Process Writing and Computer Correction: Happy Wedding or Shotgun Marriage

Yi-chien Yao, Feng Chia University

Clyde A. Warden, The Overseas Chinese College of Commerce


In Asia, the role of English is vital in many professional fields. This fact has lead to English playing an important role in most Asian education systems. English writing instructors, in Asia, however, face an apparently insoluble contradiction of wanting to teach a process approach to writing while having to face large class sizes and heavy teaching loads. In this study, the application of computers, used for creating feedback on grammar and mechanical errors, shows how one responsibility of the teacher can be shouldered by technology. This approach used low cost technology and did not require expensive computer labs. The results included: no interference with the instructors teaching methods, decrease in errors, more time for emphasis on process writing.


English instruction plays an important role in Taiwan schools, from junior high to university. Non-English majors are often required to study English for a minimum of seven years. Writing is a vital part of such studies, especially during college and university level classes. At this level, it is not difficult to find teachers who know of the "process" writing and the teaching approach it entails. However, numerous factors influence teachers' ability, or willingness, to apply such techniques. The litany of problems is not at all unique to the Taiwan experience, and include, but are not limited to: time constraints, large class size, standardized exam preparation, departmental and school "objective standards," remuneration, teaching load and students' expectations.

The result is often a large emphasis on product, but not always through teacher choice or belief that the product is of paramount importance. EFL teachers are often struggling to satisfy the basic requirements placed on them. There simply is no more room to bring in other approaches, even though such approaches will reward everyone in the end. Absolutism and dogmatism, which often seem to accompany process approach endorsements, also do not entice teachers to break with tradition.

This paper reviews an endeavor to use computers in an attempt to break free of some of the above mentioned constraints, thus making adoption of a process approach more viable. The assertion is that automated correction of some writing errors can be integrated with a process approach successfully to the benefit of both instructor and students. Testing of such integration was attempted in Taiwan with the positive results of lowering of teacher workload and decreasing students' errors (Liou, 1993). The study reviewed in this paper was a larger attempt involving 173 students and including all the variables and problems that are encountered in the real world of EFL teachers working with large class sizes.


Process Approach

The process approach to teaching writing began about twenty years ago with a look into the writing process of experts and non-experts. American writing teachers began to focus less on the writing that students produced, and more on the actual process of writing. This thrust aimed at finding the sources of students' writing problems and then aiding them to overcome those problems. Writers, such as Murray (1980), were influential with the view that through a series of drafts a writer will slowly discover what s/he wants to say.

While the process approach has been widely accepted in the EFL teaching environment, there is no one clear methodology that all teachers follow. In fact, for EFL teachers in Taiwan, process writing is much more a general approach and attitude that exists within the context of many teaching tools. While the process movement has certainly affected the attitudes of many teachers, it is questionable how many teachers would themselves label their teaching "process." This approach has been most actively advocated in the native English environment; however, there are numerous pressures during education (testing and measurements) and after graduation (single draft tasks in the work place and translation tasks) which do not jibe well with a process approach (Horowitz, 1986) in the EFL setting.

Computers in EFL

Studies have shown that students have generally good feelings about computer technology being used in the classroom and that such technology does have a positive impact (Nash et al., 1989; Brady, 1990; Herrmann, 1987; Johnson, 1988; Monohan, 1982; Phinney & Mathis, 1988). The more specific use of computers in generating feedback or advice has much less clear results. When commercial grammar checking software reached some level of maturity, in the early 1990s, researchers looked into the viability of applying such software in language learning classrooms. Pennington (1991) pointed out the danger of grammar checkers causing teachers to revert to outdated methods of language teaching. She observed that grammar checkers seem to run counter to a more student centered classroom approach of teaching and that such software was often wrong in its error detection and advice, Pennington (1992). Liou (1991), investigated the application of the commercial grammar checking program Grammatik IV and found it mostly checked style errors and often missed substantive errors. Generally, all commercial grammar checking software packages failed to live up to their stated aims (Bolt, 1992). The initial reaction to computer based grammar checking placed a cloud of doubt over such software that still exists today.

Second looks at such software revealed programs with flaws but which could be improved on and serve a useful purpose. Liou, et al. (1992), Liou (1992), Holland, et al. (1993), Coniam (1991), Webster (1991), Xu (1994) and many other EFL researchers have worked to create their own software that can more accurately analyze sentences (commonly referred to as parsers). The results are generally positive, yet have not been able to substantially surpass the accuracy and convenience of the commercially available packages (although such custom-built parsers often perform better in finding specific errors or patterns). Meanwhile, the commercial packages have continued to evolve, as has the understanding of how to use them.

Bolt's (1992) detailed analysis of the different grammar packages, reveals that they fall into differing degrees of "transparency." This is a measure of how much of the program the user can access and even modify. The more transparent a program is, the more changes the user can make, thus reshaping the program for more specific needs. Levy and Garton (1994) continued the customization of GrammatikV that Brock (1990) had attempted with GrammatikIV. Warden (1995) followed this road and also adapted GrammatikV to be more sensitive to EFL business genre writing errors and showed that in a controlled study, students using such a program did lower their error rates. Liou conducted a similar study, using the software Complete Writer's Toolkit, in a process oriented class and found positive results, although no significant differences (possibly in part due to the total small sample size).


As has been observed by researchers not only in Taiwan, but around the world, large class size is often the norm. While it is clear that teaching plays the paramount role in activating student participation in a class, other important factors continue to make large process oriented classes very difficult to handle. The ever looming pressures of standardized examinations, teacher reviews as well as students' expectations all are pushing a teacher to reach some objective goals.

Although one may not agree with the importance of the product in EFL learning, what teacher wants his/her students to leave school knowing the students' writing, although process oriented, will be chalked full of mechanical and stylistic errors (all of which were taught, but simply not reinforced). Not only is this undesirable, it is a politically untenable position for a teacher in most Asian English departments.

In this study, an application of the process approach to writing, without the loss of emphasis on product, in a large class environment is reviewed. The inclusion of computer software allowed process to be stressed, while not losing sight of the final product.


There were 173 first year university students participating in this study. These non-English major students were from four departments:
Department: Insurance Finance Chemical Eng. Intl. Trade
Label: A B C D
Number of students: 38 37 48 50

All four class were taking a Freshman English course of 3 credit hours. Students were required to submit three writing assignments according to standard school policy.

In the previous six years of English language studies, these students received little or no formal writing training in a process approach to their writing. The reason for this was that the type of composition test given on entrance exams was very predictable and even announced prior to the actual testing time. The author has observed, while scoring the English composition section of the entrance examination, typical students' strategies include:

This general profile of entering college freshmen often challenges college English instructors in their syllabus design. Therefore, paragraph development, format, and organization, which students did not have much knowledge of, yet are indispensable elements of English composition, became the focus of freshmen English writing classes.

The Software

The authors found that the normal assumptions surrounding many studies in the U.S., as well as in Taiwan, were simply not realistic. The ability to supply each student access a computer terminal for immediate feedback on his/her writing presented numerous resource problems. Additionally, the instructor was not looking to make a major change in the way students approached their work, nor in the way the teacher managed his/her classes.

The QBL (Quick Business Letters) system (Warden, 1995) was adopted since it satisfied many of the issues the instructor faced. Fundamentally, the QBL system is a software package (custom designed in Taiwan) that supplies each student with a floppy disk for completing his/her writing assignments. Rather than tie students to specific terminals or computer labs, the disks can be taken anywhere and used on any IBM compatible PC, including old 286 machines. Students can print out their work at any time and bring it to class or to their peers for review. Before a designated time, the electronic files are sent, via local area network, to the instructor. The printouts are handed into the teacher and handled just as the teacher traditionally dealt with such assignments. The electronic files, however, are sent through a Windows based program (QBL TOOLS) that automatically finds errors, tracks students' progress and then prints feedback for each student on his/her errors. QBL TOOLS uses GrammatikV as its AI engine with a totally modified interface and over 400 custom rules that address the specific errors made by Taiwanese students.

The Process

The instructor did not attempt to create a classroom devoid of any mention or emphasis on the final product. Rather, she placed her approach firmly with what many, if not a majority, of practical EFL teachers see as a happy medium. Such an approach is elaborated by Caudery (1995, A-3 p. 12) in his survey of foreign language teachers:

While some respondents held fast to the "process, not product" ideal of earlier years, explicitly down-playing the importance of final product and emphasizing the focus on the process, at least as many respondents seem to have adopted a "process AND product" approach, stressing that the writing process is a means to an end, and consequently that process-based teaching should have that end firmly in sight.

Handouts, explaining the process approach, were given to the subjects ahead of each writing class in order to encourage process behaviors during the time students completed exercises in and out of class. Every individual writing activity was pre-scheduled, so students could prepare appropriately. In the first writing class, the instructor introduced the format and rules of a typed composition, a novel requirement for the students since most of them had little or no training in typing. Additionally, the computer software, QBL, and its features were introduced. One hour was required for each introduction.

Students were required to hand in their printouts to the instructor one week after instruction and to also send their texts to their instructor's account via campus based local area network. The instructor next downloaded the students' texts from the network to her disks. The QBL software was then used to automatically find the objective mechanical errors in each student file and then print out feedback sheets. Each student received one feedback sheet. Each feedback sheet contained some summary information followed by a list of the specific errors found in the text. Any specific error was printed with a reference to its location (Check:), the type of error (Rule Class:) and short advice on how to correct the error (Advice:), as in the following example.


Check: a assistant

Rule Class: Article

Advice: You may want to use `an` before `assistant`.


The instructor next glued each student's feedback sheet to his/her original printout, which had also been turned in, and then returned the paper back to the student for self correction. Students corrected their errors in blue ink on their printout paper and marked in front of every error picked out by QBL. They were encouraged to refer to grammar books or to consult their classmates on these objective errors. The next class session was reserved for individual conferences to cover errors that students could not correct either by themselves or with the help of their classmates. This conferencing would usually take less than an hour for a whole class. The instructor collected students' assignments again for manual correction and grading with emphasis placed on students' organization ability and the focus of the specific writing assignment.

Grammatical errors and sentence structure errors that were not detected by the software would be now marked in red, by the instructor. However, most of the time and effort was spent on the evaluation of students' paragraph development and organization of ideas. The instructor returned the graded assignments to the student's and reviewed their collective strengths and weaknesses during the next class period, requiring about 30 minutes.

Surprisingly enough, none of the students complained or raised questions about the integration of computers. The only problem encountered was with the network and the printers in certain computer labs. QBL is user friendly and comes with a manual in both English and Chinese.

For the second and third assignments, students received a checklist that described the areas of their writing that were emphasized during class (See Appendix 1). This checklist was to make students aware of the aim of their writing practice.

The third assignment was a revision of their first composition. According to the review of their weaknesses in class together with the suggestions from the feedback sheet, students edited their first composition. However, the main focus had been shifted to content, organization, cohesion and style (as listed on the checklist in appendix 1)

At the first glance, the current process was seemingly time-consuming

for the instructor:

  1. Students handed in their assignments
  2. Instructor downloaded students' files from the network and printed out the feedback sheets
  3. Instructor returned the assignment plus feedback sheets to students for self correction
  4. Students handed in their assignments after correction
  5. Instructor graded students' final works

On second examination, however, such an approach consumes much less time than traditional methods. In fact, the use of computers saves time not only for the teacher but also for the students (in ease of revision).

The normal teaching load, for English university teachers, is four Freshman English classes with a total of approximately 200 students (elective classes are in addition to this minimum). If it takes 10 minutes to correct and grade a single student paper, then a teacher would spend over 100 hours per semester (for three assignments), out of class, simply improving one of the four language skills of the students. Such an approach is tedious, time-consuming, and inefficient if teachers correct every error manually.

By integrating CALL software into a process oriented classroom, a better solution for both teachers and students emerges. The analytical data has shown that total amount of errors were significantly reduced in one semester and the teacher had fulfilled the role as a facilitator by introducing a process oriented approach.


Error Decline

All four classes showed declines in their rate of errors. Such a decline is consistent with the other studies in Taiwan using similar software (Liou, 1993; Warden, 1995).

Mean Errors
Assignment 1 Assignment 2 Assignment 3
Class A 13.12 9.87 3.97
Class B 8.70 6.37 2.33
Class C 11.70 9.33 5.09
Class D 11 7.61 3.82
Line chart of mean error changes over assignments

Analysis of variance showed that the four classes' total errors did not form a single population. This was to be expected as the English abilities of differing majors varies widely. Somewhat of a surprise, however, was that through the three assignments the four classes did not converge to form a single population. Since instruction was exactly the same for all classes, it seems that the starting point must be taken into consideration when examining the progress of any single class of students.

ANOVA analysis of all 4 classes through the 3 assignments
Assignment 1 Assignment 2 Assignment 3
F Ratio (F Prob.) F Ratio (F Prob.) F Ratio (F Prob.)
All Classes: 3.24 (.02) 5.81 (.0008) 8.26 (.0000)

Looking at each class on its own, it is quite clear that each assignment brings a statistically significant reduction in total errors. The reduction in errors is quite pronounced and clearly shows that students had responded to the computer generated feedback.

Significance of mean error change between assignments (using an LSD test)
Assignment 1 Assignment 2 Assignment 3
Class A 13.12 * 9.87 * 3.97
Class B 8.70 * 6.37 * 2.33
Class C 11.70 * 9.33 * 5.09
Class D 11 * 7.61 * 3.82
* Statistically significantly different at P

Linear-trend analysis shows clearly that the average of the four classes' total errors is moving downward at a rate of -3.66 (R Squared=.98, F=.59, significant at P

This is the end of the first part. Click here for the next part.