0346-251X/93 $6.00 + 0.00 0 1993 Pergamon Press Ltd
System, Vol. 21, No. 2, pp. 151-163, 1993 Printed in Great Britain
PUPIL
SELF-CORRECTION IN ORAL COMMUNICATION ENGLISH AS A FOREIGN LANGUAGE* PETER
t University of York,
S. GREEN?
and KARLHEINZ
IN
HECHTS
York, United Kingdom; and $.Universitiit Miinchen, Miinchen, Germany
This article focuses on self-corrrection as a cognitive strategy. Is it a strategy that is typical of the more successful foreign language learner? After a look at the different forms of self-correction-covert self-correction and overt editing-as psycholinguistic processes, Krashen’s Monitor Theory is considered. Six suppositions about self-correction are then examined in the light of an empirical investigation. The self-corrections in spoken language (English) of 286 German and English pupils are analysed and the results interpreted. As the test population comprised native and non-native speakers, the extent to which their self-correction behaviour is comparable is also examined. The results throw a favourable light on the efficacy of self-correction, and relevant suggestions for the foreign language teacher are presented at the end of this article.
LEARNING
STRATEGIES
AND
SELF-CORRECTION
The question of how a foreign language is learned in a school context and what mental processes are involved has given rise in recent years to an interest in the study of learning strategies.’ Learning strategies, as opposed to production or communication strategies, are long-term plans employed by learners in different ways and with differing frequency, to control the learning process. Of the many different kinds of learning strategies [cf. Oxford and Nyikos (1989) and Oxford (1989)] we are particularly interested in the linguistic ones, such as generalisation, transfer and self-correction or monitoring. Interest in learning strategies is inspired by the hope of finding out which learning strategies are employed by successful learners. * Is self-correction a characteristic of successful pupils? Chamot and Kupper (1989: p. 21) found that self-correction was a strategy leading to more effective language learning. However, their investigation, which was based at the outset on the study of 40 American high school students learning Spanish, had only 13 learners by the end of the study period (fourth semester).
*An earlier version of this article by the same authors appeared in German in Die Neueren Sprachen 90 (1991: pp. 607-623) under the title “Schiilerselbstkorrektur beim Einsatz des Englischen in miindlicher Kornmunikationeine empirische Untersuchung”.
152
PETER
S. GREEN
and KARLHEINZ
HECHT
Self-correction may then be a learning strategy in a foreign language, but we also ourselves not infrequently when using our native language, which we have already learned. So self-correction is also part of our psychological make-up, belonging attentional processes which accompany procedures that are not fully automatic. (1989: p. 19) describes them thus:
correct largely to the Cohen
Attentional processes are under moment-to-moment control by a central processor, which monitors . . . the action sequence, modifying performance according to feedback about changes in external circumstances and internal needs and intentions. A good example of this distinction between automatic and attentional processes is provided by driving a car. Emerging from a road junction is (or ought to be) an attentional process.
Speaking and writing are processes which are not fully automatic and which are therefore accompanied by the psycholinguistic control process of monitoring.
SELF-CORRECTION
AS A PSYCHOLINGUISTIC
PHENOMENON
All self-correction derives from two basic faculties-the creative and the critical-which are applied in both language production and language reception. In this article we are concerned only with self-correction in language production and in particular the kind of monitoring of language production that in psycholinguistics is designated overt editing or overt self-correction. As, for example, Hackett (1967), Laver (1980) and Fromkin (1980) show, we first of all constantly monitor the language planning process-covert selfcorrection at the pre-articulatory level. Then, when we articulate a plan and produce a message in either spoken or written form, we check again to ensure that the realisation is actually in accord with the plan-overt self-correction at the post-articulatory level. If we notice, for example, that a sound, a word or a structure deviates from the plan, we endeavour to correct this slip of the pen or tongue .3 Hackett (1967: p. 922) describes it as follows: This suggests that ordinary speech may, at least at times, be more like reading aloud than we have thought: as though the speaker first constructed a “text” somewhere inside himself and then read it off, sometimes inaccurately. It should perhaps be added that, in spite of covert editing, the “text” itself may contain errors [cf. Motley et al. (1983: p. 84)]. Self-correction, then, is a process of comparison between a desired form (the “text”) and what we have actually produced. We discover, through implicit or explicit knowledge, discrepancies between “internal” and “external” language [cf. Levelt (1983)] and retrospectively correct any deviations in form or content.
What instance is it that gives us specific instructions as to how we are to correct? Levelt (1983: p. 50) answers the question as follows: “. . . the monitor makes the speaker aware of this (the mismatch) . . . and an alarm signal is sent to working memory”. Working memory then gives instructions on how the correction is to be carried out. This involves employing knowledge of language that may be either explicit or implicit, or some combination of the two. In this sense, monitoring is a cybernetic process in which the critical faculty controls the creative. Self-correction in the mother tongue4 appears then to be a
PUPIL SELF-CORRECTION
natural control process.
process and in foreign-language
KRASHEN’S
RESTRICTED
IN ORAL COMMUNICATION
acquisition
153
it may contribute
USE OF MONITORING:
“THE
to the learning
MONITOR”
Ever since the 198Os, when Krashen presented his Monitor Theory, self-correction in foreignlanguage learning has been a controversial topic [see, for example, Krashen (1981, 1982) and Morrison and Low (1984)]. Krashen restricted Monitoring (the initial capital is Krashen’s way of identifying his version of the concept) to the conscious application of rules in language production. According to his theory, language production is largely initiated and sustained by unconscious knowledge (acquisition) and is completely independent of conscious knowledge (learning). Conscious knowledge is generally only brought into play in Monitoring. As there is no transfer from conscious to unconscious knowledge (the so-called non-interface hypothesis), Krashen (1982: p. 15) concludes that “learning has only one function, and that is as a Monitor, or editor”. It should be pointed out that Krashen is referring to learners acquiring a foreign language after puberty.5 Krashen’s Monitor Theory goes further. He claims that, for the foreign the Monitor is only used under the following conditions: (1) the learner must have enough time to reflect; (2) the learner’s attention must be primarily on the form not the content produced; (3) the learner must have explicit knowledge of the rule.
language
learner,
of what is being
Krashen’s (1982: p. 18) view is that “. . . for most people . . . it takes a real discrete point grammar test to meet all . . . conditions for Monitor use”. Given these conditions, the following can be observed: (a) self-correction functions best in straightforward cases of violation of morphological rules (Krashen, 1982: p. 105); (b) it is less effective in violations of complex grammar (ibid.), functioning only for the easiest aspects of grammar (Krashen, 1982: p. 112); (c) not every learner makes use of the Monitor and those who do are not very successful“the effects of self-correction on accuracy is [sic] modest” (ibid.); (d) self-correction occurs only for a small percentage of errors (ibid.).
EXPECTATIONS
CONCERNING SELF-CORRECTION FOREIGN-LANGUAGE TEACHING
IN SCHOOL
Krashen’s assertions run counter to our own experience, which is that pupils constantly, and largely successfully, monitor their own productions. Moreover, his claims are based on studies involving only a small number of testees and refer exclusively to the acquisition of English as a second language in an English-speaking environment. We decided, therefore, to study the self-correction behaviour of pupils learning English as a foreign language in
154
PETER S. GREEN and KARLHEINZ
HECH’I
German-speaking surroundings, in a group of adequate size for statistical we expected to find can be formulated as six suppositions:
purposes.
What
Supposition
1. Self-correction is part of the general psychological attentional processes mentioned above and occurs therefore in L2 as well as Ll, even when the attention of the speaker is primarily directed towards meaning and there is-as is usual in oral communication-a certain time pressure.
Supposition 2. We assume that self-correction knowledge of rules, as traditional rather than feel for the language.
language
in a foreign language is primarily based on teaching generally imparts knowledge about
Supposition 3. As the learner’s
knowledge of rules is limited, self-correction is not frequent in the case of grammatical deviations, but, when it does occur, its effects are positive and the number of mistakes is reduced.
Supposition 4. Self-correction against morphology, is least complex.
occurs most frequently and effectively in the case of offences since that is the area where the rule system of the target language
Supposition
5. As self-correction, when it occurs, usually has a positive outcome, those pupils who make frequent use of self-correction have a lower error rate than their peers who self-correct less frequently.
Supposition 6. As self-correction is part of the attentional processes and therefore of the general psychological make-up of a speaker, its forms in Ll and L2 are similar. However, the number of self-corrections is lower for native speakers, because they have already learned the language. EMPIRICAL
INVESTIGATION
Test details Our long-standing research project into the school learning of English as a foreign language-“Learner language”6has built up a corpus of some 1500 communicative (message-centred) pupil productions and 4600 medium-centred tests (of grammar, vocabulary and style). For the present investigation we made use of message-centred oral productions, 286 in all. The main test population was made up of 232 pupils learning English as a foreign language in secondary schools in different parts of Bavaria. They covered the full range of ability and the age range 12-19. There was also a control group of 54 English pupils from schools in Yorkshire, aged 14-15. The distribution and numbers of the German pupils involved in the investigation make it possible to claim that they were a representative sample of Bavarian school learners of English. All the pupils were volunteers and they knew that their productions would not affect their school grades but would be evaluated purely for research purposes. The pupils were divided into two groups and within each group they performed the same task irrespective of age or ability. One group was required to make an oral statement to
PUPIL
SELF-CORRECTION
IN ORAL COMMUNICATION
155
the police regarding a case of shop-lifting. In doing so, the pupils were to assume the role of the shop owner and they drew their information about the incident from a series of cartoon pictures. This task is referred to as Picture Story Oral (PSO). The other group took part in a simulated telephone conversation with an English interlocutor (whose voice was recorded on cassette). These pupils played the role of a German pupil in England, and, with the help of a flowchart (in German), were to ask the warden of a youth hostel for information about accommodation, meals, prices, opening times, location of the hostel, etc. This task is referred to as Telephone Conversation (TC). The distribution of the pupils across school types, levels and tasks is shown in Table 1.’ Table
School
1. Distribution
of pupils across tasks
type/level
Gymnasium beginners Gymnasium intermediate Gymnasium advanced Realschule Hauptschule Total no. of German pupils English pupils Total no. of testees
school types,
levels and
PSO
TC
Totals
26 32 29 20 19 126 34 160
24 20 21 21 20 106 20 126
50 52 50 41 39 232 54 286
In the context of school foreign-language learning the simulations were as close as possible to real communication whilst permitting comparability through the control of content by the pictures or the flowchart. The testees were concerned primarily with conveying information and attention to form was in the background, as in all authentic communication. They had a chance to study the pictures or the flowchart beforehand, and could ask the teacher who conducted the test questions of a procedural but not of a linguistic nature. Each performance was recorded individually so that pupils had no opportunity of helping each other. The recorded performances were transcribed for analysis.
Examination of the first supposition (self-correction occurs in L2 even with attention to meaning and under time pressure) As regards attention to the meaning of the language, the test format-role-play and transmission of information (PSO) or performance of speech acts (TC)-ensured that the speakers were all concerned primarily with meaning and only secondarily with the formal side of language. Furthermore, pupils knew that they would not be graded on their performance. As regards time pressure, a cassette recorder was running during the pupil’s performance in the picture story, which, whilst not able to replace a live interlocutor, certainly inhibited any lengthy reflection on the correct formulation of the pupil’s speech intentions. There were, it is true, occasional long pauses, but they were the exception. In the telephone conversation, one cassette recorder recorded the pupil’s performance whilst another played the voice of the youth hostel warden as soon as the pupil had completed a speech act. Both simulations exerted a considerable time pressure on the pupils, as they frequently
156
PETER S. GREEN and KARLHEINZ
HECHT
commented at the end of the recording. Is then our assumption correct that, despite these conditions, self-correction will occur as part of the attentionalprocesses? Tables 2 and 3 show the numbers of self-corrections for the picture story and the telephone conversation, respectively.
Table 2. Text length
School
and number
No. of words
type/level
Gymnasium beginners Gymnasium intermediate Gymnasium advanced Realschule Hauptschule Total (German pupils) English pupils
Table
3. Text length
of self-corrections
2410 3648 4116 2140
1577 14,011 3064
and number
No. of selfcorrections
=every nth word
68 102 90 88 41 389 56
36th 36th 46th 24th 38th 36th 64th
of self-corrections
No. of
for PSO
for TC
type/level
words
No. of selfcorrections
=every nth word
Gymnasium beginners Gymnasium intermediate Gymnasium advanced Realschule Hauptschule Total (German pupils) English pupils
3120 2940 3234 3150 2480 14,924 2700
92 69 64 31 48 304 46
34th 43rd 51st 102nd 52nd 49th 59th
School
For the 232 German pupils there are altogether 693 self-corrections, an average of three self-corrections per pupil production. This may not seem a very high number, but the texts averaged only 125 words in length. The first supposition is in any case clearly correct.
Examination of the second supposition rule knowledge)
(self-correction in L.2 is largely based on explicit
We cannot know whether the 232 German pupils had an explicit rule for each of their self-corrections. We did, however, administer a test of explicit knowledge of grammar rules to a comparable group of 250 German and 50 English pupils (Green and Hecht, 1992). The pupils were presented with erroneous sentences originating from our corpus of pupil productions. The error in each sentence was underlined and the pupils were asked, first, to give an explanation or rule that would enable whoever had made the error to understand and correct it, and, second, irrespective of whether they had been able to give a rule, to correct the error. Although we were very liberal in deciding what was a correct rule and, although the rules called for were all in central areas of grammar and had been thoroughly taught in the school syllabus, the pupils were only able to give an acceptable formulation in 38% of the possible cases. It seems then highly unlikely that only explicit knowledge enabled pupils to achieve self-corrections in the foreign language. This view is lent strong
PUPIL SELF-CORRECTION
IN ORAL COMMUNICATION
157
support by another finding from the explicit grammar test: in no less than 50% of the cases in which the pupils arrived at a correct correction, they had given either an incorrect rule or no rule at all. At best, then, pupils could call on explicit rules for half of their correct corrections (whether they did so is unknown); for the other half they must have been dependent on their feel for the language. Our second supposition was not therefore confirmed: self-correction in a foreign language cannot be based primarily on explicit rule knowledge. Examination of the third supposition (self-correction is IargeIy successful) Self-correction may lead to an improvement of the original utterance or it may result in further error. Whether a particular self-correction was successful or unsuccessful was decided by the majority judgment of three native speakers. Table 4 shows that they judged 560 of the 693 instances of self-correction to have been successful. This success rate of 81% confirms the supposition. Indeed, for most of the individual groups, the success rate is even higher. The overall rate is reduced mainly by the low success rate of the Realschufe in the picture story (38%). There is no obvious reason why this result is so out of line with the others. Table 4. Number
School
type/level
Gymnasium
beginners
Gymnasium
intermediate
Gymnasium
advanced
Realschule Hauptschule Total
of successful No. of pupils/ text type 26 PSO 24 TC 32 PSO 20 TC 29 PSO 21 TC 20 PSO 21 TC 19 PSO 20 TC 232
self-corrections
and their effect
Successful self-corrections No.
%
No. of actual errors
63 74 96 60 78 61 33 21 35 39 560
93 80 94 87 87 95 38 68 85 81 81
380 454 388 349 441 328 283 403 290 463 3779
on error
numbers
No. of potential errors 443 528 484 409 519 389 316 424 325 502 4339
% of potential errors avoided 14 14 20 15
15 16 10 5 11 8 13
Table 4 shows further the overall extent to which pupils improved what they produced by self-correction. The 232 German pupils made a total of 3779 errors in their productions (again according to the judgment of a majority of native speakers6). As we have seen, the pupils avoided 560 further errors by self-correction, thereby achieving a 13% reduction in their potential error total. For the pupils of the Gymnasium the improvement is in all cases larger (14-20%) and largely independent of the text type. Examination of the fourth supposition (self-correction is most frequent and effective with morphological errors) Our belief that morphological rules are relatively straightforward and easily learnt was borne out by the explicit grammar test (Green and Hecht, 1992: p. 179). Did it follow then that the frequency and success of self-correction was highest for offences against morphology? Table 5 shows the frequency and effectiveness of self-correction for five
158
PETER
Table
S. GREEN
and
KARLHEINZ
HECHT
5. Frequency and effectiveness correction by linguistic category
Linguistic category Phonetics Morphology Syntax Lexis Semantics
of self-
% of selfcorrection
Success of selfcorrection (o/o)
13 23 20 28 15
92 91 86 81 96
linguistic categories: phonetics, morphology, syntax, lexis and semantics. As can be seen, morphology accounts for 23% of self-corrections but is surpassed by lexis with 28%. The picture is modified somewhat if one looks at the success rates, where morphology has the higher value (91 vs 81%). However, the highest success rate of all is found for semantics (96%). Thus, the fourth supposition is not confirmed. Interestingly, if the categories are combined into morpho-syntax and lexico-semantics, then each accounts for 43% of self-correction, and the distribution found here exactly matches that reported by Desgranges [(cited by Konigs (1992: p. 171)]. Desgranges investigated self-correction in the German spoken by a group of children of immigrant workers (Italian and Turkish). They had learnt German through going to school in Germany but without any special teaching. They produced only a few self-corrections in phonetics, and morphosyntactic and lexico-semantic corrections were about equal in number. Although the learning situation of the two groups of pupils was different-the Italian-Turkish pupils were secondlanguage learners in L2 surroundings, whereas the German pupils were foreign-language learners in Ll surroundings-the two studies have in common that they examined selfinitiated self-corrections in both monologue and dialogue speech. By contrast, in Kdnigs’ investigation of classroom correction, teachers corrected 65% of morpho-syntactic errors but only 39% of lexico-semantic errors (Konigs, 1992: p. 173).
Examination of the ffth rate)
supposition (pupils who self-correct frequently have a low error
The converse of this supposition might, of course, also be true: pupils who rarely or never self-correct have a high error rate. To examine this supposition, we looked at “underusers” and “overusers” of self-correction. As the average number of self-corrections was three per pupil production, we defined underusers as those pupils with no self-corrections and overusers as those with five or more. Thirty of the 232 pupils, or 13070, turned out to be underusers and 52, or 22%, overusers. However, when we examined the error rates of the pupils in those categories, we could find no significant correlation between the two measures (p < 0.01). Thus, for example, one of the pupils at the beginners’ level in the Gymnasium had six self-corrections and a high error rate of 21% (i.e. 21% of that pupil’s production was erroneous), whereas another pupil in the same group had seven self-corrections and a low error rate of 11%. A third pupil with no self-corrections at all achieved an extremely low error rate of only 8%. The average error rate for that group was 16%. At the advanced level in the Gymanasium, where the average error rate was 11%, there was another pupil with no self-corrections and an error rate of only 4%.
PUPIL
SELF-CORRECTION
159
IN ORAL COMMUNICATION
The absence of any relationship between the number of self-corrections and the quality of the product, as judged by the absence of error, is contrary to the finding of Chamot and Kupper (1989: p. 21) that good language learners make frequent use of self-correction. That was not true of our corpus and the fifth supposition is therefore unfounded.
Examination of the sixth supposition (self-correction, as part of the attentional processes, takes similar forms in L1 and L2 but the number of self-corrections is lower for native speakers) Self-corrections characteristics:
in the
mother
tongue
have
been
observed
to
have
a number
of
(1) They may occur without a pause. (2a) They may occur after an unfilled pause. (2b) They may occur after a filled pause. (3) They can be divided into three different classes [cf. Faerch and Kasper (1983: p. 216)]: (a) false starts unretraced-the correction occurs immediately after the error; (b) false starts retraced-the correction does not follow the error immediately but the syntactic unit in which the error is embedded is repeated; (c) new starts-a longer section of text before, and sometimes after, the trouble source is repeated in order to change or expand a syntactic unit.9 Are these characteristics found also in the self-corrections which occur in a foreign language? An analysis of the self-corrections in our corpus of oral productions revealed the following (examples all taken from the corpus): (1)
Self-corrections
occurred
without
a pause:
a man came [to my] into my shop (2a) They occurred after an unfilled pause: [he put. . .] he switched the radio on (2b) They occurred after a filled pause: another radio from the [desk em . . . em] table (3) They occurred in three variants: (a) false starts unretraced: he . . . [get] got nervoz4s
the man [wear] wore a coat (b) false starts
retraced:
running [out of the street] out of the shop [there was stand] there was standing (c) new starts: [while I was climbing up the em . . .] while I was climbing up to take the but [when Z climbed the . . .] when I climbed down from the ladder Both unretraced infrequent:
and retraced
false starts false starts new starts
false starts
unretraced retraced
were frequent
49% 42% 9%
in the corpus;
new starts
were
160
PETER
S. GREEN
and KARLHEINZ
HECHT
All three variants occur then in a foreign language, but how similar is their distribution to that of native speakers? An analysis of the 34 productions of the 14-15year-old English pupils who performed the picture story task revealed the following distribution: false starts false starts new starts
unretraced retraced
55% 41% 4%
There appears then to be a very high degree of similarity between the forms that selfcorrection takes in Ll and L2. However, as Tables 2 and 3 show, the frequency of selfcorrection is generally lower for Ll speakers. Supposition 6 is thus correct.
CONCLUSIONS
AND
IMPLICATIONS
FOR TEACHING
As we have seen, learners can improve what they produce in a foreign language with the help of their explicit or implicit knowledge about the language. A number of Krashen’s pronouncements about the Monitor-sufficent time for reflection, focus primarily on linguistic form, self-correction based on explicit rule knowledge, modest effects of selfcorrection-were not confirmed by our investigation. The forms self-correction took in overt editing are very similar to those found with the native speakers. The self-corrections that we investigated were overwhelmingly successful and helped to reduce overall error. The value of overt editing as a learning strategy was, however, unproven in the investigation, where self-correction was not necessarily associated with learners who had a low error rate. In our observations, the number of self-corrections in oral production was low (an average 125word production had three corrections), but, as their success rate was high (an average of 81%), it seems to be in learners’ interests to be self-critical. That may be a reason for encouraging self-correction, but what then is the role of teacher correction? On the one hand, it is clear that negative feedback plays a part in first-language acquisition [cf. Hirsh-Pasek et al. (1984) and Morgan and Travis (1989)] and that it prevents fossilisation of incorrect linguistic forms in second-language acquisition. On the other hand, the effectiveness of teacher correction in foreign-language learning remains open to question. Empirical investigations have had contradictory results. Whilst Horner (1988), for example, on the basis of relevant studies, pleads for careful and moderate teacher correction, Semke (1984) opposes it, since her own investigation (duration only 10 weeks) failed to demonstrate any advantages in teacher correction. Holley and King (1974) are also critical, whilst Holtwisch (1990) calls for error prevention rather than error correction. And Allwright (1988: p. 109) warns that: However, what worries me most is not only that such feedback procedures are inefficient in themselves, but that they constitute a large quantity of ‘spoonfeeding’ with the teaching taking virtually all the responsibilty for error detection and correction. Such spoonfeeding militates against the development of writer autonomy, . . . So, too much teacher correction may be at the expense of the learner’s willingness to selfcorrect. K&rigs (1992) found that self-corrections played only a small role in the classroom,
PUPIL SELF-CORRECTION
IN ORAL COMMUNICATION
161
although three-quarters of pupils in questionnaires wished to have the possibility of selfcorrection (Kleppin and K&rigs, 1991). In the same study, the pupils with the highest degree of interest in self-correction were those whose teachers most often invited pupils to selfcorrect. So learners can be encouraged to develop and make use of the strategy of selfcorrection. From a pedagogical point of view, self-correction may be seen as part of an education for autonomous learning. If our teaching aims to lead to what Dickinson (1988: p. 34) called “learning how to learn”, then self-correction is a step in that direction because it encourages responsibility and independence in the learner. But how might selfcorrection be practised in the classroom?
First of all, the teacher’s role in the correction of both oral and written work might be restricted to the identification and collation of errors, leaving the actual correction as far as possible to peer correction in group work. This is a stage at which independent and appropriate use of dictionaries and grammars can be practised. What began as group work can later become pair and individual work. Part of the foreign language learning process might also be to confront learners with oral and written texts containing unidentified errors, which can then be identified and corrected in group, pair or individual work. Rinvolucri has an “auction game” with a similar aim. One variant takes the following form:
Pair the students
off and give each pair an auction sheet. Tell them that some of the sentences on the sheet are correct and some incorrect. They are to read through and decide which sentences are correct and which incorrect. In the auction that is to follow they are going to have to bid for sentences, the aim being to buy only correct sentences. Tell them to note down in the budget column of their auction sheets how much they are willing to bid for a given sentence. They may not spend more than 0000 in the auction. The winners of the auction are the pair with the most correct sentences and the most money left (Rinvolucri, 1984: p. 18).
Some teachers may have reservations incorrect.1° They should be reassured years ago:
about confronting learners with anything that is by what Corder (1979: pp. 293-294) wrote several
Evidence from studies in concept formation show [sic] the importance of negative instances. . . . Language learning is not parrot learning: we do not ‘learn’ or ‘practise’ examples. They are the data from which we induce the systems of the language. Skill in correction of errors lies in the direction of exploiting the incorrect forms produced by the learner in a controlled fashion.
Of course, correction should not be at the expense of fluency and willingness to communicate: those criteria must take precedence. We would not wish learners to become “overusers” of monitoring from fear of making mistakes. On the other hand, communication is not only concerned with the transmission of propositional meaning, it has an affective value as well. Long-term language teaching cannot afford to undervalue a striving for linguistic correctness, because it can often be an important factor in social acceptance by native speakers. Their tolerance of error has limits.
162
PETER S.
GREEN
and KARLHEINZ
HECHT
NOTES ‘See, for example, Naiman et al. (1978), Bialystok (1979), Wenden and Rubin (1987) and Chamot and Kupper (1989). 2Rubin (1981) describes six strategies employed by successful learners in an investigation: (1) classification, (2) monitoring, (3) memorisation, (4) guessing, (5) deductive reasoning, and (6) practising. 3There is also covert self-correction at the post-articulutory level, e.g. the constant checking of the accuracy of speech sound production by means of kinaesthetic and auditory feedback. We only become aware of this checking when the feedback is disturbed in some way, e.g. by local anesthetic in the mouth or artificially created delay of auditory feedback [cf. Lee (1950) and Abercrombie (1967)]. 4Psycholinguists refer to two further kinds of self-correction, which will not be investigated here: delayed control (e.g. an interlocutor’s reaction reveals that we have said something incomprehensible) and refiexion (e.g. we hear a recording of what we have said or read what we have written after an interval and make alterations). ‘Krashen’s non-interface hypothesis has been much criticised, e.g. by Gregg (1984), Cross (1985), Ellis (1985), Horner (1987) and McLaughlin (1987). ‘A full account of the project is given in German by Hecht (1991). A brief description in English and references can be found in Green and Hecht (1992). ‘The three school types found in Bavaria (and generally in Germany) are Gymnasium, Realschule and Hauptschule. The Gymnasium is the most academic school type, where pupils are expected to remain until the age of 18 or 19, when they take the Abitur qualification for university entrance. The Reufschule is for the more practically oriented pupils, whose further studies after compulsory schooling will generally be pursued in technical or commercial colleges. The Hauptschule is for the non-academic pupils, who will generally become skilled or unskilled manual workers. There is a certain amount of movement between the three school types. There were three groups of Gymnasium pupils-beginners with 3-4 years of English, intermediate with 5-6 years and advanced with 8-9 years. The pupils from the Realschule and Hauptschufe corresponded in age and experience to the intermediate pupils. The English pupils were from comprehensive schools and a secondary modern school. They covered the full ability range. 8Error in a foreign language is by no means a clear-cut category. Apart from spelling and morphological deviations from the norm, most errors are determined by a wider context and are a matter of subjective judgment. In an earlier phase of our learner language project, five native speakers were asked to identify the errors in 60 letters written in English by German pupils. Collectively (but independently) they arrived at a substantial total of “errors”, but they only agreed (majority, not unanimous, agreement) on half of the total (Green and Hecht, 1985). For that reason, when we speak of errors in this investigation, we refer to what a majority of three native speakers identified as errors. ‘Psycholinguists speak also of repeats, in which, however, nothing is changed. Generally, a word is repeated immediately after being spoken, e.g. “I I saw that the man . . .” Repeats occurred in our corpus too but will not be pursued here [cf. Maclay and Osgood (1959: p. 24)]. toA teacher of our acquaintance, who gets pupils to identify and correct errors in texts, writes in bold capitals across the top of the texts “CAUTION: RUBBISH!“.
REFERENCES D. (1967) Elements of General Phonetics. Edinburgh:
ABERCROMBIE, ALLWRIGHT, BIALYSTOK,
R. (1988) Don’t
E. (1979) The role of conscious
Edinburgh
University
Press.
ELT Documents 129, 109-116. Hong Kong.
correct-reformulate! strategies
in second
language
Canadian Modern
proficiency.
Language Review 35, 372-379. CHAMOT,
A. U. and KUPPER,
L. (1989) Learning
strategies
in foreign language
Foreign Language
instruction.
Annals 22, 13-24. G. (1989) Memory in the Real World. Hove:
COHEN,
CROSS,
D. (1985) The Monitor
DICKINSON, ELLIS,
Lawrence
Erlbaum
S. P. (1979) Introducing Applied Linguistics. Harmondsworth:
CORDER,
teacher. British Journal ofLanguage Teaching 23,75-78. in Language Learning. Cambridge: Cambridge University Press.
Theory and the language
L. (1988) Se/f-instruction
R. (1985) Understanding Second Language Acquisition.
FERCH, FROMKIN,
C. and KASPER,
Associates. Penguin.
Oxford:
Oxford
University
G. (eds) (1983) Strategies in Interlanguage Communication.
V. A. (ed.) (1980) Errors in Linguistic Performance.
New York: Academic
Press.
London. Press.
Longman.
PUPIL
GREEN,
P. S. and HECHT,
SELF-CORRECTION
163
IN ORAL COMMUNICATION
‘K. (1985) Native and non-native
evaluation
of learners’
errors in written
discourse.
System 13, 77-91. GREEN,
P. S. and HECHT,
K. (1992) Implicit
and explicit grammar:
study. Applied Linguistics
an empirical
13, 168-184. GREGG,
K. R. (1984) Krashen’s
Monitor
HECHT,
K. (1991) Forschungsvorhaben. Fremdsprachenforschung 2, 119-126.
and Occam’s
K. et al. (1984) Brown & Hanlon of Child Language 2, 81-88.
HIRSH-PASEK, HOCKETT,
C. F. (1967) Where
Applied Linguistics 5, 79-100.
Razor.
Bericht tiber das Forschungsprojekt revisited: mother’s
“Lernersprache”.
Zeitschrtftfur
to ungrammatical
forms. Journal
sensitivity
slips, there slip I. Janua Linguarum 2, 900-936.
the tongue
HOLLEY, F. and KING, J. (1974) Imitation and correction in foreign language learning. In Schumann, J. and Stenson, N. (eds), New Frontiers in Second Language Learning, pp. 81-89. Rowley, MA: Newbury House. HOLTWISCH,
H. (1990) Fehler vermeiden
HORNER,
D. (1987) Acquisition,
learning
HORNER,
D. (1988) Classroom
correction.
KLEPPIN,
K. and KiiNIGS,
Korrekturverhalten
statt Fehler
Zielsprache Englisch 2, 7-12. System 15, 333-349.
zu verbessern.
and the monitor: Is it correct?
a critical
look at Krashen.
System 16, 213-220.
F. Cl. (1991) Der Korrektur auf der Spur. Untersuchungen zum miindlichen von Fremdsprachenlehrern. Bochum: Brockmeycr.
KONIGS, F. G. (1992) “Lernen” oder “Erwerben” revisited. Zur Relevanz fur die Sprachlehrforschung. Die Neueren Sprachen 91, 166-179.
KRASHEN,
S.
D. (1981)
der Zweitsprachenerwerbsforschung
Second Language Acquisition and Second Language Learning. Oxford:
Pergamon
Press. S. D. (1982) Principles and Practice in Second Language Acquisition.
KRASHEN,
Oxford:
LAVER, J. (1YXo) Monitoring systems in the neurolinguistic control of speech production. (ed.), Errors in Linguistic Performance, pp. 287-305. New York: Academic Press.
Pergamon
Press.
In Fromkin,
V. A.
LEE, B. S. (1950) The effects of delayed speech feedback.
Journal of the AcousticalSociety of America 22,24-26.
LEVELT,
in speech.
W. J. M. (1983) Monitoring
MACLAY,
H. and OSGOOD,
MCLAUGHLIN, MORGAN,
and self-repair
C. (1959) Hesitation
phenomena
Cognition 14, 41-104.
in spontaneous
English speech.
B. (1987) Theories of Second-language Learning. London:
J. L. and TRAVIS,
L. L. (1989) Limits on negative
information
Edward
Word 15, 19-44.
Arnold.
in language
input. Journal of Child
Language 16, 531-552. MORRISON, D. M. and LOW, G. (1984) Monitoring and the second language Schmidt, R. W. (eds), Language and Communication, pp. 228-250. London:
learner. In Richards, Longman.
J. C. and
MOTLEY, M. T. et al. (1983) Experimental verbal slip studies. a review and an editing model of language encoding. Communication Monographs 50, 79-101. NAIMAN, N., FRiiHLICH, M., STERN, H. H. and TODESCO, A. (1978) The Good Language Learner. Toronto:
Modern
Language
Center,
OXFORD, R. (1989) Use of language training. System 17, l-13.
Ontario learning
Institute
for Studies
strategies:
OXFORD, R. and NYIKOS, M. (1989) Variables affecting students. The Modern Language Journal 73, 292-300. RINVOLUCRI, RUBIN, SEMKE,
M. (1984) Grammar Games. Cambridge:
J. (1981) Study
H. D. (1984) Effects
of studies
choice of language Cambridge
with implications learning
University
strategies
for strategy by university
Press.
Applied Linguistics 11, 117-131. of the red pen. Foreign Language Annals 17, 195-202.
of cognitive
WENDEN, A. L. and RUBIN, Prentice-Hall.
in Education.
a synthesis
processes
in second
language
learning.
J. (eds) (1987) Learner Strategies in Language Learning. Englewood
Cliffs,
NJ: