A study in referential coherence: Chinese writers of English

A study in referential coherence: Chinese writers of English

Huisi Li S 1940554

MA in Applied Linguistics

Faculty of Liberal Arts

University of Groningen

Supervisors:

Dr. Marjolijn Verspoor

Dr. Wander Lowie

Abstract

Table of Contents

Abstract ……… 1

Introduction ……….. 2

Background literature ……….... 5

Different patterns of thought? ……….... 5

What is coherence? ………. 8

Referential coherence ………... 9

Mental processing instructions ……….. 14

Methods of measuring referential coherence ………. 18

Conclusion ……… 20

Methodology ……….. 21

Main study ……… 21

Clines in Brown’s study ………... 29

An example of analysis ………... 30

The follow-up study ………. 32

Result ……… 34

Main study ……… 34

The follow-up study ………... 40

Discussion ……….. 42

Main study ……… 42

The follow-up study ………. 50

The pattern of thought ………... 51

Limitations of this study ……….... 56

Conclusion ………... 58

Reference ………... 60

Appendix Ⅰ ……… 64

Introduction

English writing is difficult for most Chinese students. They spend a huge amount of time on reciting new and sophisticated words, learning advanced sentence structures and remembering different cohesive devices, but their efforts are not always successful. This phenomenon draws attention of many Chinese researchers (Liu & Brain, 2005, Shen, 2010, Zhang, 2000) who dedicate themselves to studying the reasons for poor English writing. One area of their studies is cohesion and coherence. Chinese researchers embrace different opinions, such as Halliday and Hasan’s (1976) and Kaplan’s (1966), to study this area. However, since it is difficult to measure coherence, most studies investigate only the problems of cohesion, but cohesion is not always necessary for readers to understand the content (Seidlhofer and Widdowson 1997). Moreover, using cohesive devices is not difficult for Chinese students to learn (Liu and Braine, 2005). Therefore, as difficult as it may be, coherence should be studied from an ESL perspective in China.

Based on such considerations, Shen (2010) investigates coherence from an ESL perspective. In her study, 30 non-English major students are invited as participants. They are required to rearrange 4 groups of sentences to make 4 coherent paragraphs. The first 3 groups have 5 sentences respectively and the last group has 6 sentences. One example of a group of sentences is shown below:

A. Of the effects caused by vitamin A deficiency, those involving eye diseases are the most pronounced and widespread.

C. What children eat can affect their health.

D. Several thousand children become blind each year because of this dietary deficiency, which is most prevalent in poor, non-industrialized countries.

E. Children who do not eat enough foods containing vitamin A can develop serious nutritional disorders.

(Shen, 2010:111)

The result shows that 21 participants out of 30 got a zero score, which implies that 70% of the participants do not “have any knowledge about how to produce a coherent passage by uniting several individual incoherent sentences into a coherent one” (Shen, 2010:109).

It is a disappointing result because the participants in Shen’s study (2010) have been learning English for more than 10 years, but they still do not understand how to make an English paragraph coherent. These participants are like Chinese learners of English, who may not make any grammatical mistakes in their essays, but who do not know how to make their essays coherent. As the following example from one of my own participants shows, the information does not “flow” in their English essay.

The participant is a Chinese college student majoring in metallurgy and there are hundreds of thousands of Chinese learners of English who have similar problems in writing a coherent English paragraph. This problem brings up the question: How can researchers and teachers help them? This question can be answered only by studies focusing on coherence, but so far, there are no studies that have done so.

Therefore, the current study will first try to find out how coherence is achieved in English essays by native speakers and compare these results with coherence measures in English essays written by Chinese speakers. Once the discrepancies have been accounted for, I will try to deduce possible factors that undermine coherence for Chinese writers.

Background Literature

The aim of my study is to find out the factors involved in making English essays written by Chinese writers sound incoherent. To do so, I will first explain Kaplan’s opinion. He (1966) proposes that different patterns of thought of Chinese writers may be reflected in their English essays, which undermine coherence. Then I will review the distinction between cohesion and coherence and what coherence is. To find a way to measure coherence, referential coherence is introduced later and why referential coherence should be studied is reviewed as well. At last, I will introduce how to measure referential coherence.

Different patterns of thought?

One of the first researchers to point out that ESL writers may not write like their native speaker counterparts was Kaplan. He (1966) assumed that ESL learners with different L1s may have different patterns of thought caused by different cultures, and the patterns of thought are reflected in the patterns of English writing. Kaplan (1966) analyzed about 600 essays. Based on the analyses, he proposes four patterns of writing used in a paragraph by Semetic, Oriental, Romance, and Slavic learners of English. He also describes a pattern of writing used by English native speakers. The patterns are illustrated below:

From his illustration, it is clear that foreign learners of English use different strategies to organize a paragraph compared with English native speakers. He believes these different patterns of writing cause foreign learners’ English essays to be “out of focus or lack organization” (Kaplan, 1966:3).

Kaplan’s study has been supported and criticized by many other researchers. Ostler (1988) is one of the followers. She does not focus on the pattern of writing, but she investigates more detailed characteristics of English essays written by English, Arabic, Japanese and Spanish speakers. The results show that the essays written by students with the same L1 have unique characteristics, which can be used to distinguish these essays from others written by other students with the same L1.

However, both Kaplan’s study and Ostler’s study are challenged by other researchers. They claim that there are two major problems. The first one appears in Kaplan’s study. The challengers think that not all English writers use a linear style. The writing style is different from writer to writer (Braddock, 1974). Hinds (1983) also criticizes the linear style used by English native speakers. Another problem is embedded in both Kaplan and Ostler’s studies. The problem is that Kaplan and Ostler do not take the developmental factor of English proficiency into account (Noor, 2001). The problem of writing style may not be the result of different patterns of thought, but the result of an insufficient level of English.

that the developmental factor is the cause of weak organization. Therefore, they assess organization in 3700 essays written by students in Canada between the grade 8 and the grade 12. The result demonstrates that students from grade 12 do significantly better than students from grade 8 in organization, despite the fact that the paragraph structures are still weak. Based on this finding, they assert that English proficiency should be taken into consideration. The result also shows that seventy percent of those students speak English at home. However, they do not organize the essays significantly better than the other students. Thus, they suppose that the transfer from L1 to the L2 does not exist in terms of organization. Moreover, the survey about how English teachers evaluate essays in Hong Kong and Canada indicates that the English teachers in Hong Kong tend to pay more attention to accuracy than to the organization of essays written by their Chinese students. Based on this observation, Mohan and Lo claim that Chinese students’ experience of learning English writing in school results in weak organization. In other words, the instruction, not the different patterns of thought, leads to Chinese students’ weak organization in English writing (Mohan & Lo, 1985).

To sum up, these controversial opinions about patterns of thought can confuse ESL English teachers; especially Mohan and Lo’s study has influenced Chinese ESL teachers. Whether there are different patterns of thought, which causes the weak organization in essays written by Chinese learners of English, is not entirely clear; therefore Chinese ESL teachers are confused about the instructions used to teach writing.

study.

What is coherence?

The terms “coherence” and “cohesion” are closely related, but according to several authors, coherence is a hyponym for cohesion, and whereas cohesive devices are explicit, coherence is implicit in the text. According to Halliday and Hasan (1976: 23), “A text is a passage of discourse which is coherent in these two regards: it is coherent with respect to the context of situation, and therefore consistent in register; and it is coherent with respect to itself, and therefore cohesive.” Several researchers embrace this opinion and develop the concept of coherence further. Seidlhofer and Widdowson (1997:207) claim that “cohesion is simply the textualisation of those contextual connections which it is assumed need to be made explicit, but coherence is the discourse function of realizing those connections.” In other words, they think coherence is the process of noticing the cohesion. They also propose that a coherent text does not necessarily need cohesive ties, and cohesion does not definitely lead to coherence. Sanders and Spooren indicate that “coherence has to be defined in terms of the cognitive representation people have or make of a discourse” (1997: 235). They believe that coherence is a cognitive process because even if linguistic cues and connectives are not available, the readers/hearers can still regard a discourse as a coherent one.

account when written discourses are evaluated.

However, from an ESL perspective, most studies have focused on cohesion, not coherence. Studies clearly show that ESL writings have different features in cohesion. For instance, Arab ESL learners prefer to overuse some cohesive devices but underuse others (Khalil, 1989); Chinese ESL learners show similar features (Zhang, 2000; Liu & Braine, 2005) that may affect the coherence of writings. However, studies also indicate some controversial results. Some studies show that the development of cohesive devices is not different in “good” and “weak” writings (Johnson, 1992; Zhang, 2000), while others show that cohesive devices in “good” writings are different from those in “weak” writings (Jafarpur, 1991). One possible reason for these controversial results is that the cohesive devices in good writings are not necessarily more advanced than those in weak writings. In other words, the use of cohesive devices may not be sufficient to evaluate writings (Witte & Faigley, 1981). Witte and Faigley state that coherence, not cohesion, “allows a text to be understood or used” (1981: 202). But it is not easy to analyze coherence. Khalil (1989) tries to analyze coherence with maxims (quantity, quality, relation and manner) proposed by Grice (1975). However, the analysis is more like a holistic evaluation. Clearly, the lack of studies that address coherence in an ESL perspective is related to the fact that it is difficult to analyze and operationalize coherence. However, coherence can be measured and analyzed to some extents by means of one aspect called referential coherence.

Referential coherence

includes several studies on discourse continuity by different researchers. From Givón’s perspective, continuity in discourse consists of three levels: thematic continuity, action continuity, and topic/participant continuity. Thematic continuity is “the overall matrix for all other continuities in the discourse” (1983: 8). It means the theme in the discourse is the same, which refers to “thematic paragraph” (1983: 8). Action continuity is inferior to thematic continuity but superior to topic/participant continuity. It is related to “temporal sequentiality within a thematic paragraph” (1983: 8). Topic/participant continuity is the most concrete one. If the topics/participants are consistent within one thematic paragraph, the whole paragraph is very likely “continuous” (Givón, 1983).

Since thematic continuity and action continuity are not easy to analyze, topic/participant continuity has become the focus. Topic/participant continuity has three major types within a thematic paragraph (Givón, 1983). These three types are: chain initial topic, chain medial topic and chain final topic. The detailed characteristics of these three types are as follows:

Chain initial topic:

(a) Characteristically a newly-introduced, newly-changed or newly-returned topic; thus

(b) Characteristically a discontinuous topic in terms of the preceding discourse context; but

(c) Potentially-if an important topic-a rather persistent topic in terms of the succeeding discourse context.

(a) Characteristically a continuing/continuous topic in terms of the preceding discourse context; and also

(b) Characteristically persistent-but not maximally so-in terms of succeeding discourse context, even when an important topic.

Chain final topic:

(a) Characteristically a continuing/continuous topic in terms of the preceding discourse context; but

(b) Characteristically a non-persistent topic in terms of the succeeding discourse context, even if an important topic.

(Givón, 1983:9)

The primary topic is commonly the grammatical subject. The direct object can be the second one (Givón, 1983 & 1992). If these topics in a thematic paragraph can be identified by readers/hearers easily, the paragraph will be regarded as a coherent one. However, several factors can hinder readers/hearers in identifying topics. These factors are as follows:

Factor Explanation

Length of absence from the register (register means topics appear and “register” in memory)

Indefinite and a firstly introduced topic is “maximally difficult” to be identified; the big gap between two appearances of one topic can hinder the identification as well. A shorter gap of absence from the register leads to easier identification.

vice versa.

Availability of semantic information “Redundant” semantic information, primarily from the predicate of the clause, may facilitate identification of the topic when potential interference occurs. “This information concerns generic probabilities that a particular topic could participate in the clause in the specific semantic/grammatical role”.

Availability of thematic information The thematic information in preceding discourse can facilitate identification of topics. This information “establishes specific probabilities-for this story, in this chapter, in this section or in this thematic paragraph-as to the topic identification within a particular clause and in a particular role”.

(Givón, 1983: 11)

Although four factors are identified, Givón (1983) claims that only the first two can be easily measured.

reference. Anaphoric reference refers to something in the preceding text, while cataphoric reference refers to something in the following text. Both types “contribute to the coherence of the text” (Dirven and Verspoor, 2004: 186). Moreover, Driven and Verspoor (2004:187) indicate that in West European languages, “if an object is completely new to the text, it has to be introduced” by “using an indefinite expression” first in order to maintain the coherence of the discourse.

Obviously, referential coherence is not a new topic. However, few studies of referential coherence are conducted from an ESL perspective. Apparently, ESL learners know how to use cohesive devices after training. For instance, Liu and Braine’s study (2005: 623) shows that Chinese ESL undergraduates can use “a variety of cohesive devices in their writings”. But sometimes the cohesive devices do not guarantee readers to feel the writings coherent (Seidlhofer and Widdowson, 1997). See the example below:

Unfortunately in the days to come autumn leaves will become a hazard to the elderly as they fall and become a wet soggy mess on the pavement.

(Seidlhofer and Widdowson, 1997:208)

Mental processing instructions

not found, “search of all previously processed propositions is made” (1978: 368). When the search is successful, the processing of new input continues; if not, “the inference is initiated” (1978:368).

Another researcher, Cirilo (1981), conducts a study to investigate the distance effect of referents on the speed of reading. Distance effect is a product of limited capacity of short-term memory since only a part of input text is active during the reading. There are two types of match (connections) between previous referents (topics) and newly input referents: immediate match and reinstatement match. Immediate match occurs when “a newly input referent can be identified with an earlier referent which is currently active in short-term memory” (1981: 358). Reinstatement match occurs when “a search must be made through the inactive propositions in long-term memory” if immediate match is impossible (1981:358). However, when both matches are not possible (“when the contents of memory are insufficient to establish coreference”), inference is employed (1981:359). The distance of immediate match is the smallest and the distance of inference is the biggest. It is assumed that different distances have different effects on comprehension of reading, which is proved by the results. The results show that the speed of reading are different when the distances are different, which means the difficulties of processing sentences can be affected by distances between referents (Cirilo, 1981). McCown and Miller (1986) investigate the effect of referential coherence in text processing as well. They find that referential coherence plays a significant role when participants recall the contents of reading materials.

definite topic (referent) has been within readers/hearers’ “file (or register)” already; therefore it can be retrieved quickly. However, an indefinite topic (referent) is not available in the “file”; therefore the readers/hearers have to “open the initial file” (1983:10).

Then he investigates referential coherence further. He states that referential coherence is essential because it can “ground the information” (1992:8). The information includes two types: the old information and the new information. The old information is predictable while the new information is unpredictable. Totally old or new information is not useful because the old information offers “no motivation for attending” and the new information offers “no grounding point for the information to cohere” (1992:8). Therefore, a discourse tends to include two types of information. The old information is to “serve to ground the new information to the already-stored old information” (1992:8), which is similar to Kintsch and Van Dijk’s opinion of “connection” between old and new propositions (1978).

When the attentional activation is used, there are four major choices because of manipulation “by the grammar of referential coherence” (Givón, 1992:23). These four major choices are:

a. continue activation of the current open file; b. terminate activation of the current open file; c. activate a currently inactive file

(i) open-activate-a new file;

(ii) reopen-reactivate-an existing file (Givón, 1992:23)

Givón (1992) also claims that different mental efforts are needed to process different referents. For instance, “the activation of an inactive referent requires more mental effort; and referents that are already active require minimal coding” (Givón, 1992:25). In other words, if the referents are not inactive, (the new information is not coherent with the active referent), the more mental effort is needed; and then the readers/hearers will find difficulty in processing the discourse.

Methods of measuring referential coherence

Givón (1983) designs a system to measure two factors that can hinder readers/hearers in identifying topics, which are length of absence from the register (register means referent appear and “register” in memory) and potential interference from other referents. In the system, three methods are employed. These are “referential distance (look-back)”, “potential interference (ambiguity)” and “persistence (decay)” (Givón, 1983: 13-14):

Referential distance is to measure the distance between the referent’s previous occurrence and present occurrence. The minimum of the measurement is a clause. If the distance is “1 clause”, the topic is “maximally continuous” (Givón, 1983:13). The maximum of the measurement is set at 20 clauses by Givón. If a referent is new, “20 clauses” will be assigned automatically (Givón, 1983).

Potential interference is to measure the ambiguity of the referent. The preceding referents may affect the identification of the current referent. However, if the preceding referent is far away from the current one, the effect will be very slim. Therefore Givón proposes that only the immediately preceding referents can interfere with the current one. These immediately preceding referents are “between 1 and 5 clause to the left, most commonly 3 clauses” (Givón, 1983:14). The way to code the ambiguity is mentioned by Brown (1983:319): if there is no interfering referent, that means there is only “one logically possible” referent, “1” is assigned; if two or more “logically possible” referents appear, “2” is assigned. If there is no “logically possible” referent, “3” is assigned.

which means the “lowest persistence”; and there is no assigned maximal value (Givón, 1983:15).

In 1992, he incorporates two new methods in his system to measure referential coherence. These two methods are “switch reference” and “overall frequency”. Switch reference is to investigate “whether the preceding clause does have the referent as an argument or does not” (Givón, 1992:16). If the referents (subject) are the same, then SS (same subject) is assigned; if not, DS (different subject) is assigned (Givón, 1983). Overall frequency is to find out “the total number of times the same referent appears as clausal argument in the discourse” (Givón, 1992:17).

These constructs and measures can help analyze referential coherence to some extent. Moreover, in Givón’s point of view (1981), it is better to analyze different types of referents/topics separately since different types are used to “bridge” different “gaps” and the types with smaller “gaps” are supposed to be identified more easily than the types with bigger “gaps” (Brown, 1983). Givón (1983:17) thinks “zero anaphora” is the easiest one to be identified by readers/hearers, followed by “unstressed/bound pronoun”, “stressed/independent pronoun”, “right-dislocated definite noun-phrase”, “neutral-ordered definite noun-phrase”, “left-dislocated definite noun-phrase”, “contrastive topicalization”, “cleft/focus constructions”, “referential indefinite noun-phrase ”.

Embracing the same opinion, Brown (1983) proposes that definite NP should be subdivided into 5 types: definite article + NP, demonstrative + NP, demonstrative alone, name, NP after possessives; and indefinite NP should be subdivided into referentials and existential.

language can affect it because each language has its own typical features. Therefore, it is possible that the features of referential coherence in one language do not appear in another language. As a result, it is impossible to establish a universal standard to assess topic continuity. The only feasible way is to analyze each language and find out its own characteristics of referential coherence (topic continuity). From this point of view, Givón and other pioneering researchers (1983) have investigated referential coherence (topic continuity) in many languages, such as Japanese, Amharic, Ute, Biblical Hebrew, Latin-American Spanish, English, Hausa and Chamorro, and described different features in each language.

Based on these characteristics of referential coherence, it is assumed that if different features of referential coherence are found in ESL students’ English essays compared with the features in authentic written English, the differences will probably undermine coherence in the essays. Therefore, the current study is conducted. In this study, only one method, referential distance, is employed. In other words, the current study tries to analyze referential coherence in terms of the length between current referent and its register.

Conclusion

Methodology

The main purpose of this study is to find out why English texts by Chinese writers may not sound coherent. The assumption is that it could be related to different strategies in establishing referential coherence. This section will consist of two separate studies. The first one, which is the main study, will compare the strategies used by English and Chinese writers of English texts to establish referential coherence. The second study, which is a follow-up study, will focus on the part where striking differences are found between Chinese writing style and English writing style: the positions of old information and new information

The research questions of this study are as follows:

1. What are the features of “referential distance” in a coherent text written by native speakers of English

2. What are the features of “referential distance” in English essays written by native Chinese Speakers?

3. What are the most striking differences in “referential distance” between English texts written by native speakers of English speakers and by Chinese speakers?

4. What are the possible explanations for these differences?

In this chapter, I will firstly explain the main study, and then the follow-up study.

Main study

other methods to measure referential coherence, it is believed that referential distance could show different features of referential coherence in the essays sufficiently; therefore only the referential distance in the essays are analyzed. After analyzing the essays, the result is compared with the analysis in Brown’s study to find out specific feature of referential distance in the essays. In Brown’s study, an English novel written by a native speaker (Doctor No by Fleming) is analyzed. The English novel is assumed to be a coherent one, which can represent features of referential distance in written English. Although the novel is a narrative, it is assumed that referential distance does not vary dramatically among different genres. The last step of the study is to find out possible causes of specific features in the essays written by the Chinese learners, based on which the follow-up study is conducted.

The technique used in the main study is discourse analysis. Excel is used to analyze the data and provide necessary graphs.

Participants

Data

The essay used in this study is a writing assignment in December. Students were required to write a text of about 150 words on “sources of information” after they had studied a unit about information. They wrote the essays in their spare time without any time pressure and they could use any available tool, such as the Internet and dictionaries, to help them finish their assignments. They were not informed that their essays would be collected as the data of this study. Therefore they wrote their essays in a natural situation.

Measure

The current study employs the method used in Brown’s study (1983) to measure the average distance between referents. The basic measurement is a clause. All “main, subordinate and other clauses” are analyzed. Relative clauses, noun clauses, infinitive clauses and gerund clauses are “counted as part of other clauses in which they function” (Brown, 1983:318). In other words, only adverbial clauses are counted separately; the other two kinds of subordinate clause (relative clause and noun clause) are not counted separately. In the texts written by the participants, there are 302 main, subordinate (adverbial) and other clauses. The detail of the number of clauses in each essay is shown below:

clause clause clause clause

40921133 12 40921152 12 40921173 13 40921191 20 In this study, the minimum measurement is “1 clause” and the maximum measurement is set at “20 clause”. Brown gives an example about how referential distance is measured:

Jim insisted on helping Mary with the typing. Although it was a simple task, he managed to make it seem like a major ordeal. By the time he had finished three pages, Mary had finished the other ten and φ was ready to leave.

For example, in the first sentence, Jim, Mary, and the typing would all receive a lookback value of 20 because there has been no previous mention of any of them. Both italicized its would have a lookback value of 1, the first being counted back to the typing, and the second being counted back to the first the first italicized he would receive a lookback value of 2, going back to Jim in the first clause. The second he would refer to the first he and would have a lookback value of 1. The second Mary would have a lookback value of 4, going back to Mary in the first clause and the φ would have a lookback value of 1.

(Brown, 1983:318)

regarded as the same referent. Moreover, the “parts” cannot stand for the “whole” or the “whole” cannot stand for “parts”. In this study, it is decided that “information”, different adjective/determiner plus “information” (good information, useful information, some

information, our information, etc.) and the “information” which is needed by different persons are not regarded as the same referent and are coded respectively. The same rule is applied to other nouns in the data.

Subjects and direct objects are coded since they are more like a topic in a clause (Givón, 1992) in the current study. If the same referents appear in the same clause, they are not coded repeatedly (e.g. in “we can get the information which we need…”, “which” and “we” are not coded as referents). Based on Brown’s method, all referents should be divided into 16 types when they are coded since the difficulties of processing different types of referents are different. These 16 types (Brown, 1983:320-323) are demonstrated below (all examples are from the data in the current study; the corrections are in brackets):

(1) Zero anaphora: a NP is deleted completely, which is marked with φ. (Imperative is not included since it is a special pattern)

We can choose which one is best for ourselves, and φ get information from the way.

Φ Facing them, we should know how to choose them.

(Imperative is not included since it is a special structure.)

We can get it from Newspapers, TVs, and Internet etc.

(3) Stressed pronoun: the personal pronoun which has stress. They can only be found by reading aloud. Since it is not found in the data, this type is excluded.

(4) Left dislocation: the same two referents appear in the same clause and the regular clause contains the second one. For example:

As for newspaper and books, they (are) slower than Internet, but reliablbe (reliable).

(5) Definite article + NP (noun phrase): the article the plus a modified or unmodified NP. For example:

In my opnion (opinion) we should have the ability to make a right judge to the different information.

(6) Demonstrative + NP: this, that, these or those plus a NP. For example:

So we must object those bad information and search those useful information.

example:

I believe this is impossible.

(8) Names: proper nouns such as Jim and Mary. This type is not found in the data; therefore it is excluded.

(9) NP following possessive: a NP follows a possessive pronoun. For example:

Don’t waste your time, and make full use of your time to learn useful things and to do wonderful things.

(10) Right dislocation: two same referents appear in the same clause and the regular clause contains the first one. This type is not found in the data, therefore it is excluded.

(11) Passivization: the word is normally an object but placed on the subject position. For example:

To meet demand of the public, so many information websites has been built.

(12) Y-movement: a NP is moved out of its normal position and placed at the beginning of the clause. This type is excluded since it can not be found in the data.

no particular referent. For example:

We can get enough information in a convinent (convenient) way which saves money and time for us all.

(14) Existential: there is/are type clause. For example:

There are many sources of information.

(15) Generics: a NP or NPs stands for all of a type. For example:

Books and magazines are the important ways to get information. Information plays an important role in modern society.

(16) Cleft/focus construction: it is used at the beginning of the clause as a place holder only, followed by an NP which is modified by a relative clause. This type is not found in the data, therefore it is excluded.

anaphora, unstressed pronouns, definite article plus NP, demonstrative plus NP, demonstrative alone, NP following possessive, passivization, indefinite referentials, existential and generics.

Clines in Brown’s study

Brown (1983:329) applied his analysis of referential analysis to the novel Doctor No. The following table shows the distance between references in terms of the average number of clauses, here presented in descending order:

An example of analysis

The table below shows the abbreviations of different types of referent. The essay below the table is written by one participant. The type of referent and the distance between the current referent and the previous one are showed in brackets.

Abbreviations Full Titles UP Unstressed pronoun G Generics IR Indefinite referentials Ex Existential φ Zero anaphora Da Demonstrative alone NP after Poss NP after Possessive

Pa Passivization DNP Demonstrative + NP Def Art + NP Definite Article + NP

Clauses Paragraph 1

1 We (UP, 20) all know in this centry,

2 information (G, 20) plays an important part (IR, 20) in modern society. 3 Everyday peoply (G, 20) want to know new things (IR, 20),

4

5

It (UP, 20) is important to keep up with the tend of the modern life (Def Art+NP, 20).

Paragraph 2 6 There are many sources of information (Ex, 20).

7 We (UP, 6) can get it (UP, 3) from Newspapers, TVs, and Internet etc. 8 Each way (IR, 1) has its benefits (NP after Poss, 20),

9 We (UP, 2) can choose which one (IR, 1) is best for ourselves, 10 and (φ, 1) get information (IR, 3) from the way (Def Art+NP, 1).

11

Or we (UP, 1) can choose all the ways (Def Art+NP, 4) to learn and know the information (Def Art+NP, 1).

12 But there are too many sources of information (Ex, 6). 13 (φ, 2)Facing them (UP, 1),

14 we (UP, 1) should know how to choose them (UP, 2). 15 Find what information (IR, 20) attract you (UP, 20) 16 and choose it (UP, 1).

Paragraph 3

17 In my opinion, for there are too many different sources of information (Ex, 20),

18

We (UP, 4) should choose the most useful one (Def Art+NP, 20) and most attractive one (Def Art+NP, 20).

22 Don’t waste your time (NP after Poss, 20),

23

and make full use of your time (NP after Poss, 1) to learn useful things (IR, 20) and to do wonderful things (IR, 20).

The follow-up study

This study is a follow-up study in the last step of the main study. It is to find reasons for the most striking differences found in referential coherence in the texts by English and Chinese writers. To do so, the study downloads two articles (one is in Chinese and the other one is in English). The purpose is to find out whether there are differences between Chinese and English texts in positions of old information and new information in each clause. At last, it uses Chi-square to test the significance of association with languages.

Data

For this study two articles are downloaded from the Internet. One is in Chinese, retrieved from Sina, and the other one is in English, retrieved from The Economist. Both articles are commentaries, discussing the same international issue; and both websites are leading ones in their target groups. Then the study uses 19 clauses from each article as corpus.

Measure

inferred from previous clause”. New information is operationalized as “can not be found or inferred from the previous clause”.

Method

As in the main study, only main clauses, adverbial clauses and other clauses are taken into account. Then if there is new information before the old one, “Y” is assigned; if there is no new information before the old one, “N” is assigned. If there is no old information, “Y” is assigned as well. For example (the old information is italicized and bold):

That has led to the charge by hawkish American Republicans, as well as many

Israelis, that Mr. Obama is bent on betraying Israel.

N

In fact, he is motivated by a harder-nosed appreciation of the pros and cons of America’s cosiness with Israel, and is thus all the keener to prod the Jewish state towards giving the Palestinians a fair deal.

N

and φ is thus all the keener to prod the Jewish state towards giving the Palestinians a fair deal.

N

He has condemned the building of Jewish settlements on Palestinian territory more

bluntly than his predecessors did,

N

because he rightly thinks they make it harder to negotiate a peace deal. N

Mr. Obama’s greater sternness towards Israel is for the general good—including

Israel’s.

N

Results

Main study

The result shows that the English essays written by Chinese students have different features of referential distance compared with the features in Doctor No. The overall comparison is as follows:

Types of referent

Average referential distance in Doctor No

Average referential distance in the data φ 1.00 2.39 UP 1.72 6.33 Da 2.27 7.33 Pa 9.37 20.00 DNP 9.84 1.75 Def Art + NP 16.66 17.51 Ex 18.16 17.85 NP after Poss 18.34 18.20 IR 19.17 17.56 G 19.23 13.66

Among these 10 types, demonstrative + NP appears twice; passivization appears once and demonstrative alone appears three times. Since the frequencies of these three types are low and the average referential distances cannot be representative, the study decides to eliminate them.

descending orders sorted by the average referential distances are different in Doctor No and the data. To show it clearly, the orders are as follows:

In Doctor No In the data

φ φ UP UP Def Art + NP G Ex Def Art + NP NP after Poss IR IR Ex G NP after Poss

It is clear that the orders among “definite article + NP”, “Existential”, “NP after Possessive”, “indefinite referential” and “generics” are different. However, except for “generics”, the average referential distances of these 4 types in the data are rather similar to those in Doctor No. See the table below:

In Doctor No In the data Def Art + NP 16.66 17.51

Ex 18.16 17.85

NP after Poss 18.34 18.20

IR 19.17 17.56

individual differences. Referential Distance 0.00 5.00 10.00 15.00 20.00 25.00 27 28 31 32 33 34 39 45 49 52 65 66 67 72 73 75 80 85 89 91 composition c la u se Def Art + NP

Figure 2 referential distance of definite article + NP

Referential Distance 0.00 5.00 10.00 15.00 20.00 25.00 27 28 31 32 33 34 39 45 49 52 65 66 67 72 73 75 80 85 89 91 composition c la u se Ex

Figure 3 referential distance of existential

Referential Distance 0.00 5.00 10.00 15.00 20.00 25.00 27 28 31 32 33 34 39 45 49 52 65 66 67 72 73 75 80 85 89 91 composition c la u se NP after Poss

Referential Distance 0.00 5.00 10.00 15.00 20.00 25.00 27 28 31 32 33 34 39 45 49 52 65 66 67 72 73 75 80 85 89 91 composition c la u se IR

Figure 5 referential distance of indefinite referential

From the graphs above, it is apparent that some types of reference cannot be found in several essays. For instance, “definite article + NP” cannot be found in one essay; “existential” cannot be found in two essays; and there are nine essays which have no “NP after possessive”. Furthermore, the deviation can be identified from the graphs: the referential distance of “definite article + NP” is smaller than 15 clauses in four essays; the referential distance of “existential” is extremely small in one essay, and it is smaller than 15 clauses in another two essays; the referential distance of “NP after possessive” in two essays are smaller than 15 clauses; and the referential distance of “indefinite referential” is smaller than 15 clauses in four essays. In spite of these individual deviations, the features of these four types of referents do not differ from the features in Doctor No greatly.

The other three types, “zero anaphora”, “unstressed pronoun” and “generics” show greater differences. See the table below.

In Doctor No In the data

φ 1.00 2.39

G 19.23 13.66

The referential distances of zero anaphora and unstressed pronoun in the data are dramatically bigger than those in Doctor No; and the referential distance of generics in the data is obviously smaller than in Doctor No. Since the differences in referential distances are the greatest here, these three types of referents can possibly be the culprits in undermining coherence. Therefore, each of these is first examined in more detail at the individual level as shown in the next three graphs. .

Referential Distance 0.00 5.00 10.00 15.00 20.00 25.00 27 28 31 32 33 34 39 45 49 52 65 66 67 72 73 75 80 85 89 91 composition cl au se ∮

Figure 6 referential distance of zero anaphora

Referential Distance 0.00 5.00 10.00 15.00 27 28 31 32 33 34 39 45 49 52 65 66 67 72 73 75 80 85 89 91 composition cl au se UP

Referential Distance 0.00 5.00 10.00 15.00 20.00 25.00 27 28 31 32 33 34 39 45 49 52 65 66 67 72 73 75 80 85 89 91 composition cl au se G

Figure 8 referential distance of generics

From the graphs above, it is clear that six essays do not have zero anaphora and a big referential distance of zero anaphora is caused by one exceptional essay in which the referential distance is 20 clauses. Other essays are very similar to Doctor No. This means that zero anaphora in the data is not a major problem. However, as Figure 7 and 8 show, there are striking differences in unstressed pronoun and generics, which may affect coherence. Moreover, there is variation among the writers, which implies that as a group, students may not know how to use these two types of referents.

participants do not know how to use these two types of referents correctly. Since these two types of referents may affect coherence, the causes of their different features of referential distance are discussed further in the discussion.

The follow-up study

Discussion

In this section, I will firstly discuss the result of the main study related to generics and unstressed pronoun. Then the result of the follow-up study is discussed. Finally we return to Kaplan’s different patterns of thought and end with implications for teaching English as foreign language and suggestions for Chinese EFL teachers. At last, some limitations of this study are addressed.

Main study

The referential distance of generics in Doctor No is 19.23 clauses while in the data it is 13.66 clauses, which is a difference of about 6 clauses. From the graph above, it was clear that there is a noticeable difference from one essay to the other. The 20 essays can be divided into three groups if the interval among three groups is set as 5 clauses. See the table below:

The referential distances of generics in 7 essays are between 15 clauses and 20 clauses, which are similar to the feature in Doctor No. Five essays’ referential distances are smaller than 10 clauses and the other 8 essays’ referential distances are between 10 clauses and 15 clauses. The causes of the smaller referential distances of generics in these two groups are discussed below.

If we investigate further, we see that the smaller referential distance is caused by repeated use of the same referent. There are three different kinds of referents that have this feature:

(1) “people”

Everyday peoply (people) want to know new things,

so people can use the information from various ways to learn.

(2) “information”

more and more people have a same idea: We need information.

So, we can inferred (infer) that information plays an important part in modern society.

Information tells us what we should do

Nowadays, information is playing an important role in modern society. As a tool of spreading culture and news,

information can help us know the newest things and stories in time.

(3) “sources of information”

There are many sources of information, such as the TV media and the Internet.

Internet is helpful for us to get information.

We can search all kinds of information on the Internet. It’s easy and reliable.

So I recommend the Internet to you for searching information.

There are many sources of information, such as newspaper, TV and website. We get to know the world with the information we got from them.

New paragraph

Usually I prefore (prefer) to believe the newspaper and TV than website.

Internet is a big sources (source) of information.

The first type of repeated referent only occurs once while the other two types occur many times. In some cases, it is possible to replace the generics by a pronoun. By doing this, the referential distance of type of generics will be enlarged. For instance:

Everyday peoply (people) want to know new things,

so people they can use the information from various ways to learn.

Now information is playing an (a) more and more important part in our product, study and daily life.

Information It tells us what we should do

Apparently, Chinese students like to use the same referent repeatedly, which is one cause of the smaller referential distance. One of the reasons may be that the particular essay topic, “sources of information” may lend itself to mentioning some generic types of referents several times, which reduces the referential distance of generics as a consequence. The smaller distance may make the essay awkward to read because an English reader may expect a full noun to be a new referent, but it is unlikely to affect a reader’s processing of referents because the referential distance is smaller.

Unstressed pronoun

In Doctor No, the referential distance of unstressed pronoun is 1.72 clauses, while in the data the referential distance is 6.33, which is about 5 clauses greater than it is in the novel. From the graph above, it is clear that there are individual differences. The 20 essays can be divided into three groups if the interval among three groups is set as 5 clauses. See the table below:

0 clauses to 4.99 clauses 5 clauses to 9.99 clauses 10 clauses to 15 clauses 40921145 4.83 40921127 5.93 40921131 10.80 40921165 3.63 40921128 6.35 40921172 10.83 40921166 2.91 40921132 6.38 40921173 3.33 40921133 6.00 40921180 4.29 40921134 7.78 40921139 6.36 40921149 7.25 40921152 9.29 40921167 6.25 40921175 8.00 40921185 5.23 40921189 5.60 40921191 5.54

referential distance is between 10 clauses and 15 clauses. Since the referential distance of unstressed pronoun in Doctor No is quite small (1.72 clauses), the unstressed pronoun in all essays have a potential effect on coherence.

When we look at the data more closely, two causes of a larger referential distance of unstressed pronoun are identified. The first one is switching point of view. For example:

So, we can inferred (infer) that information plays an important part in modern society. Let me take IT for example.

If you work in the IT field,

and you don’t get enough information,

you will soon fall behind others.

We can see and receive information everywhere, such as on the Internet, on the TV

programs.

I wonder what the world will be like

if the world has no information

I believe this is impossible

and I belive (believe) all people have the same idea.

There are many sources of information.

We can read newspapers, books, magazines, surf the Internet, listen to others and so on to

On newpapers (newspaper) and magazines there are usually latest news and adventisement (advertisement).

The Internet contains all kinds of information.

You can read it conveniently.

But some illegal information still exists,

we should get away from it.

The second cause of greater referential distance is that Chinese students like to use a pronoun to start a new clause and discuss old information at the end of the clause. For instance:

Internet is helpful for us to get information.

We can search all kinds of information on the Internet (old information).

Information plays an important in mordern (modern) society. Every day we need all kinds of information (old information).

Information plays an important part in mordern (modern) society.

There are many sources of information, such as newspaper, TV and website.

We get to know the world with the information we got from them (old information).

result in the switch of point of view. They are:

(1) Switches of point of view, such as from “we” to “you” is not strictly forbidden in Chinese writing. It can be illustrated by a Chinese essay (See Appendix) written by a senior high school student. He wrote the essay in the “National College Entrance Examination” and he got a full mark. In his essay, the point of view switched among “you”, “we” and “I”.

(2) Collectivism may make Chinese students tend to use “we” instead of “I”, which results in the switch between “we” and “I”. This characteristic has been discussed by a number of Chinese scholars. Cai (2001) mentions that Chinese students like to use “we” because of the influence of collectivism. He also supposes that Chinese students tend to use “we” instead of “I” to show politeness and objectiveness. It is very possible that this particular feature in Chinese writing is transferred to English writing.

In conclusion, the switch of point of view is due to characteristics of Chinese writing, Chinese culture and Chinese students’ writing strategy. These three factors cause Chinese students to choose various pronouns when they write English essays, which make the referential distance large.

In terms of eliciting old information by a new pronoun, two factors should be taken into account. The first one seems to be that Chinese students have not acquired how to use the passive voice; or they do not like to use the passive voice. In all 20 essays, passivization is almost nonexistent; therefore it was excluded form this study. If the students really do not know how to use the passive voice, their only choice is to use a pronoun to start a new sentence to discuss old information further, which leads to a large referential distance of unstressed pronoun. Then to investigate the phenomenon of eliciting old information by a new pronoun, the follow-up study is conducted, and the second factor is found.

The follow-up study

should be “placed as the end of the information unit” (Quirk et al. 1985:1361). In other words, old information is used to introduce new information because the theme can be continued and “the new information often needs to be stated more fully than the given” (Quirk et al. 1985:1361).

The follow-up study shows that in English old information indeed almost always occurs at the beginning of the clause in English. However, in Chinese the positions of “theme/old information” and “focus/new information” are not strictly distinct. It implies that “focus/new information” can occur both at the beginning and at the end of one clause in Chinese. Therefore, if Chinese students do not mind putting new information at the beginning of the clause, they will put a new pronoun at the beginning of the clause and the old information at the end of the clause. Because Chinese writers put unstressed pronouns, which have not occurred in the previous clause, at the beginning of the following clause, the referential distance of unstressed pronoun is greater than in English texts written by native speakers.

To summarize, it is possible that Chinese students do not know how to use the passive voice or they do not like to use it. Furthermore, since Chinese is not strict about where the “theme/old information” and “focus/new information” should be placed, Chinese students may use new pronouns to introduce old information. Therefore the referential distance of unstressed pronoun is large.

The pattern of thought

information to start a new clause, which discontinues the theme. It is possible that this feature makes readers feel that the English writings of Chinese students are not “straight”. The different patterns of thought may not influence the organization of an essay as a whole, but it may affect the positions of “theme/old information” and “focus/new information”. In other words, it may influence structures of clauses. Therefore, Chinese ESL teachers need to draw students’ attention to the positions of old and new information and guide them to reduce the referential distance of unstressed pronoun. Furthermore, teachers also need to guide students to avoid repeated information and switches among “we”, “I”, and “you”. One example about how to reduce referential distance of unstressed pronoun is shown below (the revised parts are bold):

Clauses Paragraph 1 Clauses Paragraph 1

1

We (UP, 20) all know in this centry,

1

As is known to all in this century,

2

information (G, 20) plays an important part (IR, 20) in modern society.

2 information (G, 20) plays an important part (IR, 20) in modern society.

3

Everyday peoply (G, 20) want to know new things (IR, 20),

3 Everyday peoply (G, 20) want to know new things (IR, 20),

4

so people (G, 1) can use the information (Def Art+NP, 2) from various ways to learn.

5

It (UP, 20) is important to keep up with the tend of the modern life (Def Art+NP, 20).

(Delete) to keep up with the trend of the modern life (Def Art+NP, 20).

Paragraph 2 Paragraph 2

6

There are many sources of information (Ex, 20).

5 There are many sources of information (Ex, 20).

7

We (UP, 6) can get it (UP, 3) from Newspapers, TVs, and Internet etc.

6 For instance, the sources (Der Art + NP, 1) can be Newspapers (G, 20), TVs (G, 20), and Internet (G, 20) etc.

8

Each way (IR, 1) has its benefits (NP after Poss, 20),

7 Each source (IR, 1) has its benefits (NP after Poss, 20),

9

We (UP, 20) can choose which one (IR,1) is best for ourselves,

8 We (UP,20) can choose which one (IR,1) is best for ourselves,

10

and (φ, 1) get information (IR, 3) from the way (Def Art+NP, 1).

9 and (φ, 1) get information (IR, 5) from the source (Def Art+NP, 1).

11

Or we (UP, 1) can choose all the ways (Def Art+NP, 4) to learn and know the information (Def Art+NP, 1).

10 Or we (UP, 1) can choose all the sources (Def Art+NP, 4) to learn and know the information (Def Art+NP, 1).

12

But there are too many sources of information (Ex, 6).

13 (φ, 2)Facing them (UP, 1), 12 (φ, 1)Facing them (UP, 1),

14

we (UP, 1) should know how to choose them (UP, 2).

13 we (UP, 1) should know how to choose them (UP, 2).

15

Find what information (IR, 20) attract you (UP, 20)

14 We (UP, 1) should find what information (IR,20) attract us 16 and choose it (UP, 1). 15 (φ, 1) and choose it (UP, 1).

Paragraph 3 Paragraph 3

17

In my opinion, for there are too many different sources of information (Ex, 20),

16 In my opinion, for there are too many different sources of information (Ex, 20),

18

We (UP, 4) should choose the most useful one (Def Art+NP, 20) and most attractive one (Def Art+NP, 20).

17 We (UP, 2) should choose the most useful one (Def Art+NP, 20) and most attractive one (Def Art+NP, 20).

19

Form the information, we (UP, 1) can learn a lot (IR, 20).

18 Form the information, we (UP, 1) can learn a lot (IR, 20).

20

And don’t choose the rubbish information (Def Art + NP, 20)

19 And we (UP, 1) shouldn’t choose the rubbish information (Def Art + NP, 20)

21

receive them (UP, 1) is just a way (IR, 20) of wasting time (IR, 20).

22

Don’t waste your time (NP after Poss,20),

21 We (UP,2) would better not waste our time (NP after Poss,20)

23

and make full use of your time (NP after Poss, 1) to learn useful things (IR, 20) and to do wonderful things (IR, 20).

22 and make full use of our time (NP after Poss, 1) to learn useful things (IR, 20) and to do wonderful things (IR, 20).

The revision includes replacing a repeated noun by a pronoun, deleting repeated information, adding missing information, changing positions of “theme/old information” and making the point of view consistent. After revision, the referential distance of zero anaphora, unstressed pronoun and generics are much more similar to the novel’s features. See the table below:

In Doctor No Types Before revision After revision

1.00 φ 1.50 1.00 1.27 UP 5.60 2.47 16.66 Def Art+NP 11.00 9.71 18.34 NP after Poss 13.67 13.67 19.23 G 13.67 20 18.16 Ex 15.33 20 19.17 IR 18.11 16.91

attention to repeated information, switches of point of view and the positions of old and new information.

Limitations of this study

The current study has several limitations, which may affect the result of this study.

The first one is about the methodology of the study. The current study tries to replicate the methodology in Brown’s study (1983). However, since Brown didn’t describe the methodology thoroughly and comprehensively, it is possible that the methodology of this study is slightly different from what Brown did. The consequence of this difference may result in slightly different referential distances. But it is believed that there is no dramatic difference.

The second one is about different genres between the data in Brown’s study and the data in the current study. In Brown’s study, the data is a novel, which is a narrative, while in the current study the data is a mixed genre of expository and argumentative writing. The different genres will have probably attributed to the different features of referential distance in Brown’s study and in the current study but it is not clear to what extent.

The third limitation is concerned with the method to find out differences of writing styles between Chinese and English in the discussion part. In this method, the old information in the previous clause included both information that had appeared and could be inferred. If inferred information is not taken into account, the result will be different. Therefore, how to define the change of “theme” needs more discussion.

coherence. Due to time limit, only referential distance is studied. Although it can be a good example to show differences of referential coherence between English essays written by native speakers and by Chinese speakers, it is better to employ all methods to study referential coherence thoroughly.

Conclusion

The current study investigates the possible causes of lack of coherence in English texts written by Chinese learners. The study was motivated by the fact that there are no studies focusing on coherence from an ESL perspective, probably because coherence is so difficult to measure. After discussing the differences between coherence and cohesion, I decided to measure coherence in English essays written by Chinese speakers by means of referential coherence. The study employs a method called “referential distance” proposed by Givón (1983) and used in Brown’s study (1983) to measure referential coherence in 20 essays which are written by 20 Chinese college students whose English proficiency level is the second level of CET. Then the referential distances in the essays are compared with those found by Brown to find out specific features in the essays that deviate from the native speaker text. The results show that there are several different features of referential distance in the essays written by Chinese students.

point of view among “we”, “I” and “you” and starting a new clause with a new pronoun to discuss old information later in the clause. The factors behind switching the point of view are assumed to be as follows: switching point of view is accepted in Chinese writing, which may be transferred to English writing; the effect of collectivism makes Chinese students prefer “we” to “I”; and Chinese students like to use “you” as a special writing strategy. The factors behind starting a new clause with a new pronoun are assumed to be due to the fact that: Chinese students may not know how to use the passive voice or they do not like to use it; and in Chinese writing, it is possible to put new information at the beginning of a clause, another factor that may be transferred to English writing.

In conclusion, unstressed pronouns may cause problems in coherence in English essays written by Chinese students. The different feature is due to Chinese writing style and Chinese culture. Possibly, it also makes the essays “spiral” in terms of clause structure. Therefore, ESL teachers in Chinese should teach their students to avoid unnecessary switches in point of view and make sure old information needs to come before new information. Finally, repeated use of the same referent should be avoided as well.

Reference

Braddock, R. (1974). The frequency of placement of topic sentences in expository prose.

Research in the teaching of English, 8, 287-302.

Brown, C. (1983) Topic continuity in written English narrative. In W. Bublitz, U, Lenk & E, Ventola (Eds), Coherence in Spoken and Witten Discourse (pp. 317-341). Amsterdam/Philadelphia: John Benjamins Publishing Company.

Cirilo, R. (1981). Referential coherence and text structure in story comprehension. Journal of

Verbal Learning and Verbal Behavior, 20, 358-367.

Dirven, R. & Verspoor, M. (2004). Cognitive Exploration of Language and Linguistic. Amsterdam/Philadelphia: John Benjamins Publishing Company.

Givón, T. (1981). Typology and functional domains. Studies in Language, 5(2), 163-194. Givón, T. (1983). Topic continuity in discourse: an introduction. In Givón, T (Eds), Topic

Continuity In Discourse: A Quantitative Cross-Language Study (pp. 5-41). Amsterdam/Philadelphia: John Benjamins Publishing Company.

Givón, T. (1983). Topic continuity in spoken English. In Givón, T (Eds), Topic Continuity In

Discourse: A Quantitative Cross-Language Study (pp. 347-361). Amsterdam/Philadelphia: John Benjamins Publishing Company.

Givón, T. (1983). Topic continuity and word-order pragmatics in Ute. In Givón, T (Eds),

Topic Continuity In Discourse: A Quantitative Cross-Language Study (pp. 145-213). Amsterdam/Philadelphia: John Benjamins Publishing Company.

Givón, T. (1992). The grammar of referential coherence as mental processing instructions.