This subsection discusses finite monoclausal structures with three verbs in which the perfect auxiliary appears as a non-finite verb. At first glance, such structures do not seem very special: (i) the perfect auxiliary governs the main verb, which appears as a past participle, and (ii) the past participle can either precede or follow the auxiliary (as in the embedded clauses with two verbs discussed in Subsection III). The examples in (21) also show that the main verb cannot be realized as an infinitive.

With regard to the order of the auxiliary and past participles, the same caveat must be made as in Subsection III, namely that the aux -part order is found only in a limited part of the Dutch-speaking area, which happens to include the more prestigious varieties of the standard language spoken in the western/middle part of that area. More generally, the part-aux order seems to be the more common one in speech, although we should mention that, to our knowledge, the variation in word order of clause-final verbs in main clauses with three verbs has not been systematically investigated. The subscripts in (22) are added for convenience to indicate whether the verb in question is finite, infinitive, or a past participle.

However, the examples in (23) show that the placement options of the past participle in embedded clauses are somewhat surprising. Since the participle is governed by the auxiliary, we would expect these verbs to be adjacent, but in fact they can be separated by the finite modal verb.

For many speakers, the three word orders can be seen as more or less free alternates, with the Vfin–aux–part order moet hebben gelezen being the more marked one. That this order is the more marked one seems to be confirmed by the regional distribution of these orders given in Table (24) for the sequence moet hebben gemaaktmust have made; whereas speakers regularly indicate that they accept only one of the orders in (24b-d), there is only one speaker who indicates that (s)he accepts only (24a). Speakers who report that they only allow (24b) are mainly found in Flanders, while speakers who report that they only allow (24c) are scattered throughout the Netherlands. The low frequency of order (24d) is due to the fact that it is only found in the northern parts of the Netherlands, which in turn may be related to the fact that this is the order usually found in Frisian (as well as standard German). The data in (24) is taken from Barbiers et al. (2008: §1.3.2.2).

	order of verbs	Total #	Total # as only order
a.	Vfinite–Aux–Part (moet hebben gemaakt)	91	1
b.	Vfinite–Part–Aux (moet gemaakt hebben)	163	48 (Flanders)
c.	Part–Vfinite–Aux (gemaakt moet hebben)	186	28 (Netherlands)
d.	Part– Aux-Vfinite (gemaakt hebben moet)	48	30 (Northern Netherlands)

The literature reviewed in Haeseryn (1990:54ff) also suggests that the order Vfin–part–aux order is particularly popular in the varieties of Dutch spoken in Flanders, while speakers from the Netherlands generally prefer the order part–Vfin–aux; cf. Stroop (2009) for the same finding based on the Corpus of spoken Dutch. The order Vfin–aux–part is again characteristic of (but not limited to) written and formal Dutch.

Clusters of more than three verbs are possible, but less common in colloquial speech. If the auxiliary immediately governs the (most deeply embedded) main verb, the principles underlying the form of the main verb and the order of the verbs are the same as in the case of three verbs: the main verb appears as a past participle, which can occur as the last verb of the verb cluster, but also further to the left. This is illustrated in (25) for the cluster zou kunnen hebben gezienmight have seen.

To our knowledge, there is not much information about the distribution of the orders in (25). The literature reviewed in Haeseryn (1990:70ff) suggests that the orders in (25a&d) are the ones usually found in the northern varieties of standard Dutch, and that the order in (25c) is considered better than the order in (25b). In the varieties of standard Dutch spoken in Belgium, on the other hand, the order in (25b) seems to be common.

The discussion above has shown for the northern varieties of standard Dutch that in perfect-tense constructions of the kind under discussion, the past participle of the main verb can follow or precede the entire verb cluster, or be placed between any two verbs in the verb cluster. This is illustrated in (26), where V* stands for zero or more verbs in the verb cluster besides the auxiliary and the main verb; the angled brackets indicate the alternative placements of the participle.

Although Barbiers et al. (2008) has shown that other orders can be found in certain dialects of Dutch, the orders in (26) exhaust the possibilities for the vast majority of Dutch speakers. In fact, most speakers will use only a subset of the word-order possibilities in (26). Recall that clusters of more than three verbs are rare in everyday speech, and even in formal speech and complex written language the number of verbs is usually limited to a maximum of 4 or 5.

This subsection discusses finite monoclausal structures with three verbs in which the perfect auxiliary appears as the finite verb. Such structures arise not only when the auxiliary governs a non-main verb, like the (semi-)aspectual verbs gaan and zitten in (27a&b), but also when it governs a main verb that selects a transparent infinitival clause, like the deontic modal verb moetenbe obliged in (27c) or the perception verb ziento see in (27d).

The most striking phenomenon in examples like (27) is the so-called infinitivus-pro-participio (IPP) effect, i.e. that the non-finite verb governed by the auxiliary appears not as a past participle but as an infinitive: the examples in (28) illustrate this by showing that substituting a past participle for the corresponding infinitival verbs in (27) leads to ungrammaticality.

Another characteristic is that the word order of the verb cluster is very strict in most northern varieties of Dutch. In main clauses like (27), the verb selected by the perfect auxiliary must precede the main verb: the examples in (29) show that reversing the order of the two clause-final verbs leads to degraded results.

In embedded clauses the word order is also very strict. This applies not only to the two non-finite verbs, which again have the order in (27), but also to the finite auxiliary and the two infinitival verbs; the auxiliary must precede them.

Any other order than in (30) leads to a severely degraded result. This implies that the perfect-tense constructions under discussion here differ markedly from the perfect-tense constructions discussed in Subsection A in that the auxiliary cannot be preceded by the verb it immediately dominates. We illustrate this in (31) for the modal construction in (30c): the auxiliary cannot be preceded by the modal regardless of the position of the more deeply embedded main verb.

In short, it seems that in the northern varieties of standard Dutch, the verb clusters under discussion can only be realized in the order in (32a); all other logically possible orders are severely degraded. This is remarkable, since Barbiers et al. (2008) show that the orders marked with a percentage sign are relatively common in certain varieties of Dutch: the order in (32e) is found in Flanders, and the order in (32f) in the northern part of the Netherlands, especially in Frisian. The order in (32b) is relatively rare, but is reported by various speakers around the IJsselmeer; it is also the order usually found in standard German. The orders marked with an asterisk are rare and do certainly not occur as the dominant orders.

After the discussion above, it will come as no surprise that in longer verb clusters with IPP, the order of the verbs is also very strict. We illustrate this in (33) and (35) for verb clusters consisting of four verbs. The examples in (33) differ from those in (30) in that we have added an epistemic modal verb, which appears as the finite verb. Any change in the order of the verbs will lead to a degraded result in the northern varieties of standard Dutch.

The southern varieties of standard Dutch, on the other hand, have more possibilities. This is illustrated by the examples in (34), taken from Haeseryn (1990:72). While the relevant northern varieties of Dutch only allow the order in (34a), the orders in (34b&c) are common in the varieties found in Belgium; the order in (34a) is reported to also be possible in these varieties.

The examples in (35) differ from those in (30) in that we have added a deontic/dynamic modal verb that appears as an infinitive (either before or after the non-finite verb originally dominated by the auxiliary); examples with two non-epistemic modals, such as (35c), are perhaps a bit marked, but can be easily found on the internet. Any change in the order of the verbs will produce a degraded result in the northern varieties of Dutch.

Some of our Flemish informants also allow the perfect auxiliary in final position. In their variety, an example such as (35a) would appear as dat Marie moeten gaan zwemmen heeft/is, where the use of zijn is due to the fact that the choice of the auxiliary is preferably determined by the more deeply embedded aspectual verb gaan; cf. Subsection II.

Dutch is an OV-language in the sense that nominal objects always precede their main verb in clause-final position: dat Jan <een boek> leest <*een boek>that Jan is reading a book. The northern varieties of Dutch have the additional restriction that nominal arguments can never permeate the verb cluster. This means that (in)direct objects can only precede the verb cluster as a whole. The examples in (39) illustrate this for cases with two verbs, the perfect auxiliary and a main verb in the form of a past participle.

Since the southern varieties of Dutch are not subject to the additional restriction that nominal arguments cannot permeate the verb cluster, one might expect that the order marked as ungrammatical above would occur in these varieties. This is not the case, however, for the independent reason that these varieties require the past participle to precede the auxiliary; the aux-part order heeft gelezen in (39b) simply does not occur in these varieties, leaving (39a) as the only option.

In the northern varieties, the object also precedes verb clusters consisting of more than two verbs. The examples in (40) illustrate this for a sequence of three verbs in which the auxiliary is an infinitive: although the past participle gelezenread can occur in several positions in the verb cluster, the nominal object must precede the verb cluster as a whole.

Since the order of the verb cluster moet gelezen hebben is acceptable in the southern varieties of Dutch, we expect speakers of these varieties to accept the order marked with a percentage sign in (40b). The judgments of our Flemish informants vary: some of them categorically reject such examples, while others accept them provided that the object is indefinite. The fact that the order marked with a percentage sign is unacceptable for all southern speakers when the object is definite may be due to the fact that definite noun phrases are more likely to be construed as presuppositional and thus more likely to be A-scrambled to a more leftward position; cf. Sections 13.2 and N21.1.4 for a discussion of this form of scrambling. Note that the order moet hebben gelezen is ungrammatical in the southern varieties due to the placement of the participle; thus, example (40c) does not tell us anything about the restrictions on permeation of the verb cluster.

Finally, example (41) provides an example with three verbs in which the perfect auxiliary is the finite verb. Although the infinitival main verb lezen can only occur at the end of the verb cluster, most speakers from the Netherlands require its object to precede the entire verb cluster.

Again the judgments of our Flemish informants vary somewhat, but they all agree that permeation of the verb cluster is possible (for some only as a marked option), provided that the object is adjacent to the main verb; if the main verb and its object in (41) are separated by the infinitive moeten, the result is unacceptable. see also Haegeman & Van Riemsdijk (1986:422ff) for further examples from West Flemish.

Prepositional objects differ from nominal objects in that they do not have to precede the main verb in clause-final position, but can also follow it: dat Jan <op zijn vader> wacht <op zijn vader>that Jan is waiting for his father. However, they are like nominal objects in that they never permeate the verb cluster in the northern varieties of Dutch. This means that prepositional objects must either precede or follow the verb cluster as a whole. The examples in (42) illustrate this for cases with two verbs, a perfect auxiliary and a main verb in the form of a past participle; in (42a) we find the part-aux order and in (42b) the aux-part order.

The examples in (43) illustrate the same for cases with three verbs in which the auxiliary is an infinitive. Although the past participle gelezenread may be placed in several positions in the verb cluster, the prepositional object must either precede or follow the entire verb cluster; the prepositional object cannot occur in the positions marked by <*> or <%>.

Since the order of the verb cluster moet gewacht hebben is acceptable in the southern varieties of Dutch, we expect that speakers of these varieties will allow the word order marked by the percentage sign in (43b). Again, the judgments of our southern informants vary: some indicate that they would not use it, while others indicate that they fully accept this word order. Note that permeation of the verb cluster is only possible if the prepositional object precedes the main verb: placing the PP immediately after the main verb in (43b) is categorically excluded. Since the order moet hebben gewacht in (43c) is not acceptable in the southern varieties, this example does not tell us anything about the restrictions on permeation of the verb cluster.

Finally, example (44) gives a similar case with three verbs in which the auxiliary is finite. Although the infinitival verb lezen can only occur at the end of the verb cluster, the northern varieties of Dutch require that prepositional objects either precede or follow the verb cluster as a whole; they cannot permeate the verb cluster.

The percentage signs between angled brackets in (44) indicate that, as in the case of indefinite noun phrases, our southern informants do accept permeation of the verb cluster. However, they disagree on whether the prepositional object must be adjacent to the main verb; all of them require it, except for one speaker who prefers the position before moetenmust.

Object clauses differ from nominal and prepositional objects in that they obligatorily follow the main verb in clause-final position; examples such as (45b) are possible, but trigger a so-called factive interpretation on the embedded clause; cf. Section 5.1.2.3. In the following we will ignore such factive clauses.

Like nominal and prepositional objects, object clauses never permeate the verb cluster, which means that object clauses can only follow the verb cluster as a whole. The examples in (46) illustrate this for cases with two verbs, a perfect auxiliary and a main verb in the form of a past participle; the positions that do not accept the object clause are marked with <*>.

Example (47) provides similar cases with verb clusters of three verbs; placing the object clause anywhere further to the left will produce an unacceptable result.

Complementives have a similar distribution as nominal objects. First, they must precede the main verb; cf. dat Els het hek <oranje> verft <*oranje>that Els is painting the gate orange. Second, in the northern varieties of Dutch, complementives usually cannot permeate the verb cluster; see below for a more precise formulation of this claim. This means that complementives usually precede the verb cluster as a whole. The examples in (48) illustrate this for cases with two verbs, a perfect auxiliary and a main verb in the form of a past participle.

The examples in (49) illustrate the same for cases with three verbs in which the auxiliary is an infinitive: although the northern varieties of Dutch allow the past participle geverfdpainted in several positions, the complementive must precede the entire verb cluster. Since the southern varieties of Dutch allow the order of the verb cluster zou geverfd hebben, we expect that they also allow the word order marked by the percentage sign in (49b), and indeed our southern informants unanimously do.

Finally, example (50) provides a similar case with three verbs in which the auxiliary is a finite verb. Although the infinitival verb vervento paint can only occur at the end of the verb cluster, speakers of the northern varieties of Dutch require that the complementive precedes the verb cluster as a whole.

Our southern informants accept permeation of the verb cluster in such constructions without any problem, although they differ a little on whether the complementive must be adjacent to the main verb; most speakers require it, but one speaker merely prefers it and thus also accepts the order marked with an asterisk.

Section 2.2 has argued that verbal particles can also be considered complementives, and we therefore expect them to have the same distribution as the adjectival complementive oranje in the examples above. This is indeed true in the sense that they must precede the main verb: cf. dat Peter zijn moeder <op> belt <*op>that Peter calls his mother. However, unlike adjectival complementives, the northern varieties of Dutch allow verbal particles to permeate the verb cluster. This is illustrated in (51b).

That verbal particles can permeate the verb cluster is also clear from the examples in (52b-c) and (53). Although speakers may have different preferences, all the given orders seem to be acceptable; cf. Bennis (1992) and Koopman (1995). Note, however, that some speakers consider the orders marked with the diacritic (?) to be less felicitous, which is reminiscent of the fact that most southern speakers allow permeation of the verb cluster only when the adjectival complementive is adjacent to the main verb; cf. the discussion of (50) above.

The contrast between examples with an adjectival complementive and with a verbal particle is perhaps not as surprising as one might think At first glance, given that some speakers of the northern variety allow adjectival complementives to permeate verb clusters if they consist of a single syllable: many of the orders marked as unacceptable in (48) to (50) improve considerably if we replace the polysyllabic adjective oranjeorange by the monosyllabic adjective geelyellow. Although the orders marked with (?) sometimes elicit a negative response, many speakers accept all of them.

Speakers’ judgments seem to diverge more for verb clusters containing more than three verbs. While Bennis (1992) claims that particles can be placed anywhere in the verb cluster in (55a) as long as they precede the main verb, Koopman (1995) does not accept the placement indicated by %. A similar divergence of judgments occurs in examples such as (55b) with complementives.

The nominal part of N + V collocations such as paardrijdento ride a horse and pianospelento play the piano can also permeate verb clusters as long as it precedes the verbal part of the collocation.

The placement options for the noun are thus more or less the same as for particles and monosyllabic complementives. This can also be seen in three-verb clusters; the noun can occur anywhere in the cluster as long as it precedes the main verb, and again some speakers seem to dislike the word orders marked with (?) in (57b''&c).

Again, judgments diverge in clusters of more than three verbs; some speakers allow all orders, while others do not accept the placement indicated by %.

Manner adverbs also seem to be part of the lexical projection of the main verb. Like nominal arguments, they must precede the main verb: dat Marie <snel> vertrok <*snel>that Marie left quickly. The northern varieties of Dutch have the additional restriction that manner adverbs never permeate the clause-final verb cluster, i.e. they must precede the entire verb cluster. The examples in (59) illustrate this for cases with two verbs, a perfect auxiliary and a main verb in the form of a past participle.

The examples in (60) illustrate the same for cases with three verbs in which the auxiliary is an infinitive: although the northern varieties of Dutch allow the past participle vertrokkenleft in several positions, the manner adverbial must again precede the entire verb cluster.

As in the previous cases, we expect our southern informants to accept permeation of the verb cluster, provided that the manner adverbial is adjacent to the main verb; however, the word-order option marked with a percentage sign is reported to be degraded if it is acceptable at all. Since the order moet zijn vertrokken in (60c) is not acceptable in the southern varieties, this example does not tell us anything about the restrictions on permeation of the verb cluster.

Example (61) provides a similar case with three verbs in which the auxiliary is a finite verb. Although the infinitival verb vertrekken can only occur at the end of the verb cluster, speakers of the northern varieties of Dutch require the manner adverb to precede the entire verb cluster.

Most of our Flemish informants indicate that, as expected, permeation of the verb cluster is acceptable provided that the manner adverbial is adjacent to the main verb; cf. also Haegeman & van Riemsdijk (1986:443) for similar examples with the negative adverb nienot from West Flemish. However, one informant prefers the order marked with an asterisk.