At first glance. it seems relatively easy to determine whether a communication verb selects a declarative or an interrogative clause. The former are selected by verbs like zeggento say and aankondigento announce, which are used in the (a)-examples in (34) to report something that has been said/announced, while the latter are selected by ditransitive verbs like vragento ask and smekento beg, which are used in the (b)-examples to report something that has been asked/requested. In short, the choice between declarative and interrogative clauses is determined by the speech act reported by the speaker.

Closer inspection reveals that the situation is more complex than this. The above only applies to cases of indirect reported speech; in other contexts verbs like zeggen and aankondigen can also select interrogative clauses, as shown by the two (b)-examples in (35). The choice between the three examples depends on the speaker’s knowledge. Example (35a) is used when the speaker knows that there will be a reorganization, but does not know whether Marie has made this public. Example (35b) would be used when the speaker does not know for sure that there will be a reorganization, and (35b') would be used when the speaker knows that there will be a reorganization but does not know when it will take place.

The examples in (36) show that the speaker has a similar choice when the sentence is negated. The choice between the three utterances again depends on the speaker’s knowledge. Example (36a) can be used to express that the speaker knows that there will be a reorganization, but that Marie did not make this public, or to express that the speaker expected that Marie would announce a reorganization, but that this expectation was not fulfilled. Example (36b) is typically used when the speaker does not know for sure that there will be a reorganization, and (36b') expresses that, while the speaker is convinced that there will be a reorganization, Marie has not given more specific information about when it will take place.

Not all communication verbs are compatible with an interrogative argument clause when they occur in an interrogative or negative sentence. For instance, the examples in (37) show that the verb aankondigento announce in (37a') does not easily allow it, which is probably due to the fact that it is factive in the sense discussed in Section 5.1.2.3. Note also that there is a contrast in acceptability between embedded yes/no and wh-questions and that the latter do occasionally occur on the internet.

For completeness’ sake, the examples in (38) show that verbs such as vragen can sometimes also be used with declarative argument clauses, in which case we are dealing with a request/demand rather than a question. The two meanings can be easily distinguished: vragen with the meaning “to ask” takes a nominal object that alternates with an aan-PP, while vragen with the meaning “to request/demand” prefers a van-PP and allows a nominal object only in formal/archaic contexts.

The examples in (39) show that the perception verbs proevento taste, ruikento smell and voelento feel can select either a declarative or an interrogative clause. The meaning of the verbs in the primed and the primeless examples differs in that in the former case the subject of the perception verb senses involuntarily (in the sense of “without conscious control”) that the state of affairs expressed by the embedded clause is true (Yuk, the soup is spoiled!), whereas in the primed examples the subject intentionally uses his/her senses to determine whether the state of affairs expressed by the embedded clause is true (No, don’t worry, the soup is still fine!).

It seems that we are not dealing with two uses of the same verb, but with true polysemy. The reason for this is that in the domain of vision and hearing there are two specialized verbs for the two meanings: ziento see and horento hear are used for involuntary perception, whereas kijkento look and luisterento listen are used for the active involvement of vision and hearing.

That the distinction between involuntary and voluntary perception is also relevant for the polysemous verbs proeven, ruiken and voelen is clear from the fact that imperatives, which imply voluntary activity, require an embedded question for these verbs.

The contrast between involuntary and voluntary perception seems quite sharp when the argument clause is introduced by the complementizer of, but is more diffuse when it is introduced by a wh-phrase. The examples in (42) seem to allow for both readings: example (42c), for instance, does not require that Jan purposefully feels how wet the laundry was, but that he might accidentally notice it while putting the laundry in the closet.

That verbs of involuntary perception are compatible with embedded wh-questions is also shown by the acceptability of the examples in (43), which contrast sharply with the primeless examples in (40).

A caveat is needed here, however, since interrogative argument clauses are generally possible with the verbs ziento see and horento hear when we are dealing with indirect perception, as shown in (44). This means that examples such as (43) can only be used to distinguish between verbs of voluntary and involuntary perception if we are dealing with direct perception and not with indirect perception (e.g. on the basis of empty spaces on book shelves or the sound of footsteps).

In addition, the examples in (45) show that ziento see and horento hear are also fully compatible with embedded yes/no questions in interrogative or negative sentences; in this respect they behave just like non-factive communication verbs such as zeggento say discussed in Subsection I.

Finally, the examples in (46) show that the addition of a deontic modal verb can have a similar effect on the selection restrictions.

We will return to verbs of involuntary and voluntary perception in Section 5.2.3.3 where we show that they differ in yet another way: the former but not the latter can occur in AcI-constructions: Jan zag/*keek de zon opkomenJan saw the sun rise.

Cognition verbs can be divided into the four groups in (47) based on whether they select a declarative or an interrogative clause.

The verb zich afvragento wonder in (47a) cannot be combined with a declarative argument clause; it only occurs with interrogative clauses introduced by the complementizer of or some wh-phrase.

The verbs in (47b) take a declarative object clause introduced by the complementizer datthat: interrogative clauses lead to degraded results. This is illustrated in (50) for the verb gelovento believe.

The situation is less clear for the verbs in (47c). The examples in (50) show that the verb begrijpento understand cannot take an interrogative verb introduced by the complementizer ofwhether, but that interrogative clauses introduced by a wh-phrase give a much better result; although example (50b) is definitely marked without the anticipatory pronoun hetit, example (50c) is perfectly acceptable. The verbs vermoedento suspect and zich realiserento realize show a similar behavior.

The examples in (51) show that verbs of the type gelovento believe and the type begrijpento understand also behave differently when they function as the head of an interrogative or negative sentence: while the former remain incompatible with interrogative argument clauses, the latter readily accept them.

Note that example (51b) with a declarative clause is normally used when the speaker wants to check whether Jan did get the relevant information that Marie is coming, whereas the use of an interrogative clause suggests that the speaker himself does not know whether/when Marie is coming and would in fact like to have more information about it (which might be available to Jan). Similarly, example (51b') with a declarative clause expresses that Jan does not understand the established fact that Marie is coming, whereas (the time of) Marie’s coming is left open when begrijpen takes an interrogative argument clause.

The verbs in group (47d) seem to be compatible with both declarative and interrogative argument clauses. We illustrate this in (52) for the verb weten. Example (52a) is used to express that Jan is aware of the fact that Marie is unable to come, and the examples in (52b&c) are used to express that Jan is capable of providing further information about whether/when Marie is able to come.

Verbs of investigation and discovery can differ in whether they select a declarative or an interrogative clause. The former seems to be the case for aantonento show, bewijzento prove, suggererento suggest and ontdekkento discover, and the latter for nagaanto examine and onderzoekento investigate. The former verbs are used especially when the proposition expressed by the argument clause refers to an established fact, and the latter when the argument clause refers to an open question.

Question formation, negation as well as the addition of a modal verb may change the selection restriction of verbs like aantonen/bewijzento prove, as is clear from the fact that the examples in (54) seem acceptable with embedded yes/no questions; in this respect such verbs behave just like the verbs of involuntary perception. The selection restrictions of nagaanto examine remain unchanged in such contexts.

Verbs of wishing like hopento hope, wensento wish, and willento want are only compatible with declarative argument clauses, regardless of whether the sentence is declarative, interrogative, or negative, as shown in (55) for the verb hopen.

The primeless examples in (56) show that verbs with an experiencer subject like betreurento regret or waarderento appreciate take declarative object clauses; interrogative clauses are excluded. The primed examples show that interrogative object clauses are also excluded when the sentence is interrogative or negative. For the sake of the discussion in Section 5.1.2.3, it should be noted that the object clauses in the primeless examples are introduced by the anticipatory pronoun hetit.

Haeseryn et al. (1997:1155) has claimed that subject-experiencer verbs such as betreuren can also take an object clause introduced by the conditional complementizer alsif; some possible cases are given in (57). However, Section 5.1.2.2 will show that there are reasons to reject this claim.

Finite object clauses occur not only as internal arguments of verbs, but also as logical subjects of complementives, i.e. in vinden and resultative constructions. The examples in (58) show that clause-final object clauses in vinden-constructions are usually introduced by the anticipatory pronoun het; omitting the pronoun results leads to a degraded result. However, the pronoun is optional when the complementive is topicalized, and even excluded when the object clause is topicalized; cf. Section 5.1.2.2 for a discussion of topicalization of object clauses, and Section 5.1.3 for a discussion of similar behavior of subject clauses.

The primeless examples in (59) show that, unlike what we found in the vinden-construction, the anticipatory pronoun can easily be omitted in resultative constructions. The primed examples show that vinden and resultative constructions behave in a similar way when the complementive or the object clause is topicalized.

The previous subsections have shown that the choice between declarative and interrogative object clauses is not simply a matter of lexical selection restrictions imposed by the matrix verb. In particular, it has been shown that question formation and negation can license interrogative object clauses with some of the matrix verbs that take declarative object clauses in positive declarative clauses.