You are on page 1of 40

Written Corrective

Feedback: Does it
Work?
Rod Ellis
Anaheim University
The International Research
Foundation for English
Language Education (TIRF)
Why investigating corrective
feedback is important?
1. Theoretical significance:
Theories differ with regard to the importance they
attach to corrective feedback (cf. Krashen and N.
Ellis).
Theories also differ with regard to the type of
corrective feedback they hypothesize is effective
(e.g. implicit vs. explicit types).
2. Pedagogic significance:
Teachers need to know whether to correct their
students’ errors and how best to do error
correction.
Oral and Written corrective
feedback
Studies or oral corrective feedback have
been undertaken by second language
acquisition researchers concerned with
whether it assists the development of
linguistic competence.
Studies of written CF have been undertaken
by writing researchers concerned with
improving the teaching of writing.
Oral and written corrective
feedback compared
Implicit or Offline or Direct or
explicit? online? indirect?

Oral CF Both Both – Both


usually
online.

Written CF Invariably Invariably Both


explicit offline.
The debate

1. Truscott (1996; 1999 argued that correcting


learners’ errors in a written composition may enable
them to eliminate the errors in a subsequent draft
but has no effect on grammatical accuracy in a new
piece of writing (i.e. it does not result in acquisition).

2. Ferris (1999) argued that it was not possible to


dismiss correction in general as it depended on the
quality of the correction – in other words, if the
correction was clear and consistent it would work.
The need for evidence
Truscott (1999):

Teachers must constantly make decisions about what to do –


and what not to do – in their classes. These decisions are
necessarily made under conditions of uncertainty: research
never puts an end to doubt. But the choices still must be made,
and made constantly. So given the world as it is, the best we can
hope for is that teachers will look seriously at the case against
grammar correction, compare it to the case for correction, decide
which is the stronger, and then incorporate that decision in their
teaching (p. 121)
The Evidence: Studies Without a
Control Group
 Chandler (2000 and 2003)
 Ferris (1995 and 1997)
 Ferris, Chaney, Komaru, Roberts and McKee 2000);
 Lalande (1982)
All these studies reported improvement in grammatical accuracy
following corrective feedback.
However, as Truscott (1996) has pointed out, such studies
cannot be used to claim that CF is effective as it is always
possible that improvement would have taken place without any
CF. A control group is essential to demonstrate that CF is
effective.
The Evidence: Studies With a
Control Group
Ashwell 2000;
 Fathman and Whalley 1990;
 Ferris and Roberts 2001;
 Kepner 1991;
 Polio, Fleck and Leder 1998;

Design problems:

 Some studies (e.g. Fathman and Whalley 1990; Ashwell 2000) did not examine
the effect of CF on new pieces of writing; that is, they only demonstrated that CF
assists learners to achieve greater grammatical accuracy in a second draft of
the written composition that had been corrected.
 Other studies included no pre-test (Kepner 1991) or used different instruments
in the pre- and post-tests (Polio, Fleck & Leder 1998).
Three Later Studies: (1) Bitchener
et al.
Bitchener, Young and Cameron (2005) investigated
the effects of direct written CF combined with
feedback in five-minute oral conferences with
individual writers on three types of error
(prepositions, the past simple and the definite
article). They found that the CF led to improved
accuracy on both the use of past simple and articles
(but not prepositions) over a 12-week period. This
study, however, only testifies to the joint effect of
written and oral CF on learners’ writing.  
Three Later Studies (2): Sheen
Sheen (2007) investigated the effects of written CF
on intermediate ESL learners’ use of English articles
in narratives. A strength of this study was that
acquisition was measured in three different ways –
by means of an error correction test, a speeded
dictation test, and a new piece of writing. The
effects of the CF were evident in statistically
significant gains on all three tests in comparison to a
control group.
Three Later Studies: Bitchener
Bitchener (forthcoming) also investigated the
effects of CF on English articles in a carefully
designed longitudinal study and similarly
showed that it resulted in statistically
significant gains in accuracy in comparison to
a control group over time.
The effects of different types of
written CF (1): Direct vs. indirect
 ‘direct’ refers to CF that supplies learners with the correct target
language form when they make an error; ‘indirect refers to
various strategies (e.g. simply indicating errors) to encourage
learners to self-correct their errors.
 Five studies have compared the effectiveness of these two types
of CF (Chandler 2003; Ferris & Helt 2000 [1]; Lalande 1982;
Robb, Ross & Shortreed 1986; Semke 1984) but with very mixed
results - two studies reported no difference, two reported in
favour of indirect CF and one reported in favour of direct CF.
 The effectiveness of ‘indirect’ CF depends on whether the
learner possesses the knowledge needed to self-correct.
The effects of different kinds of written
CF (2): Direct vs. metalinguistic
 direct feedback entails supplying learners with the
correct target form; metalinguistic feedback involves
providing some kind of metalinguistic clue as to the
nature of the error that has been committed and the
correction needed.
 Sheen (2007) compared direct CF alone and direct
CF in combination with metalinguistic CF. Direct CF
in combination with metalinguistic CF was more
effective than direct CF alone.
The effects of different kinds of written
CF (3): Unfocused vs. focused
 In unfocused CF teachers correct all (or at least a range of) the
errors in learners’ written work. Focused CF selects specific
errors to be corrected and ignores other errors.
 Learners are more likely to attend to corrections directed at a
single (or a limited number of ) error type(s) and more likely to
develop a clearer understanding of the nature of the error and
the correction needed.
 Sheen (2007) and Bitchener (forthcoming) both examined
focused CF
 No study to date has compared focused and unfocused CF
Designing a Written CF Study
Key design features:
1. There must be a control group.
2. The needs to be a pre-test, an immediate post-test and a delayed
post-test.
3. The tests used must reflect the kind of language use the CF is
designed to improve.
4. The study must investigate the effects of CF on new pieces of writing
not just on redrafting.
5. The study needs to show the effect of CF on specific linguistic
features.
6. The study needs to demonstrate that CF does not cause learners to
try to avoid using the target feature.
7. The study needs to show that the CF does not have a deleterious
effect on some other aspect of writing (e.g. fluency).
Research Questions
1. Does written CF help L2 learners to
become more accurate in the use of the
English indefinite and definite articles to
express first and second mention?
2. Is there a difference in the effect of
unfocused and focused CF directed at
using the indefinite and definite articles to
express first mention and anaphoric
reference?
The target structure
Use of articles (to express first and second mention).

Grammatical and ungrammatical examples of this taken from the data


collected for the study are as follows:

There is a dog. The dog felt very hungry.

* One day the dog stole a bone from the butcher. The dog
escaped. (‘a’ needed for first mention).

* The dog tried stealing a bone. He biting a bone. (‘the’ is


needed for second mention).
Reasons for choosing articles
 Obligatory occasions in certain types of
discourse (e.g. narratives) are plentiful
 Learners have ongoing difficulties with
articles
 The use of articles for first and second
mention permits a ‘rule of thumb’
Design
 Quasi-experimental design: two experimental groups - focused
CF (N = 18), unfocused CF (N = 18) - and a control group (N =
13).
 All three groups completed a pre-test, an immediate post-test
and a delayed post-test, where all the tests involved narrative
writing based on picture stories.
 In addition, all three groups completed an error correction test
prior to the treatment and immediately following the treatment.
 The two experimental groups received error correction on three
written narratives. One group received focused correction and
the other unfocused correction.
Participants
The participants were 49 students enrolled in
general English classes in a national
university in Japan. Their proficiency was
intermediate. They were a mixture of male
and female.
Treatment
The students completed three short animal stories:

1. The teacher handed out a short animal story together with pictures
and told the students that they were going to read the story and then
rewrite the story.
2. Students were asked to read the fable silently.
3. The teacher explained key words and discussed the story with the
class.
4. The teacher read the story aloud once to refresh their memory.
5. The teacher collected in the animal stories (but let the students keep
the pictures) and then gave out a blank writing sheet. She told the
students to write a title for the story and then to retell the story with as
much detail as they can remember.
6. The teacher collected the students’ written stories for correction.
The corrective feedback
The teacher corrected each student’s story
and handed it back to the students. The
students were given time to examine the
corrections and then wrote the next story in
the same lesson.

Two types of correction provided – focused


and unfocused.
Direct/ unfocused correction

This involved correcting article errors together with other types of error
(e.g. errors in past tense, prepositions and vocabulary) by indicating
and correcting the errors on the student’s stories.

For example in the student story below it involved inserting ‘a’ before
‘bone’ and ‘butcher’ in the first sentence and crossing out ‘having’ and
inserting ‘the’ before ‘bone’ in the second sentence and crossing out ‘at’
and writing ‘to’ over it.

A dog stole bone from butcher. He escaped with having bone.

Up to 5 errors were corrected two of which were always article errors.  


Direct/focused correction
This was the same as for direct/ unfocused
except that ONLY article errors were
corrected.
The aim was to correct between 3 and 5
article errors involving first/second mention
but this was not always possible as some
students produced few errors.
Narrative writing tests
Three different picture compositions were used, all
taken from Byrne 1967); (1) ‘The Dog and the
Bone’, (2) ‘The Blind Man’ and (3) ‘A Thief Gets
caught’.
Each picture composition consisted of four pictures.
These picture compositions were chosen because
they involved reference to two or more people/
objects and thus created contexts for the use of the
target structure.
At each testing time approximately one third of each
group completed a different story.
Error Correction Test
This consisted of 15 items, each containing
two related statements, one of which was
underlined. The underlined sentence
contained an error. The students were asked
to write out the incorrect sentence correctly.
12 of the 15 items contained sentences with
article errors (6 involving ‘a’ and 6 involving
‘the’). There were also three distractors.
Exit Questionnaire
The learners were asked to complete an exit
questionnaire after they had completed the
treatment tasks and tests. The purpose of
this was to establish whether they had
recognized that to focus of the writing they
had completed was on articles. The
questionnaire consisted of one multiple
choice question and an open question.
Scoring
1. Writing test scores were calculated using
obligatory occasion analysis (expressed as
a proportion of 1).
2. Error correction tests scores were scored in
terms of whether the learners successfully
corrected the article error (total 12).
Results (1): Exit questionnaire
The overwhelming response of the students
to the exit questionnaire was that the tasks
were intended to practice their writing skills or
general English skills. Only 1 student
responded to the multiple choice question by
indicating that the purpose was to practice
grammar
Results (2): Writing Tests
1

0.9

0.8 Focus s ed
Unfocus s ed
0.7 Control

0.6

0.5
P re-tes t P os t-tes t Delayed P os t-tes t
Statistical analysis
1. No group differences in the pre-test scores.

2. No group differences on post-test 1.

3. Group differences in post-test 2 evident; both experimental


groups outperformed the control group.

4. No differences between the unfocused and focused CF


groups.

5. Only the CF groups improved over time.


Results (3): Error Correction Test

Estimated Marginal Means of MEASURE_1

10.00 Group
focussed
unfocussed
control
9.00
Estimated Marginal Means

8.00

7.00

6.00

5.00

4.00

1 2
time
Statistical Analysis
1. No significant differences in the pre-test
scores.
2. Significant differences in the post-test
scores: both experimental groups
outperformed the control group.
3. The two CF groups did not differ.
Discussion
1. Control group’s use of articles was unstable; CF groups more
consistent.
2. Exposure to written corrective feedback helped the learners to
use articles with greater consistency in subsequent writing
and, in most learners, to manifest gains in accuracy which
were durable.
3. In the long term these gains were significantly greater than
those achieved simply as a product of practice in narrative
writing.
4. These effects were evident despite the fact that CF groups
were not aware that the focus was on articles.
An example: Focused Group
Pre-test Post-test 1 Post-test 2

One day the dog steal One day, a man One day a blind man was
the bone from the steal the handbag taking walk on the
butcher and run from a lady and sidewalk. But there is a
away. When the dog run away. The big hole and sign which
reached at the middle lady shouted said ‘Danger’. Of course
of bridge, the dog ‘Please anybody he can’t see the sign and
looked at the river … catch him!’ The he was enclosing to the
man was running hole …
away front of an
old man …
An example: Unfocused Group

Pretest Post-test 1 Post-test 2

One day a dog An old man stoled a There was a hole on a


stole a bone handbag from an old sidewalk and was a
from butcher. woman. Soon the sign that announced
When he walked man began to run. there is a danger. One
on the bridge The woman shouted day a blindman walked
over the river he ‘Help! He is a the sidewalk. Of course
looked at surface thief!’ A old man the man can’t see the
and saw himself who was sitting on a sign …
with holding a bench …
bone …
Was there an avoidance strategy?

Obligatory occasions for articles in the three tests

Delayed Post-
Pre-test Post-test
test
Groups n Total Mean Total Mean Total mean

Focussed 18 235 13.1 255 14.17 270 15.0

Unfocussed 18 215 11.9 249 13.8 257 14.3


Control 13 155 11.9 204 15.7 179 13.8
Focused vs. unfocused CF?
The study did not show that one type of CF
was better than the other.

But the distinction between the two types of


CF may not have been clear enough i.e. not
all errors in the unfocused group were
corrected and at least 2 out of 5 corrections
were of articles.
Conclusion
The case for and against written corrective feedback does not rest
entirely on what research shows about its effectiveness but the
research evidence is clearly a major factor to be considered.

Truscott’s own rejection of written CF is based largely on what he sees


as an absence of empirical evidence demonstrating that it contributes to
acquisition as manifested in improvements in grammatical accuracy in
subsequent writing.

There are now a growing number of studies showing that written CF can
be effective. They are tipping the argument in favour of the case for
correction.

But research still limited (e.g. studies that have shown an effect for CF
have all examined articles).

You might also like