The italicized conditional clauses in (131) show that verb-second is optional: if the conditional clause is introduced by the linker element alsif, the finite verb occurs in clause-final position, but if als is not present, the finite verb must be clause-initial. There is reason to believe that the latter is only possible in one specific context, viz. when the conditional clause is part of a left-dislocation construction; cf. e.g. Den Besten (1983:fn.3), Haeseryn et al. (1997: §21.8), and Beekhuizen (2008); .

That verb-first cannot apply in ordinary clause-internal adverbial clauses can be shown in at least two ways. First, the examples in (132) show that verb-first is marked when the resumptive element dan is absent. Example (132b) is given a percentage sign to indicate that this structure cannot easily be used to express the intended conditional reading; for the moment we will ignore that some speakers seem to allow this form, but we will return to this in Subsection V.

Second, the examples in (133) show that verb-first is also excluded when the adverbial clause is in clause-final position.

A generalization that more or less presents itself on the basis of the examples in (131)-(133) is that conditional adverbial clauses allow verb-first only if they are clause-external. This is the case in left-dislocation constructions such as (131), where the clause-initial position of the main clause is occupied by the resumptive element danthen, but not in examples such as (132), where the conditional clause itself occupies the clause-initial position, or examples such as (133), where it occurs in clause-final position. The structures of these examples are thus as sketched in (134).

However, the examples in (135) show that verb-first is also excluded in parenthetical conditional clauses; since it can be argued that parenthetical clauses are not structurally embedded in the main clause, this shows that being external to the main clause cannot be considered a sufficient condition for allowing verb-first; cf. Reuneker (2017) for a more detailed discussion of cases such as (135a).

Note that we can identify parenthetical clauses by adding the phrase tenminsteat least; in (131) and (132a), this would lead to severely degraded results, but it is easily possible in (135a). It is also possible in (133a), but then we are dealing with an afterthought, which is also indicated by an intonation break before the adverbial clause.

That left-dislocated phrases are indeed clause-external is also shown by examples like (136) and (137). In (136), the main clause is an imperative, and since imperative clauses always have the finite verb in first position, the als-clause cannot be clause-internal.

The same holds for (137), in which the main clauses are yes/no questions.

The V1-requirement of the main clauses in (136) and (137) makes it necessary to place the resumptive element dan in the middle field of the clause. This option is not available in the declarative main clauses in (138): the resumptive element must be placed in clause-initial position as in the acceptable examples in (131) above; cf. also Section 11.1, sub VIII, and Greco & Haegeman (2020).

The hypothesis that verb-first is only possible if the conditional adverbial clause is left-dislocated predicts that embedding the two examples in (131) will not lead to acceptable results, because left dislocation is a property of root clauses. The unacceptability of (139b) shows that this is indeed what we find for (131b). The case for (131a) is less straightforward, given the acceptability of (139a), but the fact that the addition of the resumptive element dan is impossible (regardless of its position in the matrix clause) shows that a left-dislocation analysis is not appropriate. The fact that the addition of tenminsteat least to the conditional clause is possible suggests that we are dealing with a parenthetical clause; cf. the discussion of (135). The same is shown by the fact that the clause forms a separate intonational phrase, which can be separated from the rest of the clause by intonation breaks.

For completeness’ sake, note that the addition of the resumptive linking element danthen to example (139a) leads to unacceptability, which shows that we that we cannot dealing with embedded left dislocation; the acceptability contrast in (139) is thus fully consistent with the hypothesis that verb-first is only possible in left-dislocated clauses.

Before concluding this subsection, we will briefly address two issues that may complicate the study of conditional V1-clauses, but which have received little attention in the syntactic literature. First, the argument based on embedding is complicated by the fact that, besides examples such as (139a), it is often possible to have constructions such as (140a) with two complementizers dat and the resumptive element dan. It is not clear a priori (i) whether such an example should be seen as the embedded counterpart of (131a), or (ii) whether we are dealing here with a performance phenomenon: the processing of the embedded clause in (139a) may be hampered by the lengthy interruption of the parenthetical conditional clause, and the resumption of the part preceding the parenthetical clause may therefore be seen as a repair strategy. The fact that example (131b) has no such counterpart is unexpected under the first approach and thus favors the second approach.

Note that although examples such as (140a) may seem rather odd at first, our Google search (May 30, 2024) on the string [dat als * dat ... dan], with a pronoun in place of the dots, shows that such cases are not uncommon; cf. also Reuneker (2017) for an example from the Corpus of spoken Dutch. We refer the reader to Section C37.2 for a discussion of a wider range of such utterances.

A second complicating issue is that in coordinate structures such as (141) verb-second may apply in the second conjunct if the linker als is not realized; cf. Haeseryn et al. (1997:1252). At first glance, this seems to confirm the previously established fact that the position of the finite verb in left-dislocated conditional clauses depends on the presence of als, but closer inspection reveals that the second conjunct in (141b) differs conspicuously from the cases discussed earlier in that its clause-initial position is filled by the subject; example (141c) shows that this is normally excluded in conditional clauses.

This raises the following question: are we really dealing with coordination in (141b), or should the supposed second conjunct be analyzed as a parenthetical clause, i.e. should (141b) be analyzed along the line of (142a) or along the line of (142b)? The latter analysis is supported by the fact that of ik twijfel erover can be separated phonetically from the rest of the sentence, but since we have no more insights to offer, we will leave this question to future research.

Putting these two complicating issues aside for the moment, we may conclude that the generalization that verb-first/second is excluded in embedded clauses can be maintained. However, the research question we still need to answer is not “how is it that certain types of embedded adverbial clauses sometimes exhibit verb-first/second” but rather “how is it that left-dislocated clauses can sometimes take the form of either a main clause or a non-main clause”?

The conditional construction in (143a) and the contrastive construction in (143b) are similar in that the V1-clauses are not part of the main clause. This is clear from the fact that the initial position of the main clause is filled by another constituent: the resumptive element dan in (143a) and the subject Jan in (143b). The primed examples show that the V1-clauses themselves cannot occupy the main-clause initial position; recall that we postponed the discussion of the fact that some speakers seem to allow (143b') to Subsection V.

At first glance, the primeless examples in (144) seem to show that the two V1-clauses in (143) both alternate with adverbial clauses introduced by a complementizer and with the finite verb in clause-final position. A closer look, however, shows that this is not the case. The optionality of dan in (144a) shows that the als-clause could be either left-dislocated or clause-internal, i.e. in the initial position of the main clause. Of course, only the left-dislocated clause can be considered as an alternant of the similarly left-dislocated V1-clause in (143a). The fact that the terwijl-clause in (144b) obligatorily triggers subject-verb inversion in the main clause shows that it occupies the clause-initial position and consequently cannot be seen as an alternant of the left-dislocated V1-clause in (143b).

The examples in (144) thus show that the alternation occurs only in the conditional construction. This should be related to another striking difference between the two constructions; while Subsection I has shown that the resumptive element dan is obligatory in the conditional construction, resumption does not seem possible in the contrastive construction, as shown in (145b).

This suggests that while the conditional V1-clause (indirectly) plays a semantic role in the main clause, this is not the case for the contrastive V1-clause because it is not connected to the main clause by formal means (such as resumption).

The fact that the syntactic tie between the two clauses is tighter in the conditional construction than in the contrastive construction is reflected in the semantics of the two constructions. In the conditional construction there is an intimate relationship between the truth of the propositions expressed by the V1-clause and the main clause, which can be expressed in propositional calculus by the material implication in (146a). In the contrastive construction, on the other hand, the V1-clause and the main clause are used to independently assert a proposition, as expressed by the conjunction in (146b).

The crucial difference between the two formulas in syntactic terms can be expressed in terms of subordination and conjunction: material implications are syntactically expressed by a single clause embedding the antecedent of the material implication (i.e. the conditional clause or a resumptive pronoun referring to it), whereas conjunctions are typically expressed by independent, conjoined clauses.

Subsection I has shown that the resumptive element dan in conditional constructions must occupy the clause-initial position of a declarative main clause; cf. the contrast between the examples in (147). This would imply that the initial position plays a special role in the connection of the clauses.

Although there is no resumptive element in the contrastive construction, it seems that there are also restrictions on the element in the first position of the declarative main clause. In order to clarify this, we need to elaborate a bit on the meaning of the construction. As the name of the construction suggests, the key issue is the notion of contrast. What is contained in this notion can be clarified by considering the examples in (148); the notion of contrast applies to the relation between the italicized phrases, and the underlined phrases are in the initial positions of the main clauses; cf. Beekhuizen (2008).

The italicized elements in fact function as contrastive topics in the sense that the non-italicized parts of the clauses provide mutually incompatible comments on these elements: for example, the comments in (148b) can be translated as the lambda expressions λx poor(x) and λx rich(x), which are mutually incompatible in the sense that lambda conversion cannot involve a single entity e, as is clear from the fact that the formula poor(e) & rich(e) is contradictory. The semantic function of the topical elements is to add information that resolves the contradiction, as is clear from the fact that the informal predicate logic translations of the examples in (148) given in (149) are fully coherent; t and p stand for time and place, respectively.

Beekhuizen (2008) notes that in some cases the relevant notion is not contrast but unexpectedness (we will not use Beekhuizen’s term concessiveness here to avoid confusion with the construction discussed in Subsection IV). For instance, the comments in example (150a) are not contradictory but tautological in nature. For example, the formula good soprano(e) & able to sing well(e) is tautological in the sense that the denotation of good soprano is contained in the denotation of able to sing well. Again, the topical elements resolve the tautology, as shown in the informal predicate logic translation in (150b). Note that concessive examples can often be recognized by the fact that the topical element in the main clause is or can be preceded by the focus particle ooktoo; adding this particle to the contrastive examples in (148) leads to a semantically incoherent result.

In his newspaper corpus, Beekhuizen found that the topics usually refer to entities (including individuals) and aspects of the spatiotemporal settings of the propositions expressed by the two clauses. Given the semantic discussion above, this is not surprising, as these settings are particularly suitable for resolving the contradictory or tautological nature of the comments. Beekhuizen also found that in more than 90% of the attested cases the initial position of the declarative main clause is occupied by the topical element. That this position is a designated position for such elements is also clear from the fact, illustrated in (151), that changing the word order of the main clauses leads to less felicitous results. We have used the diacritic “$” to express this, because the main clauses are perfectly acceptable without the contrastive V1-clauses, and consequently there is no a priori reason to assume that the examples in (151) are syntactically ill-formed; italics and underlining are used in the same way as in (148).

The fact that the topic can occupy the initial position of the declarative main clause is again not surprising, since contrastive topic/focus elements are often found in this position; cf. Section 11.3.2. Nevertheless, it is noteworthy that the use of contrastive accent does not improve the examples in (151), while contrastive accent in contrastive coordinate structures of the kind in (152), with small caps indicating the accent, allows the contrasted elements to occur in the middle field of the clause.

This contrast between the two construction types suggests that we are dealing with two different kinds of contrastive constructions: while the declarative clauses in (148) are contrastive topic constructions, the declarative clauses in (152) are probably contrastive focus constructions. The difference between the two constructions is that contrastive topic constructions have not only a contrastive accent on the topic element, but also an additional accent in the comment of the clause, as shown in (153).

The crucial observation is that it is difficult to get this intonation pattern with two contrastive accents in examples such as (152b), where the first contrasted element occupies a position in the middle field of the clause: this is shown in (154).

We conclude that the acceptability contrast between the examples in (148) and (151) is crucially linked to the difference between constructions with a contrastive element in the clause-initial position and constructions with a contrastive element in the middle field; we leave the details of the analysis to future research.

Beekhuizen further found that the associate of the topical element in the contrastive V1-clause often precedes the subject, as in (155a). The contrast between the two examples in (155) shows that this is not always possible, but that it depends on the information-structural status of the subject: while definite subjects can follow the adverbial phrase in 2013 if they are part of the discourse-new information, this is impossible for subject pronouns like hijhe, which are always part of the presupposition (discourse-old information). This fits well with the word-order generalizations discussed in Section 13.2.

In the examples above, the contrastive topics in the main clause have the same syntactic function as their associate in the contrastive V1-clause. However, example (156) shows that this need not be the case, as we find here that a subject is contrasted with an agentive door-PP. This shows that it is sufficient for the contrasted topics to have a similar semantic function.

This section has shown that contrastive V1-clauses are external to the main clause and therefore do not constitute counterexamples to the generalization that dependent clauses do not allow V-first/second. We have also seen evidence that such V1-clauses differ from conditional V1-clauses in at least two ways: they are not left-dislocated, and they do not alternate with adverbial clauses introduced by a linker. Due to the lack of resumption, contrastive V1-clauses have a less tight syntactic relation to the following main clause than conditional V1-clauses.

The exclamative constructions in (157) are slightly adapted from Van der Horst & Van der Horst (1999:266). Examples like these are characterized by a typical exclamative intonation pattern; small caps indicate contrastive accent and the exclamation mark the exclamative intonation contour. Exclamative constructions are normally used to express an emotional attitude of the speaker towards the propositional content: amazement, vexation, indignation, etc.

At first glance, examples such as (157) look very similar to the marked conditional constructions in (143a'), with a conditional V1-clause but without the resumptive element danthen, the discussion of which we have postponed to Subsection V. However, this is a visual deception: in speech, the intonation pattern would immediately distinguish the two, and they also express quite different meanings. A nice illustration of this is given in Van der Horst & Van der Horst; they quote an advertising slogan for Croma, a brand of frying fat:

Of course, the conditional use in (158a) was the one intended; if this slogan were given an exclamative intonation pattern, it would lead to a reading expressing disapproving amazement, which is what we tried to express by the translation in (158b). The translation also expresses that the exclamative construction has no conditional meaning: the speaker is simply asserting that the propositions expressed by the two clauses are both true. There is a relation between the two propositions, however, in that it is the truth of the proposition expressed by the first clause that makes the truth of the proposition expressed by the second clause so surprising; cf. Beekhuizen (2008: §4) for more discussion. Note in passing that the second-person pronoun je as (158a) can easily be given a generic interpretation leading to the interpretation “Anyone who likes meat fries in Croma” but that the second-person pronoun in (158b) must refer to the addressee (but note that Beekhuizen found a fairly large number of generic exclamative constructions in his newspaper corpus).

Exclamative examples such as (157) never involve a resumptive element, which might indicate that the first V1-clause is in the initial position of the second clause. However, this would contradict our earlier conclusion, based on conditional and contrastive constructions, that V1-clauses are always clause-external. Let us then consider the alternative that the first clause is external to the second clause, although it is not easy to find convincing arguments for/against the two options. It would be an argument for the first option if the V1-clause could also appear in a clause-internal position, but the examples in (159) show that this is not the case.

Instead, the examples in (159) suggest that the first V1-clause is external to the second. It would be an argument for this analysis if the contrastive V1-clause could also occur syntactically independent of the second. The coordination constructions in (160), based on the examples in (157) and (158b), show that this is indeed possible.

In fact, the examples in (161), again modeled on example (160b), show that the first clause need not even be coordinated with a declarative clause, but can also be coordinated with an interrogative clause, or a nominal phrase preceded by dan.

It would be another argument for assuming that the first clause is external to the main clause if the main clause could be used as an independent exclamative V1-clause in other contexts. The examples in (162) show that this is also possible.

Examples (160) and (162) strongly suggest that the exclamative constructions in (157) and (158b) are juxtaposed clauses. This fits well with the observation that such exclamative constructions are typical of speech, since exclamative V1-constructions of the type in (160) and (162) are typical of narration. If the juxtaposition analysis is indeed correct, then exclamative V1-clauses are well-behaved with respect to our hypothesis that V1-clauses cannot occur clause-internally.

The introduction to this section has shown that concessive clauses come in at least two varieties, repeated here in a slightly different form as (163a&b). The concessive clause in (163a) is an ordinary adverbial clause: the impossibility of including the particle toch in the initial position of the main clause shows that it must occur clause-internally, and in line with our hypothesis that V1-clauses cannot occur clause-internally, it is introduced by the complementizer-like element hoewelalthough and has the finite verb in clause-final position. The concessive clause in (163b), on the other hand, must be external to the main clause, as is clear from the fact that the particle toch in the first position of the main clause cannot be omitted. Concessive main clauses such as (163b) differ from the conditional clauses discussed in Subsection I in that they do not have an alternant with the finite verb in clause-final position; examples such as (163b') are unacceptable.

The reason for the ungrammaticality of (163b') may be that examples such as (163b) cannot be analyzed as left-dislocation constructions; cf. the discussion of contrastive construction in Subsection II. If (163b) were a case of left dislocation, we would expect the particle toch to be analyzed as a resumptive element linked to the concessive clause, but this analysis is rather unlikely, given that example (164a) shows that this particle can also be used in examples with a clause-internal concessive clause: if toch were a resumptive element, example (164a) would have two constituents performing the same syntactic function. Furthermore, example (164b) shows that toch differs from conditional dan in that it does not have to be clause-initial in declarative clauses; it can even be omitted altogether, although Haeseryn et al. (1997:1391) claims that this is a feature of written texts.

The ungrammaticality of (163b') is also related to the status of the element (ook) aleven though. The fact that (ook) al does not block verb-second shows that it is a regular phrase in clause-initial position and not a complementizer-like element. This is consistent with the fact, illustrated in (165), that some other constituent is usually moved into this position when (ook) al is omitted. The conclusion that (ook) al is a phrase occupying the clause-initial position of the concessive clause correctly predicts that it cannot license the clause-final placement of the finite verb in (163b').

Note in passing that Haeseryn et al. (1997:1392) claims that the omission of (ook) al does not require some other constituent to be moved into the clause-initial position: they consider Was de reclame groot, toch bleef het succes maar klein possible in the formal register. We ignore this case, because we find this example artificial and obsolete; see Van der Horst (2008) for a similar example from Old Dutch (p.337) and the claim that the construction with al is already common in Middle Dutch (p.773-4).

That concessive clauses introduced by (ook) al have the hallmarks of regular main clauses seems to fit well with our earlier conclusion that a left-dislocation analysis is not possible; they must therefore be analyzed as independent main clauses. This is also suggested by another difference with conditional clauses. The (a)-examples in (166), repeated from Subsection I, show that conditional clauses in extraposed position must be introduced by als and therefore do not allow movement of the finite verb. Example (166b), on the other hand, shows that placing the concessive clause last does not affect its form; this again shows that it cannot function as a regular adverbial clause.

The above discussion leaves us with the question of what kind of structure is plausible for the concessive constructions under discussion. The first thing that comes to mind is that we are dealing with two juxtaposed main clauses, and this may indeed be a plausible analysis for examples such as (165b), given that (167a) shows that we may also coordinate the two clauses by the conjunctive coordinator maarbut and that the first clause can readily be used independently. This is not true for examples such as (165a): the use of maarbut in (167b) gives rise to a degraded result and the independent use of the first clause in (167b') has some sense of incompleteness (indicated by the diacritic “$” and the series of dots).

The examples in (167) suggest that a simple juxtaposition analysis may not be the right answer. Since we have no further insights to offer at this point, we leave open the question of the internal structure of the concessive construction under discussion, while concluding that this does not jeopardize the generalization that verb-first/second is excluded in clause-internal dependent clauses.

The previous subsections have shown for a number of adverbial-like V1/2-clauses that they are clause-external, thus supporting the hypothesis that verb-first/second is impossible in the case of (clause-internal) adverbial clauses. This subsection considers some possible counterexamples to this hypothesis. The first case was already mentioned in our earlier discussion, but set aside. Consider again the examples in (168). Example (168b) is normally considered infelicitous, but we have marked it with a percentage sign because Van der Horst & Van der Horst (1999:256ff) provides a large number of attested conditional V1-clauses without resumptive dan from various written sources such as newspapers, belles-lettres, advertisements, etc.

Van der Horst & Van der Horst claims that examples of this type are a recent innovation that became particularly popular in the 1980s although they also found some cases from the 14th century onwards; the examples in (169) show that there are even a number of proverbs of this form.

Van der Horst & Van der Horst (1999:256ff) provides an analysis according to which conditional V1-clauses are clause-internal if dan is not present, and claim that this has become possible in analogy to constructions with als-clauses. They further suggest that the rise of clause-internal conditional V1-clauses is to be expected, since this eliminates an irregularity from the system by allowing all dependent clauses to occur clause-internally. From our point of view, however, such a change would introduce an irregularity into the system because it goes against the well-supported hypothesis that V1-clauses are categorically rejected in clause-internal position. This hypothesis can be saved, however, if we assume that constructions with conditional V1-clauses but without resumptive dan are not part of the Dutch core grammar.

Two possibilities present themselves. One possibility, also considered by Van der Horst & Van der Horst, is based on the observation that the use of resumptive dan is a property of spoken language that is disfavored in written language; its omission in constructions with conditional V1-clauses may therefore be a case of hypercorrection. Another possibility is that some speakers allow the omission of resumptive elements in clause-initial position. If so, the analysis of constructions with conditional V1-clauses without resumptive dan would be as given in (170a). This would lead to the expectation that speakers who allow (170a) also allow “preposition stranding” in examples such as (170b), provided that apparent preposition stranding results from the deletion of the resumptive pronominal part of the discontinuous pronominal PP daar ... op, indicated by strikethrough.

Since we are unable to test whether this expectation is borne out, we must leave it to future research, while noting that we believe a correlation is likely to be found. The reason for this optimism is that, according to Van der Horst & Van der Horst (1999:270), the rise in popularity of the two constructions in (170) occurred more or less simultaneously (in the second half of the 20th century). Whatever the outcome of such an investigation, we can conclude from the discussion above that it is not at all obvious that the occurrence of conditional V1-clauses without resumptive dan refutes the hypothesis that V1-clauses do not occur clause-internally: an appeal to hypercorrection or an analysis along the line in (170a) would be entirely consistent with this hypothesis.

A second possible problem is posed by adverbial-like V1-clauses containing the modal verbs willen and mogen. We will limit the discussion to cases with willen, as in (171). At first glance, these examples seem to be regular conditional constructions of the type discussed in Subsection I: the optionality of the resumptive element danthen in (171a) suggests that the als-clause is a regular adverbial clause, which can either occupy the clause-initial position of the main clause or be left-dislocated; the obligatoriness of dan in (171b) further suggests that we are dealing with a proper V1-clause in the sense that it occurs clause-externally.

However, a closer look shows that, at least in some cases, we are dealing with a slightly different type of construction. First, the acceptability contrast between the examples in (172) shows that the alternation between the als-clause and the V1-clause is not always readily possible.

Second, example (172b) does not express a material implication: the eventuality of “the project becoming a success” as expressed in the first clause is not presented as a sufficient condition for the eventuality of “we working hard” as expressed in the second clause. In fact, the relationship is reversed: the second eventuality can be seen as a condition for the first to take place; cf. Boogaart et al. (2007:240), where examples such as (172b) are characterized as teleological. This interpretation is directly related to the fact that the modal verb willen in (172) does not have a deontic (volitional) but rather an epistemic meaning; cf. Section 5.2.3.2, sub IIIA. A third difference is illustrated by the primed examples in (173), which show that teleological (but not conditional) V1-clauses can occur in clause-final position.

Since we have assumed that clause-final adverbial clauses are clause-internal, example (173b') is a possible counterexample to our hypothesis that V1-clauses can only occur clause-externally. A possible solution can be found in Beekhuizen (2008:46), where it is suggested that V1-clauses in examples such as (173b') are actually parenthetical clauses. Indeed, there are reasons to think that this is the case: Subsection I has shown that parenthetical clauses have the characteristic property that they can contain tenminsteat least and this possibility is also available for clause-final teleological V1-clauses. Note the contrast between the examples in (174a&b), which seems to show that a clause cannot be both left-dislocated and parenthetical. For completeness’ sake, we have added example (173b') to show that the parenthetical clause can also appear in the middle field of the clause.

The presence of dan proves that the V1-clause in (174a) is clause-external, and the possibility of tenminste in the (b)-examples makes it plausible that we are dealing with a parenthetical clause, and these two facts in turn strongly suggest that teleological V1-clauses behave in accordance with our hypothesis that adverbial-like V1-clauses only occur clause-externally. But of course further investigation of this construction is needed to establish this conclusion more firmly; we refer the reader to Beekhuizen (2008: §5) as well as the discussion between Beekhuizen and Leuschner (2016) for a good starting point.

Subsections I to IV have shown that the italicized V1/2-clauses in (129), repeated here as (175), are clause-external; in the conditional construction in (175a), this is clear from the fact that most speakers require the use of the resumptive element danthen in the initial position of the main clause; in the contrastive and concessive constructions in (175b&d), this is clear from the fact that the initial position of the main clause is occupied by some other constituent. For the exclamative construction in (175c) this is somewhat harder to show, but a juxtaposition analysis is quite plausible, since the first clause can also be used as the first conjunct in the near-synonymous coordinate construction Helpt Marie iemand en dan wordt ze door hem beroofd! Imagine: Marie helps someone and then that person robs her!.

The discussion supported the hypothesis that verb-first/second is impossible in regular clause-internal adverbial clauses. Subsection V concluded with a number of possible problems for this hypothesis; however, we have shown that it is plausible that the V1-clauses discussed there are not clause-internal either.