Restrictive relative clauses are infelicitous with proper nouns and uniquely referring noun phrases as antecedents; since there is no need for additional information to identify the intended referent, restrictive relative clauses are simply superfluous. In fact, the use of a restrictive relative clause in such situations will only lead to confusion, since it will suggest a referent set with more than one member. Consider the sentences in (270). Sentence (270a) is acceptable in any context because in the default case there is only one sun in the domain of discourse (domain D), so no further identifying information is needed for the hearer to pick out the intended referent. Adding a restrictive relative clause, as in (270b), has the effect of overriding the default value by suggesting that the set of suns in domain D has a cardinality greater than one.

Something similar holds for proper nouns: (271a) is acceptable in any context, since in the default case there is only one person with the given name in domain D: adding a restrictive relative clause usually leads to unacceptability, as shown by (271b).

The only context in which a proper noun can be followed by a restrictive relative clause is when the proper noun fails to uniquely identify the intended referent in the given context. In this case, the referent set denoted by the proper noun is indeed greater than one, which means that the function of the restrictive relative clause is to allow the hearer to select the intended referent. Thus, in the given situation where both father Joseph and son Isaac are well-known painters, sentence (272a) is perfectly acceptable. Note, however, that in such cases the proper noun no longer functions as a proper noun but as a common noun phrase, as indicated by the obligatory use of the definite determiner dethe. The acceptability of (272b), which also involves a proper noun modified by a restrictive relative clause, can be explained in a similar way. See also Sections 15.2.1 and 19.1.2.1 for a more detailed discussion of proper nouns.

Example (273b) shows that, in general, restrictive relative clauses are not easily used in noun phrases containing a demonstrative determiner. This is due to the fact that the demonstrative suggests that there are other (linguistic or extra-linguistic) means to uniquely identify the referent in question.

In contrastive contexts like those in (274), demonstratives can co-occur with restrictive relative clauses. This is to be expected, since in such contexts there are always two or more referents to be distinguished: the relative clause serves to distinguish the contrasted referents from all other elements, while the demonstrative serves to distinguish the contrasted elements from each other.

In non-contrastive contexts, the distal demonstratives dat/diethat/those can be used in combination with a restrictive relative clause provided that the demonstrative does not have its usual (uniquely) identifying function. In (275a), for example, the demonstrative dat is used to introduce an entity into the discourse (comparable to English this), while in (275b) the deictic force of the demonstrative die is not sufficient to identify the referent in question. In (275c) the stressed demonstrative is used to refer to an as yet unidentifiable generic referent set. In all these sentences it is possible to replace the demonstrative by the definite article, while omitting the relative clause leads to unacceptable results in the given contexts.

Noun phrases with a possessive pronoun or genitive possessor behave in much the same way as noun phrases with a demonstrative pronoun. In general, the use of a possessor is sufficient to pick out the intended referent, and as a result they typically do not co-occur with a restrictive relative clause. This is shown in example (276a). Again, exceptions must be made for cases where the possessor does not uniquely identify the intended referent in the given context. This is typically the case with family names in examples such as (276b); cf. also Section 19.2.2.2, sub I. Moreover, as with the demonstratives in (274), constructions with a possessor can be more or less felicitously restricted by a relative clause in a contrastive context such as (276c).

A personal pronoun can be modified by a restrictive relative clause only if the pronoun itself does not provide sufficient identifying or restrictive information in the given context; cf. also Subsection IIE. Consequently, restrictive relative clauses cannot be used to modify the singular first-person pronoun ik in (277), whose referent is contextually identified as the speaker.

Plural first-person pronouns can be relativized, but only in generic contexts. Thus, the pronoun wijwe in example (278a) has a generic reference: it denotes the whole class of Dutch people. In example (278b), on the other hand, wij refers to a contextually determined set of persons, and in this case the use of a restrictive relative clause is unacceptable. Note that the pronoun in (278a) has to be stressed, which is probably due to the fact that the referent of a reduced pronoun is always recoverable from the linguistic context.

Second-person pronouns can also be relativized only in exceptional cases, i.e. when the use of the pronoun alone does not sufficiently narrow the set of possible referents. In (279a), the pronoun is used to address a person standing among other people, and the restrictive relative clause is used to correctly identify the intended person as the person who is laughing louder than the others. In (279b), on the other hand, the pronoun refers to a uniquely identified hearer, and the addition of a restrictive relative clause is infelicitous. Note that the pronoun in (278a) must be stressed again.

Third-person pronouns accept relativization more freely, provided they are non-reduced. This is illustrated in example (280) for pronouns with (regular) specific reference.

The primeless examples of (281) show the same thing for the more special use of pronouns with non-specific singular reference. Since the pronouns in these examples lack a specific referent and have very little semantic content, the resulting constructions are very similar to the so-called semi-free relatives discussed in Section 17.3.2.2, sub IA3. Thus, the antecedent personal pronouns in these constructions can be replaced by the element degene(n)the one(s), although this results in the loss of the gender information expressed by the masculine and feminine pronouns in (281).

In the primeless examples of (282) the same is shown for pronouns with non-specific universal reference. In this case, the pronoun with universal reference can be replaced by quantifiers such as iedereeneveryone, in which case the universal reference of the antecedent is emphasized; in (282b') this requires the number specification of the verbs to be adapted to the singular feature of the quantifier iedereen.

The examples in (283) show that there is no restriction on the syntactic function of the relativized personal pronoun in the matrix clause: in these examples the antecedent pronoun functions as subject, direct object, and indirect object, respectively. Note that, like the subject pronoun, the object pronouns must be non-reduced. As will be clear from (283c), the syntactic function of the antecedent pronoun in the main clause need not correspond to that of the relative pronoun in the relative clause: the former functions as the indirect object of the main clause and the latter as the subject of the relative clause. Note that pronouns functioning as antecedents must have the form required by their syntactic function in the matrix clause; using the form required by a pronoun with the syntactic function of the relative pronoun, as in (283c'), leads to severe ungrammaticality.

We have repeatedly pointed out above that modification by a restrictive relative clause is only possible with the strong pronouns. Since the neuter singular third-person pronoun is normally pronounced in its reduced form ’tit, it will not be surprising that modification of this pronoun is not possible; as shown in example (284c), the demonstrative form datthat is used instead (with the pronoun wat as the relative element).

First, the examples in (292) show that extraposition from direct object DPs is possible. Of course, this does not mean that extraposition is always possible: for example, Subsection B will show that extraposition from the direct object requires that the relative clause contains salient/new information, and Subsection D will show that the antecedent is not scrambled, i.e. belongs to the focus (new information) of the clause.

The examples in (293) show that extraposition from a prepositional indirect object is also easily possible.

However, this does not hold for the nominal indirect objects in (294): the (a) and (b)-examples in (294) show that extraposition of the relative is possible, but only if the direct object is moved to a position preceding the indirect object. It seems that this fact has to do with the definiteness of the direct object, since example (294b) improves considerably if we replace the demonstrative with the indefinite noun phrase adviesadvice; this is shown in (294c).

Example (295) shows that the acceptability of extraposition from a direct object can also be influenced by the presence of material to the right of the direct object. The examples in (295) show that it is easier to extract a restrictive relative clause from a direct object in a double object construction when the direct object is preceded by a nominal indirect object than when it is followed by a prepositional indirect object.

Example (296a) shows that extraposition from PP-complements of a verb is perfectly acceptable, as is extraposition from prepositional indirect objects in (293). The same seems to hold for extraposition from the PP-complement of a noun or an adjective, although some people may consider such examples to be somewhat marked, which may have to do with the fact that the primeless examples compete with the primed examples in which the full PP-complement is in extraposed position.

The primeless examples in (297) show that extraposition of a relative clause from a subject also yields a perfectly acceptable result. Extraposition is prohibited, however, when the subject occupies the regular subject position right-adjacent to the complementizer (or finite verb), which is clear from the fact that the corresponding primed examples are degraded under neutral intonation.

The reason for this contrast is probably related to the information structure of the clause. Subsection B will show that extraposition of relative clauses requires the noun phrase to be focal, whereas subjects in the regular subject position are typically the topic of discourse; subjects that are part of the new information of the clause usually occupy a more rightward position in the clause, following clause adverbs like gewoonlijk in (297b). In this sense, the contrast concerning the subject in the primeless and primed examples is parallel to the contrast between non-scrambled and scrambled objects; cf. Section V13.2. In generative grammar, this parallel is accounted for by assuming that, like the scrambled position of an object, the regular subject position is a derived position; the base position of the subject is claimed to be lower in the structure as part of the lexical projection of the verb (VP), whereas the regular position is part of the functional structure of the clause (IP).

Extraposition from a PP-adjunct also seems acceptable, although the result is occasionally marked. Some examples are given in (298), which involve a comitative, a spatial, and a temporal adverbial phrase, respectively. Note that the markedness of the primeless examples may again have something to do with the fact that they compete with the primed examples in which the complete adverbial PP is in extraposed position.

The examples in (299) show that in some cases (e.g. when the relative clause is not too long) the relative clause can appear both adjacent to its antecedent and in extraposed position. The difference between the two orders can usually be explained in terms of end focus, i.e. by appealing to the general tendency to have the focal elements in sentence-final position: in (299a) the relative clause is interpreted as containing the new or focal information, and as such is given main emphasis; in (299b), on the other hand, the most neutral reading is one in which both the relative clause and the past participle ontmoetmet express new information, and both are given main emphasis.

The examples in (300) show that extraposition is strongly preferred when the relative clause carries contrastive focus: the order in (300b) is only acceptable with accent on the past participle, an option not available in the contrastive example given.

The preference for extraposition correlates with the length of the relative clause: the longer the relative clause, the greater the preference for its placement in sentence-final position. Thus, the order in (301a), with the lengthy relative clause in extraposed position, feels more natural than the order in (301b), with the relative clause adjacent to the antecedent.

Again, end focus is at work: the longer the relative clause, the more likely it is to contain new or otherwise salient information. This also means that length in itself is not always enough to make extraposition of the relative clause possible. There may be other weighty or salient information in the sentence with which the relative clause has to compete for the sentence-final position. An example is given in (302), where extraposition of the relative clause is prevented by a relatively weighty VP uitnodigen voor een kopje koffie.

Whether or not there is a preference for extraposition of the relative clause also depends on the definiteness of the antecedent: the (a)-examples in (303) show that with definite antecedents both orders are acceptable, whereas the (b)-examples show that with indefinite antecedents extraposition is clearly preferred. When the antecedent is an existential quantifier like iemandsomeone, as in the (c)-examples, the contrast seems even clearer. These observations are, of course, consistent with the assumption that extraposition requires the presence of focus, since indefinite DPs are more likely to contain focal/new information than definite ones.

A similar contrast can be found for subjects in expletive constructions: since the subject in these constructions receives a non-specific indefinite interpretation, we expect extraposition to be the preferred option. That this expectation is indeed borne out is illustrated in (304) by an intransitive and an unaccusative construction.

Differences in the acceptability of extraposition of relative clauses also occur in those cases where two restrictive relative clauses are in extraposed position, one extracted from a direct object and one from a subject. The result of such extraposition is clearly unacceptable in those cases where the relative clause extracted from the subject precedes the one extracted from the direct object, as in (305b'). When the relative clause extracted from the subject follows that extracted from the direct object, as in (305b), the result is slightly better, but still highly questionable; cf. Rochemont & Culicover (1997:280) for similar English data.

The contrast becomes clearer in (306) when the subject is given extra emphasis. Two cases can be distinguished. Example (306a) involves contrastive focus, implying that there are other students who read the article but are not interested in linguistics. Example (306b) involves restrictive focus, implying that there is only one student who is interested in linguistics, and that only this student has read the article. The order of the relative clauses in the primeless examples seems to produce an acceptable result (although the sentences may still be difficult to process), whereas the order of the relative clauses in the primed examples is completely excluded.

The examples in (310) show that extraposition of a restrictive relative clause from a direct object is possible, provided that the antecedent of the relative clause has not been scrambled or topicalized.

This is often described in terms of freezing: a phrase that has been moved (the scrambled/topicalized noun phrase) is an island for extraction (extraposition of the relative clause). However, appealing to the freezing principle is problematic because we will see in the next subsection that wh-movement of the antecedent is possible when the relative clause is in extraposed position. For this reason, it seems better to appeal to the information-structural effects of scrambling and topicalization. Since the kind of scrambling we are discussing here (cf. Section V13.2) is only possible with noun phrases that are part of the presupposition (discourse-given information) of the clause, we can rule out extraposition in (310b) by appealing to our earlier conclusion in Subsection B that the noun phrase must be sufficiently focal in order to license extraposition. Since topicalization generally involves discourse topics, we can give a similar account of the impossibility of (310c).

Note, however, that scrambled and topicalized phrases can sometimes be contrastively focused; although judgments about the precise status may differ among speakers, on this reading examples like (310b&c) become more acceptable. The relative acceptability of the examples in (311) provides direct support for the claim that an account in terms of the information-structural properties of the movements involved is superior to an account in terms of freezing.

That an account in terms of freezing is not adequate is also clear from the fact that it is easily possible to have simultaneous wh-movement of the antecedent and extraposition of the relative clause. For example, the discontinuous construction in (312a) is clearly preferable to that in (312b), where wh-movement applies to the DP as a whole. The data in (312) are consistent with the information-structural account, since wh-phrases are focal by definition.

However, not all questioned constructions allow extraposition of the relative clause. To allow this, the wh-phrase must not be D-linked, i.e. it must not presuppose that the set of questioned entities in domain D is non-empty. This requirement is indeed met in (312); this question does not presuppose that there are persons in domain D who play the piano well. A minimal contrast arises if we replace the wh-phrase in these examples by the partitive phrase hoeveel van de mensen [RC die echt goed piano spelen]how many of the people who play the piano well.

The examples in (313) presuppose that domain D contains a non-empty set of people who play the piano well, and as a result the split pattern is worse than the unsplit pattern. Wh-phrases headed by the interrogative determiner welke are also D-linked, and again the result of splitting the noun phrase is a degraded result.

The fact that wh-movement of a D-linked antecedent is not allowed is again in line with the general assumption that preposed antecedents must be focal: although wh-phrases are by definition focal, the D-linked ones presuppose knowledge of a particular set related to domain D and are as such less focal than their non-presupposed counterparts. Extraposition is also less acceptable in example (315), again due to the fact that the wh-phrase is D-linked in this case (which is clear from the use of the definite noun phrase het orkestthe orchestra). Note that (315b) becomes acceptable when the antecedent carries emphatic or contrastive focus, but in this case the construction is likely to be used as an echo question

Example (312b) shows that the split pattern is preferred when the interrogative noun phrase has the syntactic function of an object. The same holds for the subject of an expletive construction, although in this case the contrast between the split and unsplit patterns is less pronounced. This is illustrated in (316a&b) for intransitive and unaccusative constructions, respectively.

This subsection has shown that leftward movement of the antecedent is possible when the relative clause is in extraposed position, but that this option is restricted by the information-structural condition that the complex noun phrase is sufficiently focal; cf. Subsection B. The fact that wh-movement is possible when the restrictive relative clause is extraposed clearly shows that an analysis in terms of freezing cannot be sustained. This is in line with approaches to extraposition that reject the claim implied by the freezing account that scrambling and topicalization intrinsically precede extraposition. Such proposals reanalyze extraposition as stranding: the object is claimed to be base-generated in postverbal position, and what appears to be extraposition of the relative clause is actually stranding of the relative clause in the base position of the object; cf. Kayne (1994). Note that such an analysis would still allow us to appeal to the freezing principle to account for the unacceptability of leftward movement with stranding of the relative clause in preverbal position (cf. the discussion in Subsection C), since in that case the relative clause would be stranded in the derived preverbal position of the object. Another strand of research compatible with the findings in this subsection assumes that the postverbal phrase is not part of the preverbal phrase, but is base-generated as an independent phrase; cf. Koster (2000), De Vries (2002:§7/2011a) and the references cited there for interesting proposals of this kind.