Effortless English Archives

Automatic English For The People

Monday, May 09, 2005

Error Correction- Ineffective

[More good stuff from SDKrashen.com]

by Stephen Krashen


The correction controversy is closely related to the grammar controversy. As I understand it, correction helps us fine-tune and adjust our consciously learned grammar rules. In his review of the literature, Truscott (1996) has concluded that correction has no effect on grammatical accuracy; in a previous ETA paper, I also reviewed this research and came to similar conclusions – correction only seems to help when students are tested on tests in which the conditions for Monitor use appear to be met, e.g. a grammar test.

Another way of determining whether grammar correction is effective is to look at studies in which students are corrected on their writing and then are asked to rewrite the same paper, taking the corrections into consideration. I have found four studies of this kind. In three studies, Fathman and Whalley (1990), Ashwell (2000), and Chandler (2003), subjects were fairly advanced students of EFL who had had considerable instruction in formal grammar, and who, we can assume, believed in conscious learning. In a fourth, Gascoigne (2004), subjects were first year university students in the US studying French. In these studies, the students had the advantage of having the corrections in front of them and had plenty of time. Because the paper was already written, students did not have to think about meaning at all but could focus on form, and they were graded on their grammatical accuracy. In these cases, correction was given the maximum chance to work; all conditions for the use of the conscious Monitor were met. Even under these optimal conditions, the impact of correction was very modest.

Subjects in Fathman and Whalley (1990) were intermediate ESL college students in the US. Students wrote compositions that described a series of pictures. We examine here two groups that were corrected: One group received correction only, the other correction plus feedback on content. Correction was limited to grammar, and consisted "solely of underlining all grammar errors (e.g. verb forms, tenses, articles, agreement). Thus students were told the location of their errors only and were not given information on the kinds of errors or shown the correct forms" (p. 182). Students wrote their compositions in class (they were given 30 minutes), the corrected versions were returned "a few days later" (p. 182) and students were given 30 minutes to rewrite.

Students wrote approximately the same number of words on each version, about 220 words in the first draft and about 250 words in the correct draft. They were able to correct only about half of their errors.

Ashwell (2000) compared the effect of correction on form with comments on content to determine if there was an optimal order (which should come first). Here, I focus only on the effect of correction, ignoring whether correction came before or after comments on form. I focus specifically on two of the subconditions. In both, subjects wrote 500 word compositions outside of class, and errors were then corrected, with correctors spending 12 minutes on each paper. The correction was "indirect feedback," that is, "underlining or circling grammatical, lexical, and mechanical errors or … using cursors to indicate omissions" (p. 233). Students had a full week to return their revised papers. The assignment was part of regular classwork.

In both conditions, students were able to correct only about one third of their errors (table 2).

Students clearly paid attention to the corrections. For all conditions of the study, students acted on 75% of the formal corrections, and 88% of the formal changes they made were in response to the corrections.

One of the conditions in Chandler (2003) also appears to be a case of students’ rewriting the same paper after correction. In this study, students were taking advanced ESL classes at a music conservatory in the US, and all "had had quite a bit of training in English grammar" (p. 272). Students had every reason to be careful: Accuracy in writing was a component of their grade in the class. Students had several days to make corrections.

Students wrote about eight pages of text and received four different kinds of feedback. In the "correction" condition ("full correction" in table 3), students were provided with the correct form, in the "underline" condition only the location of errors was indicated, as in the previous two studies. In the "describe" condition, a margin note was written indicating the kind of error made in the line it was made (e.g. "punc"), but the precise location was not given. All abbreviations had previously been explained in class and students received a list of the abbreviations. Finally, in the underline/describe condition, both the kind of error made and its precise location were indicated.

As indicated in table 3, with full correction students were able to correct nearly 90% of their errors. It should be noted, however, that all students had to do was copy the teacher’s correction. The other conditions produce results that are quite similar to what we have seen before.

Summary of Correction Studies

These studies represent the most optimal conditions for correction to work: All students were university-level and were able to understand grammar. All were motivated to do well, in some cases grades were at stake. All had plenty of time, from 30 minutes to one week to make corrections and all had access to their grammar texts. All they were asked to do was rewrite their own corrected essay. Thus, all conditions for Monitor use were met.

When students are told only where the error is, they can only correct from 1/3 to 1/2 of their errors. They get better when given more information, but even when they are given the actual rule, and need only copy, they still miss 10% of the errors. This is hardly a compelling case for correction.

Ferris (2004) claims that successful editing of one’s text in the short term is "likely a necessary, or at least helpful, step on the road to longer term improvement in accuracy" (p. 54). It is considered a given that students’ accuracy improves when editing from one draft to the next. The "big question," according to Ferris, is whether correction helps students improve over time. My conclusion is that we have not even provided a positive answer to the "little question," whether correction under optimal conditions works even in the short- term.