Studies of Writing Involving Different Proficiency Groups

writers’ use of cohesive devices.

2.3.1 Studies of Writing Involving Different Proficiency Groups

Pritchard (1980) conducted a research on lexical and grammatical ties in 44 eleventh graders and discovered that the average use or frequency of total lexical or grammatical ties did not differentiate the good from the poor essays. In addition to Pritchard's study (1980), other scholars like Witte and Faigley (1981) also explored the internal characteristics that distinguished high-rated essays from low-rated ones.

With this purpose in mind, the researchers collected 90 freshmen essays from the University of Texas. After all the essays were graded holistically by two raters, ten essays were chosen. Among these ten essays, five represented the highest score and five represented the lowest. Later, Witte and Faigley analyzed these selected essays by using two taxonomies offered by Halliday and Hasan (1976). The first taxonomy concerned the employment of grammatical and lexical cohesive ties in the written discourse while the second taxonomy involved the textual distance between the presupposed and presupposing elements of a given tie. From the analysis of the student essays, the high score essays were denser in cohesion than the low score essays. Further, high-rated writers used a substantially higher relative percentage of immediate³ and mediated cohesive ties whereas low-rated writers used more mediated-remote and remote ties. As for the control of cohesive ties, two-thirds of the cohesive ties were subsumed under the category of lexical cohesion. And there was a comparatively larger percentage of repetitions used by low-rated writers although over 50% of the total lexical ties were classified as the same item category in all student essays. With regard to the occurrence of collocation, high-rated writers used

3 Immediate cohesive ties refer to presupposed and presupposing cohesive ties occur in adjacent T-units; mediated cohesive ties means that between the presupposed cohesive ties and presupposing cohesive ties still exist one or more T-units that including the same cohesive ties. Therefore, both immediate and mediated cohesive ties join cohesive items in adjacent T-units. On the other hand, remote ties occur when the presupposed and presupposing ties are separated by one or more intervening T-units. Last, ties which are both mediated and remote are called mediated-remote.

24

more lexical collocations when compared with low-rated writers. The differences in the textual distance of cohesive ties and in the number of cohesive ties indicated that high-rated writers tended to establish stronger cohesive bonds than the low-rated writers. Immediate adjacent ties displayed high-rated writers’ lexical or conceptual capability to expand or elaborate concepts in their essays. This stood in contrast to low-rated writers whose lack of sufficient vocabulary to extend the ideas they introduced in their essays resulted in their writing containing superfluous redundancy.

This analysis further pointed to the fact that low-rated writers’ cognitive skills were still under development as they failed to perceive and articulate abstract ideas or refer to instances beyond their immediate context. Based on the above-mentioned analysis, the researchers suggested that analysis of cohesion may serve as a tool to distinguish between stages of writing development for it measured more sophisticated aspects of language development than error analyses and syntactic analyses do. With regard to the writing quality, no statistical tools were applied to compare good and poor writing.

Instead, Witte and Faigley only implied that cohesion seemed to be an important property of writing although no evidence suggested that large or small numbers of ties in themselves would affect writing quality. Besides, the small number of subjects analyzed also rendered their conclusion a tentative one.

McCulley (1985) also carried out similar research exploring the relationship between cohesion, coherence and writing quality in persuasive writings composed by 17-year-olds and discovered that some cohesive devices, whether grammatical or lexical ones, seemed to influence writing quality. Despite the similar results with Witte and Faigley (1981), McCulley’s (1985) research still could not provide solid statistical evidence that counts of individual cohesive ties correlated with writing quality and sufficed to analyze, evaluate and teach compositions.

In view of the deficiencies in McCulley’s study, Neuner (1987) replicated the

25

study by adding new variables, using a more complete list of types of ties, and examining the relative distances between the presupposed and presupposing, the mean length of cohesive chains and the diversity and maturity of the vocabulary within chains. Similar to Pritchard, and Witte & Faigley’s studies, participants were required to write on a single topic. Among the collection of 600 college freshman papers, twenty good and poor essays, respectively, were randomly selected for analysis. The results of the analysis confirmed Pritchard's (1980) findings. That is, a simple calculation of the cohesive ties did not discriminate good from poor writing and the percentages of types of ties do not vary greatly from good to poor essays. Besides, the average words per tie in each essay and the average number of lexical ties per essay also failed to show statistically significant differences between the two groups of essays. However, cohesive chains, which referred to a string of semantically related lexical collocations, reiterations, synonyms, super-ordinates and their reference pronouns were found to show differences in good and poor essays. In good essays, cohesive chains were able to extend over greater distances and involved greater proportions of the whole text which contributed to the texture of the text (Halliday&

Hasan, 1976). Finally, a Standard Frequency Index (SFI)⁴ was employed to assess the maturity of word use in good and poor essays. The results showed that good writers used significantly more words, whether across all chains or in each individual chain, than poor writers did. From Neuner’s findings, we can see that cohesive ties are not the discriminating factor for good or poor writing. Rather, longer chains, greater variety and maturity of words are the primary features of good writing.

Ferris (1994) adopted a multidimensional approach and examined 160 ESL

4 This SFI measurement is one of the computational analysis tools, which is meant to measure the frequency of words appearing in the language and it is logarithmic with values ranging from 90 to below 10. A word with an SFI 90 is likely to occur once in every 10 words while a word with an SFI 80 means a word will only occur in every 100 words.

26

students’ written samples. In her study, both lexical and syntactic features of ESL students’ writing—twenty-eight variables in total—were identified and counted.

These variables served as an index to discriminate between two proficiency groups.

Moreover, these variables were correlated with students’ writing proficiency and the holistic scores assigned by the raters to see whether there was correlation between them. The result of her study showed that higher proficiency writers used a greater variety of cohesive devices such as lexical and referential ones while lower proficiency writers relied more on repetition to promote textual cohesion. That is, the more advanced students were more flexible in the use of lexical and referential cohesive devices such as synonymy, antonymy, definite article reference and deictic reference. Moreover, the fact that various syntactic structures were applied in advanced student writing indicated that students of higher proficiency level were more sensitive to the pragmatic use of the language, which resulted in promoting the overall textual coherence of those students’ writing (Bardovi-Harlig, 1990). In addition, from the result of the multiple regression analysis, text variables employed to distinguish two proficiency groups also correlated positively with holistic scores.

Chang (1995) conducted a quantitative and qualitative study of freshmen English compositions. First, the frequency of occurrence of the cohesive devices in 241 compositions was calculated by a computer program. After that, he randomly selected twenty compositions from students’ placement test at the beginning of the data collection and a random selection of ten compositions was done again among the twenty samples to further observe whether students had made progress over the time of instruction. Later, students’ compositions were graded holistically by six Chinese-born instructors of the Freshman English Writing course, in terms of the content, organization, diction, grammar, spelling and punctuation. According to the scores assigned by the graders, he classified these compositions into three groups: HI,

27

MID, and LOW. The analysis of the data was discourse-oriented because all the students’ errors in the composition were examined from the perspective of their discourse function in the whole paragraph rather than in a single sentence. The result of the study showed that the definite article, demonstratives and deictics constituted the most frequently employed category of grammatical cohesive devices. In the aspect of lexical cohesive devices, the frequency and percentage of each cohesive tie were compared, with repetition showing the highest frequency and general word the lowest.

However, no statistical tools were implemented to justify whether or not the differences in percentage between the high and low groups were significant. Moreover, the examination of reiteration appeared to be oversimplified because only one sample was selected to reach the conclusion that “these writers [the more proficient writers in the HI group] employ more and longer reiteration chains and their use of the reiteration ties shows greater variety” (Chang, 1995, p. 44). For any future study, an investigation of cohesive errors could be conducted to measure which area is the most problematic for EFL writers.

In response to the room for improvement in Chang’s study (1995), Liang (1997) conducted a replicated study by analyzing cohesion in freshmen English compositions quantitatively and qualitatively. About 61 freshmen English compositions were graded and divided into high and low groups. Later, twenty samples representing high and low groups respectively were randomly selected for both quantitative and qualitative analysis. A quantitative analysis was implemented to count the grammatical and lexical cohesive devices and R-chains⁵. In addition, t-test was used to reconfirm the difference in means between the two groups. Meanwhile, a qualitative analysis was used to illustrate the most frequent words/phrases and the errors committed by the

5 The term R-chain was coined by Liang (1997) and it referred to words of the following

semantic/lexical relationship: Repetition, (Near-) Synonym, Hyponym or General Words, were tied together.

28

subjects. The results revealed that reference was employed most frequently, followed by conjunction whereas repetition possessed the highest frequency in the category of lexical cohesion. However, in comparing the use of cohesive devices in writing by high and low groups, only General Words and the number of R-chains were found to show significant differences. Further, despite the fact that high group writers had higher cohesion scores⁶ than those in the low group, the discrepancy was insignificant. Overall, the results of the study revealed that “there is a positive but low correlation between the number of cohesive devices employed and the writing quality of the samples.” Although the result obtained was very insightful for teaching writing, the researcher did not explore how high and low groups differ in the application of lexical devices.

Chou (2000) intended to explicate the relationship between lexical cohesion and the writing quality of EFL students’ essays regardless of genre. Hence, four essays representing two genres—argumentation and description— were assigned to four juniors of the English department of Chinese Culture University. Next, he adopted Cameron’s model⁶ to calculate the lexical cohesive scores and adopted Liang’s R-chain (1997) to identify and count the lexical ties. The results he obtained showed that higher density in lexical cohesion did not guarantee a better piece of writing.

Besides, a further investigation into the students’ lexical cohesion revealed that more repetition was found in the two low-rated essays than in the one high-rated essay.

Chou claimed in the conclusion that this study showed similar results arrived at by Witte and Faigley (1984) that “cohesion-based” distinctions between the quality of high-rated and low-rated essays may be questionable and the results also confirmed Tierney and Mosenthal’s (1983) argument that cohesion analysis cannot be a predictor

6 Cameron calculated cohesive scores by first dividing a text into T-units, which is one independent clause with all subordinate elements attached to it and by dividing the recorded number intersentential cohesive ties by the number of T-units in a text.

29

of a text’s coherence and that “a cohesion index is causally unrelated to a text’s coherence.” Despite the fact that these results seemed to coincide with those of previous studies, we should be cautious of the validity and reliability of the study in that the analysis involved only a small number of samples and is therefore less persuasive and convincing.

Chen (2003) conducted a quantitative and qualitative study to analyze the lexical cohesion in senior high school students’ compositions. In her analysis of students’ writing between high and low groups, she used six categories to examine whether students had different uses of them (e.g., repetition, synonym, hyponym, antonym, superordinate, and collocation). And her findings showed that no significant differences between high and low groups although repetition was the most common tie used in both groups. In addition, from the observation of means she found that high-level students were better with the use of synonyms and collocations and that they used more synonyms and less general words in the compositions than low-level students. Finally, some teaching implications were provided to raise the teacher and student’s awareness of lexical cohesion.

Liu and Braine (2005) investigated the use of cohesive devices in 50 argumentative compositions created by Chinese undergraduate non-English majors by applying Halliday and Hasan’s (1976) framework for analysis. A diversity of cohesive devices were found in students’ writing, among which lexical cohesive devices formed the largest percentage of the total number of cohesive devices, followed by references and conjunctives. The quality of writing was significantly correlated with the total number of cohesive devices and it was even highly correlated to lexical devices. Last, certain problems with reference and lexical cohesive devices were identified. They included the shifted use of pronouns, omission or misuse of the definite article, underuse of comparatives and overuse of the phrase ‘more and more’,

30

restricted choice of lexical items and incorrect use of collocation.

Studies of student writing were not restricted to comparison within the same age group. Along the line of the research on the cohesion system that characterized the good and poor writing, some researchers conducted the analysis of the cohesive devices in student writing by different age groups. In Cherry and Cooper ‘s study (1980), they examined the student writer’s performance in both the distribution of cohesive ties and the average distance of ties in writings by grades four, eight, twelve, and college students. The result of their study revealed a positive correlation between writer’s maturity and the use of lexis in writing and a negative correlation between writer’s maturity and the use of grammatical cohesive ties.