Summary

Cohesion theory and the framework of cohesive devices proposed by Halliday and Hasan (1976) were recognized and adopted for the analysis of cohesion in writing through the years. Even though the model was not without flaws and was modified by Stostky (1983) and Hasan (1985), it still served a pivotal role in the discourse analysis.

Further, Stostky’s framework was limited because she only included lexical cohesive devices in her framework and did not provide clear criteria for the judgment of collocation. As for Hasan’s model, conjunctive elements were not included in the three kinds of relationship—co-reference, co-classification and co-extension. In view of this, the study will mainly adopt Halliday and Hasan’s (1976) framework but modify the model according to Stostky (1983) and Hasan’s (1985) opinions.

Previous studies involving the performance of different proficiency groups in cohesion revealed that there was no consistent conclusion in whether analysis of cohesion may serve as a tool to distinguish good and poor writers (Chen, 2003; Chou, 2000; Ferris, 1994; Liu & Braine, 2005; McCulley, 1985; Prichard, 1980; Tierney &

Mosenthal, 1983; Witte & Faigley, 1981). In spite of this, it was proved that high and low proficiency groups differed in the use of cohesive devices. The high score essays were denser in cohesion and used a greater variety of cohesive devices (Ferris, 1994;

40

McCulley 1985; Witte & Faigley, 1981). Secondly, high proficiency writers were better with the use of synonym, collocation and used less general words than low proficiency ones (Chen, 2003; Crowhurst, 1987; Jin, 2001). Additionally, high proficiency writers used significantly more personal reference and temporal conjunctions (Jin, 2001). Contrary to high proficiency groups, lower proficiency groups relied more heavily on repetition to promote textual cohesion (Chou, 2000;

Ferris 1994) and lacked sufficient vocabulary to extend the ideas they introduced in their essays, which ended up containing superfluous redundancy in the writing (Witte

& Faigley, 1981). Concerning the quantitative analysis of cohesive devices, it was found that reference was employed most frequently, followed by conjunction (Chang, 1995; Liang, 1997; Liu & Braine, 2005; Norment, 1982). In addition, lexical cohesive devices formed the largest percentage of the total number of cohesive devices (Liu &

Braine, 2005; Witte & Faigley, 1981). To be more specific, repetition showed the highest frequency in lexical cohesive devices (Chang, 1995; Chen 2003; Liang, 1997).

With regard to errors in cohesive devices, shift use of pronouns, omission or misuse of definite article, underuse of comparatives, over use of the phrase “more and more“, and restricted choices of lexical items were found in Liu and Braine’s (2005) study.

Previous studies across different genres showed that no significant differences in cohesion devices were found across genres (Jin, 2001; Norment, 1982, 2002). Besides, more sentences were used in narrative mode than in expository mode (Norment, 1982). With regard to cohesive devices, more references (pronominals, demonstratives) and conjunctions (temporal conjunctives) were used in narration than in exposition or in argumentation (Crowhurst, 1987; Norment, 1982, 2002; Tierney &

Mosenthal, 1983). Further, similar frequencies of usage in conjunctions and references occurred in exposition and narration (Norment, 2002). On the evidence of previous studies, it seems that writers of different levels and different modes of

41

writing may influence the pattern and frequency of cohesive devices in written discourse.

42

CHAPTER THREE METHOD

This chapter presents the method of the present study. Thus below, the participants, instruments, pilot study and procedures of the formal study are presented in turn.

The purpose of the study is to investigate whether two levels of students, namely high and low proficiency, performed differently in narrative and expository writings in terms of the use of cohesive devices. The participants in this study were divided into high and low proficiency groups and both groups were requested to produce two modes of writing. After that, the number of cohesive devices in the writing of two proficiency levels was compared to see whether writers of different levels will perform differently with respect to writing proficiency or genre difference. The design of the study is both quantitative and qualitative in nature. Hence, in addition to the quantitative comparison, a qualitative analysis of the cohesive errors will be presented as well to find out the problematic areas of students' use of cohesive ties.

3.1 Participants

Participants in this study were freshmen and seniors from National Taiwan Normal University. The freshmen were non-English majors enrolled on the Freshman English course while the seniors were English majors. The participants numbered 60 in total and the number of participants representing each of the two levels was equal.

As Mandarin speakers, all of them had learned English for at least six years. To divide the participants into high and low proficiency groups, a screening selection based on their writing scores was done to validate the grouping,

3.2 Instruments

Three instruments were used to analyze the data of the study: (a) the scoring guide of the TOEFL Test of Written English. The scoring scheme provided by TOEFL

43

involved a six-category analysis of students’ writing skills, rating from 6-0. Each of the six categories includes the following components: suitable word usage, content organization, examples/evidence, formal writing style, correct grammar and sentence variety (See Appendix A); (b) a combination model of cohesive devices adopting Halliday and Hasan’s (1976), Stotsky’s (1983) and Hasan (1985)’s framework was employed to identify cohesive elements occurring in the written text that the participants composed and (c) Independent-Samples T-test. T-test was run on one or more independent groups; to specify the grouping variables and check for any difference between or among the groups. In the study, the number of cohesive devices will be regarded as continuous variables, and thus assumed to be qualified for a T-test analysis.

3.3. Pilot Study

The purpose of the pilot study was to assure the feasibility of the research and to foresee any possible problems in the future study. There were 14 participants in the pilot study, and the recruited members included freshmen as well as junior English majors. The number of the participants from the two groups was equal and all of them voluntarily participated in this study. All of them had learned English for at least six years and Mandarin was their mother tongue. In the pilot study, participants were required to produce two genres of writing—narration and exposition. Before they started to write, some instructions were given and participants had to write according to the instructions.

In the pilot study, 14 pieces of writing samples in each mode were collected. The highest score was 5 and the lowest score was 3. Therefore, those who scored highest were grouped as high proficiency writers while those who scored the lowest were grouped as low proficiency writers. To answer the research questions, all cohesive devices in both high and low proficiency level writers’ writing were calculated and

44

analyzed accordingly. To investigate whether high and low proficiency writers would perform differently in the narrative and expository essays, the percentage of the each cohesive item was calculated and the statistical tool, t-test, was implemented to see whether there was difference in the cohesive pattern in terms of different proficiency groups or genres. Later, a more detailed and qualitative analysis of the cohesive errors in the participants’ writing was conducted from the perspective of the discourse. The cohesive items examined in the pilot study were reference, substitution, ellipsis, conjunction, repetition, synonym, antonym, superordinate, and meronym. However, in the follow-up study, meronym will be deleted from the framework not only for its rare occurrence but for ease of comparison between grammatical and lexical cohesive devices.

The result of the pilot study was discussed from four aspects. First, the proficiency level was set as a variable in the narrative writing. The result showed that high and low proficiency writers showed no differences in their use of grammatical cohesive devices but significant differences in their use of lexical cohesive devices.

Besides, additive and repetition exhibited the highest frequency in high and low proficiency writers’ essays. Moreover, high and low writers showed significant differences in the use of referent elements for high group writers used more personal references while low group writers used more demonstrative references. Second, the proficiency level was still set as a variable but the genre was changed to the expository writing. The result showed that high and low students had no difference in grammatical and lexical cohesive devices. Nevertheless, additive and repetition still had the highest frequency in two groups.

Thirdly, the genre was set as a variable. High proficiency writers showed no differences in their employment of grammatical and lexical cohesive devices in narration and exposition. However, there were significant differences in the use of

45

reference elements. Personal reference was weighed highest in both genres; but still personal references had higher frequency in narration than in exposition. In addition, repetition and additive had the highest proportion in both genres. Lastly, low proficiency writers showed no difference in the employment of grammatical cohesive devices but significantly difference in the lexical cohesive devices. Different from the high group, there was no significant difference in the use of reference elements in two genres. However, repetition and additive still has the highest proportion in the low groups’ writing in two genres.

Through qualitative analysis, additive conjunction was widely used in high and low group's narrative writing. Further, there was a higher percentage of adversative conjunction such as however in exposition than in narration. As for high proficiency students' errors, it was found that the number of repeated words may not guarantee good writing. Writers may use the inappropriate description repeatedly in their writings. In addition to the wrong choice of words, the tense was also a problem.

Writers may choose the correct word but choose the wrong tense and in so doing make their writing incoherent. Besides, writers may have extra addition of cohesive devices. Writers tended to add the definite article the before a noun though in actual fact no referent could be found in the previous passage.

With regard to the low group's errors, there was lack of reference the to refer to something that was mentioned before or there were extra addition of references the to refer to something that was not mentioned before. Besides, wrong choices of pronouns could also be found. Writers may use the singular pronoun to refer to the plural objects in the preceding discourse. Further, the repeated cohesive items made the article sound redundant.

In response to the results of the pilot study, several items were modified. First, even though the classification of the grouping was based on the participants’ writing

46

scores, from the collected data, it was found that there was a short gap because most writers were graded 3-5 points, which made the result of the pilot not consistent.

Therefore, in the formal study, the participants were changed. Freshmen non-English majors were compared with senior English majors, in the hope of collecting more samples with greater distinction. Second, the lexical item, meronym, was deleted because of its rare occurrence in the pilot study. Further, after the deletion of meronym from the analytic framework, we obtained four grammatical and four lexical cohesive devices. The equal number of cohesive devices in the two categories would make it easier to compare these two categories.

3.4 Procedures of the Formal Study

This section presents the procedures of the formal study. In what follows, the background of the participants and data collection, coding scheme of cohesive devices and data analysis.

3.4.1 Participants and Data Collection

Among the 60 participants, half of them were senior English majors while the other half were freshmen non-English majors. In each group, the participants were requested to produce narrative or expository writing. To make the study more feasible, half of the senior English majors (15 students) produced narrative essays whereas the remaining fifteen students produced expository essays. This applied to the freshmen group, too.

The title of the narrative writing is ‘A Special Day to Remember’ while the title of the expository writing is ‘Cell Phone’. Before requesting participants to write on the two topics, writing prompts were provided with some clues about what to write in their essays. In the narrative writing prompt, for example, the prompt specified the genre of the writing first. Next, it offered some questions for brainstorming some ideas of writing the topic and some instructions for emphasizing the important

47

elements that a narrative writing should include in describing a whole event such as setting and character. (For more detailed description of the two writing prompts, see Appendix B.) The word limit for each essay was 200 words per essay and the time limit was sixty minutes for each piece of writing. During the time when participants were writing, no dictionary was allowed to be used or referred to.

After all the student samples were collected, they were divided into two piles, of narrative and expository writing respectively, and were graded holistically by two raters⁹ according to the criteria set by TOEFL (Test of English as a Foreign Language) writing test, ranging from 6-0 (See Appendix A). In both narrative and expository writing, those who scored more than 4 points were classified as the high proficiency group and those below 3 points were classified as the low proficiency group. In addition, only 9 narrative samples were graded above 4 points, but 14 expository examples were graded above 4 points. Thus, a random selection of 9 expository samples was done and a total of 18 narrative and expository samples were classified as the high proficiency group. Similarly, 21 narrative samples and 16 expository samples were graded under 3 points (See Appendix C). Therefore, a random selection was done to select 10 narrative and 10 expository samples as the low proficiency group. Therefore, a total of 38 samples would be examined in the study (See Appendix D & E).

3.4.2. Coding Scheme of Cohesive Devices

In order to address the four research questions, all cohesive devices in both high and low proficiency groups’ writing were calculated and analyzed accordingly. The coding scheme for the calculation of cohesive devices includes two major parts—

grammatical and lexical cohesive devices. The grammatical cohesive devices include

9 The two raters are experienced raters teaching writing for many years in National Taiwan Normal University. One of them graded the narrative writing while the other one graded the expository writing.

Both of them assigned the scores based on the same assessment criteria set by TOEFL writing test.

48

10 In Halliday and Hasan's model (1976), the first and second personal pronouns were not counted in the cohesive framework. However, in this study, both are counted as cohesive devices because they are indispensable in the narrative writing.

49

iii. Causal

so therefore consequently on account of this in consequence with this in mind for because it follows arising out of this

to this end in that case under the circumstances in this respect apart from this in other respects

iv. Temporal

and then next just then before that in the end first…then at first/originally finally/now at once soon next time next day meanwhile until then at this moment up to now at this point from now on

As for the lexical cohesive devices, they include repetition, synonym, antonym and superordinate. The lexical cohesive framework mainly followed Halliday and Hasan’s framework (1973), but some adjustments were made according to Stotsky’s (1983) and Hasan’s (1985) model. Originally, in Halliday and Hasan’s framework;

repetition, synonym, superordinate, general word and collocation were included. In this study, however, collocation was not included because of the lack of clear criteria for collocational judgment (Halliday & Hasan, 1976, Hasan 1985). Even though Stotsky (1983) tried to reformulate the category of collocation by counting words that tend to appear together in texts with similar topics, she was still not able to propose a set of clear-cut criteria. The judgment of collocational cohesion still depended heavily on the reader’s reading experience. Secondly, antonym was added to the analytic framework. Stotsky (1983) doubted what linguistic principles could be employed to classify antonym as collocation, and synonym under the category of reiteration. In Hasan’s 1985 model, she also added antonym to the reiteration and disregarded collocation as cohesive devices.

Thirdly, superordinate and general words were classified as one category.

50

According to Halliday and Hasan’s definition of superordinate, the meaning of the superordinate “includes that of the earlier one” or “any item that dominates the earlier one in the lexical taxonomy” (Halliday & Hasan, 1987). And the difference between these two is “degree of generality”. Therefore, it is sometimes difficult to judge whether it is a superordinate or a general word. In view of this difficulty in judging the degree of generality of words, in this research, the entry “general word” is subsumed under the category of superordinate. Further, the semantic relationship of subordinate was categorized into the hierarchical category—superordinate. From Stotsky’s critique, presenting the general idea first followed by the discussion of examples is one of the features of essay writings. Therefore, the inclusion of a cohesive tie by a subordinate item that follows a superordinate one can verify whether different genres will show differences in the writers’ employment of cohesive devices.

Based on the adjustments above, lexical cohesive devices, like grammatical cohesive devices, consist of four categories—repetition, synonym, antonym and superordinate.

1. Repetition: In this category, simple repetition of the same word, words with different grammatical inflections and derivatives are counted as one instance of repetition. For example, walk, walked and walking are counted as one lexical item and are classified as the same word.

2. Synonym: Synonyms are different phonological words which have the same or very similar meanings (Saeed, 1997).

3. Antonym: Words that have the opposite meaning are classified as antonyms like cold/hot, high/low, etc.

4. Superordinate: Superordinate is a relation held between a general class and its sub-classes. In the lexical taxonomy, there are degrees of generality along the scale of the word meaning. Hence, words which include the meaning of the earlier

51

ones in the text are classified as superordinate. At the highest end of the scale of generality are words like thing, human, creature. With the companion of the reference item the, they can anaphorically refer to the previous text, which makes them function like a grammatical cohesive device. Therefore, words with the most general meaning have the characteristics of both grammatical and lexical cohesive function and are called general word in Hasan and Halliday’s scheme. In this study, the so-called general word will be incorporated into the category of superordinate.

Aside from this, originally in Halliday and Hasan’s scheme, any item whose meaning anaphorically includes the earlier one is classified as a superordinate. That is, a cohesive tie is created by subordinate plus superordinate. Nevertheless, whether the exchange of the position between superordinate and subordinate constructs a cohesive relation is not discussed (Stotsky, 1983). In this study, it will be counted as cohesive devices as well.

3.4.3 Data Analysis

After the calculation of cohesive devices in high and low proficiency writers’

samples, both quantitative and qualitative analyses were indispensable for answering

samples, both quantitative and qualitative analyses were indispensable for answering