• Dutch
  • Frisian
  • Saterfrisian
  • Afrikaans
Show all
7.2.The hierarchical order of verbs in verb clusters
quickinfo

Verbs in a verb cluster are in a selectional relationship, and thus also in a certain hierarchical (structural) relation. In order to clarify the notion of hierarchy in verb clusters, consider (37a): since we know that the modal verb must selects a bare infinitival and that the perfect auxiliary to have selects a participle phrase, the base-generated hierarchical structure of this example must be as indicated by the bracketing. This bracketing shows that the modal verb is superior to the auxiliary (as well as the participle), and that the auxiliary is superior to the participle. Example (37b) also shows that in English the superiority relation between verbs is straightforwardly reflected by their linear order: superior verbs precede the structurally lower ones.

37
a. John [must [have [seen that film]]].
b. John must have seen that film.

This is not the case in languages like Dutch, however: the processes involved in the creation of verb clusters may disrupt the one-to-one correspondence between hierarchical and linear order. For example, verb clustering may linearize the hierarchical structure in (38a) in various ways, as indicated in the (b)-examples.

38
a. Jan [moet [hebben [de film gezien]]].
b. dat Jan die film moet hebben gezien.
b'. dat Jan die film moet gezien hebben.
b''. dat Jan die film gezien moet hebben.

Subsection II will therefore propose a procedure for mechanically determining the underlying hierarchical order of verbs in verb clusters. This procedure will show, for instance, that in (39a) the modal verb willen'to want' is superior to the perfect auxiliary hebben, whereas in (39b) the auxiliary is superior to the modal.

39
a. dat Jan dat boek morgen <gelezen> wil <gelezen> hebben <gelezen>.
  that  Jan that book  tomorrow    read  wants  have
  'that Jan wants to have read that book by tomorrow.'
b. dat Jan dat boek altijd al heeft willen lezen.
  that  Jan that book  always  already  has  wanted  read
  'that Jan has always wanted to read that book.'

Subsection III will show that the investigation of superiority relations reveals certain systematic hierarchical restrictions between verbs entering a single verb cluster; the contrast between the two examples in (40), for instance, will be argued to show that perfect auxiliaries may select verbal projections with an aspectual verb as their head, but that aspectual verbs are not able to select verbal projections with a perfect auxiliary as their head.

40
a. dat Jan dat boek is gaan lezen.
  that  Jan that book  is go  read
  'that Jan has started to read that book.'
b. * dat Jan dat boek gaat hebben gelezen.
  that  Jan that book  goes  have  read
readmore
[+]  I.  Notational conventions

Before we start our investigation, we want to introduce a number of notational conventions that may facilitate the discussion. If possible, we will distinguish the verbs in our schematic representations of verb clusters by means of denominators like Aux(iliary) for auxiliary verbs, Asp(ectual) for aspectual verbs, Modal for modal verbs, and Main for the most deeply embedded main verb. By using en-dashes to indicate linear order, we can schematically represent the verb clusters in (39) as in (41).

41
a. Modal–Aux–Main
wil hebben gelezen
a'. Modal– Main–Aux
wil gelezen hebben
a''. Main–Modal–Aux
gelezen wil hebben
b. Aux–Modal–Main
heeft willen lezen

Furthermore, we will use numeral indices to indicate the hierarchical order; Vi+1–Vi expresses that Vi+1 is superior to Vi, due to the fact that the former verb selects the projection of the latter verb as its complement. This means that we can now simultaneously express the linear and the hierarchical order of the verbs in the verb clusters in (39) by means of the representations in (42).

42
a. Modal3–Aux2–Main1
wil hebben gelezen
a'. Modal3–Main1–Aux2
wil gelezen hebben
a''. Main1–Modal3–Aux2
gelezen wil hebben
b. Aux3–Modal2–Main1
heeft willen lezen

Observe that the use of shorthand "Main" in (41) and (42) is somewhat misleading because we have argued that modal verbs like willen'to want' are also main verbs. By restricting the use of the most deeply embedded main verb (that is, by not using "Main2", "Main3", etc), this will probably not lead to any misinterpretations.
      In order to avoid confusion, it is also important to note that the numbering convention is not used consistently in the linguistic literature: in many studies on verb clusters, counting does not start with the most deeply embedded verb, but with the most superior one, e.g., the finite verb in main clauses. We opt for the former option for practical reasons, more specifically because it will enable us to compare examples like (43a) and (43b) while keeping the numeral indices constant.

43
a. dat Jan dat boek heeft willen lezen.
Aux3-Modal2-Main1
  that  Jan that book  has  wanted  read
  'that Jan has wanted to read that book.'
b. dat Jan dat boek wil lezen.
Modal2-Main1
  that  Jan that book  wants  read
  'that Jan wants to read that book.'
c. dat Jan dat boek leest.
Main1
  that  Jan that book  reads
  'that Jan is reading that book.'
[+]  II.  A procedure for determining hierarchical order

Detecting the hierarchical relations between verbs is easy in English as they can be read off the linear order of the verbs. Things are different, however, in the Germanic OV-languages, as these seem to allow the verbs in verb clusters to be linearized in various language-specific orders. For example, the cluster formed by the verbs in examples such as dat Jan dat liedje heeft moeten zingen'that Jan has had to sing that song', with the hierarchical order indicated in the header of (44) surfaces in various linear orders depending on the language in question:

44
[...Aux3 [...Modal2 [... Main1 ...]]]
a. Aux3Modal2Main1: Dutch
b. Aux3Main1Modal2: German
c. Modal2Aux3Main1: —
d. Modal2Main1 Aux3: Afrikaans
e. Main1Aux3Modal2: —
f. Main1Modal2Aux3: Frisian

Example (44) shows that four out of the six logically possible linear orders occur as a neutral order in some major Germanic OV-language. There are only two linear orders that do not occur as such: the orders in (44c&e) are rare and occur in stylistically/intonationally marked contexts only; see Schmid & Vogel (2004) for a selection of German dialects, and Barbiers et al. (2008:ch.1) for Dutch dialects.
      The variation we find shows that the linear order of verbs in verb clusters does not necessarily reflect their underlying hierarchical order. Fortunately, there is a simple procedure to establish the latter order, which is based on the assumption that the most superior (structurally highest) verb in the cluster shows up as the finite verb in finite clauses: by omitting this verb, the next most superior verb will surface as the finite verb, etc. By applying this procedure to example (45a), we can provide syntactic evidence for the hierarchical structure proposed in the header of (44); omission of the finite auxiliary forces the modal verb to surface as the finite verb in (45b), and by also omitting this modal, the verb zingen will surface as the finite verb in (45c).

45
a. dat Jan dat liedje heeftfinite moeteninf zingeninf.
Aux3–Modal2–Main1
  that  Jan  that song has  must  sing
  'that Jan has had to sing that song.'
b. dat Jan dat liedje moetfinite zingeninf.
Modal2–Main1
  that  Jan  that song  must  sing
  'that Jan has to sing that song.'
c. dat Jan dat liedje zingtfinite.
Main1
  that  Jan  that song  sings
  'that Jan is singing that song.'

As it happens, the linear order of the verbs in (45) reflects their hierarchical order in a one-to-one fashion. We will therefore apply the same procedure to example (46a), in which the linear order does not correspond in a one-to-one fashion to the underlying hierarchical order [... Modal [... Aux [... Main ...]]].

46
a. dat Jan dat liedje zoufinite gezongenpart hebbeninf.
Modal3–Main1–Aux2
  that  Jan that song  would  sung  have
  'that Jan would have sung that song.'
b. dat Jan dat liedje gezongenpart hadfinite.
Main1–Aux2
  that  Jan that song  sung  had
  'that Jan had sung that song.'
c. dat Jan dat liedje zongfinite.
Main1
  that  Jan that song  sang
  'that Jan sang that song.'

Although the hierarchical order of the verbs in a given verb cluster will normally also be clear from the selection restrictions imposed by the verbs involved, it is certainly useful to be able to support analyses proposed on the basis of such restrictions independently by means of the simple omission test proposed here.

[+]  III.  Restrictions on hierarchical order

This section discusses a number of restrictions on the hierarchical order of verbs in verb clusters. The main issue is: What types of verbal projections can be selected by what types of verbs? Subsection A starts with a discussion of the basic cluster types of two verbs that can be created by embedding a main verb under a non-main verb or some other main verb that triggers verb clustering. The investigation in the later subsections in a sense inverts the procedure for determining the hierarchical organization of verb clusters proposed in Subsection II by considering the question of how the basic cluster types discussed in Subsection A can be extended by embedding them under some non-main verb, or an additional main verb that triggers verb clustering. The discussion will show that it is not the case that anything goes: there are certain restrictions on what counts as acceptable verb combinations. The existence of such restrictions is clearest in clusters of three or more verbs with just one single main verb, and Subsection B will therefore discuss these first, subsequently, Subsection C and D will address verb clusters of three or more verbs with, respectively, two and three main verbs. It is possible to construct clusters with four or more main verbs, but such clusters are rarely attested in actual language use and resist syntactic investigation due to the fact that the meanings expressed by such clusters are normally quite far-fetched; for this reason, we will not attempt to discuss such cases in a systematic way.

[+]  A.  Verb clusters of two verbs

An absolute restriction on verb clusters is that the most deeply embedded verb must be a main verb. In our examples we will generally use the transitive verb lezen'to read' for practical reasons instead of an intransitive or an unaccusative verb: (i) some of the superior verbs may impose an animateness restriction on the subject of their verbal complement; (ii) the placement of the direct object of lezen provides a clue for the analysis of the construction—verb clustering requires that it precede the superior verb; (iii) infinitival transitive verbs like lezen can be passivized whereas intransitive and unaccusative verbs cannot.
      Sections 5.2 and Chapter 6 have shown that main verbs can be selected by various types of main and non-main verbs. In what follows, we will discuss a small, representative sample of such verbs triggering verb clustering. We will take the subject control verb proberen'to try' and the subject raising (SR) verb schijnen'to seem' in (47) as representatives of the class of main verbs selecting te-infinitivals, and the modal verb moeten'must/be obliged', the perception verb zien'to see', and the causative/permissive verb laten'to make/let' in (48) as representatives of the class of main verbs selecting bare infinitivals. The verb clusters in these examples are in italics, and the superior main verbs are underlined.

47
a. dat Jan dat boek probeert te lezen.
Control2–Main1
  that  Jan that book  tries  to readinf
  'that Jan is trying to read that book.'
b. dat Jan dat boek schijnt te lezen.
SR2–Main1
  that  Jan that book  seems  to readinf
  'that Jan seems to be reading that book.'
48
a. dat Jan dat boek moet lezen.
Modal2–Main1
  that  Jan that book  must  readinf
  'that Jan must/is obliged to read that book.'
b. dat Jan haar dat boek ziet lezen.
Perc2–Main1
  that  Jan her  that book  sees  readinf
  'that Jan sees her read that book.'
c. dat Jan haar dat boek laat lezen.
Caus2–Main1
  that  Jan her  that book  makes  readinf
  'that Jan makes/lets her read that book.'

Non-main verbs can also be divided into several classes. First, the examples in (49) show that perfect and passive auxiliaries select verbs in the form of a participle. Example (49c) contains the ditransitive particle verb voorlezen'to read aloud', since krijgen-passivization requires that an indirect object be promoted to subject. Note that the participles may also follow the auxiliaries; we will ignore this here but return to it in Section 7.3, where we will discuss the linearization of verb clusters. The verb clusters in (49) are again in italics, and the non-main verbs are underlined.

49
a. dat Jan dat boek gelezen heeft.
Main1–Perf2
  that  Jan that book  readpart  has
  'that Jan has read that book.'
b. dat dat boek gelezen wordt.
Main1–Pass2
  that  that book  readpart  is
  'that that book is being read.'
c. dat het kind dat boek voorgelezen krijgt.
Main1–Pass2
  that  the child  that book  prt-readpart  gets
  'that the child is being read that book aloud.'

Second, the examples in (50) show that there are also non-main verbs selecting infinitival complements: aspectual verbs like gaan'to go' select bare infinitivals, whereas semi-aspectual verbs like zitten'to sit' select te-infinitivals (if they are finite).

50
a. dat Jan dat boek gaat lezen.
Asp2–Main1
  that  Jan that book  goes  readinf
  'that Jan is going to read that book.'
b. dat Jan dat boek zit te lezen.
Semi-asp2–Main1
  that  Jan that book  sits  to readinf
  'that Jan is reading that book.'
[+]  B.  Larger verb clusters with one main verb

The verb clusters in the examples discussed in Subsection A can be extended by adding one or more verbs that triggers verb clustering. That it is not a random affair can readily be observed in larger verb clusters with a single main verb, that is, extensions of the verb clusters in (49) and (50) with a non-main verb. We start our discussion with extensions of the (semi-)aspectual examples in (50), after which we will proceed to the perfect/passive examples in (49). The examples in (51) first show that aspectual verbs like gaan'to go' and semi-aspectual verbs like zitten'to zit' may co-occur, but that the former must then be superior to the latter–cases like (51b), in which a semi-aspectual verb is superior to an aspectual verb, are unacceptable.

51
a. dat Jan dat boek gaat zitten lezen.
Asp3–Semi-asp2–Main1
  that  Jan that book  goes  sit  readinf
  'that Jan is going to read that book.'
b. * dat Jan dat boek zit (te) gaan lezen.
Semi-asp3–Asp2–Main1
  that  Jan that book  sits  to go  readinf

The primeless examples in (52) show that (semi-)aspectual verbs can also co-occur with the perfect auxiliaries; aspectual verbs take the auxiliary zijn, whereas semi-aspectual verbs take the auxiliary hebben (just like their main verb counterparts). The primed examples show, however, that the perfect auxiliary must be superior to the (semi-)aspectual verb; they do not seem to be able to take a perfect phrase, that is, a phrase containing a perfect auxiliary as their complement (although examples such as (52a') do occasionally occur on the internet). Example (52c) shows that examples such as (51a), which contain both an aspectual and a semi-aspectual verb, can also occur in the perfect tense; the auxiliary must then again be the most superior one in the cluster.

52
a. dat Jan dat boek isgaan lezen.
Perf3–Asp2–Main1
  that  Jan that book  is go  readinf
  'that Jan has been going to read that book.'
a'. * dat Jan dat boek gelezen gaat hebben.
Asp3–Perf2–Main1
  that  Jan that book  readpart  goes  have
b. dat Jan dat boek heeft zitten (te) lezen.
Perf3–Semi-asp2–Main1
  that  Jan that book  has  sit  to readinf
  'that Jan has been reading that book.'
b'. * dat Jan dat boek gelezen zit (te) hebben.
Semi-asp3–Perf2–Main1
  that  Jan that book  readpart  sits  to have
c. dat Jan dat boek is gaan zitten lezen.
Perf4–Asp3–Semi-asp2–Main1
  that  Jan that book  is go  sit  readinf
  'that Jan has started to read that book.'

Although it is not possible to have more than one perfect or more than one passive auxiliary in a single clause, the examples in (53) show that it is possible for perfect and passive auxiliaries to co-occur. Example (53a) is marked with a percentage sign given that it is restricted to certain southern varieties of Dutch, but example (53b) is generally accepted.

53
a. % dat dat boek gelezen is geworden.
Main1–Perf3–Pass2
  that  that book  readpart  is been
  'that that book has been read.'
b. dat het kind dat boek voorgelezen heeft gekregen.
Main1–Perf3–Pass2
  that  the child  that book  prt-readpart  has  got
  'that the child has been read that book aloud.'

The hierarchical order of the two auxiliaries is very strict: the perfect auxiliary is always superior to the passive auxiliary. In fact, it seems that passive auxiliaries are always very low in the structure, as is clear from (54a) in which the passive auxiliary is embedded under the aspectual verb gaan'to go'. Similar examples with semi-aspectual verbs like zitten'to sit' seem rare though, and mainly restricted to main verbs and verbal expressions denoting acts of deception like bedriegen/belazeren'to deceive' and om de tuin leiden'to lead down the garden path' in the (b)-examples; in such cases, the semi-aspectual verb is again clearly superior to the passive auxiliary.

54
a. dat Jan per maand betaald gaat worden.
Main1–Asp3–Pass2
  that Jan  per month  paid  goes  be
  'that Jan is going to be paid per month.'
b. dat ik hier bedrogen/belazerd zit te worden.
Main1–Semi-asp3–Pass2
  that  here  deceived/deceived  sit  to be
  'that Iʼm being deceived here.'
b'. dat ik om de tuin geleid zit te worden.
Main1–Semi-asp3–Pass2
  that  around the garden  led  sit to be
  'that Iʼm being led down the garden path.'

The discussion in this section has shown that there is a strict hierarchical order between the non-main verbs in verb clusters. This order is as given in (55), in which the connective ">" stands for "is superior to".

55
Hierarchical order in verb clusters with one main verb: perfect auxiliary > aspectual > semi-aspectual > passive auxiliary > main verbHierarchical order in verb clusters with one main verb: perfect auxiliary > aspectual > semi-aspectual > passive auxiliary > main verb
[+]  C.  Larger Verb clusters with two main verbs

This section discusses larger verb clusters with two main verbs, As our point of departure we will take examples in (47) and (48) from Subsection A, which are repeated here as (56) and (57) for convenience.

56
a. dat Jan dat boek probeert te lezen.
Control2-Main1
  that  Jan that book  tries  to readinf
  'that Jan is trying to read that book.'
b. dat Jan dat boek schijnt te lezen.
SR2–Main1
  that  Jan that book  seems  to readinf
  'that Jan seems to be reading that book.'
57
a. dat Jan dat boek moet lezen.
Modal2–Main1
  that  Jan that book  must  readinf
  'that Jan must/is obliged to read that book.'
b. dat Jan haar dat boek ziet lezen.
Perc2–Main1
  that  Jan her  that book  sees  readinf
  'that Jan sees her read that book.'
c. dat Jan haar dat boek laat lezen.
Caus2–Main1
  that  Jan her  that book  makes  readinf
  'that Jan makes/lets her read that book.'

We will extend these constructions by an additional non-main verb. In principle, this can be done in two different ways: we can add the non-main verb to the superior main verb, but we can also add it to the structurally lower one. The discussion in the following subsections will show that there are various restrictions. These are, however, normally not of a syntactic, but rather of a semantic or a pragmatic nature.

[+]  1.  Perfect auxiliaries I: Perf3-Main2-Main1

It seems easily possible to add a perfect auxiliary to the superior main verbs in (56) and (57) with the exception of the subject raising verb schijnen: most people consider examples such as (58b) at least marked. Observe that all examples exhibit the infinitivus-pro-participio (IPP) effect, which is of course not surprising given that we have seen that this is a hallmark of verb clustering; cf. Section 7.1.1. For convenience, we will underline the added non-main verbs in the examples to come.

58
a. dat Jan dat boek heeft proberen te lezen.
Perf3–Control2–Main1
  that  Jan that book  has  try  to readinf
  'that Jan has tried to read that book.'
b. ? dat Jan dat boek heeft schijnen te lezen.
Perf3–SR2–Main1
  that  Jan that book  has  seems  to readinf
  'that Jan has seemed to read that book.'
59
a. dat Jan dat boek heeft moeten lezen.
Perf3–Modal2-Main1
  that  Jan that book  has  must  readinf
  'that Jan has had to read that book.'
b. dat Jan haar dat boek heeft zien lezen.
Perf3–Perc2-Main1
  that  Jan her  that book  has  see  readinf
  'that Jan has seen her read that book.'
c. dat Jan haar dat boek heeft laten lezen.
Perf3–Caus2-Main1
  that  Jan her  that book  has  make/let  readinf
  'that Jan has made/let her read that book.'
[+]  2.  Perfect auxiliaries I: Main3-Perf2-Main1

At first sight, it seems that control and subject raising verbs differ with respect to the question as to whether they are able to take a perfect te-infinitival as their complement: whereas (60b) is impeccable, example (60a) seems infelicitous.

60
a. $ dat Jan dat boek gelezen probeert te hebben.
Main1-Control3-Perf2
  that  Jan  that book  readpart  tries  to have
  'that Jan tries to have read that book.'
b. dat Jan dat boek gelezen schijnt te hebben.
Main1-SR3-Perf2
  that  Jan  that book  readpart  seems  to have
  'that Jan seems to have read that book.'

There is reason, however, to assume that the infelicitousness of (60a) is not due to some syntactic selection restriction imposed by proberen, but is related to the fact that proberen triggers an irrealis reading of its complement: the eventuality expressed by the te-infinitival must be located in the non-actualized part of the time interval evoked by the present/past tense of the matrix clause—in the present, the eventuality is located after speech time. This seems to clash with the default reading of the perfect, which locates the completed eventuality in the actualized part of the relevant tense domain. The present perfect example (61a), for example, locates the eventuality before speech time by default; it normally expresses that Jan has read the book at speech time. It must be observed, however, that this default reading of the perfect is pragmatic in nature and can readily be canceled by adding an adverbial phrase like morgen'tomorrow' that refers to a time interval in the non-actualized part of the tense domain; example (61b) locates the completed eventuality after speech time; see Section 1.5.4 for extensive discussion.

61
a. Jan heeft het boek zeker gelezen.
  Jan has  the book  certainly  read
  'Jan has certainly read the book.'
b. Jan heeft het boek morgen zeker gelezen.
  Jan has  the book  tomorrow  certainly  read
  'Jan will certainly have read the book by tomorrow.'

This suggests that the default reading of the perfect tense makes the assertion expressed by (60a) incoherent, Example (62) shows, however, that (60a) also becomes fully acceptable if we add the adverb morgen'tomorrow'. This suggests that the unacceptability of (60a) is not due to some syntactic (or semantic) selection restriction either but is simply an effect of pragmatics: the addition of morgen provides additional temporal information that cancels the default reading of the perfect, as a result of which the eventuality expressed by the infinitival clause can be located in the non-actualized part of the present domain and the message becomes fully coherent.

62
dat Jan het boek morgen gelezen probeert te hebben.
  that  Jan  the book  tomorrow  prt-readpart  tries  to have
'that Jan tries to have read the book by tomorrow.'

Note in passing that we cannot appeal to the IPP-effect in order to establish that we are indeed dealing with a verb cluster of three verbs in examples such as (62), given that it is impossible to add a second perfect auxiliary associated with the superior verb proberen: cf. *dat Jan dat boek morgen gelezen heeft proberen/geprobeerd te hebben. It seems, however, very unlikely that (62) can be analyzed as a remnant extraposition construction: under such an analysis, the fact that the participle gelezen precedes the verb proberen can only be derived if we extract this participle from the verb cluster gelezen te hebben of the extraposed te-infinitival clause, but such movements have not been attested (or even considered as a possible option) in the existing literature. Nevertheless, we should note that we did find a small number of cases on the internet such as gehoord/gezien beweert te hebben'claims to have heard/seen', despite the fact that there is strong evidence for assuming that beweren normally triggers (remnant) extraposition; we will ignore this problem here and leave the question as to whether or not these cases should be seen as accidental writing errors for future research.
      Subsection C1 has shown that the perfect auxiliary can be readily added to the superior verb in clusters like Modal2–Main1; the relevant example is repeated here as (63a). Example (63b) shows that it is equally possible to add a perfect auxiliary to the embedded main verb.

63
a. dat Jan dat boek heeft moeten lezen.
Perf3–Modal2–Main1
  that  Jan that book  has  must  read
  'that Jan has had to read that book.'
b. dat Jan dat boek moethebben gelezen.
Modal3–Perf2–Main1
  that  Jan  that book  must have  read
  'that Jan has to have read that book.'

The two examples do, however, exhibit a conspicuous difference in interpretation: whereas the modal in (63a) receives a (directed) deontic "obligation" reading, the modal in (63b) receives an epistemic "necessity" interpretation; we refer the reader to Section 5.2.3.2, sub III, for a discussion of these types of modality. This contrast can also be demonstrated by the fact illustrated in (64) that the hierarchical order Perf3–Modal2–Main1requires the subject of the sentence to be able to control the eventuality expressed by Main1, whereas the hierarchical order Modal3–Perf2–Main1 does not require this.

64
a. * dat dat huis heeft moeten instorten.
Perf3–Modal2–Main1
  that  that house  has  must  prt.-collapse
b. dat dat huis moet zijn ingestort.
Modal3–Perf2–Main1
  that  that house  must be  prt.-collapsed
  'that that house must have collapsed.'

It is not clear whether the difference in interpretation between the two examples in (63) has a syntactic origin. The past perfect counterpart of (63a) in (65a), for example, seems to be compatible both with a directed deontic and with an epistemic reading of the modal verb. That this is indeed the case is supported by the fact that the past perfect counterpart of (64a) in (65b) is also fully acceptable.

65
a. dat Jan dat boek had moeten lezen.
Perf3–Modal2–Main1
  that  Jan that book  has  must  read
  'that Jan had been obliged to read that book.'
b. dat dat huis had moeten instorten.
Perf3–Modal2–Main1
  that  that house  has  must  prt.-collapse
  'that that house had had to collapse.'

Section 5.2.3.2, sub IIIC, has further argued that the epistemic reading of example (63b) is related to the default reading of the perfect tense, namely that the completed eventuality is placed in the actualized part of the present-tense interval (that is, before speech time). This correctly predicts that the deontic interpretation of the modal is possible in (66), in which we cancelled this default reading by adding an adverb like morgen'tomorrow', which locates the eventuality in the non-actualized part of the present-tense interval.

66
dat Jan dat boek morgen moet hebben gelezen.
Modal3–Perf2–Main1
  that  Jan  that book  tomorrow  must  have  read
'that Jan must have read that book tomorrow.'

This leads to the conclusion that there does not seem to be any syntactic restriction that blocks the extension of the cluster Modal2–Main1 by adding a perfect auxiliary associated with either Modal2 or Main1.
      This leaves us with the constructions containing perception and causative verbs, subsection C1 has shown that perfect auxiliaries can be readily added to these verbs, but it seems impossible to add them to the embedded main verb; examples such as (67) are infelicitous.

67
a. $ dat Jan haar dat boek ziet hebben gelezen.
Perc3–Perf2–Main1
  that  Jan her  that book  sees  have  readpart
  Compare: 'that Jan sees her have read that book.'
b. $ dat Jan haar dat boek laat hebben gelezen.
Caus3–Perf2–Main1
  that  Jan her  that book  makes  have  readpart
  Compare: 'that Jan makes/let her have read that book.'

The use of the dollar signs indicates that it is again not a priori clear whether the unacceptability of these examples is due a syntactic or a semantic/pragmatic restriction. We believe that there is reason to think of a constraint of the latter type. In the case of (67a), the reason for this is that examples such as dat Jan haar dat boek ziet lezen'that Jan sees her read that book' express a notion of simultaneity: the eventuality of seeing occurs simultaneously with the eventuality expressed by the embedded bare infinitival, and the default reading of simple present locates these eventualities at speech time. This seems to clash with the default reading of the perfect tense in examples such as (67a), which locates the completed eventuality expressed by the infinitival complement in the actualized part of the present-tense interval, that is, before speech time.
      Under its causative interpretation, the construction in (67b) is an irrealis construction in the sense that the eventuality expressed by the embedded bare infinitival is located after speech time, which again clashes with the default interpretation of the perfect, which locates the completed eventuality before speech time. Under its permissive interpretation, the eventuality expressed by the embedded bare infinitival is either located at or after speech time, and this again clashes with the default interpretation of the perfect. It should be noted, however, that the addition of an adverb like morgen'tomorrow' does not seem to improve the result: ??dat Jan haar morgen dat boek laat hebben gelezen, perhaps because this construction is blocked by the simpler construction dat Jan haar morgen dat boek laat lezen'that Jan will make her read that book tomorrow'. We will not pursue this issue any further.
      The main finding of this subsection is that there is no reason for assuming a syntactic restriction that prohibits the selection of a perfect infinitival construction by the superior main verbs in (56) and (57). In some cases this leads to infelicitous results, but this seems due to semantic/pragmatic reasons.

[+]  3.  Passive auxiliaries I: Pass3-Main2-Main1

It seems impossible to passivize the superior verbs in the examples in (56) and (57) from the introduction to this subsection (p.). The fact that control verbs like proberen'to try' resist passivization if they are part of a verb cluster strongly suggests that this is due to some syntactic constraint. Consider the examples in (68). The primeless examples illustrate again that proberen is not only able to select transparent te-infinitivals, which gives rise to verb clustering, but also opaque te-infinitivals, which gives rise to extraposition. The primed examples show that passivization is only possible if the complement is opaque/extraposed; cf. Koster (1984b). Observe that substituting an infinitive for the participle geprobeerd or changing the order of the verb cluster (or a combination of the two) will not affect the status of (68b').

68
a. dat Jan probeert (om) het boek te lezen.
extraposition
  that  Jan tries  comp  the book  to read
  'that Jan is trying to read the book.'
a'. dat er geprobeerd wordt (om) het boek te lezen.
  that  there  tried  is  comp  the book  to read
  'that it is tried to read the book.'
b. dat Jan het boek probeert te lezen.
verb clustering
  that  Jan the book  tries  to read
  'that Jan is trying to read that book.'
b'. * dat er het boek geprobeerd wordt te lezen.
Control2–Pass3–Main1
  that  there  the book  tries  be  to read

The fact that proberen can be passivized if it selects an opaque te-infinitive suggests that there must be something special going on if proberen selects a transparent te-infinitive. However, there is no reason for assuming that this is due to some selection restriction, given that this can also be accounted for in terms of obligatory and optional control; see Section 5.2.1.3, sub III, for these notions. First, the implicit PRO-subject of opaque infinitival clauses is optionally controlled; it does not require an antecedent in the matrix clause, as a result of which the passive construction in acceptable. Second, the implicit PRO-subject of transparent infinitival clauses is obligatorily controlled in that it does require an antecedent in the matrix clause, as a result of which the passive construction is unacceptable. We refer the reader to Section 5.2.2.1 for a more detailed discussion of this. For completeness' sake, we should note that the discussion above has ignored the fact that (68b) can in principle also be analyzed as a remnant extraposition construction, that is, as a case with a semi-transparent te-infinitival; this does not really affect the argument given that Section 5.2.2.3 has shown that PRO-subjects of such infinitival clauses are also obligatorily controlled.
      The fact that subject raising verbs like schijnen'to seem' cannot be passivized is expected; Section 5.2.2.2, sub IC, has shown that such verbs are unaccusative; since unaccusative verbs always resist passivization, there is nothing special to discuss here.
      The examples in (69) show that main verbs selecting a bare infinitival cannot be passivized, regardless of whether the passivized verb surfaces as an infinitive or a participle. We give the examples with clusters in the order Pass3–V2–Main1; changing this order will not affect the acceptability judgments.

69
a. * dat (er) dat boek wordt moeten/gemoeten lezen.
Pass3–Modal2–Main1
  that  there  that book  is mustinf/mustpart  readinf
b. * dat (er) haar dat boek wordt zien/gezien lezen.
Pass3–Perc2–Main1
  that  there  her  that book  is  see/seen  readinf
c. * dat (er) haar dat boek wordt laten/gelaten lezen.
Pass3–Caus2–Main1
  that  there  her  that book  is  madeinf/madepart  readinf

Because constructions with perception and causative verbs are often analyzed as heads of AcI-constructions, the unacceptability of the (impersonal) passive constructions in (69b&c) need not surprise us as we may expect that passivization must involve promotion of the subject of the bare infinitival to subject of the matrix clause. This expectation is, however not borne out; the examples in (70) are also unacceptable.

70
a. * dat zij dat boek wordt zien/gezien lezen.
Pass3–Perc2–Main1
  that  she  that book  is  see/seen  readinf
b. * dat zij dat boek wordt laten/gelaten lezen.
Pass3–Caus2–Main1
  that  she  that book  is  madeinf/madepart  readinf

The fact that perception and causative verbs are not normally analyzed as control verbs (see Petter 1998:ch.4 for an alternative view) suggests that the unacceptability of passivization in (69) and (70) cannot be accounted for by an appeal to control theory. Since there is no obvious semantic/pragmatic reason for the impossibility of passivization either, it seems likely that we have to account for the unacceptability of these examples in terms of verb clustering; for one attempt of this type we refer to Bennis & Hoekstra (1989b).

[+]  4.  Passive auxiliaries II: Main3-Pass2-Main1

This subsection discusses passivization of the more deeply embedded verbs in the examples in (56) and (57) from the introduction to this subsection (p.). Let us first consider the case in (71), in which the superior verb is a control verb; (71a) involves regular passivization and (71b) krijgen-passivization. The reader can easily identify the two main verbs by keeping in mind that the most deeply embedded main verb (Main1) appears as a participle; the control verb is the finite verb.

71
a. % dat Jan gelezen probeert te worden.
Main1–Control3–Pass2
  that  Jan  readpart  tries  to be
  'that Jan tries to be read.'
b. dat Jan dat boek voorgelezen probeert te krijgen.
Main1–Control3–Pass2
  that  Jan that book  prt-readpart  tries  to get
  'that Jan is trying to be read that book aloud.'

The fact that the krijgen-passive can readily be embedded under proberen in (71b) suggests that the infelicitousness of (71a) has little to do with syntactic selection restrictions imposed by the verb proberen. Instead, it is reasonable to assume that it is due to the fact that we are dealing with an obligatory subject control construction; the animate subject of proberen is simply not a suitable antecedent for the implied PRO-subject of the infinitival passive construction: $Jani probeert [PROi gelezen te worden]. Observe that we used a percentage sign in (71a) in order to express that some people may accept this example in the reading in which Jan is attempting to make other people read a body of work written by him; cf. Louis Couperus wordt nog veel gelezen'Louis Couperus is still read a lot'. That our account in terms of obligatory subject control is on the right track is also suggested by the fact that completely parallel examples are acceptable if subject control leads to a result compatible with the selection restriction imposed by the passive construction on the subject. This is illustrated in the primeless examples in (72), in which the animate subject of proberen is a suitable antecedent for the PRO-subject of the infinitival passive construction: Jani probeert [PROi ontslagen/verkozen te worden]. The primed examples are added to show that we get the IPP-effect in the perfect tense, from which we may conclude that we are indeed dealing with verb clustering—the same is, of course, supported by the fact that the passive participles precede the verb proberen.

72
a. dat Jan ontslagen probeert te worden.
Main1–Control3–Pass2
  that  Jan dismissed  tries  to be
  'Jan Jan is trying to get dismissed.'
a'. dat Jan ontslagen heeft proberen/*geprobeerd te worden.
  that  Jan dismissed has  try/tried  to be
  'Jan Jan has tried to get dismissed.'
b. dat Jan verkozen probeert te worden.
Main1–Control3–Pass2
  that  Jan elected  tries  to be
  'that Jan is trying to get elected.'
b'. dat Jan verkozen heeft proberen/*geprobeerd te worden.
  that  Jan elected  has  try/tried  to be
  'that Jan has tried to get elected.'

      The examples in (73) show that subject raising verbs like schijnen are quite capable of taking a passivized te-infinitival: (73a) involves a regular and (73b) a krijgen-passive. The most deeply embedded main verb (Main1) again appears as a participle, while the subject raising verb is the finite verb.

73
a. dat dat boek door Els gelezen schijnt te worden.
Main1–SR3–Pass2
  that  that book  by Els  prt-readpart  seems  to be
  'that that book seems to be read by Els.'
b. dat Marie dat boek voorgelezen schijnt te krijgen.
Main1–SR3–Pass2
  that  Marie that book  prt-readpart  seems  to get
  'that Marie seems to be read that book to.'

      The examples in (74) show that modal, perception and causative verbs are able to select a passivized bare infinitival. The acceptability of the results sometimes depends on the embedded main verb, for which reason we replaced the verb lezen by the main verb slopen'to demolish' in (74b&c).

74
a. dat dat boek morgen gelezen moet zijn.
Main1–Modal3–Pass2
  that  that book  tomorrow  readpart  must  have.been
  'that that book must have been read by tomorrow.'
b. dat Jan het huis (door Els) gesloopt zag worden.
Main1–Perc3–Pass2
  that  Jan the house  by Els  demolished  saw  be
  'that Jan saw the house be demolished (by Els).'
c. % dat Jan het huis (door Els) gesloopt liet worden.
Main1–Caus3–Pass2
  that  Jan the house  by Els demolished  let  be
  'that Jan made/let the house be demolished (by Els).'

A percentage sign is added to (74c) because some speakers object to this example, and the same seems to hold to a lesser extent for (74b). There is reason for assuming that this is not related to a selection restriction imposed by the causative/perception verb. Instead, it seems related to the fact that there is an alternative way of expressing the passive meaning with the help of AcI-verbs. This is illustrated in the examples in (75), which show that the subjects can simply be omitted or be replaced by agentive door-PPs.

75
a. dat Jan Els het huis zag slopen.
  that  Jan Els the house  saw  demolish
  'that Jan saw Els demolish the house.'
a'. dat Jan het huis (door Els) zag slopen.
  that  Jan the house   by Els  saw  demolish
b. dat Jan Els het huis liet slopen.
  that  Jan Els the house  made  demolish
  'that Jan made/let Els demolish the house.'
b'. dat Jan het huis (door Els) liet slopen.
  that  Jan the house   by Els  let  demolish

One possible account for the markedness of (74c) is an appeal to syntactic blocking; for one reason or another, speakers simply value the structure in (75b') higher than the one in (74c). If so, we may conclude our discussion by saying that the superior main verbs in (56) and (57) do not impose any restrictions on the voice of their infinitival complement.

[+]  5.  Semi-aspectual and aspectual verbs I: Asp3-Main2-Main1

The examples in (76) show that the addition of a (semi-)aspectual verb on top of control structures such as (56a) seems possible. Some people may find example (76b) somewhat marked as given, but it becomes completely acceptable if we add an adverbial phrase of duration like al de hele dag'already the whole day': dat Jan dat boek al de hele dag zit te proberen te lezen.

76
a. dat Jan dat boek gaat proberen te lezen.
Asp3–Control2–Main1
  that  Jan that book  goes  try  to readinf
  'that Jan is going to try to read that book.'
b. dat Jan dat boek zit te proberen te lezen.
Semi-Asp3–Control2–Main1
  that  Jan that book  sits  to try  to readinf
  'that Jan is trying to read that book.'

A problem with the examples in (76) is that we cannot prove that we are dealing with verb clusters of the type Asp3-Main2-Main1. The reason is that perfect auxiliaries must be superior to the (semi-)aspectual verbs; cf. example (55) in Subsection B. This means that we can make sequences of the form Perf4–Asp3–Control2-Main1, as shown in (77), but not of the form Asp4–Perf3–Control2–Main1, which are needed to check whether the IPP-effect applies to the control verb proberen.

77
a. dat Jan dat boek is gaan proberen te lezen.
  that  Jan that book  is go  try  to readinf
  'that Jan has been going to try to read that book.'
b. dat Jan dat boek heeft zitten proberen te lezen.
  that  Jan that book  has  sits  try  to readinf
  'that Jan has been trying to read that book.'

Given that the examples in (78) must be analyzed as extraposition constructions (with the extraposed clause underlined), we must leave the option open that the examples in (76) are not instances of verb clustering but remnant extraposition constructions.

78
a. dat Jan gaat proberen dat boek te lezen.
extraposition
  that  Jan goes  try  that book  to readinf
  'that Jan is going to try to read that book.'
b. dat Jan zit te proberen dat boek te lezen.
extraposition
  that  Jan sits  to try  that book  to readinf
  'that Jan is trying to read that book.'

Nevertheless, in the absence of solid reasons for claiming that the examples in (76) cannot be analyzed as a verb-clustering construction, we will provisionally assume that it is in fact a possible analysis (besides the remnant extraposition analysis).
      The examples in (79) show that the addition of a (semi-)aspectual verb on top of subject raising constructions such as (56b) is impossible. It may be the case that the infelicitousness of these examples is related to the earlier noted fact that subject raising verbs do not readily appear as non-finite verbs, but it seems equally plausible to assume that it is due to the evidential modality expressed by schijnen: the infinitival clause simply does not satisfy the semantic selection restriction of (semi-)aspectual verbs that their complements refer to an activity controlled by the subject of the clause (cf. Section 6.3.1, sub II).

79
a. * dat Jan dat boek gaat schijnen te lezen.
Asp3–SR2–Main1
  that  Jan that book  goes  seem  to readinf
b. * dat Jan dat boek zit te schijnen te lezen.
Semi-Asp3–SR2–Main1
  that  Jan that book  sits  to seem  to readinf

The examples in (80) show that the addition of a (semi-)aspectual verb on top of modal structures such as (57a) also gives rise to degraded results. For completeness' sake, it should be noted that examples such as (80a) do occasionally occur on the internet, but we tend to think that these are just erroneous forms that are used instead of the fully acceptable form moet gaan lezen'must go read'. The question as to whether this suggestion is indeed on the right track we will leave to future research. It is again plausible to assume that the unacceptability of the examples in (80) is due to the fact that the modal phrases do not satisfy the semantic selection restriction of (semi-)aspectual verbs that their complement refer to an activity controlled by an agent.

80
a. * dat Jan dat boek gaat moeten lezen.
Asp3–Modal2–Main1
  that  Jan that book  goes  must  readinf
b. * dat Jan dat boek zit te moeten lezen.
Semi-Asp3–Modal2–Main1
  that  Jan that book  sits  to must  readinf

The results are different for the AcI-constructions with the perception verb zien and the causative verb laten. The examples in (81) show that it is possible to add the aspectual verb gaan to the examples in (57b&c). We added the particles nog and wel to facilitate the intended posterior-to-speech-time reading; without these particles some speakers may have problems with this construction.

81
a. dat Jan haar dat boek nog wel gaat zien lezen.
Asp3–Perc2–Main1
  that  Jan her  that book  yet aff  goes  see  readinf
  'that Jan will eventually see her read that book.'
b. dat Jan haar dat boek gaat laten lezen.
Asp3–Caus2–Main1
  that  Jan her  that book  goes  make  readinf
  'that Jan is going to make/let her read that book.'

The examples in (82), on the other hand, show that it is not possible to add the semi-aspectual verb zitten to the examples in (57b&c).

82
a. * dat Jan haar dat boek zit te zien lezen.
Semi-asp3–Perc2–Main1
  that  Jan her  that book  sits  to see  readinf
b. * dat Jan haar dat boek zit te laten lezen.
Semi-asp3–Caus2–Main1
  that  Jan her  that book  sits  to let  readinf

The acceptability of the examples in (81) strongly suggests that the unacceptability of the examples in (82) is not due to some syntactic constraint, given that the semi-aspectual verb zitten can normally be embedded under aspectual gaan. It is therefore more likely that the unacceptability of (82) is due to some semantic incompatibility between the semi-aspectual verbs and the verbs zien and laten This might be independently supported for the verb zien by the contrast between the two examples in (83), in which zien takes a nominal object.

83
a. Welke film ga je zien?
  which movie  go  you  see
  'Which movie are you going to see/watch?'
b. * Welke film zit je te zien?
  which movie  sit  you  to see
  Intended reading: 'Which movie are you watching?'

The acceptability contrast indicated in (83) is confirmed by a Google search (7/16/2013): whereas the (colloquial) question in (83a) occurs more than twenty times on the internet, the question in (83b) does not occur at all. We cannot provide similar evidence for causative laten as this verb does not allow nominal complements.

[+]  6.  Semi-aspectual and aspectual verbs II: Main3–Asp2-Main1

Generally speaking, it seems possible to add a (semi-)aspectual verb associated with the structurally lower main verbs of the examples in (56) and (57) from the introduction to this subsection (p.), although we will see that there are certain complications which deserve attention. The examples in (84) show that whereas the addition of an aspectual verb such as gaan is fully acceptable, the addition of a semi-aspectual verb such as zitten gives rise to a degraded result.

84
a. dat Jan dat boek probeert te gaan lezen.
Control3–Asp2–Main1
  that  Jan that book  tries  to go  readinf
  'that Jan is trying to start to read that book.'
b. $ dat Jan dat boek probeert te zitten lezen.
Control3–Semi-asp2–Main1
  that  Jan that book  tries to sit  readinf
  Compare: 'that Jan is trying to be reading that book.'

The acceptability of example (84a) strongly suggests that the unacceptability of example (84b) cannot be due to some syntactic constraint, given that semi-aspectual verbs like zitten can normally be embedded under aspectual verbs like gaan, but that there must be some semantic/pragmatic reason for it. This is quite plausible: the fact that semi-aspectual zitten locates the eventuality expressed by the infinitival clause dat boek lezen'to read that book' in a temporal interval that includes speech time clashes with the fact that the verb proberen triggers an irrealis reading, that is, locates the eventuality expressed by the infinitival clause after speech time.
      The acceptability of the examples in (85) show that it is easily possible to add a (semi-)aspectual verb associated with the structurally lower main verbs in subject raising contexts. Nothing special needs to be said here.

85
a. dat Jan dat boek schijnt te gaan lezen.
SR3–Asp2–Main1
  that  Jan that book  seems  to go  readinf
  'that Jan seems to be going to read that book.'
b. dat Jan dat boek schijnt te zitten (te) lezen.
SR3–Semi-asp2–Main1
  that  Jan that book  seems  to sit  to readinf
  'that Jan seems to be reading that book.'

      The examples in (86) show that the (semi-)aspectual verbs may also occur embedded under the modal verb moeten. The translations suggest that the modal can be interpreted either as an epistemic or as a directed deontic modal, but judgments are not very sharp.

86
a. dat Jan dat boek moet gaan lezen.
Modal3–Asp2–Main1
  that  Jan that book  must  go  readinf
  'that Jan must go reading that book.'
b. dat Jan dat boek moet zitten lezen.
Modal3–Semi-asp2–Main1
  that  Jan  that book  must  sit  readinf
  'that Jan must be reading that book.'

That the epistemic reading is possible seems clear and it can also be supported by the fact that it is the most conspicuous reading of the examples in (87), in which the (semi-)aspectual verb is preceded by an additional perfect auxiliary. The representation of the clusters in (87) is: Modal4–Perf3–AsP2/Semi-asp2–Main1.

87
a. dat Jan dat boek moet zijn gaan lezen.
  that  Jan that book  must  be  go  readinf
  'that Jan must have started to read that book.'
b. dat Jan dat boek moet hebben zitten lezen.
  that  Jan that book  must have  sit  readinf
  'that Jan must have been reading that book.'

The deontic reading of the examples in (86) seems less prominent, which is perhaps also suggested by the fact that the examples in (88), in which the modal verb is preceded by a perfect auxiliary, seem less acceptable. We leave the status of clusters of the type Perf4–Modal3–Asp2/Semi-asp2–Main1 as an issue for future research.

88
a. ? dat Jan dat boek heeft moeten gaan lezen.
  that  Jan that book  has  must  go  readinf
b. ? dat Jan dat boek heeft moeten zitten lezen.
  that  Jan that book  has  must  sit  readinf

      The acceptability of the verb clusters in (89) with the perception verb zien'to see' seems to depend on tense marking. Example (89a) is somewhat odd which may be related to the fact that AcI-constructions with perception verbs normally express a notion of simultaneity; examples such as dat Jan haar dat boek ziet lezen'that Jan sees her read that book' express that the eventuality of seeing is simultaneous with the eventuality expressed by the embedded bare infinitival. The problem with (89a) is due to the fact that while the simple present on the verb zien locates these eventualities at speech time, the aspectual verb gaan locates the eventuality expressed by the infinitival clause after speech time. The past-tense example in (89b') seems more acceptable; it expresses that Jan witnessed the beginning of the eventuality of her reading the book.

89
a. $ dat Jan haar dat boek ziet gaan lezen.
Perc3–Asp2–Main1
  that  Jan her  that book  sees  go  readinf
  'that Jan sees her start reading that book.'
b. dat Jan haar dat boek zag gaan lezen.
Perc3–Asp2–Main1
  that  Jan her  that book  saw  go  readinf
  'that Jan saw her start reading that book.'

The fact that examples (90a&b) are both acceptable shows that the semi-aspectual verb zitten does not raise similar problems as gaan, which is consistent with the fact that zitten locates the eventuality expressed by the bare infinitival in a temporal interval that includes speech time.

90
a. dat Jan haar dat boek ziet zitten lezen.
Perc3–Semi-asp2–Main1
  that  Jan her  that book  sees  sit  readinf
  'that Jan sees her reading that book.'
b. dat Jan haar dat boek zag zitten lezen.
Perc3–Semi-asp2–Main1
  that  Jan her  that book  saw  sit  readinf
  'that Jan saw her reading that book.'

For completeness' sake, note that the account of the markedness of example (89a) given above receives further support from the acceptability of present-tense ACI-constructions with an illusory reading in (91), which were briefly discussed at the end of Section 5.2.3.3, sub I; in such constructions, simultaneity is not implied and the contradiction does not arise.

91
Ik zie haar dat boek nog wel een keer gaan lezen.
  see  her  that book  prt  prt  a time  go  read
'I envisage that sheʼll eventually start reading that book.'

      The examples in (92) with laten'to make/let' are also acceptable. However, the fact that the aspectual verb gaan locates the eventuality after speech time in the non-actualized part of the present-tense interval favors the causative interpretation of (92a). The fact that the semi-aspectual verbs locate the eventuality in a temporal interval that includes speech time makes the permissive reading of (92b) the most plausible one.

92
a. dat Jan haar dat boek laat gaan lezen.
Caus3–Asp2–Main1
  that  Jan her  that book  makes  go  readinf
  'that Jan makes her start reading that book.'
b. dat Jan haar dat boek laat zitten lezen.
Caus3–Semi-asp2–Main1
  that  Jan  her  that book  lets  sit  read
  'that Jan lets her read that book.'

To conclude, we want to note that some speakers find AcI-constructions with (semi-)aspectual verbs somewhat harder to get with transitive than with intransitive verbs. This also seems reflected by our Google searches; it is relatively easy to find examples with intransitive verbs like werken'to work' or slapen'to sleep', but more difficult to find examples with transitive verbs like lezen'to read'.

[+]  7.  Conclusion

The subsections above have shown that there do not seem to be many syntactic restrictions on the formation of larger clusters with two main verbs. The superior main verbs in the verb clusters in (56) and (57) from the introduction to this subsection (p.), which are given in abstract form in the leftmost column of Table 2, seem to allow their complements to contain all the non-main verb types which we discussed, and which are mentioned in the top row of Table 2. This is indicated in the cells by means of the numeral 1, which indicates that the auxiliary in the header of the relevant column can be associated with the embedded main verb (= Main1): in general we were able to account for the less felicitous cases by appealing to semantics and/or pragmatics, which we indicated in the table with a number sign before the numeral 1. In as far as there are syntactic restrictions, these seem to involve the superior main verbs: see the shaded cells without the numeral 2. First, passive auxiliaries are special in that they can only be associated with lower main verbs; we suggested that this is directly related to verb clustering (in a way that perhaps still has to be discovered). Second, subject raising verbs are special in that they normally cannot occur in a non-finite form; this may reflect some deeper syntactic or morphological property of verb clusters, but we have seen that there are reasons to attribute this to more accidental semantics/pragmatics properties of the constructions involved; we leave this open for future research. There are a number of other cases that seem infelicitous, but these seem to have a semantic/pragmatic origin and are therefore marked with a hash sign before the numeral 2. The passive AcI-constructions that we indicated to be marked may be disfavored by some speaker due to syntactic blocking.

Table 2: Verb clusters with two main verbs and one non-main verb
perfect passive aspectual semi-aspectual
Control2-Main1 2/1 1 2/1 2/#1
SR2–Main1 1 1 1 1
Modal2–Main1 2/1 1 1/#2 1/#2
Perc2–Main1 2/#1 1 (marked) 2/1 1/#2
Caus2–Main1 2/#1 1 (marked) 2/1 1/#2

The conclusion that the restrictions related to the embedded main verb are not always syntactic in nature is important because it is often claimed that the superior main verbs impose lexically encoded, syntactic selection restrictions on the substantive verbal contents of their infinitival complements. Our survey above does not corroborate this point of view; the formal restrictions imposed by these verbs are restricted to the morphological form (te-infinitive or bare infinitive) of the verbs they govern. All other restrictions seem to be semantic/pragmatic in nature.

[+]  D.  Larger verb clusters with three main verbs

This subsection investigates the hierarchical structures of verb clusters with three main verbs by extending the structures in (47) and (48) from Subsection A, which are repeated here as (93) and (94), with one main verb that selects a transparent infinitive. We begin with constructions containing an additional subject raising verb since such verbs seem to be the most permissive.

93
a. dat Jan dat boek probeert te lezen.
Control2-Main1
  that  Jan that book  tries  to readinf
  'that Jan is trying to read that book.'
b. dat Jan dat boek schijnt te lezen.
SR2–Main1
  that  Jan that book  seems  to readinf
  'that Jan seems to be reading that book.'
94
a. dat Jan dat boek moet lezen.
Modal2–Main1
  that  Jan that book  must  readinf
  'that Jan must/is obliged to read that book.'
b. dat Jan haar dat boek ziet lezen.
Perc2–Main1
  that  Jan her  that book  sees  readinf
  'that Jan sees her read that book.'
c. dat Jan haar dat boek laat lezen.
Caus2–Main1
  that  Jan her  that book  makes  readinf
  'that Jan makes/lets her read that book.'
[+]  1.  Clusters of the type SR3-V2–Main1

Clusters with three main verbs, in which the highest verb is a subject raising verb like schijnen'to seem' seem to exhibit few restrictions, the main one being that stacking of subject raising verbs, as in (95), is prohibited.

95
a. $ dat Jan dat boek lijkt te schijnen te lezen.
SR3–SR2–Main1
  that  Jan that book  appears  to seem  to readinf
b. $ dat Jan dat boek blijkt te schijnen te lezen.
SR3–SR2–Main1
  that  Jan that book  turns.out  to seem  to readinf

We did not mark these examples with an asterisk because it is not a priori clear whether their unacceptability is due to a syntactic restriction. One argument in favor is that the subject raising verbs lijken'to appear', schijnen'to seem', and blijken'to turn out' seem to resist appearing as non-finite forms more generally; see the discussion in Subsection C. It is also likely, however, that the examples in (95) are excluded because the raising verbs express incompatible or even contradictory evidential information; see Section 5.2.2.2, sub II, for a discussion of the evidential meanings of these verbs.
      Example (96) shows that schijnen can take a projection of a control verb as its complement. Note that there are two options: one in which proberen takes a transparent te-infinitival, represented in (96a), and one in which it takes a semi-transparent te-infinitival, represented in (96b).

96
a. dat Jan dat boek schijnt te proberen te lezen.
SR3–Control2–Main1
  that  Jan  that book  seems  to try  to read
  'that Jan seems to try to read that book.'
b. dat Jan <dat boek> schijnt te proberen <dat boek> te lezen.
SR2–Control1
  that  Jan   that book  seems  to try  to read
  'that Jan seems to try to read that book.'

The examples in (96) show that the linear string dat Jan dat boek schijnt te proberen te lezen can in principle receive two analyses: one analysis in which the te-infinitive is part of the verb cluster, and one with remnant extraposition of the infinitival complement clause of proberen. It raises the question as to whether they are both available. This can be checked by taking into consideration their perfect-tense counterparts in (97a&b): examples (97a) exhibits the IPP-effect, which requires that the te-infinitive be part of the verb cluster; example (97b) does not exhibit this effect, which shows that we are dealing with remnant extraposition. Although these examples are somewhat marked due to their complexity, speakers tend to accept them both. For completeness' sake, we added the fully acceptable extraposition example in (97c): that this example is preferred to the other two may be due to (i) the fact that it has a smaller verb cluster than (97a), and (ii) the fact that it differs from (97b) in that it does not involve the marked option of leftward movement of the object from the extraposed clause. The verb clusters in these examples are in italics.

97
a. ? dat Jan dat boek schijnt te hebben proberen te lezen.
verb clustering
  that  Jan  that book  seems  to have  try  to read
  'that Jan seems to have tried to read that book.'
b. ? dat Jan dat boek schijnt te hebben geprobeerd te lezen.
remnant extraposition
  that  Jan that book  seems  to have  tried  to read
  'that Jan seems to have tried to read that book.'
c. dat Jan schijnt te hebben geprobeerd dat boek te lezen.
extraposition
  that  Jan seems  to have  tried  that book  to read
  'that Jan seems to have tried to read that book.'

Despite the fact that (97a&b) are judged as marked, we provisionally conclude from the discussion above that all three options occur. The examples in (98) show that it is easily possible to embed the constructions with modal, perception and causative verbs in (94) under the subject raising verb schijnen.

98
a. dat Jan dat boek schijnt te moeten lezen.
SR3–Modal2–Main1
  that  Jan that book  seems  to must  readinf
  'that Jan seems to be obliged to read that book.'
b. dat Jan haar dat boek schijnt te zien lezen.
SR3–Perc2–Main1
  that  Jan her  that book  seems  to see  readinf
  'that Jan seems to see her read that book.'
c. dat Jan haar dat boek schijnt te laten lezen.
SR3–Caus2–Main1
  that  Jan her  that book  seems  to make/let  readinf
  'that Jan seems to make/let her read that book.'

We conclude from the discussion above that subject raising verbs do not impose any syntactic restrictions on their infinitival complement; the fact that stacking of subject raising verbs, as in (95), is impossible may have a semantic reason.

[+]  2.  Clusters of the type Control3-V2–Main1

It seems possible to stack control verbs, although the resulting structures may be somewhat "heavy" semantically. An example is constructed in (99a) by adding the verb weigeren'to refuse' on top of the verbs in (93a). The IPP-effect in (99b) shows that the verb weigeren can indeed participate in verb clustering.

99
a. dat Jan dat boek weigert te proberen te lezen.
  that  Jan that book  refuses  to try  to readinf
  'that Jan refuses to try to read that book.'
b. dat Jan dat boek heeft weigeren te proberen te lezen.
  that  Jan  that book  has  refuse  to try  to read
  'that Jan has refused to try to read that book.'

It seems harder to establish, however, that weigeren and proberen together can be verb-clustering verbs in the same structure at the same time, that is, that the te-infinitive te lezen can indeed be part of the verb clusters in (99). This is because we can also analyze these examples as cases in which proberen takes a semi-transparent infinitival clause, that is, in which it is involved in remnant extraposition; cf. the examples in (100) with leftward movement of the direct object of the extraposed complement clause of proberen.

100
a. dat Jan <dat boek> weigert te proberen <dat boek> te lezen.
  that  Jan that book  refuses  to try  to readinf
  'that Jan refuses to try to read that book.'
b. dat Jan <dat boek> heeft weigeren te proberen <dat boek> te lezen.
  that  Jan    that book  has  refuse  to try  to read
  'that Jan has refused to try to read that book.'

Unfortunately, we cannot appeal to the IPP-effect to show that both analyses are available due to the fact that the infinitival complement of weigeren'to refuse' cannot be used in a perfect-tense construction: cf. Jan weigert te zingen'Jan refuses to sing' versus $Jan weigert te hebben gezongen'Jan refuses to have sung'. However, in the absence of evidence that examples (99) cannot be analyzed as verb-clustering constructions, we will provisionally assume that this is in fact a possible analysis (besides the remnant extraposition analysis). There are in fact more complexities involved in the analysis of example (99b). For example, the infinitival complement of weigeren need not be transparent but may also be semi-transparent, as is unambiguously shown by the lack of the IPP-effect in the examples in (101a&b). In these cases it is again not clear whether the verb proberen selects a transparent or a semi-transparent te-infinitival clause. That the latter is at least possible is clear from the fact that the direct object may also follow the verb proberen; this is shown in (101c), in which the complement clause of proberen is underlined.

101
a. dat Jan dat boek heeft geweigerd te proberen te lezen.
remnant extraposition
  that  Jan  that book  has  refused  to try  to read
  'that Jan has refused to try to read that book.'
b. dat Jan heeft geweigerd dat boek te proberen te lezen.
extraposition
  that  Jan  has  refused  that book  to try  to read
  'that Jan has refused to try to read that book.'
c. dat Jan heeft geweigerd te proberen dat boek te lezen.
  that  Jan  has  refused  to try  that book  to read
  'that Jan has refused to try to read that book.'

The discussion above has shown that example (99a) is at least four ways ambiguous with respect to verb clustering and remnant extraposition: the verbs weigeren and proberen can both trigger verb clustering, they can both be involved in remnant extraposition, or they can have different values in this respect. The choice between verb clustering or remnant extraposition can be decided by the IPP-effect in the case of weigeren, but not in the case of proberen, as the latter is not selected by a perfect auxiliary. The postulate of structural ambiguity evoked by the verb proberen is therefore based on the fact that we do not have any compelling reason for assuming that it is not there. For completeness' sake, we want to note that acceptability judgments on the perfect-tense examples in (99) to (101) are not uniform: the (presumed) uniform verb-clustering order in (99b) and the uniform extraposition order in (101c) seem to be best, whereas all other cases are judged to have some intermediate status.
      Example (102) shows that proberen'to try' cannot embed a subject raising construction with schijnen. We mark this example with the dollar sign, because it is not clear whether we are dealing with a syntactic restriction. The reason may again be due to the fact that such control verbs trigger an irrealis reading on their infinitival complements in the sense that they assert something about a potential future event. The unacceptability of (102) may therefore be due to the fact that the verb schijnen does not denote an event, but expresses that the speaker has evidence of some kind for assuming that the proposition embedded under it is true.

102
$ dat Jan dat boek probeert te schijnen te lezen.
Control3–SR2–Main1
  that  Jan that book  tries  to seem  to readinf

      The examples in (103) show that embedding the constructions with modal, perception and causative verbs in (48) under proberen gives rise to varying results. Because examples such as dat Jan haar probeert te zien optreden'that Jan tries to see her perform' seem fully acceptable, we will assume that verb clusters of the form in (103b&c) are syntactically well-formed, and concentrate below on the infelicitous example (103a).

103
a. $ dat Jan dat boek probeert te moeten lezen.
Control3–Modal2-Main1
  that  Jan that book  tries  to must  readinf
  Compare: 'that Jan tries to have to read that book.'
b. ? dat Jan haar dat boek probeert te zien lezen.
Control3–Perc2-Main1
  that  Jan her  that book  tries  to see  readinf
  'that Jan tries to see her read that book.'
c. dat Jan haar dat boek probeert te laten lezen.
Control3–Caus2-Main1
  that  Jan her  that book  tries  to make/let  readinf
  'that Jan tries make/let her read that book.'

The unacceptability of the modal example in (103a) may again be due to the fact that such control verbs trigger an irrealis reading on their infinitival complement, but now we have to consider two different options: one in which the modal has an epistemic "necessity" reading and one in which it has a deontic "obligation" reading. Providing an account for the first case seems relatively simple given that the infinitival modal clause dat boek te moeten lezenis incompatible with an irrealis interpretation: the modal moeten does not denote an eventuality but expresses that the proposition embedded under it will occur in the non-actualized part of the present tense domain. The fact that the deontic "obligation" reading is also impossible in (103a) may be due because it is simply implausible in this context. That there is no general ban on the embedding of infinitival deontic modal clauses under proberen is clear from the contrast between the two examples in (104): whereas it is plausible for Jan to try to obtain permission to come, it seems less plausible for him to try to obtain an obligation to come. The fact that (104a) is fully acceptable shows that verb clusters of the type Control3–Modal2-Main1 are possible.

104
a. dat Jan probeert te mogen komen.
Control3–Modal2-Main1
  that  Jan tries  to be.allowed.to  come
  'that Jan tries to be allowed to come.'
b. $ dat Jan probeert te moeten komen.
Control3–Modal2-Main1
  that  Jan tries  to must  come
  Compare: 'that Jan tries to have to come.'

We conclude from the discussion above that control verbs do not impose any syntactic restrictions on their infinitival complement in verb-clustering contexts; the fact that infinitival complements with an evidential or epistemic modal reading are impossible is due to the fact that they do not satisfy the semantic selection restriction that infinitival complement clauses of control verbs like proberen impose an irrealis interpretation on their infinitival complement.

[+]  3.  Clusters of the type Modal3-V2–Main1

The examples in (105) show that modal verbs like moeten'must/be obliged' can take an infinitival complement containing the control verb proberen, but cannot take infinitival complements containing the subject raising verb schijnen. The unacceptability of (105b) may be due to the fact that subject raising verbs like schijnen resist appearing as non-finite forms more generally but may also point into the direction that their evidential meaning is semantically incompatible with the meaning of modal verbs like moeten. We will leave this issue open here.

105
a. dat Jan dat boek moet proberen te lezen.
Modal3–Control2-Main1
  that  Jan that book  must  try  to readinf
  'that Jan must try to read that book.'
b. $ dat Jan dat boek moet schijnen te lezen.
Modal3–SR2–Main1
  that  Jan that book  must  seem  to readinf

The examples in (106) show that modal verbs of the type moeten'must', kunnen'can' and willen'want' can co-occur in various kinds of combinations: in (106a) we see two dispositional modal verbs, in (106b) an epistemic/dispositional and a dispositional modal verb, in (106c) an epistemic and a non-directed deontic verb, and even more combinations are possible.

106
a. dat Jan die sonate morgen wil kunnen spelen.
M3–M2–Main1
  that  Jan that sonata  tomorrow  wants  can  play
  'that Jan wants to be able to play that sonata tomorrow.'
b. dat Jan dat probleem snel moet kunnen oplossen.
M3–M2–Main1
  that  Jan that problem  quick  must  can  prt.-solve
  'that Jan must be able to solve that problem quickly.'
c. dat het probleem snel opgelost moet kunnen worden.
M3–M2–Main1
  that  the problem  quick  prt.-solved  must  can  be
  'that it must be possible to solve the problem quickly.'

The examples in (107) show that modal verbs can also be combined felicitously with perception and causative/permission verbs.

107
a. dat Jan haar dat boek kan zien lezen.
Modal3–Perc2–Main1
  that  Jan her  that book  can  see  readinf
  'that Jan can see her read that book.'
b. dat Jan haar dat boek moet laten lezen.
Modal3–Caus2–Main1
  that  Jan her  that book  must make  readinf
  'that Jan must make/let her read that book.'
[+]  4.  Clusters of the type Perception3-V2–Main1

The examples in (108) show that perception verbs like zien'to see' can take an infinitival complement containing the control verb proberen as its complement, but not infinitival complements containing the subject raising verb schijnen. The unacceptability of (108b) may be due to the fact that subject raising verbs such as schijnen resist appearing as non-finite forms more generally but may also point into the direction that they are semantically incompatible with perception verbs given that they do not refer to an eventuality that can be directly observed by means of the senses (here: vision).

108
a. dat Jan haar een fiets ziet proberen te stelen.
Perc3–Control2-Main1
  that  Jan her  a bicycle  sees  try  to readinf
  'that Jan sees her try to steal a bicycle.'
b. $ dat Jan haar een fiets ziet schijnen te stelen.
Perc3–SR2-Main1
  that  Jan her  a bicycle  sees  seem  to stealinf

The examples in (109) show that adding a perception verb to the constructions with modal, perception and causative verbs in (48) gives rise to varying results. Example (109a) is unacceptable, and this may have a similar reason as the unacceptability of (108b): bare infinitival complements with a modal verb do not refer to eventualities that can be directly observed by means of the senses. The markedness of examples like (109b&c), which are normally taken to be grammatical (cf. Kroch and Santorini 1991), does not seem syntactic in nature, but is probably due to the computational complexity of these examples: in the English renderings of these examples it is very easy to identify the relevant verb-subject pairs on the basis of linear order because the subject is always left-adjacent to the verb it belongs to, but in Dutch the identification requires for each pair that various elements be skipped. The difficulty of establishing the relevant pairs (even in reading) suggests that there might be a psycholinguistic reason for the markedness of Dutch examples like (109b&c).

109
a. $ dat Jan haar dat boek zag moeten lezen.
  that  Jan her  that book  saw  must  read
  Compare: 'that Jan saw her be obliged to read that book.'
b. ? dat Jan Marie de merel zag horen zingen.
  that  Jan Marie the blackbird  saw  hear  leave
  'that Jan saw Marie hear the blackbird sing.'
c. ? dat Jan Marie haar hond zag laten zwemmen.
  that  Jan Marie her dog  saw  make  swim
  'that Jan saw Marie make/let her dog swim.'
[+]  5.  Clusters of the type Causative3-V2–Main1

The examples in (110) show once again that causative laten'to make/let' can take an infinitival complement with the control verb proberen as its complement, but that infinitival complements with a subject raising verb such as schijnen are impossible. The unacceptability of (110b) may be due to the fact that subject raising verbs like schijnen resist appearing as non-finite forms more generally but may also point into the direction that they are semantically incompatible with causative laten due to the fact that they do not denote activities.

110
a. dat Jan Marie dat boek laat proberen te lezen.
Caus3–Control2-Main1
  that  Jan Marie that book  makes  try  to readinf
  'that Jan lets Marie try to read that book.'
b. $ dat Jan Marie dat boek laat schijnen te lezen.
Caus3–SR2-Main1
  that  Jan Marie that book  makes  seem  to readinf

The examples in (111) show that adding the causative verb to the constructions with modal, perception and causative verbs in (94) gives rise to varying results. Example (111a) is unacceptable and this may again have a semantic reason: epistemic modal verbs do not refer to an eventuality, and the use of deontic modals simply seem to give rise to implausible scenarios. Examples such as (111b) are perhaps somewhat marked but we do find examples of this type (often with one or more omitted lower subjects) on the internet: cf. Die twee jongens die ze laten horen zingen with the intended reading "those two boys that they make (us) hear sing". Examples such as (111c) are again difficult to interpret, although we found surprisingly many cases of this sort on the internet (mainly with the two lower subjects omitted) with infinitives like registreren'to register', vastleggen'to record', and onderzoeken'to investigate' by searching for the string [laten laten]: cf. Die buurgemeente heeft in 2007 laten laten onderzoeken hoe ... (approximately: In 2007 the local authority has made someone make someone investigate how ...').

111
a. $ dat Jan haar dat boek laat moeten lezen.
  that  Jan her  that book  makes  must  read
  Compare: 'that Jan lets/makes her be obliged to read that book.'
b. ? dat Jan Marie de merel laat horen zingen.
  that  Jan Marie the blackbird  makes  hear  sing leave
  'that Jan makes Marie hear the blackbird sing.'
c. ?? dat Jan Marie haar hond laat laten zwemmen.
  that  Jan Marie her dog  makes  make  swim
  'that Jan makes Marie make/let her dog swim.'
[+]  6.  Summary

Table 3 summarizes the findings of this subsection. Virtually all verb clusters in the examples in (93) and (94) from the introduction to this subsection (p.) can be extended by means of an additional main verb, although clusters with embedded subject raising verbs are exceptional. The exceptional status of raising verbs may be due to a restriction that disfavors these verbs to appear as non-finite forms (although they do occur in more formal texts), but we have seen that there are also reasons for assuming that semantic restrictions are at play; we leave it to future research to investigate this option. Modal verbs are exceptional in that they cannot be embedded under perception and causative verbs; we have seen that there are again reasons for assuming that this is due to the fact that perception and causative verbs require that their infinitival complement refers to an eventuality. Clusters marked with one or more question marks seem grammatical but are not very felicitous, and we have suggested that this may be due to the computational complexity that they involve.

Table 3: Verb clusters with three main verbs
Control3 SR3 Modal3 Perc3 Caus3
Control2-Main1 + + + + +
SR2–Main1
Modal2–Main1 + (deontic) + +
Perc2–Main1 + + + ? ?
Caus2–Main1 + + + ? ??
[+]  E.  Conclusion

This section has investigated the restrictions on the hierarchical order of verbs in verb clusters. The main issue is: What types of verbal projections can be selected by what types of verbs? The discussion has shown that there are restrictions on what counts as acceptable verb combinations. It seems, however, that these restrictions are not of a syntactic, but of a semantic/pragmatic nature. For example, we have seen that the fact that subject raising verbs like schijnen normally cannot be embedded under other verbs may be related to the fact that such raising verbs do not denote eventualities but express evidential modality. We have seen that other restrictions may be of a pragmatic nature, or may even be related to computational complexity.

References:
  • Barbiers, Sjef, Bennis, Hans, Vogelaer, Gunther de, Devos, Magda & Ham, Margreet van de2008Syntactic atlas of the Dutch dialectsAmsterdamAmsterdam University Press
  • Bennis, Hans & Hoekstra, Teun1989Why Kaatje was not heard sing a songJaspers, Danny, Klooster, Wim, Putseys, Yvan & Seuren, Pieter (eds.)Sentential complementation and the lexiconDordrechtForis Publications21-40
  • Koster, Jan1984Anaphoric and non-anaphoric controlLinguistic Inquiry15417-459
  • Kroch, Anthony S. & Santorini, Beatrice1991The derived constituent structure of the West Germanic Verb-Raising constructionFreiding, Robert (ed.)Principles and paramaters in comparative grammarCambidge, Mass/LondonMIT Press269-338
  • Petter, Marga1998Getting PRO under control. A syntactic analysis of the nature and distribution of unexpressed subjects in non-finite and verbless clausesAmsterdamFree University AmsterdamThesis
  • Schmid, Tanja & Vogel, Ralf2004Dialectal variation in German 3-verb clusters: a surface-oriented optimality theoretic accountThe Journal of Comparative Germanic Linguistics7235-274
report errorprintcite