The dependent clauses in (225) and (226) are optional. Since logical subjects are usually obligatory, it is often assumed that, syntactically speaking, these clauses are not the real logical subjects of the adjective; they are interpreted as the subject by virtue of their relation to the anticipatory pronoun hetit, which functions as the syntactic subject of the adjective. The relation between the pronoun and the clause is expressed by coindexation, as in (227).

The anticipatory pronoun acts as a kind of “placeholder” for the subject clause, which is normally placed at the right edge of the matrix clause. However, this placeholder must be omitted when the subject clause is placed in clause-initial position, as in (228). This provides additional evidence for the claim that the clauses in (227) are the logical subjects of the adjective.

The judgments on the (a)-examples in (228) would follow immediately if we were to assume that the clauses are located in the subject position of the sentence; this would obviate the need for the anticipatory pronoun het as a placeholder for the subject clause. However, there are reasons to believe that they are located in the sentence-initial position, which can also host wh-phrases and topicalized elements. The first reason is illustrated in (229): subject clauses cannot follow the finite verb in yes/no questions, which would be the rule in the case of nominal subjects; cf. Is Jan leuk?Is Jan nice?.

The second reason is that preposing of the clause in embedded clause is not a favored option for finite subject clauses and is downright impossible for infinitival subject clauses, as in (230). If preposing of subject clauses targets the canonical subject position, this would not be expected.

We find essentially the same for object clauses; movement of such clauses into the regular object position (i.e. the position of het) usually leads to an unacceptable/marked result. We have illustrated this in (231) for the vinden-construction in the (b)-examples in (227); note that we have ignored here that the results improve when pronounced with a marked intonation contour that triggers a factive reading; cf. Section V5.1.2.3 for discussion.

Although the examples in (229) to (231) show that the clausal subjects in (228) clearly do not occupy the same position as the anticipatory pronouns in (227), the fact that the anticipatory pronoun het cannot be used in (228) suggests that topicalization of the subject clauses does not take place in one fell swoop, but proceeds via the position occupied by the anticipatory pronoun het in (227); if so, this position is occupied by a trace of the clause, and consequently the insertion of the “placeholder” cannot take place. We refer the reader to Den Dikken and Næss (1993) for arguments, based on English and Norwegian locative inversion constructions, in favor of the claim that topicalization of clauses must proceed via the regular argument (i.e. subject/object) positions of the clause.

When the adjective is preposed, as in (232), the anticipatory pronoun is optionally present, although the two cases differ slightly in intonation and meaning. If the anticipatory pronoun is present, it is followed by a short intonation break and the sentence simply expresses that the state of affairs to which the clausal subject refers can be characterized by the adjective leuknice. If the anticipatory pronoun is absent, on the other hand, there is no intonation break and the sentence expresses that from among the things under discussion the state of affairs expressed by the subject clause can be characterized as leuknice; the sentence is contrastive, as is clear from the fact that in this case the adjective must be assigned contrastive accent.

The examples in (233) show that the adjective and the clausal subject cannot be preposed as a whole; we added the adverbial zekercertainly to make the resulting sentence more phonologically balanced. The unacceptability of the examples in (233) suggests that the adjective and the clausal subject do not form a constituent.

There is one exception to this general rule: adjectives that take an interrogative clause as their logical subject do allow this kind of topicalization when they are negated. This will be discussed in Subsection II.

That the adjective and its clausal subject do not form a constituent is also supported by the fact that in embedded clauses the clausal subject is not adjacent to the adjective but must follow the verb(s) in clause-final position (if present). This is illustrated in (234).

Consider again the examples in (225a) and (235b), repeated in a slightly different form in (239), in which the anticipatory pronoun het acts as the nominative subject of the sentence.

There is a striking difference between such cases when the anticipatory pronoun het follows the finite verb in second position, as in the primeless examples in (240); the pronoun het can be omitted if the adjective is duidelijk, but not if the adjective is leuk. A similar difference can be observed in the primed examples, where the anticipatory pronoun follows the complementizer dat of an embedded clause.

The examples in (241) show that in vinden-constructions the anticipatory object pronoun het is usually obligatory, regardless of which adjective is used.

Nevertheless, example (242) shows that in writing and formal speech the anticipatory object pronoun can be dropped when the verb achten is used.

Although it is not obvious what causes the difference in the distribution of het, it can be taken as evidence that the constructions with leuk and duidelijk are also different from each other, so it can be taken as a first piece of evidence for the hypothesis in (238) that the subject clauses have a different syntactic status as internal/external argument in the two constructions.

Another difference between the constructions with leuk and duidelijk manifests itself when the adjectives are negated: the declarative subject clause can then be replaced by a dependent interrogative clause in the case of duidelijk, but not in the case of leuk. This is illustrated by the acceptability contrast between the (a) and (b)-examples in (243). For completeness, note that the (b)-examples are acceptable regardless of whether the negation is expressed syntactically by the negative adverb nietnot or morphologically by the negative prefix on-.

Example (244) shows that adjectives morphologically derived from verbs that select an interrogative complement clause, such as twijfelachtiguncertain (derived from twijfelento doubt), also select a dependent interrogative complement.

Because dependent interrogative clauses typically occur as complements of specific verbs, it is usually assumed that they are internal arguments of the verb: for the same reason, the interrogative subject clauses in (243b&b') should be considered complements (i.e. internal arguments) of (on)duidelijk; cf. the hypothesis in (238b). The next two subsections provide further arguments for this conclusion.

A final difference between leuk and duidelijk is that only the latter can be used in a resultative construction. This may not be related to the hypothesis in (238), but rather to the fact that duidelijk optionally selects a dative argument: an adjective like evidentobvious, which is probably of the same type as duidelijk (see Table 2 below) but does not select an additional argument, cannot enter the resultative construction either (alth.

As for consider the class of adjectives that can take a finite clause as their logical subject, it is not always easy to determine which type they belong to. This is mainly because adjectives that allow the omission of the anticipatory pronoun in constructions like (240) do not always allow an interrogative clause in negative contexts. Furthermore, the results of the wh-extraction test are not always clear, because many speakers do not allow wh-extraction from the subject clause easily anyway. Table 2 provides the results for a small sample of adjectives, where pronoun-drop indicates whether the anticipatory pronoun het can be dropped in the relevant contexts, interrogative indicates whether the finite clause can be an interrogative clause in negative contexts, and wh-extraction indicates whether wh-extraction is possible in the absence of an anticipatory pronoun.

The adjectives in Table 2 can all take either a noun phrase or a finite clause as their logical subject, but there are also some cases in which a clausal subject is preferred, with the proviso that subjects of the nominal type are restricted to the personal pronoun het or the neuter demonstratives dit/datthis/that, which can refer to propositions, and (often marginally) to deverbal nouns such as vertrekleaving. Some examples are the evaluative adjectives jammer/spijtigunfortunate in (249) and the modal adjectives mogelijkpossible and zekercertain in (250).

For completeness’ sake, note that the examples in (251) with the time adverb nunow seem acceptable, which may be due to fact that the adverb allows the indefinite noun phrase to be interpreted as an eventuality, i.e. “the occurrence of a bomb attack”.

The inf-nominalizations in (252) can also be used as subjects of these adjectives, probably for the same reason.

Obligatory control adjectives select a van or voor-PP, and the nominal [+animate] complement of the PP controls the implied subject of the infinitival clause. A small sample is given in (256).

Examples of adjectives taking a van-PP are given in (253a) and (257): the adjective attributes a property to (the behavior of) the referent of the nominal complement of van. The van-PP may be omitted, in which case an arbitrary interpretation results along the lines indicated in (254).

Examples of adjectives taking a voor-PP are given in (253b) and (258). The referent of the nominal complement of voor acts as an “experiencer”: example (258) expresses that Jan has (i.e. experiences) problems in admitting mistakes. If the voor-PP is not expressed overtly, PRO again gets an arbitrary interpretation.

Optional control adjectives select a voor-PP whose nominal complement optionally controls the implied PRO-subject of the infinitival clause. Two subcases should be distinguished: adjectives that select a PP with a [+animate] nominal complement, and adjectives with a PP whose nominal complement can be either [+animate] or [-animate]. Examples of the two kinds are given in (259).

The PP-complements of the adjectives in (259a) do not have to control the implied subject of the infinitival clause, as is illustrated by the examples in (260). Example (260a) is ambiguous between at least two readings: either it can be that Jan takes the book (to someone), or it can be that someone else brings the book to Jan. The latter reading can be enforced by adding the indirect object hem to the infinitival clause, as in (260b): if this pronoun is interpreted as coreferential with Jan, the implied subject PRO must be construed as disjoint in reference from Jan, since otherwise it would be illicitly bound by it; cf. Section N22.1 for a discussion of the binding conditions on pronouns. As before, coindexing indicates coreference.

Similar observations can be made with respect to the adjectives in (259b): (261a) is ambiguous between a reading in which it is Jan himself who discloses the secret and a reading in which the disclosure is done by some other person. As in (260), the latter reading can be enforced by adding a pronoun to the infinitival clause that is interpreted as coreferential with Jan, as in (261b).

At first glance, control readings are difficult to obtain when the PP takes a [-animate] complement: for instance, example (262a) does not allow a control reading. This does not mean that control is excluded; the impossibility of the control reading is due to the fact that the noun phrase het milieuthe environment is simply not a suitable subject for the predicate vuilnis stortento dispose of waste. If the [-animate] PP-complement is a suitable subject for the infinitival predicate, as in the passive construction in (262b), control is perfectly acceptable. Note that the arbitrary reading of PRO is not possible in (262b); cf. Section V4.3 for further discussion.

Optional control adjectives differ from obligatory control adjectives in that the referent of the nominal complement of voor is not an experiencer: the examples in (261) do not express that telling a secret is experienced as dangerous by Jan. Instead, Jan refers to an entity that is potentially affected by the state of affairs expressed by the infinitival clause. This difference also accounts for the acceptability of the examples in (262), since the [-animate] complements of voor would be incompatible with an experiencer reading.

The arbitrary control adjectives do not take a PP-argument (although for some speakers this depends on the context). Consequently, there is no controller and the implied subject of the infinitival clause must be construed arbitrarily. Adjectives that belong to this class are: afkeurenswaardigcondemnable, gebruikelijkcommon, onnodignot needed, overbodigunnecessary, raadzaamadvisable. Some examples are given in (263).

That PRO is construed arbitrarily is clear from the presence of the generic possessive pronoun jeoneʼs in (264), which is interpreted as coreferential with arbitrary PRO.

We conclude this discussion of infinitival clausal subjects by noting that there are also adjectives that can take a finite but not an infinitival clause. This holds for epistemic modal adjectives like waarschijnlijkprobably and zekercertain.

Bennis and Hoekstra (1989a) suggest that the unacceptability of (265b) is due to the fact that these modal adjectives do not select a PP, leaving PRO unidentified. A problem for their proposal is that it incorrectly predicts that arbitrary control adjectives of the types discussed in the previous subsection do not occur at all. The claim that PRO can be given an arbitrary interpretation, on the other hand, raises the question as to why examples such as (265b) are unacceptable. Since we have no insight to offer here, we leave this question to future research.

It has been suggested that the so-called easy-to-please construction in (266b) is derived by NP-movement from the construction in (266a), which we will henceforth conveniently call the het-construction; cf. Chomsky (1973). This means that the relationship between (266a) and (266b) is claimed to be similar to the relationship between the two examples in (267), in which the noun phrase Jan arguably originates in the subject position of the embedded infinitival clause; it replaces the anticipatory pronoun het as a result of its movement into the subject position of the main clause, which is often referred to as subject raising (cf. Section V5.2.2.2).

It has been claimed that example (266b) is derived from (266a) by a similar movement of the noun phrase dit probleem from the object position of the embedded clause into the subject position of the matrix clause, as a result of which it replaces the pronoun het. This raising-to-subject derivation is given in (268b).

The main reason for assuming that the het and the easy-to-please constructions are related by movement is that the examples in (266a&b) seem to be almost synonymous, just like the subject-raising examples in (267). However, Subsection 1 will show that similar constructions also occur with adjectives such as leuknice and that in such cases there are differences in meaning. Subsection 2 will further show that there are also a number of syntactic differences, which suggests that the raising-to-subject approach presented in (268b) cannot be maintained and that we simply have to assume that the subject is base-generated as the logical subject of the adjective; cf. Bennis & Wehrmann (1987) and Chomsky (1995:ch.3). This analysis raises the question as to why the direct object of the infinitival clause cannot be expressed; this question is discussed in Subsection 3. Subsection 4 concludes the discussion of the easy-to-please construction by showing that the adjectives entering the construction always express properties that are subject to subjective evaluation.