The set of verbs taking opaque te-infinitival complement clauses includes nom-dat verbs such as berouwento regret, PO-verbs such as neigen (tot)to be inclined (to), and particle verbs such as opdragento order in (53); cf. Evers (1975:39ff) and Den Besten et al. (1988). The primeless examples show that such verbs do not allow clause splitting; like om + te-infinitivals, the complete te-infinitival clause is in extraposed position, i.e. they follow the matrix verb in clause-final position. The primed examples show that the matrix verbs appear as past participles in the perfect tense; as with om + te-infinitivals, there is no IPP-effect. We will italicize the te-infinitivals in the following examples and, for simplicity, refrain from indicating their implicit PRO-subjects; it should be added immediately after the clause-final main verb.

Opaque infinitivals seem to be characterized by the fact that they do not have the syntactic function of direct object of the matrix verb, or are not assigned a thematic role by it. The infinitival clauses in the (a)-examples function as subjects introduced by the anticipatory subject pronoun hetit. The infinitival clauses in the (b)-examples correspond to the nominal part of a PP-complement of the matrix verb, as can be seen from the fact that they are introduced by the anticipatory pronominal PP ertoeto it. Finally, the infinitival clauses in the (c)-examples are not arguments of the verb at all, but are licensed as logical subjects of the verbal particle op; cf. Section 2.2.1.

Verbs that select a transparent infinitival complement often have a modal or aspectual interpretation. Examples are the modal verbs schijnento seem, lijkento appear and blijkento turn out. That the infinitival complements of these verbs are transparent is clear from the fact that they are obligatorily split; while the te-infinitive in (54) must follow the matrix verb in clause-final position, its object must precede it.

That the infinitival complement in (54) is transparent is also clearly shown by the fact that we are dealing with subject-raising constructions; the subject of the infinitival clause becomes the nominative subject of the higher clause. This becomes clear when we consider the almost equivalent examples in (55): the subject of the finite complement clause in (55a) appears as the nominative subject of the whole sentence in (55b), in which the complement clause is infinitival.

Unfortunately, it is more difficult to show that the modal verbs schijnento seem, lijkento appear and blijkento appear trigger the IPP-effect, for the simple reason that these verbs cannot normally be used in perfect-tense constructions; cf. (56a). Some speakers, however, do allow the perfect tense marginally in subject-raising constructions, provided that they exhibit the IPP-effect; cf. (56b).

Other examples of transparent te-infinitivals mentioned by both Evers (1975:5) and Den Besten et al. (1988) are the somewhat formal/obsolete semi-modals dienento be obliged to, plegento tend/be accustomed to and wetento be able to, which seem to have a deontic interpretation and are probably best analyzed as control structures. However, it is difficult to find independent support for the control analysis: in principle, pronominalization of the infinitival clause [PRO dat boek te lezen] would be a good test, were it not for the fact that it leads to the loss of the modal interpretation of the matrix verbs in favor of their more common lexical meaning “to serve/commit/know”. The transparent nature of the te-infinitivals in (57) is evident from the fact that clause splitting and the IPP-effect are obligatory in these examples.

Evers (1975) suggested that te-infinitivals functioning as theme arguments and appearing as direct objects can be either opaque or transparent (in our terms), but he also noted that some verbs, his class IIIb, are not very particular in the sense that they can select either type. We illustrate the latter case in (58) with perfect-tense constructions containing the matrix verb proberento try. That the verb appears as a participle in (58a) but as an infinitive in (58b) suggests that we are dealing with an opaque and a transparent infinitival clause, respectively. This is also supported by the fact that the infinitival clause is split in (58b) but not in (58a). Following the standard hypothesis of the time that Dutch has an underlying OV-order, Evers accounted for this by assuming that the direct object clause is base-generated to the left of the matrix verb, and that (58a) and (58b) are derived by extraposition of the whole clause and verb raising of the infinitival verb te lezento read, respectively.

However, the examples in (59) suggest that it is not sufficient to assume that certain verbs optionally trigger extraposition or verb raising. The unacceptability of example (59a) first shows that extraposition does indeed require the matrix verb to appear as a past participle in perfect-tense constructions; there are no extraposition constructions with IPP in standard Dutch (although Barbiers et al., 2008: §2.3.6.1.3, shows that there are Flemish and Frisian dialects that accept examples such as (59a)). However, Den Besten et al. (1988) found that clause splitting is very common when the matrix verb occurs as a participle; clause splitting does not require IPP, as can be seen from the fact that example (59b) alternates with (58b); see also Haeseryn et al. (1997:950-2) and taaladvies.net/taal/advies/vraag/674, as well as Gerritsen (1991: Map 25) and Barbiers et al. (2008) for this usage in the Dutch dialects.

Note that the verb proberen is really special in that it seems to be equally acceptable with opaque, transparent, and semi-transparent infinitival complement clauses. Many verbs are more restrictive in this respect (although there is always some variation in what speakers do or do not accept): besluitento decide, for example, can only take opaque or semi-transparent te-infinitivals, which are incompatible with the IPP-effect, as illustrated in (60b').

The main conclusion to be drawn from Subsections I to III is that, from an observational point of view, we can distinguish the three types of te-infinitivals in (61) on the basis of whether or not clause splitting and IPP are possible.

However, semi-transparent te-infinitivals differ from transparent ones in that the former do not require that all nonverbal constituents of the infinitival clause precede the matrix verb; cf. the contrast between the two examples in (62). This bears out that clause splitting of semi-transparent te-infinitivals is not the result of verb clustering in the technical sense defined in the introduction to this section, i.e. the formation of an impermeable series of verbs in clause-final position.

Constructions with semi-transparent te-infinitivals like (59b) and (62b) were called third constructions in Den Besten et al. (1988), but later became known as remnant extraposition constructions. Den Besten et al. (1988) derived the construction by a combination of extraposition of the te-infinitival and leftward movement of one or more of its constituents. As a result, the extraposed phrase consists only of a remnant of the original te-infinitival; cf. also Reuland (1981). If we adopt the leftward movement analysis (while leaving open the question of whether extraposition involves rightward movement of the infinitival clause), the representations of (59b) in (62b) are as given in (63).

The fact that the direct object een kusa kiss in (62b) can either precede or follow the clause-final verbs implies that the postulated leftward movement is optional (i.e. subject to some constraints). This means that it is no longer obvious that the te-infinitivals in examples such as (64) should be considered opaque, since they can also be analyzed as semi-transparent clauses without the postulated (optional) leftward movements in (63).

All this may indicate that Den Besten et al. (1988) were wrong in assuming that there are opaque te-infinitivals, and that we should rather assume that all te-infinitivals are (semi-)transparent. If this is the case, then the “opaque” cases discussed in Subsection I cannot be described by appealing to the label clause type. Since (semi-)transparent infinitival clauses differ crucially from the opaque infinitival clauses discussed in Subsection I in that (i) they are selected as internal arguments of a verb, and (ii) they have the syntactic function of direct object, this may be the key to the solution. This will be one of the issues addressed in our detailed discussion of te-infinitivals in Section 5.2.2.