Impact of Recasts On The Accuracy in Efl Learners' Writing (Ma Thesis)

GIRNE AMERICAN UNIVERSITY
FACULTY OF EDUCATION
DEPARTMENT OF ENGLISH LANGUAGE TEACHING

IMPACT OF RECASTS ON THE ACCURACY IN EFL LEARNERS` WRITING

MA THESIS
BY
OLGA DEGTEVA
Email: olga.williamson(bilingualage.com

SUPERVISOR:
ASST. PROF. DR. SONUC DIMILILIER

GIRNE
2011

TABLE OF CONTENTS
A8S18AC1v
ACknCWLLuCLMLn1Svll
LlS1 Cl 1A8LLSvlll

CHAPTER 1 INTRODUCTION .......................1
1.1.Background oI the problem.......................1
1.2. Purpose oI the study .............................2
1.3. Problem Statement and Research Question...................3
1.4. SigniIicance oI the study..........................4
1.5. Limitations and Delimitations oI the Study .................................................................... 5

CHAPTER 2 LITERATURE REVIEW ..................................................................................... 6
2.1. Present State oI AIIairs in the Research on Written Feedback in SLA ........................... 6
2.2. Direct vs. Indirect Feedback...........................8
2.3. Focused vs. UnIocused Feedback.........................9
2.4. Research on Recasts and ReIormulations....................10

CHAPTER 3 METHOD...........................17
3.1. Research Design...........................17
3.2. Operationalisation............................18
3.3. Participants................................19
3.4. Tasks, treatment and procedures....................20
3.4.1. Baseline.........................20
3.4.2. Treatment phase.........................20
3.4.3. Post-test.........................21
3.4.4. Questionnaire Ior research question N 3..................22
3.5. Data Analysis..........................22
3.5.1.Scoring.........................24
3.5.2. Statistical Procedures.......................24
3.5.3. Pre-test and Post-test Scores.....................26
3.5.4. Qualitative Analysis......................27

CHAPTER 4. RESULTS AND DISCUSSION..................28
4.1. Do Focused Recasts Encourage Learners to Attend to Form and Increase
Accuracy?......................................................................................................28
4.2. Are Some Types oI Errors More AIIected by Focused Recasts Than
Others?....................................................................30
4.3. How Do Learners Perceive Focused Recasts? What Is Their Attitude to Such
Kind oI Feedback?................................................................................40

CONCLUSION........................................41

REFERENCES................................44

APPENDICES.................................50
ALnulx 1 CPA81 1 1oLal scores of errors arLlclpanL !30
ALnulx 2 CPA81 2 1oLal scores of errors arLlclpanL k31
ALnulx 3 CPA81 3 lLembased and rulebased errors arLlclpanL !32
ALnulx 4 CPA81 4 lLembased and rulebased errors arLlclpanL k33
ALnulx 3 CPA81 n3 1he daLa dlsLrlbuLlon arLlclpanL k34
ALnulx 6 CPA81 n6 1he daLa dlsLrlbuLlon arLlclpanL !33
ALnulx 7LxAMLL Cl A 1Lx1 Wl1P CCuLu L88C8S36
ALnulx 8LxAMLL Cl A 1Lx1 Wl1P 8LCAS1S38

ABSTRACT
IMPACT OF RECASTS ON THE ACCURACY IN EFL LEARNERS` WRITING

Olga Degteva
MA Thesis, 2011
Supervisor: Asst. ProI. Dr. Sonuc Dimililier, ELT Dept.

Since the Iamous Truscott`s 'The case against grammar correction in L2 writing class
(1996) there has been an ongoing debate in SLA research about the value oI corrective Ieedback
and its diIIerent Iorms. A growing number oI empirical research is now investigating the
question, and although more and more evidence is obtained against Truscott`s statement, there
are still no deIinite conclusions about whether the Ieedback should be given, and iI yes, in what
Iorm.
The present study, designed as a longitudinal single-subiect study with two participants,
contributes to this research base, investigating one particular Iorm oI written corrective Ieedback
Iocused recast. During seven weeks thirteen written texts oI each participant (Iirst three served
as a pre-test, the last one as a post-test) were given Ieedback in the Iorm oI Iocused recasts and
then analysed Ior errors. All types oI errors were targeted in the study. Also item-based and rule-
based errors were considered separately to Iind out whether Ferris`s (2002) assumption about
treatable and untreatable errors could be conIirmed.
The results showed signiIicant decrease in the number oI errors immediately aIter the
baseline, and then steady downtrend throughout the treatment phase up to the post-test. The
comparison oI the pre-test and post-test scores let to conclude that recasts signiIicantly assisted
in increasing accuracy oI writing. Quantitative analysis showed that the number oI rule-based
errors decreased more than the number oI item-based errors.
Qualitative analysis oI the data oI one oI the participants showed, that item-based errors
were treatable. It also brought up the suggestion that item-based Ieatures cannot be treated as a
group. Each item is a single phenomenon which is not a part oI any grammatical system oI the
language, and unlike rule-based Ieatures, no generalization can be applied to item-based Ieatures.
The study showed that iI to take item-based errors as single phenomenon, then both rule-based
errors and item-based errors are equally treatable, thus questioning the ground Ior classiIying
errors as treatable and untreatable.
The study also suggests a direction oI the Iurther research on the eIIect oI recasts on the
complex Ieature systems, such as Conditional III or Modal Verbs Ior expressing possibility in the
past. These Ieatures Iailed to be corrected through recasts, but due to their complexity a longer
study is needed to investigate the possibilities oI recasts.

Key words: EFL writing, corrective Ieedback, recasts.

ACKNOWLEDGEMENTS

I would like to express my gratitude to Asst. ProI. Dr. Sonuc Dimililer Ior her guidance
during the term oI my candidature and her invaluable assistance and encouragement in
completing this work.
I would also like to express my indebtedness to Assoc. ProI. Dr. Sleyman D. Gker Ior
helpIul comments on my Iinal draIts.
Also I owe special thanks to Assist. ProI. Dr. UzuI Azmun Ior his essential help and
assistance.

LIST OF TABLES
Table Page
Table 1. Total scores oI errors with mean values (MV), coeIIicient oI
variation (CV) and standard deviation values (SD). Participants J and
K................................................................................................ 25

Table 2. Item-based and rule-based scores oI errors with mean
values, coeIIicient oI variation and standard deviation values.
Participant J........................................................................................ 25

Table 3. Item-based and rule-based scores oI errors with mean
values, coeIIicient oI variation and standard deviation values.
Participant K...................................................................................... 25

Table 4. Level change evaluation...................................................... 26

Table 5. Pre-test and post-test scores................................................ 26

Table 6. CoeIIicient oI variation in the baseline and treatment........ 29

Table 7. Pre-test - post-test number oI errors in articles, prepositions
and tenses. Participant K............................................... 31

Table 8. WS error (RB). Participant K.............................................. 32

Table 9. Present and past participles as adiectives (RB). Participant
K....................................................................................................... 33

Table 10. IndeIinite pronouns (RB). Participant K..........................34

Table 11. Modal verbs, possibility in the Iuture (RB). Participant
K........................................................................................................ 34

Table 12. Participle clause (RB). Participant K................................ 35

Table 13. Complex obiect. Participant K.......................................... 35

Table 14. Examples oI corrected IB Ieatures. Participant K............. 36

Table 15. Examples oI IB errors that were used incorrectly in
subsequent writings aIter they had been recasted. Participant K...... 39

CHAPTER 1. INTRODUCTION

1.1. Background of the Problem

Corrective recasts have been studied as a corrective Ieedback in speaking in more than 60
descriptive, quasi-experimental and experimental studies. As Long (2007) claims, there is
mounting support Irom research in SLA that recasts Iacilitate language development (p. 76).
Studies by Doughty and Varela (1998), M. Ishida (2002), Choi (2000), Ortega and Long (1997)
researched recasts in speaking in SLA and obtained results Iavouring recasts.
Focused recast is a corrective negative Ieedback that iuxtaposes a deviant non-target
language Iorm with its correct target-language Iorm without explicit explanation, which allows a
learner to contrasts two Iorms and see an error while attending to meaning thus assisting
acquisition as opposed to learning. In speaking Iocused recasts consist oI the repetition oI the
deviant learner utterance with rising intonation, Iollowed immediately by a corrective recast that
is always delivered with Ialling intonation (Long, 2007).
While Iocused recasts in speaking have been investigated actively and there is a bigger
number oI studies that Iound them quite an eIIective way oI Ieedback (Iwashita, 2003; Leeman,
2003; Long, Inagaki & Ortega, 1998;) than a number oI studies that claimed recasts less eIIective
than other types oI Ieedback (Lyster, 1998b; Lyster & Ranta, 1997), writing recasts unIortunately
cannot boast with such Ilourishing research. However, both writing and speaking being
productive skills share much in common. It is quite logical to suppose that what works Ior the
one can as well work Ior the other.
The eIIicacy oI recasts Ior promoting language development lies in the immediate
iuxtaposition oI the learner`s error and the correct reIormulation provided by the more advanced
speaker (Farrar, 1990; Long, 1996; Saxton, 1997, 2005). Unlike direct error corrections, recasts
are contextualised, they allow to Iocus on a problematic item not in isolation, but in a context.
Such contextualised Iocus can lead to acquisition oI item-based phenomena (like prepositions,
collocations and set phrases) and generalisation oI rule-based phenomena (like articles, tenses,
etc.).
Computer Mediated Communication (CMC), which is actually Internet-mediated
communication, has at last attracted attention oI SLA researchers to its huge educational
potential. Pelletieri (2000) states: "Because CMC Iosters negotiation oI meaning and Iorm-
Iocused interaction and because students communicating through this medium have more time to
process and monitor the interlanguage, I believe that CMC can play a signiIicant role in the
development oI grammatical competence." ( p. 83). The emergent environment oI interactive
internet-mediated written discourse is still a largely unexplored pedagogical context Ior recasts.
As Long (2006) notes, there is a need Ior Iurther research in this area because the written
modality is a robust environment Ior manipulating the degrees oI saliency oI target items, and
Iindings Irom this line oI research may have important pedagogical implications Ior teachers and
materials developers in distance language instruction programs.

1.2. Purpose of the Study
The present study attempts to investigate the eIIect oI corrective Ieedback in the Iorm oI
Iocused recasts on the development oI L2 learners literacy in writing.
The main Iocus oI the study is on learners' linguistic errors, which are subiect to
consistent treatment by means oI Iocused recasts and whether such treatment can lead to
improved literacy over time. Also a developmental change oI each problematic item that comes
into contact with a recast is investigated, thus, because the study is contextual, it allows a deeper
insight into a process oI acquisition.
Also the learners' attitude and response to such type oI corrective Ieedback is targeted in
this study. Attention to the learner's perspective on corrective Ieedback is oI a primary
importance, Ior it is through comprehension oI how a learner processes and uptakes Ieedback,
that a teacher can better understand when a certain type oI Ieedback is more eIIective and
suitable.
The study assumes that the Iindings might contribute to the improvement oI the
techniques oI teaching writing in an Internet-mediated teaching/learning environment.

1.3. Problem Statement and Research Question

Out oI all types oI written corrective Ieedback, recasts alongside with reIormulations stand
out as the types oI Ieedback that while Iocusing on errors are also Iocused on the context.
There is an apparent similarity between reIormulation and Iocused recast, since Iocused
recast 8 a reIormulation. The diIIerence however lies in the Iocused character oI recasts as
opposed to indirect reIormulation. ReIormulations happen at a text level, so that to notice an
error a learner has to compare two texts - original and a reIormulated variant and Iind original
and reIormulated items. Whereas recasts happen at a sentence level. An underlined sentence
containing an error is immediately iuxtaposed to a correct sentence, and although errors are not
directly marked, to Iind problematic items within an original sentence by comparing it with its
recasted variant is easier, thus it is easier to notice and process the errors. Yet, unlike error
corrections, which draw attention to a grammatical Iorm as it is, excluding the context, recasts
provide the correct Iorm within the context, pointing at interdependence oI Iorm and meaning.
In the long run continuous written recasts oI the same erroneous item in diIIerent contexts can
enable a learner to generalise about rule-based language items and acquire item-based
phenomena. Moreover, like reIormulations, recasts allow Ior native-likeness oI style that is
especially valued by learners (Santos, Serrano & Manchon, 2010). So preserving all positive
sides oI reIormulations, recasts at the same time should be easier to notice and process.
These assumptions Iorm the Ioundation oI the present study.
This study considers the Iollowing research questions:

1. Do Iocused recasts encourage learners to attend to Iorm and increase accuracy?
2. Are some types oI errors more aIIected by Iocused recasts than others, or not?
3. How do learners perceive Iocused recasts? What is their attitude to such kind oI
Ieedback?

The Iirst two questions deal with the eIIect the recasts may or may not have on the
development oI L2 literacy. To put it simpler, the questions imply whether Iocused recasts can
lead to a change oI L2 knowledge, and oI what kind this change can be.
The second question was prompted by the Ferris's (2002) argument that iI a grammatical
Ieature is clearly rule-based, it is more treatable than when a Ieature is item-based.
The third question deals with the learners' reaction to recasts. It probes into learners'
response to such kind oI Ieedback, what kind oI response it is and whether this response Iacilitate
learning

1.4. Significance of the Study
The signiIicance oI the study lies in its attempt to discover a new way oI teacher's response
to learners' writing which can render help in improving L2 learners accuracy.
It seems especially relevant now, when the vast and unceasing argument is carried on
between supporters and opposers oI a Iorm-oriented written Ieedback, while teachers-
practitioners are seeking ways to deal with grammar illiteracy in their writing classes. There is
also no agreement among the supporters on which types oI Ieedback are more eIIective. So, a
deep study into a less researched Iorms oI written Ieedback can bring results oI a great practical
value.
There is also another important issue to be taken into consideration. In the modern world
oI electronic communication Iorms oI language teaching/learning are being reconsidered. With
the appearance oI Internet-mediated teaching/learning appropriate teaching methods are
demanded in diIIerent spheres including writing. The Iindings oI the present study contribute to
the development oI teaching/learning writing via Internet communication.

1.5. Limitations and Delimitations of the Study
Limitations oI the present study come with the chosen design. Single subiect research may
produce results that have strong internal validity, when all internal validity threats are addressed.
However, due to the small number oI study participants, single subiect research tends to have
poor external validity, limiting the ability to generalise the Iindings to a wider audience. Indeed,
the results oI the study may reveal positive eIIect oI a treatment Ior a particular subiect, but the
same conclusion cannot be made Ior all L2 learners.
However, a single subiect study involves careIul examination and description oI
participants and conditions, as well as detailed deIinition oI a target behaviour (in our case
accuracy in writing) and a used treatment. The assumption is that under the same conditions
(Internet-mediated teaching/learning environment) other subiects oI the same proIiciency level
and with the same level oI motivation can achieve the same results iI the same treatment is
applied.
Moreover, the study can be replicated under diIIerent conditions to investigate whether
the same treatment proves to be eIIective Ior example in a classroom setting.
On the whole, the research is cumulative. More studies oI the same nature will provide
opportunity Ior wider generalisation.

CHAPTER 2. LITERATURE REVIEW

Perhaps oI all the Iour skills that are taught in ELT writing is the most challenging due to
many Iactors inIluencing literacy development in L2. However, the importance oI writing as a
language skill is diIIicult to overestimate and over years the interest oI researchers to writing has
increased. The present chapter is reviewing some oI the most interesting studies on written
Ieedback in SLA.

2.1. Present State of Affairs in the Research on Written Feedback in SLA

The question oI corrective Ieedback in SLA Ior years has been one oI the most conIusing
problems both Ior students and teachers due to the contradictory attitude to error correction. The
opinions stay divided into two main streams: against corrections (Krashen, 1982; Truscott, 1996;
2007) vs Ior thoughtIul appropriate corrections (Chandler, 2003; Bitchener & Knoch, 2008;
Sheen, 2007). Within the second group the research on diIIerent types oI Ieedback is Ilourishing
to discover what thoughtIul appropriate corrections actually are.
The attitude to writing has always been as to "a secondary Iorm oI language - highly
dependent upon the more primary oral Iorms (listening and speaking) (MacArthur, Graham, &
Fitzgerald, 2006, p. 171). Also students' writing has been always interpreted as either a Iinished
product to be evaluated by a teacher (product-oriented approach) or a series oI draIts each
commented by a teacher on Iorm, vocabulary, organisation and content ideally leading to
creating a good balanced written composition (process-oriented approach) (Ferris, 2003, p. 20).
Hence, the two lines oI discussion about the role oI teacher's Ieedback and how it can aIIect (iI it
can) the development oI writing skills: discussion about content Ieedback and error Ieedback (or
corrective Ieedback on Iorm).
By the end oI the last century Ieedback on Iorm was widely unpopular "due no doubt to
the prominence oI the process-writing paradigm in ESL writing classes at the time with its
consequent de-emphasising oI sentence-level accuracy issues" (Ferris, 2003, p. 42). Many
writing teachers were discouraged to Iocus on Iorm, persuaded by Krashen's Natural Approach
which stated that iI students' content were emphasised, appropriate Iorm would Iollow naturally,
as it does in children's L1 oral acquisition (Krashen, 1984). Perhaps the most severe opponent to
corrective Ieedback on Iorm, or grammar correction, has been John Truscott. In his review article
" The Case Against Grammar Correction in L2 Writing Classes" (1996) he straightIorwardly
calls to teachers to totally abandon grammar correction, because: " (a) Substancial research
shows it to be ineIIective and none shows it to be helpIul in any interesting sense; (b) Ior both
theoretical and practical reasons, one can expect it to be ineIIective; and (c) it has harmIul
eIIects" (p.328). The author argues precisely against grammar correction as "correction oI
grammatical errors Ior the purpose oI improving a student's ability to write accurately" (p. 329).
His conclusion is that no matter what Iorm or type oI grammar correction a teacher uses, it will
not have any positive eIIect on literacy improvement.
However, with time it became more and more obvious that such content-oriented
approach in writing was not resulting in disappearance oI students' errors, that students' lack oI
accuracy might well be held against them in various academic and proIessional contexts, and that
students themselves were Irustrated by the lack oI grammar Ieedback instruction.
That has brought teachers and researchers back to the issue oI corrective Ieedback on
Iorm.
A number oI studies on the eIIect oI corrective Ieedback have suggested that the Iorm-
Iocused Ieedback can be eIIective (Fatham & Walley, 1990; Ferris, 2002; Chandler, 2003), and
even desired by learners (Grami, 2004). Also now researchers have been looking at diIIerent
types oI Ieedback to evaluate each oI them. The Iindings about traditional types oI Ieedback on
Iorm are contradictory.
There are Iour main types oI traditional corrective Ieedback that have been studied:
explicit correction (the error is Iully corrected by a teacher), marking mistakes without
explanations (the problematic issue is underlined, no explanation is given), number oI errors per
line in a margin (no explanation or location oI the errors is given), and correction codes (special
signs to indicate diIIerent types oI mistakes, e.g W. W stands Ior a wrong choice oI word), that
label an error, but do not correct it, so that a learner has to correct them himselI.
In the study by Robb, Ross and Shortreed (1986) all Iour types were tested. The students
were assigned to Iour groups with diIIerent types oI Ieedback and were told then to rewrite their
essays. No group showed signiIicant improvement. Some authors thereIore have drawn a
conclusion that corrections in writing do not work (Grey.R, 2004). Others say that grammar
Ieedback that indicates the place but not the type oI errors (marking mistakes without
explanation) give better results (Fathman & Walley,1990) or that only written Ieedback coupled
with student-teacher conIerencing is eIIective (Brender, 1998).
ThereIore, taking into consideration the two contradictory conclusions that corrective
Ieedback is necessary, but traditional types oI error Ieedback do not always show signiIicant
impact on the improvement oI literacy, there has obviously been a need Ior investigation oI other
types oI corrective Ieedback that could be helpIul in literacy development oI L2 students.

2.2. Direct vs. Indirect Feedback

However, not all researchers and teachers are so categorical about corrections.
Indeed, despite the Iact that the Ieedback on Iorm was severely iudged and "Iound guilty"
oI uselessness and even causing harm, most teachers do use corrective Ieedback in class as well
as most students expect them to do so (Zacharias, 2007).
The supporters oI corrective Ieedback are now actively investigating the eIIect oI
diIIerent types oI correction on L2 learners accuracy. Especially big attention is drawn to direct
vs indirect Ieedback research. The Iindings are controversial, Irom "research in general has not
demonstrated that direct correction oI errors by teachers is eIIective in helping students improve
either the accuracy or substance oI their writing... indirect techniques such as noting the location
oI errors helps students improve their overall accuracy, both on subsequent draIts oI the same
paper and later assignments" (Ferris & Hedgcock, 1998, p. 127) to the conclusion that direct
error corrections lead to improved accuracy in immediate revisions and subsequent writing
(Chandler, 2003).
In the quasi-experimental study Bitchener and Knoch (2010) investigated the extent to
which written corrective Ieedback can help advanced L2 learners, who already demonstrate a
high level oI accuracy in two Iunctional uses oI the English article system and the extent to
which there may be a diIIerential eIIect Ior diIIerent types oI Ieedback on any observed
improvement. Sixty-three advanced L2 learners at a university in the USA Iormed a control
group and three treatment groups: (1) those who received written meta-linguistic explanation; (2)
indirect circling oI errors; and (3) written meta-linguistic Ieedback and oral Iorm-Iocused
instruction. On three occasions (pre-test, immediate post-test, delayed post-test) the participants
were asked to describe what was happening in a picture oI a diIIerent social setting. SigniIicant
diIIerences were Iound in the level oI accuracy on (1) the immediate post-test piece oI writing
between the control group and all three treatment groups; and (2) on the delayed post-test piece
between the control and indirect groups and the two direct treatment groups. The results oI the
study Iavoured direct written Ieedback over indirect one.
A longitudinal study by Lalande (1982), studies by Ferris (2003), Fratzen (1995) revealed
the advantage oI indirect Ieedback over direct corrections, while Chandler (2003) discovered that
direct Ieedback lead to the biggest gains in accuracy. Thus the results oI the studies exploring
direct vs indirect Ieedback are still inconclusive.

2.3. Focused vs. Unfocused Feedback

Recently a Iew attempts to study Iocused Ieedback have been undertaken (Sheen, 2007;
Rouhi & Samiei, 2010; Ellis, Sheen, Murakami, & Takashima, 2008). Among supporters oI
corrections research into such kind oI Ieedback (also known as selective correction as opposed to
comprehensive correction) has become widespread.
The researchers have deIined Iocused Ieedback as a Ieedback that selects speciIic errors
to be corrected and ignores the others (Ellis at all., 2008). The assumption is that iI attention and
understanding is important Ior acquisition, as cognitive theories oI SLA argue (Ellis, 2005;
Schmidt, 1994) then Iocused Ieedback must be more beneIicial to learners, because " they are
more likely to attend to corrections directed at a single error type and more likely to develop
clearer understanding oI the nature oI the error and the correction needed" (Rouhi & Samiei,
2010, p.5).
However, the results oI the studies are contradictory. While Sheen (2007) and Zacharias
(2007) reported the eIIicacy oI Iocused corrective Ieedback, others (Bitchener & Knoch, 2009;
Fazio, 2001; Truscott and Hsu, 2008; Rouhi & Samiei, 2010) Iound no greater improvement oI
accuracy in learners' writing when Iocused Ieedback was implemented compared to other types
oI Ieedback.
For example Ellis et al. (2008) undertook a comparative study to investigate the eIIect oI
Iocused vs unIocused written corrective Ieedback on the accuracy with which Japanese
university students used the English indeIinite and deIinite articles to denote Iirst and anaphoric
reIerence in written narratives. The research was designed as a pre-test - immediate post-test -
delayed post-test study. The Iocused group received correction oI iust article errors on three
written narratives while the unIocused group received correction oI article errors alongside
corrections oI other errors. Both groups gained Irom pre-test to post-tests on both an error
correction test and on a test involving a new piece oI narrative writing and also outperIormed a
control group, which received no correction, on the second post-test. The conclusion is that
corrective Ieedback is equally eIIective Ior the Iocused and unIocused groups and there is no
evidence in Iavour oI Iocused Ieedback.
It has also been reported that both students and teachers preIer comprehensive error
Ieedback (Lee, 2004).

2.4. Research on Recasts and Reformulations

Written recasts have been the least investigated type oI written Ieedback, though as it was
mentioned previously it has been quite heavily researched in speaking with the results mostly
Iavouring recast. In writing there are quite Iew studies probing into such kind oI Ieedback.
Long (2007) deIines corrective recast as "reIormulation oI all or part oI a learner's
immediately preceding utterance in which one or more non-target like items are replaced by the
correspondent target language Iorms, and where, throughout the exchange, the Iocus oI the
interlocutors is on meaning, not language as obiect...so error correction is implicit and
incidental" (p.77).
OI course taking into consideration the nature oI writing, written recasts are less implicit
and incidental than oral recasts, yet less explicit than direct error corrections, since the Iocus is
on a sentence as a unit, not on an error as such. There is one more issue that seems to give
written recasts an advantage over an oral recast. While oral recasts have been Iound by some
researchers ambiguous and thus ineIIective (Lyster, 1998b), written recasts cannot be taken by
learners Ior mere repetition oI their utterance, so the question oI ambiguity is eliminated here.
ReIormulations as well as recasts also involve replacement oI non-target items by the
target Iorms, but because they deal with rhetorical Iactors as well as grammar, lexis and syntax
(Levenston, 1978), the diIIerence as Cohen (1989) explains it, is that a reIormulator should
'rewrite the paper so as to preserve as many oI the writers` ideas as possible, while expressing
them in his/her own words so as to make the piece sound native-like (p. 4). Not only does it
present a big time-consuming work Ior teachers, but also a tremendous diIIiculty Ior learners to
detect errors, since in a reIormulated text problematic items can be completely paraphrased by a
teacher or even omitted.
In the study by Sachs and Polio (2007) reIormulations and error corrections have been
researched Ior their eIIect on improving learners' grammatical accuracy. The study was designed
as a three-stage composition-comparison-revision task. Concurrent verbal protocols were
employed during the comparison stage in order to study the learners' reported awareness oI the
more targetlike reIormulations. The reactivity oI think-alouds as a research tool was also
assessed. First, 15 adult learners oI English participated in a repeated-measures study with three
experimental conditions: error correction, reIormulation, and reIormulation think-aloud.
Participant reports oI awareness in the reIormulation think-aloud condition suggested that
noticing oI Ieedback was related to the accuracy oI subsequent revisions. A second nonrepeated-
measures study was then carried out with 54 participants; a control group was added and the
design was modiIied in an attempt to eliminate the reported tendency oI learners to develop and
use memorisation strategies while processing the written Ieedback. In both experiments,
participants perIormed signiIicantly better in the error correction condition than in the
reIormulation condition. Error corrections were Iound to be more eIIective.
The Iindings are not surprising however,taking into consideration the complexity oI error
noticing and detecting with such kind oI Ieedback. The advantage oI reIormulations is in their
ability to oIIer learners a sample oI native-like written text that speaks their ideas and thoughts,
this is what learners themselves value in such kind oI Ieedback. Santos et al.(2010) attempted to
compare the eIIect oI direct error correction vs reIormulation on the Ioreign language learners oI
intermediate level oI proIiciency in L2. ReIormulation in the research presented an original
learner's text reIormulated by a teacher to make it look native-like. ReIormulations concerned
erroneous grammar Iorms and inappropriate choice oI words. Their Iindings were as Iollowed:
Written corrective Ieedback on the whole has positive eIIect on noticing and uptake.
Error correction has clear advantage over reIormulation.
There are other important Iindings to be noticed in connection with the mentioned study
by Santos et al. The Iirst is the existence oI individual diIIerences between the learners in
processing corrective Ieedback. This issue was oI a primary importance when the research design
was considered, and will be discussed later.
The other interesting Iact mentioned by Santos is learners' attitude to error corrections
and reIormulations. The participants were interviewed about both types oI Ieedback and
maniIested that:
Their preIerence oI error corrections over reIormulations is explained by the Iact that
error corrections are easier to detect and notice
ReIormulation was a completely new type oI Ieedback Ior them, and they would need
time to get used to it.
All participants noted, that reIormulations could greatly improve their mastery oI
language, since "they will be processing their own writing with near-native style" (p.149).
Thus, despite the complexity and diIIiculty Ior error detection, reIormulations have the
advantage oI presenting native-like language.
In a quasi-experimental study by Doughty and Varela (1998) the Iirst time the eIIect oI
Iocused recasts in writing alongside with speaking was tested. In the case oI written works oI the
learners errors involving the target structures were circled, and written recasts iuxtaposed. The
learners oI the treatment group showed signiIicant gains in speaking and in writing.
Written recasts have been investigated to check their eIIect on the acquisition oI speciIic
language Ieatures, like Ior example in the study by Ayoun (2001) in which written recasts were
tested against models and traditional grammar instructions in acquisition oI 5a88G com5o8G and
m5arfat in French with a pre-test, repeated exposure, and post-test design. Participants were
randomly assigned to 1 oI 3 conditions: R (recasting: implicit negative Ieedback), M (modelling:
pre-emptive positive evidence), and G (grammar: explicit positive evidence and negative
Ieedback). The M and R groups read a diIIerent story with illustrations each week. The M group
was presented with a sentence corresponding to the illustration Ior 3 seconds, then the
participants were asked to answer a related question. The R group had to Iorm a sentence using
given elements based on the illustration, then the participants were exposed to the correct answer
Ior 3 seconds. The G group read traditional grammar lessons, took a short practice, and were
presented with the answer key. Post-test results revealed that the R group perIormed signiIicantly
better than the G group but not the M group. The conclusion was drawn that recasting is the most
eIIective Iorm oI Ieedback alongside with models.
In 2004 Ayoun carried out a Iollow-up study on the eIIectiveness oI written recasts vs
models and traditional grammar instruction, and the development oI temporality in the
interlanguage oI French college students as second language learners. The results oI the study
weakened the previous Iinding. The G group outperIormed the R and M groups in accuracy and
Irequency in the 5a88G com5o8G all three groups decreased in their accuracy and Irequency in the
m5arfat, but the G group outperIormed the other two groups; the three groups were practically
identical in the production oI various predicate types in the 5a88G com5o8e and m5arfat on the
pre- and post-tests. Overall, it was the G group's perIormance that showed a greater aspectual use
oI predicate Irequency and type in the m5arfat. thus questioning the conclusions oI the
previously conducted research.
Z. Han (2002) investigated recasts in a small-scale empirical study designed s a quasi-
experiment. Eight L2 learners oI English were randomly assigned to a recast and a nonrecast
group. The study adopted a pretest, posttest, and delayed posttest design, with eight pedagogical
recast sessions between the pretest and the posttest Ior the recast group paralleled by eight
regular sessions Ior the nonrecast group. Data collected consisted oI written and oral narratives
primed by cartoon strips and produced by the subiects in both groups. Recasts appeared to be
successIul in this study in that they heightened the L2 learners' awareness and led to considerable
improvement in their tense consistency during oral and written perIormance. Importantly, the
study identiIied Iour conditions that may be necessary Ior recasts to Iacilitate learning:
individualized attention, consistent Iocus, developmental readiness, and intensity.
Mohammadi and Javadi (2009) researched the eIIect oI written recasts vs explicit
negative Ieedback in the acquisition oI the auxiliary verb 'do/did and the irregular vs. regular
Iorm oI verbs in the simple past tense. Twenty students and a teacher participated in the study.
Students were divided into two groups. The learners in the treatment group received recasts
while learners in the control group received the same instruction as the experimental group plus
explicit negative Ieedback. Pictures were used to elicit written output and a written test was
perIormed to collect the data. The analysis oI the data indicated out-perIormance oI participants
in the experimental group over that oI the control group.
Some studies have Iound recasts less eIIective when compared to other types oI Ieedback
(Ellis, Loewen & Erlam, 2006), or even ineIIective (Ayoun, 2004; Sauro, 2007).
Sauro, Ior example, compared recasts and metalinguistic Ieedback through computer
mediated communication on the development oI L2 knowledge and production accuracy. High
intermediate to advanced adult Swedish university learners oI English (n23) Irom an intact
class were randomly assigned to one oI three conditions (two Ieedback conditions and one
control) and were randomly paired with English native speakers. During task-based interaction
via text-chat, the Swedish learners received Iocused corrective Ieedback on omission oI the zero
article with noncount nouns oI generic reIerence (e.g. employment, global warming, culture).
Pretests, post-tests and delayed post-tests oI knowledge (acceptability iudgments) and production
accuracy (short-answer questions) measured learning outcomes. Results showed a signiIicant
advantage Ior metalinguistic inIormation on the immediate development oI target Iorm
knowledge with noncount nouns oI generic reIerence that had been introduced during the
intervention.
Among studies that suggest a beneIit Ior Ieedback which contains positive evidence is
Leeman's (2003) investigation oI two components oI recasts (negative evidence and the
enhanced salience oI positive evidence), which Iound an advantage Ior Ieedback that contained
only positive evidence over Ieedback that contained only negative evidence.
Research that suggests a superior beneIit Ior corrective Ieedback that generates modiIied
or pushed output or repair includes Lyster's (1998a) study oI French immersion classes. Written
posttest results showed a signiIicant advantage Ior students receiving prompts (written Ieedback
which prompts learners to attempt selI-repair) while students in the recast condition perIormed
similarly to students in the no Ieedback condition. Similarly, McDonough's (2005) study oI Iour
corrective Ieedback combinations included two groups that received types oI corrective Ieedback
that allowed them to modiIy output and two groups that did not. The results indicated that the
number oI learners who progressed to a more advanced level oI question Iormation was greater
Ior the Iirst two groups than Ior the latter.
The second oI these two studies also investigated whether corrective Ieedback Iacilitated
the development oI implicit and explicit knowledge. Ellis, Loewen and Erlam (2006) examined
learners' use oI the English past tense marker -ed Iollowing exposure to either explicit corrective
Ieedback (metalinguistic inIormation) or implicit corrective Ieedback (recasts).
Findings indicated that learners who received corrective Ieedback containing
metalinguistic inIormation signiIicantly outperIormed learners in the recast and control groups
on tests oI both implicit (oral elicited inIormation) and explicit (grammaticality iudgments) L2
knowledge. Furthermore, metalinguistic corrective Ieedback and not recasts also seemed to
promote generalization oI the -ed Iorm to new contexts.

The study that have been brieIly reviewed above present diIIerent, sometimes
contradictory conclusions. What is clearly seen Irom the review is that, although the research on
oral recast is plentiIul, it is still inconclusive. Scarcer and even less inconclusive is the research
on written recasts as a Iorm oI corrective Ieedback. To generalize, more studies on recasts are
needed.

The literature review has shown that:
a) there is a continuous argument about whether corrective Ieedback should be
implemented in SLA or not,
b) among the supporters oI written corrective Ieedback there is still no agreement as to
what type oI Ieedback is most eIIective,
c) numerous study produce diIIerent results, hence no possibility Ior deIinite conclusions,
d) most oI the studies conducted on written recasts so Iar have been oI group
experimental design, probing into one grammatical phenomenon, but no single-subiect
longitudinal study has been yet undertaken to see how sustained written corrective Ieedback n
the Iorm oI recasts can aIIect learners` accuracy on an individual learner level.

CHAPTER 3. METHOD

3.1. Research Design

Single-subiect AB research design has been chosen as a Irame Ior the study. The design
allows Ior deep qualitative analysis oI the data, natural setting oI the study, attention to context
and individual characteristics oI the participants oI the study. There are several reasons Ior
choosing such design.
Firstly, a deep and true commitment oI the author to personal individualised approach in
pedagogy. As Kumaravadivelu (2001) notices: "language pedagogy, to be relevant, must be
sensitive to a particular group oI teachers, teaching a particular group oI learners pursuing a
particular set oI goals within a particular institutional context embedded in a particular
sociology-cultural milieu". These particularities play a decisive role in language learning,
making doubtIul the necessity oI external validity oI the language research. Indeed, "it is
reasonable to claim, that generalisations oI the Iindings and a-one-size-Iits-all perspective
towards the Iindings in SLA research is a wrong assumption" (Navidinia & Eghtesadi, 2009,
p.58) and a comprehensive approach must be considered when studying a phenomenon in SLA
(Ortega, 2005).
Secondly, to choose iust one grammatical phenomenon as an obiect Ior recast treatment is
a common point oI most oI the studies on written recast: tenses were investigated by Han (2002),
auxiliary /o/// by Mohammadi and Javadi (2009), grammatical gender by Dasse-Askildson
(2008), the Past Tense by MaItoon, Ahmadi and DaIIariIard (2010), the Past Tense and
conditional Iorm by Doughty and Varela (1998), articles by Bitchener and Knoch (2009), etc.
But no study has yet been undertaken to investigate eIIects oI extensive longitudinal written
recast treatment on L2 accuracy, though such study would allow not only Ior comparison oI how
diIIerent error categories are aIIected by recast treatment, but, what is more important, Ior
getting a whole picture oI dynamic developmental process oI learning writing. The task that a
researcher undertakes is huge and involves continuous treatment and Iollowing the
developmental change oI every erroneous item in a learner's writing Ior a certain period oI time,
but it can help obtain interesting results.
And the last reason, but not the least: Doughty and Williams (1998) argue that the eIIects
oI any corrective Ieedback are "gradual and cumulative rather than instantaneous and
categorical" (p. 40). Lyster and Ranta point out how complex the process oI L2 learning is
(1997). Taking into consideration the slow, non-linear and partial nature oI many oI the
processes involved in SLA other kinds oI studies should be undertaken, "Iuture studies should
thereIore be longitudinal in nature in order to allow Ior longer periods oI exposure than has been
the case to date" (Tatawy, 2006, p. 15).
Hence a longitudinal single-subiect AB research design was adopted Ior this study.

3.2. Operationalisation

Errors/ mistakes were operationalised as non-target Iorms or/and non-target-like usage oI
target Iorms in the written samples oI the participants. No diIIerentiation was made between
errors and mistakes due to the impossibility to deIine the nature oI a problem with each
erroneous item in the writing (whether a problem was due to the lack oI knowledge, or due to the
other Iactors involved).
Accuracy was operationalised as usage oI target language Iorms and target-like usage oI
target Iorms in the written samples oI the participants.
Recast was operationalised as a reIormulation oI a non-target Iorm or a non-target-like
usage oI a target Iorm within a sentence into a target Iorm or target-like usage oI the target Iorm
within the sentence. Recasts reIormulated ill-Iormed utterances in their entirety (no partial or
segmented recasts was provided). Unlike other types oI Iocused corrective Ieedback which are
Iocused because they concentrate on treating one or two grammatical errors, recast in the present
study was deIined as Iocused, because it Iocused learner's attention on a language unit containing
an error; in the study such unit was a sentence.
Rule-based errors were operationalised as errors relating to Ieatures that occur in "a
patterned, rule-governed way" (Ferris, 1999, p. 6).
Item-based errors were operationalised as errors relating to unsystematic Ieatures oI the
language that are not governed by the rules.
All the errors in the participants' writing were subiect to corrective Ieedback in the Iorm
oI Iocused recasts.
No additional emphasis was placed on the source oI the error.

3.3. Participants

Reviewing research on the eIIectiveness oI recasts in Iirst and second language
acquisition Nicholas, H., Lightbown, P. M., and Spada, N. (2001) concluded that recasts appear
to be most eIIective in contexts where it is clear to the learner that the recast is a reaction to the
accuracy oI the Iorm, not the content, oI the original utterance. This conclusion was later
conIirmed by Dasse-Askildson (2008) in the experimental study oI the eIIect oI written recasts
on grammatical gender acquisition by L2 learners. Thus to be eIIective recasts should be
addressed to learners oI quite a high level oI proIiciency. This conclusion deIined the choice oI
the participants Ior the study: two upper-intermediate English L2 leaners were invited to part-
take in the study. Both participants were native Russian speakers. In 2008 they took a 5 month
course oI English in a language school in Russia, then aIter a short break they started individual
English lessons up to the end oI the year. In 2009 they took a course oI Internet-mediated
English lessons, which was Iinished in autumn. Since then both participants have been trying to
maintain their level oI English on their own.
There were also other requirements to participants which were determined by the research
design. In the AB type oI design a baseline "A" is tracked, and then some treatment "B" is
implemented. II there is a change then the treatment is said to have had an eIIect. By its nature, a
single-subiect design has quite high internal validity. Out oI 8 threats to internal validity in
experimental studies, only one is a concern in adult longitudinal studies - history (Abrahams,
1997). To control a history variable, ABAB and multiple baseline variant oI a single-subiect
design have been introduced.
However, in the present study neither ABAB nor multiple baseline design is applicable
due to the impossibility oI unlearning in the Iirst case, and a limited number oI participants in the
second. Hence the requirement Ior the participants with a minimised possibility oI learning Irom
sources other than the treatment "B" implemented in the study. Both participants were not taking
English lessons or communicating with other English speakers besides the experimenter up to
the last session. Between the last session and the test the participants had limited short-time
communication with English speakers while visiting Cyprus. The length oI the gap between the
let session and the test was too short Ior any learning to occur.
The participants were told that the study was Iocused on writing, but no detail about
Ieedback or recasts was provided to prevent the participants Irom being inIluenced by the
obiectives oI the study.

3.4. Tasks. Treatment and Procedures

3.4.1 Baseline

In the "A" stage three samples oI writing oI each participant were examined Ior errors,
thus serving as a pre-test. The Iirst sample (numbered as N0) was a piece oI writing done prior to
the beginning oI the study during the inIormal correspondence between the researcher and the
subiects. This was done Ior two reasons:
1) to have more material Ior error analysis in the baseline,
2) to see whether the awareness that they participate in the study could inIluence the
subiects' accuracy (i.e whether there would be a gap in score between sample N0 and samples N1
and N2).
3) to obtain three measures Ior establishing a valid baseline. Three measures are usually
named as a minimum data points required Ior a baseline (Alberto & Troutman, 2006).

3.4.2. Treatment phase

The writing samples oI the phase "B" consisted oI diary-entries and topic-based
compositions; each sample was submitted once a week.
AIter samples N1 and N2 (the last in the "A"phase) were submitted they were analysed
Ior errors, sentences containing erroneous items were recasted using Google Docs soItware (each
sentence containing an error was highlighted, recasted variant was placed in the margin opposite
the highlighted sentence) and both samples containing recasts were emailed back to the
participants beIore they started their next task. The same procedure was implemented Ior each
Iollowing sample.
The participants were not instructed how to process the Ieedback, nor were they required
to do revisions. It was up to them to decide in what way they could use the Ieedback

3.4.3. Post-test

The last session was designed as a post-test. Unlike the previous sessions it was neither a
diary-entry, nor a topic-based composition. The task represented a set oI questions, which the
participants were to answer in a Iree manner. The questions were designed to elicit a maximum
number oI structures and items that had been identiIied as problematic Ior the participants (that
had been used incorrectly in any oI the previous sessions) some time during the study.
The test consisted oI two parts. AIter the Iirst set oI questions was answered, the answers
were analysed in relation to the number oI elicited problematic structures and items. The purpose
oI the analysis was to Iind out which oI the problematic items the participants had used in their
answers and which not. Then the second set oI questions was presented to the participants in
the attempt to encourage them to use those problematic structures that they avoided to use while
answering the Iirst set oI questions.
The questions reIerred to the liIe experience oI the participants, no grammar questions or
any questions concerning language knowledge were asked. Thus the attention oI the participants
was still Iocused on the content as it had been all the way through the study.
The score Ior the test was calculated exactly as the scores Ior the previous sessions, and
was also mapped onto the graph.
3.4.4. Questionnaire for research question N 3

To address the third research question the participants were presented with a set oI
questions that they were to answer in writing aIter the test was completed. The questions were:
1) How do you Ieel about the teacher's Ieedback that you encountered in this study? Is it
helpIul? Does it help to notice errors?
2) What does such kind oI Ieedback lack in your opinion?
3) How did you handle the Ieedback? What exactly did you do with the teacher's
comments in the margin?

3.5. Data Analysis

The primary data Ior the study consisted oI thirteen written texts Irom each participant
(26 texts in total), the Iirst three serving as data Ior baseline condition, the next nine - as data Ior
treatment condition and the last one - as an immediate post-test.
Each text in the treatment condition was written aIter receiving corrective Ieedback in the
Iorm oI Iocused recasts on a previous text.
Each text was analysed Ior errors.
The errors were coded as Iollows:
O WW - wrong word (e.g ;ew meaning look)
O Spr - set phrase (using a set phrase incorrectly (e.g. 3 the o3e ha3/... o3 the
other ha3/ meaning fr8tlv... 8eco3/lv. or using direct translation Irom Russian: t /oe83t care
meaning I /o3t care)
O P - preposition (using a wrong preposition, omitting a preposition or using a
preposition where it should not be used)
O WS - wrong structure (omitting a subiect, incorrect construction oI any
grammatical structure apart Irom those mentioned separately in the present list)
O WO - word order
O Cond - conditional clause
O Partcl - participle clause
O Interr - wrong Iormation oI interrogative sentence
O Seq - sequence oI tenses
O Indirs - indirect speech
O CO - complex obiect
O Copulab - copula "be"
O PastSN - wrong Iormation oI Past Simple negative
O TF - wrong tense Iormation
O Gen - genetive
O Missedw - missed word
O Quant - quantiIier
O Demonp - demonstrative pronoun
O Determ - determiner
O Indp - indeIinite pronoun
O A - article
O WF - wrong Iorm (using an incorrect part oI speech: vou ha;e to iu8t g;e
vour warm to me. meaning warmth, incorrect verb Iorm: I ca3 fou3/ etc.)
O Nasa - noun as adiective (lea;e8 ocea3. 8tu/e3t8 lfe etc.)
O Partpre - participle as premodiIier
O InI - inIinitive
O Gerund
O Adv - adverb aIter "be" and " sound"
O S/Pl - singular or plural
O Poss - possessive
O T - tense (incorrect choice oI tense)
O Mod - modal verb
O SVOA - subiect verb obiect agreement (wrong subiect - verb or verb - obiect
agreement)
O Sp - spelling
Each sample was printed out, errors were underlined and a corresponding code Ior each
error was placed in the margin. At the end oI each sample a total number oI words and a total
numbers oI errors were indicated.
Also to address the third research question all the coded errors were classiIied into two
types: item-based Ieatures (prepositions, set phrases, wrong words and spelling) and rule-based
Ieatures (all the rest). The number oI errors oI each type was calculated in every sample.

3.5.1. Scoring

The score Ior each sample was calculated as a percentage, with total number oI words in
a sample taken as 100.
The scores Ior item-based and rule-based Ieatures were also calculated in the same
manner (as a percentage oI each type with the total number oI words in a sample taken as 100).
The obtained data were mapped onto 4 graphs (see appendix):
1) Total scores oI errors. Participant J.
2) Total scores oI errors. Participant K.
3) Item-based and rule-based (IBRB) scores oI errors. Participant J.
4) Item-based and rule- based (IBRB) scores oI errors. Participant K.

3.5.2. Statistical procedures

Visual and statistical analysis oI the obtained data Ior the interpretation oI experimental
eIIects was implemented. Analysis addressed three changes in the data pattern (Wolery & Harris,
1982): variability, trend, and level.
Variability and trend Ior the data were calculated using EXCEL analysis toolpak and
GraphPad SoItware Quick Calcs.
1) To analyse variability oI the data, standard deviation (SD) and coeIIicient oI variation
(CV) was calculated Ior the data in each phase. CV was calculated as the ratio oI the SD to the
mean value. The calculated standard deviation values, coeIIicient oI variation and level values
were included into the tables oI scores and then mapped onto the graphs (see Appendix Chart N5
and Chart N6).

Table 1. Total scores of errors with mean values (MV). coefficient of variance (CV) and
standard deviation values (SD). Participants 1 and K.

oase||re l(ealrerl
0 1 2 V\ 30 C\ 3 1 5 Z 8 9 10 11 T V\ 30 C\
J Z.Z Z.8 8.2 Z.9 0.2 0.03 1.1 3.1 1 1.8 2.5 3.9 2.9 3.1 1.9 1.1 3.2 1.09 0.31
K 9 8. 9.2 8.9 0.30 0.03 5.Z .3 1.3 5 5.1 3.2 1.3 1.5 3 2.3 1.1 1.2Z 0.28

Table 2. Item-based and rule-based scores of errors with mean values. coefficient of
variance and standard deviation values. Participant 1.

oase||re l(ealrerl
0 1 2 V\ 30 C\ 3 1 5 Z 8 9 10 11 T V\ 30 C\
l8 2.9 3.Z 2.8 3.1 0.19 0.1 2.2 1.5 1.Z 1.9 1.2 0.5 1 1 0. 0.Z 1.2 0.58 0.18
R8 1.Z 1 5.1 1.Z 0.Z 0.15 2.2 2 2.1 3 1.2 3.1 1.8 2.1 1.3 0. 2 0.81 0.12

Table 3. Item-based and rule-based scores of errors with mean values. coefficient of
variance and standard deviation values. Participant K.

oase||re l(ealrerl
0 1 2 V\ 30 C\ 3 1 5 Z 8 9 10 11 T V\ 30 C\
l8 2.Z 3 2.1 2.Z 0.3 0.11 1. 2.2 1.3 2. 1.Z 1 1.5 1.5 1.2 1 1.5 0.51 0.31
R8 .3 5.9 .Z .3 0.52 0.08 1.1 1 1.3 2.1 3.Z 2.1 2.8 3 1.8 1.3 2.9 0.91 0.32

The analysis revealed that all the data Iell within normal distribution. The coeIIicient oI
variance Ior each set oI data Iell in the range oI low-variance distribution (CV1). However, Ior
all three data sets oI each participant (total scores. IB scores and RB scores) there was observed
an increase in CV value Irom A phase to B phase.

2) Trend lines Ior each set oI data were calculated and mapped onto the graphs in EXCEL
analysis toolpak. Then the trends were analysed visually to establish the movement in the data
series. InsuIIicient uptrends were observed in A phase Ior each data set, and downtrends (with
diIIerent angle oI decrease) in B phase. (See Appendix Charts NN1-4).

3) Level was calculated as a mean Ior each phase. To analyse whether the diIIerence in
levels oI A phase and B phase was statistically signiIicant and scientiIically relevant, two- tailed
unpaired t test was run Ior the total scores (TS), item-based (IB) and rule-based scores (RB) in
the phases A and B. The results are shown in table N4 (the Iirst letter beIore a score type
indicates a participant).

Table 4. Level change evaluation

!1S k1S !l8 !88 kl8 k88
pvalue p00001 p00001 p00003 p00004 p00041 p00003
Mean of (A8) 47000 43000 19000 27000 1140 3330
Confldence lnLerval
93
3262361373
93
2823361747
93
1081827182
93
1317238828
93
04461834
93
19334747

The results oI the t test revealed that the drop in the level oI errors aIter the treatment was
implemented is statistically signiIicant (p0.05) Ior all the scores Ior both participants. Since
even the low ends oI the conIidence interval represent a large enough diIIerence to be considered
important, there is a diIIerence between condition means Ior both total and IBRB scores and the
diIIerence is large enough to be scientiIically relevant.

3.5.3. Pre-test and post-test scores

To analyse the scores oI pre- and post-tests the means oI the three scores oI the baseline
and the scores oI the post-test were compared. The results are shown in Table N5.

Table 5.. Pre-test and post-test scores

I15 k15 II8 Ik8 kI8 kk8
retest 79 89 31 47 27 63
osttest 14 23 07 06 1 13

The results show that the error scores Ior the post-test are lower than the mean error score
Ior the pre-test.

3.5.4. Qualitative analysis

Also qualitative analysis oI the written texts oI one oI the participants was undertaken. It
consisted oI examining the changes (or their absence) oI each type oI errors through all the
thirteen texts with the three texts oI the phase "A" taken as one. The purpose oI the analysis was
to discover a pattern oI evolution oI non target-like items into target-like items, and also to Iind
out whether there is a diIIerence in reacting to recasts between rule-based errors and item-based
errors. The data was organised in two ways:
1) A number oI each type oI error Ior every text was entered in the EXCEL spreadsheet,
1) Examples oI each coded type oI errors were organised in the EXCEL spreadsheet, thus
allowing to Iollow the progress oI the erroneous item Irom text to text providing it was used by
the participant aIter it had been recasted.

CHAPTER 4. FINDINGS AND DISCUSSION

The analysis oI the results is structured according to the research questions that guided
the study.

4.1. Do Focused Recasts Encourage Learners to Attend to Form and Increase
Accuracy?

The analysis oI the obtained data showed considerable decrease in the level oI errors aIter
the beginning oI treatment. It does contradict to the Truscott conclusion about ineIIiciency oI
written corrective Ieedback or even its harmIul inIluence. The error level drops suIIiciently aIter
the texts written in A phase were examined by the researcher, returned to the participants
containing recasts, and reviewed by the participants.
There are two possible explanations Ior this phenomenon:
1) Learning occurred aIter Texts N1-2 with the included recasted erroneous items texts
were reviewed by the participants,
2) Or, which is more probable, written Ieedback increased suIIiciently attention oI the
participants.
Here it Ieels appropriate to raise the question oI errors and mistakes. Although in the
study there was no distinction made between the two phenomena, aIter the results oI the baseline
and treatment phases were analysed, the question was brought up how errors could be eliminated
or decreased with the level oI attention raised. It is obvious that attention to a certain language
phenomenon can occur only when there is knowledge about the phenomenon. In other words,
when a learner is about to use a phenomenon that he knows is a part oI the language system, he
can retrieve some inIormation about the phenomenon either Irom his memory or Irom external
sources and use this inIormation Ior utterance construction. Whereas iI there is no knowledge
about the phenomenon at all, an increased level oI attention cannot possibly help to increase
accuracy in constriction an utterance containing this phenomenon.
So it is suggested that the drop oI error level immediately aIter the phase "A" (error here
is as it is operationalised in the study, i.e. errormistake) actually indicated the decrease in a
number oI mistakes (as opposed to errors) due to the increased attention oI the participants to
Iorm.
In the phase "B" the direction oI the trend lines Ior all the data sets indicates decline,
showing a steady increase in accuracy Ior both participants Ior total scores as well as Ior IB and
RB scores. With the increase oI accuracy we can also observe a decrease in stability oI the
perIormance.

Table 6. Coefficient of variation in the baseline and treatment

hase A hase 8
arLlclpanL ! Cv003 Cv034
arLlclpanL k Cv003 Cv028

In the treatment the error scores oI the participants were not stably reducing Irom session
to session, though the trend lines were generally downtrends.
Several Iactors can be hold accountable Ior the increase oI variability:
1) A given topic Ior the assignment can aIIect the language choice (more complex or
more simple language is chosen Ior a text). Descriptions Ior example can involve usage oI simple
tenses and structures, while event accounts usually demand more elaborate tense choice, and in
narrations learners may tend to use quite sophisticated structures. Thus the possibility oI error is
diIIerent Ior diIIerent types oI text.
Since the participants oI the study were not instructed precisely about what they were
expected to write, and only general outlines were given to them to deIine the topic, they Ielt Iree
to choose any type oI text they wished. This was done on purpose to insure that the learner's
personal interest in what they wrote about was maintained at a high level during the study.
2) Some learner-oriented variables can be at play, such as tiredness, low concentration,
low mood, etc.
The analysis oI the pre- and post-test scores revealed a suIIicient improvement in
accuracy Ior both participants (see Table N5, Chapter 3.5.3). Total error score Ior participant J.
reduced Irom the mean 7.9 Ior the pre-test to 1.4 Ior the post-test. Total error score Ior
participant K. reduced Irom the mean 8.9 Ior the pre-test to 2.4 Ior the post-test.
On the whole, the obtained results allow Ior a deIinite "yes" as the answer to the Iirst
research question. As a result oI corrective Ieedback in the Iorm oI Iocused recasts the
participants showed signiIicant decrease in the number oI errors right aIter the treatment was
implemented, and the decreasing tendency was maintained through the whole treatment phase up
to last session (the post-test).

4.2. Are Some Types of Errors More Affected by Focused Recasts Than Others?

Ferris (2002) argues that iI a grammatical Ieature is clearly rule-based, it is more treatable
than when a Ieature is item-based.
To examine this assumption the data were analysed in terms oI rule-based and item-based
errors.
As it is shown in Table N 4, Ior both participants the error level oI both IB and RB errors
drops suIIiciently in the phase "B". But the mean oI (A-B) IB errors is less than the mean oI (A-
B) Ior RB errors (JIB 1.9000, JRB 2.7000; KIB 1.140, KRB 3.350). The conclusion can be
drawn that RB errors are more treatable than IB errors as Ferris and a Iew other researchers
argue (Bitchener, Young, & Cameron, 2005).
Indeed, the numerical data obtained in the present study show how diIIerently IB and RB
errors reacted to treatment. As an example let us compare a number oI certain errors in the pre-
test and in the post-test. The errors taken Ior example are the same errors that Bitchener et al.
(2005) targeted in their study, that conIirmed Ferris's conclusion (2002) about treatable RB and
untreatable IB errors: prepositions (IB), articles (RB; zero, a and the articles were targeted in the
present study) and tense (RB; in the present study all tenses were targeted).

Table 7. Pre-test - post-test number of errors in articles. prepositions and tenses. Participant K.

Mean number of errors of ln Lhe
preLesL
number of errors ln Lhe posLLesL
A 286 3
96 4
1 76 1

The Iall in the number oI errors in articles and tenses is indeed more dramatic than the
Iall in the number oI erroneous prepositions. But there is one question to be considered. While
there is no doubt that the Iive erroneous articles and one erroneous tense in the post-test are the
same errors oI articles and tenses made in the pre-test (errors that were repeated during the whole
study, because there is a limited number oI rules that one can violate using tenses or articles),
and it is iust a number oI occurrences oI these errors that has actually come down, we cannot be
so sure with prepositions. Are the erroneous prepositions in the post-test the same that were in
the pre-test? The same question can be asked Ior other IB errors: wrong words, set phrases,
spelling. There are no deIinite rules to Iollow with IB Ieatures. While we can take an RB Ieature
as a sum oI several phenomenon that makes a system (e.g articles deIinite article indeIinite
article zero article), there is no system in set phrases or such errors as wro3g wor/, and even
prepositions are minimally systematic in English. So we cannot take them as a sum oI
phenomenon; each wrong word, or set phrase or usage oI a preposition should be considered
separately.

The qualitative analysis oI the texts revealed the Iollowing:
1) With RB Ieatures the smaller the system oI the Ieature is the more treatable the errors
are.

Table 8. WS error (RB). Participant K.

1exL Lxample
8asellne 1here was fresh calmlv
1here was fun
1here ls so sunnv ln Lhe room
non LaraeLllke usaae
n7 8uL here ls noL verv cold
ln sprlna Lhere ls noL much dlrLllv
ln summer Lhere ls noL sLuffv
1esL
1here lL was wlndv and abouL 23
dearees
1here lL was noL crowded on Lhe
beach
SomeLlmes lL was verv hoL
near LaraeLllke usaae

1araeLllke usaae

As it is shown in the table, the target-like usage oI the structure in question appears only
in the last session, but it is still accompanied with the two examples oI near target-like usage.
The shiIt to the target-like usage is obvious, but it is still not complete. The system oI this
language Ieature is not very simple, it consists oI the subiect t as an empty word, dummy, which
is necessary to maintain the strict word order oI the English language, plus there is a clear
interIerence oI the structure There to be. So what we can observe here is a separation oI the
two conIused structures and building two separate Ieature systems. Also both present a deIinite
degree oI diIIiculty due to partial overlapping with the L1.
When a Ieature system is simpler, and there is a direct correlation with the L1, the shiIt to
the target-like usage can take less number oI sessions, as it was the case with participles and
indeIinite pronouns in negative sentences in the study.

Table 9. Present and past participles as adjectives (RB). Participant K.

1exL Lxample
8asellne She and her moLher looked llke
soldlers no movlna no emoLlons
whlLe skln bla scarlna eves
n3 8lrds some anlmals walked Lhere
So scared and charmed
n11 She has a verv verv scared vlew 1araeLllke usaae
1esL 1he mosL exc|t|ng Lhlna ln mv Lrlp
was Lhe whole Lrlp from Lhe flrsL
dav Llll Lhe lasL one
8uL on Lhe oLher hand l had Lhe
mosL bor|ng Lhlna 3 hours bv
plane
8uL on Lhe oLher hand l was
scared a llLLle blL
l was noL scared aL all
1araeLllke usaae

Also with simple Ieatures iust one recast sometimes was enough to ensure the correct
usage oI the Ieature throughout the study. IndeIinite pronouns in negative sentences were used
correctly aIter the Iirst recast. As it can be seen Irom the table, one recast had quite a long lasting
eIIect on the Ieature.

Table 10. Indefinite pronouns (RB). Participant K.

Whereas articles and tenses Ior example constitute quite complex systems, and though the
number oI these types oI error decreased suIIiciently by the end oI the study, the participants still
made errors oI these types.

2) With RB Ieatures a learner adds a new constraint to his hypothesis oI the Ieature system
construction aIter each recast and tests the hypothesis in the subsequent writing, thus building his
knowledge oI the system, providing the Ieature is used Irequently in a set oI subsequent writings.

1exL Lxample
8asellne She dldnL care abouL her and
never Lold someLhlna abouL her
llfe
n3 1o look around and see nobodv 1araeLllke usaae
n3 8uL l see nobodv
l see noLhlna around
1araeLllke usaae
n6 l donL sav LhaL l undersLood
noLhlna
1araeLllke usaae
n8 l do noLhlna and some horrlble
LhouahLs sLop me
1araeLllke usaae
Table 61. Modal verbs. possibility in the future (RB). Participant K.

3) II a recasted RB Ieature is not used in the subsequent writing, and the interval between
the recast and the next time the Ieature is used is long, an error can be repeated in the same or
slightly modiIied way, no matter whether the Ieature is simple or not.

Table 12. Participle clause (RB). Participant K.

1exL Lxample
n3 Was walklna Lhrouah Lhe foresL
Lhe beam saw somebodv
Was looklna aL Lhelr boLh Lhe
beam LhouahL lll alwavs llsLen Lo
mv mum
1esL 8een walklna Lhrouah Lhe SL
Pllarlon CasLle l Lrled Lo lmaalne
Lhelr llfe

4) Some Ieatures were observed to shiIt Irom target-like usage to a non target-like usage (U-
shaped learning).

1exL Lxample
n4 Lvervone can do lL and l wlll able
Lo do
n3 ?ou are small veL and lL wlll be
able Lo be danaerous enouah for
vou
n6 lf vou are 8usslan lL |ght be
easler Lo undersLand vour Lnallsh
speech
1araeLllke usaae

Table 13. Complex object. Participant K.

The pattern may be a part oI the U-shaped learning, when an incorrect usage can happen
with the Ieature that used to be used correctly.

5) II an RB Ieature constitutes a very complex system a learner can Iail to uptake the
recast at all, as it was the case with Conditional III and modal verbs to express past possibilities
in the Paricipant K. texts. However, the more complex the system is, the longer it takes to
incorporate it in a learner`s interlanguage.
When there is no explicit grammar inIormation provided, a learner leans onto his L1
knowledge to construct L2 grammar systems. Such grammatical Ieatures as Conditional III or
modal verbs to express past possibilities cannot be paralled to the like Ieatures in the
participants` L1 Russian, because in Russian they are contextual Ieatures, not grammatical.
Since there are no special grammatical markers Ior these Ieatures in Russian, it is extremely hard

1exL Lxample
n7 l donL wanL mv chlldren be woken
up
n8 l [usL wanL mv blabla famllv be
happv
1esL
I want you to show us a place ln
Cvprus where vou can leave vour
auesLs on Lhe whole dav and [usL
relax

l wanL Lhe weaLher wlll be llke
before
1araeLllke usaae

Ior Russian-speaking EFL learners to understand their meaning and diIIer them Irom other
modals and conditionals in the English language.

6) There is no deIinite pattern with IB errors in the analysed texts. Some IB errors were
successIully corrected aIter they had been recasted, others not. The examples are given below in
the tables NN14-15.

Table 14. Examples of corrected IB features. Participant K.

1exL Lxample
n11 And pollcemen are looklna for me
It doesnt care
1esL I w||| not care lf l come back home
wlLhouL sunburn
1araeLllke usaae
8asellne Cur arannv d|v|ded us manv vears
aao
n4 As a [oke we separated our caLs a
alrl ls mv parenLs own and a bov
ls mlne (meanlna dlvlded)
1esL l have d|v|ded mv monev
approxlmaLelv on 3 equal parLs
1araeLllke usaae
8asellne 8efore l had sa|d her !usL no
Lears!
n3 Cnce one woman sa|d e l was
ln Cvprus lasL vear
n8 Pe wlll noL say e whaL should l
do Lomorrow
1esL
When l come back home l wlll say
to y fa||y
1araeLllke usaae
1exL Lxample
n10
l sLood ahead hlm non LaraeLllke usaae
n11
Ch l see a skeleLon ln fronL of me 1araeLllke usaae
n3
?ou can marrv on Cvprus onlv 3
davs afLer vour arrlval Lhere
1esL l wanL vou Lo show us a place ln
Cvprus where vou can leave vour
auesLs on Lhe whole dav and [usL
relax
1araeLllke usaae
n3
l have flown on Lhe LarLh
(dlrecLlon)
n7
l ao down Lhe sLalrs Lo Lhe klLchen 1araeLllke usaae
n10 AlmosL everv mornlna he aoes Lo
hls work
l wenL Lo Lhe markeL
1araeLllke usaae
n4
Pe Lhrew Llle ouL from Lhe
wlndow
n7
l look aL mv bla happv famllv
Lhrouah Lhe wlndow
1araeLllke usaae
n3
8uL aL LhaL Llme he wanLed Lhe
same Lo come back aL home
1esL
When l come back home 1araeLllke usaae
1exL Lxample
8asellne Cn Lhe one slde we llked her for
dolna noLhlna 8uL on Lhe oLher
slde l undersLood LhaL she dldnL
alve us necessarv knowledae
Cn Lhe one hand lLs mv broLher
Cn Lhe oLher hand lLs hls wlfe
(meanlna flrsLlv And secondlv)
1esL 1he mosL exclLlna Lhlna ln mv Lrlp
was Lhe whole Lrlp 8uL on Lhe
oLher hand l had Lhe mosL borlna
Lhlna 3 hours bv plane

1araeLllke usaae

Some IB errors were transIerred to the next texts despite the recasts.
Table 15. Examples of IB errors that were used incorrectly in subsequent writings after they had been
recasted. Participant K.

1exL Lxample
8asellne So plLv! non LaraeLllke usaae
n9 lLs verv plLv non LaraeLllke usaae
n9 8uL mavbe farther Lhls maLh Lasks
l would undersLand
n11 l donL know whaL wlll be farther non LaraeLllke usaae

Some words that had been used incorrectly (wro3g wor/ type oI error) and then recasted,
were not used by the participants in the next texts, so it is impossible to say whether they were
aIIected by the recast or not.
In general, the study clearly showed that both item-based and rule-based errors can be
equally responsive to corrective written Ieedback in the Iorm oI recast. The degree oI treatability
oI RB errors depends on the complexity oI the system the treated Ieature makes. The simpler the
Ieature system is, the more treatable the errors oI this Ieature are. Thus, Ior example,
demonstrative pronouns and copular be were successIully treated by recasts; articles and tenses,
though suIIicient improvement was observed through the study, were occasionally used
incorrectly up to the end oI the study; conditional III and modal verbs expressing possibility in
the past were not used correctly aIter they had been recasted.
Item-based errors, since they are not part oI a system, are more an issue oI memorising.
Once they are noticed by a learner, they are the subiect oI memorising. Sometimes they are
successIully learned aIter the Iirst recast, in other cases several recasts are needed Ior a Ieature
to be memorised. Thus it is the Irequency oI use that can help improve the accuracy oI IB
Ieatures more than anything else.

4.3. How Do Learners Perceive Focused Recasts? What Is Their Attitude to Such
Kind of Feedback?

Both participants were required to answer a set oI questions to Iind out in what way they
had used recasts and what their opinion about such kind oI Ieedback was.
Participant J. revised her writings aIter receiving the text with recasts.
Participant K. looked through a text and recasts trying to memorise the recasted variants,
then she did it again next day.
In the study both participants showed signiIicant improvement in accuracy, so it was not
oI a great importance how exactly the participants had dealt with recasts. More probably, an
important Iact was that each participant was leIt Iree to Iind her own most suitable Ior her way to
beneIit Irom the Ieedback.
As to their attitude to the recasts, both participants expressed their wish to get explicit
explanations alongside with the recasts, so that they could understand why a certain Ieature was
used in a certain way. Such wish was particularly expressed about the rules oI the language, i.e
rule-based Ieatures.

CONCLUSION

The study attempted to investigate the eIIect oI written Iocused recasts on accuracy oI
writing. The obtained results indicate the Iollowing:
Contrary to Truscott (1996) opinion, written corrective Ieedback in the Iorm oI recasts
does help to improve accuracy. The results oI the study revealed dramatic improvement in
accuracy oI both participants, which contradicts Truscott argument about useless or even harmIul
eIIect oI corrective Ieedback.
Improvement in accuracy can happen aIter the Iirst session already due to the simple
increase oI attention on the part oI a learner, which eliminates mistakes made by neglect or
carelessness. Sustained treatment can improve accuracy Iurther, eliminating simple errors that
are easy to systemise and helping to build knowledge about more complex systems.
Both item-based and rule-based Ieatures are treatable by recasts. This Iinding questions
Ferris`s classiIications oI errors as treatable and untreatable based on whether the erroneous
Ieatures are rule-based or item-based. The qualitative analysis clearly showed the improvement
oI accuracy Ior both Ieatures. The way recasts aIIect them is diIIerent though. With RB Ieatures
a learner Iollows a path oI hypothesis testing, where aIter each recast a new boundary (rule) is
added to the system, thus allowing a gradual construction oI the system oI rules Ior each RB
Ieatures. Only too complex Ieatures such as Conditional III and modal verbs to express
possibility in the past, Iailed to be eIIected by recasts in the study. Still we cannot conclude that
these Ieatures are untreatable. The diIIiculty oI these Ieatures lie in their unique nature that
cannot be compared to similar Ieatures in the participants` L1. It is not grammatical markers oI
these Ieatures that present diIIiculties Ior learners, it is the meaning that they convey (i.e. not
HOW to use the Ieatures, but WHEN to use them). We suppose that more writing practice with
corrective Ieedback in the Iorm oI recasts can help solve this predicament, on condition that the
tasks are careIully designed to elicit the usage oI the complex systems in question. A Iurther
longer study is needed to analyse how recasts could inIluence the acquisition oI such complex
Ieature systems, especially whether recasts are able to help learners single out the meaning that
such systems convey.
IB Ieatures depend more on a learner's memorising ability, than systemising. Since IB
Ieatures are usually simple in construction, it is their uniqueness that make them diIIicult.
Recasts provide a necessary input (reactive by the nature oI recasts), oI what unique IB Ieature
should be used in a deIinite context. As the study showed some IB errors could be corrected aIter
the Iirst recast, some took longer, some Iailed to be corrected. But the longer the treatment is, (i.e
more reminders as to what is the correct IB item to use in the context are given to a learner) the
higher possibility oI memorising is. This assumption can be a Ioundation Ior a Iuture research on
IB Ieatures and corrective Ieedback.
In general attributing to IB and RB errors a diIIerent degree oI treatability seems wrong.
Both types oI errors are treatable, but in diIIerent way.
Although recasts provide learners with target-like written language, learners can still lack
explicit explanations, especially with the RB Ieatures, as the study showed. Our assumption is
that explicit explanations could be oI great value with Ieature systems construction, thus helping
learners to understand a target-like usage oI rule-based Ieatures, while with item-based Ieatures it
is Irequency oI use that is oI more importance. Here lies one more suggestion Ior a Iuture
research to compare written recasts vs. written recasts plus explicit explanation, especially in
the process oI the acquisition oI complex Ieatures.

The study has practical implications Ior teaching and learning processes.
Firstly, corrective Ieedback is a useIul tool in EFL learning/teaching process.
Secondly, corrective written Ieedback in the Iorm oI Iocused recasts can be used by
teachers who give writing instruction not only in internet-mediated learning/teaching
environment, but also in a traditional class setting providing students submit their written works
via internet. The process oI recasting using modern soItware is quick, which is an important
Iactor Ior teachers oIten overloaded with a time-consuming iob oI correcting texts manually, and
the recasts in the margins are easy Ior students to work with.
Also the recasts that a teacher leaves in the margin make a history oI a learner`s success
or Iailure with a certain Ieature. It is easy to detect by looking through the marginal comments oI
several subsequent texts oI a learner. It is also oI help when considering individual conIerencing
with a learner to discuss his achievements.
Thirdly, contrary to Ferris`s (2002) conclusion both IB and RB Ieatures are treatable and
consequently should be treated by a teacher. There is no point in classiIying errors into treatable
and untreatable and leaving the latest without a teacher`s attention.

References
Abrahams, D. (1997). I3tro/uctorv 8ocal re8earch metho/8. Retrieved Irom http://www.SoIia
research ethics.net/tutorial/Abrahams/SRM main.HTML
Alberto, P. A., Troutman, A. C. (2006). Applied behaviour analysis Ior teachers (7th ed.). Upper
Saddle River, NJ: Person Education, Inc. Retrieved May 4, 2011, Irom Questia database:
htpp://www.questia.com
Ayoun, D. (2001). The Role oI Negative and Positive Feedback in the Second Language
Acquisition oI the Passe Compose and ImparIait. The Mo/er3 La3guage Jour3al, 85.
226243. doi: 10.1111/0026-7902.00106
Ayoun, D. (2004). The eIIectiveness oI written recasts in the Second Language Acquisition oI
aspectual distinctions in French: A Follow-up Study. The Mo/er3 La3guage Jour3al. 88
(1), 31-55.
Bitchener, J., Knoch, U. (2009). The relative eIIectiveness oI diIIerent types oI direct written
corrective Ieedback. Sv8tem. 37, 322-329.
Bitchener, J., Knoch, U. (2010). Raising the linguistic accuracy level oI advanced L2 writers
with written corrective Ieedback. Jour3al of Seco3/ La3guage Wrt3g Jolume.19 (4),
207-217.
Bitchener,J., Young, S., Cameron, D. (2005). The eIIect oI diIIerent types oI corrective Ieedback
on ESL student writing. Jour3al of Seco3/ La3guage Wrt3g.14. 191-205.
Brender, A. (1998). ConIerencing. An Interactive Way to Teach Writing. The La3guageTeacher
3l3e. Retrieved March, 5, Irom
http://www.ialtpublications.org/tlt/Iiles/98/iul/brender.html
Chandler, J. (2003). The eIIicacy oI various kinds oI error Ieedback Ior improvement in the
accuracy and Iluency oI L2 student writing. Jour3al of Seco3/ La3guage Wrt3g. 12 (3),
267-296.
Choi, M.Y. (2000). EIIects oI recasts on irregular past tense verb morphology in Web-chat.
Unpublished master's thesis, University oI Hawai'i, Honolulu. Available Irom Wiley
online library http://onlinelibrary.wiley.com/doi/10.1111/i.1540-4781.2010.01143.x/pdI
Cohen, A. D. (1989). ReIormulation: A technique Ior providing advanced Ieedback in writing.
Gu/el3e8. 11 (2), 1-9.
Dasse-Askildson, V. (2008). How learners' aIIective variables impact their perception oI recasts
in the acquisition oI grammatical gender in L2 French. Arzo3a Work3g Pa5er8 o3 SLA
a3/ Teach3g. 15. 1-35. Retrieved Irom
http://w3.coh.arizona.edu/awp/AWP15/AWP155BDasse5D.pdI
Doughty, C., Varela, E. (1998). Communicative Iocus on Iorm. In C. Doughty & J. Williams
(Eds.), Focus on Iorm in classroom second language acquisition (pp. 114-138).
Cambridge: Cambridge University Press. Retrieved May 13, 2011, Irom Questia
database: htpp://www.questia.com
Doughty, C., Williams, J. (1998). Pedagogical choices in Iocus on Iorm. In C. Doughty, J.
Williams (Eds.), Focu8 o3 form 3 cla88room 8eco3/ la3guage acqu8to3 (pp. 197-261).
New York: Cambridge University Press. Retrieved May 4, 2011, Irom Questia database:
Ellis, N.C. (2005). At the interIace: Dynamic interactions oI explicit and implicit language
knowledge. Stu/e8 3 Seco3/ La3guage Acqu8to3. 27. 305352. Retrieved May 23,
2011, Irom Questia database: htpp://www.questia.com
Ellis, R., Loewen, S., Erlam, R. (2006). Implicit and explict corrective Ieedback and the
acquisition oI L2 grammar. Stu/e8 3 Seco3/ La3guage Acqu8to3. 28(2), 339-368.
Ellis, R., Sheen, Y., Murakami, M., Takashima, H. (2008). The eIIects oI Iocused and unIocused
written corrective Ieedback in an English as a Ioreign language context. Sv8tem. 36, 353-
371.
Farrar, M. J. (1990). Discourse and the acquisition oI grammatical morphemes. Jour3al of Chl/
La3guage 77, 607-624.
Fathman, A.K., Whalley, E. (1990). Teacher response to student writing: Focus on Iorm versus
content. In B. Kroll (Ed.), Second Language Writing (pp. 178-190). Cambridge, UK:
Cambridge University Press. Retrieved May 8, 2011, Irom Questia database:
Fazio, L.L. (2001). The eIIect oI corrections and commentaries on the iournal writing accuracy
oI minority and maiority language students. Jour3al of Seco3/ La3guage Wrt3g. 10 (4),
235-249.
Ferris, D. (1999). The case Ior grammar correction in L2 writing classes: A response to Truscott
(1996). Jour3al of Seco3/ La3guage Wrt3g. 8, 1-10.
Ferris, D. (2002). Treatme3t of error 3 8eco3/ la3guage wrt3g cla88e8. Ann Arbor: University
oI Michigan Press.
Ferris, D. R., Hedgcock, J. (1998). Teach3g ESL com5o8to3. Pur5o8e. Proce88. a3/ Practce.
Mahwah, NJ: Lawrence Erlbaum.
Ferris, D.R. (2003). Response to student writing: Implications Ior second language students.
Mahwah, NJ: Lawrence Erlbaum Associates. Retrieved April 19, 2011, Irom Questia
database: htpp://www.questia.com
Fratzen, D. (1995). The eIIect oI grammar supplementation on written accuracy in an
intermediate Spanish content course. Mo/er3 La3guage Jour3al.79 (3), 329-344.
Grami, M. (2005).The eIIect oI teachers written Ieedback on ESL students perception: A study in
a Saudi ESL university-level context. A33ual Re;ew of E/ucato3. Commu3cato3 a3/
La3guage Sce3ce8.2. Retrieved April, 4, Irom
http://research.ncl.ac.uk/ARECLS/vol2documents/Grami/grami.htm
Grey, R. (2004). Grammar Correcto3 3 ESL/EFL Wrt3g Cla88e8 Mav Not Be Effect;e.
Retrieved February, 18, Irom http://itesli.org/Techniques/Gray-WritingCorrection.html
Han, Z. (2002). A Study oI the impact oI recasts on Tense consistency in L2 output. TESL
Quarterlv. 36 (4), 543-572. Retrieved Irom http://www.istor.org/stable/3588240
Ishida, M. (2004), EIIects oI Recasts on the Acquisition oI the Aspectual Form -te -(ru) by
Learners oI Japanese as a Foreign Language. La3guage Lear33g. 54, 311394.
doi: 10.1111/i.1467-9922.2004.00257.x
Iwashita, N. (2003). Negative Ieedback and positive evidence in task-based interaction:
DiIIerential eIIects on L2 development. Stu/e8 3 Seco3/ La3guage Acqu8to3. 25(1),
1-36.
Krashen, S. (1982). Principles and Practice in Second Language Learning and Acquisition.
OxIord: Pergamon.
Krashen, S. (1984). Writing: Research, Theory and Applications. Laredo: Beverly Hills.
Kumaravadivelu, B. (2001). Toward a postmethod pedagogy. TESL Quarterlv. 35, 537-560.
Lalande, J.F. (1982). Reducing composition errors: An experiment. Mo/er3 La3guage Jour3al.
66 (1), 140-149.
Lee, I. (2004). Error correction in L2 secondary writing classrooms: The case oI Hong-Kong.
Jour3al of Seco3/ La3guage Wrt3g. 13 (4), 285-312.
Leeman, J. (2003). Recasts and second language development. Stu/e8 3 Seco3/ La3guage
Acqu8to3. 25(1), 37-63.
Levenston, E. A. (1978). Error analysis oI Iree composition: The theory and the practice. I3/a3
Jour3al of A55le/ L3gu8tc8. 4 (1), 1-11.
Long, M. (2006). SLA: Breaking the siege. In M. Long. Problems in SLA. Mahwah, BJ:
Lawrence Erlbaum.
Long, M. H. (1996). The role oI the linguistic environment in second language acquisition. In W.
C. Ritchie & T. K Bahtia (Eds.), Handbook oI second language acquisition (pp. 413-468).
New York: Academic Press.
Long, M. H. (2007). Problem8 3 SLA. Mahwah, NJ: Lawrence Erlbaum Associates.
Long, M., Inagaki, S., Ortega, L. (1998) The role oI implicit negative Ieedback in SLA: Models
and recasts in Japanese and Spanish. The Mo/er3 La3guage Jour3al 82, 357 371.
Lyster, R. (1998a). Negotiation oI Form, Recasts, and Explicit Correction in Relation to Error
Types and Learner Repair in Immersion Classrooms. La3guage Lear33g. 48. 183218.
doi: 10.1111/1467-9922.00039
Lyster, R. (1998b). Recasts, repetitions, and ambiguity in L2 classroom discourse. Stu/e8 3
Seco3/ La3guage Acqu8to3. 20. 51-81.
Lyster, R., Ranta, L. (1997). Corrective Ieedback and learner uptake. Stu/e8 3 Seco3/
La3guage Acqu8to3. 19. 37-66. Retrieved Irom http://iournal.tc-
library.org/index.php/tesol/article/viewFile/160/158
MacArthur, C. A., Graham, S., Fitzgerald, J. (2006). Ha3/book of Wrt3g Re8earch, GuilIord
Press.
MaItoon, P., Ahmadi, M., DaIIariIard, P. (2010). The eIIect oI recast vs selI-correction on
writing accuracy: the role oI awareness. Broa/ Re8earch 3 Artfcal I3tellge3ce a3/
Neur8ce3ce. 2 (1), 17-28
Mohammadi, M., Javadi, T. (2009). Application oI recasts in TEFL: 3 Effect;e3e88 of Wrtte3
Reca8t ;8. Ex5lct Negat;e Fee/back. Retrieved March, 23, Irom
http://riau.academia.edu/MoitabaMohammadi/Papers/238695/ApplicationoIRecastsin
TEFLOnEIIectivenessoIWrittenRecastvs.ExplicitNegativeFeedback
Navidinia, H., Eghtesadi, A. R. (2009). SLA research in postmethod era: Neglect,
misunderstandings and alternatives. E/ucato3al Re8earch a3/ Re;ew. 4 (2), 57-62.
Nicholas, H., Lightbown, P. M., Spada, N. (2001). Recasts as Feedback to Language Learners.
La3guage Lear33g. 51. 719758. doi: 10.1111/0023-8333.00172
Ortega, L. (2005). For what and whom is our research? The ethical as transIormative lens in
instructed SLA. Jour3al of Mo/er3 La3guage8. 89, 427-442.
Ortega, L., Long, M. H. (1997). The eIIects oI models and recasts on the acquisition oI obiect
topicalization and adverb placement in L2 Spanish. S5a38h A55le/ L3gu8tc8. I. 65-86.
Pellettieri, J. (2000). Negotiation in cyberspace: The role oI chatting in the development oI
grammatical competence in the virtual Ioreign language classroom. In M. Warschauer &
R. Kern, (Eds.), Networkbased language teaching: Concepts and practice (pp. 59-86).
Cambridge, England: Cambridge University Press.
Robb, T., Ross, S., Shortreed, I. (1986). Salience oI Ieedback on error and its eIIect on EFL
writing quality. TESL Quarterlv. 20, 8393.
Rouhi, A., Samiei, M. (2010). The eIIects oI Iocused and unIocused indirect Ieedback on
accuracy in EFL Writing. The Socal Sce3ce8. 5 (6), 481-485.
Sachs, R., Polio, C. (2007). Learners' uses oI two types oI written Ieedback on L2 writing
revision task. Stu/e8 3 Seco3/ La3guage Acqu8to3. 29, 67-100.
doi:10.1017/S0272263107070039
Santos, M., Lopez Serrano, S. & Manchon, R. (2010). The diIIerential eIIects oI two types oI
direct written corrective Ieedback on noticing and uptake: ReIormulation vs. error
correction. I3ter3ato3al Jour3al of E3gl8h Stu/e8, 10, 131-154.
Sauro, S. (2007). A comparative study oI recasts and metalinguistic Ieedback through computer
mediated communication on the development oI L2 knowledge and production accuracy.
Dissertation available Irom ProQuest. Paper AAI3271810.
http://repository.upenn.edu/dissertations/AAI3271810
Saxton, M., (1997). The contrast theory oI negative input. Jour3al of Chl/ La3guage 24,139-
161.
Saxton, M., (2005). 'Recast' in a new light: Insights Ior practice Irom typical language studies.
Chl/ La3guage Teach3g a3/ Thera5v. 21, 23-38.
Schmidt, R. (1994). Deconstructing consciousness in search oI useIul deIinitions Ior applied
linguistics. In J. Hulstiin & R. Schmidt (Eds.), Co38cou83e88 3 8eco3/ la3guage
lear33g. AILA Re;ew. 11. 1126.
Sheen, Y. (2007). The eIIect oI Iocused written corrective Ieedback and language aptitude on
ESL learners acquisition oI articles. TESL Quarterlv. 41, 255-283.
Tatawy, M. E. (2006). Corrective Ieedback in second language acquisition. Retrieved Irom
Working Papers in TESOL & Applied Linguistics http://iournals.tc-
library.org/index.php/tesol/article/viewArticle/160.
Truscott, J. (1996). The case against grammar correction in L2 writing classes. La3guage
Lear33g. 46, 327-369.
Truscott, J., Hsu, Y. A. (2008). Error correction, revision and learning. Jour3al of Seco3/
La3guage Wrt3g. 17, 292-305.
Wolery, M., Harris, S.R. Interpreting results oI single-subiect research designs. Jour3al of
Phv8cal Thera5v. 62 (4), 445-453. Retrieved April, 30, Irom:
ptiournal.apta.org/content/62/4/445.Iull.pdI
Zacharias, N. T. (2007). Teacher and Student Attitudes toward Teacher Feedback. Jour3al of
La3guage Teach3g a3/ Re8earch. 38, 38-52, doi:10.1177/0033688206076157.

ALnulCLS
1oLal scores of errors arLlclpanL !

0
1
2
3
4
3
6
7
8
9
hase A
hase 8
Llnear (hase A)
Llnear (hase 8)
1oLal scores of errors arLlclpanL k

0
1
2
3
4
3
6
7
8
9
10
hase A
hase 8
Llnear (hase A)
Llnear (hase 8)
lLembased and rulebased error scores arLlclpanL !

0
1
2
3
4
3
6
7
8
9
l8A
l88
88A
888
Llnear (l8A)
Llnear (l88)
Llnear (88A)
Llnear (888)
lLembased and rulebased error scores arLlclpanL k

0
1
2
3
4
3
6
7
l8A
l88
88A
888
Llnear (l8A)
Llnear (l88)
Llnear (88A)
Llnear (888)
uaLa dlsLrlbuLlon arLlclpanL k

0
1
2
3
4
3
6
7
8
9
10
hase A
hase 8
6 uaLa dlsLrlbuLlon arLlclpanL !

0
1
2
3
4
3
6
7
8
9
hase A
hase 8

Impact of Recasts On The Accuracy in Efl Learners' Writing (Ma Thesis)

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Impact of Recasts On The Accuracy in Efl Learners' Writing (Ma Thesis)

Uploaded by

Copyright:

Available Formats

GIRNE AMERICAN UNIVERSITY

You might also like