Subject-raising verbs differ from control verbs in that they do not take om + te-infinitivals. The unacceptability of the subject-raising construction in (487b) is easy to explain, because Section 5.2.2.1 has independently established that om + te-infinitivals are syntactic islands for movement, and can therefore be assumed to block subject raising. However, it is less clear why (487c) is unacceptable, especially since (480a) has shown that similar constructions are possible with finite clauses; this unacceptability may be due to the fact that there is no suitable controller for the implied subject PRO (Bennis & Hoekstra 1989a).

Such an account of the unacceptability of (487c) would leave unexplained, however, why the (c)-example in the parallel set of examples in (488) is unacceptable as well, given that the experiencer object me of lijkento appear could in principle function as a controller for PRO. We will not pursue this issue here and leave it to future research.

The examples in (487a&b) in the previous subsection show that infinitival clauses of subject-raising constructions must be transparent to NP-movement. This is consistent with the fact that such clauses are transparent infinitivals in the sense defined in Section 4.4.3: subject-raising constructions exhibit verb clustering (and thus require the embedded infinitival clause to be split), and the te-infinitive seems to trigger the infinitivus-pro-participio (IPP) effect on the matrix verb in perfect-tense constructions. The former can be illustrated by the contrast between the two examples in (489); the infinitival clauses are italicized.

It is less easy to show that subject-raising constructions exhibit the IPP-effect, because many speakers tend to reject perfect-tense constructions with evidential modal verbs; cf. Haeseryn et al. (1997:958), and also Schmid (2005:27), where it is claimed that subject-raising constructions tend to resist perfectivization cross-linguistically. Nevertheless, it seems that some speakers at least marginally accept perfect-tense constructions like those in (490) and then always prefer the IPP-effect; replacing the infinitives schijnen ‘to seem’, lijken ‘to appear’ and blijken ‘to turn out’ in (490) by the corresponding participial forms geschenen, geleken and gebleken indeed worsens the results considerably; cf. Reuland (1983: §3.2) and Rutten (1991:70).

Van der Horst (2008) claims that constructions with schijnen have exhibited the IPP-effect since the 18th century, which he illustrates with only one example (p.1464&1796), and also provides a number of recent IPP-cases with blijken (p.1769). These claims can be substantiated by Google Book searches on the strings [heeft schijnen te] and [heeft blijken te], although there are also a number of relevant hits for [heeft geschenen te] and [heeft gebleken te]. Van der Horst does not discuss cases with the verb lijken, and a Google Book search on the strings [heeft lijken/geleken te] also yields no relevant hits, although Haegeman (2006) claims that lijken occurs in the perfect tense. The results of our searches are given in Table (491); the reported results have been manually checked and exclude hits from linguistic sources.

	infinitive	participle
schijnen ‘to seem’	12	2
blijken ‘to turn out’	11	6
lijken ‘to appear’	0	0

The results in (491) are based on older written sources and are not representative of current usage. Unfortunately, the results of Google searches on the strings [heeft modalinf/part te] are too polluted with irrelevant cases (often machine translations from English) to say anything meaningful about the frequency of genuine cases of constructions such as (490) with and without IPP on the internet, except that the numbers are small. We therefore leave this issue for future research and tentatively assume that, to the extent that perfect-tense forms of subject-raising constructions are possible at all, they tend to exhibit the IPP-effect.

A more technical question raised by the assumption that subject-raising verbs are main verbs concerns the status of the infinitival clause: Is it an internal or an external argument of the modal verb, i.e. are we dealing with intransitive or unaccusative verbs? The unaccusative analysis seems plausible; since the subject of the infinitival clause indisputably appears as the nominative subject of the matrix clause, it seems unlikely that the infinitival clause is generated as the external argument of the matrix verb, because such arguments must appear as subjects of active constructions; this would make subject raising impossible. However, if the infinitival clause is generated as an internal argument of the verb, there is no external argument, and as a result the subject of the infinitival clause can raise to the subject position of the higher clause.

That we are dealing with unaccusative verbs is also supported by the fact that blijken takes zijn in the perfect tense (in non-IPP contexts): Dat is/*heeft geblekenThat has turned out, since the selection of the perfect auxiliary zijn is a sufficient condition for assuming unaccusative status. The complementive constructions in (492) show that schijnen and lijken do not allow zijn in the perfect tense; that they seem to prefer hebben is not a problem, however, since the selection of zijn is not a necessary condition for assuming unaccusative status; cf. Section 2.1.2.

The conclusion that subject-raising verbs are unaccusative correctly predicts that such verbs do not allow impersonal passivization. This is illustrated in (493) for the verb lijken in the three syntactic contexts in which it can occur. The reason why the nominative subjects cannot be suppressed in the primed examples is that they are not arguments of the passivized verb, but originate as arguments of the complements of that verb; for convenience, the (split) complements are italicized in the primeless examples.

The (b)-examples in (494) show that passivization of the embedded infinitival clause is possible; the (a)-examples are given for comparison. As predicted by the subject-raising analysis, passivization of the infinitival clause also affects the nominative subject of the subject-raising construction as a whole; the internal argument of the infinitival verb, de auto, becomes the subject of the infinitival clause and thus also appears as the nominative subject of the subject-raising construction as a whole.

Finally, consider the examples in (495) adapted from Bennis & Hoekstra (1989c:172); the judgments hold only for speakers who allow passivization of the idiomatic expression de strijdbijl begravento bury the hatchet/to make peace. The fact that the idiomatic reading is preserved in (495b') can be taken as evidence for the claim that the noun phrase de strijdbijl is base-generated as part of the infinitival clause: since phrasal idioms are listed in the lexicon, the expression de strijdbijl begraven must be inserted into the structure as a unit.

Subject raising requires the te-infinitival clause to be a complement of a verb; the primed examples in (496) show that while non-raising constructions such as (496a) have nominal counterparts, subject-raising constructions such as (496b) do not.

This suggests that te-infinitival complements of nouns differ from those of verbs in that they are not transparent to movement. This is consistent with Koster’s (1984b) claim, discussed in Section 5.2.2.1, that te-infinitival complements of nouns do not involve obligatory control either.

The previous subsections have shown that subjects of subject-raising constructions cannot be analyzed as arguments of the subject-raising verb, but originate as arguments of the embedded infinitival verb. Subject raising from te-infinitival complement clauses occurs with certain unaccusative verbs (but not with their corresponding nominalizations).

The verb blijken expresses that there is factual evidence in support of the proposition expressed by the argument clause. The use of this verb further suggests that the truth of the proposition can be at least intersubjectively established on the basis of the evidence available; most people who carefully consider this evidence would come to the same conclusion. Example (498a) shows that the nature of the factual evidence presented can be specified in an adverbial uit-PP when the clause is finite (or pronominalized by e.g. het ‘it’/dat ‘that’), but not in the corresponding subject-raising and complementive constructions in (498b&c); the latter examples nevertheless seem to imply that the truth of the proposition expressed by the infinitival/small clause can be established intersubjectively. Such adverbial uit-PPs are not usually found with the verbs lijken and schijnen; cf. Vliegen (2011: Table 3).

Note in passing that examples such as (498) are perfectly fine if the preposition uit is replaced by volgensaccording to: this may be due to the fact that the complement of the volgens-PP does not refer to the evidence on which the speaker bases his judgment about the truth of the proposition, but to the “judgment” provided by another source. While example (498a) expresses that the speaker concludes from Jan’s statement that Jan is the perpetrator, an example like Volgens zijni verklaring blijkt dat Jani de dader is attributes this conclusion to Jan himself.

It often seems to be implied that there is a particular group of people who drew the conclusion from the available evidence. With a finite complement clause the person(s) responsible for the conclusion can be expressed by a dative object (often the first-person plural pronoun onsus), usually referred to in the literature as the experiencer. The verb blijken should therefore be considered a nom-dat (i.e. dyadic unaccusative) verb. The addition of an experiencer leads to a degraded result in the corresponding subject-raising and complementive constructions.

It should be noted that the use of an experiencer object is limited even in the case of finite argument clauses: although it seems possible in perfect-tense constructions, it is generally rejected in simple past/present constructions. This is also clear from a Google search (October 17, 2023): the string [er is ons gebleken dat] yielded 24 hits, while the strings [er blijkt/bleek ons dat] yielded no more than 9 relevant hits. The use of the dative experience may therefore be obsolete, which may be supported by the fact that many of these examples come from older or more formal texts.

The verb lijken indicates that the claim expressed by the argument clause in (497a) is based on (unmentioned) available evidence; in a sense, we are dealing with a subjective assessment of the evidence by a certain set of individuals, which by default includes the speaker. However, example (500a) shows that this set can also be made explicit by using an optional experiencer object, in which case the default reading can easily be overridden. The availability of an experiencer object shows that the verb lijken, like blijken, should be considered a nom-dat (dyadic unaccusative) verb. However, lijken differs from blijken in that the experiencer can easily appear in the corresponding subject-raising and complementive constructions in (500b&c).

The verb lijken differs from the other two verbs in that it can easily take a finite clause introduced by the linking element (als)ofas if; the judgments on examples (501b) with schijnen vary from speaker to speaker, which is indicated by the percentage sign; we will return to this issue briefly in Subsection 4.

However, it is not obvious that we can conclude from this that the modal verb lijken can take an alsof-complement. An argument against this conclusion is that lijken also occurs as a PO-verb with the meaning “to resemble”; cf. example (502a). Since Section 2.3.1, sub VI, has shown that anticipatory pronominal PPs can often be omitted, it seems plausible to assume that example (501a) is a shorter form of example (502b) and thus does not involve the modal verb lijken; note in passing that the declarative complementizer dat ‘that’ is also possible in example (502b).

On the other hand, example (503a) shows that (501a) can be extended with an experiencer object; the fact, illustrated in (503b), that the experiencer and the anticipatory pronominal PP cannot occur together thus argues against the elision analysis. For completeness’ sake, note that Google searches (10/15/2023) for the strings [het lijkt mij/me alsof] and [het lijkt mij/me erop alsof] also show that only the former is common, but that some speakers also consider example (503a) to be marked.

The above discussion seems to lead to the conclusion that the evidential modal verb lijken is compatible with an alsof-complement after all.

An experiencer object is unlikely with the verb schijnen in examples such as (497c), and the same goes for the corresponding subject-raising and complementive constructions in (504). The reason is that schijnen indicates that the truth of the proposition is based on hearsay/rumors; unlike blijken and lijken, the assumption of the truth of the proposition is not based on evidence available to any identifiable individual in the domain of discourse, and may even be lacking.

Moreover, the examples in (505) show that schijnen differs from blijken and lijken in that it does not easily allow the pronominalization of its complement clause. This shows that there is no evidence for the claim that the raising verb schijnen is a main verb; main verb status can only be argued on the basis of the assumption that schijnen belongs to the same class as blijken and lijken.

The three modal verbs blijkento turn out, schijnento seem and lijkento appear differ in the type of evidence available for testing the truth of the proposition expressed by the complement of the verb: blijken suggests that there is strong evidence on the basis of which the truth of the proposition can be established objectively or intersubjectively, lijken suggests that the evidence is weaker in the sense that it is not conclusive and can only be interpreted in a subjective way, while schijnen suggests that the nature of the evidence is unclear or even absent; cf. Sanders & Spooren (1996) for experimental support of these findings.

It seems that speakers often have difficulties in judging examples with the modal verbs blijken, lijken and schijnen. One reason for this may be the interference of other constructions. We have seen, for example, that the dyadic modal verb lijken has a closely related counterpart that functions as a PO-verb with the meaning “to resemble”; these verbs are diachronically derived from the same source and are still quite close in meaning. Furthermore, the situation in Dutch is rather special in that Dutch has two verbs, viz. lijken and schijnen, while German has only one verb, scheinen. This suggests that the meanings of these verbs are quite close, which may cause some confusion among speakers, especially since the introduction of lijken is quite recent (Vliegen 2011 suggests the 17th century).

To conclude this brief discussion of blijken, lijken and schijnen, we would like to note that they often occur in examples such as (506), where they are part of a phrase headed by the preposition naar; the pronoun hetit is often optional (especially with the verb blijken). Vliegen (2010) calls such naar-phrases parenthetic. Such an analysis may indeed be appropriate for cases such as Jan is (naar het schijnt) de beste leerling van zijn klasAs it seems, Jan is the best pupil in his class, but it is clearly wrong for examples such as (506), because this phrase is used in the initial position of the sentence and must therefore be considered a clausal constituent.

To our knowledge, examples such as (506), which can be easily found on the internet using the search string [naar (het) Vmodal], have not yet been analyzed in detail, and we therefore leave them for future research.

Belovento promise and dreigento threaten are generally used as verbs with an illocutionary meaning. The examples in (507) show that the illocutionary verb beloven is a triadic verb that can select a noun phrase or a clause as its direct object. For our discussion in the following subsections it is important to note that the complement in (507c) is an opaque infinitival clause in the sense of Section 4.4.3; it is in extraposed position and the infinitival verb does not trigger the IPP-effect, i.e. the participle beloofd cannot be replaced by the corresponding infinitive beloven.

The examples in (508) show that the illocutionary verb dreigento threaten is normally an intransitive PO-verb, and that the complement of the PP can be either nominal or clausal; in the latter case the clause is optionally introduced by an anticipatory pronominal PP. For our discussion in the following subsections, it is again important to note that the complement in (508c) is not a transparent clause: it is in extraposed position and the infinitival verb does not trigger the IPP-effect, i.e. the participle gedreigd cannot be replaced by the corresponding infinitive dreigen.

Now that we have discussed the regular uses of belovento promise and dreigento threaten, we can continue with their use as subject-raising verbs in the examples in (509). That we are dealing with subject raising in these examples is clear from the fact that the inanimate noun phrases do not have the proper semantic properties to be assigned the agent roles of the illocutionary verbs beloven and dreigen.

That the nominative subjects are not arguments of the verb beloven and dreigen is further supported by the fact, illustrated by the primeless examples in (510), that the infinitival clause cannot be pronominalized without the subject of the sentence. In fact, the primed examples show that the anticipatory elements het/ermee cannot be used to introduce the infinitival clause either.

A third argument in favor of subject raising is that the complementizer om is not allowed: the ungrammaticality of the examples in (511) is as expected given that om + te-infinitivals are islands for extraction and thus block the indicated NP-movements. Note that in this respect the modal verbs beloven and dreigen behave conspicuously differently from the corresponding illocutionary verbs in (507c) and (508c), which readily allow om + te-infinitivals as their complements.

A fourth argument is that beloven and dreigen are like the subject-raising verbs blijkento turn out, schijnento seem and lijkento appear discussed in Subsection A in that they are often not accepted in the perfect tense, but trigger the IPP-effect when speakers do accept them; this is clear from the fact that Barbiers (2006) marks (512) as unacceptable, while Van Dreumel and Coppen (2003) assign it a question mark to indicate that not all speakers consider it grammatical.

In subject-raising constructions, beloven and dreigen take on an evidential or perhaps epistemic modal meaning; they express that the available evidence is sufficient for the speaker to conclude that the eventuality expressed by the infinitival clause may come to pass. The original illocutionary meaning of these verbs is lost: they no longer denote the illocutionary acts of promising or threatening, but express a positive (in the case of beloven) or negative (in the case of dreigen) evaluation by the speaker of the eventuality expressed by the infinitival clause; cf. Verhagen (2005) and Vliegen (2006).

That we are dealing with modal verbs is supported by the fact that the choice of present or past tense can affect the implications concerning the question as to whether the embedded proposition is actually realized; cf. Section 1.5.2 for similar observations concerning epistemic modals like moetenmust and kunnenmay. Consider the examples in (513). Example (513a) leaves entirely open whether Marie’s promising career will actually lead to her becoming a great writer. Example (513b), on the other hand, strongly suggests that something unforeseen happened: for example, Marie would have been a great writer had she not been killed in an accident.

The reason for the negative implication in (513b) is pragmatic and follows from Grice’s (1975) maxim of quantity: if the speaker knows at the moment of speaking that Marie is already a great author, he can be more precise by simply using a present tense: Marie is een groot schrijver (geworden)Marie is/has become a great author. That we are dealing with pragmatics is also clear from the fact that any negative inference can be overridden by contextual information: for example, adding the adverbial phrase al vroegalready early in her career to example (513b) results in the positive implication that Marie is a great author at the moment of speaking; cf. Marie beloofde al vroeg een groot schrijver te wordenAlready early in her career Marie promised to become a great author. Of course, we find this pragmatic effect not only in the case of beloven, but also (and perhaps more pervasively) in the case of dreigen.

The corpus study in Vliegen (2006) has shown that in the vast majority of cases the infinitival clauses embedded under the modal beloven are copular constructions; cf. the examples in (514).

The infinitival clauses embedded under the modal dreigen, on the other hand, can be more varied, as shown by the fact that the examples in (515) are both perfectly acceptable in a modal reading.

The higher degree of productivity of the subject-raising construction with modal dreigen may be related to the fact, also noted by Vliegen, that it arose earlier in the language than the corresponding construction with beloven, with the result that the illocutionary reading of the latter may be more prominent than that of the former.

The previous subsection has shown that constructions with beloven and dreigen can be ambiguous when the nominative subject is animate; cf. Bennis & Hoekstra (1989c:174-5). The verb beloven in examples such as (516a) can be interpreted as a control verb or as a subject-raising verb, because there are no syntactic clues favoring one interpretation over the other. Of course, we abstract from the fact that the (extra-)linguistic context may disambiguate (516a) by favoring a particular interpretation.

Example (516a) can be disambiguated by adding the complementizer om, or (perhaps) by adding the anticipatory pronoun hetit (which leads to a somewhat marked result here), since these additions both exclude the subject-raising reading; for convenience, the elements originating within the infinitival clause are italicized in (517).

Furthermore, the examples can also be disambiguated when they are used as embedded clauses. When the infinitival clause is in extraposed position, as in (518a), we usually interpret the construction as a control structure with an illocutionary verb (but see Subsection C for more discussion). When we are dealing with verb clustering (i.e. a split infinitival clause), as in (518b), we usually have a subject-raising construction with a modal verb (although Section 5.2.2.3 will show that we should be careful not to jump to conclusions when clause splitting seems possible).

Another way of disambiguating (516a) is the addition of an indirect object; example (519) does not allow a modal interpretation of the verb beloven. This is also clear from the fact illustrated in the (b)-examples that the infinitival clause cannot be split in embedded contexts The disambiguating effect of the addition of an indirect object shows that control and subject-raising verbs differ not only in meaning but also in adicity. Note that the number sign is used in (519b') to indicate that, surprisingly, many speakers find this example acceptable in a control reading; we return to this issue in Section 5.2.2.3.

The verb dreigen in examples such as (520a) can be interpreted as a control verb or as a subject-raising verb, again because there are no syntactic clues favoring one interpretation over the other; again, we are abstracting from the fact that the context may disambiguate (516a) by favoring a particular interpretation.

Like example (516a) with beloven, example (520a) can be disambiguated by adding the complementizer om or by adding an anticipatory pronominal element, here appearing as the PP ermeewith it; both options exclude the subject-raising reading. For convenience, the elements originating within the infinitival clause are again italicized in (521).

The examples are also disambiguated when they are used as embedded clauses: when the infinitival clause is in extraposed position, as in (522a), we are usually dealing with a control structure; when we find clause splitting, as in (522b), the subject-raising reading is preferred (we return to this issue in Section 5.2.2.3).

Another way to disambiguate example (520a) is passivization. In the control construction, the nominative subject is an agentive argument of the verb dreigen, and consequently we expect impersonal passivization of this verb to be possible in the control reading; example (523a) shows that this expectation is indeed borne out. In the subject-raising construction, the nominative subject is an argument of the infinitival verb, and we expect passivization to result in the promotion of the object of the infinitival verb to subject of the main clause, with the concomitant suppression of the nominative subject of the corresponding active construction; example (523b) shows that this expectation is again borne out.

Although we would in principle expect the same passivization possibilities for beloven, we have not been able to construct examples of the type in (523b) with it. This is clearly related to the fact noted earlier that examples such as (524a) cannot be construed as subject-raising constructions.

The fact that (524a) does not allow a subject-raising reading correctly predicts that passivization of the verb beloven is possible, but that passivization of the infinitival verb is impossible. The former is due to the fact that the implied PRO-subject in the resulting structure in (525a) can be controlled by the noun phrase in the agentive door-PP (which can of course be left implicit, in which case PRO receives an arbitrary interpretation). Explaining the latter is a bit more complicated, because we have to consider two different structures. First, the control structure in (525b) is probably ruled out because the noun phrase de gemeentethe municipality is not a suitable antecedent for the implied PRO-subject of the passive infinitival clause for semantic reasons. Second, the subject-raising construction in (525b') is of course excluded because beloven simply does not allow subject raising; cf. (524b').

In short, since in the vast majority of cases the modal verb beloven takes an infinitival copular construction as its complement, and copular constructions do not allow passivization, we predict that subject-raising constructions with embedded infinitival passive clauses will be rare (if existing at all).