English Restrictive Relative Clauses English Language Essay

Published: Last Edited:

This essay has been submitted by a student. This is not an example of the work written by our professional essay writers.

This chapter reviews the studies on English restrictive relative clauses from four different perspectives, namely pragmatics, syntax, semantics, and discourse. The review includes the grounding function of English restrictive relative clauses, information status of NPs, the syntactic-semantic aspects of relative clauses and discourse structure.

2.1 Grounding and Restrictive Relative Clause

English relative clauses are categorized into two sub-categories: Restrictive relative clause and non-restrictive relative clause. Look at the following examples:

The boy who works at the Chinese restaurant studies hard.

The boy, who works at the Chinese restaurant, studies hard. (Stockwell et al. 1973, Comrie, 1981)

The first sentence is restrictive relative clause; while the second one is on-restrictive relative clause. in the first sentence, the function of 'who works at the restaurant' is to identify the head noun 'the boy' which identities one boy among others. The relative clause, inserted into this sentence, helps make the referent clear and provide more specific meaning to the matrix sentence: The boy studies hard. The relative clause that behaves in this way, to identify or to restrict the meaning of the head noun, is known as a restrictive relative clause. (Master, 1996) In this paper, the author focus only on restrictive relative clause.

Scholars such as Schachter (1971), Bernardo (1979), Tabasowka (1980), Beaman (1984), Hann(1987), Fox (1987), Fox and Thompson (1990), and Nugsamee (2003) present full discussion on the functions of English restrictive relative clauses, in the following discussion, restrictive relative clauses will be shortened as RRC. It is showed in their founding that there are two main functions of RRCs: grounding and description. RRCs serive grounding function in the following circumstance: If the information coded in a RRC has already been known to the hearer/reader, the RRC is used to ground the referent and help the listener/reader identify the referent mentioned in the previous discourse. Those researchers propose that RRCs are frequently used to help hearers identify the mentioned referents in conversation, as in "The man who I met yesterday is my boss." However, many RRCs are also employed for making the new referents a relevant part in the discourse, as shown by the example "A guy who I met in the street yesterday said he knew your sister." In this example, the RRC is used not for the identification of old referents, but for making the new referent A guy a contextual relevance through the given entity I in the RRC. Thus, in this study, the term 'grounding' is used to refer to RRCs used both for the identification of old referents and for making new referents relevant to the discourse. The concept 'grounding' will be made clear in the subsequent parts of this study. Leech and Svartvik also state this function in their definition of RRC. They define the English Relative Clause as: (a study of discourse function of relative clause from a functional semantic sentence framework)

"Various types of sub-clauses are linked to part of all of the main clause by back-pointing elements usually relative pronoun (wh-pronouns: who, whom, whose, which, that, and zero). The principal function of a relative clause is that of postmodification in a noun phrase, where the relative pronoun points back to the head of the noun phrase (the antecedent)". (P285)

The other function of RRCs is that they may make assertion about their head NPs and provide new information for the newly-introduced head NP. RRCs of this kind perform description functions.

In addition, in Bernardo (1979) and Beanan (1984)'s studies, they classify RRc into two categories: information-bearing and identificatory. In Beanan's terminology, information-bearing refers to a relative clause that carries new information, while identificatory refers to a relative clause that carries old/given/known information. These two categories, however, can only be understood within the context of the entire story. It would be helpful to look at some examples from Beanan: (1984, P73)


(1) There was a man (who was picking pears).

(2) The three boys pass the man (who had just come down from the ladder).

In the above sentences, the parts in brackets are mentioned for the first time in the context.


(3) And then the man (who was picking pears) comes down from the ladder.

(4) The man (who had been picking the pears) is just getting down from the ladder.

In the above sentences, the parts in brackets have been previously mentioned in the context.

The first two examples can be considered as information-bearing because they provide new information to the hearer about the "the man". The second pair of examples is identificatory relatives because they single out "the man who was picking pears" from other members of the category "man" present in the previous context. As already stated, this distinction can only be understood within the context of the entire story.

In Fox and Thompson (1990) further distinguishing the types of grounding, they propose three central kinds of grounding to elaborate on how relative clauses make their head NPs a relevant part in the whole discourse. The first is what Prince(1981:236) has called anchoring: "a discourse entity is anchored if the NP representing it is linked, by means of another NP, or "Anchor", properly contained in it, to some other discourse entity". She gives an example to illustrate this type of anchoring.

(5) A guy I work with says he knows your sister. (Prince, 1981:233)

In (5), I, a given referent in the speech context, in the relative clause serves as an anchor to link the new discourse entity a guy. It thus makes the head NP relevant to the whole discourse.

The second kind of grounding is proposition-linking, in this situation a new referent is linked under the frame evoked in the previous discourse. Consider text (6).

(6) The mother's sister is a real bigot. Y'know and she hates anyone who isn't a Catholic. (Fox and Thompson, 1990:301)

In this example, the entire NP anyone who isn't a Catholic is grounded by its association to the previous proposition characterizing 'mother's sister as a bigot'. Therefore, the information stated in the RRC serves proposition-linking grounding function.

The third kind of grounding is done by main clauses instead of relative clauses. According to Fox and Thompson, relative clauses of this kind provide no grounding; that is, it does not relate a head NP to any given referent in the relative clause. Instead, the new referent is grounded by the given entity in the main clause instead of being made relevant by a given NP in the relative clause. In such a case, the RRC is used to provide new information for the new head NP, as shown in (7).

He's got a spring that comes, way up. (Fox and Thompson, 1990:301)

In example (7), the new referent a spring is grounded by the subject of the main clause he instead of the given referent in the RRC. The RRC in this example is not used to ground its head NP, but simply used to provide new information for its head NP.

Fox (1987) further proposes the other discourse function of relative clauses and states that relative clauses are structured to provide information for their firstly-introduced head NPs, thereby providing description function. This is accomplished by providing a stative description of some aspect of the new referents, as shown in the following example.

(8) I did not notice it but there is a woman in my class who is a nurse and, she said to me she said that did you notice he has a handicap and I said what. You know I said I do not see anything wrong with him, she says his hand. (Fox, 1987:861)

Fox states that the woman's 'nurseness' is critical to her relevance to the following story. It is through her training as a nurse that she is presumably more attuned to physical handicaps-an attunement which the teller describes himself as lacking. Thus, the RRC here plays a crucial role in introducing a new referent and making it relevant to the ongoing discourse. Thus, from this example, it is apparent that the RRC is employed to provide description for the head NP, paving the way for the head NP and making it a contextual part in the ongoing discourse.

From the perspectives of Givon (1993,1995), Fox (1987), and Fox and

Thompson (1990), a RRC may be used for any of two reasons: grounding and description. When a RRC serves the function of linking the current referent to the preceding utterance, it does a grounding task. In such a case, it grounds its head NP, as in (2)-(5). When the information coded in a RRC is associated with the prior proposition frame, the RRC does a proposition-linking grounding work, as shown in (6). Furthermore, when a RRC is not used to ground a new discourse entity, the purpose of it is to provide new information for its newly-introduced head NP. RRCs of this kind provide description function, as in (7)-(8).

2.2 Information Status of NPs

Several scholars present their views stating distinction of given and new information.

Halliday adopts the Prague school view of information as comprising two categories: new information, which is information that the addresser believes is not known to the addressee, and given information, which is information that the addresser believes is known to the addressee.

In his discussion of information structure, Halliday (1967) is particularly concerned with spoken English. He claims that besides a thematization analysis, a sentence may also be divided into information units, some of which would be focal. On the other hand, the focal unit carries new information, and the non-focal units convey information that is recoverable in the preceding discourse (given).

Compared to Halliday, Chafe (1974, 1976) has provided a slightly different interpretation of the given/new information definition. He claims that given information can be referred to as information that the speakers assume to be in the hearer's consciousness at the time of the utterance. New information, in contrast, is thought of as information that is believed by the speaker not to be in the hearer's consciousness, either because the hearer is no longer aware of it, or because it has not been introduced into awareness.

In his book "Discourse, Consciousness, and Time", Chafe (1994) elaborates his ideas on given/new distinction by describing various states of consciousness that affect the way in which people produce speech, which he calls "intonation units". Consciousness, according to Chafe, is analogous to vision, which can be classified into two groups: foveal vision, the object of people's focus, and peripheral vision, anything else on the scene, which people are aware of, but not necessarily focusing on.

Similar to the vision metaphor, Chafe proposes that human beings also have both focal and peripheral consciousness, and that our mind is in a constant state of change as we move from one focus to another. He identifies the state of change as consisting of three information types in the mind: active, semiactive and inactive. While active information is information that is in the focus of people's attention (given), semiactive information is something that is known before, but not in the current focus of attention (accessible). Finally, inactive information is information that is not stored in people's unconscious mind (new).

Another well-known perspective towards the notion of given/new information was developed in 1977 by Clark and Haviland in "Comprehension and the Given-New Contract". They claim that "the processing demands entailed by sentence information structure are formulated in a given-new strategy". When speakers and hearers are engaged in discourse, they follow certain cooperative communicative principles: to be informative, relevant, and unambiguous. In doing so, the speaker provides background (given) information which he assumes is known to the hearer. To this base, he adds new information which is presumably unknown to the hearer.

According to the Clark and Haviland's Given-New Strategy, the process of comprehending a sentence in discourse context comprises three sub-processes or stages: (1) identifying the given and new information in the current sentence, (2) finding an antecedent in memory for that given information, and (3) attaching the new information to this spot in memory. The primary usefulness of this strategy has focused on various possibilities that can happen during stage 2. For instance, sentences construed as given but have no obvious antecedent from previous sentences should impose difficulties in comprehension.

Vande Kopple also highlights an importance of discourse context as a major criterion for given/new information definition. According to Vande Kopple, given/new information can be defined as follows:

Given information centers on elements that have been mentioned prior to a particular point in a text or that are recoverable from the text or the extralinguistic situation. New information includes the elements not meeting these criteria. (P53)

(Chafe, 1974, 1976, 1987, 1994). Chafe (1976:30) defines given information as "knowledge which the speaker assumes to be in the consciousness of the addressee at the time of utterance", and new information as "what the speaker assumes he is introducing into the addressee's consciousness." His work (1994) further recognizes a three-way breakdown into given, accessible, and new information in place of the simple binary distinction of given and new information. Chafe proposes that given information can be characterized as already active at the time of utterance, accessible information as semiactive at the point of utterance and new information as inactive at the time of utterance. The distinction between given and accessible information can be illustrated in the following examples.

(9) A: We got some beer out of the trunk.

B: The beer was warm. (Chafe, 1994:170)

(10) A: We checked the picnic supplies.

B: The beer was warm. (Chafe, 1994:170)

The beer in B's response of (9) has been treated as given information since it has been established as a shared referent with A's utterance. The beer in (10) is accessible because of the knowledge that the picnic supplies are likely to contain beer.

Another scholar does major contribution to the information status studies is Prince. She categorizes entity into given ad new. When referring to new, she introduces a new concept "anchored". A new entity might be newly introduced and has no connection to the context; however, the real case is that the new entity certainly has some kind of connection with some already existent entities in context. She considers that this new entity is anchored, meaning that its relative position in the context has already been settled. Anaphora can introduce a new concept into a discourse; also it can anchor this entity in many ways. Anchoring means that the newly introduced entity builds some connection with the preexistent entities in the discourse dynamic context.

e.g: Adolf Hitler, who was the leader of the largest party in Germany- National Socialists, was demanding for himself the chancellorship of the democratic Republic he had sworn to destroy.

As a new concept in this text, Adolf Hitler is anchored by the descriptive anaphora "the leader of the largest party in Germany". This anchoring enables readers to have a clear knowledge of Hitler's social position at that time. New information includes introduction of new entity or new characteristics as well as establishment of a new connection.

It is very important to anchor the new entity and it is also the essential demand of coherence of discourse. The information value of new entity lies in the connection built by new entities with existent entities. The additional descriptive modifiers are not helping to access to the existent entities, as a matter of fact, there is no accessing issue, because the related entity is not given, but rather new, the information provided by the related modifier is new as well, the only function here is to anchor.

Hwang's (1994) study also revealed a tendency of new information in her relative clause. In "Relative Clauses, Adverbial Clauses and Information Flow in Discourse", Hwang investigates discourse-pragmatic functions of relative clauses and adverbial clauses in association with information flow in English narrative stories. Based on data obtained from this analysis, Hwang found that perhaps the most common functions of relative clauses in any language is "providing background information that is often new but descriptive, rather than advancing the event line of a story". (P680) Consider the example below:

(11) Once upon a time there was a little girl ½›who lived alone with her father½.

Information in the relative clause in (11) gives background information about "a little girl". It is new and descriptive, and introduces additional characters "her father", who is related to the character introduced by the head noun. This function of introducing related participants, according to Hwang, is found more in her English relative clauses.

2.3 Theme and Rheme of Relative Clauses in English

<a study of discourse function of relative clause from a functional sentence perspective framework>

Mathesius was among the first to introduce this distinction. According to his classification, the theme can be referred to as the segment that the speaker is talking about in the sentence, whereas the rheme is what the speaker says about the theme. In other words, the theme mostly corresponds to the first constituent or the grammatical subject in a clause, and the rheme often corresponds to the grammatical predicate. The third concept of FSP, namely, topic/comment or given/new is similar to the second and it was also introduced by Mathesius. His distinction is based on the analysis of sentences into two major elements: the given and new or the topic and the comment. Generally speaking, the topic is perceived as a segment in a sentence that conveys given, known or old information, "information that is expressed in, recoverable from or is relatively more accessible in prior sentences of the text". On the other hand, the comment is considered as a segment that carries new information, information that is now expressed in, is not recoverable from, or is relatively less accessible in prior sentences.

Kuno (1976) is another linguist who take into account theme and rheme

in the study of English relative clause. In his famous article, "Subject, Theme, and the Speaker's Empathy"-A reexamination of Relativization Phenomena, Kuno claims that only NPs which are considered as the theme of a relative clause can be relativized. Following are some of his examples:

(12) Here is the singer that Susan painted a picture of.

(13) Here is the door that Smith know the people at

(14) Here is Smith, who know the people at the front door.

In sentences (12) and (13) the relative pronoun is an oblique object (object of preposition), yet (12) is acceptable whereas (13) is not. Kuno states that (12) is acceptable because it is easy to interpret "the singer" as the theme of the relative clause while (13) is unacceptable because "the poor" is not the theme of the clause. Instead, the theme of the relative clause in (13) is "Smith". Since it is "Smith" which is relativized in (14), the construction is acceptable.

Similar to Kuno, Bak (1981) is another researcher who applies the concept of theme/rheme and topical progression to the study of Korean and English relative clauses. Bak explores English relative clause in terms of topic-comment articulation or the thematic structures. Three major types of relative clauses are investigated: simple relative clauses, multiple relative clauses and self-embedded relative clause. Consider the following example from Bak (P356):

(15) This is the dog ½›that chased the cat½½›that killed the rat½.

Bak suggests that the thematic structure in this example reveals the most fundamental way of forming a relative clause in many languages. That is, each topic is developed from the preceding comment.

He then concludes that relative clauses and text sequence are quite similar in terms of the thematic structure. However, the only acceptable thematic structures of relative clauses are of two types: the simple linear thematic progression and the thematic progression with a continuous constant theme.

2.4 Accessibility Hierarchy of Restrictive Relative Clause <A Study of Discourse Functions of Relative Clauses from a functional sentence perspective framework>

Many linguists have also accepted that the English relative clauses conform to one specific order known as the Accessibility Hierarchy. The accessibility hierarchy hypothesis claims that more relative clause structures are more easily processed than others.

The NP Accessibility Hierarchy is intended to capture the availability of noun phrase argument positions for relative clause formation. Keenan & Comrie (1977) suggest the following hierarchy, where the positions at the top are universally more accessible for relativization:

(1) Accessibility Hierarchy (henceforth AH)

subject (SU) > direct object (DO) > indirect object (IO) >oblique object or object of preposition (OPREP)> genitive object (GENO) >object of comparison (OCOMP)

Informally, this means that it is easier to relativize subjects than any other

NP position, easier to relativize direct objects than any lower position, and so on.

This hierarchy suggests that relative clause in (16) is more accessible than (17), which is itself more accessible than (18), and etc.:

(16) I kicked the girl {who ate my chocolate cake}.

(17) I kicked the girl {whom the children liked very much}.

(18) I kicked the girl {whom Sam sent the invitation to}.

(19) I kicked the girl {whom Sam was standing beside}.

(20) I kicked the girl {whose sister Sam liked}.

(21) I kicked the girl {whom Sam liked more than John}.

Proposed by Keenan and Comrie (1977), the accessibility hierarchy is influenced by the semantic and grammatical role of the relative pronoun which is co-referential with the relative clause head NP. In (16), for example, the head NP of the relative clause "the girl" is the subject of the relative clause verb "eat". On the contrary, "the girl" in (17) is the direct object of "like" and in (18) is the indirect object of "send" and so forth.

Based on this hierarchy, two important claims have been made. First all languages have subject relative clauses (17), and second, if a language has, for instance, a relative clause of the type "IO", then it will also have relative clause types lower on the hierarchy, which is "SU" and "DO". In other words, we can make claims about all other relative clause types that a language will have when the lowest type is identified.

Nonetheless, it should be pointed out that the levels such as "DO", "IO" or "SU" do not tell about their positions in sentences. For example, "who" in (23) belongs to the "SU" level, but the relative clause can occur in two positions in the Matrix; one position is when the head NP is the subject of the Matrix as in (23), and the other is when the relative clause modifies an NP in the predicate part of the Matrix as in (22):

(22) I kicked the girl who ate my chocolate.

(23) The girl who ate my chocolate is going to pay for it.

Based on the above examples, (22), which involves right branching is generally easier to process than sentence (23) which involves center-embedding. As we can see, what Keenan and Comrie are concerned with is only the functions of the relative pronouns within the relative clause, and not the functions of the head noun phrases within the main clause.

Concerning these semantic and grammatical roles of relative pronouns within the subordinate clause, Keenan (1975) and Keenan and Comrie (1977) make another claim that tends to draw a lot of linguists' attention. They state that when going up the hierarchy, the difficulty of understanding, which implies the difficulty of learning increases.

As in the main clauses, relative clause heads are likely to refer to given information if they are the subject of the relative clause verb and tend to refer to new information when they are the objects of one. Consider the following example:

(24) Susan ate the apple ½›that fell off the tree.½

(25) Susan ate an apple ½›that fell off the tree.½

(26) Susan ate an apple ½›that Sophie had picked.½

(27) Susan ate the apple½›that Sophie had picked.½

Based on these sentences, (24) and (25) should be easier to process than (26) and (27). Since the head of these relative clauses is an object of the main clause, and should thus encode new information. (26) should be the most preferred of all. Based on these examples,

Keenan then emphasizes that, all other things being equal, when subjects are relativized in the insert or dependent clause, the sentence is easier to process than when objects are relativized.

2.5 Cohesion function

Cohesion is one of the most important contents of discourse analysis. In a cohesive discourse, the semantic connections between sentences are very tight; this is because without semantic connections between sentences, a discourse can not be formed. When a language is used for communication, some kind of cohesive means are applied to form close semantic connections so as to construct cohesive discourse and let audiences or readers understand the meaning behind words.

The pre-text and the post-text need cohesive devices to set up semantic relations, and these cohesive devices will take certain forms. In the literature concerning cohesion, only which cohesive devices can occur is discussed, but not why they should take the forms that they take, and not other forms. The present paper discusses the motivations behind the various forms the cohesive devices of text take, and proposes eleven general cohesive principles.

The basic components of text are sentences. How a sentence is formed is decided by grammar system. As long as the grammar system is understood, forming sentences are difficult. But text as a semantic party has no constrains from the perspective of its forms. In this way, it is hard to describe the formation principle of the cohesive mechanism of text. How does the cohesive mechanism of text work? It seems that this basic question of many language phenomenons is seldom studied.

The function of reference is to establish meaning connection. Cohesion distance, more precisely, is a psychological distance, so that cohesion in the short temporary disruption can be continued by nouns, rather than the form of pronouns. It can be called the principle of psychological distance.

In cohesion, sometimes the same items are used to express the same object or concept, sometimes, different items are used to refer to the same object or concept. These are closely related to the communication purposes. In order to display continuity, logical rigidity and precision, relatives are used to show the interdependency.

Another view centers on predictability. Kuno (1972, 1977, 1980) is a proponent of this view. In Kuno's term, linguistic referents are given if they are predictable, and if the addressee could predict them when they are deleted. On the other hand, new referents are unpredictable.

The other view is closely related to what Halliday (1967, 1974, 1981, 1994) calls recoverability. According to Halliday, information that is presented as recoverable is given; if not, it is new. To this, he adds that there are a number of elements in language that are inherently given in the sense that they are not interpretable except by the reference to some previous mentions or some features of the situation. The reference includes anaphoric elements, which refer to things mentioned before, and deictic elements, which are interpreted by reference to the "here and now" of the discourse (Halliday, 1994:298).

The most complex category of discourse entities is what Prince calls inferables. She notes that an entity is inferable if the reader/listener can infer it by means of logical and plausible reasoning from the discourse entities which have been already evoked. For example, readers/listeners are able to infer the driver from the fact thate very bus has a driver in "I got on a bus yesterday and the driver was drunk". Prince further proposes a subclass of inferable entities, which she calls containing inferables. A typical example is "Hey, one of these eggs is broken!"(Prince, 1981:233). The NP one of the eggs is a containing inferable which is contained within the other NP the see ggs which is situationally evoked.

Prince's taxonomy then includes the following seven subclasses: unanchored brand-new, anchored brand-new, unused, inferables, containing inferables, textually evoked and situationally evoked. The first three can be classified as new, the second two as inferable, and the last two as evoked. Thus, based on Prince's taxonomy interms of assumed familiarity, we can make a very fine information-status distinctionof NPs.

All the above views on given and new information are very closely related and not independent of each other. However, for practical analysis that writing researchers are likely to carry out, we do not need to distinguish them so finely all the time. Based on the previous studies, several writing researchers suggest that given information covers elements that have been mentioned prior to a particular point in a text, recoverable from the extralinguistic situation or texts, or inferrable from the other mentioned discourse entities through logical and plausible reasoning or world knowledge. On the other hand, new information refers to firstly introduced discourse entities only (Vande Kopple, 1986, Biber, 2000, Nuamthanom, 2003). Along the line with the arguments of these researchers, we classify inferable and evoked entities into discourse entities representing given information as opposed to new information which contains firstly introduced discourse entities. The corresponding forms of given/new information are defined as follows:

Given information: synonyms, pronouns, repetitions, ellipsis or inferable entitiesfrom other mentioned entities through logical reasoning or world knowledge.¼ŽNew information: first introduced entities.

2.6 Syntactic-Semantic Aspects of Restrictive Relative Clauses

Du Bois (1987) notes that arguments comprising new information appear preferentially in the S or O roles, but not in A role, which leads to formulation of a Given A Constraint, where A role is the 'transitive subject', S role the 'intransitive subject' and O role the 'transitive object'. He formulates Given A Constraint as "avoiding introducing a new referent in the A-role argument position" (p. 827). Fox (1987) and Fox and Thompson (1990) apply Du Bois's notion to examine the distributional pattern of relative clauses in English conversation, and find that S-relatives and O-relatives predominate over A-relatives, where S-relatives represent that the relative pronoun is the subject of an intransitive verb of the relative clause, A-relatives mean that the relative pronoun is the subject of a transitive verb of the relative clause and O-relatives express that the relative pronoun is the object of a transitive verb of the relative clause, as shown in the following examples respectively.

(a) S-relatives

and he's got a spring that comes way up. (Fox, 1987:859)

(b) A-relatives

No in fact I know somebody who has her now. (Fox, 1987: 859)

(c) O-relatives

This man who I have for linguistics is really too much. (Fox, 1987:859)

According to Fox (1987), the preponderance of S-relatives and O-relatives over A-relatives in her study receives the support from the following facts. First, O-relatives predominate over A-relatives since the NPs in subject positions of O-relatives mainly consist of exphoric pronouns such as the speaker I and the hearer you in spontaneous English conversation, they serve as better anchors than the NPs in the object positions of A-relatives which are usually made up of full NPs that usually represent new information in conversation. According to Fox's study, A-relatives are more likely to be chosen when the NPs in the object positions of A-relatives perform the function of linking the current utterance to the preceding discourse, as shown by her in (12). However, the NPs in the object positions of A-relatives rarely represent given information in English conversation since they frequently are the position for new information to be introduced (Givon, 1979; Du Bois, 1987; Fox, 1987 and Fox and Thompson, 1990). A-relatives thus tend to be fairly rare in spontaneous conversation. Second, the preponderance of S-relatives over A-relatives may be caused by the reason that the subject positions of intransitive verbs tend to be chosen to introduce new discourse entities (Du Bois, 1987). Thus, the subject positions of intransitive verbs in S-relatives usually form the position where new information is likely to be given, as shown by a spring in (11). Fox and Thompson (1990) add that the head NPs of S-relatives tend to be indefinite nonhuman objects in the main clauses, and by the time we hear the object head NPs, we have already heard the main-clause subjects which are typically pronouns in speech contexts. In such cases, the object head NPs are already grounded by the main-clause subjects. This is what Fox and Thompson refer to as main-clause grounding. S-relatives here are used not for grounding but for providing new information for their new head NPs.

2.7 Discourse Structures

In this section, first the link between discourse structures and textual relationships within discourse will be established. Then, the definition of discourse unit will be introduced and the occurrences of RRCs in discourse units in the present study will be examined.

Discourse structures are very much concerned with the semantic relations holding between the segments of discourses. The functional relations between discourse segments is one approach to the overall configuration of discourse structures. This approach is emphasized by Hoey (1983) and Hoey and Winter (1986).Emphasis of their works is placed on the interpretive acts involved in relating one textual segment(s) to the other(s) through logical sequence or matching relations. Examples under the heading of logical sequence relation include condition-consequence, instrument-achievement, cause-consequence, and problem-solution, as shown below respectively.


(28) If the royal portrait was not used [on stamps], the arms of the country or reigning house were often taken as a suitable symbol. (Hoey, 1983:19)

In (28), the subordinate clause is the condition and the main clause is the consequence.


(29) [1] It was over, it was known, it was decided, there was nothing at all, ever, to be done about it. [2]He might as well, now, go to bed. [3] So he stood up, put down his empty glass, looked at himself with some curiosity in the mirror, to see if he looked different for having understood, and went to bed. (Hoey, 1983:21)

[1], [2] and [3] all together form a cause-consequence relation where [1] and [2] jointly form the cause for the consequence given in unit [3].


(30) The trouble with this country's economy used to be that there were too many farthings being made. Once we stopped making them everything turned out all night. (Hoey, 1983:51)

The linguistic device the trouble indicates a problem-solution relation between two discourse segments.

When segments of a text are compared or contrasted with one another, they can be brought under the heading of matching relations. Consider examples (29) and (30).


(31) People think of Birmingham in different way. Alderman Frank Price sees the city as a sort of anvil, whereas my barber thinks of it as 'a neutral sort of place built by people who worked hard for generations'. (Hoey, 1983:23)

In this example, two kinds of people's thought about Birmingham are under contrast.


(32) Upstairs Fred thumped and bumped and tossed and turned. And downstairs Ted moaned and groaned and crashed and thrashed all over the bed. (Hoey, 1983: 20)

In text (32), Fred and Ted are being compared for their similarity of response to the sleeping arrangements.

According to Winter and Hoey, logical sequence and matching relations are the two basic categories of clause relations between segments of texts. Besides, they add that another common clause relation is general-particular, including two subcategories general statement-example and preview-detail. The following text characterizes general statement-example relation.

(33) It is interesting to note that iconic models only represent certain features of that portion of the real world which they simulate. For example, a map will only contain those features which are of interest to the person using the map. Similarly, architects' model will be limited to include only those features which are of interest to the person considering employing the architect. (Hoey, 1983:113)

The first sentence of (33) is a general statement for which the subsequent two matching sentences provide examples.

Winter and Hoey note that the texts above often contain clues to help the reader/listener interpret the clause relations. The most apparent means whereby a clause relation may be signaled to the reader/listener is by the use of subordinators, conjuncts, and vocabulary items. Take (31) and (33) for examples, in (31), the conjunct so occurring in the beginning of sentence [3] tells us that three sentences form a cause-consequence clause relation. In (33), the adjective different first tells us that people's thought about Birmingham is to be contrasted and then whereas reveals that the contrast is under way.

From previous studies discussed above, we can see that discourse often contains linguistic devices to facilitate the interpretation of textual relations between units of discourse. Linguistic devices are supporting evidence to the cognitive activity of deducing the relations. Subordinators, conjuncts and lexical items are apparent means whereby textual relations can be signaled to the reader/listener. However, some texts may not involve explicit devices to express the functional relations between discourse segments. In such cases, the reader needs to infer the functional relations.

Guided by previous studies, in the present study, based on Hoey (1983) and Winter (1986)'s classification of textual relation, a discourse unit can be defined as a discourse segment which bears a functional relation with one another. In the present study, the defining of a discourse unit relies on the author's own interpretative act of textual relations existing between discourse units. Thus, this division may be subjective, but the author follows one general agreed principle: discourse units must have a textual relationship with one another.

(34) The story of a poor family that acquired fame and fortune overnight, dramatically illustrates the power of the press. The family lived in Aberdeen, a small town of 23,000 inhabitants in South Dakota. As the parents had five children, life was a perpetual struggle against poverty. They were expecting their sixth child and faced with even more pressing economic problems.

--------- Lesson 45

The functional relation between the first sentence and the rest of the paragraph is general statement-example. The first sentence is the topic sentence and the rest of the paragraph provides an example for the statement. In this text, we can see that the RRC occurs in the beginning position of the discourse unit and also introduces the topic of the whole discourse . Then, consider the following example.

(35) [1] In many ways, this is unfortunate for the poor actors who are required to go on repeating the same lines night after night. One would expect them to know their parts by heart and never have cause to falter. [2] Yet this is not always the case.

--- Lesson 22

In (35), the whole paragraph consists of two discourse units: Discourse units [1]and [2] are linked by the contrast relationship. And the S-relative occurs in the beginning position of discourse unit [1].

A further investigation is conducted on how the discourse functions of RRCs are related to their occurrences in discourse units. Take texts (34) and (35) for examples, in (34), the RRC is an A-relative and the NP the introduction of a poor family in the RRC has not been mentioned in the previous discourse, it thus serves as an introductio to the new head NP the poor family. The new head NP the poor family then becomes the topic in the following discourse. The RRC thus performs decription function and further presents the new head NP the poor family as the topic in the whole discourse. In (35), the RRC occurs in the beginning position of discourse unit [1], and it is an S-relative. The function of the S-relative is to provide new information for the head NP. The information stated in the RRC here involves the opening of the new referent and the anticipation of the information that will be constructed in the ongoing discourse. Thus, the RRC introduces the head NP as the topic in the subsequent discourse. From texts(34) and (35), we can see that the purpose of using RRCs in the beginning of discourse units is to introduce the new head NPs for further development. Such development is achieved in two ways. The new head NPs are further developed either by making them grounded through the given entity in the RRC. Or new information in the RRC may be added to the newly-introduced head NP and further make the new referent become the topic in the following discourse. In either way, RRCs in the beginning of discourse units are used to develop new head NPs for the subsequent discourse.

From the discussion above, it can be seen that the discourse functions of RRCs bear close relation to their occurrences in discourse units. When RRCs occur in the initial positions of discourse units, they primarily serve as a device to formulate new head NPs as topics in the subsequent discourse. How the discourse functions of RRC sare related to discourse structures of written texts will be fully investigated in the following chapter. From the brief discussion above, it is obvious that in addition to the aspects of information flow such as given/new information status of NPs, 'discourse structure' factor is crucial in deciding the discourse functions of RRCs in written argumentations and expositions.

2.8 Summary and Conclusion

From the previous studies, it has been investigated how English RRCs are operated both in spontaneous conversation and written texts. RRCs are used for two reasons: grounding and description. RRCs achieve grounding function by either performing as proposition-linking or anchoring the new referents through the given referents in RRCs. Besides, RRCs fulfill the description function by providing new information for their newly-introduced head NPs. Moreover, RRCs are realized by different types of syntactic relatives: A-relatives, S-relatives, and O-relatives. Different syntactic types may contribute to different discourse functions of RRCs.

The findings of previous studies are mainly from studies on spontaneous conversations and written text. However, here is one thing that differs from speaking texts and written texts. The typical pattern of given/new information status of discourse entities in spoken discourse is different from those in written discourse. Most referring expressions in spontaneous conversation present given information, and among these given information, exophoric referents such as you and I account for over half of all given referents in conversation, while this kind of given referents are not present quite often in written texts given the features of written texts. (Prince, 1981 and Biber, 2000). Second, the preferred nominal forms used for given referents are also different in these two registers: Spoken texts have greater reliance on pronouns and written texts use full lexical NPs most of the time. Compared with spoken discourse, written discourse consists of higher proportion of full lexical NPs expressing new information (Biber, 2000). So it is clear that there exists discrepancy between spoken and written languages with respect to information status of referring expressions. Based on this fact, we have reasons to believe that written texts may not rely on as many human pronouns in grammatical subject positions as spoken texts do to anchor new referents. O-relatives in written languages thus may not occur as frequently as in spoken languages.

Besides, written languages often display discourse structure markers that are greatly different from those used in spoken languages. Written texts tend to rely on an extensive set of linguistic markers such as logical connectors moreover and temporal markers when or rhetorical organizers of larger stretches such as firstly and in conclusion. Thus in this way, written texts usually have very clear discourse structures most of the time. As for spoken languages, due to limited time, spoken languages frequently depend on extralinguistic clues to interpret the functional relation between clauses, so they often use many incomplete sentences and the boundary of discourse units tend to be marked by non-linguistic elements. Therefore, written languages display their typical patterns of discourse structures very differently from spoken languages. Based on the analysis of difference between speaking and written text, it is assumed that in written text, linguistic markers will be more often used, since relative clauses are an obvious linguistic marker, so restrictive relative clauses may be used quite often.

To get a comprehensive understanding of discourse structures of written texts, we establish the link between discourse structures and textual relationships within discourse. The functional relations holding between discourse segments may be included under two broad categories: logical sequence and matching relations.

In this paper, it is believed that in addition to the factors such as information status of NPs, coherence requirement of the text, the difficulty of accessibility and syntactic-semantic types of RRCs, discourse structures may also play a crucial role in affecting the discourse functions of RRCs in written texts. Thus, notions 'information status of NPs', 'coherence', 'syntactic types of RRCs' and 'discourse structures' will be employed to investigate the discourse functions of RRCs and their relations to their occurrences in discourse units.