Shawn Loewen

Grammar Correction in ESL student writing: How effective isit?

     The benefit of grammar correctionis a topic of continued interest to English as a Second Language(ESL) writing instructors because we are faced with thepracticality of this issue on a regular basis. What kind offeedback should I give? How much? How often? Does it help thestudents? In an article reviewing current research on grammarcorrection, John Truscott (1996) answers these questions by arguingthat grammar correction (which he defines as the "correction ofgrammatical errors for the purpose of improving a student’s abilityto write accurately") is not only completely ineffective but alsoharmful and should consequently be abandoned. Truscott claims thatno studies have proven that grammatical feedback on student writingleads to greater accuracy; those studies that make this claim are,according to him, faulty or inconclusive. Intrigued by such adefinitive solution to this issue, a colleague and I decided toinvestigate this matter for ourselves in an action researchproject. This paper will examine the research that led us to thisstudy, detail the method used to investigate, describe the results,and discuss the limitations and implications of the study.

LITERATURE REVIEW

     As previously mentioned,Truscott (1996) claims that no studies have proven (to hissatisfaction) that grammar correction is effective. In fact,several studies have found that students who received content-basedfeedback improved in various aspects of writing, particularlyfluency. This leads Truscott to claim that grammar correction isharmful because it discourages students from writing fluently.Three of the studies Truscott uses to support this claim are Semke(1984), Sheppard (1992) and Keppner (1991). The results of thesestudies will now be discussed briefly.

     Semke (1984) investigated fourgroups of students which received different types of feedback.Group one received feedback on content only, group two receivedfeedback on errors, group three received feedback on both contentand errors, and group four had errors pointed out but notcorrected. No significant difference was found in accuracy;however, group one, which received only content feedback, didsignificantly better on fluency and fill in the blank tests. Groupfour, which had errors indicated but not corrected, had the worstresults on the fluency test. The content feedback group was theonly one to improve.

     A similar study by Sheppard (1992)consisted of two groups, one receiving both written and oralgrammar correction, the other receiving feedback only on content.Again, there was no significant difference between the two groupsin accuracy; however, Sheppard found that the content group didbetter on sentence boundaries and that the grammar correction groupdid worse on a test of complexity involving the use of subordinateclauses (Sheppard attributes the low complexity scores to studentschoosing to use simple structures and thereby avoiding more complexstructures which they might use incorrectly.)

     Finally, Keppner (1991) found nosignificant difference in the accurate use of structures betweengroups receiving feedback on grammar and content. The onesignificant difference that was found was that the content groupimproved on higher level propositions.

     In contrast to these studies whichfound no significant effect for grammar correction, Fathman andWhaley (1990) found that both grammar and content feedback, whethergiven separately or together, resulted in improved writing onrevisions. This study, however, focuses only on revisions, not onsubsequent essays, and Truscott argues that there is no reason toassume that improvement on revisions due to grammar correction willlead to students becoming better writers

     Not only does Truscott argueagainst the efficacy of grammatical feedback, he also cites studieswhich examine practical problems in grammar correction. Cohen andCavalcanti (1990) found that teacher bias played a large role indetermining the type of feedback a student was given. Teachers feltthat certain students had certain types of problems which needed tobe commented on when this may or may not have actually been thecase. In addition, the researchers found inconsistencies in theteachers’ feedback; the teachers often did not notice many of thestudents’ errors. Similarly, Zamel (1985) reported that teacherswere not consistent or systematic in giving feedback and that thequality of the feedback was often poor.

     While teachers are often not adeptat giving feedback, students’ responses to feedback are alsoproblematic. Cohen (1987) shows that students have limitedstrategies for dealing with teacher feedback on essays. Many of thestudents in the study simply made mental notes of the feedback. Inaddition, many times the students did not understand the teacher’scomments; thus rendering the comments ineffective. Truscott notesthat Cohen does not question the benefit of grammatical feedback;instead, Cohen concludes that students need to be better trained inmaking use of teacher comments and that teachers need to be clearerin their comments.

     One study not addressed byTruscott that makes a case for a specific type of grammarcorrection is Aljaafreh and Lantolf (1994). In this study theresearchers used the Vygotskian Zone of Proximal Development (ZPD)to guide students to grammatical errors by narrowing the focus onerrors from a page, to a paragraph, to a sentence, to a line,according to the ability of the students. Eventually, if thestudents were not able to locate and correct the errors forthemselves, the researchers indicated them. The goal of thisgraduated approach to error correction was to move students towardsa greater ability to self-correct, instead of relying on teachercorrection. The study did find that students were more adept atself-correcting after receiving this method of correction.

     In light of these conflicting, andoften strident (primarily on the part of Truscott), claims aboutthe value of grammar correction, my colleague and I devised thefollowing research question: Does correcting grammatical errors instudents’ essays result in improved accuracy in the use of thosestructures in subsequent revisions as well as in new essays?

METHOD

     The subjects of this studywere all Temple University students in two classes of Composition50 for ESL students. There were two classes which are designated asSection 1 (n=5) and Section 2 (n=7). All the students in Section 1were Asian, either speakers of Gujarati, Vietnamese, Korean orChinese. Section 2 consisted of two Spanish speakers, and the restwere Urdu, Chinese, or Korean speakers. The ages ranged from 18 to30 with an average age of 22. Nine of them were first yearstudents; the other three were in their second year. The length ofU.S. residence ranged from 3 to 12 years with an average of 7years.

     The topic of the course was‘cultures in America’ and the readings focussed on such issues asmulticulturalism, immigration, bilingual education and officiallanguage policy. The students were required to write seven essayswith an optional eighth. The essays were expository in nature,asking the students to defend a position on an issue which had beenread and discussed in class.

There were two revisions, as shown the chart, and the length ofthe essays varied depending on where they were written. By way ofexplanation, Essays 6 and 7 were not returned to the studentsbecause they were turned into the department for furtherassessment; consequently, there was no possibility of feedback onthese two essays.

     It seemed fairly clear from theliterature that global correction of errors was ineffective;therefore we decided to focus on two specific grammatical errorsper section on which we would give feedback. This was notinconsistent with the type of error correction I had been giving inprevious semesters since my method for marking essays was to focusprimarily on organization and content which were also the main fociof the course. I would mark some grammatical errors if they wereparticularly egregious or if there was a consistent error beingmade. My grammar correction, however, was quite haphazard andinconsistent. For the purposes of this study, we decided to focuson two specific errors per section. From the students’ first essay,we ascertained that plurals, simple past tense, 3rd person -s, andpassive voice were grammatical features that the students had somedifficulty with. We used the first essay as a baseline from whichto measure, and we also began giving feedback on the first essay.The two sections received feedback as follows:

     Section 1: Plural -s and SimplePast Tense
     Section 2: Third Person -s andPassive Voice

Since each section received feedback on two differentstructures, they served as controls for each other.

     The form the correction took wasalso important. We decided that the error would be underlined andthe correct form would be written above. This way there would be noambiguity about the nature of the error or the correct form. Also,after almost every essay, the students met with the teacher for aconference, and the teacher would point out the errors which he hadmarked. However, there was generally little more than a passingremark about grammatical errors unless the student asked for moreinformation which seldom occurred.

     For the actual coding of theessays, my colleague and I marked the errors independently, andthen compared our results. For any discrepancies, we negotiatedbased on a set of guidelines until we agreed on whether or not itwas an error. We counted both obligatory occasions where thestructure must be used and overgeneralizations in which thestructure was used when it should not have been.

In the essays which were returned (three of the essays were notreturned to the students due to departmental policy) theappropriate grammatical errors were underlined and the correct formwas written above. No other grammatical feedback was given;however, feedback on the sentence, paragraph and global level wasgiven as usual. Primarily the other feedback pertained toorganization and structure of the essay.

     Error analysis is fraught withdifficulty, and one of those difficulties was determining what typeof error was involved. For example, there were quite a few caseswhere a singular noun was used incorrectly. One way to correct theerror would be to pluralize the noun. Another way, however, wouldbe to add an article (a, the) before the noun. Since articles arenotoriously difficult for ESL students, we did not count these asplural errors when the presence or lack of an article would makethe sentence correct.

     Also, in examining passiveconstructions, both the copula and participle had to be correct orit was counted as a passive error. However, if both copula andparticiple were incorrect, it was still only counted as one passiveerror.

Plural -s:
Can America afford this many immigrant?
Students shouldn't be offercourses in their native language.
Third Person -s:
Learning English help immigrants tocommunicate with everybody around them.

RESULTS

     First let’s look at thefrequency of the structures under investigation. Charts 1a and 1bshow the mean frequency of each of the four structures in the eightessays. The scores reflect both obligatory occasions andovergeneralizations, and therefore only illustrate only thefrequency of each structure, not accurate use of the structure. Ifthe structure was used, either correctly or incorrectly, that useis reflected in these charts.

     As is apparent from the charts,both sections used plurals quite frequently. The scores range froman average of 20 plurals (Section 1, Essays 6 and 7) to an averageof around 65 (Section 1, Essay 3 and much of Section 2). However,the mean frequencies for the other three structures are much lower.The past tense averages above ten occurrences only once for Section1 (Essay 8) and three times for Section 2 (Essays 2, 5 and 8). Therest of the essays have a past tense frequency score of below ten,and sometimes below five. The occurrences of third person -s andpassive voice are also below ten and often below an average of fiveoccurrences per essay.

     Turning now to the accuracy scoresin Charts 2a-d, we can see that the accuracy score for the plural-s for Section 1 begins at 45% on Essay 1. Then in the subsequentessays it ranges between 70 and 90% with the three in-class essays(4, 6, and 7) being the lowest scores besides Essay 1. Section 2’sscores range between 80 and 95%, again with the lower scoresoccurring on the in-class essays.

     The third person -s accuracyscores show Section 1 beginning at 70% accuracy and ending at 85%.Again Essay 4 which was in-class shows the lowest accuracy scorewith 55%. Section 2 begins just under 85% and ends at 100% afterbottoming out at 70% for Essays 6 and 7.

     The past tense scores for Section1 begin at 70% and then decrease with the revision in Essay 2. Bythe end of the semester the scores have risen to 95%. Section 2begins at 85% accuracy and also decreases in the second essay.After bottoming out at just under 70%, this group finishes wherethey started, at 85%.

     The accuracy scores for the use ofthe passives are based on such low frequencies in the essays thatthese scores are not considered.

     After plotting the accuracyscores, a Fisher Exact Probability Test was performed on the datato determine statistical significance. This statistic allowed usobserve any statistical significance in our small sample size bylooking at scores above and below the mean. We compared the scoreson the first and final essays (1 and 8) to determine if there wasany improvement after receiving error correction. We also comparedSection 1 and Section 2 to determine if error correction resultedin an improvement in the two structures being marked for eachsection. No significant difference was found for any of thesecomparisons at the .05 level. (See Appendix for Fisher ExactTables.)

DISCUSSION

     According to the results ofthe study, the answer to the research question is that correctinggrammatical errors in students' essays does not result in improvedaccuracy in the use of those structures in subsequent revisions orin new essays.

     However, there are severallimitations of the study which must be considered before grammarcorrection is abandoned completely. One such limitation is thesmall sample size of the study. Obviously with such few students itis difficult to make generalizations about the treatment, and wewould like to see this study replicated with a larger sample size.We had hoped for a larger sample size since the class was capped attwelve students; however, these sections were added to the courselistings at the last minute and did not receive fullenrollment.

     Another limitation to consider isthe infrequent occurrence of three of the structures underinvestigation. As mentioned earlier, the obligatory occasions forpast tense, third person -s, and passive voice were very low. Thismay have been due in part to avoidance, but it was also due inlarge part to the nature of the writing under investigation. Theexpository writing the students were doing did not present manyobligatory occasions for the use of past tense or third person -s.In regards to the passive voice, students may have avoided it dueto the complexity of the structure. In order to determine thenormal frequency of these structures in expository writing, abaseline count should be conducted using the texts which were readby the students.

Obligatory Occasions

     The one structure that did see ahigh number of obligatory occasions (i.e. plurals) was the one inwhich Section 2 demonstrated an acquisition-like level of accuracywhich made comparison between sections impossible. One explanationfor this consistently high accuracy score might be that, ingeneral, Section 2 seemed to consist of more competent writers.This is reflected in the final grade averages for the two sections.In Section 1 the average final grade was C+ while in Section 2 itwas B. One other reason for Section 2's high plural performance mayhave been the presence of two native Spanish speakers who have asimilar plural marking system. On the other hand, Section 1consisted only of native speakers of Asian languages which do nothave pluralization rules similar to English.

Obligatory Occasions

     One interesting observation aboutthe accuracy scores was the consistently lower scores on thein-class writings. This is not surprising given that studentswriting at home do not have the same time constraints and that theyalso have the opportunity to edit their essays more carefully. Thisevidence, however, calls into question the use of in-class writingsif one is concerned primarily with grammatical accuracy.

     Besides simply investigating ourresearch question, we also learned a few other things during thisstudy. One is how difficult it is to be consistent in markingerrors, especially under real-time constraints. My colleague and Itried to cross-check the errors which received feedback before Ireturned the essays; however, this was not always possible. For onein-class essay, as we did an error analysis at the end of thesemester, we found that several errors which should have receivedfeedback did not. Even though we were looking for these errors, theconstraints of reading student handwriting, the pressure of needingto return the essays promptly, and the general craziness ofmid-semester prevented us from catching all of them. So even ifconsistent correction of one or two grammatical errors isbeneficial, it can be difficult to catch all of them in the realclass environment.

     Another consideration in choosingwhich error(s) to focus on is the importance of the error. Manytimes as we were correcting errors, we had the feeling that thesewere really insignificant errors which did not hindercommunication. If a teacher chooses to correct all the occasions ofone grammatical feature there will be instances where the error isminor or perhaps even debatable. In this instance it does not seemthat the students’ best interests are being served by asking themto focus on minor points of grammar.

Accuracy Scores

     One other thing to consider isstudents' attitudes to grammar correction. Many students want orexpect grammar correction on their essays. One of my students wrotein her journal, "I don't mind if you correct my paper when you readbecause I think when you correct my paper that help me to learnmore about writing." Also several times at the beginning of aconference, students would ask how their grammar was. The lack ofgrammatical feedback seemed to make them nervous. Truscott'ssolution to this is to retrain students to realize that grammaticalcorrection is not beneficial and to free them from this tyranny.Leki (1991) says that foreign students may come to the US withpreconceived ideas about grammar correction based on their previouslearning experiences. "ESL students may see the purpose of theirwriting classes in the U.S. as primarily to help them eliminateerrors." My students also felt that error correction was important;however, when asked what their English teacher usually does in thisregard, more than half could not identify the type of correctionused on their essays.

IMPLICATIONS

     One of the reasons given byteachers for disagreeing with the findings about the usefulness oferror correction was the large number of variables in grammarcorrection which they felt have not been considered. One variablethey mentioned was the type of correction given. In our study weunderlined and corrected the error, but perhaps this is not aneffective means of giving grammar feedback. Indeed Aljaafreh andLantolf (1994) indicate that giving narrowing spirals ofgrammatical feedback in the Vygotskian Zone of Proximal Developmentdoes seem to be effective. Teachers then argue that it is the kindof feedback which is ineffective, not correction as a whole.

     Another variable mentioned wastime. Students might not benefit from grammar correction in asemester, but over a longer time period they might show somebenefit. However, The question then becomes how much time isneeded? Is a student's improved accuracy in a grammatical structureattributed to error correction or to continued exposure to thetarget language?

     A whole host of individualdifferences was mentioned such as topic, student interest, transferof skills, level of students and learning styles. In fact we hadwanted to investigate learning styles initially, but foundoperationalizing the construct too difficult in the limited timeavailable. Our hypothesis was that analytic learners might benefitfrom grammar correction because it would be consistent with theirmethod of language learning. Holistic learners, on the other hand,might not benefit from such a method. This is an area for furtherresearch.

     One final thought to consider iswhat the student wants. Truscott argues that just because studentswant grammatical correction does not mean that we should give it tothem. Why give them something harmful even if they want it, is hisrationale. But as has been shown, many teachers and students dofeel that grammar correction is necessary. Do we as teachers havean obligation to give them what they want?

     In the end, the many questionssurrounding error correction call for teachers and researchers tocontinue to give critical attention to this issue. It is my hopethat this article has contributed to this inquiry, even if in onlya small way.

APPENDIX

Fisher Exact Probability Test

Number of individual students above and below the meanfrequency of correct use

I. Comparisons of First and Final Essays