Examples (402a&a') show that when a coordinated noun phrase with only one article is relativized, the coordinated antecedent reading is usually strongly preferred to the non-coordinated one. The non-coordinated antecedent reading is only possible (and then still somewhat marked) if the second conjunct is given a generic interpretation, as in (402b'), which requires that the second conjunct be given extra emphasis.

Note that the coordinated antecedent reading in the primeless examples can be derived without appealing to backward conjunction reduction, since the coordination does not take place at the level of the DP, but at some lower level in the NP-domain. This is illustrated for example (402a) in (403).

When only the second conjunct is preceded by an article, as in (404), a coordinated antecedent reading is excluded: only the non-coordinated antecedent reading in (404b) is available, meaning that all men, and the boys who are late, will be denied admission.

Since the article is not overtly expressed in indefinite plurals, there is no point in trying to distinguish between a one-article and a two-article reading. However, when numerals are used instead of the null article ∅, the distinction does become relevant. The coordinated antecedent reading in (405a), according to which the noun phrase refers to a set of four persons consisting of both men and boys, is clearly preferred to the non-coordinated antecedent reading in (405b), according to which there is a set of four men as well as a non-qualified set of boys restricted by the relative clause.

With example (406b) we aim at triggering a generic reading of the noun phrase jongens. The fact that this example is degraded is probably due to the fact that this is only possible if the first conjunct is also interpreted generically, which is blocked by the presence of the numeral vierfour; if this numeral is omitted, the example becomes perfectly acceptable with the intonation pattern typical for the non-coordinated antecedent reading. Note in passing that, in contrast to (405a), the non-generic example in (406a) seems to require a partitive reading, which may be due to the fact that the sentence has a future interpretation, so that at the moment of speech it is still unknown how many men and boys will actually be late.

As with the definite article, a coordinated antecedent reading is not possible when only the second conjunct is preceded by a numeral, as in (407); only the non-coordinated antecedent reading in (407b) is available, meaning that all men as well as four boys who were late were denied admission.

Omitting the second article leads to a degraded result with both indefinite and definite singular antecedents. This is not related to the presence of the relative clause, since we see the same thing in examples like *?Een/de man en jongen werden niet toegelaten and *?Jan las een boek en artikel. Insofar as the examples in (408) are acceptable, they seem to be interpreted with a coordinated antecedent reading: the non-coordinated antecedent reading is completely excluded. Note that the gender of the nouns does not play a role since the gender of the conjuncts is the same: non-neuter in (408a) and neuter in (408b). Nor is the syntactic function of the modified noun phrase relevant: it functions as subject in (408a) and as object in (408b). Thus, the reason for the degraded status of the primeless examples must be that a single determiner cannot be used to modify two singular nouns.

Now that we have considered the normal case in which coordination results in a plural noun phrase, we need to discuss a special case with coordinated singular elements in which the conjuncts need not refer to two different entities. In example (409a), for instance, the coordinated structure must be interpreted as referring to a single (non-specific) person. On the other hand, when the second determiner is present, as in (409b), the only possible interpretation is one in which only the second conjunct is modified by the relative clause.

Example (410) shows that the same contrast occurs in the case of a coordinated antecedent with specific reference, although the singular reference reading in (410a) may be slightly marked. Example (411) shows that the singular reference reading is also less readily available with [-human] referents.

The examples in (412) show that the singular reference reading of coordinated antecedents is also available in the case of a definite determiner. Notice that in accordance with the fact that the coordinated antecedent has only one referent, the finite verb of the main clause in (412a) appears in the singular form moethas to. In (412b), where the subject consists of two separate DPs, the finite verb of the main clause appears in the plural form moetenhave to. As usual, the (b)-example is ambiguous for some people between a coordinated and a non-coordinated antecedent reading.

When the modified noun phrase functions as the object of the clause, we find the same contrast, although now the number marking on the verb does not help to distinguish the two readings. The example in (413) is ambiguous between the singular reference reading, with the structure in (413a), and the non-coordinated antecedent reading, with the structure in (413b). Note that the coordinated antecedent reading is ruled out by the fact that the relative pronoun triggers singular agreement on the finite verb of the relative clause.

The structure of the examples with the singular reference reading is as given in (414), where the relative pronoun takes some lower coordinated nominal projection as its antecedent.

This structure immediately accounts for the fact that extraposition of the relative clause is possible; since the two coordinated elements act as the antecedent of the relative clause, extraposition does not violate the coordinate structure constraint. In this respect, relative constructions with a singular reference reading behave like constructions with a non-restrictive relative clause; cf. Subsection IB.

When the coordination involves plural elements, the coordinated antecedent reading is the most readily available one; the non-coordinated antecedent reading is at best marginally possible in a (very unlikely) generic interpretation. Due to the past tense, such a generic interpretation is not available in (416a'), and for this reason the non-coordinated antecedent reading is completely ruled out. The present tense in the two (b)-examples favors a generic reading, but the non-coordinated antecedent reading is still very hard to get.

The coordinated antecedent reading in the primeless examples can again be derived without appealing to backward conjunction reduction. The only difference between these examples and the structure in (403) is that here the coordination does not take place at the NP-level, but at some higher level within the DP. This is illustrated for the primeless examples in (416) in (417).

If only the second conjunct is preceded by an article, a coordinated antecedent reading is impossible. The non-coordinated antecedent reading is again quite degraded.

Since in plural indefinite noun phrases the article is not overtly expressed, there is no point in trying to distinguish between a one-article and a two-article reading. However, when numerals are used instead of the indefinite null article ∅, the distinction does become relevant. This is illustrated in (419). Only the coordinated antecedent reading in (419a) is available, according to which the antecedent refers to a set of four persons, consisting of both men and boys; it is this entire set that is subsequently restricted by the restrictive relative clause. On the unavailable non-coordinated antecedent reading in (419b), it would be claimed that there is a set of four men and a non-qualified set of boys, with only the latter set being modified by the relative clause.

The generic example in (420b) is degraded, as is the corresponding case with a restrictive relative clause, which can probably be attributed to the fact that the first conjunct vier mannen must also be interpreted generically, which is incompatible with the presence of the numeral. In contrast to the corresponding example in (405b) with a restrictive relative clause, (420a) does not seem to allow a partitive interpretation, which may be due to the fact that at speech time it is not yet known how many men and boys will be late, so that the reader is actually not able to provide any additional information about the referent set.

If only the second conjunct is preceded by a numeral, as in (421), neither the coordinated nor the non-coordinated antecedent reading is available.

When the conjuncts are singular, non-restrictive relative clauses do not differ from restrictive ones. Omitting the second article leads to a severely degraded result, both with indefinite and with definite singular antecedents. This is not due to the presence of the relative clause since we observe the same thing in examples such as *?Een/de hond en kat werden niet toegelaten. To the extent that example (422a) is acceptable, it must be interpreted with a coordinated antecedent reading, signaled by the plural finite verb in the relative clause: the non-coordinated antecedent reading in (422b), signaled by a singular finite verb in the relative clause, is completely ruled out.

As in the case of restrictive relative clauses, the two elements of the antecedent need not refer to two different entities if the second conjunct is not preceded by an article. For example, the coordinated antecedent in (423a) can only be interpreted as referring to a single (non-specific) person. On the other hand, when the second determiner is present, as in (423b), the only possible interpretation is that only the second conjunct is modified by the relative clause.

The examples in (424) show that the same contrast seems to occur in the case of a coordinated antecedent with specific reference, although the singular reference reading in (424a) is slightly marked. Example (425a) serves to show that singular reference readings are less readily available with [-human] referents.

The examples in (426) show that the singular reference reading of coordinated antecedents is also available in the case of a definite determiner. In accordance with the fact that the coordinated antecedent has only one referent, the finite verb of the main clause (426a) appears in the singular form wonwon. In (426b), where the subject consists of two separate DPs, the finite verb of the main clause appears in the plural form wonnenwon.

We find the same contrast when the noun phrase functions as the object of the clause, although the number marking on the finite verb in the main clause does not help to distinguish the two cases: (427a) has the singular reference reading, and (427b) has the non-coordinated antecedent reading. Note that the coordinated antecedent reading is excluded in (427b) by the fact that the relative pronoun triggers singular agreement on the finite verb of the relative clause.

The structure of the examples with the singular reference reading is as given in (428), which differs from the one in (414) with restrictive relative clauses in that the relative pronoun is external to the NP and takes some higher coordinated nominal projection as its antecedent.

This structure explains why extraposition of the non-restrictive relative clause is possible; since the two coordinated elements act as the antecedent of the relative clause, extraposition does not violate the coordinate structure constraint.

If the two conjuncts of a coordinated antecedent are singular, ambiguity between the coordinated and non-coordinated antecedent reading will not arise if the relative pronoun functions as the subject of the relative clause, since the number marking on the finite verb of the relative clause will then have a disambiguating function; cf. Subsection II. If the two conjuncts differ in gender, the difference between the two readings becomes even clearer, since in this case differences in gender are directly reflected in the choice of the relative pronoun: whereas dat is used for neuter singular antecedents, die is used in all other cases. Thus, in example (430a), both the form die of the relative pronoun and the plural finite verb of the relative clause preclude a reading in which the relative clause would modify only the neuter singular noun meisje, since this noun would require the pronoun dat and trigger singular agreement on the finite verb of the relative clause, as in example (430a'). The same is true for the examples in (430b&b'), where the potential ambiguity is also resolved on the basis of the form of the relative pronoun and the finite verb in the relative clause.

If we replace the indefinite article w a definite one, the result of coordinated antecedent reading in the primeless examples is degraded for many speakers.

Note that the primeless examples in (431) create problems for the backward conjunction-reduction analysis similar to those we encountered in Subsection II. According to this analysis these examples have the structure in (432), but there are at least two reasons to think that these structures are ungrammatical. First, the presence of a plural finite verb in the relative clause requires a plural subject, and since it is the relative pronoun that performs this syntactic function, that pronoun cannot take a singular antecedent. Second, the form die of the relative pronoun precludes a structure in which the relative clause restricts the two coordinated elements separately, since it cannot take the neuter singular noun meisje as its antecedent. We will not pursue the question of how such constructions can be adequately represented, but refer to Subsection II for relevant discussion.

The discussion above does not imply that backward conjunction reduction is completely excluded when the two conjuncts differ in gender. Conjunction reduction is, however, restricted in such a way that the omitted part is completely identical to the second part, as in the primeless examples in (433). Although these examples are rather complex from a perceptual point of view and are more likely to be found in written text than in speech, they seem well-formed. The most plausible analysis for these examples is the one given in the primed examples, where part of the first relative clause is omitted as a result of backward conjunction reduction.

Subsection III has shown that in cases where the second conjunct is not preceded by an article, a coordinated antecedent can have a single reference reading. The examples in (434) show, however, that this requires that the two conjuncts have the same gender: the diminutive neuter noun vriendjefriend and the non-neuter noun kameraadpal cannot share the same relative clause, since they require different relative pronouns, although it should be noted that those constructions in which the relative pronoun matches the gender of the second conjunct seem slightly better than those in which it matches the gender of the first.

Of course, the examples in (434) become perfectly acceptable with the relative pronoun die, if we replace the neuter diminutive form vriendje with the non-neuter form vriend; cf. Ik zoek een vriend en kameraad die dezelfde hobby heeft als ik and Ik zoek een kameraad en vriend die dezelfde hobby heeft als ik.

Non-restrictive relative clauses also allow the two elements of a coordinated antecedent to differ in gender. When the two elements are singular and the second conjunct is headed by a neuter noun, as in (435), the difference between the coordinated and non-coordinated antecedent reading can be detected on the basis of the choice of relative pronoun: die is used for [-neuter] and dat is used for [+neuter] singular antecedents. In the primeless examples in (435), the form die of the relative pronoun (as well as the plural form of the finite verb of the relative clause) precludes a reading in which the relative clause would modify only the second conjunct. In the primed examples, the presence of the singular [+neuter] pronoun dat (as well as the singular form of the finite verb of the relative clause) forces the non-coordinated reading, in which the relative clause takes the neuter noun meisje as its antecedent.

Subsection III has shown that coordinated antecedents preceded by a single article with a single reference reading can be modified by a non-restrictive clause. The examples in (436) show that, as in the case of non-restrictive relative clauses in (434), the two conjuncts must be of the same gender: the neuter noun vriendjefriend and the non-neuter noun kameraadpal cannot share the same relative clause, since they require different relative pronouns. Again, these examples become perfectly acceptable if we replace the neuter diminutive form vriendje with the non-neuter form vriend.

When the conjuncts differ in number and the relative pronoun functions as the subject of the relative clause, as in (437), the relative order of the conjuncts can have a disambiguating effect. The (a)-examples show that when the plural conjunct precedes the singular one, the form of the finite verb in the relative clause resolves the potential ambiguity: if it is plural we are dealing with the coordinated antecedent reading; if it is singular we are dealing with the non-coordinated antecedent reading. On the other hand, if the singular element precedes the plural element, ambiguity does arise: since only the plural form of the verb is allowed, both the coordinated and the non-coordinated antecedent reading are possible. Note that, as usual, intonation will serve to disambiguate the two readings in speech.

For many speakers, the coordinated antecedent reading is less acceptable in the case of coordination of two definite phrases; in this respect, the primeless examples in (438) behave like cases in which the conjuncts do not differ in number. The non-coordinated reading in the primed examples is fully acceptable.

The examples in (439) and (440) show that non-restrictive relative clauses can easily take as their antecedent a coordinated antecedent with conjuncts that differ in number: all combinations and interpretations are easily possible in both indefinite and definite phrases.

When two coordinated elements differ in definiteness, it is virtually impossible to interpret the restrictive relative clause as modifying both elements. Thus, the primeless examples in (441), in which the plural form of the finite verb of the relative clause forces a coordinated antecedent reading, are clearly unacceptable. The non-coordinated antecedent readings in (441), on the other hand, are perfectly acceptable. The examples in (442) also clearly favor the non-coordinated antecedent reading.

The examples in (443) and (444) show that non-restrictive relative clauses with coordinated antecedents consisting of a definite and an indefinite conjunct do not differ significantly from restrictive relative clauses. Again, only the non-coordinated antecedent readings are perfectly acceptable.

Example (445a) shows that quantified elements that are formally plural can form a coordinated antecedent of a restrictive relative clause, although the marked status of (445b) suggests that acceptability may depend on the combination of quantifiers. The primed examples show that the non-coordinated antecedent reading is always readily available.

If the quantified elements are formally singular, the coordinated structure is also formally singular, as shown by the fact that the coordinated subject in (446) triggers singular agreement on the verb. Note that it is possible to omit the second quantifier.

This implies that if the relative pronoun functions as the subject of the relative clause, the finite verb of the relative clause must also be singular, as in the examples in (447). If the second quantifier is not present, the non-coordinated antecedent reading is completely impossible. On the other hand, when the second quantifier is present, the non-coordinated antecedent reading seems to be preferred. The reason for this may be that the structures in (447a) and (447b) compete, and the former is preferred because it does not require the postulation of elided structure; cf. the discussion of example (380) in Subsection I.

The fact that (447b) is still reasonably acceptable may follow from the fact that backward conjunction reduction leads to the acceptable representation in (448a). An argument for such an analysis is that the two conjuncts in examples such as (447b) must be of the same gender; the (b)-examples (448) show that if the conjuncts are of different genders, the first relative pronoun cannot be deleted by conjunction reduction, but must be realized overtly. Example (448b') may be slightly marked, but improves if the verbs are replaced by verbs in the present tense, which would allow generic reading: Elke jongen die en elk meisje dat te laat komt, wordt gestraftany boy or girl who is late will be punished.

When a quantified and a non-quantified element are coordinated, as in (449), the coordinated antecedent reading in the primeless examples seems questionable. The non-coordinated antecedent reading in the primed examples, on the other hand, is perfectly acceptable.

It is difficult to judge whether non-restrictive relative clauses can take a quantified antecedent at all. In many potential cases, an appositive interpretation seems more readily available, although it must be said that it is often difficult to distinguish the two cases.

Given this, it is not surprising that non-restrictive relative clauses with coordinated quantified antecedents never give rise to a very felicitous result: both the coordinated antecedent reading in (451a) and the non-coordinated antecedent reading in (451b) are somewhat marked. Note that the latter seems to become fully acceptable when the modifier is given an appositional interpretation, as in (451b').

When quantified and non-quantified elements are coordinated, the coordinated antecedent reading is not available. The examples in (452) show that the construction is only marginally acceptable with the non-coordinated antecedent reading in (452b), or with the appositional interpretation in (452b').