Dr. Rod Ellis: TESOL Written Corrective Feedback

Professor Rod Ellis, gave a presentation which is available on YouTube.com. In it, he focuses on written corrective feedback. I’ve written a basic summary below. Get a drink, a snack, your notebook, make yourself comfortable and enjoy an allusive, informative explanation of the current state of affairs regarding written corrective feedback; the types and strategies, what we know, what we don’t know and what we should do.

[jwplayer file=”http://www.youtube.com/watch?v=wn35iHCljC8″]

Running time: 1:09:08

Why do we give written corrective feedback?

  1. To enable learners to revise their own writing, i.e. produce a better second draft
  2. To assist learner to acquire correct English

A Typology of corrective feedback types

  1. Strategies for providing corrective feedback
  2. How learners respond to the feedback

Written corrective feedback strategies

1. Direct written corrective feedback

Teachers provide correct form, i.e. crossing out an unnecessary word, phrase or morpheme, inserting a missing word, phrase or morpheme, inserting a missing word or morpheme, and writing the correct form above or near to the erroneous form (Ferris 2006)

  • Advantage – Provides learners with explicit guidance about how to correct their errors. Ferris and Roberts (2001) suggest direct¬†written corrective feedback is probably better than indirect written corrective feedback with writers of low levels of language proficiency.
  • Disadvantage – It requires minimal processing on the part of the learner and thus, although it might help them to produce the correct form when they revise their writing, it may not contribute to long-term learning.

However, a recent study by Sheen (2007)* suggests that direct written corrective feedback can be effective in promoting acquisition of specific grammatical features (Low intermediate level learners).

* The effect of focused written corrective feedback and language aptitude on ESL learners.

2. Indirect written corrective feedback

Involves indicating that the learner has made an error but without actually correcting it. This can be done by underlining the errors or using cursors to show omissions in the learners’ text or by placing a cross in the margin next to the line containing the error. In effect, this involves deciding whether or not to show the precise location of the error, i.e. just indicate which line of text the error is on.


  • Caters to ‘guided learning and problem solving’ (Lalande 1982) and encourages learners to reflect on linguistic forms
  • Considered more likely to lead to long-term learning (Ferris and Roberts 2002)


  • Learners cannot correct if they do not know the correct form
  • Learners may be able to correct but will not be certain that they are correct

The results of studies that have investigated direct vs. indirect written corrective feedback are very mixed (cf. Lalande 1982 and Ferris and Robert’s 2002). No study to date (2012) has compared the effects on accuracy in new pieces of writing.

3. Metalinguistic written corrective feedback

Provides learners with some form of explicit comment about the nature of the errors they have made.

  • Use of error codes, i.e. abbreviated labels for different kinds of errors placed over the location of the error in the text or in the margin. e.g. art = article, prep = preposition, sp = spelling, ww = wrong word, t = tense, etc.
  • Metalinguistic explanations of their errors, e.g. numbering errors and providing metalinguistic comments at the end of the text.

Informal poll: learners were in favour of metalinguistic explanations but teachers were not. Rod Ellis suggested that it had something to do with hard work on the teachers’ part.

Studies on use of metalinguistic error codes

  • Lalande (1982) – A group of learners of L2 German that received correction using error codes improved in accuracy in subsequent writing whereas a group receiving direct correction made more errors. However, the difference between them was not statistically significant.
  • Robb at al (1986) – The use of error codes no more effective that three other types of written corrective feedback they investigated, i.e. direct feedback and two kinds of indirect feedback.
  • Ferris (2006) – Error codes helped learners to improve their accuracy over time in only two of the four categories of error she investigated, i.e. in total errors and verb errors but not in noun errors, article errors, lexical errors or sentence errors (e.g. word order errors).
  • Ferris and Roberts (2001) – Error codes helped learners to self-edit their writing but no more so than indirect feedback.

Overall then, there is very limited evidence to show that error codes help writers to achieve greater accuracy over time and it would also seem that they are no more effective than other types of written corrective feedback in assisting self-editing.

Studies on use of metalinguistic error explanations

This is less common than error codes. It’s time-consuming and calls for the teacher to be able to write clear and accurate explanations for a variety of errors.

Sheen (2007) compared direct and indirect metalinguistic written corrective feedback. Both were effective in increasing accuracy in the learners’ use of articles in subsequent writing completed immediately after the written corrective feedback treatment but the metalinguistic written corrective feedback proved more effective that the direct written corrective feedback in the long term, i.e. in a new piece of writing completed two weeks after the treatment.

Rod Ellis speculated that metalinguistic written corrective feedback forces learners to formulate some kind of rule about the particular grammatical feature and then they use this rule but it takes time for them to be able to use this rule effectively. Direct feedback might have an immediate effect but learners soon forget the correction, whereas if they’ve learned the rule, maybe it’s going to have a longer term effect on learners’ ability to avoid the errors.

4. Focus of the feedback

Focused vs. unfocused written corrective feedback

1. Focused written corrective feedback advantages, i.e. correcting just one type of error

  • provides multiple corrections of the same error
  • is more likely to be attended to by learners
  • is more likely to help learners to develop understanding of the nature of the error

2. Unfocused written corrective feedback advantage, i.e. correcting all or most of the errors

  • addresses a range of errors, so while it might not be as effective in assisting learners to acquire specific features as focused written corrective feedback in the short term, it may prove superior in the long term.

The distinction of focused and unfocused written corrective feedback applies to all of the previously discussed options. The bulk of written corrective feedback studies completed to date have investigated unfocused written corrective feedback. Sheen (2007) – Focused written corrective feedback, i.e. errors in the use of articles for the first and second mention, proved effective in promoting more accurate language use of this feature. However, to date (2012), there have been no studies comparing the relative effects of focused and unfocused written corrective feedback.

5. Electronic written corrective feedback

Extensive corpora of written English can be exploited to provide learners with assistance in their writing. Electronic resources provide learners with the means where they can appropriate the usage of more experienced writers.

An example of electronic written corrective feedback

“Mark My Words” (Milton 2006)

  1. An electronic store of approximately 100 recurrent lexico-grammatical and style errors that he found occurred frequently in the writing of Chinese learners
  2. A brief comment on each error an with links to resources showing the correct form
  3. Teachers use the electronic store to insert brief metalinguistic comments into learners’ text
  4. Learners consult the electronic resources to compare their usage with that illustrated in the samples of language made available. This assists learners to self-correct.
  5. An error log for each piece of writing, drawing learners’ attention to recurrent linguistic problems is generated

There has been no research to investigate whether this is effective or to investigate whether it has any actual effect on language acquisition, as measured in new pieces of writing.


  • Removes the need for the teacher to be the arbiter of what constitutes a correct form. Teachers’ intuitions about grammatical correctness are often fallible; arguably a usage-based approach is more reliable
  • Allows learners to locate the corrections that are most appropriate for their own textual intentions and encourages learner independence

6. Reformulation written corrective feedback

This involves native-speakers rewriting learners’ texts in such a way as ‘to preserve as many of the writers’ ideas as possible, while expressing them in their own words so as to make the pieces sound native-like’ (Cohen 1989: 4) The writers then revise their writing by deciding which of the native-speakers’ reconstructions to accept. In essence then, reformulation involves two options ‘direct correction’ + ‘revision’ but it differs from how these options are typically executed in the whole of the learners’ texts are reformulated thus laying the burden on learners to identify the specific changes that have been made.

Sachs and Polio’s (2007) study

  • This study compared reformulation and direct correction.
  • Learners were shown their reformulated/corrected stories and asked to study them for 20 minutes and take notes if they wanted.
  • One day later, they were given a clean sheet of paper and asked to revise their stories but without access to either the reformulated/corrected texts or the notes they had taken.
  • Both the reformulation and direct correction groups outperformed a control group. However, the correction group produced more accurate revisions than the reformulation group.
  • It should be noted, however, that reformulation serves also to draw learners’ attention to higher order stylistic and organisational errors.

Types of learner response

  1. Revision required
  2. No revisions required
    • Learners asked to study corrections
    • Learners just given back corrected text

Rod Ellis notes that learners may only look at their grade and nothing more if they aren’t required to study their corrected texts.

Ferris (2006) study

Ferris (2006) identified a number of revision categories in the re-drafts of 146 ESL learners’ essays. Out of the corrected errors:

  • 80.4% were eliminated in the redrafted compositions either by correcting the error or by deleting the text containing the error or by making a correct substitution.
  • 9.9% of the errors were incorrectly revised
  • 9.9% no change was made

Overall, research shows that written corrective feedback assists revision. Ferris’ descriptors were as follows:

Label Description
Error corrected Error corrected per teacher’s marking.
Incorrect change Change was made but incorrect.
No change No response to the correction was apparent.
Deleted text Student deleted marked text rather than attempting correction.
Substitution, correct Student invented a correction that was not suggested by the teacher’s marking.
Substitution, incorrect Student incorrectly made a change that was not suggested by the teacher’s marking.
Teacher-induced error Incomplete or misleading teacher marking caused by student error.
Averted erroneous teacher marking Student corrected error despite incomplete or erroneous teacher marking.

An important theoretical issue

Theories of language learning differ in the importance they attach to:

  • Noticing the feedback in input
  • Revising the correct linguistic forms in output

But no research has addressed this issue.

Chandlers’ (2002) study

This compared indirect written corrective feedback plus the opportunity to revise with indirect written corrective feedback with no opportunity to revise. Results:

  • Accuracy improved from the first to the fifth piece of writing significantly more in the group that was required to correct their errors than in the group that just received indication of their errors
  • This increase in accuracy was not accompanied by any decrease in fluency

However, this study cannot be used to claim that written corrective feedback with revision contributes to L2 learning as there was no control group, i.e. a group that received no written corrective feedback. Rod Ellis notes that a great weakness of studies that have investigated written corrective feedback is that the studies have had no control groups and this makes it very difficult to say whether the written corrective feedback is actually having any effect on learning.


The situated nature of written corrective feedback

Hyland and Hyland (2006) commented, ‘it may be … that what is effective feedback for one student in one setting is less so in another’ (p.88).
A sociocultural perspective on written corrective feedback would emphasise the need to adjust the type of written corrective feedback offered to learners to suit their stage of development (Aljaafreh and Lantolf 1994) although how this can be achieved practically remains unclear in the case of written corrective feedback.

By teachers

Teachers need to consider the various options and formulate an explicit policy for correcting errors in learners’ written work. They also need to subject their policy to evaluation by evaluating the effects of their error correction, e.g. through action research.

By researchers

There is an obvious need for carefully designed studies to further investigate the effects of written corrective feedback in general and of different types of written corrective feedback. Guenette (2007) observed that is is important that studies are conducted in a way that make them comparable but sadly that has not typically been the case. A typology of written corrective feedback provides a classification of one of the key variables in written corrective feedback studies – the type of written corrective feedback – which can serve as a basis for research.