Chapter 5 and Chapter 6 discussed cases in which a main or non-main verb takes a clause or a smaller verbal projection as its complement. We have seen that finite clauses introduced by the finite complementizer datthat or ofwhether and infinitival complement clauses introduced by the complementizer-like element om are usually in extraposed position, i.e. after the matrix verb in clause-final position. This is illustrated in the examples in (4), where the matrix verb is underlined and the complement clause is given in square brackets with the categorial label CP (= complementizer phrase). Since CPs do not allow splitting, we can disregard such cases for our present purpose.

The optionality of the complementizer-like element om in (4c) shows that infinitival clauses without this element can sometimes be in extraposed position, but in certain other cases they also allow or even require clause splitting. The latter is what we typically find when the complement clause is a bare infinitival clause. This is illustrated by the examples in (5); in (5a) the matrix verb is a main verb, while in (5b) it is the aspectual non-main verb gaanto go. We have italicized the dependent bare infinitival clause and underlined the verbs to emphasize that clause splitting leads to a clustering of the verbs in clause-final position.

The examples in (6) show that in the northern varieties of Dutch verb clusters are usually impermeable: the verbs in the cluster cannot be interrupted by a dependent of the embedded main verb. This is clear from the fact that the objects of the bare infinitives in (5) cannot occur between the matrix verbs and the embedded main verb; cf. Section 7.4 for a discussion of a number of exceptional cases. However, the percentage signs indicate that permeation of the cluster is possible in some southern varieties of Dutch, especially in West-Flanders; cf. Sections 5.2.3 and 6.0, and Barbiers (2008: §2). For ease of exposition, in the following we will present the northern intuitions and simply mark permeated verb clusters as unacceptable.

Examples like those in (5) are special not only because they require verb clustering, but also because they exhibit a peculiar behavior in the perfect tense; whereas verbs governed by a perfect auxiliary normally appear as past participles, the non-finite verbs in (5) appear as infinitives when governed by a perfect auxiliary. This so-called infinitivus-pro-participio (IPP) effect is illustrated in (7); note in passing that all three verbs in clause-final position must cluster together.

Section 5.2.2.3 has shown that verb clustering and the IPP-effect go hand in hand, and argued that they can be used as diagnostic properties of structures that exhibit monoclausal behavior, i.e. structures that consist of two separate clauses but behave as if they were a single clause; cf. Table 1.

	monoclausal	biclausal
verb clustering	+	—
infinitivus-pro-participio	+	—

If so, the notions of clause splitting and verb clustering do not have the same extension: the extension of the latter is a subset of the extension of the former. That clause splitting need not involve verb clustering in the technical sense is shown by the examples in (8). Example (8a) first shows that te-infinitival complement clauses can also be split by the finite verb in clause-final position. However, it differs from those in (5) in that the object of the infinitive can also follow the matrix verb in clause-final position, as shown in (8b). Example (8a) also differs from those in (5) in that it does not show the IPP-effect; in the perfect-tense example in (8c) the verb bewerento claim appears in its participial form and cannot appear as an infinitive.

If verb clustering and the IPP-effect go hand in hand, we have to conclude that (8a) does not involve verb clustering. Section 5.2.2.3 therefore analyzed (8a) as a case of remnant extraposition; cf. also Section 4.4.3, sub IV. This means that the infinitival clause is in extraposed position, but its object is extracted from it by leftward movement across the finite verb; the structure of this example is thus as given in (9); cf. Reuland (1981), Den Besten & Rutten (1989), Rutten (1991), Broekhuis et al. (1995), and many others.

Support for the analysis in (9) comes from the fact that if the embedded te-infinitive has two (or more) dependents, they can occur on different sides of the matrix verb beweren, although this option is generally considered to be marked compared to the two alternative orders.

Example (10b) cannot be derived by extraposition of the full te-infinitival, nor by verb clustering, but it can be derived by remnant extraposition, i.e. by extraposition of the clause plus leftward movement of the indirect object Marie. If this analysis is viable, example (10c) can of course be analyzed in a similar way to (10b) by leftward movement of both the indirect and the direct object. The examples in (10) can thus be analyzed as in (11).

Section 5.2.2.3 has also shown that some cases of splitting of te-infinitival clauses do involve verb clustering in the technical sense. Consider the primeless examples in (12), which at first glance suggest that proberen is just like beweren.

However, when we consider the perfect-tense counterparts of these two examples in (13), we see that they behave differently with respect to the IPP-effect: whereas the verb proberen can appear either as a participle or as an infinitive in the split pattern, it must appear as a participle in the non-split pattern.

If verb clustering and the IPP-effect are indeed two sides of the same coin, we should conclude that (12a) is structurally ambiguous: it involves remnant extraposition when proberen appears as a participle in the corresponding perfect-tense construction in (13a), but verb clustering when it appears as an infinitive. This conclusion is supported by a consideration of cases where the infinitive has two or more dependents. The primeless examples in (14) show that the IPP-effect can only occur when all dependents precede the finite verb in clause-final position.

The examples in (15) show that all examples in (14) become acceptable if we replace the infinitive proberen by the participle geprobeerd, but then we are no longer dealing with verb clustering but with (remnant) extraposition. Note that, as in (10), placing the dependents of the infinitive on different sides of the matrix verb (here: proberen) is generally considered to be marked, compared to the alternative orders.

The discussion above has shown that there are two tests that can be applied to determine whether we are dealing with verb clustering or (remnant) extraposition: verb-clustering constructions (i) exhibit the IPP-effect, but (ii) do not allow permeation of the verb sequence by the dependents (e.g. arguments or modifiers) of the embedded main verb (at least in the northern varieties of Dutch). Sections 7.2 and 7.3 will only consider cases that fit this description.