Presentation on theme: "Re-examining Individual Differences in Working Memory , Learner Awareness of L2 Forms and L2 Development through Recasts on Task-basked Interaction Good."— Presentation transcript:
1Re-examining Individual Differences in Working Memory , Learner Awareness of L2 Forms and L2 Development through Recasts on Task-basked InteractionGood afternoon, everyone, I am really feeling so honored to have such an opportunity to present my own research here today. Actually it is a small part of data from my Phd research I have just finished analysing recently. The title is…..DAI Binbin AmyThe Chinese University of Hong KongThe 3rd international conference on TBLTUniversity of Lancaster, Sep
2What have been done? However, Besides, Therefore, Empirical studies linking working memory (WM) capacity & noticingWeak relationship(p=0.051)(1) Mackey et al. (2002)Noticing: stimulated recall & exit questionnaireWorking memory: English non-word recall, L1&L2 listening span testsNo relationship(2) Trofimovich et al. (2007)Noticing: visually cued discriminationWorking memory: letter-number sequencing testBased on schmidt’s famous noticing hypothesis, various empirical studies have been conducted and some evidence has been got to prove the important effect of noticing on learners’ second language learning. however, only a limited number of studies have investigated the relationship between WM capacity and noticing.The first one was conducted by Mackey and her colleagues, which has been considered as an exploratory study in this field. In this study, “19 participants’ noticing data were from stimulated recall after immediate posttest, 11 participants’ noticing data were only from questionnaire with only 5 questions ………” , unequal detection methods of noticing may lead to some problems to measure noticing. We may have the same question, “How can we give the persons noticing scores only through 5 simple open questions” and also this questionnaire was answered two weeks after the on-line treatment sessions, we cannot deny that it’s really a serious test of participants’ memory, how could they remember exactly what happened two weeks ago?.In the second study conducted by trofimovich and his colleagues, “visually cued discrimination” was used to detect noticing in this study. After Participants’ performance, there was a recast appeared on the screen and participants only need to answer yes or no each time when asking “did you notice any difference?” this way also can not be regarded as an effective one to detect noticing because this kind of yes or no question is just like a reminder to let participants deliberately notice the difference but not their real cognitive process.Fortkamp and Bergsleithner measured noticing by means of the accuracy of the two sentences the participants had produced in their oral protocols, one sentence correction got 50 points, two 100 points, others are zero point, which is not an effective way to know participants cognitive process.As to the testing of wm capacity, three studies used three different ways. But from the construct-validity point of view in testing, Mackey et al.’s study may be more comprehensive.Now comes to the results of the three studies, weak relationship with p value 0.51, no relationship has been found in the next two studies.It is not difficult to see that we can not simply draw a conclusion on whether the relationship exists or not. The investigation into the link between WM capacity and noticing is still needed with a reasonable and effective refinement of research methods to detect noticing and measure participants’ working memory capacity.However,Very few studies have made an attemptto link the understanding level with WM capacity !!Besides,Roberts (1995) & Mackey et al. (2007):L2 proficiency noticing & understanding(1) To make an attempt to explore the relationshipbetween WM capacity and understanding(2) L2 proficiency levels will be regarded as anindependent variableTherefore,facilitative for language learning
3The effect of WM capacity on interaction- driven L2 development Ando et al. (1992)Grammar Approach (Higher WM, More development)VS.Communicative Approach (Lower WM, More development)Mackey et al. (2002) Lower WM, More development(immediate posttest)Sagarra (2007) Higher WM, More development(delayed posttest)Trofimovich et al. (2007)No relationship in immediate posttestTalking about the effect of working memory capacity on learners’ second language development, Ando et al.’s study has shown a very interesting result, that is in explicit learning (GA) learners with high wm obtained more improvement, however, in implicit learning conditions, the result was reversed. To be more focused on recast-driven development, both mackey and sagarra’s studies have found two different stories in the immediate posttest and delayed posstest. However, Trofimovich et al.’s study didn’t find the relationship between WM and L2 development in the immediate posttest. Therefore, research results on the effect of wm capacity on recast-driven L2 development are only suggestive but not confirmative and still needs further exploration.
4Research QuestionsIs there a relationship between WM capacity and learner awareness of recasts in interactional feedback at respective two levels (noticing and understanding)?Is there a relationship between learners’ L2 proficiency levels and their awareness of recasts in interactional feedback at respective two levels (noticing and understanding)?What are the effects of learners’ L2 proficiency and WM capacity on their L2 improvement?So here comes the research questions of the study, I’ve tried to find whether there is a relationship between….., whether there is a relationship between….. And what the effects of ….are.
5How did I design my research? Learner participants: (non-English major/ undergraduates/ mainland China)Age(Number of participants)18 (4)19 (6)20 (4)21 (6)22 (3)23 (1)GenderMaleFemale816MajorScienceArts186GradeOneTwoThree12210Mother tongueAll MandarinOther languagesAll English( including two have learnt some Japanese)Age of starting learning English7-9 (3)10 (5)11(3)12 (3)13 (8)14 (2)Learning experience abroadAll NoneThere are 24 participants in this study, actually it’s one of experimental groups in my research project. As to their detailed background information, you can refer to this page in the handouts, because of very limited time, I will not explain for you here. But one point I have to make is that all of them had very similar English learning background.The background information of learner participants (24)
6Learner participants Proficiency levels Voluntary participationProficiency levelsC-test (mean=46.13, SD=6.05) (high≧47 vs. low ≦46)WM capacity levelsComposite score= z (Non-word) + z (L2 listening span)(high>0 vs. low<0)51 freshmen in XJTU: two classes, one English teacherFour extracts of English articles (Dörnyei and Katona, 1992)Cronbach's Alpha=.770Concurrent validity(1) C-test (June 17) & Term Proficiency Test (May 25)( r =.583, p<0.01)(2) C-test (June 17) & CET-4 (June 21)( r =.633, p<0.01)24 participants selected among a large number of students for:Assumptions for two-way ANOVA:Shapiro-Wilk normality tests & Homogeneity of variances testsNormal distribution of both scores of c-test and WM testLevene’s Test of Equality of Error Variances (ns)No significant difference among each group at the beginningParticipants were recruited on the basis of the following three aspects voluntary participation, English proficiency levels and WM capacity levels. A validated C-test was conducted to measure their proficiency levels, Alpha has reached .77 and concurrent validity was also relatively high. Two different tests were used to test their working memory capacity: non-word recall test, 42 English non-words were selected from prof. skehan’s research project onn language aptitude sponsored by CUHK. And english listening span test in which a set consisted of two, three, four, or five sentences, three sets per each sentence span level, 42 sentences in total. All the test processes were strictly following the literature on working memory tests. As Mackey et al.’s study, composite scores were obtain through the sum of z scores of two different tests.And actually these 24 participants were selected among a large number of students in order to satisfy assumptions for two-way ANOVA. SW normality tests showed the normal distribution of both scores of c-test and WM test, homogeneity of variance test showed ….. And their scores had no significance difference among each cells at the very beginning.The phonological loopCentral executive componentNon-word recall test42 English non-words from Prof. Skehan’s project in CUHKL2 listening span test3 sets per each sentence span level (2-5), 42 sentences in total
7Native speaker interlocutors Two male experienced interlocutorsFour carefully designed training proceduresWatching the video of the instructionof recastsDemonstrating b&g examplesfrom the video clips of pilot studyRole-playing all tasks involved(video-taped)Reflecting the role-playing processAs Prof. Skehan said, if the interlocutors cannot provide recasts in an appropriate way, the research will be dead. Because of their crucial roles, careful individual training for both interlocutors was conducted before the experiment. Four procedures were included. First, they watched the video about the instruction of recasts, then bad and good examples selected from the video clops of pilot study were demonstrated, after that interlocutors had the role-play performing all tasks in the experiment with the whole process video-taped. Finally, they had to reflect what they had done through watching their own performance.Individual training
8Working memory test & C-test (1) Immediate posttest (2/4) ProcedureWorking memory test & C-test (1)Pretest (1/5)Treatment 1(2/1)Treatment 2(2/2)Treatment 3(2/3)The whole procedure of the study lasted for about five weeks. After the selection of participants, pretest was held in the fifth day of the first week, Friday. Three treatment sessions were conducted in the successive three days of the second week and then followed the first posttest on Thursday, immediate stimulated recall was right after this posttest. The delayed posttest was held three weeks after immediate posttest.Immediate posttest (2/4)Stimulated recall (2/4)Delayed posttest (5/5)Procedure (Week/ Day)
9Treatment and assessment tasks MaterialsTreatment and assessment tasksTaskLinguistic targetTypeDirectionof informationEnglishquestionsInformationexchangeSpot-the-differenceTwo-wayFor both treatment and tests, there were three tasks used to elicit the linguistic targets in the study, they were English questions and past tense. The duration of the first two tasks, spot-the-difference and picture-drawing, was controlled within 20 minutes, 10 per each. You can find the examples of pictures which had been used in the study in the handout. And task sequence was all the same in both treatment and tests. Interlocutors were required to provide recasts of questions and past tense through their interaction with the participants when doing tasks in the three treatment sessions.(10 mins)Picture-drawingEnglishquestionsInformation gapOne-way(10 mins)Englishpast tenseInformation gapOne-wayStory-tellingTask sequence was all the same in both treatment sessions and tests.
10Stimulated recallhas been applied as an introspective measure of L2 learners’ cognitive processes, especially noticing.Immediately after the first posttestVideo clips of nearly all of LREs (Language-related episodes)15%-20% distracters & self-initiated recall allowed at any time(recasts of non-linguistic targets / correct responses etc.)Pausing at the end of each LRE and asking “what were you thinking at that time?” (strict training for the researcher)L1 of recall commentsStimulated recall was applied into the experiment to detect noticing and understanding. Video clips of nearly all of language-related episodes were selected as the stimuli to elicit the recall comment. here it means the episodes which include the interlocutors’ recasts of non-targetlike usage of questions and past tense. The researcher asked “what were you thinking at that time” when pausing at the end of these episodes. In order to distract their attention and prevent them from knowing researcher’s real intention, there are 15-20% distracters included in the recall interview. Besides, participants were allowed to control the computer mouse and stop at any places where they wanted to recall their thoughts. Participants used their first language in this process to guarantee no barrier of presenting their thoughts.
11Coding and scoring: stimulated recall comments Stimulated Recall Comments (LRES)Others (Other, No Thoughts, Thoughts Forgotten)Focus on MeaningFocus on FormNoticing L2 FormUnderstanding L2 FormThe figure presented here provides us a clear way to know how I coded all the stimulated recall comments. Three main categories, they were “focus on meaning, focus on form and others”, and noticing and understanding L2 forms were two different levels of focus on form. Noticing was coded as a verbal reference to the target structures without or with mention of rules. Understanding was coded as an explicit formulation of the rule underlying the target structures.Noticing: a verbal reference to the target structureswithout or with mention of rules.Understanding: an explicit formulation of the rule underlyingthe target structuresScoring: “one noticing/understanding, one point” policynumber of N/Unoticing/understanding ratio=total number of comments
12Scoring and coding : task performance Question formation6 Stages based on Pienemann & Johnston (1987) and adapted from a series of studies2 different higher level structures in two different taskscoded as developmentPast tenseTargetlike forms in obligatory contexts were counted — accuracy of productionQuestions participants produced in three tests were coded from stage 1 to 6. The coding scheme was based on Pienemann & Johnston’s developmental stages and adapted from a serial of empirical studies. Only when participants could produced 2 questions with different higher level structures in two different tasks in the posttest or delayed postte would be coded as development, otherwise, no development occurred.For another linguistic targets, scores were given based on accuracy of their production. Suppliance of targetlike past tense forms of verbs in obligatory contexts were counted, error-free ratios was obtained as their final scores.
13What did I find in my research? The relationship between WM & AwarenessQuestions:Understanding data: none from recall commentsNoticing(percentage)WMCMeanSDHigh21.8926.65Low12.2918.24p=0.27nsNext comes to the results. First of all, what we have to bear in mind is that understanding data from English questions was unavailable from recall comments, nobody recalled the detailed rules of question formation. Therefore, research questions related to understanding level could only be answered through the other target: past tense.As to the relationship between WM & noticing, no significant difference has been found for both two linguistic target, but participants’ noticing of past tense may has relationship with noticing because of large effect size. P value was more than .05 may result from the small sample size.There was no relationship between WM and understanding. However, from means presented here, we may find some implications, high wm high noticing, and high wm high understanding.Past tenseNoticing(percentage)WMCMeanSDhigh40.2727.43low22.1914.42p=0.056d=0.87Understanding(percentage)WMCMeanSDhigh18.3221.26low11.1510.12p=0.64ns
14Past tense The relationship between L2 proficiency & Awareness QuestionsNoticing (%)ProficiencyMeanSDhigh10.0914.77low24.0827.72p=0.11, nsPast tenseRegarding the relationship between proficiency and awareness, no significant difference has been found for both questions and past tense. However, look at the means here, participants with lower proficiency level noticed more than those with high proficiency. And for the understanding level, the result was reversed, those with high proficiency level understand more than low proficiency level participants.Noticing (%)ProficiencyMeanSDhigh29.0626.30low33.3920.89p=0.44, nsUnderstanding(%)ProficiencyMeanSDhigh19.2618.01low10.2214.61p=0.64, ns
15L2 proficiency, WM & interaction-driven development QuestionsQuestionsQuestionsPost-testPost-testPost-testPost-testWMCL2 developmentHighLowTotalDevelopment3811No development94131224WMCL2 developmentHighLowTotalDevelopment3811No development94131224WMCL2 developmentHighLowTotalDevelopment3811No development94131224WMCL2 developmentHighLowTotalDevelopment3811No development94131224p=0.041for questions, Chi-square analysis showed that low WM capacity participants achieved more development than those with high WM capacity and the difference was significant in both two posttests.Delayed Post-testDelayed Post-testDelayed Post-testWMCL2 developmentHighLowTotalDevelopment2810No development4141224WMCL2 developmentHighLowTotalDevelopment2810No development4141224WMCL2 developmentHighLowTotalDevelopment2810No development4141224p=0.013
16L2 proficiency, WM & interaction-driven development QuestionsPost-testProficiencyL2 developmentHighLowTotalDevelopment6511No development7131224p=0.68nsAnd no significant difference in their development was found based on two groups of the variable “proficiency”. Posttest and delayed posttest were all the same.Delayed Post-testProficiencyL2 developmentHighLowTotalDevelopment5611No development7131224p=0.68ns
17L2 proficiency, WM & interaction-driven development Past tenseVariablesWM(Means)Sig.EffectSize(d)ProficiencyInteractionWM*ProHighLowPosttest58.0246.75.26.4946.47220.127.116.11.04Delayed posttest67.9618.104.22.16861.2766.01.54However, for the linguistic target English past tense, the results presented a more complicated picture. We can see from this table that there were no significant main effects but WM and proficiency had an interaction effect on learners’ development in both two posttest.
18Interaction (WM*Pro) in the posttest In order to get a clearer picture, let’s look at the line graphs of interaction effect. It is clear here that among participants with low proficiency level, those with high wm capacity achieved marginally significant more in the posttest. Among participants with high WM capacity, those with low proficiency level achieved significant more.18
19Interaction (WM*Pro) in the delayed posttest Results of delayed posttest were the same with the immediate posttest, except that p values .05 and .07 were bit higher, but effect sizes were very high here, so the non-significance may also result from the small size.19
20Development over time Past tense Variables Sig. Test of Sphericity .000Sig.=.149(assumption was met)Test*Proficiency.598Test*WMC.572Test*WM*Proficiency.092
21Provisional Conclusion The relationship between WM capacity & noticingQuestions: showing the trend towards “High WM-High Noti” (ns)Past tense: showing the trend towards “High WM-High Noti”(ns but large effect size)The relationship between WM capacity & understandingQuestions: data unavailablePast tense: showing the trends towards “HWM-HUnder” (ns)The relationship between proficiency level & noticingQuestions & Past tense: showing the trendtowards “LPro-HNoti” (ns)The relationship between proficiency level & understandingQuestions: data unavailablePast tense: showing the trend towards “HPro-HUnder” (ns)The effects of proficiency and WM on L2 developmentQuestions: Low WM capacity — more development (both posttests)Past tense: LProHWM —more development (both posttests)
22ReferencesAndo, J., Fukunaga, N., Kurahashi, J., Suto, T., Nakano, T., & Kage, M. (1992). A comparative study on two EFL teaching methods: The communicative and the grammatical approach. Japanese Journal of Educational Psychology, 40,Dörnyei, Z., & Katona, L. (1992). Validation of the C-test amongst Hungarian EFL learners. Language Testing, 9 (2),Mackey, A. , AI-Khalil, M., Atanassova, G., Hama M., Logan-Terry, A., & Nakatsukasa, K. (2007). Teachers’ intentions and learners’ perceptions about corrective feedback in the L2 classroom. Innovation in Language Learning and Teaching 1 (1),Mackey, A., Philp, J., Egi, T, Fujii, A., & Tatsumi, T. (2002). Individual differences in working memory, noticing of interactional feedback and L2 development. In P. Robinson (Ed.), Individual Differences and Instructed Language Learning, (pp ). Amsterdam/ Philadelphia: John Benjamins Publishing Company.
23ReferencesPienemann, M., & Johnston, M. (1987). Factors influencing the development of language proficiency. In D. Nunan (Ed.), Applying Second Language Acquisition Research (pp. 45–141). Adelaide: National Curriculum Resource Centre, AMEP.Roberts, M.A. (1995). Awareness and the efficacy of error correction. In R. Schmidt (Ed.), Attention & awareness in foreign language learning (pp ). Hawaii: Second Language Teaching & Curriculum Center.Sagarra, N. (2007). From CALL to face-to-face interaction: the effect of computer-delivered recasts and working memory on L2 development. In A. Mackey (Ed.), Conversational interaction in second language acquisition (pp ). Oxford: Oxford University Press.Trofimovich, P., Ammar, A., & Gatbonton, E. (2007). How effective are recasts? The role of attention, memory, and analytical ability. In A. Mackey (Ed.), Conversational interaction in second language acquisition (pp ). Oxford: Oxford University Press.