Relative Clauses Research Paper

(1)       (a)       The woman [who we met yesterday] owns the local supermarket.

(b)       We saw a lizard [which had a bright blue tongue].

(c)       The developer sent an update to every user [who registered the program].

The phrase ‘limits the potential reference’ needs some further discussion. In (a), the subject of the sentence is a definite NP which is presumed to refer to a single individual, who is being said to own the supermarket. The RC provides a substantial portion of the in- formation needed to identify that individual (that we met this person yesterday), the ‘head’ noun woman providing the rest (that the person is an adult female). In (b) on the other hand there is no presumption of unique identification, but the content of the relative clause, ‘having a bright blue tongue,’ must be true of the object NP, so is limiting its potential reference. Example (c) is the most complex. Here, the NP that the RC is delimiting doesn’t refer to any specific individual, but we can still say that it is delimiting the reference of the NP, because the quantifier every indicates that if an entity satisfies the conditions indicated by the RC and the noun user (is a potential referent of the NP), then the developer sent that entity an update.

A problem for defining RCs is provided by so-called ‘Nonrestrictive relatives’ (NRCs) such as this: (2)                   Lois was introduced to Clark, [who said that he had read all her books].

Here the bracketed clause is not really limiting the reference of the NP of Clark, since this is a proper name and presumed to identify an individual on its own. In some languages, such as English, NRCs show significant differences from ordinary RCs, while in others, such as Japanese, they don’t (Kuno 1973, p. 233, Andrews 1985, pp. 28–9). It is not entirely clear whether NRCs should be regarded as a kind of RC or as a closely related construction.

1. Essential Structure

Since an RC delimits the reference of an NP by specifying the role taken by the referent of the NP in a situation, it needs to contain an internal position corresponding to the role of that NP, which can be called the NP_rel position. Many but not all con- temporary syntactic theories postulate that the NP_rel position is occupied by an NP, which may be unpronounced, or moved from the normal location for an NP with that semantic role. In the RC (1a), for example, the NP_rel would be described as having the direct object role, and identified as the word who at the beginning of the RC, preposed from its usual postverbal location.

From the viewpoint of formal semantics, one can propose that RCs have the logical form of λ-abstracts on a variable in the clause, occupying the NP_rel position. For example, the logical form of who we met yesterday would be along the lines of:

Such an expression is taken to denote the set of all entities such that if x refers to that entity, the sentence ‘we met x yesterday’ is true. If RCs are defined as clauses with this logical form, then NRCs can be treated comfortably as a type of RC.

2. Typology

The typology of RCs across languages can be organized in terms of two major parameters involving the position of the relative clause, and the treatment of position of NP_rel within it, a variety of minor parameters involving other aspects of the form of RCs, and the extremely important topic of Accessibility, which will be deferred to the section on theoretical significance below.

2.1 Major Parameters

The first major parameter is the relationship of the RC to the NP whose reference it delimits. There appear to be five possibilities, falling into two major groups:

(3)       (a)       The clause forms a constituent with the NP it modifies:

(i)        positioned after the head N of that NP,

(ii)       positioned before the head N of that NP, and

(iii)      appearing in an NP with no head N (alternatively formulated by saying that the head appears internally to the RC).

(b)         The clause doesn’t form a constituent with the modified NP, but is a separate constituent:

(i)        at the front of the clause containing that NP, and

(ii)       at the end of the clause containing that NP.

Types 3(a)(i) and 3(b)(i) can be illustrated clearly from English; all the examples above are 3(a)(i), while 3(b)(i) is illustrated by so-called ‘extraposed’ relatives like this:

(4)       A person was standing on the corner [who had a cockatoo on his shoulder].

Japanese provides perhaps the best-known examples of 3(a)(ii), which is widely found in verb-final languages, and sometimes in verb-medial languages such as Chinese

(5)       (a)       Japanese (Andrews 1985, p. 27):

Kore    wa       [ano     hito      ga        kai-ta]

this      _THEME that      person _NOW      write-_PAST

hon      desu.

book    is

This is a/the book which that person has written.

(b)         Chinese (Li and Thomson 1981, p. 116):

[Zhangsan       maı de] qıche hen guı.

Zhangsan buy NOM car very expensive

The car that Zhangsan bought was very expensive.

Example 3(c)(iii) is more exotic, but is commonly found in Native American languages. 6(a) is an example from Ancash Quechua (Cole 1987 p. 279), while 6(b) below is the corresponding 3(a)(ii) construction, also available in Ancash:

(6)       (a)       [Nuna bestya-ta         ranti-shqa-n]

man     horse-ACC     buy-PERFECT-3

alli                   bestya-m         ka-rqo-n.

good                horse-EVID    be-PAST-3

(b)       [Nuna ranti-shqa-n] bestya

man     buy-PERFECT-3 horse (NOM)

alli                   bestya-m         ka-rqo-n.

good                horse-EVID    be-PAST-3

The horse the man bought was a good horse.

In 6(a), bestya-ta ‘horse’ is playing the same basic function as the head N in an English RC, or in the 6(b) example, but both its position and its accusative case-marking are those appropriate to its function as NP_rel. Such constructions are often called ‘head internal’ RCs.

Finally, type 3(b)(i) is especially characteristic of (Indo Aryan) South Asian languages such as Hindi and Marathi (Srivastav 1991), although it is also found elsewhere, such as in Aboriginal Australia (Hale 1976):

(7)       (a)       [Jo laRkii khaRii hai] vo lambii hai.

REL girl standing is DEM tall is

The girl who is standing is tall.

(b)         [Nyuntulu-rlu kutja-npa wawirri

you-ERG RC-you(SUBJ) kangaroo

pantu-rnu], ngatyulu-rlu

spear-PAST, I-ERG

kapi-rna purra-mi.

FUT-I(SUBJ) cook-NONPAST

I will cook the kangaroo that you speared.

These structures are often called ‘correlative clauses,’ and one can question whether they should be seen as a type of RC or a closely related construction.

Srivastav, for example, argues that they are in fact quantifying expressions which include reference delimitation in their function, rather than mere reference delimiters.

The second major parameter of variation in RCs lies in the treatment of NP_rel. There are three important and partially independent parameters of variation:

(8)       (a)       Reduction of NP_rel:

(i)        no reduction (NP_rel remains a full NP),

(ii)       pronominalization (NP_rel is expressed as a pronoun), and

(iii)      omission (NP_rel is not expressed).

(b)       Marking of NP_rel:

(i)        no special marking (NP_rel looks like an ordinary NP of a form that also appears in main clauses), or

(ii)       a special ‘relative’ form is used for pronoun or determine of NP_rel.

(c)       Change of position of NP_rel:

(i)        remain in situ, or

(ii)       move (usually to the front).

In both the internally headed RC of 6(a) and the correlatives of (7), NP_rel is unreduced. In standard English, NP_rel is either deleted or expressed as a special pronoun, which is furthermore moved to the front, perhaps accompanied by some surrounding material

(9)       (a)       The people [we talked with e] were interesting.

(deletion from position marked with e)

(b)       The people [who(m) we talked with e] were interesting.

(movement of relative pronoun who(m) from e-position), and

(c)       the people [with whom we talked e] were interesting.

(movement of relative pronoun together with containing PP).

In modern Hebrew (Borer 1984, p. 220), NP_rel can be omitted or reduced to an ordinary pronoun, which can either remain in situ or move to the front:

(10)     (a)       Ra?iti ?et ha-yeled sherina ?ohevet (?oto).

saw-I ACC the-boy that Rina loves (him)

(b)         Ra?iti ?et ha-yeled (she-) ?oto rina

saw-I ACC the-boy that him Rina

?ohevet.

loves

I saw the boy that Rina loves.

Pronominal NP_rel that remain in situ are often called ‘resumptive pronouns.’

2.2 Minor Parameters

These are numerous; only two will be discussed here, clause introductory particles, and nominalization.

There may or may not be a clause introductory particle, which can appear instead of or in addition to a preposed NP_rel. In Modern Hebrew as illustrated in (10) above, the introductory particle is either she- (cliticized onto the next word) or ?asher, and it is optional if NP_rel is preposed. In modern English, there is an RC-introducing word that which is optional, but can’t co-occur with proposed NP_rel:

(11)     (a)       The person (that) I saw was tall.

(b)       The person who I saw was tall.

(c)       *The person who that I saw was tall.

(d)       *The person that who I saw was tall.

A somewhat more subtle form of variation is in clause-reduction. In the examples we have seen so far, the RCs are full-fledged finite clauses, but it is possible for them to become reduced. In Turkish, for example, when NP_rel is neither the subject nor contained in the subject (this is an approximation; see Underhill 1972 for further discussion), the subject of the relative clause is put into the genitive case, and the verb of the relative clause takes a nominalizing affix followed by cross-referencing as is normally applied to possessed nouns. The relative clause of 12(a) thus bears a considerable resemblance to the possessed noun phrase of 12(b):

(12)     (a)       [oglan-in git-tig-i] mektep

boy-GEN go-NOM-3SG school

The school that the boy goes to

(b)       Ahmed-in sapka-si

Ahmed-GEN hat-3SG

Ahmed’s hat.

Lehmann (1986) observes that such reduction is especially characteristic of structures of type 3(a)(ii) (RC preceding head with NP), and is apparently absent for correlatives (type 3(b)(i)).

Many further potential features of relative clauses are discussed in typological surveys such as Downing (1978), Keenan (1985), Andrews (1985), and Lehmann (1986), and in the papers in Peranteau et al. (1972).

3. Theoretical Significance

The theoretical significance of RCs has derived largely from the study of ‘accessibility,’ the possibilities for the role of ‘underlying position’ of NP_rel within the RC. Historically, this seems to have been divided into two largely independent investigations. Ross (1967) (following on from earlier work by Chomsky) studied systematically what might be called the ‘long distance’ aspect of the function of NP_rel, the capacity of NP_rel to be located in a subordinate clause structure within the relative clause itself. Later, Keenan and Comrie (1977) investigated the question of what kinds of functions NP_rel can bear in a monoclausal RC.

The long-distance aspect of accessibility revolves around the ability of NP_rel to occur, and be moved, specially marked, or deleted in a wide range of environments, including from arbitrarily deep subclauses of the RC (NP_rel position marked with e):

(13)     (a)       the book (that) Mary bought e,

(b)        the book (that) Mary thinks Jane discussed e,

(c)       the person (that who) Mary thinks Jane spoke with e,

(d)       the book that Mary thinks that Susan says that Sam borrowed e.

This carries implications about the kind of grammatical theory required to describe languages; simple phrase-structure rules in particular are clearly not adequate, some kind of augmentation being required (in early generative grammar, it was assumed that grammatical transformations were required; subsequently, a wide range of alternatives was found).

Most important is the fact that in spite of the long distance possibilities for NP_rel, its positioning is, nonetheless, limited by principles which tend to be similar across languages. In most languages that allow movement or deletion out of subordinate clauses, for example, this cannot happen from clauses inside NPs (the complex NP constraint):

(14)     *the book that Mary believes the claim that Jane discussed e.

These constraints have been especially important for linguistic theory because:

(15)     (a)       They apply to more than just relativization, but rather a wide range of apparent deletion and movement phenomena.

(b)       They appear in similar forms in many different languages (in both Irish and Hebrew, for example, the Complex NP constraint is observed when NP_rel is moved or deleted, but not when it is expressed as a resumptive pronoun).

(c)       It is not clear how they could be learned on the basis of direct evidence of the kind presumably available to children learning their native language.

They therefore raise fundamental questions about universal grammar and what it is that makes human languages learnable by children.

One of the more striking discoveries is that long-distance relationships appear to be composed of shorter ones; this was first proposed on theoretical grounds by Chomsky (1973), direct empirical evidence appearing in McCloskey (1977), where it was shown that in Modern Irish, clause introductory particles assumed a special form if they were in an RC, and contained that RC’s NPrel (McCloskey 1977, p.17):

(16)     (a)       Meas me gur thuig me an t-ursceal.

thought I that understood I the novel

I thought I understood the novel.

(b)       an t-ursceal a mheas me a

the novel REL thought I REL

thuig me

understood I

the novel I thought I understood.

In 16(a) we see the normal form gur of the particle introducing a complement of the verb think, but in 16(b), when the object of this complement is a (deleted) NP_rel, the form a is used instead. Similar ‘intermediate position’ effects were discovered in other languages and collected in Zaenen (1983).

The other aspect of accessibility is the function of NP_rel in monoclausal RCs. Keenan and Comrie (1977) proposed that there was an accessibility hierarchy:

(17)     Accessibility hierarchy

SU > DO > IO > OBL > GEN > OCOMP

such that any RC strategy would apply to NP_rel whose functions were a continuous segment of the hierarchy, and languages could always relativize subjects. In Modern Hebrew, for example, SU (subject) and DO (direct object) can be relativized by the deletion strategy, while pronouns (in situ or moved) can only be used when NP_rel is DO or below.

The validity of the hierarchy corroborates the universal significance of the grammatical relation concepts used in formulating it, and raises significant issues about their nature. For example, Fox (1987) observed that, in spoken discourse, it was relatively rare for transitive subjects to be NP_rel (10 percent), most NP_rel being either objects (45 percent) or intransitive subjects (45 percent). The high frequency of relativing objects seems discrepant with a tendency for relativization to be restricted to subjects.

A solution is afforded by a reinterpretation of grammatical relations, in which the traditional ‘subject’ concept is split into one concept, associated with the semantic role of ‘Agent,’ and another with the semantic role of ‘Topic.’ This idea goes back to early work by Schachter (1997), in which relative clauses in Tagalog constitute some of the evidence, and has been incorporated into many contemporary syntactic frameworks (see Manning 1996, Van Valin and LaPolla 1997 for recent discussion). Fox’s observations then indicate that the grammatical relation at the top of the accessibility hierarchy should not be identified with the traditional concept of subject, but with the topic-oriented subject of this more recent work.

Bibliography:

1. Andrews A D 1985 Studies in the Syntax of Relative and Comparative Clauses. Garland Press, New York
2. Borer H 1984 Restrictive relatives in Modern Hebrew. Natural Language and Linguistic Theory 2: 219–60
3. Chomsky N 1973 Conditions on transformations. In: Anderson S R, Kiparsky P (eds.) A Festschrift for Morris Halle. Holt, Rinehart and Winston, New York, pp. 232–86
4. Cole P 1987 The structure of internally headed relative clauses. Natural Language and Linguistic Theory 277: 302
5. Downing B T 1978 Some universals of relative clause structure. In: Greenberg J H (ed.) Universals of Human Language. Stanford University Press, Stanford, CA, Vol. 4, pp. 357–418
6. Fox B A 1987 The noun phrase accessibility hierarchy reinterpreted: Subject primary or the absolutive hypothesis. Learning 63: 856–70
7. Hale K L 1976 The adjusted relative clause in Australia. In: Dixon R M W (ed.) Grammatical Categories in Australian Languages. Australian Institute of Aboriginal Studies, Canberra, pp. 78–105
8. Keenan E L 1985 Relative clauses. In: Shopen T (ed.) Language Typology and Syntactic Description, Vol. II: Complex Constructions. Cambridge University Press, Cambridge, UK, pp. 141–70
9. Keenan E L, Comrie B 1977 Noun phrase accessibility and universal grammar. Linguistic Inquiry 8: 63–99
10. Kuno S 1973 The Structure of the Japanese Language. MIT Press, Cambridge, MA
11. Lehmann C 1986 On the typology of relative clauses. Linguistics 24: 663–81
12. Li C N, Thompson S A 1981 Mandarin Chinese: A Functional Reference Grammar. University of California Press, Berkeley, CA
13. Manning C D 1996 Ergativity: Argument Structure and Grammatical Relations. CSLI Publications, Stanford, CA
14. McCloskey J 1977 Transformational Syntax and Model Theoretic Semantics. Reidel, Dordrecht
15. Peranteau P M, Levi J N, Phares G C (eds.) 1972 The Chicago Witch Hunt: Papers from the Relative Clause Festival. Chicago Linguistic Society, Chicago
16. Ross J R 1967 Constraints on Variables in Syntax. Ph.D. thesis, MIT
17. Schachter P 1977 The subject in Philippine languages: Topic, actor, actor-topic, or none of the above. In: Li C N (ed.) Subject and Topic. Academic Press, New York, pp. 491–518
18. Srivastav V 1991 The syntax and semantics of correlatives. Natural Language and Linguistic Theory 9: 637–86
19. Underhill R 1972 Turkish particles. Linguistic Inquiry 2: 87–100
20. Van Valin R D, LaPolla R J 1997 Syntax: Structure, Meaning and Function. Cambridge University Press, Cambridge, UK
21. Zaenen A 1983 On syntactic binding. Linguistic Inquiry 14: 469–504