Correlative coordinators can easily be confused with correlative adverbial phrases such as enerzijds ... anderzijds ...on the one hand ... on the other (hand) ... and niet alleen ... ook ...not only ... also .... The crucial difference is that correlative coordinators are external to the coordinands, whereas correlative adverbials are part of them. This can be easily illustrated by means of clausal coordinands: in (502a), the initial positions of the coordinated main clauses (in square brackets) are occupied by their subjects, which means that the two parts of the correlative coordinator en ... en ... are clause-external; in (502b), on the other hand, the initial positions are occupied by the correlative adverbial phrases, as can be seen from the fact that they trigger subject-verb inversion. Note that in (502b) we are dealing with an asyndetic construction, but that it is possible to replace the phonetically empty coordinator by the coordinator maarbut.

One problem with applying the word order test in (502) is that not all correlative coordinators can link main clauses, as shown in (503a) for zowel ... als ...both .... and .... However, example (503b) shows that the two parts of the correlative coordinator cannot occupy the initial positions of the coordinated clauses either, providing somewhat weaker evidence for the claim that they are not clausal constituents.

Nevertheless, it has been claimed that there is reason for assuming adverbial status for the initial part of the correlative coordinator, zowel: Haeseryn et al. (1997:1515), for instance, claim that it can penetrate into the first coordinand. This claim is crucially based on their assumption that (504b) involves conjunction reduction (indicated by strikethrough). We have marked the structure in (504b) with an asterisk because the analysis suggested by Haeseryn et al. is highly problematic in light of the fact that the deleted part cannot be realized overtly; a more natural alternative analysis would be that we are simply dealing with coordination of PP-modifiers, as indicated in (504b').

It should also be noted that some speakers consider the order in the (b)-examples to be marked compared to (504a). This order is even more degraded when the noun is singular: (505b) is at best marginally acceptable when the string zowel van Marie als van Jan is parenthetical, i.e. preceded and followed by an intonation break. The unacceptability of (505b) on the non-parenthetical reading would be unexpected in a conjunction reduction analysis, which should therefore be rejected.

Note in passing that Neijt (1979:6-7) has shown that examples such as (504b') are impossible with other correlative coordinate structures: replacing zowel ... als ... by en ... en ...and ... and ..., of ... of ...either ... or ..., or noch ... noch ...neither ... nor ... leads to severely degraded results (on the intended, non-parenthetical reading). Again, this is unexpected in a conjunction reduction analysis. Haeseryn et al. (1997:1517) also provide a conjunction reduction analysis for example (506a), with coordination of main clauses (CPs). An alternative analysis would involve coordination of verbal predicates (VPs), as indicated in (506b).

The conjunction reduction approach suggested by Haeseryn et al. is again problematic because it cannot account for the unacceptability of example (507a), since the (presumed) clausal coordinands are both syntactically well-formed. The VP-coordination analysis, on the other hand, fares better here because example (507b) is unacceptable for two reasons: (i) the finite verb snoeit has been extracted from the first VP-coordinand by verb-second in violation of the coordinate structure constraint discussed in Section 38.3, sub II, and (ii) the finite verb plant in the second coordinand cannot undergo verb-second at all because the verb-second position is already occupied by snoeit. That the unacceptability of (507a&b) should be accounted for in terms of verb-second is supported by the fact that their embedded counterpart (without verb-second) in (507c) is fully acceptable; cf. Neijt (1979:7ff.).

The discussion above has shown that the two parts of correlative coordinators must be external to the coordinands. This casts doubt on the generally accepted claim that noch ... noch ...neither ... nor ... is a correlative coordinator. The examples in (508) show that the two occurrences of noch are internal to the clausal coordinands: the (a)-examples show that noch obligatorily triggers subject-verb inversion and (508b) shows that noch can even occur in the middle field of the first clause.

The examples in (508) thus conclusively show that noch ... noch ... can be used as a correlative adverbial, which need not surprise us, since Section 38.3, sub IIIG, has already shown that noch can be used as an adverbial with the meaning (en) nietand not. Of course, it is not the case that these examples provide conclusive proof that the sequence noch ... noch ... cannot occur as a correlative coordinator in other contexts: the unacceptability of (508a') may simply be due to some idiosyncratic restriction on the coordinands of correlative noch ... noch ... It will be clear, however, that the burden of proof is on those who wish to maintain the traditional analysis. Those who reject this analysis cannot rest their case either, since they should provide a better alternative for cases such as Ik heb noch Jan noch Marie gezienI have seen neither Jan nor Marie, in which noch ... noch ... seems to behave as a coordinator. We will return to this issue in Subsection III.

Haeseryn et al. (1997:1517) claim that the intended meaning of the unacceptable example in (507a) can be expressed by example (509a). They rate this example as fully acceptable, but our informants rate it as unacceptable, or at least quite marked, and the same holds for example (509b), which is claimed to occur alongside (506a). The examples in (509) are regularly cited in the literature as counterexamples to the generalization that the coordinands in a coordinate structure must be of the same kind in the sense formulated in Section 38.3, sub I.

We believe that theoretical claims based on these marked, constructed examples should be approached warily: we would be happy to assume that they are not part of the core syntax and should be regarded as a quirk of the formal register, i.e. as a relic of the older adverbial use of zowel als as described in the Woordenboek der Nederlandsche Taal (item Zoowel, sub 1); see also Van Zonneveld (1992:340).

The examples in (510) show that similar structures with even min/(net) zo min ... als ...neither ... nor ... seem acceptable (although some speakers object to (510b), which we have indicated with the percent sign). However, there is no a priori reason for analyzing these sequences as coordinators; an adverbial analysis may also be viable, since that they have the form of an equative; cf. even/(net) zo aardig als ...(just) as kind as; we will return to this in Chapter 40. Note in passing that, for this reason, we will not follow the orthographical rule that that even min and zo min are written as single words; elsewhere, when their status as head/phrase is not at issue, we do follow the orthographical rule.

We also have good empirical reasons for excluding even min/zo min ... als ... from the set of correlative coordinators and for analyzing them as ordinary adverbially used equatives of the form even/zo A als ...as A as. First, note that zo min is typically modified by the degree modifier netjust as, which is impossible for a coordinator but common for adjectives premodified by zoas; cf. net zo aardig (als)just as nice (as).

Second, if we were dealing with correlative coordinators, we would expect the als-part to be obligatory; the cases in (512) show that this is borne out by zowel ... als ... but not by even min/zo min ... als ...; just as in the case of even/net zo aardigjust as kind the als-part can be omitted if its content can be reconstructed from the context.

Third, we have seen in Subsection A that the two parts of the correlative coordinator zowel ... als ... cannot occupy the initial position of clausal coordinands. Example (513b) shows, however, that this is possible for the initial part of the supposed coordinators even min/zo min ... als ... The fact that even min and net zo min trigger subject-verb inversion shows that they are clausal constituents.

Fourth, example (514b) shows that even min/zo min ... als ... would be unique among coordinators in being unable to coordinate nominal phrases. The intended meaning can be expressed by (514b'), but in such cases even min/zo min als-phrases clearly have an independent adverbial function. The (a)-examples show that zowel ... als ... again exhibits exactly the opposite behavior.

The claim that the even min/zo min als-phrase in (514b') has an independent syntactic function seems uncontroversial, since Haeseryn et al. (1997:1517) provide the same analysis for similar examples. They also analyze net zo min ... als Peter in example (515b) as adverbial. This analysis can be supported by the fact that net zomin in (515b) can be replaced by other degree adverbials, as shown by Jan is net zo goed/zeer een schurk als PeterJan is just as good a scoundrel as Peter; a similar replacement of zowel is never possible, probably because it is a single word.

The (b)-examples in (511) to (515) all provide evidence for the claim that even min and zo min are clausal constituents; therefore it seems safe to conclude that even min/zo min ... als ... are not correlative coordinators, but should be analyzed as adverbially used equatives of the form even/zo A als ...as A as. The fact that the (a)-examples show that zowel ... als ... behaves systematically differently can be taken as evidence for analyzing it as a true correlative coordinator. Readers interested in further discussion of the internal structure of even min/zo min ... als ... phrases are referred to Section 40.1, sub I.

This subsection has examined the set of putative correlative coordinators in (501) from the point of view that such coordinators are like simplex coordinators in that they are external to their coordinands. This has led to the conclusion that the two members of the sequence noch ... noch ... should be analyzed as adverbial phrases in the case of clausal coordination; nevertheless, we will not exclude this sequence from the list in (501) for the simple reason that noch ... noch ... may still function as a correlative coordinator in other contexts. We will exclude the sequences even min/zo min ... als ... from the list because there is compelling evidence that these are not correlative coordinators but adverbially used equatives of the form even/zo A als ...as A as. The result of all this is the reduced list in (516).

One of the characteristic properties of the correlative coordinator en ... en ...and ... and ... is that its initial and non-initial parts are accented, which is why it is often written with an accent: Ik heb én Jan én Marie gezienI have seen and Jan and Marie. This subsection will show that coordinate structures with correlative en ... en ... differ from those with simplex en in that they (i) exhibit additional restrictions on the coordinands, (ii) cannot be interpreted cumulatively, and (iii) exhibit special agreement properties.

The examples in (517) show that the correlative en .. en ... behaves like simplex en in that it can link clauses (CPs), noun phrases (DPs), APs, and PPs.

The correlative and simplex forms differ, however, in that the former links only declarative main clauses: linking interrogative (Q), imperative (Imp) or exclamative (Excl) clauses leads to a degraded result; see also Neijt (1979:16). For comparison, consider the corresponding examples with simplex enand in Section 38.4.1, sub IA1.

Dependent declarative and interrogative clauses, on the other hand, behave alike in that they both can be conjoined; imperatives are excluded for the independent reason that they cannot be embedded at all.

Correlative en ... en ... also differs from simplex en in that it is more restricted when it comes to conjunction of extended verbal projections smaller than clauses: while example (520a) is fully acceptable, example (520b) is very much worse. The difference seems to be related to verb-second of the main verb, since comparable coordination is possible when the main verb does not have to undergo verb-second: this is illustrated in the primed examples for embedded clauses and for main clauses with complex verb constructions such as the perfect tense.

For completeness, the examples in (521) are added to show that the same observations can be made for constructions with monadic verbs.

The simplex and the correlative coordinators also exhibit different behavior in embedded clauses with complex verb constructions: while in (522a) simplex en can be used to conjoin main verbs, correlative en ... en ... cannot. Example (522b) shows that they do show similar behavior when they conjoin a larger verbal projection that includes the auxiliary.

The differences in the behavior of simplex en and correlative en ... en ... show that the simplex coordinator is able to conjoin a larger set of constituents than the correlative one; cf. Neijt (1979:§1.1). This conclusion can also be drawn on the basis of non-verbal coordination. First, the correlative coordinator is more restricted when it comes to conjunction of nominal projections smaller than DP; example (523b) shows that while simplex en is able to do this, correlative en ... en ... is not.

Second, the examples in (524) suggest that the same holds for adjectival phrases. Example (524a) asserts that Jan has the two independent properties of being very young and being (very) inexperienced. The sentence Jan is erg jong en onervaren, on the other hand, can be interpreted in such a way that Jan has the (single) complex property of being young and inexperienced, which Section 38.4.1, sub IA1, accounted for by assigning it the structure in (524b). The primed examples show that the correlative coordinator is only compatible with the multiple-property reading, i.e. it cannot coordinate the smaller adjectival projections without the degree adverbial; see Corver (1990:53) for more examples.

Third, as shown in (525), it appears that while both the simplex and the correlative coordinator are capable of coordinating full PPs, as in (525a), they differ in that the simplex but not the correlative coordinator can coordinate smaller projections of the preposition, as in (525b), or the nominal complement of the preposition, as in the (c)-examples.

Finally, we see in (526) that coordination of attributive modifiers by correlative en ... en ... gives rise to a marked result, while coordination by simplex en is easy; see Corver (1990:51) for more examples.

Neijt concludes from data of the above kind that correlatives can only be used to link major phrases, i.e. the set of “fully expanded” projections of the lexical categories N, A, and P that function as clausal constituents, as well as to specific smaller, non-clausal verbal projections (“VPs”).

There do not seem to be any restrictions on the type of clausal constituent. As shown in (527), correlative coordinate structures with en ... en ... can be used as arguments, as complementives and (to a lesser extent) supplementives, and in various adverbial functions; there is no clear difference in this respect with the corresponding constructions with simplex en given in Section 38.4.1, sub I.

Coordinate structures with correlative en ... en ... differ semantically from those with simplex en in that they cannot be interpreted cumulatively: while example (528a) is ambiguous between a distributive and a cumulative reading, as can be seen from the fact that the modifiers beidenboth and samentogether can both be used, example (528a') has only a distributive reading. That coordinate structures with correlative en ... en ... cannot be interpreted cumulatively is also clear from the fact that such coordinate structures cannot act as antecedents for the reciprocal elkaareach other.

The two (a)-examples in (528) also differ in subject-verb agreement: while the coordinate structure with simplex en triggers plural agreement on the finite verb, the one with correlative en ... en ... usually triggers singular agreement. The judgments are not always sharp but Haeseryn et al. (1997:1501) claim that singular agreement is always the preferred option, which would be consistent with the hypothesis discussed in Section 38.1, sub IVD, that coordinate structures with an inherent distributive reading must trigger singular agreement on the finite verb when their coordinands are both singular; another clear example illustrating this is given in (529a). Example (529b) shows that, not surprisingly, correlative coordinate structures with en ... en ... trigger plural agreement when the two nominal coordinands are plural. Mixed cases are usually less good, although De Vries & Herringa (2008:§3) claim that there is a tendency for agreement with the coordinand closest to the verb; see also G. de Vries (1992:§2.4). Since such examples are not used in colloquial speech and native speakers tend to reject them categorically, it is difficult to evaluate this claim; we therefore simply mark the degraded examples with a percent sign.

Similar problems with subject-verb agreement arise with mixed person features: if the coordinands trigger the same morphological form on the finite verb, as in (530a&b), the result is generally deemed acceptable but if they trigger different forms, as in the (c)-examples, the result is severely degraded.

Subsection I has already shown that the correlative zowel ... als ...both ... and ... differs from the correlative en ... en ...both ... and ... in that it cannot link clausal coordinands. The other instances in (531) show, however, that the two coordinators do not differ when it comes to coordinating coordinands of other categories.

The examples in (531b-d) also show that correlative coordinate structures with zowel ... als ... can be used as argument, complementive or adverbial phrase; the same is shown by the fact that the substitution of zowel ... als ... for en ... en ... in the examples in (527) in Subsection A does not affect the acceptability judgments in any significant way; we leave it to the reader to construct the relevant examples.

Coordinate structures with correlative zowel ... als ... also behave like those with en ... en ... in that they are usually used as major phrases in Neijt’s sense: they are always clausal constituents or larger verbal projections. The latter is illustrated by the examples in (532) and (533), which correspond to the examples in (521) and (522) from Subsection A with correlative en ... en ....

The fact that non-verbal coordinate structures are usually clausal constituents is clear from the fact that the substitution of zowel ... als ... for en ... en ... in the examples in (523) to (526) from Subsection A does not materially affect the acceptability of these examples; again, we leave it to the reader to construct the relevant examples. Judgments about examples with attributive modifiers are not always sharp, however, and it is not impossible to find examples like those in (534) on the internet.

Such (potential) counterexamples to the claim that correlative coordinate structures are usually clausal constituents often sound formal or artificial; the more natural way of expressing the same thoughts would be as indicated in (535). These examples involve coordination of full noun phrases with backward conjunction reduction (indicated by strikethrough); we refer the reader to Section 39.1 for relevant discussion.

Another potential counterexample with postnominal modifiers is (536), adapted from Haeseryn et al. (1997:1573). However, this coordinate structure has a parenthetical ring to it, which is also corroborated by the fact that it can be used in postverbal position; we refer the reader to Subsection IA, example (504b), for a more detailed discussion of a similar counterexample.

We will therefore leave aside examples like (534) and (536). We note, however, that examples like those in (537) are perfectly natural, which is surprising if correlative coordinate structures are usually clausal constituents. At the present moment, we see no way to account for this.

Nominal coordinate structures with correlative zowel ... als ... are like those with en ... en ... in that they must be interpreted distributively: they cannot license adverbials such as samentogether and they cannot act as antecedents for the reciprocal elkaareach other.

The hypothesis discussed in Section 38.1, sub IVD, that coordinate structures with an inherent distributive reading must trigger singular agreement on the finite verb when their coordinands are both singular thus correctly predicts that (539a) exhibits singular agreement. Haeseryn et al. (1997:1516) claim that in some cases a plural finite verb is preferred, but these examples seem to be artificial and intuitions seem to differ among speakers; cf. De Vries & Herringa (2008:12). De Vries & Herringa also note that some cases of plural agreement seem to be of a semantic nature, which is especially clear when at least one of the nominal coordinands is a collection noun, as in the quite natural (b)-examples in (539), which are taken from taaladvies.net/taal/advies/vraag/949.

There is little new to add about subject-verb agreement, since we find the familiar pattern: cases in which the two coordinands trigger the same morphological form on the finite verb are fully acceptable, while cases in which the coordinands trigger different forms are degraded to varying degrees: since judgments may vary from speaker to speaker and from case to case, we simply mark the less felicitous cases with the percent sign.

Correlative of ... of ...either ... or ... behaves more or less like en ... en ...and ... and ..., since the substitution of the former for the latter in the examples in Subsection A does not affect the acceptability in a significant way. We will illustrate this here for just a few of the cases. The examples in (541) first show that of ... of ... is able to coordinate coordinands of various categories as well as main clauses.

In (541a), the correlative of ... of ... precedes the initial position of the clausal coordinands and thus can be safely assumed to be external to them. This may be different in the case of the correlative ofwel ... ofwel ...either ... or .... The examples in (542) show that ofwel can not only be used external to the coordinands, but can also occupy the initial position of the coordinated clauses. This suggests that the sequence ofwel ... ofwel ... can be used both as a correlative coordinator and as a correlative adverbial. Haeseryn et al. (1997:1508), who analyze the two examples in (542) as correlative coordinate structures, claim that the order in (542a) is especially found in the Netherlands while (542b) is typically found in Belgium, but according to us both are equally acceptable in the standard variety. We represent (542b) here as an asyndetic construction, but other analyses are conceivable, such as of [wel ...] of [wel ...]; see Bredsneijder (1999:13) for discussion.

The examples in (541b-d) show that correlative coordinate structures with of ... of ... can be used as an argument, a complementive or an adverbial phrase. Coordinate structures with correlative of ... of ... also behave like those with en ... en ... in that they must be used as major phrases in Neijt’s sense: they are always clausal constituents or larger verbal projections. The latter is illustrated in (543) and (544), which correspond to the examples in (521) and (522) with correlative en ... en ... from Subsection A.

There is little new to say about subject-verb agreement, as we find the familiar pattern: cases in which the two coordinands trigger the same morphological form on the finite verb are fully acceptable, while cases in which the coordinands trigger different forms are degraded to varying degrees. Since judgments may differ from speaker to speaker and from case to case, we mark the degraded cases with the percent sign.

The highly formal correlative hetzij ... hetzij/of ... differs from of ... of ... in that it cannot link main clauses: cf. *Hetzij ik kom naar jou hetzij ik ga naar oma. In fact, Haeseryn et al. (1997:1511) suggest that this correlative coordinator is mainly used for linking adverbial phrases: Ik kom hetzij vanmiddag, hetzij vanavondI will come either this afternoon or this evening. However, they then discuss several types of cases, suggesting that hetzij ... hetzij/of ... has roughly the same potential as of ... of ....

Subsection IA has already shown that the correlative coordinator noch ... noch ... cannot coordinate clauses. It is able, however, to coordinate smaller constituents of various categories.

These instances also show that correlative coordinate structures with noch ... noch ... can be used as arguments (546b), complementives (546c), and adverbials; again, replacing en ... en ... in the examples in (527) from Subsection A by noch ... noch ... has no bearing on acceptability. Correlative coordinate structures with noch ... noch ... also behave like those with en ... en ... in that they are major phrases: they are always clausal constituents or larger verbal projections. The latter is illustrated by the examples in (547)/(548), which correspond to the examples in (521)/(522) with the correlative en ... en ... from Subsection A.

With respect to subject-verb agreement, we find the by now familiar pattern: cases in which the two coordinands trigger the same morphological form on the finite verb are fully acceptable, while cases in which the coordinands trigger different forms are degraded to varying degrees. Since judgments may differ from speaker to speaker and from case to case we mark the degraded cases with the percent sign.

A promising solution to the problem that the initial parts of a correlative coordinator cannot be analyzed as two-place linkers would be to assume that the two subparts Co1 (= the initial part) and Co2 constitute subparts of a single (complex) head; cf. Larson (1985). The base structure of correlative coordinate structures would then be as in the left-hand side of Figure 32, while the surface structure is derived by moving the initial part to some position higher in the structure; the dotted line indicates an indeterminate number of nodes external to the coordinate structure CoP.

Larson’s main argument in favor of the analysis in Figure 32 is semantic in nature in that the presumed landing site of Co1 restricts the semantic scope of Co2, which was argued mainly on the basis of an observation about English either ... or ... We will not review the semantic arguments in the following, but instead focus on a number of syntactic arguments for and against this analysis.

A second line of investigation starts from the assumption that the constituting parts of correlative coordinators are all heads. On this assumption, there seem to be two obvious ways of representing correlative coordinate structures: either the initial part of this structure takes the first coordinand as its complement, as in the left-hand representation in Figure 33, or it takes the full coordinate structure as its complement, as in the right-hand representation: see Kayne (1994:58) including footnote 2.

Both representations in Figure 33 are problematic in that the initial part of the correlative coordinator does not behave as a two-place linker: it takes only a single complement, viz. the noun phrase Marie in the left-hand structure and the coordinate structure Co2P in the right-hand structure. The left-hand representation also violates the co-occurrence restrictions on coordinands, according to which the coordinands must be syntactically similar, because als links the CoP zowel Marie and the noun phrase Peter. In the right-hand representation, zowel does not enter into a relation with either coordinand. The two double-head analyses are therefore not very promising in their present form, but there are slightly more sophisticated versions of them that may be more promising.

The analyses discussed in Subsections A and B assume that the initial part of a correlative coordinate structure is (part of) a head. There are also proposals which assume that the initial part is phrasal. The rationale for this assumption is that the initial part exhibits similarities with focus particles such as restrictive alleenonly; cf. Section V13.3.2, sub IC. For example, Hendriks (2001a) has observed that the initial part resembles focus particles in that it must have an emphatically accented phrase in its domain. Since Section 38.1, sub V, has shown that coordinands usually cannot be unaccented (cf. Ik bezocht Peter en haar/*ʼrI visited Peter and her), this is not a conclusive argument for claiming that the initial parts of correlative coordinate structures are focus particles, but Subsection 1 will show that these elements do indeed exhibit striking similarities in syntactic behavior; see also Hendriks (2001b/2004) and Johannessen (2005).

Hendriks’ conjecture that the initial parts of correlative coordinate structures are focus particles makes the same predictions as the CorP-analysis discussed in Subsection B: the initial and the non-initial part(s) of “correlative coordinators” can have a different form because the initial part is not a coordinator but a focus particle; the fact that the non-initial parts must have the same form shows that the coordinators in the domain of the focus particle must be identical (both in meaning and in form). However, we will see in Subsections 1 and 2 that Hendriks’ conjecture also accounts for various facts that are not expected by the CorP-analysis. Subsection 3 provides a more speculative discussion based on data suggesting that correlative structures are not (or at least need not be) a single phrase.