oral production and error correction amongst arab learners of english haifaa faqeih, phd student...
TRANSCRIPT
Oral Production and Error Correction Amongst Arab
Learners of English
Haifaa Faqeih, PhD Student &
Dr Emma Marsden, Lecturer in Second Language Education
University of York
Contact: [email protected] & [email protected]
Summary of study
• Experiment comparing effectiveness of different corrective feedback techniques:
Recasts v metalinguistic information v oral-tasks-only group
• Arabic L1
• Learning English modals: must, can, will
• Pre Intermediate learners
• Phase 1: ESL in UK language institute
• Phase 2: EFL in Saudi university
Previous studies: recasts effective• Definition: “reformulation of a learner’s utterance by altering one or
more incorrect forms therein while retaining semantic content” (Révész 2009)
• Supporting effectiveness:– Leeman 2003; Doughty 2001; Doughty & Varela 1998; Leeman 2003;
Long 1996; Ohta 2000; Oliver 1995; Egi 2007; McDonough & Mackey 2006; Sheen 2006; Ellis et al 2006.
• But effectiveness depends on range of factors: – e.g.: proficiency, length of recast, number of changes, target form, prior
experience of participants to error correction, attitudes to error correction, working memory
• Theoretical support & accounts:– Interaction hypothesis (negotiation of meaning, timely intervention)– Noticing the gap– Priming– Implicit learning from positive evidence – Induction of explicit knowledge e.g. helps in explicit hypothesis
formulation and testing?
Gaps we aim to address 1. Tasks done by control group
• Ellis et al 2006; Sheen 2006 did not have a task+test control group
2. Different contexts – UK ESL (oral interaction + error correction normal)
versus – Saudi EFL (little oral interaction + proactive
metalinguistic info.) • Sheen 2004 investigated contexts, but learning not
measured• Most CF studies done in contexts where oral interaction
fairly normal.
3. Metalinguistic information only• Ellis repeated error + information; Sheen recast +
information– though prosody not controlled
Gaps we aim to address
• New linguistic feature: English modals • *cans go, *can goes, *can going, *will can go
• NNS teacher with group • ecological validity
• Measures not always used in previous studies: – Uptake during interaction + achievement + attitudes– Delayed post tests (7 weeks)– Time & communicative pressures
• Oral production test• Timed grammaticality judgement test
The Study
• Phase 1: Saudis learning English in York, ESL– Recast (10); Metalinguistic (8); Control (6)– Randomly assigned (not matched)
• Phase 2: Saudis learning English in Saudi, EFL – Recast (20); Metalinguistic (20); Control (20)
Intervention
• 3 hours • over 4 sessions, over 2 weeks
• Oral interaction tasks• Eliciting English modals: can, must, will
• Task design: – outcome clear– meaning focussed– Pair & group & teacher/student
Eliciting modals in intervention
• Session 1 Can
• Session 2 Will
• Session 3 Must
• Session 4 Mix
• Declarative, negative & interrogative
Example of intervention, eliciting “can” and “can’t”
• Have a conversation with your friend about whether you can do the activities shown on your cards & find out what s/he can do.
• You then have to talk to the group about what your friend can and can’t do.
Flip forwards Basketball Tractor Tennis
Error correction during oral production
• Recast groupStudent: He cans go on holiday
Teacher: He can go on holiday
• Metalinguistic groupStudent: He cans go on holiday
Teacher: You must not change the modal – modals do not agree with the subject
• Control groupStudent: He cans go on holiday
Teacher: and your next picture?
Outcome measures• Pre, post and delayed post test (7 weeks)• ALL back to same instructor between post & delayed post• Timed grammaticality judgement test
• Timings from natives + 20%• Participants corrected incorrect items• 3 warm-up, 9 correct, 9 incorrect, 9 correct fillers, 9 incorrect fillers• Separate analyses for grammatical & ungrammatical items (Ellis, 2005)
• Written Gap fill• 11 items, 6 fillers
• Oral production• 7 tasks • Pictures eliciting advice, explanations, suggestions• e.g. Use the picture prompts to tell John what the rules are if he wants
to go to the Mosque [for must]• Not same as format as intervention • Suppliance in Obligatory Context• Interlanguage scoring
Results: Gap fill test
Gap fill results
Table: P values in paired Wilcoxon tests
pre – post pre – delayed
post – delayed
Metalinguistic <.05 . <.05 .798
Recast <.05 <.05 .324
Control ANOVA not significant so no paired tests
Results: Grammaticality Judgement test
• Overall (grammatical + ungrammatical target items): No differences between conditions
• though paired tests do look significant…
• Correct items: no differences between conditions
• So, incorrect items…
Results: Grammaticality Judgement test
• Incorrect items
Incorrect items in GJTpre – post pre –
delayedpost – delayed
Metalinguistic
(8)
<.05 <.05 .236
Recast
(10)
.214 <.05 .091
Control
(6)
Friedman ANOVA not significant so no paired tests
Results: Oral production
Pre-post Mean raw oblig contexts
pre post
Metalinguistic <.05 31 42
Recast .086 33 38
Control .686 27 35
Effect sizes (Cohen’s d)pre - post pre – delay post
Gapfill M 1.72 1.35
R 1.35 0.95
C 0.21 0.27
GJT incorrect items M 0.70 1.11
R 0.44 1.08
C 0.26 0.18
Oral production M 1.01
R 0.51
C -0.39
Overall Patterns: Explicit knowledge???
• Yes, both R & M made gains on gapfill &
incorrect items on GJT Ellis, 2005
• But, after GJT and gapfill all participants were
asked whether they knew what the test was
testing
– Nobody thought modals had been tested!
Overall patterns: Implicit knowledge??
• No, neither made gains on correct items on GJT • But yes, M, and R to lesser extent, made gains
in oral production– M accessed explicit knowledge during oral
production? It had become automatised?– R needed more consolidation to induce knowledge
sufficiently to access it during oral production? Delayed tests will tell…
• Ellis et al found gains only at delayed post test on implicit measures
Implications to date• Demonstrates importance of …
– Need for control group + task group: gains, but no sig. self-induced attention to form or incidental or implicit learning
– Need for different measures: explicit v implicit knowledge
– Need for delayed post tests: recasts lead to raised awareness for processing future exemplars?
• Tentatively: – Noticing at level of understanding more reliably
beneficial, in short term at least– Recast group may have induced explicit knowledge
Next steps• Analyse uptake & delayed post test oral production &
attitudinal data• Phase 2 in Saudi Arabia• All conditions innovative in that context:
– M group: Normally, metalinguistic info provided proactively– R group: Normally, very few oral recasts– Control task-only group: Normally, little oral task work
• Sheen 2004 found uptake greater in FonF type contexts• We could find
– learners interpret recasts as CF and benefit from metalinguistic because FonF normal
– All, including control, groups benefit equally because the effect of the oral interaction is so marked
– OR no group benefits because oral interaction, recasts and reactive metalinguistic info are ‘not normal’
Thank you for listening.
Study funded by a Saudi post-graduate scholarship
References
Ellis, R. (2005). Measuring implicit and explicit knowledge of a second language: A Psychometric Study, Studies in Second Language Acquisition, 27, 141–172
Ellis, R., Loewen, Sh., and Erlam, R. (2006). Implicit and explicit corrective feedback and the acquisition of L2 grammar. Studies in Second Language Acquisition, 28, 339- 368.
Sheen, Y. (2004). Corrective feedback and learner uptake in communicative classroom instructional settings. Language Teaching Research, 8, 263.
Sheen, Y.(2006) Corrective Feedback, Individual Differences, and the acquisition of English Articles by Second Language Learners. Unpublished doctoral thesis. University of Nottingham.