Restrictive modifiers restrict the set of referents of the noun phrase, whereas non-restrictive modifiers do not. This means in the case of definite noun phrases that the restrictive modifier is needed to enable the listener to pick out the intended (possibly singleton) set of referents, whereas the non-restrictive modifier simply provides additional information about the intended referent of the noun phrase. Consider the examples in (11).

In example (11a), the cat referred to by the noun phrase is assumed to be identifiable to the hearer in the given context, and the information provided by the PP naast me op het bedbeside me on the bed simply provides additional information about the location of the cat in question; if the modifier were omitted, the sentence would be less informative, but still perfectly acceptable. In example (11b), on the other hand, the PP-modifier provides information that the hearer needs in order to correctly identify the intended referent of the noun phrase; there are several entities in the domain of discourse (domain D) that are part of the denotation of the head noun katcat, and the restrictive modifier provides the additional information that the hearer needs in order to pick out the intended referent.

Restrictive modifiers can thus be said to be necessary for successful communication: if domain D contains several cats and the speaker would say de kat ligt te spinnenthe cat is purring, the hearer will not be able to identify the intended referent of the noun phrase and will most certainly ask for additional information. Not surprisingly, then, a commonly used test for distinguishing between restrictive and non-restrictive modifiers is whether the modifier can be omitted. This test is pragmatic rather than syntactic, since the result is, strictly speaking, always grammatical, so that the acceptability of the resulting construction should be judged in terms of its felicitousness in the given context: omitting a non-restrictive modifier merely results in a less informative but unambiguously interpretable sentence, whereas omitting a restrictive modifier results in either an insufficiently informative and therefore infelicitous sentence, or an incorrect/unintended (overgeneralized) statement.

That restrictive modifiers are used to enable the hearer to pick out the intended referent(s), while non-restrictive modifiers provide additional information about the intended referent, is also clear from the fact that uniquely referring noun phrases cannot be modified by the latter. This can be illustrated by considering examples like (12), where the relative clause modifies a proper noun. Example (12a) is acceptable because the non-restrictive relative clause simply provides information about the mother of the intended referent. Example (12b), on the other hand, would only be acceptable in the exceptional case that the hearer knows that there is more than one person named Jan de Jong, only one of whom happens to have an Argentine mother. And actually, even under this interpretation, the example is marginal, since the proper noun would then preferably be treated as a count noun and preceded by the definite article: De Jan de Jong die een Argentijnse moeder heeft, spreekt vloeiend Spaans (see also Section 15.2.1, sub II).

Something similar holds for the examples in (13), where the referent of datzelfde boekthat same book can be assumed to be known to the addressee without the information provided in the relative clause: using a non-restrictive relative clause, as in (13a), is possible because it simply provides some additional information about the book in question; adding a restrictive relative clause, as in (13b), is impossible because it suggests that one and the same book can have different contents.

The examples in (12) and (13) show that restrictive modifiers cannot be used if the referent of the definite noun phrase is already uniquely determined without them. The opposite restriction seems to hold for the non-restrictive modifier: the referent of the noun phrase must be uniquely determined in domain D in order for a non-restrictive modifier to be licensed. This is illustrated by (14), where the non-restrictive relative clauses can be used, but only if the referents of the noun phrases have been previously introduced into the discourse. If this is not the case, the sentences must be pronounced/have the punctuation associated with a restrictive relative clause.

Finally, we want to emphasize that although non-restrictive modifiers do not play a role in determining the proper referent set of the noun phrase, they do play an important role in discourse. For example, the information given in the relative clauses in (14) can be construed as the motivation or reason for the proposition expressed by the main clause: (14a) suggests that the speaker bought the car on credit because it was very expensive, and (14b) suggests that the fact that the student spoke fluent Spanish is due to the fact that he has an Argentine mother. In some cases, omitting a non-restrictive modifier can even lead to pragmatically infelicitous sentences. This may be the case in sentences that contain an element that can only be properly interpreted on the basis of the information provided by the non-restrictive modifier. Consider example (15a): if the information provided by the non-restrictive modifier net uit het ziekenhuisjust out of the hospital is new to the hearer, the adverbial phrase natuurlijkof course in example (15a) cannot be properly understood without it: the sentence is not unacceptable without the modifier, but it is still infelicitous in the given context. The same holds for (15b), where the correct interpretation of the element tochstill requires knowledge of the fact that my friend has six children: “Despite the fact that my friend has six children, she still has time for a job.”

When the noun phrase is indefinite with a specific referent, the function of the non-restrictive modifier is again to provide additional information about the referent of the noun phrase, the main difference with definite noun phrases being that this referent is not assumed to be identifiable by the hearer. The function of the restrictive relative clause, on the other hand, changes: although its main function is still to restrict the referent set of the noun phrase, it no longer serves to enable the hearer to uniquely identify this referent (set). Consider the indefinite version of the examples in (12), given in (16).

The indefiniteness of the noun phrase suggests that the hearer is not assumed to be able to pick out the intended referent: the speaker may be introducing a new entity into domain D or there may be several cars in this domain. The difference between the restrictive and non-restrictive sentences is that the former provide additional information, whereas the latter restrict the set of possible referents. Since identifiability is not at stake here, omitting the relative clause leads to a felicitous result in both cases. However, omitting the restrictive relative clause changes the presuppositions of the main clause, whereas this is not the case in constructions with non-restrictive relative clauses. Thus, in (16a), the core message conveyed is that the speaker bought only one car, and in addition it is said about this one car that it is an expensive car; leaving out the non-restrictive relative clause does not affect the core message. In (16b), on the other hand, the use of the restrictive relative clause suggests that the speaker bought more than one car, and that only for the expensive one payment was deferred. This suggestion that the speaker bought several cars is lost if the restrictive relative clause is omitted.

The difference in function between non-restrictive and restrictive modification is particularly clear in constructions with generic noun phrases introduced by an indefinite article (eena in the singular and the empty form Ø in the plural), since these can only be modified by restrictive modifiers. Example (17a) gives constructions with relative clauses, example (17b) constructions with prepositional postmodifiers, and example (17c) constructions with adjectival postmodifiers.

The restrictive modifiers in the primeless examples restrict the referent set of the noun phrase. This means that in all these sentences, the noun phrase without the modifier refers to a larger set of entities than the modified noun phrase: in (17a), it is predicated only of a subset of students that they will not get a pass; in (17b) the set of all cities is restricted to those with more than one million inhabitants and it is to this subset that the predication applies; example (17c) asserts that it is only children over the age of five who have to pay.

One way to account for the unacceptability of the primed examples in (17) is to appeal to our knowledge of the world by saying that since the modifier does not restrict the referent set, it is taken to provide additional information about the full sets denoted by the nouns. This would mean that these sentences express two propositions, both of which are said to be true of the full denotations of the nouns. For instance, example (17a') asserts that everyone who is a student will fail the exam, while at the same time all students are said to be lazy. In (17b'), both propositions expressed are clearly false: not all cities have more than one million inhabitants, and not all cities are metropolises. Similarly, the implication in (17c') that all children are older than five is false. However, it is not entirely clear whether appealing to our knowledge of the world is sufficient to account for the unacceptability of the primed examples in (17), since this incorrectly predicts that examples in which both propositions are true should be perfectly acceptable: this is the case in (18a&b), which are nonetheless dubious. For completeness’ sake, note that the (textbook) example in (18c), in which a definite noun phrase refers to the species/family of whales, is perfectly acceptable.