As a logical consequence of their non-restrictive function, non-restrictive relative clauses can easily be used in combination with antecedents with unique referents. Since these referents can be assumed to be identifiable, the relative clauses need not, and typically cannot, contain identifying information. This is illustrated in (324) for an antecedent in the form of a proper noun, a noun with unique reference, and antecedents containing a demonstrative or possessive determiner.

Non-restrictive relative clauses can also be used to modify personal pronouns, provided that they have the full, non-reduced form. If the antecedent and the relative pronoun have the same syntactic function in the matrix and the relative clause, such constructions are perfectly acceptable. This is illustrated in (325) for cases where both the antecedent and the relative pronoun function as subjects; note that the finite verb of the relative clause agrees in number with the antecedent pronoun.

In (326) the same thing is shown for object pronouns: the pronouns haarher and onsus function as direct objects in the matrix clauses, with the relative pronoun die fulfilling the same function in the relative clauses, and the result is fully acceptable.

When the antecedent and the relative pronoun do not have the same syntactic function, the results are generally marked. The examples in (327) show this for cases where the personal pronoun functions as the direct/indirect object of the matrix clause or as the complement of a preposition, while the relative pronoun is the subject of the relative clause. In (327a&b), some speakers allow and even prefer the third singular heeft to the first singular heb.

The examples in (328) show the same thing for cases where the personal pronoun acts as the subject of the matrix clause and the relative pronoun acts as the direct or indirect object of the relative clause.

The examples in (327) and (328) marked with a question mark are all cases where the personal pronoun functions as a subject and the relative pronoun as an object, or vice versa. When they function as direct and indirect objects, respectively, the constructions are perfectly acceptable. Examples are given in (329).

The data on personal pronoun antecedents suggests that the personal pronoun can act as the antecedent of a relative pronoun with a different syntactic function, as long as the personal pronoun has the morphological form that “matches” the syntactic function of the relative pronoun: if this is not the case, a marked result arises. This would explain the fact that the examples in (330) are perfectly acceptable despite the fact that the plural pronoun jullieyou acts as the subject in the main clause, while the relative pronoun acts respectively as a direct object, an indirect object and the complement of a preposition. This may be due to the fact that the form jullie can be used in all these functions. We will return to pronouns modified by a non-restrictive relative clause in Subsection IIID.

The examples in (331) show that non-restrictive relative clauses can also have an indefinite antecedent, i.e. an antecedent whose referent is assumed to be unidentifiable to the hearer. The relative clauses do not serve to restrict the set of possible referents, but simply provide additional information about the referent of the antecedent.

In (331) the antecedent is interpreted as specific; the identity of the intended referent(s) is known to the speaker but not to the hearer. Indefinite antecedents of non-restrictive relative clauses can also be generic, as in (332), in which case the relative clause is interpreted as providing generic information; in both the case of a plural antecedent and in the case of a singular antecedent, the information given in the relative clause must be taken to apply to the entire class of entities denoted by the antecedent, i.e. to all students.

It is less clear whether non-restrictive relative clauses can be used to modify non-specific indefinite antecedents, i.e. noun phrases referring to entities unknown to the speaker. Example (333a) is perfectly acceptable, but it is not immediately clear whether we should construe the modified noun phrase as non-specific or generic; cf. Section 19.1.1.5, sub IC2. The most prominent reading of example (333b) is one in which the noun phrase is construed as specific, i.e. as known to the speaker; the non-specific interpretation of the noun phrase seems to yield a marked result, and to favor an appositional reading of the relative clause. Although the judgments are somewhat subtle, we conclude from this that it is impossible to modify non-specific indefinite noun phrases by means of a non-restrictive relative clause, which could be attributed to the fact that speakers cannot provide additional information about entities not familiar to them.

Non-restrictive relative clauses can be used to modify nominal predicates, provided that the relative pronoun functions as the predicate of the relative clause. The examples in (334) show that the relative pronoun in such cases is always of the form wat.

When the relative pronoun functions as an argument in a non-restrictive relative clause, it is sometimes difficult to determine what the antecedent of the relative clause is. Example (335a), for example, can easily be misanalyzed as a case in which a non-restrictive relative clause modifies the predicate een dwaasa fool. The correct analysis is the one in which the relative clause provides some specific information about the noun phrase die manthat man, which means that the relation of the relative clause to the nominal predicate is more indirect: the fact that the man always does as he is told is the reason why he is considered a fool. This use of the relative clause is characterized by a primary accent on the relative clause, which provides new information about the antecedent. That the relative clause does not modify the predicate in examples such as these is clear from the fact, illustrated in (335b), that the pronoun die is replaced by its neuter counterpart dat when the non-neuter subject die man is replaced by the neuter noun phrase het meisjethe girl. From this we can safely conclude that we are dealing with a relative clause in extraposed position that takes the subject of the clause as its antecedent, which is also supported by the fact that the primed examples are also acceptable.

Essentially the same thing is shown in (336), where the nominal predicate is the neuter noun geniegenius. Again, the form of the relative pronoun depends on the gender of the subject of the clause, not on that of the predicate.

A complicating factor for the examples in (335) and (336), which we have ignored in the discussion above, is that it is not entirely clear whether we are really dealing with non-restrictive relative clauses: Subsection IIIA, will show that it is normally impossible to extrapose such clauses from subjects in the clause-initial position, which means that we may actually be dealing with appositions. However, this does not affect the conclusion we can draw from the data discussed so far, namely that a nominal predicate cannot be the antecedent of a non-restrictive relative clause when the relative pronoun functions as an argument.

A potential problem for such a claim is presented by the somewhat marked examples in (337). In these examples, the relative clause is generic in the sense that it provides information about the whole class of fools/genial people: this use of the modifying clause is characterized by a primary accent on the (adverbial) element expressing the generic nature of the relative clause. The fact that the relative clause provides information about the class denoted by the predicate makes it plausible to assume that it is not the subject but the predicate that functions as the antecedent of the relative clause.

Note, however, that the relative pronoun does not agree in number with the nominal predicate: the predicate is singular, whereas the relative pronoun, which functions as the subject of the relative clause, triggers plural agreement on the finite verb. Note also that full agreement between the relative pronoun and the neuter nominal predicate een genie in (337b) would require the former to have the form dat (and not the plural form die). This lack of number and gender agreement suggests that we are not dealing with a relative construction in (337) at all, but with a construction of some other kind. In this context it may be useful to consider the sequences in (338), where the anaphoric plural pronoun in the second sentence also refers to all members of the class referred to by the singular generic subjects of the first sentence.

From the discussion in this subsection, we conclude that non-restrictive relative clauses can only be used if the relative pronoun also functions as a predicate, in which case the pronoun must be of the form wat.

The additional information provided by the non-restrictive relative clause is not always restricted to the referent of the antecedent; often the relative clause has an implicit adverbial-like relationship with the matrix clause. In example (340a), for instance, the relative clause can be construed as the reason for the immediate purchase of the book. Likewise, the relative clauses in (340b-d) are all likely to be given a similar adverbial-like interpretation, expressing cause in (340b), concessive contrast in (340c), and a temporal relation in (340d).

In examples such as these, the non-restrictive relative clause is necessary for the correct interpretation of other elements in the matrix clause; for example, the adverbs nog steedsstill in (340b) and the modal particle tochafter all in (340c) can only be interpreted on the basis of the information given in the relative clause; similarly, the correct interpretation of the adverbial phrases weeragain and de volgende dagthe next day in (340d) depends on the information given in the relative clause. Omitting the relative clauses in these cases yields a grammatically correct but infelicitous result (unless the context provides the relevant information).

Non-restrictive relative clauses are usually used to present additional or background information about the antecedent, as in (339), or about the antecedent and the event described in the matrix clause, as in (340). In both cases, the role of the relative clause is restricted to the sentence and does not play a crucial role in the development of the discourse (conversation, story, argument, etc.). In some cases, however, non-restrictive relative clauses in sentence-final position may have almost the status of a matrix clause, both in terms of importance and discourse continuity. Such non-restrictive relative clauses are often called “continuative” or “consecutive”. Although from a purely syntactic point of view such relative clauses can be left out, the omission of the relative clause would lead to an information gap and thus to an incoherent discourse. First, consider the example in (341), where the information provided by the relative clause is clearly background information, as shown by the fact that the addition of the modifier overigensby the way is perfectly acceptable.

In (342a), on the other hand, the relative clause is a crucial link in the discourse chain. As such, the use of overigens is infelicitous, while a modifier like vervolgenssubsequently, which serves to enhance the discourse coherence, is perfectly acceptable. The sequence in (342a) is therefore very close to the sequence in (342b), where the same information is provided in a matrix clause.

This subsection briefly mentions some of the properties of the cleft construction, because this construction contains a phrase that closely resembles a relative clause. Despite the fact that there is no intonation break between the antecedent and the modifying clause, we will analyze this modifying clause as non-restrictive, because it does not restrict the (possibly singleton) referent set of the antecedent, but modifies this antecedent as a whole. Such an analysis is supported by the fact that the antecedent can in any case take the form of a proper noun or a uniquely referring expression (Smits 1989:203).

As can be seen from the examples in (343), cleft constructions characteristically contain the copular verb zijn and the impersonal pronoun hetit. The modifying clause seems to contain a relative pronoun which takes the non-pronominal phrase (which need not be a DP) as its antecedent. The function of the cleft construction as a whole is to emphasize the referent set of the antecedent, which is always given focal/contrastive accent.

The relative clause fulfills the crucial function of linking this antecedent to the ongoing discourse by providing additional information. The relative clause in (343a), for example, clearly does not function to restrict the set of all Americans, but instead provides additional information about this set as a whole. This additional information links the antecedent to the preceding discourse, which is clear from the fact that the relative clause contains the deictic demonstrative pronoun ditthis, which can only be interpreted by appealing to information from the preceding context. If we abstract from the contrastive function of the cleft construction, (343a) provides more or less the same information as the main clause De Amerikanen hebben dit voor het eerst ontdektThe Americans discovered this first. This means that the omission of the relative clause renders the construction infelicitous, since it deprives the addressee of the information needed to properly relate the Americans to the topic of discussion, and would leave the addressee wondering why reference is being made to the entities denoted by the nominal predicate.

The possibility of extraposition of non-restrictive relative clauses seems to be more or less the same as in the case of restrictive relative clauses discussed in Section 17.3.2.3.2, although the result always tends to be slightly marked. Furthermore, it should be noted that giving judgments is often complicated by the fact that the resulting strings are generally also acceptable on an appositive reading, in which case the clause is preceded by a very distinct intonation break (a pause and usually a falling intonation, much more pronounced than in the case of non-restrictive modifiers), which separates it from the preceding material and emphasizes its parenthetical nature.

Section 17.3.2.3.2, sub II, has shown that it is possible to extrapose a non-restrictive relative clause from a subject, provided that the latter does not occupy the canonical subject position to the immediate right of the complementizer. The examples in (345) show that the same is true for non-restrictive relative clauses. While (345a) is acceptable only if pronounced with the intonation pattern typical of an appositional reading, the examples in (345b'&c') do not require this.

Example (346b) shows that extraposition of a non-restrictive relative clause from a direct object antecedent seems to be possible: of course, we may also be dealing with an apposition here, but it seems that we do not have to pronounce this example with the intonation pattern associated with appositions. In this respect, example (346b) differs crucially from the (c)-examples in (346), which involve scrambling and topicalization of the direct object, respectively, and which are acceptable only with the intonation pattern associated with appositions.

Extraposition of non-restrictive relative clauses seems to lead to a slightly marked result when the antecedent is the complement of a preposition. This is illustrated in (347b) for a prepositional indirect object and in (348b) for a PP-complement of the verb. The (c)-examples show that topicalization of the PP makes the result unacceptable on the intended non-appositional reading. Note that, as in the case of extraposition of restrictive relative clauses, the markedness of the (b)-examples may be due to the fact that the (b)-examples compete with constructions in which the full PP is in extraposed position.

Section 17.3.2.3.2, sub II, has also shown that extraposition of restrictive relative clauses from nominal indirect objects is possible, provided that the indirect object is preceded by the direct object. The examples in (349) show that the same is true for non-restrictive relative clauses: the examples in (349c&d), in which the direct object precedes the indirect object as a result of scrambling and topicalization, respectively, are considerably better than example (349b), in which the direct object follows the indirect object.

The data in (349) suggests that a non-restrictive relative clause in extraposed position must be construed with the first noun phrase to its left. This is further supported by the contrast between the primeless and primed examples in (350). The unacceptability of (350a') can of course be accounted for by appealing to the freezing principle, since the word order in this example is derived by leftward movement of the prepositional indirect object. The unacceptability of (351b), on the other hand, cannot be accounted for in the same way, since we are dealing here with the underlying order of the two arguments. The fact that leftward movement of the prepositional indirect object in (350b'&b'') makes this example fully acceptable thus supports the claim that an extraposed non-restrictive relative clause must be construed with the first noun phrase to its left.

The conclusion that an extraposed non-restrictive relative clause must be construed with the first noun phrase to its left predicts that if a sentence contains two non-restrictive relative clauses modifying different DPs, only the second relative clause can be in extraposed position. The examples in (351) show that this prediction is correct: example (351a) simply gives the unmarked order without extraposition; the (b)-examples show that, as expected, extraposition of RC2 leads to an acceptable (though marked) result, whereas extraposition of RC1 leads to an unacceptable result; the (c)-examples show that extraposition of both relative clauses is impossible, regardless of their order.

If two non-restrictive relative clauses modify a single antecedent, extraposition is also impossible. Section 17.3.2.3.4, sub I, will show that stacking of non-restrictive relative clauses is severely restricted, but not impossible: if the two stacked relative clauses are introduced by different relative pronouns and if the relation between the two relative clauses is specified, as in (352a), the result may be more or less acceptable, although a structure in which the two relative clauses are coordinated is much preferred; cf. the perfectly acceptable Ik heb Els uitgenodigd, die hiernaast woont en met wie ik (bovendien) bevriend ben. The two (b)-examples show that extraposition of the relative clauses is categorically impossible.

Scrambling of the antecedent and relative clause together is possible, as shown by (353b) for direct object antecedents. Example (353c) shows that scrambling of the direct object cannot strand the non-restrictive relative clause, and (353d) illustrates that scrambling of the antecedent is also impossible when the relative clause is in extraposed position. The examples in (354) illustrate the same for a prepositional indirect object antecedent.

Topicalization of both the antecedent and the restrictive relative clause is possible. This is true regardless of the syntactic function of the antecedent. This is illustrated in the (b)-examples of (355) and (356) for antecedents functioning as direct object and indirect object, respectively. The (c)-examples show that topicalization cannot strand the relative clause in the original position of the object. The (d)-examples show that, unlike in constructions with restrictive relative clauses, splitting the antecedent and the relative clause by topicalizing the former and extraposing the latter is excluded: (355d) is acceptable, but only on a (restrictive) appositive reading (cf. Section 17.1.3), and (356d) is completely unacceptable.

Subsection IA has shown that non-restrictive relative clauses can be used to modify personal pronouns: when the antecedent and the relative pronoun have the same syntactic function (in the matrix clause and the relative clause, respectively), such constructions are perfectly acceptable; when the antecedent and the relative pronoun do not have the same syntactic function, the result is somewhat marked when either the relative pronoun or the antecedent functions as a subject.

In the case of topicalization, however, the results are somewhat different: the examples in (357) seem to indicate that topicalization of the object of the matrix clause is possible only in those cases where the antecedent and the relative pronoun have the same syntactic function. For instance, the examples in (357) in which the antecedent hemhim and the relative pronouns function as direct objects are perfectly acceptable.

If, on the other hand, the relative pronoun has the function of the subject of the relative clause, as in (358), the result is highly marked.

However, example (359a) shows that examples such as (358) improve when the topicalized object pronoun takes the subject form corresponding to the function of the relative pronoun in the relative clause. As example (359b) shows, this form (hijhe) is not acceptable when the direct object is in its regular position in the middle field of the clause.

Note also that this possibility of using the nominative form only arises with the direct object; when the antecedent functions as the indirect object of the matrix clause, the subject form can never be used, regardless of whether the object is in topicalized position or in its regular position in the middle field of the clause. This is shown by (360).

Finally, the examples in (361) show that similar problems do not arise in other cases where the antecedent and the relative pronoun have different syntactic functions. For example, in (361a) the antecedent has the function of indirect object while the relative pronoun functions as direct object, but topicalization is still possible. And in (361b) the antecedent functions as a direct object while the relative pronoun is part of the adverbial phrase and does not even function as an argument in the relative clause. From this we can conclude that topicalization is possible when the antecedent pronoun has the morphological form required by the syntactic function of the relative pronoun in the relative clause. In this respect, non-restrictive relative clauses behave like restrictive ones; cf. Section 17.3.2.3.2, sub IIE.