Mistake correction

Keith Johnson

Downloaded from https://academic.oup.com/eltj/article/42/2/89/372446 by University of Leeds user on 29 November 2020

The aim of this article' is to explore what a particular view of language,
language learning, and language teaching suggests with regard to a
specific language-teaching problem—the problem of what to do when
students get things wrong.

Language as skill The particular view of language, of language learning, and of language
teaching that is presented in this article is one which is prepared to make a
parallel between language and other complex skills like playing tennis,
piloting an aircraft, or playing a musical instrument. The justification for
such a parallel is that all these behaviours, including language use, involve
performing complex sequences of activities. The type of knowledge the
performer needs to develop for all these behaviours, including language
use, is knowledge concerned with how to (what Anderson 1980 calls 'pro-
cedural knowledge'), rather than knowledge about (what he calls 'declara-
tive knowledge'). The knowledge of a skilled language user and the
knowledge of a skilled oboe player have in common that they both involve
forms of procedural knowledge.
The view of language as skill, of language acquisition as skill acquisition,
and of language teaching as skill training, will offend many, who may find
the comparisons this article makes between language learning and learning
(for example) how to ride a horse, inappropriate if not offensive. The view
certainly needs more justification than can be given here.2 What may be
said here is that it is by no means a new view, though it is one which has
gone rather out of fashion in recent years (hence perhaps feelings of
inappropriateness and offence). It has gone out of fashion doubtless largely
through the influence of Chomsky and the view that language is unique
among human behaviours, acquired in a unique way by means of a
language-specific acquisition device (the LAD) which does not appear to
contribute much towards the acquisition of other, non-linguistic skills. This
view of language as 'unique and uniquely acquired' strongly suggests that if
we wish to know anything about how languages are learned, we shall get no
useful information from looking at how other skills are learned. According
to this view, the proper study of language acquisition is indeed language
acquisition.
Of course, this Chomskyian view both can be and has been challenged.
As Anderson (1980:398) says: 'little direct evidence exists to support the
view that language is a unique system'. And once language is deprived of its
unique status, then the acquisition of skills other than language becomes an
area of study likely to be of interest to the language teacher. Under the
Chomskyian influence, such interest has waned somewhat; this article is
part of an attempt to show how looking at language learning in terms of
skills may be fruitful in both theoretical and practical terms.

EL TJournal Volume 42/2 April 1988 © Oxford University Press 1988 89

F—dback The concept of feedback is central in the literature on skill acquisition. It is
. recognized that though there is a place in training for initial guidance in
skill learning, there is also an important place for feedback (viewed as the
provider of information, rather than as a reinforcer). It seems intuitively
true that a great deal of learning how to serve in tennis for example, comes
after any initial guidance the teacher might give, when the learner picks up
the ball, serves, and notes the outcome. The sequence of events, in this case,
is not learn - • perform, but learn —* perform —• learn. This sequence
correctly suggests that when we speak about feedback, we are speaking

Downloaded from https://academic.oup.com/eltj/article/42/2/89/372446 by University of Leeds user on 29 November 2020

about something that potentially contributes to the learning process. For
lengthy discussion on the concept of feedback, see Annett (1969).
Though the situation is better today, much language teaching of the past
exemplifies the learn —> perform sequence. We teach, and the students
learn; they then perform, exemplifying, we hope, the learning that has
taken place. During or following performance, error correction is used to
plug the holes. But approaching language teaching as skill training sug-
gests that feedback may have more of a role to play. A central aim of this
article is to suggest that more attention should be given to the issue of how
we can best provide feedback.
That more attention needs to be given to this issue is further suggested by
what most teachers will see as the comparative failure of the feedback
measures we employ. Our students leave the 's' off the third singular of the
simple present; we put it back on for them, and at the next opportunity they
leave it off again. One of my problems as a novice horse rider is that I lean
forward on the horse; the teacher tells me to sit up straight; a moment later I
am leaning forward again. In these cases our methods of feedback do not
seem to meet with much success.

Error* and mistakes To consider how things might be improved, we might begin by asking why
it is that students get things wrong. There are at least two reasons.3 One is
that the student either does not have the appropriate knowledge, or has
some false knowledge. He or she may either not know how a tense of English
works, or have the wrong idea. In this case, we may say that the student's
interlanguage knowledge is faulty. The result is what Corder (1981) calls an
error.
There is, however, a second reason for a student getting something
wrong. It may be a lack ofprocessing ability. I know I should not lean forward
on the horse, and when simply trotting round the paddock I do not do so.
My problem comes when approaching a small jump. My feet may fall out of
the stirrups, the horse may begin to get difficult, and one result (there may
be other more painful ones!) is that I lean forward. It is not my 'knowledge'
that is at fault here; it is my ability to 'perform my competence' (the phrase
is taken from Ellis 1985a) in difficult operating conditions. The result is
what Corder (1981) calls a mistake.
In recent years a number of writers, dealing with different areas in the
language learning/teaching field, have made distinctions which can be
related to Corder's between errors and mistakes. Bialystok (1982), for
example, takes the area of language testing as her starting point. She
observes that we have tended to assess language mastery quantitatively,
providing statements that 'the learner simply knows more or less of the
language, or knows some of the formal properties and not others' (p. 181).
But we should also, she argues, ask qualitative questions, about the condi-
tions under which these formal properties can be correctly manipulated.4

90 Keith Johnson
Two examples, one from a non-linguistic skill and one from language
use, will illustrate. A footballer may, in normal circumstances, be a good
goal scorer. But when we assess his mastery, we will need to take into
account circumstances which are far from normal. Can he, for example,
score in the Mexico World Cup, at an altitude of six thousand feet, against a
good side, knowing that spectators at home will bay for the blood of the
defeated? Similarly, when we come to judge a student's linguistic ability, we
would be foolish to pronounce that she has mastered the present perfect
tense simply on the grounds that she has managed to use it correctly in a

Downloaded from https://academic.oup.com/eltj/article/42/2/89/372446 by University of Leeds user on 29 November 2020

gap-filling task, done under 'ideal' conditions. Can she, one would need to
ask, use the tense correctly over a bad intercontinental telephone line, with
all attention focused on getting the message across in the shortest possible
space of time?
Returning now to the error/mistake distinction, having noted that it is
one manifestation of a more general distinction between knowledge and
processing ability, one might claim that we have paid more attention in
language teaching to errors than to mistakes. What is less arguably true is
that techniques (like, perhaps, explanation) for handling errors spring
more readily to mind than techniques for handling mistakes. It may further
be the case that we have tended to treat mistakes as if they were errors.
Since the two are different, it seems likely that they will need to be handled
in different ways.
Corder (1981:10) argues that 'mistakes are of no significance to the
process of language learning'. But if we use the word 'mistake' to describe a
malformation due to inability to process under difficult sets of operating
conditions, then it is likely that a good percentage of our students' malfor-
mations are mistakes and not errors. If this is the case, the subject of mistake
correction becomes an important one in language teaching.

Mlstmkm correction How can mistakes be eradicated? One might propose that in order to
eradicate a mistake, a student will need at least four things. These are:
a. The desire or need to eradicate the mistake. It is likely that a number of
mistakes do not get eradicated simply because students know they can get
by without eradicating them. The simple present 's' (which has little
communicative value) probably falls into this category.5
b. An internal representation of what the correct behaviour looks like. The
student needs, in other words, the 'knowledge' that makes the malforma-
tion a mistake and not an error. It is unlikely, of course, that the 'knowledge'
is possessed in a form in which the linguist will possess it; which is why it is
referred to here as an 'internal representation' (begging the question of
what that internal representation will look like).
c. A realization by the student that the performance he or she has given is
flawed. The learner needs to know that a mistake has occurred. Some form
of feedback will provide this.
d. An opportunity to repractise in real conditions.
In learning how to serve in tennis, then, the learner who has just served
badly needs (a) a desire to serve properly, (b) to know what a good service
looks and feels like, (c) a realization that the service was bad, and (d) the
chance to practise again.
This article will not deal with the first of these conditions, important

Mistake correction 91
though it is. It will consider how the remaining three might be provided in
the classroom.

Providing internal Initial guidance should help the student to form an internal representation
representation of what the behaviour is like (for example, how a particular structure
operates and is used in English). How such guidance is best given is another
area where the skills literature has much to offer. There is extensive
discussion (for instance, in Holding 1965) on the relative merits of explana-

Downloaded from https://academic.oup.com/eltj/article/42/2/89/372446 by University of Leeds user on 29 November 2020

tion and demonstration, and a look at the techniques used by trainers of
non-linguistic skills is likely to offer the language teacher exciting and fresh
perspectives. There is, for example, the Suzuki method of violin playing
where the learner is at an early age saturated with violin music, providing
an internal representation of the behaviour which can be 'proceduralized'
later. A further method is discussed in Gallwey (1971) where it is suggested
that learning to become a good tennis player may be helped by mimicking
the movements and even the idiosyncracies of a great player. 'Pretending to
by Jimmy Connors' may in part help one to play tennis like Jimmy
Connors. Full discussion of initial guidance techniques is beyond the scope
of this article, but there is much wisdom to be gleaned from the practices of
the skill trainer.
Though initial training may help to provide internal representation,
much can also be done 'after the event'—after, that is, the learners have
performed the behaviour for themselves. One technique which 'models
after the event' is reformulation. This technique, discussed by Levenston
(1978), Cohen (1983), and Allwright et d. (1984), is usually used for the
teaching of writing. There are several versions of reformulation, but the
basis is diat a native speaker rewrites a student essay, as far as possible
preserving the intended meaning. Reformulation is different from recon-
struction, which is what most of us do to student essays. In reconstruction,
errors and mistakes are simply corrected. The result will be sentences free
from gross malformations, but ones which may not remotely resemble
sentences a native speaker would produce to express the same content.
Because reconstruction focuses on errors and mistakes, it may well provide
the learner with information on where he or she went wrong. What refor-
mulation offers, and reconstruction fails to offer, is information on how a
proficient speaker would have said the same thing. Reformulation provides
a model of what the behaviour should look like; and though its clearest use
is for writing, there is no reason why spoken language should not be
reformulated.6

Realization of flawed It is interesting to note that according to Bartlett (1947:879), 'maybe the
performance best single measure of mental skill lies in the speed with which errors are
detected and thrown out. . .'. Knowing what has been done wrong (and
what to do about it) is something which, for example, distinguishes the
skilled tennis player from the novice. Six points will be made about this
stage:

1 It cannot automatically be assumed that the learner will be aware of

having made a mistake. The very conditions which produce the mistake
may prevent its detection. The fact that I have so many things to attend to
at the jump, on a difficult horse, with feet out of the stirrups, may make me
lean forward; it may also prevent me from knowing that I have leaned

92 Keith Johnson
forward. Therefore some positive action needs to be taken to make me
aware.
2 The positive action of being told by the teacher is probably not enough.
Learners seem to need to see for themselves what has gone wrong, in the
operating conditions under which they went wrong. There are various ways
of achieving this. My leaning forward on the horse is brought home best
when I see a video of myself doing it. As a second best, it is useful for me to
see others making the same mistake in the same conditions; and where

Downloaded from https://academic.oup.com/eltj/article/42/2/89/372446 by University of Leeds user on 29 November 2020

other learners are not available, teachers often provide the information by
mimicking the learner to indicate what is being done wrong. 'Monitoring
yourself in difficult operating conditions' suggests putting classroom lan-
guage on tape or video.
3 Explanation is probably not the best way to give mistake feedback.
There is evidence in the skills literature (for example, in Holding 1965) that
explanation is a procedure to be used warily anyway. Performance can be
positively harmed by elaborate explanation, as when the tennis coach
provides a lengthy lecture on how to hold the racket during service; the
result may simply be to inhibit the novice who, trying to serve, attempts to
bear in mind all the points of the explanation (cf. Gallwey 1971). One might
further argue that any benefit that explanation might provide would be for
errors rather than mistakes. The defining characteristic of a mistake is that
the student knows what should be done; explanation could therefore be
seen as providing what he or she already has.
4 It may again be that the best way of providing the necessary realization is
by confronting the learner with the mismatch between flawed and model
performance. This again points to reformulation. I want to see what the
teacher looks like going over the jump on a difficult horse (i.e. injidl operating
conditions—the importance of this will be touched on later), then to compare
this with what I looked like, injidl operating conditions.
5 When reformulation takes place, it may be that the most useful feedback
comes from those areas of mismatch which students are themselves able to
identify, because those areas will accord with the stage of their skill (or
interlanguage) development. A further example from riding; I was having
problems doing a good trot, and the teacher was demonstrating what it
should look like. During her demonstration, I noticed something about the
position of her legs which she had never drawn my attention to; it was not
on her 'teaching programme'. Once I held my legs in the same position,
several of the things which I was getting wrong and which she had drawn
my attention to suddenly became right. In that situation I was learning
something she had not set out to teach. Language teachers may find in their
experience similar examples of where 'point learned' is at odds with 'inten-
ded teaching point'; one of the benefits of reformulation is that if, without
comment, one merely presents students with a model performance to be
compared with their flawed performance, it is left up to them to note and
learn what they will from the comparison.

6 But in conjunction with (5) above a further point needs to be made.

There is one sense in which language skill is like ice skating. In ice skating,
learning the rudiments of survival—being able to stand up, move forward,
turn etc., without falling over—is a comparatively small part of becoming
an accomplished performer. A large part of the task involves learning to

Mistake correction 93
conform to an accepted model, established over time by tradition, of what
good skating looks like. In terms of'getting by on ice' many details of the
accepted model (how the legs and body should be held, for example) are
mere frills. The same is true of language, and the rudiments of linguistic
survival can be met by a form of pidgin. The learner who says 'Please give
beer' is unlikely to go thirsty; but he or she will have failed to conform to
externally imposed norms about language behaviour, norms which in pure
survival terms are frills. The skills literature's distinction between intrinsic
and extrinsic feedback (cf. Annett 1969) is relevant here. Intrinsic feed-

Downloaded from https://academic.oup.com/eltj/article/42/2/89/372446 by University of Leeds user on 29 November 2020

back, springing from the situation itself, is likely to provide information on
whether the rudiments of survival have been met; it occurs when the skater
falls over or the learner fails to get his beer. But such feedback is unlikely to
provide information on whether externally imposed norms have been
adhered to. For this, extrinsic feedback (from an outside source) is needed,
and to provide it the teacher will find it necessary to draw conscious
attention to mistakes and errors.7

Opportunity to practise The sequence being discussed in this article is one of mistake occurrence
again, in real operating —* corrective action —* retrial. There is some evidence in the skill
conditions literature (e.g. Annett 1969) that the relationship between the second two is
important. In terms of time, for example, it may be more important how
soon retrial takes place after corrective action than how soon after mistake
occurrence corrective action occurs. We therefore need to speak not just
about feedback after performance, but also about feedback before retrial.
It seems important that real operating conditions should be present in
retrial. The following exemplifies why, first in relation to a non-linguistic
skill, then in relation to language.
A novice pilot may well be able to land in clear weather when the plane
has no mechanical defects. The problem may be landing the plane in fog
and when the flaps are not working correctly. In this situation, to practise
landing in clear skies in a perfect plane is clearly of restricted value. What
the pilot needs to practise is, precisely, landing in fog with faulty flaps. For
this an aircraft simulator is provided. What the simulator offers is various
configurations of operating conditions.
The student may be able to form the present perfect correctly in a gap-
filling task. His or her problem may be with getting it right over the
intercontinental telephone line referred to earlier. In this situation, simply
giving more gap-filling tasks is of as restricted a value as landing in clear
weather. What the learner needs is some form of'present perfect simulator'
which will vary the operating conditions, to simulate just those types of
conditions which are presenting difficulties.
What does a 'present perfect simulator' look like? Perhaps work like that
of Brown et ol. (1984)—which may be interpreted as an attempt to identify
some parameters of difficulty in operating conditions—will provide a way
of grading tasks in terms of operating condition complexity. Whether or not
this is so, it is clear that in important respects, free practice offers a form of
'present perfect simulator'.8 What free practice provides is ready-made sets
of operating conditions; these will vary from moment to moment, and will
place variable demands on the learner's ability to process. Sometimes the
interaction will require speedy response, sometimes not; different interac-
tions will involve different amounts of language; the demands of message
(and hence the degree of attention the learner must give to what he or she is
saying rather than how he or she is saying it) will change. There will be other

94 Keith Johnson
types of variation, not least in affective conditions (the degree of anxiety
felt, attitude towards interactant, etc.) which will affect the performer's
processing efficiency. Bad conditions along parameters like these are the
language user's equivalents of fog and faulty flaps. Free practice will go a
long way towards simulating, over time, the operating conditions in which
mistakes occur.
The stages of corrective action and retrial are both seen as crucial to
mistake eradication. It is optimistic to suppose that once corrective action
has been taken, a mistake (as opposed perhaps to an error) will disappear.

Downloaded from https://academic.oup.com/eltj/article/42/2/89/372446 by University of Leeds user on 29 November 2020

Part of learning to land in fog involves landing in fog; part of learning to use
the present perfect on an intercontinental phone involves phoning intercon-
tincntally and using the present perfect. It is, however, equally optimistic to
suppose that retrial alone will efficiently eradicate mistakes. Both stages are
seen as necessary but not, taken alone, sufficient for mistake eradication to
occur.

Conclusion This article begs many questions. It may be said merely to switch the focus
of attention from initial learning to feedback. The question of how to
provide successful feedback is no less perplexing than the question of how to
facilitate successful initial learning. But perhaps a willingness to pursue the
metaphor of language learning as skill learning will provide interesting new
perspectives on both these questions and many others. •
Received January 1987

rence of fossilization when communicative needs are

1 This article arose out of talks I gave at the Univer- being adequately met.
sity of Lancaster and at EaJing College of Higher 6 The suggestion that reformulation might be used for
Education. Many useful points made during discus- spoken language does raise some practical, logistical
sion after these talks have been incorporated into problems which would need discussion.
this version. 7 This similarity between language and ice skating
2 Johnson (1986) attempts to provide such a was pointed out to me by Dick Allwright, whose
justification. comments on a number of points made in this article
3 This point is made by Bialystok and Sharwood are gratefully acknowledged.
Smith (1985), who use the terms knowledge and con- 8 The term 'free practice' is here intended in a general
trol to describe the distinction discussed here. See sense to refer to the kinds of activity which Byrne
Note (4) below for further reference to their work. (1976) associates with the 'production stage'. A
4 Many of the issues discussed in this article directly central characteristic of such practice is that learn-
relate to issues arising in the literature on variability ers are given considerable freedom to choose what
in intcrlanguage—e.g. in Tarone (1982 and 1983), they say and when they say it. Many kinds of role
Bialystok (1982), Ellis (1985b) and Bialystok and play and simulation exercises are 'free practice' in
Sharwood Smith (1985), among others. It is by no this sense.
means the case that all these researchers would
support the position being developed here. This
position is closest to Bialystok and Sharwood Rmftrwtcms
Smith's, though they avoid association with any Allwright, R. L., M-P. Woodley, and J. M. All-
general model of skill acquisition, specifically the wright. 1984. 'Investigating Reformulation as a
one which informs this article--the model of Ander- Practical Strategy for the Teaching of Academic
son (1982). For discussion of interlanguage vari- Writing.' Paper presented at the BAAL Annual
ability within Anderson's framework for skill General Meeting, September 1984.
acquisition, see Johnson (forthcoming). Anderson, J. R. 1980. Cognitive Psychology and Its Impli-
5 The literature on pidginization and fossilization cations. San Francisco: W. H. Freeman.
contains relevant discussion on this point. See Smith Anderson, J. R. 1982. 'Acquisition of cognitive skill.'
(1972) for the idea that pidgins are simplified and Psychological Review 89/4: 369-406.
reduced because used for restricted functions. Annett, J. 1969. Feedback and Human Behaviour.
Selinker and Lamendella (1978) discuss the occur- Harmondsworth: Penguin Books.

Mistake correction 95
Bartlett, F. C 1947. 'The measurement of human Johnson, K. (forthcoming). 'Cognitive skill acqui-
skill.' British Medical Journal 4510:835-8 and sition and second-language acquisition.' Available:
4511:777-880. K.Johnson, Department of Linguistic Science, Uni-
Bialystok, E. 1982. 'On the relationship between versity of Reading, Whiteknights, PO Box 218,
knowing and using linguistic forms.' AppludLinguis- Reading RG6 2AA.
tics 3/3A81-206. Levenston, E. A. 1978. 'Error analysis of free com-
Bialystok, E. and M. Sharwood Smith. 1985. 'Inter- position: the theory and the practice.' Indian Journal
language is not a state of mind: an evaluation of the of Applied Linguistics 4/1:1-11.
construct for second-language acquisition.' Applied Selinker, L. and J. Lamendella. 1978. 'Two perspec-

Downloaded from https://academic.oup.com/eltj/article/42/2/89/372446 by University of Leeds user on 29 November 2020

Linguistics 6/2:101-117. tives on fossilization in interlanguage learning.'
Brown, C , A. Anderson, R. Shillcock, and G. Yule. Inttrlanguage Studies Bulletin 3:143-91.
1984. Teaching Talk. Cambridge: Cambridge Univer- Smith, D. M. 1972. 'Some implications for the social
sity Press. status of pidgin languages' in D. M. Smith and R.
Byrne, D. 1976. Teaching Oral English. London: W. Shuy (eds.): Sociolinguistics in Cross-Cultural Analy-
Longman. sis. Washington DC: Georgetown University.
Cohen, A. D. 1983. 'Reformulating second-language Tarone, E. 1982. 'Systematicity and attention in inter-
compositions: a potential source of input for the language.' Language Learning 32/1:69-84.
learner.' ERIC ED 228 866. Tarone, E. 1983. 'On the variability of interlanguage
Corder, S. P. 1981. Error Analysis and Inttrlanguage. systems.' Applud Linguistics 4/2:142-63.
Oxford: Oxford University Press.
Ellis, R. 1985a. Understanding Second Language Acqui-
sition. Oxford: Oxford University Press.
Ellis, R. 1985b. 'Sources of variability in inter-
language.' Applied Linguistics 6/2:118-31. Thmmuthor
G a l l w e y , W. T. 1971. The Inner Gamt of Tennis. New Keith Johnson lectures in the Department of Linguis-
York: Random House. tic Science at the University of Reading. He was a
Holding, D. H. 1965. Principles of Training. Oxford: founder member of the Centre for Applied Language
Pergamon Press. Studies at the University of Reading, where his work
Johnson, K. 1986. 'Language Acquisition as Skill included materials production, presessional course
Acquisition.' Paper delivered at the C.A.L.S. Collo- organization, and teacher training. He has published
quium, 1986. Available: K.Johnson, Department of in the area of communicative language teaching, and is
Linguistic Science, University of Reading, White- at present interested in viewing language teaching
knights, PO Box 218, Reading RG62AA. within a cognitive skills framework.

96 Keith Johnson