Oral Corrective Feedback in

Language Pedagogy and SLA

Rod Ellis
University of Auckland
What is corrective feedback?
Corrective feedback (CF) takes the form of
responses to learner utterances that contain (or
are perceived as containing) an error.
It occurs in reactive form-focused episodes
consisting of a trigger, the feedback move and
(optionally) uptake.
Speaker Utterance Move
Student I went to the train Trigger
station and pick up my
aunt.
Teacher Use past tense Feedback
consistently

Student I went to the train Uptake

station and picked up
my aunt
Why has corrective feedback
attracted so much interest?

1. In language pedagogy – the

importance of grammatical
correctness
2. In SLA – the role of negative
evidence and ‘pushed output’
An interface issue
The study of corrective feedback in
SLA allows for an evaluation of
common pedagogical claims about
whether, when and how to correct
learners’ errors.
Pedagogical positions
The importance of positive feedback
Teacher guides often emphasise the need to provide
positive feedback.

Positive feedback serves two functions (Nunan, 1991:195)

 ‘to let students know they have performed correctly’
 ‘to increase motivation through praise’

Correcting students is seen as potentially dangerous

because it can damage learners’ receptivity to learning. It
needs to be given ‘in an atmosphere of support and warm
solidarity’ (Ur, 1996; 255).
Five central questions
1. Should learners’ errors be corrected?
2. When should learners’ errors be corrected?
3. Which errors should be corrected?
4. How should errors be corrected?
5. Who should do the correcting?

(Hendrickson, 1978).
Should learner errors be corrected?
The importance attached to correcting error varies in
different methods:
 audiolingualism ‘negative assessment is to be
avoided as far as possible since it functions as
‘punishment’ and may inhibit or discourage learning’
 humanistic methods ‘assessment should be positive
or non-judgemental’ in order to ‘promote a positive
self-image of the learner as a person and language
learner’
 skill-learning approaches ─ ‘the learner needs
feedback on how well he or she is doing’
(Ur 1996; 243).
Not in fluency work
Harmer (1983) argued that when students
are engaged in a communicative activity, the
teacher should not intervene by ‘telling
students that they are making mistakes,
insisting on accuracy and asking for
repetition etc.’ (p. 44).
Correcting errors should be largely restricted
to accuracy work.
When should learner errors be
corrected?
Teachers have the option of either correcting
immediately an error occurs or making a note of
the errors and delaying correction until later.
 Immediately in accuracy activities
 Delayed in fluency activities

Teacher notes accompanying course books

frequently instruct teachers to leave correction
until the end of fluency activities (Hedge, 2000).
Which errors should be corrected?
‘Learners can only use just so much
feedback information: to give too much may
simply distract, discourage and actually
detract from the value of learning’ (Ur, 1996;
255).
Selective correction is both more practical
and more supportive of students’ feelings
(Katayama, 2007)
Deciding which errors to correct
Teacher guides discuss this question in terms of:
 Errors versus mistakes (Corder, 1967)
 Global versus local errors (Burt, 1975)
 Errors involving simple versus complex features
(Krashen, 1982).
 Persistent vs. occasional errors
But none of their proposals are easy to implement.
How should errors be corrected?
Teacher guides suggest a variety of strategies for
correcting errors.
 Questioning the learner (e.g. “the teacher may say ‘Is that
correct?’”–Harmer, 1982:63).
 Direct indication (e.g. “Tell the students that there is an
error”–Scrivener, 2005:300).
 Requesting clarification (e.g. “the teacher looks puzzled
and requests clarification”–Hedge, 2000:291).
 Requesting repetition (e.g. “the teacher simply asks the
student to repeat what he has just said”–Harmer, 1982:62).
 Echoing (e.g. “the teacher may echo what the student has
just said with a questioning intonation”–Harmer, 1982:62).
Corrective feedback strategies in
the teacher guides
The guides:
 simply provide lists (i.e. they do attempt to
classify the strategies into general types)
 do not provide examples of these strategies
taken from actual classroom interaction (i.e.
they just provide simple descriptions of
them).
What the guides advise
 Use a variety of corrective strategies (Hedge,
2000)
 Use strategies that require learners to correct
their own errors.
‘The object of using correction techniques is to give the
students a chance to get the new language right’
(Harmer, 1983:63).
‘People learn more by doing things themselves rather
than being told about them’ (Scrivener, 2005:3).
 the importance of ‘encouraging, tactful
correction’ (Ur, 1998:249).
Who should do the correcting?
Three possibilities:
 the teacher,
 the student who made the error
 another student.
What the guides recommend
Hedge (2000) and Scrivener (2005)
advised giving students the opportunity to
self-correct and, if that fails, inviting
another student to perform the correction.
The least favoured option in the guides is
teacher correction
A general picture
1. Teachers need to carry out correction sensitively to
avoid a negative emotional response in learners.
2. They should delay correction in oral fluency work..
3. They need to be selective in the errors they correct.
Various proposals for deciding which errors to correct
have been put forward but none are easy to implement
in practice.
4. Teacher educators are reluctant to recommend which
strategies teachers should use but favour those that
induce learners to correct their own errors.
5. As far as possible, it is the students who should do the
correction not the teacher.
A final comment
The pedagogical advice found in the
teacher guides is based on the authors’
own experience of teaching and ‘received
opinion’ about corrective feedback.
There is no reference to any research on
corrective feedback.
Corrective Feedback in SLA
Theoretical positions
1. UG-based accounts of corrective
feedback
2. Cognitive-interactionist accounts
3. Sociocultural theory and corrective
feedback
UG-based accounts
Universal Grammar - a highly abstract set of
linguistic principles that act as constraints on the
form that the specific rules of a language can
take.
The poverty of stimulus argument - the input that
learners are exposed to is insufficient to ensure
full acquisition of a target language grammar and
thus UG is required to provide an ‘explanation of
how it is that learners come to know properties of
grammar that go far beyond the input’ (White
2003:20).
The role of negative evidence
Different positions:
 UG can only operate on positive evidence
(Schwartz, 1993) so corrective feedback (=
negative evidence) plays no role in language
learning
 Negative evidence can trigger UG principles
and may be needed to help learners overcome
persistent errors.
 Negative evidence only plays a role in the
intermediate stages of L2 acquisition (Carroll,
2001)
Cognitive-interactionist accounts
Cognitive-interactionist theories emphasise
that CF is most likely to assist acquisition
when learners are focused primarily on
meaning in the context of producing and
understanding messages in
communication, commit errors and then
receive feedback that they recognize as
corrective.
Two functions of CF
CF facilitates the processes responsible for
acquisition in two ways:
1. by providing learners with positive evidence
of target language forms
2. by pushing learners to self-correct their
errors (i.e. though output).
Recasts provide positive evidence
T: When were you in school?
L: Yes. I stand in the first row? (trigger)
T: Oh, you stood in the first row. (corrective
move)
L: Yes, in the first row.
Prompts push learners to self-
correction

S: Why does he fly to Korea last year? (trigger)

T: Pardon? (corrective move)
S: Why did he fly to Korea last year? (uptake)

(Yang & Lyster, 2010:234-4).

A controversy
Long (2006) – recasts are the more effective
because they provide learners with both
negative and positive evidence. They point out
that unless learners receive positive evidence it
will be impossible for them to acquire ‘new’
linguistic forms.
Lyster (2004) – according to skill-learning theory
prompting learners to self-correct is more
effective because it helps learners to gain
greater control over partially acquired linguistic
features.
Combining prompts and recasts
L: I think that the worm will go under the soil.
T: I think that the worm will go under the soil?
L: (no response)
T: I thought that the worm would go under the
soil.
L: I thought that the worm would go under the
soil.
(Doughty & Varela, 1998)
Explicit versus implicit feedback
Implicit corrective feedback
the corrective force of the feedback is not
overt (e.g. recasts; clarification requests)

Explicit corrective feedback

the corrective force is signalled linguistically
(e.g. metalinguistic explanation or
elicitation)
Explicit feedback – an example
L: He kiss her
T: No, kissed - past tense.
L: He kissed her
Another controversy
The case for implicit types of CF is based
on the claim that they do not interrupt the
communicative flow of an interaction to
the same degree as explicit types.
Explicit CF, however, has the advantage of
being more likely to be attended to by the
learner.
Sociocultural theory and corrective
feedback
 Correction is not something done to learners
but rather something carried out with learners.
 It enables the joint construction of a Zone of
Proximal Development – a sociocognitive state
where learners are assisted to use linguistic
features that they cannot employ
independently.
 Corrective feedback needs to be ‘graduated’ –
that is, it must provide the learner with the
minimal level of assistance needed to construct
a ZPD.
A regulatory scale
Aljaafreh and Lantolf (1994) developed a
‘regulatory scale’ to reflect the nature of the
graduated assistance that occurred when a
tutor helped learners to identify and correct
errors in an oral conference. This scale was
based on a continuum of corrective
strategies employed by a tutor, reflecting
how explicit or implicit the strategies were.
 Types of corrective feedback
Implicit Explicit
Input-providing Conversational recasts Didactic recast
Explicit correction
Explicit correction +
metalinguistic explanation
Output-prompting Repetition Metalinguistic comments
Clarification requests Elicitation
Paralinguistic signal

(adapted from Lyster, Saito & Sato, 2013:3)

Corrective feedback research
1.Does CF assist L2 acquisition?
2.Which type of CF is most effective in
assisting L2 acquisition?
3.Does learner self-correction following
CF (i.e. uptake) contribute to L2
acquisition?
4. How do linguistic, contextual and
learner variables influence the
effectiveness of CF?
Does CF assist L2 acquisition?
Li’s (2010) meta-analysis (33 studies involving 1,773 learners):
 corrective feedback had a medium effect on acquisition.
 this effect was evident in tests immediately following the
treatment involving CF and over time.
 the effect was much greater in studies carried out in a
laboratory than in a classroom.
 the effect of CF was greater in foreign language than in
second language settings
 CF proved more effective in treatments that involved discrete-
item practice of grammatical structures (e.g. in drills) than in
communicative activities.
 the effects of CF were evident in both tests that measured
controlled language use and free production.
Aljaafreh and Lantolf’s study
The degree of scaffolding provided by the tutor for
a particular learner diminished over time (i.e.
whereas at one time the instructor needed to
correct quite explicitly to enable a learner to self-
correct, at a later time more implicit correction
sufficed). In accordance with how learning is
conceptualized in sociocultural theory, they argued
that this demonstrated that learning was taking
place.
Which type of CF is most effective in
assisting L2 acquisition?
 Research conducted within a cognitive-interactionist
framework has investigated input-providing vs. output-
prompting and implicit vs. explicit CF. It assumes that not
all types of CF are equally effective and therefore, the
primary goal of CF research is to establish which type
works best.
 Research conducted within a sociocultural framework is
based on the assumption that for CF to be effective it needs
to be systematically tailored to the learner’s developmental
level. From this perspective there is no one type of CF that
will work best
Cognitive-interactionist research
Some key studies:
 Ellis, Loewen & Erlam (2006)
 Lyster (2004)
 Lyster & Mori (2006)
 Mifka-Profozic (2012)
These studies suggest that the context of the
instruction influences which type of CF is most
effective. In general, though, explicit CF is more
effective than implicit CF.
Sociocultural research
Key studies:
 Aljaafreh & Lantolf (1994)
 Nassaji & Swain (2000)
 Erlam, Ellis & Batstone (2013)

General conclusion – graduated feedback is

effective but possibly not more so than simple
explicit feedback.
Does learner self-correction following CF
(i.e. uptake) contribute to L2 acquisition?
 Lyster & Ranta (1997) – learners more likely to
repair their errors following prompts than recasts.
 But little research on whether this facilitates
acquisition.
 Loewen (2005) – uptake predicted adult ESL
learners’ test scores.
 A possible conclusion – learners benefit from CF
even if they do not repair their errors but when they
do ‘deeper processing’ may occur which may assist
learning.
 Mediating variables
1. Linguistic targets
• CF more likely to be attended to if it targets vocabulary or
pronunciation than morphosyntactical features
• Regular versus irregular forms (Yang & Lyster, 2010)

2. Instructional context
• Recasts more salient in a form-focused instruction context

3. Learner variables:
• Age
• Proficiency (prompts more effective for less proficient
learners but prompts and recasts equally effective for more
proficient learners – Ammar & Spada (2006)
• Anxiety (Sheen, 2008)
Re-examining the role of corrective
feedback in language pedagogy
Should learners’ errors be
corrected?
 The reservations that some teacher
educators have expressed about CF is
not supported – CF ‘works’
 The claim that CF is best kept for
accuracy-based activities is not
supported – CF is needed in fluency-
based activities as well
When should learner errors be
corrected?
 There is no support for the proposal that
correction should be delayed until after a
fluency-activity has been completed.
 The research supports providing CF in
communicative interactions.
But no research to date has investigated whether
‘immediate’ or ‘delayed’ CF is more effective.
Which errors should be corrected?
 There is no basis in the research for
focusing only on ‘global’ errors – CF
directed at ‘local’ errors has been
shown to be effective.
 The research indicates the
effectiveness of a ‘focused’ approach to
CF
How should errors be corrected?
 The SLA research is of greatest value in
helping to address this question:
 The treatment of CF in the teacher guides
provides no theoretical justification for the
choice of strategy.
 In SLA, the classification of strategies into two
key dimensions (i.e. input-providing vs. output-
prompting and implicit vs. explicit) is
theoretically driven.
Some general guidelines for conducting CF
 Aim to provide intensive CF.
 Explicitness is important.
 Do not rely predominantly on recasts.
 Teachers should vary how they correct according to instructional
context. In a communicative activity, brief explicit forms of correction
may be needed. In a grammar exercise, recasts can be effective.
 Combine input-providing and output-prompting CF strategies in a
systematic way - e.g. corrective recasts.
 Encourage uptake with repair. (In this respect, the SLA research
lends support to the recommendation of the guides which emphasize
the need to ensure learners successfully uptake the correction).
Who should do the correcting?
 Teachers need not be wary of other-initiated/
other repair (i.e. CF initiated and completed by
the teacher). However, leaving time for learner
uptake of the correction can assist learning.
 Descriptive studies of CF in classrooms show
that despite the recommendation that teachers
should make use of peer-correction, this rarely
occurs.
Final Comment
CF is an aspect of instruction where the concerns
of teachers and interests of SLA researchers
coincide and, as such, constitutes an ideal
construct for examining the contribution that SLA
can make to language pedagogy.
The approach I have followed is not to ‘apply’
SLA research to language pedagogy but rather to
draw on it to evaluate common pedagogic claims
and thereby to encourage a reconsideration of
these.
 References

Ellis, R. (2012). Language Teaching Research and Language

Pedagogy. Malden, MA.: Wiley Blackwell.

Ellis, R. (2013). Exploring Language Pedagogy through

Second Language Acquisition Research. London: Routledge.

Lyster, R., Saito, K. & Sato, M. (2013). State of the art article:
Oral corrective feedback in second language classrooms.
Language Teaching 46, 1-40.
Thank you!
r.ellis@auckland.ac.nz