Investigation of Using Text-Critiquing Programs in a Process-Oriented Writing Class

Investigation of Using Text-Critiquing Programs in a Process-Oriented Writing Class

Hsien-Chin Liou
National Tsing Hua University, Taiwan, Republic of China

Abstract:
While the cost vs. gain of text-analysis CALL programs has been discussed (e.g., Brock, 1990) and major drawbacks of commercial packages such as Grammatik for specific learner groups were pointed out (Liou, 1991), few research studies on integrating available text-analysis programs into realistic writing classroom activities have been formally conducted so far. This research report addresses the issue by documenting the college EFL writers' revision process in which human teachers' feedback, students' own revision, and the use of two commercial packages, Grammatik and Complete Writer's Toolkit, were incorporated. A small-scale quasi-experimental study was conducted to examine the effect of the packages while assessing the writing performance of 39 subjects placed in either the control group or the experimental group. In addition, an interview was conducted to elicit the subjects' response to this CALL strategy. A comparison was made regarding (a) the performance of the two-group subjects, (b) the effects of the programs vs. subjects' peer comments (peer editing), and (c) the differences between the effects of critiquing of Grammatik and Complete Writer's Toolkit. Other factors which may affect the gain while using the program were qualitatively documented. Results showed that each of the two packages has a role to play for students of various proficiency levels and that weak subjects benefited more from Writer's Toolkit and liked such programs better. No group difference in writing quality was found regarding students who used or did not use the programs. Attitudes toward the use of such programs tend to be positive. It seems that text-analysis programs may be beneficial to learners in writing revision given careful classroom design and individual attention to learners' writing proficiency. Several pedagogical implications are raised.

INTRODUCTION

Focusing on accuracy of form while not sacrificing expressiveness of composition, this paper addresses the issue of writing revision from the perspectives of (a) kinds of help which computer-assisted language learning (CALL) may provide and (b) pragmatic concerns which exist in the Taiwanese context.

Writing revision tasks can be divided into an examination of five aspects in an essay: contents, organization, language use, mechanics, and vocabulary (Jacob, Zinkgraf, Wormuth, Hartfiel, and Hughey, 1981 provided an evaluation profile with these five categories). It is argued in this paper that human teachers and CALL should cooperate to make writing revision cost-effective. In this cooperation, CALI, may serve well in assisting grammar, spelling, and punctuation editing/revision tasks. These tasks mainly deal with accurate form in writing because computers, as machines, will function best in mechanical aspects in the foreseeable future of technological development. Truly, correct form alone, or accuracy of writing, does not constitute a good paper. Further, a shift in focus on fluency and appropriateness occurred when communicative language teaching started to influence the teaching of English in Taiwan several years ago. However, first, communicative approaches do not neglect accuracy at all but try to add another dimension of consideration by counter-balancing¹ the predominant single focus on accuracy of form alone as in the past. In addition, recurrent errors in production in a piece of student writing, demonstrate the learner's immature English proficiency. Thus, the pursuit of correctness of form in EFL writing can be justified, though research is required in terms of how the gravity of grammatical mistakes influences the total evaluation of a paper.

In Taiwan, the English-as-a-foreign-language (EFL) writing class has been regarded as one of the most formidable courses to teach due mainly to the task of correcting student papers. It occupies much of the writing teachers' evaluation time, among other revision or commenting tasks. Most of the students tend to make grammatical mistakes repeatedly regardless of the teacher's efforts in correcting the same errors several times. In addition, correcting composition mistakes has become not only the writing teachers' headache in formal schools but also has created an urgent demand for semi-government-sponsored institutes. Researchers, professors, and graduate students in the fields of science and technology, with a great need to frequently publish scholarly papers, resort to writing revisors in some sort of writing center. National Tsing Hua University is a good example of this phenomenon. The University is famous in Taiwan for its science and engineering programs. Most of its graduates are required to finish their theses in English. However, this is never an easy task for either the graduate students or the advisors, the latter suffering from correcting errors in student papers. In view of this, in 1988 the Dean of Academic Affairs in our university launched a project which was aimed at promoting English proficiency of the students. A writing center was set up where graduate assistants are responsible for correcting the research papers of graduate students in the fields of science and engineering. One of the other ideas was to use computer programs to correct the grammatical errors on students' papers. It would save a great deal of time if computer programs could help with some revision tasks. During the period of 1989 to 1992, we purchased four text-critiquing packages, Right Writer (Rightsoft, 1988), Grammatìk IV (Price, 1989), Complete Writer's Toolkit (System Compatibility Co., 1990), and Power Edit (Artificial Linguistics, Inc., 1991). Since the version of Right Writer we obtained did not allow an interactive mode of processing and could not satisfy our needs in terms of error detection, it was given up for widespread use. Power Edit, rated as the first (Rabinovitz, 1991), is hard to learn to use in a short period of time due to its complicated functions. Furthermore, as a pilot test, these packages were first used with students of Foreign Languages, who had little computer literacy. Power Edit was regarded as inappropriate for the moment. Grammatìk IV was examined by the author but found ineffective for the students in a self-access context (Liou, 1991). Thus, Complete Writer's Toolkit is the main package which is investigated in this study.

REVIEW OF THE LITERATURE

Since word-processing packages gained popularity in writing classes, text-critiquing programs, or grammar checkers, have become widespread. However, common criticisms against text-critiquing programs stem from several arguments. First, they tend to foster student dependence on programs. Second, most of the commercial programs generate incorrect analysis due to their limitations in using primitive computational techniques. Then, there is a concern that students — with deficient language proficiency — tend to accept the incorrect analysis. Last, the use of such programs encourages a product-oriented approach to writing. In spite of the claims that appear to be against the use of text-critiquing programs, empirical data are needed to validate them.

In second language (L2) contexts, several studies on the use of text-critiquing have been conducted. Reid (1986), using Writer's Workbench found that generally students liked such programs, but she did not examine whether or not the quality of writing was influenced by the use of the program. Pennington and Brock (1989) compared the writing quality of two small groups of students with one applying tutor-facilitated process-oriented revision and the other, the feedback of Critique alone. They found that the Critique group performed worse in producing more short sentences, shorter drafts, and fewer revisions. However it is hardly fair to compare a human tutor who understands content and organization with a grammar-sustained program, which has inherent limitations due to its inability to analyze the semantic or pragmatic aspects of writing. Brock (1990), using the new self-formulated rule sets provided as facilities in Grammatik, argued that though there is the possibility of customizing text-critiquing programs for English-as-a-second-language (ESL) writers, important questions remain as he explored the cost versus gain:

• For what purpose is the text analysis program to be used?

• To what extent will a grammar checking program, such as Grammatik, improve student writing?

• What messages about writing are teachers conveying when they encourage ESL students to use text analysis?

• Could the time students spend with a text analysis program be given to other writing tasks that offer greater returns?

• If text analysis is utilized, where in the writing process should this occur? (p. 59)

Liou (1991), investigating the usefulness of Grammatìk IV (G4) for English-as-a-foreign-language writers in a study in which no control group was used, found that only 14 percent of the errors G4 detects are substantive though student writers felt the process interesting. She then launched a project to develop a grammar checker specifically for Chinese students.

A thorough investigation of text-critiquing programs should be conducted as not enough empirical evidence has been provided yet to determine its usefulness. Previous research has either lacked a control group for comparison of writing quality or a valid comparison, or was not conducted in a realistic class setting. It is believed that CALL should be integrated into regular classroom activities so that it becomes a mode of learning (also supported by Levy, 1992). Further, as a teacher-researcher, I believe that conducting classroom-oriented research can directly benefit classroom practice because the findings can provide immediate feedback for teaching practices. Thus, to address the effectiveness of text-critiquing programs through empirical investigation and, if possible, to suggest a good way of employing them in EFL writing classrooms are the purposes of the present study.

RESEARCH STUDY

The study was designed in such a way that each of the procedures was integrated into regular classroom activities in a first-year writing course — Grammar and Writing (G & W) for EFL majors in a university in Taiwan, Republic of China. The design is a quasi-experimental study with 19 students in the CALL group and 20 in the control group. Three months before the research study was conducted (Spring 1992), a pretest measure including a 20-question sentence-level test and an essay-writing task (titled "My College Life") were used to see if there was a difference between the groups. The verification of the lack of difference is seen in Table 1.

In the G & W class, in addition to one-on-one conferences, freshman students were asked to write five essays according to the five units/topics designed for this semester. Two-week (four hours) activities form one unit, during which students had to write up one paper. The G & W course was designed to teach writing through a process-oriented approach whose activities can be summarized in Table 2.

In the first two hours of each unit, classroom activities involved lecture and practice of grammatical points - through sentence combining exercises, vocabulary presentation by a small group of students about a particular topic, group editing of a student essay selected from the previous unit, and editing practice of another essay from the previous unit. During the second two hours, some kind of pre-writing activity was carried outside before the task of composing the writing assignment started in class. Lastly, all the students formed pairs to do a peer editing task where students had to critique their partners' papers regarding clarity and structure. This was done sometimes out of the class due to time constraints. There are forty² first-year students in our department; they were evenly divided into two sections. Instructors of the two sections followed the same course design and materials. Students of the two sections served as the CALL group and the control group.

For this study of the writing assignment a choice was given of "The Advantages and Disadvantages of Being a/an ______ (name of a profession)," or "Some Career Tips for College Graduates." The research procedures are summarized in Table 3.

When the students finished the first draft, they did the peer editing task. Then they went to a computer laboratory, typed in the paper, and submitted it to the instructor. For their first drafts, the instructor looked closely at the content and organization and made general comments like ,! re-write the five lines" or "You need a conclusion," etc. In addition, the instructor underlined each of the problematic areas (ungrammatical points, wrong word form, inappropriate word usage, etc.) without any other marks or directions for correction. It took about 11 to 12 minutes to comment on each essay. In the second week of the unit, students were given their own essays back and did the group editing and individual editing activities; then they were asked to revise the first drafts after class on the computer. After one or two days, each of the students came back to the instructor's office and ran Complete Writer's Toolkit (CWT) one by one with the instructor observing the process (about 15 minutes for each subject). Because they had never used any text-critiquing programs like CWT, the instructor told them how to proceed. Some of them were not satisfied with the performance of CWT; then, the investigator introduced G4 to them and let them run the essay again using G4³. After the grammar checking process, their essays were printed out to await final grading. Then, each of the students was orally interviewed by the instructor about how they felt about the package. Notes were taken during the process. The following six questions were asked:

1. Have you used CWT or G4 before?

2. Do you like packages such as CWT or G4?

3. Do you think they are useful in writing revision?

4. How do they compare with the effectiveness of peer editing?

5. Do you think their use can save you time in writing revision?

6. Do you think they are useful for EFL learning?

Meanwhile, the control group followed the same procedures except that they skipped the commenting task done by CWT. Both groups turned their final drafts in after either the students' own or CWT's and the students' own revision was completed on the first drafts. The investigator ran CWT on the control group's drafts for comparison.

Data Analysis

Error analysis was done on each of the subjects' drafts; the total numbers of errors and words in each of the drafts were recorded. The types of errors were marked but were not quantitatively considered for this study. The CALL group's drafts analyzed by CWT or G4 were recorded and analyzed. Final drafts were graded according to the following criteria: 30% of the total score was allocated to grammar/sentence structures and 70% to the other aspects. This resulted in a score out of 100 points. The reason for special allocation to grammar and sentence structures is that the course was designed to strengthen student writing and in particular grammatical accuracy in sentence-level performance. In grading, when a paper consisted of many errors or simple and naive sentence structures, it was scored low in the 30% portion. Papers of the two groups were scored by the same instructor. The T-test procedure was used to compare the means of writing performance of the CALL group and the control group. Most of the interview data were coded on a scale of being positive to negative. Responses from the 19 subjects were tallied according to the scale and percentage for each question. Idiosyncratic comments were recorded.

Results

A comparison of total words used in the subjects' first drafts and final drafts showed that the mean difference was small (first drafts, 339 words on average; final drafts, 359 words). This indicated that most of the subjects tended to keep their content and possibly organization (no further analysis of organization was done).

Comparison of writing performance

Errors were located in the drafts of the subjects' first version, the CWT version, and the final version. Throughout the revision process, subjects might have created new errors when attempting to revise content organization, or expressions.⁴ The newly added errors and those in the first drafts became the grand total number of errors for each group. The results are shown in Table 4.

More than half of the total errors remained in the final drafts. It is clear that the control group made more errors and did not correct them at all for the second drafts — only 0.5% of them were rectified. In contrast, the CALL group made fewer errors and knew how to rectify them in the second drafts - 7% of the errors were removed. It is evident that subjects were not able to correct most of their mistakes by themselves even after some devices to raise their consciousness as to form, such as marks, were used.

Further, about one tenth of the errors came from the student revision process - see added errors. It seems that the more they wrote, the more errors they tended to make.

Most importantly, the results indicated that CWT had exerted some influence on the CALL group: 38% of the total errors could be detected by CWT. It would have eased one fifth of the instructor's burden if it had been used in the control group (20.5%).

As for the final evaluation of the papers, no difference in writing performance was found regardless of whether or not CWT had been used. Results from the T-test procedure run on the final scores to compare the group means between the CALL group and the control group indicated that there is no significant group difference (P > .05; see Table 5), though the mean of the CALL group (68.3 /100) is larger than that of the control group (66.4 /100). This may suggest that error reduction due to use of CWT does not contribute to the gain in the final grade of a paper, but it may be attributed to the fact that a small sample of subjects participated in this study.

COMPARISON BETWEEN USE OF CWT AND G4

One of the major drawbacks in the commercial packages like CWT, other than the fact that they use primitive computational techniques, is that they are designed for first language (L1) writers. Among other differences, it is noted that native speakers do not make frequent grammatical mistakes like EFL learners regarding types and frequency. However, CWT and G4, as used in this study, seem to suggest some usefulness.

As mentioned, when subjects were not satisfied with the critique from CWT, the observer asked if they would like to try another and introduced G4 if the offer was accepted. This happened mostly with the more advanced student writers. No quantitative measure was made to compare the performance between that of CWT and that of G4; however, more than 30 cases of error detection examples were taken from the running sessions of each package. An impressionistic summary of these samples indicated that CWT did outperform G4 in terms of accuracy of error detection in that CWT covered more error types and generated fewer false alarms. Sophisticated student writers did not think CWT useful in substantive ways, but the less proficient students found it useful. While G4 is less accurate in grammar and mechanics, it is superior in stylistic checking (see examples in Table 6), which may benefit advanced students more.

The two examples taken from G4 critiques caused an advanced learner to look up the usage of rather and fairly in an off-line dictionary before making the final word choice.

Other than these considerations, the two packages are comparable regarding user-friendliness and easy-to-operate interface design, the focus of error types to be detected (see the following) as well as false alarms and the misleading messages they tend to generate (detailed examples are listed in Appendix A).

EXAMPLE: to unconsciously suggest

message generated by G4: Avoid splitting the infinitive to suggest. Try changing the position of the intervening modifier(s).

message generated by CWT: The sequence {to ... suggest} may be a split infinitive. It is preferable to avoid split infinitives by placing adverbial material before or after the verb.

Furthermore, false alarms may not become detrimental for learners. For example, in one case CWT generates a false alarm with irrelevant feedback, as shown in one of the subjects' drafts:

"As teachers, our great achievement and satisfaction are that all of our students can get somewhere," said my teacher and aunt. Indeed, this feeling is exceedingly strong in my father's mind whose is usually proud of his students' success.

CWT's feedback: Some writers prefer to use that rather than which in restrictive (defining) relative clauses. If the relative clause is non-restrictive, which should be preceded by a comma.

(See Appendix B for an elaborate long message for this point)

CWT flags whose. Even though it did not give an accurate feedback message, this consciousness raising caused the subject to revise the sentence drastically into the following, which is beyond her regular revision habits.

"As teachers, our great achievement and satisfaction are that all of our students can get somewhere," said my teacher and aunt. Naturally, they are always proud of their students' successes.

For this grammatical point, there is a corresponding message in G4.

Check: which

Advice: That is almost always preferred to which in this situation. If you really mean which, then it usually needs to be replaced by a comma. Press the Help key for more information.

Replace: that

Likewise, the result of the following false alarm in G4 was the following:

To forge ahead toward what they intend to do without considering the inner and outer factors does not guarantee successful; [Advice: Consider using an adverb instead of the adjective {successful}.]

The advice message, though misleading, raised the subject's consciousness of form and finally caused her to replace successful with success.

Attitudes Toward the Use of Text-Critiquing Programs

As to the question of student perceptions of the use of such a tool, results from the interview were summarized in the following figures (see detail in Appendix C). Fifty-two percent of the subjects liked such packages; 69% of them thought the programs were useful for writing revision; 69% of them found the programs could save revision time; and 52% thought the programs useful for learning. Generally speaking subjects had a positive attitude toward the use of such programs. However, the role of peer editing in our syllabus needs to be re-assessed; at least the way we designed its use needs to be improved. Because the peer editing task requires good enough English proficiency and accommodation of scheduling to arrange for the meeting of pairs, the interview results suggested that use of CWT may save more time than asking the partner to criticize the paper. In addition, the partner tended to be lenient about the peer's paper, but CWT faithfully pointed out the mistakes. This also suggests that pair dynamics in the syllabus design may influence the degree of usefulness of the programs.

CONCLUSION AND IMPLICATIONS

In conclusion, though no difference was found in the quality of writing regardless of the use of text-analysis programs, the error number counts showed that the programs may help remove some portion of mistakes in a paper, if one uses them. Either CWT or G4 has a role to play in text-analysis for learners with a certain proficiency level. Moreover, students have a generally positive attitude toward the use of such programs. Given sound design and preparation from classroom teachers, the use of text-analysis programs does not have to lead to a product-oriented writing class.

The cost versus gain in using a text-analysis program to assist in writing revision involves many variables over which classroom teachers may have most of the control. Several implications can be drawn from this study. First, it is suggested that student attention should be called to the use of the program by a writing instructor, if s/he decides to incorporate the programs into the students' writing. Teachers may warn students that such programs have limitations; thus, students will not have too high expectations nor completely depend on them for revision. Besides, instructors should give detailed instructions before they ask students to use such programs or even give a demonstration to familiarize them with terms or explanations used in the packages. Brock (1990) warned that Grammatìk III may not be a suitable addition to ESL composition pedagogy because students may believe the incorrect analyses and adopt them into their revision. Indeed, some of the false alarms from such programs are very misleading to weak student writers. It is safer if the instructor can observe the process several times before having students use it on a widespread basis. Whether the programs encourage an orientation toward product depends on how a teacher designs the writing class and when or where to encourage students to use the program. In our case, it is used to refine students' final drafts, at a time when form becomes important after the content and organization are fixed.

Another implication is for the developers of text-analysis programs. Overt feedback for an error in such systems may not help students develop their own revision strategies. The system may as well produce messages telling student users how many errors of a certain type are detected in the paper with/out marks. That is, it should summarize the error types and frequencies) but not direct them how to correct the error  (communication with Liz Hamp-Lyons at TESOL'92 Vancouver). Students should learn how to correct their own mistakes if we hope they will become independent learners and writers. Another nice feature the programs could incorporate is a tutorial or elaborate explanation, such as those in CWT, to be at the user's disposal. Those done in CWT are barely satisfactory because they tend to address an irrelevant point for the error detected; moreover, the examples and explanation messages are not very helpful. Better instructional design in these aspects is what language teachers can contribute. For non-EFL majors in the Taiwanese context, it may as well provide explanation messages in Chinese to facilitate understanding.

Equally important are the teachers' efforts. Classroom teachers should be responsible for teaching students strategies to enhance clarity and quality of writing, which text-analysis programs cannot achieve in the foreseeable future. It is suggested that the text-analysis programs be used in process-oriented writing instruction, given clear instructions and sound design, when the purpose of their use is to call upon students' explicit knowledge in editing their final drafts.

Last, this study suggests that given human efforts in the revision efforts of contents and organization — from either peers or the teacher — text-analysis programs can help with the revision of grammar, spelling, and punctuation. Such use encourages a collaboration task where humans and CALL complement each other to make the revision task cost-effective.

APPENDIX A

Examples of Output from CWT and G4

OUTPUT from CWT

False alarms

• may be various from person to person -- > people or persons

• Not only is teaching a stable career but I like children very much. Why do they have this feeling?

• --> not contain a main clause

• Some of them pay no attention to what the teacher teaches but disturb the learning atmosphere of whole class. -- > disturbs.

• the disadvantage of being a teacher is that ethical morality between teachers and students is gradually disappearing. --> are

• many guiltiness young students [The word {many} does not agree with {guiltiness}.]

• First of all --> [Consider firsts instead of {First}.]

• First, you must find out what kind of career you interest in most automatically. [Advice: The word {First} does not agree with {interest}.]

Errors detected

• the contact with unsophisticated children often keep a young mind --> keeps

• Like my father and relatives often tell me "I can get more achievement of being a teacher than any other career." [Advice: Consider {As} or {As if} instead of {like}.]

• Take an example, a man who likes young students will be more patient with his students so he will like to be a teacher. [Advice: This appears to be a run-on sentence.]

• may refused --> refuse

• The author claims that "an important issues that has involved journalist in legal procedures (and even put it in jail) is the matter of protecting the identities of sources." [Advice: The quoted material appears to be improperly punctuated.] -->issue

• must goes -- > go

• an news --> a

• upholded --> upheld

System failure

Message: Sentence was too long to process for grammatical structure. (as compared with related phenomenon detected in G4: long sentences can be difficult to read and understand. Consider revising so that no more than one complete thought is expressed in each sentence.)

OUTPUT from G4

False alarms

• Thirdly, most teachers agree with the fact that getting along with many young and active students is a great pleasure of being a teacher which also brings them young spirits.

• [Be sure you are using is with a singular subject. (It is.).] [The singular noun teacher may be used incorrectly with the plural form of the verb bring.]

• several career tips --> several careers tips

• as being famous, making a lot of money --> as

Errors detected

• more direction ("direction" should be in plural form.)

• You may also

• a unsuitable one.

• amounts of their work does

• the both (word order)

• Well, I am very fond of this job because what waiting for me in everyday life are challenges, not routines, --> is

• "As teachers, our great achievement and satisfaction are that all of our students can get somewhere," said my uncle and aunt. (CWT caught this, too.)

Style

• A paragraph contains more than one sentence.

APPENDIX B

On-Line Long Tutorial-Like Message in CWT

Some writers prefer to use that rather than which in restrictive (defining) relative clauses. If the relative clause is non-restrictive, which should be preceded by a comma.

Relative clauses provide additional information about a noun. They are classified as restrictive or non-restrictive depending upon whether they are essential to the definition of the noun or supplementary to it.

A restrictive clause is essential to the meaning of the sentence in that it restricts the definition of the noun to the class described by the clause. For example,

The painting that won first prize is hanging in the foyer.

(The relative clause "that won the first prize" specifies exactly which painting is hanging in the foyer.)

A non-restrictive clause contains information that is not essential to the specification of the noun it modifies. It is more like a parenthetical comment. For example,

The use of seat belts, which can prevent serious injury in auto accidents, is now mandatory in several states.

(The relative clause "which can prevent serious injury in auto accidents" provides supplementary information about seat belts, but is not essential to the statement that their use is now mandatory in several states.)

When a relative clause is non-restrictive (and the noun begin modified is not a person), the relative pronoun is "which," and the clause must be separated from the rest of the sentence by commas.

When a relative clause is restrictive (and the noun being modified is not a person), the relative pronouns "which" and "that" are often used interchangeably. If "which" introduces a restrictive relative clause you can substitute "that" without changing the meaning. For example,

A black hole is a star which has collapsed

Some writers argue that the use of "which" in restrictive relative clauses may be confusing. They prefer to use "that" for restrictive clauses and to limit "which" to non-restrictive clauses.

APPENDIX C

Results of the Interview Data (case = 19)

1. Have you used CWT or G4 before?

0% of them have.

2. Do you like packages such as CWT or G4?

Yes 52% No 36% Don't know yet 11%

(Note: The figures were rounded off to the nearest whole number.)

Other comments

• acceptable

• like it if it is useful

• funny but confusing (because subject didn't know much of the grammar checker like a spelling checker)

3. Do you think they are useful to writing revision?

Yes 69% very little 16%

No 5% don't know yet 11%

Other comments

• better than nothing

• It takes time to run but functions are few and unsatisfactory.

• little on contents and organization

• Only computer thinks it incorrect in using "secondly;" nobody thinks so

4. How are they compared with the effect of peer editing? (prompt more about the peer editing task)

Positive comments

• partners are fine

• fine but lenient; courteous, did not correct word usage (only read for clarity)

• corrected my paper to a limited extent

• corrected grammar

• conscientious partner, commented on every aspect of paper

Negative comments

• reluctant to comment (because the subject's paper is at least twice as long as that of average subjects for most of her writing assignments) — impatient to read a long essay

• partner nastily criticizes

• partner not easily accessible

• neither partner nor computer helps

• partner often tired, could not match mutual schedules

• almost didn't revise any

Other comments

• peer corrected semantics, not mechanics, depending on peer's attitude and proficiency level (if peer is good or compatible, the commenting task tends to be satisfactory)

• peers are humble but comments not useful, would rather replace a partner, like teacher's comment, very helpful; if he corrected others, others would feel unhappy

5. Do you think their use can save your time in writing revision?

Yes 69% No 11%

Barely 5% Don't know yet 16%

Other comments

• In using the program, we do not have to find a classmate to comment on our paper.

6. Do you think they are useful for EFL learning?

Yes 52% No 22%

Don't know yet 21%

Other comments

• To some extent. But we should read more books instead of using CALL programs (due to limited experience with CALL).

• No, but would like to try it later.

ACKNOWLEDGMENT

Special thanks go to Ms. Yuli Yeh, who taught the control group for this study and helped with data collection. I would like to acknowledge Ms. Kui-Pen Hsu's contribution in helping analyze part of the data.

NOTES

¹ Correctness of form has been a focus in our EFL education; one of the reasons is that learners do develop variable interlanguage systems, or errors at every stage of language development, especially at the beginning level; the other reason is long-term pedagogical emphasis on accuracy.

² A subject in the CALL group dropped out in the middle of the study; thus, there were 39 subjects in total.

³ For those who only used CWT, the investigator herself alone ran their papers again using G4 and recorded the output.

⁴ The T-unit (Hunt, 1965) was used to analyze four subjects' final drafts with an increase of more than 30 words from their 1st drafts. Two of them added one to four T-units without an increase in error number, whereas one, interestingly, increased 5 T-units with an increase of 3 errors, and another one increased 4 T-units with an increase of 5 errors. This may have indicated that their interlanguage systems are still developing.