The following text is a modified reproduction of:

Lehmann, Christian 2003, “Relative clauses”. Frawley, William J. (ed.), International encyclopedia of linguistics. Oxford: Oxford University Press; 3:460-462.
https://www.researchgate.net/publication/297918505_Relative_clauses

Basic concepts

A relative construction involves a superordinate proposition p, a subordinate proposition q and a term playing a role in both propositions, called the nucleus. q is understood as modifying the nucleus (in order either to narrow down its concept or to pin down its referent); p is understood as applying to the nucleus as modified by q. A syntactic construction coding such a constellation is a relative construction to the extent that the operations involved – subordinating q, forming it into a modifier of the term and signalling the role of the nucleus in q – are part of its grammatical structure. If so, then the nucleus appears as the head nominal (or ‘domain nominal’) and the subordinate clause as the relative clause.

If the nucleus is represented by a lexical nominal, the relative clause may be its attribute. It may also be null, in which case the relative clause is a free relative clause.

If the relative clause has an external head or is headless, it is oriented towards one of its syntactic or semantic positions. This position may or may not be occupied by some pronominal element. The semantic counterpart of this representative of the nucleus inside the relative clause is neither a concept nor a referent (except for non-restrictive constructions). It is best thought of as a variable as it is used in predicate calculus.

In , girl is the head noun, the bracketed constituent is a relative clause, whose represents the nucleus in the relativized position, and girl whose name I can't remember is a relative construction.

As indicated, a relative construction is based on three grammatical operations:

• Subordination-nominalization subordinates the relative clause and desententializes it to some degree.
• Attribution/head formation combines head and relative clause into one notion.
• Anaphora-empty-place formation forms an empty place in the relative clause which refers to the head.

These operations take different shapes in the different types of relative construction. Their semantic effect may also be achieved without grammatical means; then no relative construction results.

Subordination-nominalization

Subordination of the relative clause may be signalled, inter alia, by a conjunction (E3, ), by a relative pronoun (E1 with which, E2, ), by a subordinating affix on its verb (, ), or merely by its embedded position (E1 without which).

A major subdivision of relative clauses is based on the syntactic relation of the relative clause to the main clause. If the relative construction forms a nominal in the main clause, it is embedded. In the embedded examples above, the relative construction is the subject of the main clause. If the relative clause does not form a nominal with its head, it is adjoined to the main clause, as in E3. In .a, the relative clause is preposed; in b, it is postposed. The most common variant of the adjoined relative construction is the correlative construction (). Here the relative clause contains a relative pronoun and the main clause a correlative demonstrative pronoun. As E3, the correlative construction is generally invertible.

In relative clauses as elsewhere, subordination may involve varying degrees of desententialization down to full nominalization. Typically, the adjoined relative clause is syntactically most like an independent clause. Since embedded relative clauses function as adjectivals or nominals, they generally show restrictions on tense/aspect/mood, genitive case on the logical subject and, ultimately, non-finiteness of the verb (relative participle).

Attribution/head formation

Attribution modifies a nominal head by a dependent expression, the attribute. In relative constructions, the attribute is a clause. Otherwise, it may be an adjective (heavy book) or a dependent nominal (book on the table). relative clauses allow greater freedom in identifying entities and forming complex concepts than other attributes.

Attribution presupposes that the head and the modifier are syntagmatically separate (E1). In E2 and .a, the subordinate clause is a relative clause, but it is not an attribute to its head. As an alternative to attribution, we here have head formation. Accordingly, relative clauses are subdivided into external head and internal head relative clauses. Internal head relative clauses predominate in languages such as Hittite, Bambara, the Indic and Yuman languages. Preposed relative clauses and indifferent (indefinite, generalized) relative clauses (E2) are generally of the internal head variety.

The most common way of signalling head-formation is by a relative pronoun (E2, ). Attribution may also be signalled by a relative pronoun (E1 with which), by attributors such as the indefinite suffix in or even by the agreement of the relative clause with the head. Often it is only conveyed by constituent order. There are three possibilities of sequencing an embedded relative clause as against its head: it may be prenominal (), postnominal (E1), or circumnominal (). A circumnominal relative clause is thus an embedded internal headed one. While circum- and postnominal relative clauses are generally only partially desententialized, the prenominal relative clause is at the end of this scale; exhibits all the nominalization phenomena mentioned before.

Anaphora-empty-place formation

The role that the head plays inside an external head relative clause is signalled by diverse strategies. The most common one is to leave an empty place in the relativized position (E1 without which, .b, ). Also common is the use of an anaphoric pronoun to represent the head (). It presupposes at least a moderate degree of sententiality, since strongly nominalized relative clauses, in particular prenominal ones, are like simple attributes in not bearing a normal anaphoric relation to their head.

A relative pronoun is used to represent the head in several, primarily Indo-European, languages (E1 with which). The prototypical relative pronoun thus functions in all of the three operations constitutive of a relative construction. An example of a less prototypical relative pronoun is the Arabic one, which only signals subordination and attribution. The relative pronoun is, thus, not a defining feature of relative clauses, but one of a set of alternative devices for forming them.

In a non-sentential attribute, the head plays the role of subject (or absolutive) (cf. heavy book with book which is heavy). This is also the most frequent function of the head in the relative clause. A complex hierarchy of syntactic functions codetermines which syntactic functions are relativizable. Its primary subhierarchy is as follows:

1. subject / absolutive
2. direct object / ergative
3. indirect object and other valency-governed arguments
4. adjuncts

Secondary functions such as the genitive attribute and mediated functions such as the function of the relativized position in are on different subhierarchies which are at lower levels than the adverbal one. In general, if a relative clause forming strategy in a language can relativize a given position on this hierarchy, it can relativize all positions higher up. Basque, e.g., can relativize only down to the indirect object, Quechua can relativize only valency-governed arguments, and English includes adjuncts. Very few strategies allow constructions such as . The achievement of an relative clause forming strategy on this hierarchy generally diminishes with desententialization.