The examples in (267) show that the coordinator enand is very 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), imperative (Imp) clauses can all be coordinated. The case of declaratives was already illustrated in (267a), examples with different kinds of interrogative clauses are given in the (a)-examples in (268), and the case of imperatives in (268b). 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, as shown in (268c).

The examples in (267a) and (268) are main clauses, but (269) shows that dependent clauses can also be coordinated. 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 (270) show that differences in the size of the coordinands can evoke differences in meaning. The clausal conjunction in (270a) 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 (270a'), 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 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 their execution; there are two (different) persons in (270b), and only one in (270b').

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 (271) 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 these projections may also be larger than the projections for which we have used the term NP elsewhere in this work.

Again, there are clear differences in meaning. Example (271a) 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 (271a'), on the other hand, has only one reading: we have 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 (271b) refers to two different persons and triggers plural agreement on the verb, while the coordinate structure in (271b') refers to a single person and triggers singular agreement on the verb; for a more detailed discussion of such cases, see Section N16.3.2.4. It will be clear that such differences in meaning and agreement 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 based on the examples in (272). Example (272a) 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 (272b) asserts that Jan has the two independent properties of being very young and of being inexperienced, while (272b') 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: (272b) allows an emphatic accent on en, while (272b') does not.

Complex properties are typically composed of atomic properties that “fit” together in a natural way. This is true 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 (273) are not sharp, but seem to go in this direction.

Note further that in periphrastic comparative constructions such as (274) the complex-property reading is the only one possible. This may be due to the coordinate structure constraint: the postverbal dan-phrase can apparently be associated with the full AP minder jong en onervaren in (274a), but not with the first AP-coordinand minder jong in (274b). We will not digress on this here but leave it to future research.

The primeless examples in (275) 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 an alternative analysis of the primed examples presents itself, 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 (276), which cannot be analyzed as involving conjunction of prepositions. But even if this argument is not adequate for (275), we can still conclude from (276) that prepositional projections smaller than full PPs can be conjoined, since the modifier vlak is arguably an integrated part of PPs (as can be seen from the fact that it is pied-piped under topicalization).

Examples like (275) and (276) raise the question whether syntactic heads (V, N, A, and P) can be conjoined at all. Because of the complexity of this question, and the fact that head conjunction analyses compete with conjunction reduction analyses, we will not discuss this question here, but in 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 (277).

There are certain adverbial types that do not easily allow coordination, such as the modal adverbials in (278a), 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 for which reason one might want to argue that we are dealing with a fixed collocation.

Coordinate structures with en also occur below the level of clausal coordinands. The examples in (279) 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 be constructed with ease for adjectival and prepositional phrases, so we need not repeat this here.

Finally, note that coordinate structures can be used not only as complements of verbs, but also of nouns, adjectives, and prepositions. Example (280) provides three simple cases to illustrate this: the prepositional coordinate structures in (280a) and (280b) function as complements of the noun oudersparents and the adjective geïnteresseerdinterested, respectively, and in (280c) the nominal coordinate structure functions as the complement of the preposition op.

Note in passing that the interpretation of (280a) may depend on the addressee’s knowledge of Jan and Els: if they are brother and sister, we have one set of parents, whereas we have 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 (281) show that it is not easy to combine 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 the number sign is used to indicate that it can receive a special interpretation, which will be discussed 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 (281c') is perfectly acceptable when 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 is going on vacation! 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 (282a-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 (282c) seems to be of a similar type, but now the answer to the question is given not by the speaker but (hopefully) by the addressee. Examples of the type in (282) seem to play an important role in the organization/development of a narrative or discourse.

Second, examples such as (283a) 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 (283a) performs a similar relation to the first clause as the manner adverb hard in the so-called specifying conjunction construction in (283a); cf. 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 (283a).

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 (284).

The restriction on mixing clauses of different types raises the question whether other types of mixed conjunctions are possible. The examples in (285) provide evidence that enand can link phrases of different syntactic categories.

Section 38.3, sub I, has already shown that this kind of mixed conjunctions is only possible if the coordinands have the same syntactic/semantic function (here: complementive) and can occur in the same position as the coordinate structure as a whole. This can be seen from the fact that example (286c) is excluded; the reason is that nominal and clausal objects occupy different positions relative to the verbs in clause-final position.

The nominal coordinate structures in (287) show that the coordinands can 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 complete 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 (288) show that nominal coordinands in conjunctions 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 (288) are used only when necessary for clarity; if the context makes it sufficiently clear who are meant, the use of the examples in (289) is much preferred.

The fact that the examples in (288) and (289) can be equivalent in certain contexts shows that the coordinate structures in (288a) and (288b) are first and second person, respectively. This is confirmed by the binding facts in (290), which show that the coordinate structures in (288) and their corresponding pronoun in (289) 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 are easily captured by the “elsewhere” rules in (291). Elsewhere rules apply in a fixed order: if 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 in in (288) match rule (291a), so that the search is terminated and first person is assigned to the coordinate structure as a whole. The (b)-examples do not match rule (291a), so the search continues with (291b); 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 (291a&b), and the coordinate structure is assigned third person in accordance with the default rule (291c).

Corbett (1983/2000) claims that the rule for number resolution is simple: conjunctive nominal coordinate structures with enand 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 interpreted as distributive or cumulative; cf. the examples in (292).

The examples in (293) show that conjoined definite and indefinite noun phrases also trigger plural agreement. Note, however, that Haeseryn et al. (1997:1478) gives 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 indeed allow singular agreement in such examples with a kind of group reading (“a man with a woman”), while other speakers (including ourselves) consider such examples to be quite artificial. The group reading seems harder to get with a dynamic verb, as in (293).

The formulation of Corbett’s (1983/2000) resolution rule makes it clear that he intends it to apply to coordinate structures with [+count] coordinands. That this restriction is necessary is 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 (294a) can trigger singular or plural agreement on the finite verb, and in expletive constructions such as (294b) singular agreement is even preferred.

The choice of plural or singular agreement in (294a) 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 (295a), or share a determiner, as in (295b); the coordinate structure in (295a) strongly favors plural agreement on the finite verb, suggesting that we are dealing with two “separate” entities, while (295b) favors singular agreement, suggesting that we are dealing with one “complex” entity; we will return to the choice of article in (295b) in the next subsection.

Coordinate structures with non-count nouns denoting “complex” entities can differ in agreement behavior from logically equivalent expressions with plural count nouns, as shown by the contrast between the two examples in (296); the formal plural marking of the noun talen in (296b) 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 (297): the singular agreement in (297a) 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 (297b) 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 (294) 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 (298): the coordinate structure in (298a) refers to two persons, while the coordinate structure in (298b) refers to a single person who happens to be the speaker’s colleague as well as his best friend.

The above examples suggest that Corbett’s resolution rule for number, i.e. that all ([+count]) nominal coordinate structures are plural, does not do full justice to the actual situation. Even if we decide to ignore examples like (294a), (297) and (298) because the conjunctions zout en peper, roken en drinken and collega en beste vriend are more or less fixed collocations, we still cannot adopt this rule, because it clearly does not apply when the coordinands contain distributive quantifiers like elk(e)each or ieder(e)every in (299); the primed examples show that in such cases, we usually find singular agreement on the verb.

Similar examples are given in (300) for singular noun phrases with geenno, and the negative pronouns niemandnobody and nietsnothing. Note, however, that these examples sound somewhat forced; the intended meanings would normally be expressed by the shorter (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) claims that plural agreement is also possible as a marked option. It should be noted that the primeless examples in (301) 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) suggests that the “exceptional” agreement patterns in (299) to (301) should be accounted for by appealing to the obligatory distributive (non-collective) interpretation of these examples.

Haeseryn et al. (1997:1477-8) notes 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 (but note that the latter can also be used as plural “these” and “those”, as this will become relevant shortly).

Example (303) 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 it differs from the coordinate structure deze and die in that it does not have to be used to refer to entities, but can also be used to refer to collectives or masses. For example, if Jan and Marie divide a collection of CDs among themselves, dit and dat in (304a) can simply refer to a stack of indeterminate CDs, whereas deze and die in (304b) would rather refer to a collection of specifically selected items.

The contrast between the (a)-examples in (305) further 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.

Now that we have established that the neuter demonstratives dit and dat do not have to refer to contextually determined entities, but can also refer to collectives, masses, and kinds, it is easier to understand why the coordinate structure in (302a), 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; however, it remains a mystery why singular agreement is obligatory. For completeness, the examples in (306) show that the coordinate structure dit en datthis and that is less restrictive when the demonstratives are modified by hierhere and daarthere; then both singular and plural agreement are possible (although 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 these modifiers hier and daar 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 (307a), (which speakers often tend to confuse with the indefinite noun phrase het een en andersome things) also triggers singular agreement, especially because the more or less synonymous example in (307b) has plural agreement. We leave this issue to future research.

The primeless examples in (308) show that coordinated subject clauses normally trigger singular agreement; note that this also holds when the anticipatory pronoun hetit is present.

The examples in (309) show that in some cases clausal coordinate structures can trigger either singular or plural agreement. The fact that this correlates with the number expressed on the nominal predicate een vervelende zaak/vervelende zaken suggests that the choice of singular or plural agreement is again related to semantics, i.e. the question of whether the two clauses refer to a single state of affairs or to separate states of affairs. Notice 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 applies to 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 (310) is therefore to be expected, and for this reason we will not digress on this here, but refer to Cremers (1993: §2.2.2) for a number of apparent exceptional cases.

Recall that example (297) 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. This is irrelevant in the present context, since these are noun phrases (and not clauses).

Subject-verb agreement with AP-subjects again depends on interpretation: when talking about certain designs, example (311a) can be used to express that simple designs in blue are the most beautiful, while (311b) would be used to express that blue designs and simple designs are the most beautiful; cf. Haeseryn et al. (1997:1464) for similar examples.

The same is true for PP subjects, though in a more indirect way. Section A28.6 has argued that the pronoun hetit in examples like (312) is an anticipatory pronoun which differs from those found in examples like (309) in that it does not introduce subject clauses but subject PPs. Again we see that agreement on the finite verb depends on contextual information.

This subsection has shown that Corbett’s (1983/2000) resolution rule for number correctly states that conjunctive coordinate structures with definite nominal [+count] coordinands are plural. This does not mean, however, that coordinate structures with enand are always plural; we have seen that this depends on various properties of the coordinands and/or the reference of the coordinate structure as a whole. Finally, we note that Haeseryn et al. (1997) presents other cases that we have decided to omit 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 (313) 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 two examples is that (313a) can be interpreted either distributively or cumulatively, while (313b) has only a cumulative reading.

The acceptability of (313b) is to be expected in view 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 take different determiners raises the question of what happens when such nouns are used. The (b)-examples in (314) make 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 (315b) indicates that such cases do occur on the internet but are considered marked by our informants.

The examples in (314) and (315) show that count nouns are not suitable for studying the resolution rules for gender, but this may be different for mass nouns. A Google search (August 28, 2024) showed that the strings in (316) all occur on the internet except for the one in (316d), which seems extremely odd indeed. A similar pattern was found for cases in which the neuter noun afvalgarbage was replaced by the neuter noun stofdust (but no cases were found with the neuter noun vuildirt as the first coordinand in (316a)). The results of this search suggest that there may be no generally accepted resolution rules for cases like these: the article is selected by the first conjunct, as in (316b&c), or the coordinate structure may take the non-neuter article de (probably as a less preferred option, since we found very few cases), as in (316a).

We have no data to show that the same pattern arises with demonstratives, since we found only three relevant 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 (316) does not arise for all mixed cases; this is clear from the fact that all the word groups in (317) 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 (316d) and (317d) 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 (318), 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 (317d).

We therefore tentatively conclude that agreement pattern I in (316) is the correct one; the determiner in coordinate structures with [-count] NPs of “mixed” gender is either non-neuter de, or is selected by the first coordinand (which is probably the preferred option). The above excursus is based on sparse and inconclusive data, and is intended to encourage further research that can give a better picture of the problem.

The discussion above has shown that it is not an easy task to determine the resolution rules for gender in the case of determiner selection. It is 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 A27.1.1, 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 (319) 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 (320) further show that attributive modification is simply not allowed when the two nouns share the indefinite article. Note, however, that speakers who allow (319b) may (marginally) accept these examples when the adjective aardigkind is construed with the first conjunct only; they may find (320a) more acceptable with the form aardige selected by jongen, and (320b) more acceptable with the form aardig selected by meisje.

For speakers who reject (319b), the effect of gender can only be established with conjoined mass nouns, as in (321); again, making reliable judgments on these examples is not easy. It seems that the inflected form cannot be used in (321b) and the non-inflected form cannot be used in (321b'), which again leads to the conclusion that agreement with the second noun is excluded. The non-inflected form can be used in (321b), 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 [stinkend [vuil en troep]]. The same goes for the non-inflected form in (321b'): it is likely that we have the structure [[stinkende troep] en [vuil]], not [stinkende [troep en vuil]].

If these 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.