The examples in (263) show that the coordinator enand is highly productive as a linker; it can link clauses (CPs), noun phrases (DPs), APs and PPs.

The illocutionary type of clauses does not seem to affect the acceptability of the resulting coordinate structures: declarative (Decl), interrogative (Q), and imperative (Imp) clauses can all be coordinated. The case of declaratives was already illustrated in (263a), and the cases of interrogatives, imperatives and wh-exclamatives are illustrated in (264). Although not all linguists consider exclamatives to be a special type of illocutionary clause, it is worth noting that exclamative clauses can also be coordinated.

The examples in (263a) and (264) are main clauses, but (265) shows that dependent clauses can also be conjoined. Again, the clauses can be declarative, interrogative or exclamative, but imperatives are excluded for the independent reason that they cannot be embedded.

Extended verbal projections smaller than clauses (CPs) can also be conjoined; we will refer to such smaller projections as VPs for convenience, but the reader should keep in mind that these projections may be larger than what is called VP elsewhere in this work. The (a)-examples in (266) show that differences in size of the conjoined projections can evoke differences in meaning. The clausal conjunction in (266a) can be interpreted as referring to two independent events: Jan may have gone to Amsterdam for sightseeing, while he may also have bought a computer in his hometown. The conjunction in (266a'), on the other hand, is more likely to refer to a single composite event: Jan went to Amsterdam and bought a computer there (or, perhaps, in order to buy a computer there). When the subject is quantified, as in the (b)-examples, the difference is even more striking, since the two examples differ not only in the number of events, but also in the number of persons involved in performing them.

Meaning differences of this kind can be used to argue against approaches that derive the primed examples from the primeless ones by conjunction reduction, because the difference follows more naturally if we assume that the coordinands in the conjunctive coordinate structure in the primed examples are smaller, predicative projections of the main verbs gaanto go and kopento buy; we return to this issue in Section 39.1.

Nominal conjunctions of different sizes are also possible. The fact that the nouns in the primeless examples in (267) are both preceded by the article dethe shows that we are dealing with fully expanded nominal projections (DPs), while the fact that the nouns in the primed examples share a single article suggests that we are dealing with smaller nominal projections, which we will provisionally refer to as NPs; note that, again, these projections may be larger than the projections for which we have used the notion NP elsewhere in this work.

Again there are clear differences in meaning. Example (267a) may be ambiguous, because the conjunction en can be taken to be either distributive or cumulative: cf. Section 38.1, sub IVD. In the distributive case, we have two separate events in which a group of women and a group of men each sing a song, while in the cumulative case we have a single event in which the men and women sing a song together. Example (267a'), on the other hand, has only one reading: we are dealing with a single group of men and women singing a song together. The semantic difference in the (b)-examples is even clearer, as it is also reflected in subject-verb agreement: the coordinate structure in (267b) refers to two different persons and triggers plural agreement on the verb, while the coordinate structure in (267b') refers to a single person and triggers singular agreement on the verb; for a more detailed discussion of such cases, see Section N17.3.2.4. It will be clear that such meaning and agreement differences conflict with approaches that derive the primed examples from the primeless examples by conjunction reduction.

It is less easy to argue that adjectival coordinands can differ in the same way as the verbal and nominal ones. One argument could be built on the examples in (268). Example (268a) is unambiguous and asserts that Jan has two independent properties: he is very young and he is also very inexperienced. The sentence Jan is erg jong en onervaren, on the other hand, is ambiguous in the way indicated by the bracketing in the two (b)-examples: the reading in (268b) asserts that Jan has the two independent properties of being very young and of being inexperienced, while (268b') asserts that he has the (single) complex property of being young and inexperienced; the modifier erg modifies this complex property, with the resulting entailments that Jan is very young and that he is very inexperienced. The structures in the two (b)-examples can be distinguished by intonation: (268b) allows an emphatic accent on en, while (268b') does not.

Complex properties are typically composed of atomic properties that “fit” together in a natural way. This holds for the properties referred to by young and inexperienced but not for those referred to by old and inexperienced, which would lead to the expectation that the complex-property reading is harder to get in coordinate structures with the adjectives old and inexperienced. The judgments on the (b)-examples in (269) are not sharp but seem to go in this direction.

Note further that in periphrastic comparative constructions such as (270) the complex-property reading is the only one possible. This can perhaps be attributed to the coordinate structure constraint: the postverbal dan-phrase can apparently be associated with to the full AP minder jong en onervaren in (270a), but not with the first AP-coordinand minder jong in (270b). We will not digress on this here but leave it to future research.

The primeless examples in (271) suggest that adpositional coordinate structures also allow modifiers to take scope over both coordinands: the use of strikethrough indicates ellipsis as the result of backward conjunction reduction. Note, however, that alternative analyses of the primed examples present themselves, according to which we are dealing with two conjoined prepositions and only a single prepositional phrase.

We will opt for the conjunction reduction analysis in the primeless examples, because we find the same phenomenon in examples such as (272), which cannot be analyzed as involving conjunction of prepositions. But even if this argument is not adequate for (271), we can still conclude from (272) that prepositional projections smaller than full-fledged PPs can be conjoined, since the modifier vlak is arguably an integrated part of PPs (as is clear from the fact that it must be pied-piped under topicalization).

Examples like (271) and (272) raise the question as to whether syntactic heads (V, N, A and P) can be conjoined at all. Because of the complexity of this issue and the fact that head conjunction analyses compete with conjunction reduction analyses, we will not discuss this issue here, but postpone it to Section 39.1, where the discussion will find a more natural place.

Coordinate structures with en can be used in virtually all conceivable syntactic functions: the previous subsection has already shown that they can be used as complete sentences, but they can also be used as arguments, complementives or supplementives, and various types of adverbial phrases. Some typical examples are given in (273).

There are certain adverbial types that do not easily allow coordination, such as the modal adverbials in (274a), but it seems that the degraded status of the sentence is for purely semantic reasons: the modalities expressed by the two coordinands do not overlap, and so we end up with a contradiction. If the modalities expressed do overlap, coordination is possible (with an intensifying reading as a side effect).

Coordinate structures also occur below the level of clausal coordinands. The examples in (275) show that they can occur as nominal modifiers, as in the (a)-examples, but also as smaller nominal projections, as in the (b)-examples, where the modifiers have scope over both coordinands. Subsection 1 has already shown that similar examples can also be constructed with ease for adjectival and prepositional phrases, so we need not repeat this here.

Finally, we note that coordinate structures can be used not only as complements of verbs but also of nouns, adjectives and prepositions. Example (276) provides three simple cases to illustrate this: the prepositional coordinate structures in (276a) and (276b) function as complements of the noun ouders and the adjective geïnteresseerdinterested, respectively, and in (276c) the nominal coordinate structure functions as the complement of the preposition op. Note in passing that the interpretation of (276a) may depend on the addressee’s knowledge of Jan and Els: if they are brother and sister, we are dealing with one set of parents, whereas we are dealing with two sets of parents if they are not.

Subsection 1 has shown that conjunction of clauses is possible regardless of their illocutionary force. However, the examples in (277) show that it is not easy to conjoin clauses of different illocutionary types (although judgments vary somewhat from speaker to speaker, and may also depend on contextual factors). The (a)-examples show that declaratives and interrogatives cannot normally be mixed: the interrogative part of these conjunctions is a polar yes/no-question but the same result would occur with the wh-question Wie is ziek?Who is ill?. The (b)-examples illustrate that declaratives and imperatives cannot normally be mixed (although we will discuss some exceptional/special cases in Subsection C). The (c)-examples show that mixing declaratives and wh-exclamatives is not easy either; for convenience we have italicized the exclamative pronoun wat as well as the adjective it modifies. Note that (277c') is perfectly acceptable if the two clauses are both pronounced with an exclamation contour. This is possible because declaratives preceded by en can be used as exclamatives; cf. Jan is ziek. Oh, en Marie gaat op vakantie! Vreselijk!Jan is ill. ... And Marie will go on holiday! Terrible!.

Although coordination of a declarative clause with a clause of another illocutionary type is usually not possible, there are exceptional cases. First, it is quite common in narratives to find examples like (278a-b), where a declarative is followed by a rhetorical question. The acceptability of such examples may be due to the fact that the question is immediately answered by the speaker, so that together they simply add a new proposition to the discourse. Semantically, then, we are dealing with the coordination of two declaratives. Example (278c) seems to be of a similar type, but now the answer to the question is given not by the speaker himself but by the addressee. Examples of the type in (278) seem to play an important role in the organization/development of a narrative or a discourse.

Second, examples such as (279a) are perfectly acceptable. On the plausible assumption that the exclamative hoe is part of a reduced clause, we may have a potential counterexample to the claim that declaratives and exclamatives cannot be coordinated. The reason for this may be that the exclamative in (279a) performs a similar relation to the first clause as the manner adverb hard in the so-called specifying conjunction construction in (279a); see Section 38.4.1, sub ID, for more discussion of this construction. We assume that this modification relation between the two coordinands is sufficient to license the coordinate structure in (279a).

Since in the prototypical case declaratives cannot be combined with clauses expressing a different, more marked illocutionary force, we may expect the same to hold for all other combinations. That this expectation is indeed borne out is illustrated in (280).

The restriction on mixing clauses of different types raises the question of whether other types of mixed conjunctions are possible. The examples in (281) provide evidence that enand can link phrases of different categories.

Section 38.3 has already indicated that this kind of mixed conjunctions is only possible if the coordinands can have the same syntactic/semantic function (here: complementive) and occupy the same position as the coordinate structure as a whole: e.g. example (282c) is excluded, because nominal and clausal objects occupy different positions relative to the verbs in clause-final position. For further examples, see Section 38.3, sub I.

In nominal coordinate structures, the coordinands may differ in all nominal features: number, person and gender. Again the main restriction is that they have the same syntactic function and appear in the same syntactic position as the coordinate structure as a whole. However, we will see in Subsection B that such mixed cases are sometimes difficult to integrate into the clause.

The limited set of examples in this subsection has made it clear that there are hardly any syntactic restrictions on coordinate structures with en. First, there are no clear restrictions on the categorial status of the coordinands. Second, coordinate structures can fulfill many syntactic functions: they can be full-fledged clauses, clausal constituents of various kinds, but also parts of clausal constituents. There are various syntactic and semantic co-occurrence restrictions on the coordinands: they must be able to have the same function and occupy the same position as the coordinate structure as a whole, and clausal coordinands must usually be of the same illocutionary type. However, they do not have to be similar in all respects, as is clear from the fact that nominal coordinands can differ in number, person, and gender.

The examples in (284) show that the nominal coordinands can easily differ in person specification: a first person coordinand can co-occur with a second or third person coordinand, and a second person coordinand can co-occur with a third person coordinand. The order of the coordinands is not syntactically significant, although there are conventions of politeness and modesty that dictate that first person pronouns come last, and (less strictly) that second person pronouns follow third person pronouns.

Coordinate structures such as those in (284) are used only when needed for clarity; if the context makes it sufficiently clear who are intended, the use of the examples in (285) is much preferred.

The fact that the examples in (284) and (285) can be equivalent in certain contexts shows that the full coordinate structures in (284a) and (284b) are first and second person, respectively. This is confirmed by the binding facts in (286), which show that the coordinate structures in (284) and their corresponding pronoun in (285) both agree with the same reflexive pronoun. Similar facts arise in the case of pronominal deixis, but we will not demonstrate this here.

The above observations can easily be captured by the “elsewhere” rules in (287). Elsewhere rules apply in a fixed order: if a rule n applies, the search is terminated; if rule n does not apply, the search continues with rule n+1. The following rules adapted from Corbett (1983:176) provide an adequate description of the Dutch data in this subsection; cf. Van Koppen (2005:28-29).

The (a)-examples match rule (287a), so that the search is aborted and first person is assigned to the coordinate structure as a whole. The (b)-examples do not match rule (287a), so the search continues with (287b); this rule matches these examples, and second person is assigned to the coordinate structure as a whole. The (c)-examples do not match either rule in (287a&b), and the coordinate structure is assigned third person in accordance with the default rule (287c).

Corbett (1983/2000) claims that the rule for number resolution is relatively simple: conjunctive nominal coordinate structures are non-singular. For languages such as Dutch, which do not have dual forms, this means that they are plural. This holds for coordinate structures with proper names or referential personal pronouns regardless of whether the coordinate structure is construed as distributive or cumulative; see the examples in (288), as well as in (284)/(286) from the previous subsection.

The examples in (289) show that conjoined definite and indefinite noun phrases also trigger plural agreement. Note, however, that Haeseryn et al. (1997:1478) give an example with indefinite noun phrases for which they claim that agreement can be either singular or plural (Er zijn/is een man en een vrouw aan de deur geweestthere have/has been a man and a woman at the door); some speakers do allow singular agreement of such examples with a kind of group reading of the noun (“a man with a woman”), while other speakers consider such examples quite artificial. The “group” reading seems harder to get when the verb is dynamic, as in (289).

Corbett’s (1983/2000) formulation of the resolution rules makes it clear that he intends them to apply to coordinate structures with [+count] coordinands. That this restriction is necessary is also supported by the Dutch data in the remainder of this subsection; cf. Haeseryn et al. (1997:1476ff.), Cremers (2001), and Herringa & De Vries (2008). A first illustration of this is that conjoined non-count noun phrases do not necessarily follow this rule: examples such as (290a) can trigger singular or plural agreement on the finite verb, and in expletive constructions such as (290b) singular agreement is even preferred.

The choice of plural or singular agreement in (290a) may be related to semantics: zout en peper can be seen as referring to either “separate” or “complex” entities. This can be supported by the fact that the nouns can each take their own definite determiner, as in (291a), or share a determiner, as in (291b); the coordinate structure in (291a) strongly favors plural agreement on the finite verb, suggesting that we are dealing with two “separate” entities, while (291b) favors singular agreement, suggesting that we are dealing with one “complex” entity.

Coordinate structures with non-count nouns denoting “complex” entities can differ in agreement behavior from logically equivalent expressions with plural count nouns, as is shown by the contrast between the two examples in (292); the formal plural marking of the noun talen in (292b) triggers obligatory plural marking of the verb. This can also be seen from the fact that plurale tantum non-count nouns like hersenenbrains and ingewandenintestines also trigger plural subject-verb agreement: Er zitten/*zit hersenen in je hoofdthere are brains in your head.

Similar agreement facts can be observed with bare-inf nominalizations (without a determiner) like rokensmoking and drinkendrinking in (293): the singular agreement in (293a) is due to the fact that the habit of smoking and drinking is presented as a characteristic of a certain lifestyle, while the plural agreement in (293b) is due to the fact that this example expresses that the two habits often go together. The two primed examples show that we find a similar correlation as in (290) with respect to the number of determiners that can be used in the corresponding det-inf nominalization in the primed examples and subject-verb agreement.

The contrast is not limited to non-count nouns. Similar examples with count nouns are given in (294): the coordinate structure in (294a) refers to two persons while in the coordinate structure in (294b) refers to a single person who happens to be the speaker’s colleague as well as his best friend.

Suppose that it is justified to set aside examples like (290a), (293) and (294) because the conjunctions zout en peper, roken en drinken and collega en beste vriend are more or less fixed collocations. Even then, we cannot adopt the more general rule that all nominal coordinate structures are plural, since this clearly does not apply to coordinands containing the distributive quantifiers elk(e)each or ieder(e)every, which usually trigger singular agreement on the verb.

Similar examples are given in (296) for singular noun phrases with geenno, and the negative pronouns niemandnobody and nietsnothing. These examples sound somewhat forced; the intended meanings would normally be expressed by the shorter (and logically equivalent) disjunctions in the primed examples.

Something similar holds for conjoined generic noun phrases: the coordinate structures in the primeless examples trigger singular agreement, although Haeseryn et al. (1997:1478-9) claim that plural agreement is also possible as a marked option. It should be noted that the primeless examples in (297) are highly formal, and that the same meanings would be expressed in more informal speech by the shorter forms in the primed examples.

De Vries & Herringa (2008; section 3) suggest that the “exceptional” agreement patterns in (295) to (297) should be accounted for by appealing to the obligatory distributive (non-collective) interpretation of these examples.

Haeseryn et al. (1997:1477-8) note that the neuter nominal coordinate structure dit en datthis and that always triggers singular agreement, whereas this does not hold for its non-neuter counterpart deze en diethis and that (we will ignore here the fact that the pronouns deze and die can also be construed as plural, i.e. as “these” and “those”, although this will become relevant shortly).

Example (299) shows that the contrast arises only when the demonstratives are used as pronominal arguments, and not as modifiers of a noun; in the latter case plural agreement is obligatory.

Herringa & De Vries (2008) suggests that the special behavior of the coordinate structure dit en dat is related to the fact that dit and dat differ from deze and die in that they do not have to be used to refer to entities, but can also be used to refer to collectives or masses. Suppose that Jan and Marie divide a collection of CDs between each other. The demonstratives dit and dat are then used to refer to a subset, while deze and die are used to refer to a collection of individual items.

Furthermore, the contrast between the (a)-examples in (301) shows that die can only be used to refer to a specific contextually determined quantity of a substance denoted by a non-neuter noun; in all other cases dat must be used. In other words, the primeless examples show that only dat can be used to refer to a specific kind of substance.

The fact that the neuter pronouns dit and dat do not have to refer to contextually determined entities, but can also refer to indeterminate collectives and masses, may give us a handle to understand why the coordinate structure in (298a), dit en dat kan/*kunnen opgeruimd worden, behaves like a coordinate structure with indefinite non-count nouns in that it can trigger singular agreement on the finite verb, but it remains mysterious why singular agreement is obligatory. For completeness, note that the examples in (302), in which the conjoined demonstratives ditthis and datthat are qualified by the modifiers hierhere and daarthere, respectively, are less restrictive in that both singular and plural agreement are possible: the acceptability judgments may differ from speaker to speaker and also seem to depend on the nature of the predicate and probably various other factors. The reason for this may be that the modifiers emphasize the individual nature of the referents of the demonstratives.

A puzzle that seems similar to the previous one is that the nominal coordinate structure het een en het anderthe one thing as well as the other thing in (303a), (which speakers often tend to confuse with the indefinite noun phrase het een en andersome things) also triggers singular agreement, especially in view of the fact that the more or less synonymous example in (303b) has plural agreement. We leave this issue for future research.

The primeless examples in (304) show that coordinated subject clauses normally trigger singular agreement; note that this also holds when the anticipatory pronoun hetit is present.

The examples in (305) show that in some cases clausal coordinate structures can trigger either singular or plural agreement. The fact that this correlates with the number imposed by the nominal predicates een vervelende zaak/vervelende zaken suggests that the choice of singular or plural agreement is again related to semantics, viz. the question of whether the two clauses refer to a single state of affairs or to separate states of affairs. Observe from the primed examples that we find the same contrast when the anticipatory pronoun hetit is present.

It is more difficult to test whether the same holds for infinitival subject clauses because it seems more difficult for them to occur in subject position anyway: cf. Het is verboden [alcohol te drinken]It is forbidden to drink alcohol versus *[Alcohol te drinken] is verboden. The contrast between the two examples in (306) is therefore to be expected, and, for this reason, we will not digress on this issue here but refer to Cremers (1993:§2.2.2) for a number of apparent exceptional cases.

Example (293) has already shown that bare-inf nominalizations used as subjects allow both singular and plural subject-verb agreement; cf. [[Alcohol drinken] en [wiet roken]] is/zijn verbodenDrinking alcohol and smoking pot is/are forbidden. However, this is irrelevant in the present context, since these are noun phrases (and not clauses).

Subject-verb agreement in examples with AP subjects again depends on the interpretation: when talking about certain designs, example (307a) can be used to express that simple designs in blue are the most beautiful, while (307b) would be used to express that blue designs and simple designs are the most beautiful ones; see Haeseryn et al. (1997:1464) for similar examples.

The same is true for PP subjects, though in a more indirect way. Section A29.6 has argued that the pronoun hetit in examples such as (308) is an anticipatory pronoun which differs from those found in examples such as provided in (305) in that it does not introduce subject clauses but subject PPs. Again we can see that the agreement on the finite verb depends on contextual information.

This subsection has shown that Corbett (1983/2000) correctly restricts the resolution rule stating that conjunctive coordinate structures are plural to nominal [+count] coordinands; in other cases number agreement is semantically determined in the sense that it depends on the reference of the coordinate structure as a whole. Haeseryn et al. (1997) present a larger number of examples illustrating this, which we have skipped here because they seem formal, idiosyncratic, or obsolete.

Resolution rules for gender can be expected in two cases: determiner selection and agreement of nouns with attributively used adjectives. We start with the selection of definite articles and demonstratives: hetthe and dit/datthis/that are used for singular neuter nouns, while dethe and deze/die are used in all other cases. The examples in (309) show that it is possible to use a single article or demonstrative for coordinate structures with more than one plural noun. The semantic difference between the examples in (309) is that while (309a) can be interpreted either distributively or cumulatively, (309b) has only a cumulative reading.

The acceptability of (309b) is expected in light of the fact that the determiner de/die is used for both neuter and non-neuter plurals. The fact that neuter and non-neuter singulars select different determiners raises the question of what happens when such nouns are conjoined. The (b)-examples in (310) make it clear that it is impossible for them to share a determiner; these examples illustrate this for the definite articles de/het but the same results occur when we replace them with the demonstrative pronouns die/dat.

It is tempting to attribute the unacceptability of the (b)-examples to problems with the gender specifications of the conjoined nouns. This is not tenable, however, since we find a similar contrast between plural and singular forms when the coordinands are of the same gender; the percent sign in (311b) indicates that such cases do occur on the internet but are considered marked by our informants.

The examples in (310) and (311) show that count nouns are not suitable for investigating the resolution rules for gender, but this may be different for mass nouns. A Google search (February 24, 2017) showed that the strings in (312) all occur on the internet except for the one in (312d), which indeed strikes us as odd. The same pattern was found for strings in which the neuter noun afvalgarbage was replaced by the neuter nouns vuildirt and stofdust. The results of this search therefore suggest that there are no generally accepted resolution rules for cases like these: the coordinate structure may take the non-neuter article de, as in (312a), or the article is selected by the first conjunct, as in (312b&c); the third possibility, viz. that the article is selected by the second conjunct, is not attested.

We do not have sufficient data to show that the same pattern arises with demonstratives, since we found only three cases of the string die troep en afval, which is neutral with respect to the two competing rules; saying that the coordinate structure takes a non-neuter determiner and saying that the determiner is selected by the first conjunct both lead to the selection of the demonstrative die. We should also note that the pattern in (312) does not arise for all mixed cases; this is clear from the fact that all the word groups in (313) occur frequently on the internet and that they can all be inserted quite naturally on the dots in: Zet ... even op tafel!Please, put ... on the table!.

The contrast between (312d) and (313d) may be related to the fact that peper en zout is a more or less fixed collocation or perhaps even a morphologically complex nominal form. The latter view can be supported by the fact that this collocation often occurs as an initial part in nominal compounds like peper-en-zoutstel(letje) and peper-en-zoutsetpepper and salt set. Note that zout-en-peperstel(letje) also occurs, but less frequently. In complex nominal forms of the N1-en-N2 type, the determiner is determined by N2, as can be seen from examples such as (314), which is often cited as a prototypical example of this type. Consequently, an analysis of peper en zout as a morphological compound would account for the acceptability of (313d).

We therefore tentatively conclude that agreement pattern I in (312) is the genuine one; coordinate structures with [-count] NPs of “mixed” gender take either the determiner de or the determiner selected by its first conjunct. It should be clear, however, that the data underlying this conclusion are inconclusive and that further research is needed.

The discussion above has made it reasonably clear that establishing the resolution rules for gender is not an easy task as far as determiner selection is concerned. It is in fact even more difficult in the case of attributive adjectival inflection. This is because gender only affects the inflection of singular indefinite noun phrases: Table 13, taken from Section A28.1.1, shows that the inflectional -e ending is omitted with singular indefinite neuter nouns such as boekbook but not with singular indefinite non-neuter nouns such as stoelchair.

	singular		plural
	de-nouns	het-nouns	de-nouns	het-nouns
definite	de oude stoel the old chair	het oude boek the old book	de oude stoelen the old chairs	de oude boeken the old books
indefinite	een oude stoel an old chair	een oud-∅ boek an old book	oude stoelen old chairs	oude boeken old books

Furthermore, the examples in (315) show that the indefinite article een is like the definite articles het/dethe in that it cannot easily be used with conjoined singular noun phrases: many speakers prefer the second article to be present.

The examples in (316) further show that attributive modification is simply not allowed when the two nouns share the indefinite article. Note, however, that speakers who allow (315b) may find the examples in (316) acceptable if the adjective aardigkind is construed with the first conjunct only; they may find (316a) acceptable with the form aardige selected by jongen, and (316b) with the form aardig selected by meisje.

For speakers who reject (315b) the effect of gender can only be established by conjoined mass nouns, as in (317); again, making reliable judgments on these examples is not easy. It seems clear that the inflected form cannot be used in (317b) and that the non-inflected form cannot be used in (317b'), which leads to the conclusion already established earlier that agreement with the second noun is excluded. The non-inflected form can be used in (317b), but it does not seem possible to modify the second noun, so it is likely that we have the structure [[stinkend vuil] en [troep]], not with [stinkend [vuil en troep]]. The same holds for the non-inflected form in (317b'): it is likely that we are dealing with the structure [[stinkende troep] en [vuil]], not with [stinkende [troep en vuil]].

If our semantic intuitions are correct, we should conclude that there are no resolution rules for attributive adjectives: they simply cannot be used to modify conjoined nominal projections. Resolution rules for determiners, on the other hand, may be available: the non-neuter form de is used, or the form selected by the first conjunct. That the resolution rules for gender are less clear than those for person and number may not be accidental; it may simply bolster up Corbett’s (1983) claim that the resolution rules for person and number are widespread across languages while those for gender are more idiosyncratic and language-specific.

Two prototypical cases of asymmetric coordination are given in (320). Although all examples are syntactically sound, the primed examples seem a bit odd in that they clash with our knowledge of the world. This is due to the fact that the linear order of the coordinands seems to be interpreted as coinciding with the temporal order of the eventualities they express: cf. Dik (1968:56-7). Example (320a') is odd, because it refers to the unconventional state of affairs of Jan getting dressed in bed (before getting up), and (320b') is odd because it refers to the unconventional state of affairs of Jan undressing in the bathtub (after getting in).

Asymmetric coordination usually occurs only when the coordinands stand in a certain semantic relation and form an integrated semantic whole in the sense that “we understand the two events to be related as part of a larger event”; cf. Culicover & Jackendoff (1997). This is only possible if the eventualities referred to by the coordinands are understood as being inherently related, which is why Zhang (2010) refers to such cases as “natural” coordination. All of this amounts to saying that the temporal interpretation is a pragmatic effect triggered by our knowledge of the world. Of course, the temporal order can also be made explicit by a deictic temporal adverbial phrase, as in (321), but such cases differ from temporal asymmetric coordination in that the temporal order of the eventualities expressed by the coordinands does not have to coincide with the linear order of the coordinands: it does when daarnaafter that is used, but not when daarvoorbefore that is used.

The fact that the temporal adverbial phrase daarvoorbefore that in (321b) can be used to override the default interpretation of (320a) provides additional support for the claim that the temporal (consecutive) order should be attributed to pragmatics. For completeness’ sake, it should be noted that “natural” coordination sometimes seems to give rise to syntactic reanalysis (see 38.3, sub IIC for a possible case from English), but since this does not seem to have occurred in Dutch, we will not digress on this issue here.