction in which the infinitive is transitive; for the latter case we have also shown that passivization is impossible regardless of the position of the object of the embedded infinitive.

It seems to be a robust generalization for Dutch that “intermediate” verbs in verb clusters of more than two verbs cannot appear as participles; cf. Section 7.1. This is clear from the so-called infinitivus-pro-participio effect found in the perfect-tense counterpart of examples such as (714a).

However, the examples in (715) show that this is probably not the reason for the unacceptability of the primed examples in (713): the result of passivization is also unacceptable if we replace the participles in these examples by infinitives.

The examples in (716) further show that impersonal passivization of constructions in which the subject of the bare infinitival clause is left implicit is also impossible. Note in passing that impersonal examples such as dat er snurken werd gehoordthat someone heard snoring are perfectly acceptable; this could be used as an additional argument for the claim in Subsection III that examples like dat Jan snurken hoordethat Jan heard snoring can involve a bare-inf nominalization.

The unacceptability of the primed examples in (716) shows that the unacceptability of the primed examples in (713) is not related to the promotion of the accusative object to subject, and that this cannot be used as an argument against the standard “exceptional case marking” approach to Dutch AcI-constructions, although the unacceptability of passivization of AcI-constructions remains somewhat puzzling in itself; cf. Bennis & Hoekstra (1989b) for an attempt to explain this. The examples in (717) finally show that the primed examples in (713) cannot be saved by substituting te-infinitives for the bare infinitives. In this respect, Dutch differs sharply from English, which is otherwise similar to Dutch in that it does not allow passivization of bare infinitival constructions; cf. Burzio (1981:319) and Bennis & Hoekstra (1989b).

The discussion above thus shows that there is no clear evidence that the subject of the bare infinitival clause is assigned case by the perception verb; the main reason for assuming this is that it solves the problem that subjects of infinitival clauses cannot normally be assigned case by any element internal to infinitival clauses.

Although subjects of infinitival complement clauses of perception verbs can always be realized as nominal phrases, in some cases they need not be present. A typical example of this is given in (718a). Since example (718b) shows that the omitted subject can be overtly expressed by an agentive door-PP, De Geest (1972), Bennis & Hoekstra (1989b), and Bennis (2000) argue that the non-realization of the subject is the result of a passive-like process. De Geest further supports this proposal by pointing out that (718b) has more or less the same reading as the clumsy (perhaps unacceptable) morphological passive example in (718c).

A possible problem with analyses of this kind is that they lead to the expectation that subjects of infinitival clauses headed by intransitive verbs such as klagento complain are also optional, given that Dutch allows impersonal passivization: cf. Er werd geklaagd over de kouPeople complained about the cold (lit: there was complained about the cold). Examples like (719a&b) show that this expectation is not fulfilled: omitting the subject of the infinitival clause produces a degraded result. However, it t is not very clear what this shows, since the morphological passive in (719c) is also impossible.

Recall that we have seen in Subsection III that the acceptability of omitting the subject also depends on the matrix verb: examples in the literature typically involve the perception verb horento hear, and the examples in (720a&b) show that many speakers are less willing to accept similar examples with zien. This would of course be surprising if we were dealing with a productive syntactic process. It may be interesting to note in this context that the infinitival morphological passive in (720c) is perfectly acceptable; more examples can be found on the internet with a google search using the search string [zag * worden], although this search will also yield various cases with bare infinitival copular constructions.

Since the differences noted above have been little studied, it is clear that more research is needed before we can draw firm conclusions; for example, it is not clear to us to what extent the tendencies noted above are systematic and/or shared by larger groups of speakers.

Haeseryn et al. (1997:1053ff) has noted that perception verbs can be complemented by the progressive aan het + Vinf phrase. The primeless examples in (721) show that in this case the logical subject of the infinitive is realized as an accusative object. Although the primeless examples are not explicitly analyzed as AcI-constructions in Haeseryn et al., such an analysis is suggested by relating the primeless examples to the primed examples, which are clearly cases of AcI-constructions.

However, the primed and primeless examples differ in several respects. First, we notice that the two constructions can differ markedly in word order: the primeless examples in (722) show that in embedded clauses aan het + Vinf phrases must precede the perception verb in clause-final position, while the singly-primed examples show that they can also follow it.

In fact, the order in which the bare infinitive precedes the perception verb is a marked one, which is only allowed when the perception verb is finite, as is clear from the fact that in the perfect-tense constructions in (723) the bare infinitives must follow the perception verbs.

One might want to account for the unacceptable orders in (723) by hypothesizing that the bare infinitives are nominalizations. This would predict that the judgments on these examples would be reversed if the infinitive zien were replaced by the participial form gezien. The examples in (724) show that this prediction is incorrect, since the participle cannot be used at all.

Taken together, the examples in (723) and (724) show that the constructions with a bare infinitive exhibit the IPP-effect, and this brings us to the second difference between the constructions with aan het + Vinf phrases and bare infinitives: the examples in (725) show that only the latter trigger this effect. This strongly suggests that the aan het + Vinf phrases differ from the bare infinitives in that they are not verbal, which is of course supported by the fact that they also differ from the bare infinitives in that they must precede the clause-final verb sequence.

Finally, the examples in (726) show that the aan het + Vinf phrases cannot be moved to a more leftward position in the middle field of the clause. This strongly suggests that such phrases function as complementives; we will return to this issue in the more general discussion of progressive aan het + Vinf phrases in Section 1.5.3, sub I.

Subsection III argued that AcI-constructions such as Jan hoort kinderen/hen lachenJan hears children/them laugh cannot be analyzed in such a way that the perception verb takes a bare-inf nominalization as its complement, because in such nominalizations the subject of the input verb (here: kinderen/hen) cannot be realized as a prenominal noun phrase. What we did not discuss, however, is that the complete AcI-construction can be the input for bare-inf and det-inf nominalizations; the singly-primed examples in (727) are cases of the former, and the doubly-primed examples are cases of the latter.

Examples like these may prove to be very important in the final analysis of the nominalization process, since they show that nominalization does not only involve the conversion of a simplex verb into a noun, but can in fact take as its input a complex syntactic object, in this case the phrase consisting of the perception verb and the bare infinitive. This in turn may favor an approach in which bare-inf and det-inf nominalization are seen as processes that take place in syntax, since under standard assumptions syntactic objects such as verbal complexes are not stored in the lexicon; if so, this disfavors any approach that claims that bare-inf and det-inf nominalization are morphological processes that take place in the lexicon; we leave this to future research and refer the reader to Lieber (2018) for a relevant discussion.