Generally speaking, subject clauses do not occur with intransitive and transitive verbs. The reason is that such verbs normally take an external argument with the function of agent. Given that clauses refer to propositions/questions/etc., and not to agentive entities, it is expected on semantic grounds that subject clauses cannot occur with such verbs. The examples in (216) show that using subject clauses with (in)transitive verbs indeed gives rise to complete gibberish.

There are potential counterexamples to the claim that transitive verbs do not take subject clauses. Example (217a), for instance, shows that the transitive verb bewijzen'to prove' can easily be combined with a clausal subject. Such cases are special, however, in that they involve factive clauses, that is, clauses the truth of which is presupposed by the speaker. Section 5.1.2.3 has shown that normally such clauses can be paraphrased by means of a noun phrase het feit dat ...'the fact that ...', as in (217b), and that they exhibit a number of nominal properties.

      Subject clauses are possible if they are internal arguments of the verb, as is clear from the fact that a transitive sentence such as (218a) is easy to passivize. The (b)-examples show that the passive counterpart of this sentence may either contain the expletive er or the anticipatory pronoun het: this is a reflex of the fact that the anticipatory pronoun is optional in (218a).

Since the examples in (218) show that DO-subjects may be clausal, it should not come as a surprise that we also find subject clauses with unaccusative verbs. The examples in (219) show that this is quite common with nom-dat verbs; cf. Section 2.1.3. We illustrate this in the (a)-examples by means of a nom-dat verb that takes zijn in the perfect tense, and in the (b)-examples by means of a nom-dat verb that takes hebben.

Subject clauses are also common with psychological predicates that take an object experiencer; cf. Section 2.5.1.3. This holds both for (220a) with the psych-verb ergeren'to annoy' and for (220b) with the periphrastic expression kwaad maken'to make angry'.

Note in passing that psych-verbs such as ergeren'to annoy' and many nom-dat verbs are object experiencer verbs; consequently, they can be combined successfully with conditional als-clauses; see the examples in (221). Since Section 5.1.2.1, sub VI, has shown on the basis of similar examples with subject experiencer verbs that such als-clauses are not arguments of the verb, we need not elaborate on this here; as illustrated in the primed examples, the fact that preposed als-phrases can be followed by the resumptive element dan'then' suggests that we are dealing with conditional adverbial clauses.

A conclusive argument for assuming that the als-phrases in (221) are not subject clauses is that the subject pronoun dat in the primed examples cannot be dropped when they occupy the sentence-initial position (that is, when dan'then' is not present). The examples in (222) show that this is compulsory when run-of-the-mill subject clauses introduced by the complementizer dat'that' occupy the initial position, for the simple reason that inclusion of the pronoun dat leads to a clause with two subjects.

      Subject clauses are also very common if they function as the subject of copular constructions, as in (223a). This is expected because such subjects are not the external arguments of the copular, for the same reason that the direct object in the vinden-construction in (223b) is not an internal argument of vinden. In these two cases we are dealing with subjects of the complementive; cf. Section 2.2.2.

The copular constructions in (224) show that the adjective bekend may take either a declarative or an interrogative subject clause. The former is always possible, but the latter only occurs if the matrix clause is negative and/or interrogative. The complementizer of is used in the (b)-examples if the relevant decision has not been made public yet, the complementizer dat if the decision has been made public but has (not yet) reached the intended public.

Again, it should be noted that we occasionally encounter als-clauses that can easily be misanalyzed as subject clauses. That we are not dealing with subject clauses here is clear from the fact, illustrated in (225), that such als-clauses differ from run-of-the-mill subject clauses introduced by the complementizer dat'that' in that a subject pronoun must be present if the als-clause occupies the sentence-initial position; we must therefore be dealing with conditional clauses.

Finally, we want to point out subject clauses are possible with epistemic modal verbs; we will return to this in Section 5.2.2.2 and 5.2.3.2.

Subject clauses normally follow the verbs in clause-final position, as shown by the primed examples in (219), which are repeated here for convenience as (227).

Subject clauses may also occur in sentence-initial position, in which case they are optionally followed by the resumptive demonstrative pronoun dat'that'.

The examples in (229) show that it is not possible to have the subject clause in the middle field of the clause; see De Haan (1974) and Koster (1978). The main clauses in the primeless examples have a non-subject in sentence-initial position and the subject clauses of (227) and (228) in the middle field; the primed examples provide the corresponding embedded clauses. Such examples are deemed ungrammatical.

We should note, however, that the examples seem at least marginally acceptable if the clause is interpreted as factive: (het feit) dat Jan erover klaagde. If this is the case, it would not be surprising, considering that Section 5.1.2.3 has shown that factive clauses are more generally used in nominal argument positions. Example (230) provides instances in which the subject clause is more clearly factive, and we believe that these cases are indeed possible (provided that the clause does not become too lengthy).

      Koster (1978) concludes from the fact that subject clauses cannot occur in the regular subject position in the middle field of the clause that subject sentences do not exist. He also proposes that the clauses in (228) are not sentence-internal, but function as sentence-external satellites that bind a (possibly phonetically empty) subject pronoun; actually, according to Koster, we are dealing with a kind of left-dislocation constructions. If we assume that pronouns are moved from the regular subject position into sentence-initial position, examples such as (228a) are analyzed as in (231a) if the demonstrative pronoun is present, and as in (231b) if it is not.

Koster's proposal was challenged in Klein (1979). An important reason is that the prosody of the examples with and without the resumptive pronoun dat differ markedly: while in the former case the clause is normally separated from the sentence by an intonation break, the clause can be prosodically integrated in the sentence in the latter case, as indicated in (232), in which the comma indicates the obligatory intonation break.

If Klein's conclusion that the clause in (232b) is sentence-internal is correct, we should account for the fact that the clause cannot occur in the regular subject position in the examples in (229) by claiming that clauses cannot surface in nominal argument positions. This is in fact the same conclusion drawn for object clauses in Section 5.1.2.2, sub III, to which we refer the reader for further discussion. We will investigate the pros and cons of Koster's proposal in our discussion of topicalization in Section 11.3.2.

Like object clauses, subject clauses cannot be preposed in sentences that contain the anticipatory pronoun het, as shown in (233b). This would follow immediately from Koster's left-dislocation analysis as the object pronoun must be replaced by the resumptive pronoun dat or its phonetically empty counterpart pro. The structures in (231) show that the position of het in (233b) is already occupied by a trace.

The analysis must be slightly different if we accept Klein's conclusion that the subject clause occupies the sentence-initial position if the demonstrative pronoun dat is not present. We then have to assume that the subject clause has not been moved into clause-initial position in one fell swoop but has moved via the regular subject position; the anticipatory pronoun is then blocked given that the subject position is occupied by a trace of the clause. See Section 5.1.2.2, sub III, for a more extensive discussion of this option.
      The (b)-examples in (234) show that subject clauses cannot be preposed in clauses that contain expletive er either; er can only be interpreted as an adverbial phrase of place in these examples. The reason for this is different, however, than in the case of het; expletive er can only be used if the subject is part of the focus (new information) of the clause, whereas preposed subject clauses are normally interpreted as being part of the presupposition of the clause.

      The option of having the anticipatory pronoun het or the expletive er is not only affected by the position of the subject clause. In examples with a complementive, the position of the secondary predicate may also be relevant. With a sentence-initial predicate het is preferably dropped and er becomes completely impossible.

The examples in (236) show that we may find the same phenomenon in perfect-tense constructions with monadic unaccusative verbs taking subject clauses like blijken'to turn out': with topicalized participles, het and er cannot be properly realized. Examples with het and er do occur on the internet but are very rare.

Although we are not aware of any theoretical account for the markedness of the primeless (b)-examples in (235) and (236), we hypothesize that examples of this type involve some kind of locative inversion of the type we find in English. Den Dikken and Næss (1993) have argued that in examples like Down the hill rolled a baby carriage the predicative PP down the hill has been topicalized via the regular subject position, and that the subject occupies its base position in the small clause headed by the moved predicate; [CP Down the hilli [TP ti rolled [SC the baby carriage ti]]]. If we assume something similar for examples such as (235b), insertion of the pronoun het may be blocked given that the regular subject position is occupied by a trace of the moved predicate. A potential problem for this analysis is that this leaves unexplained why insertion of het seems to be marginally possible. Furthermore, it remains to be seen whether an analysis of this sort can be extended to examples such as (236b). The degraded status of the primed (b)-examples can again be related to the information structure of the clause if left dislocation/topicalization of the predicate is only possible if it is part of the presupposition of the clause. We leave it to future research to investigate whether proposals along these lines are viable.
      Example (218) in Subsection I has shown that in passive constructions the choice between het and er is related to the question as to whether the object clause in the corresponding passive construction can be combined with the anticipatory pronoun het. It seems that, as in English, clause-final subject clauses in active sentences can always be introduced by anticipatory het and that in many cases they can also be combined with expletive er. The semantic difference between the two options is not always clear, but we may suppose that the choice between the two options depends on whether the subject clause is presented as part of the presupposition or the focus of the sentence. A Google search (1/27/2012) shows that the frequencies of het and er in examples such as (237) are both high.

An appeal to the information structure of the sentence seems supported by examples like those in (238). Given that interrogative clauses are less likely to be interpreted as presuppositional than declarative clauses, we expect examples such as (238a) to be extremely rare (despite being definitely grammatical). A Google search (3/22/2013) on this string shows that this expectation is indeed borne out: it resulted in no more than 4 hits. Strings such as (238b), on the other hand, are very frequent.

Given the result of our Google searches mentioned above, one would also expect the frequency of examples such as (239a) to be much lower than examples such as (239b). This expectation is, however, not borne out: we found 225 cases of the two strings in (239a) and only 13 of the two strings in (239b).

The results of our Google searches on the examples in (239) show that there must be other factors, yet to be identified, that must be involved in the choice between het and er. One factor that springs to mind is that the choice is related to the type of predicate, but we leave this for future research.