Main and non-main verbs play different semantic roles in the clause. The former function semantically as n-place predicates and are therefore typically the (semantic and syntactic) head of a clause; sentences containing two main verbs are thus usually biclausal. The fact that the addition of a non-main verb to a clause such as (80a) does not affect the number of arguments that can be expressed is usually taken as evidence that non-main verbs are not predicates. Instead, they are assumed to add e.g. temporal, aspectual or modal information to the meaning expressed by the main verb.

Let us assume for the moment that non-main verbs must combine with another main verb in a structure exhibiting monoclausal behavior, but that main verbs cannot, and that we can test this for infinitival constructions by assuming that mono and biclausal structures differ systematically in terms of verb clustering and the infinitivus-pro-participio (IPP) effect, as indicated in Table 2.

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

The examples in (81) illustrate the monoclausal properties of structures containing the semi-aspectual verb zitten. First, (81a) shows that the semi-aspectual verb and the main verb lezen form a verb cluster, as a result of which the infinitival verb zitten is separated from its direct object het boekthe book. Second, (81b) shows that the IPP-effect is obligatory.

We should note again that verb clustering is somewhat obscured in the varieties of Dutch spoken in Belgium, since these allow permeation of the verb cluster by various elements; for example, the order in (81a) marked by a percentage sign is acceptable in some of these varieties. Note also that passive constructions are exempt from the IPP-effect; we will ignore this here, but return to it in Section 6.2.2.

The examples in (82) illustrate the biclausal properties of structures containing the main verb bewerento claim: (82a) shows that the object het boekthe book of the verb lezento read can intervene between beweren and lezento read, and (82b) shows that the IPP-effect does not occur.

Now consider example (83a), in which the verb proberento try semantically functions as a two-place predicate with an agentive subject and an infinitival direct object clause. That we are dealing with a regular direct object clause is clear from the fact, illustrated in (83b), that the infinitival clause can be pronominalized or replaced by a referential noun phrase.

Example (83a) also shows that the infinitival complement of proberen can be either an om + te-infinitival or a te-infinitival without the complementizer om. We will see shortly that these infinitival complements behave somewhat differently, but first the examples in (84) show that the two types of infinitival clause can be placed after the verb proberen in clause-final position, and that proberen must occur as a past participle in the corresponding perfect-tense construction. This is fully consistent with the earlier claim that proberen is a main verb.

However, the examples in (85) show that the te-infinitival without om is special in that it is also compatible with the IPP-effect, provided that the infinitival clause is split, i.e. that the object of the infinitival verb lezen precedes proberen: the word order in (85b) is unacceptable.

This shows that proberen can also trigger monoclausal behavior, from which we can conclude that it does not always behave like an ordinary main verb, but can be of a hybrid nature in the sense that it can also exhibit properties of non-main verbs. The fact that proberen is not an isolated case, and that there are more unsuspected main verbs that can enter a verbal complex and thus trigger monoclausal behavior, strongly suggests that having this option is not a defining property of non-main verbs. This is confirmed by the fact that constructions with bare infinitivals always exhibit monoclausal behavior, regardless of whether the selecting verb is a main or a non-main verb: this is illustrated in (86) for the aspectual verb gaan and the main verb horento hear.

All this implies that the hypothesis that main verbs differ from non-main verbs in that they cannot combine with another main verb in a structure that exhibits monoclausal behavior is refuted, and that we must therefore look for other means of distinguishing main from non-main verbs.

This subsection examines two other syntactic properties that seem to be related to the predicative nature of main verbs, as opposed to the non-predicative nature of non-main verbs. The predicative nature of main verbs like bewerento claim and proberento try is clear from the fact that they do not require a projection of a main verb as their complement; the primed examples in (87), in which the italicized infinitival clauses of the primeless examples are pronominalized or replaced by a noun phrase, clearly show that we are dealing with two-place predicates, i.e. regular transitive main verbs.

Non-main verbs such as the aspectual verb gaan in the (a)-examples in (88), on the other hand, are clearly not predicative, as is clear from the fact that they normally do not allow pronominalization of the projection of the infinitival main verb: the verb gaan is not able to license the subject and the object pronoun, which clearly shows that it does not behave like a transitive verb. A possible problem, however, is that the (b)-examples show that modal verbs exhibit unexpected behavior in this respect; (88b') shows that pronominalization is possible (cf. Section 4.5, sub II, where the same point was made).

Another possible problem is that we expect that main verbs always allow pronominalization of their infinitival complement. This is borne out in the case of the perception verb ziento see in the (b)-examples in (89): the primed example clearly shows that zien functions as a regular transitive verb with a nominal subject and an object clause. However, the seemingly similar construction with the causative verb laten in the (c)-examples does not allow the pronominalization of its infinitival complement. If the (b) and (c)-cases are indeed the same, this would be surprising. We leave this problem aside, while noting that Section 5.2.3.4 will show that there is indeed a non-trivial difference between constructions with perception verbs such as zien and the causative verb laten, which perhaps provides a first handle to explain the unacceptability of (89c').

We have seen that there are two ways to determine whether a verb that combines with an infinitival verb is predicative in nature. The simplest way is to examine whether it is able to introduce an argument that is not licensed by the embedded main verb; if this is the case, the matrix verb clearly has an argument structure of its own. The second way is to investigate whether the projection of the infinitival verb can be pronominalized; if so, we can conclude that the pronoun must be semantically licensed and therefore functions as an argument of the verb. Table 3 provides the results of these tests for the verbs in (88) and (89).

verb type	additional argument	pronominalization	example
aspectual	—	—	(88a)
modal	—	+	(88b)
perception	+	+	(89a)
causative	+	—	(89b)

Assuming that the distinction between main and non-main verbs is really determined by the question of whether the verb is predicative in nature, we have to conclude that, of the four verb types discussed here, only the aspectual verbs can be considered non-main verbs. This implies that the dividing line between these two sets will be slightly different than what is usually assumed in more traditional grammars. For example, while modal verbs are usually considered non-main verbs, we must conclude that they are main verbs, as proposed in Klooster (1984/1986).

For completeness’ sake, we conclude by noting that the pronominalization test must be applied with care; not all structures containing the pronoun dat/het can be used to show that the verb under consideration is predicative in nature. There seem to be two complications. First, the examples in (90) show that secondary predicates can also be pronominalized by the pronoun dat; the intended interpretation of the pronoun is indicated by coindexing. The acceptability of the second conjunct in these examples does not show that the copular verb zijn is a two-place predicate; as Section 2.2 has shown, it is simply a verb taking a predicative small-clause complement.

Second, the examples in (91) show that left dislocation constructions should also be set aside. The fact illustrated in (91a) that the pronoun dat can be used to refer to the left-dislocated participle phrase does not show that the perfect auxiliary hebben is a two-place predicate. In fact, if we were to take example (91a) as evidence for the assumption that the perfect auxiliary hebben is a two-place predicate, we would be forced to conclude on the basis of examples like (91b&c) that it can also be a three-place or even a four-place predicate, a conclusion that is clearly untenable.

We normally use the term complement as equivalent to the term internal argument; it refers, for example, to arguments of verbs that are assigned a thematic role like goal or theme. Since Section 4.6 argued that main and non-main verbs differ in that only the former are predicative, while the latter are not able to select arguments, we could restrict the term verbal complement such that it refers only to verbal arguments of main verbs. However, we will adopt a somewhat looser notion of verbal complement, which also includes the verbal projections in the domain of non-main verbs. The main reason for this is that we have seen that non-main verbs impose certain morphosyntactic selection restrictions on the main verb: perfect auxiliaries, for example, must combine with past participles, aspectual verbs only combine with bare infinitivals, and semi-aspectual verbs usually combine with te-infinitivals. By stating that non-main verbs select the projection of the main verb as their complement, these selection restrictions can be accounted for.

By discussing verbal complements of main and non-main verbs in similar terms, it will also be easier to compare the behavior of such verbal complements. That this is desirable is clear from the fact that Subsection II has shown that besides obvious cases of main and non-main verbs, there are also verbs of a more hybrid nature; we will see numerous other cases in Sections 5.2.2 and 5.2.3.

This section has discussed a number of properties of main and non-main verbs. Main verbs function semantically as n-place predicates and are therefore typically the (semantic and syntactic) head of a clause; if the sentence contains two main verbs, they are prototypically expressed in a biclausal structure. Non-main verbs, on the other hand, are not predicates, but provide additional information to the meaning expressed by the main verb. As a result, non-main verbs must combine with a main clause in a verbal complex and thus trigger monoclausal behavior; they exhibit the two properties in Table 4, repeated from Subsection I.

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

However, it is not always easy to determine whether we are dealing with a main or a non-main verb, since some verbs exhibit a somewhat hybrid behavior. Subsection II was devoted to the question of how we can distinguish main from non-main verbs. We argued that it is not sufficient to show that a verb enters a verbal complex with an infinitival main verb to conclude that we are dealing with non-main verbs, because main verbs such as proberento try also have this property. Therefore, we decided that we need to examine the predicative nature of the verb in question: if the addition of this verb results in the addition of an argument that is not licensed by the non-finite main verb, or if the projection of the non-finite main verb can be pronominalized, then we are dealing with a main verb. This leads to a somewhat different classification than is usually assumed in descriptive grammars. We have illustrated this for modal verbs such as willen, which are normally classified as non-main verbs, but must be classified as main verbs according to our criterion. Section 5.2 will show that this also holds for a number of other verb types.