This subsection discusses cases in which verb-second does not seem to apply to main clauses. The focus will be on clausal comparative correlative constructions of the form hoe A1-er ... hoe A2-er ...the more A1 ... the more A2 ...’. example (109a) shows that such constructions have the finite verb of the (italicized) main clause in clause-final position. We will also pay some attention to cases such as (109b), which occur only as idioms.

Comparative correlative constructions typically express a positive (and sometimes negative) degree correlation between two or more entities, properties, events, etc. Two idiomatic examples are given in (110): example (110a) expresses a positive correlation between the number of people and the degree of enjoyment, and (110b) relates the time of the evening to the quality of the people present (typically used in jest, as an ironic way of expressing a negative correlation between the two).

In formal usage, the second occurrence of the modifier hoe in (109a) is sometimes replaced by des te, which clearly does not belong to the core grammar (i.e. the automatically acquired part) of the language. It rather belongs to its periphery (the consciously learned part), given that des is the old genitive form of the definite article. In such cases the first occurrence of hoe is sometimes also replaced by des te. Cases in which only the first occurrence of hoe is replaced by des te seem unacceptable; this means that we find only the three alternants in (111a-c).

That the form in (111a) is the one most commonly found in colloquial speech is also supported by the corpus research reported in Assendelft (2019), which is based on a selection from the OpenSoNaR corpus (https://opensonar.ivdnt.org), consisting of Dutch and Belgian newspapers (the formal part of the corpus) and internet forums (the informal part). The results are shown in Table (112): constructions with hoe ... des te ... are definitely outnumbered (especially in Belgium), while those with des te ... des te ... are even rare.

	hoe… hoe…	hoe… des te…	des te… des te…	total
newspapers (B)	2006 (94,8%)	96 (4,5%)	13 (0,6%)	2115 (100%)
News papers NL	767 (76,4%)	230 (22,9%)	7 (0,7%)	1004 (100%)
discussion forums (B)	767 (91,7%)	62 (7,4%)	7 (0,8%)	836 (100%)
discussion forums (NL	240 (88,9%)	24 (8,9%)	6 (2,2%)	270 (100%)
total	3780 (89,5%)	412 (9,8%)	33 (0,8%)	4225 (100%)

Although the forms with des te are arguably not part of the core grammar of Dutch, we will include them in our discussion because they may clarify the central issue: the non-application of verb-second in main clauses. Consider the clausal comparative correlatives in (113), which are problematic under the hypothesis that main clauses require the finite verb to be in first or second position. If these examples are complete (i.e. non-reduced) sentences, we should conclude that at least one of the two constituent clauses functions as a main clause, but both clauses have the finite verb in clause-final position.

According to Haeseryn et al. (1997:566), examples like those in (113) alternate with examples like those in (114), where the second clause has the finite verb in second position: the finite verb occurs right-adjacent to the clause-initial phrase hoe meer N. For this reason, the first clause is analyzed as an adverbial phrase; cf. also Vandeweghe (2005).

The percentage signs in (114) indicate that we think that the verb-second order in these examples is highly marked; cf. Den Dikken (2003) for similar judgments. This appears to be supported by the results in Assendelft (2019: Table 2): no more than 8 (0.2%) cases with the verb in second position were found. However, the verb-second order seems perfectly acceptable if the second occurrence of hoehow is replaced by the more formal form des te, as in (115).

Haeseryn et al. and Den Dikken claim that the verb-second orders in (115) are the preferred ones, but the results in Assendelft (2019: Table 3) show that the two orders occur more or less on a 50/50 basis, although there seems to be a slight preference for the verb-second order in the Dutch newspaper corpus. This may not reflect a general preference, but rather the effect of the style guidelines used to edit the articles. It may therefore be relevant that a Google search (May 29, 2024) on the string [hoe meer je * des te meer] showed that the verb-final order is by far the most common in the first 50 hits. We will leave this issue to future research; we simply follow Assendelft in assuming that both orders in (115) are perfectly acceptable. Since verb-second order is normally restricted to main clauses, its availability in (115) makes it plausible that the second clause functions as a main clause. This is consistent with the fact, illustrated in (116), that embedding such examples blocks verb-second in the second clause; cf. Den Dikken (2003), where it is also noted that embedded constructions such as (116) are only possible with bridge verbs like zeggento say, denkento think, vindento find, etc.

For completeness, note that verb-second is categorically rejected in the first clause; cf. Den Dikken (2003) for data showing a similar contrast between the two parts of the English comparative correlative construction the more A1 ... the more A2.

The conclusion that the second clause is the actual main clause seems solidly based, but it also raises the question as to the exact structure of the construction as a whole. The fact that the phrase des te meer (N) is left-adjacent to the finite verb in verb-second position in examples such as (116) suggests that it occupies the initial position of the main clause, as a result of wh-movement; cf. Section 11.3.6 for further discussion. If correct, this would imply that the first clause is external to the main clause (and thus may belong to the class of elements to be discussed in Chapter C37). This is represented schematically in (117).

Being external to the main clause is in fact not exceptional for adverbial-like clauses, as we also find it in conditional and concessive clauses like (118a&b), which will be discussed in more detail in Section 10.3.2. Such examples are also similar to comparative correlatives in that the main clauses are introduced by (or at least contain) an element that establishes a relation between the material external to the main clause and the main clause itself; the linking elements danthen and tochstill/nevertheless.

However, the similarity between comparative correlatives and conditional/concessive constructions is not perfect: the linking elements danthen and toch are resumptive in nature, so we may be dealing with left dislocation (cf. Section C37.2); the linking element in comparative correlatives (i.e. hoe/des te) simply expresses the correlation, i.e. has no obvious resumptive function. This makes a left-dislocation analysis less plausible for the comparative correlative construction; cf. Den Dikken (2005) for an alternative proposal, based on comparison with correlative constructions in other languages, according to which the first clause is adjoined to the second.

The analysis proposed in (117) raises at least two non-trivial problems. The first problem is that we need to account for the fact that verb-second is not obligatory in the comparative correlative construction with des te meer in the clause-initial position (cf. (115)), and that verb-second even seems even impossible in the corresponding constructions with hoe (cf. (114)). The second problem, which is probably closely related to the first, concerns the possibility of adding complementizers to the examples in (113). The resulting examples in (119) are perhaps less favored than those in (113), especially for speakers from the Netherlands: Assendelft (2019: Table 6) shows that they are common for Belgian speakers but rare for speakers from the Netherlands; examples such as (119) are also accepted in Paardekooper (1986:350/658) and Den Dikken (2003)’ .

Den Dikken (2003:9) claims that the primeless and doubly-primed examples in (119) cannot be replicated in the corresponding constructions with des te; more specifically, it is claimed that the complementizer dat cannot be used in the first clause. That complementizers are rare in constructions with des te is confirmed by Assendelft (2019: Table 6): there were no cases from Dutch origin. There were three cases of Belgian origin (unfortunately, no information about the placement of dat is given), but this may be somewhat suspect in view of the low frequency of des te in the Belgian part of the corpus; cf. Table (112). However, a Google search (March 16, 2025) on the search string [des te langer dat + subject pronoun] yielded several cases of the form in (120a), which we therefore gave a percentage sign; cf. Des te langer dat ze staan, des te bitterder ze wordenThe longer they stand, the more bitter they become (https://blog.seniorennet.be/keukenweetjes/archief.php?ID=176). What we could not find were cases such as (120c) with a complementizer in both clauses, which we indicated with a dollar sign, because the fact that des te is part of the formal register makes it somewhat difficult to give acceptability judgments on these cases.

In addition to the two problems for the analysis in (117) regarding verb-second and the use of the complementizer, there is the problem that the complementizer does not have to be present in the verb-final construction. The presumed main clause can thus occur in three different forms; the form in (121b) is the expected one, while the forms in (121a&c) are the unexpected, deviating ones.

Den Dikken (2003) relates the grammaticality contrast between hoe and des te in the verb-second construction in (121b) to the fact that the phrase hoe A-er cannot normally occur in the main-clause initial position, whereas des te A-er can. This is demonstrated in (122) by coordinate constructions with the conjunctive coordinator maarbut, which expresses a negative correlation.

If this line of thinking is correct, the availability of the structures in (121a&c) could be seen as the result of a repair strategy for the comparative correlative construction with hoe, which was subsequently extended by analogy to the corresponding construction with des te. We will leave this suggestion for future research.

As far as we know, the theoretical literature is silent on the fact that comparative correlatives can sometimes be reduced. It seems at least plausible to provide a sluicing-like analysis to account for the similarity between the (a) and (b)-examples in (123): the fact that the two hoe-phrases may precede the complementizer dat (if present) shows that they occupy the initial position of the main clause (i.e. the specifier of CP) as a result of wh-movement, and we may therefore be able to derive the (b)-examples from (123a) by deleting the remainder of the clause (i.e. IP). For a general outline of such an analysis, see Section 5.1.5.

The discussion above has shown that comparative correlative constructions of the type in (121a&c) are indeed problematic for the otherwise robust generalization that main clauses have obligatory verb-second. As far as we know, there are no other productive constructions that violate this generalization, but there are idiomatic constructions that systematically violate it. Some examples are given in (124).

The examples in (124), which should probably be regarded as historical relics, all have the same overall structure: the first clause is a free relative and the second clause is introduced by a resumptive (demonstrative) pronoun linking the free relative to the main clause; note, incidentally, that there is variation in the form of the demonstrative pronouns (wat versus dat and wie versus die). The parallel form of the two clauses might suggest that they have the same internal structure, but this is not the case, since the non-idiomatic present-day counterparts of the examples in (124) would have verb-second in the second clause; this is illustrated for (124d) in (125), with and without a resumptive pronoun.

The examples in (125) suggest that the second clauses in the proverbs in (124) should also be analyzed as main clauses. We refer the reader to Den Dikken (2005: 500-1) for the claim that the analyses of examples such as (124) and comparative correlatives can be unified by assuming that the non-main clauses in the latter construction are in various ways similar to free relatives, a property that is claimed to be a cross-linguistic feature of comparative correlatives.

To conclude our discussion of the comparative correlative construction, we want to point out that comparative correlatives such as (123b'') should not be confused with hoe ... hoe-phrases of the type in (126a), which can be used as complementives in e.g. copular constructions; cf. Section A26.3.2.1, sub II. Since (126b) shows that finite clauses cannot be used as complementives in copular constructions, the sluicing analysis proposed for (123b'') is clearly not suitable for the hoe ... hoe-phrase in (126a).