Sunday, October 28, 2007

weekly reflection (week 10) by Ping and Yao

I have to thank Yao for taking careful notes in class. I only tweaked it by adding my own notes. I hope the commentary below grasps most of the main ideas we discussed in class. Please feel free to comment on anything I missed. Thank you.

Q: How refined should the transcription be?

From a CA perspective, any pause can carry interactional meaning. It's impossible to determine what a pause means beforehand. The entire interactional context, including prosody, gestures, pauses, and gaze, should be taken into account. Therefore, a careful transcription is necessary. The decision on how refined the transcription should be is really based on the researcher's analytical purposes. Transcriptions are always selective. The bottom line is to be as much faithful as possible to your data. And the transcription system should be consistent in order to avoid confusion. The final transcription is always the result of a series of compromises between faithfulness to the data and the readability of the transcription.


CA's emic interests: studying behavior from inside a particular system, looking at the subsequent turns to interpret the intended actions to be achieved

Emic vs. Etic

This emic vs. etic distinction comes from the study of anthropology.
emic: looking at the data, let the category emerge from the data. The term 'emic' comes from phonemic: study of sound as they represent category that can form contrast (e.g., very vs bury) v and b doesn't make phonemically difference but probably make phonetic difference. Meaning comes from within the participants in the context. From the participants' own perspective, analysts examine what is achieved in the sequential development and how meaning is made relevant to participants. In other words, meaning is situated in the context and can't be described outside the context.

etic: researchers impose the categorization onto the data. It comes from phonetics, study of sound as pronounced physically. The etic viewpoint refers to meanings from the outside perspective but not from the participants.

Hauser (2005): there are possibilities of interpretation and the teacher impose his/her interpretation but we don't know whether it's the intended interpretation is not known. Whether the meaning maintained depends on how the meaning is negotiated.

Q: how can we interpret the action of the participations?

Lourdes: It's like postmodernism: There is no truth out there. There is not the meaning of utterance independent of the context. Without looking at the context, it's impossible to understand how meaning is constructed through the sequence of utterance. In Hauser (2005), he is saying “this is one possibility” “this is my interpretation of this” He's trying to make the best proximal interpretation, but in the end there is no truth. Conversation analyst walks a fine line, it causes the tension how to use the evidence to make the interpretation, which can not be validated.

Ping: It helps to have outsiders to review the data together. It helps me to refine the data and consider alternative interpretations.

Yuki: Two types of approaches are possible:
interpretive approach
emic approach
Lourdes: However, critical conversation analysis is not real conversation analysis.

Subject: intersubjectivity between participants
objectivity is ideal.
Lourdes: Though there is no single true interpretation, but we need to interpret by the best methodology possible. It's a tension between taking “analyzing turn” as the universal methodology to understand “interaction”,while at the same time “turn” can not be interpreted without tightly linked to the context.
In CA studies, there's a lot of hedging in the interpretation, which indicate the author is presenting one interpretation, not the other.

Sangki: the main idea of the article is that meaning is always co-constructed and can not viewed without looking at the context.

Ping: For CA studies, there's a lack of longitudinal works, looking at the same phenomena over time. Also, there is this concern of how to apply results to the pedagogical context..

Lordes: Koshik (2002) is very pedagogically oriented. There's a lot more to meaning. We shouldn't underestimate the actual richness the interactions of how corrections are given and taken. It's a healthy reminder that we shouldn't just do counting. So we have one extreme of analyzing every second of turn to analyze meaning and another extreme of completely deviating from meaning and just counting the number of corrective feedback.
Koshik (2002) talks about how the teacher upgrades and downgrades assistance including the prosody cues. The idea that assistance is incremental is interesting. The utterance was designed to be incomplete to prompt self-correction. CA is all about local sequential context of interaction: How things unfold and build-up to something. For this reason, CA could be used to look at when the assistance is needed, when it should be upgraded, and when it should be withdrawn in the learning process. This is relevant to our reading for next week.


Ending comments from Lourdes: It's interesting that we don't treat other approaches (statistics, cognitive) as marked, but treat CA as marked (too much jargon). The truth is all approaches have their own jargon and they are equally ratified. So when we choose our approach, we don't treat it as default. The approach has to be a good match with the researcher.

Tuesday, October 23, 2007

Weekly Reflection (W9) by Hung-Tzu

This week, we continued our oral update on the projects. Four presentations with diverse topics were presented.

Project by Sangki and Mune

Conceptualizing agents in discourse and frequency effects in English L2 learners’ overpassivization errors: A replication and extension of Ju (2000)

Based on cognitive explanation proposed by Ju (2000), overpassivization errors will be examined against three independent variables, causation types ( 2 levles: external vs. internal causers), word token frequency (2 levels, high frequency verbs vs. low frequency verbs), and types of unaccusatives (2 levels, alternating vs. nonalternating unaccusatives). A grammatical judgment test will be used to test intermediate learners (N=20), advanced learners (N=20), and native speakers (N=10).

The class had a brief discussion on the grammatical judgment test including what the participants were asked to judge the sentences, the specific items on the test, and also the distracters included. Since overpassivization is an error commonly found in advanced learners, learners will not have overpassvization errors untill they have passive knowledge. The distracters are designed to test learners’ past knowledge. No errors on overpassivization might either mean that learners are so advanced that they have no problem, or that learners may simply have not acquired passives yet.

Project by Luciana

Focus on form and self repair: Some insights into foreign language learning

Luciana gave us a brief report on the dissertation that she is currently working on. The study asked the following five research questions (1) How do task types influence focus on form and self-repair? (2) To what degree does learners’ proficiency level affect their focus on from and self-repair? (3) What is the nature of linguistic knowledge targeted in focus on form and self repair? (4) How does the interaction depth influence focus on from and interaction pattern? (5) Do learners perform similarly in focus on form and self repair?
This is a very rare study since Luciana looked at group interaction instead of dyads. From her preliminary data analysis, the class suggested using medium to look at group distribution and also individual learner participation within group activities. We also discussed the term ‘depth of LRE’ and it was mentioned that a lower inference label closer to how the study is operationalized such as length of LRE might be able to avoid mis-interpretation from readers.

Project by David

David presented an interesting CALL project in which an alien interacted with learners and give feedback on errors through negotiation of meanings. Suggestions on how the alien project could be expanded included tracking student responses after feedback is given, choosing more generative target structure for the study and providing theoretical ground for this alternative way of giving feedback (i.e. justifying the pedagogical reasons). Yao mentioned that this type of study might be related to human computer interaction or ethnographic research within computer environment. The followings are references that Yao sent to the class list.

Hampel, R. (2003). Theoretical perspectives and new practices in audio-graphic conferencing for language learning. ReCALL, 15(1), 21-36. CALICO (Vol. 20, No. 3); PujolĂ  (2001) and Bangs (2003); Toole and Heift (2002); Heift (2003).

Project by Dan

Promoting grammar awareness with color-coded feedback

Dan reported his pilot study of color-coding method to give feedback on student writing. The interview from this pilot study revealed that students perceived the system as beneficial in terms of raising meta-awareness. With the short treatment period, Dan suspected that there might not be significant improvement in accuracy of writing, yet, the followings are possible ways to demonstrate the benefit of the color-coding system. (1) Looking into students’ self revision ability at the beginning and at the end of the color-coding treatment might be a way to quantify student learning. (2) Survey designed with Likert scale asking students’ preference in receiving feedback both before and after the treatment. (3) Semi-structured interviews eliciting information on how learners engaged in the revision process using the color-coding system.

Followed by the discussion on the project, we did a hands-on activity applying Dan’s color-coding system to student writing. The activity led to discussion on realistic classroom problems such as how much feedback to give and the many decisions that teachers go through when giving feedbacks.

Sunday, October 14, 2007

Weekly Reflection (W8) by BoSun

This week we started with oral presentation for final project. On Tuesday, the presentation was ordered as follows: Hung-Tzu, Kevin, Yun-Deok and BoSun. On Thursday, Ping, Sorin, Yuki, Myong Hee presented their research proposals. The topics varied and here are the projects sorted by mode of feedback, i.e. oral feedback vs. written feedback.
-------------------------------------------------------------------------------------
Research project addressing written feedback: Hung-Tzu, Kevin and Yuki

(1) Error correction in L2 writing: How successful are students in revising lexical errors? by Hung-Tzu
Her research will deals with written error correction and students revision of lexical errors with use of three different strategies (thesaurus, online dictionary and collocation dictionary). The rationale for her research is that it is necessary to examine lexical errors separately form grammar errors since written feedback literature revealed that the effectiveness of feedback types and learner’s ability to revise differ depending on the types of errors, i.e. lexical vs. grammar errors (Ferris &Robers, 2001; Gaskell &Cobb, 2004; Ferris, 2006).

The participants are 44 ESL students of UHM at two different levels (20 intermediate and 11 advanced learners) and they were taking academic reading courses with a focus on vocabulary learning. The procedure is as follows; 1) the participants completion of writing task followed by reading activities, 2) teacher’s provision of indirect feedbacks (underlining) for five lexical errors 3) student’s revision of their own writing using one of the three tools (thesaurus, online dictionary and collocation dictionary) 4) students’ reflection and evaluation of their writing and revision process. The data is 155 first drafts and 155 revised drafts including 775 lexical errors. The data will be analyzed by use of concordancer to examine distribution of the errors and learner’s repair depending on their strategy.

(2) Indirect error correction and improving grammar in L2 writing By Kevin
His research questions are 1) can indirect errors correction lead to improved performance on certain grammatical constructions on first drafts in an intermediate L2 writing class? 2) does indirect correction affect different grammatical constructions differently? His assumption is that indirect feedback involves depth of processing which encourages students correct their errors better.
The participants are 15 university-level ESL students aged 18-24 with various L1 background. At the time of data collection, they were enrolled intermediate writing class in the HPU focusing on grammatical accuracy. Three drafts were collected; teacher gave indirect feedback for first and second draft, students revised their first and second drafts and resubmit the drafts (1st draft-feedback-2nd draft-feedback-3rd draft). He reported that a rage of grammatical errors included verb form, verb test, incorrect articles, etc and the number of errors reduced by the third draft.

(3) Enculturation into academic discourse: focus on deficiency or agency by Yuki.
She is planning to conduct two studies focusing on the writers’ enculturation into academic English writing with two different data; one from university level- ESL writing classes in Hawaii and the other from her own writing. At the first project, she is addressing how contextual factors shape students into the academic discourse community. The participants are divided into two groups: 21 graduate students enrolled at advanced college academic ESL writing course and 22 undergrad freshmen taking freshmen composition course. The data will be analyzed for 1) types of feedback, 2) incorporation of feedback by types, 3) thematic analysis of students’ perception of their language, content, and rhetorical style development.

At the second project, she is carrying out longitudinal study of her own negotiation and enculturation process into disciplinary scholarly writing. For data analysis, she is employing autoethnography.
-------------------------------------------------------------------------------------
Research projects about oral feedback: YunDoek, BoSun, Ping, Sorin and Myong Hee

(1) Which on can language learners rely on best, recasts or prompts, with relation to learner’s perception? By YunDoek
Her study addresses relative effectiveness of recasts vs. prompts on L2 learning in accordance with learner perception of the feedback in classroom settings on both short and long term basis. Her research questions is tackling following issues 1) the level of learner’s immediate uptake in response to recasts and prompts 2) the level of learner’s uptake for recasts and prompts on a long term basis 3) difference between immediate and delayed-post test performance for recast and prompts 4) similarities and differences between teachers’ and students’ preference for different feedback types across different linguistic items.
The participants will be students from HELP in UHM. The design will be both descriptive and experimental, adopting treatment and pre-post test. The participants are divided into control and treatment group, take the pre and post test, and go through treatment either recasts or prompts between pre- post tests.

(2) Reexamination of sub-categories of recasts and learner uptake by BoSun
She assumes that recasts constitute continuum with explicit and implicit end based on oral feedback literature (Sheen, 2007) and is tackling following issues 1) do 4 different types of recasts adopted from Lyster & Ranta (1998) enhance the acquisition of L2 grammatical structure 2) what characteristics of recasts lead to learner uptake and repair better? 3) Do 4 different types of recasts result in different effects depending on the learners’ proficiency?
The research will be descriptive study with two different levels (intermediate and advanced) of English classes (one from HELP and the other from ELI) in UHM. The data will be analyzed by use of coding scheme from Ryster and Ranta (1998) which sub-categorized recasts into 4 types: isolated declarative, isolated interrogative, incorporated declarative, and incorporated interrogative, depending on intonation (falling vs. rising) and existence of additional information (with or without additional information). The measure for acquisition is uptake, which is defined as learners’ immediate response followed by teachers’ recast (Lyster & Ranter, 1997). Uptake is again sub-divided into two categories: repair (correction) and needs repair (acknowledgement of errors).

(3) Implicit/ Explicit recasts, learner’s responses to recasts and linguistic development by Sorin.
She is examining 1) which types of recasts (implicit vs. explicit) leads to more learner uptake and subsequently more linguistic development of the target structure 2) whether primed production in response to recasts occur, if so, which types of priming leads to more frequent primed production.
The participants are KSL learners in Korea. Her research design is quasi-experimental wit pre-post-delayed posttest. The target structure hasn’t been finalized yet, and she is considering relative clause to be one of the good candidates. Coding scheme will follow the one from Lyster and Ranta (1998) and two out of the four types of recast will be chosen for implicit and explicit recast. The measurement will be uptake and primed production. Uptake will be operationalized as a student’s utterance immediately following the teacher’s feedback (Lyster & Ranta, 1997) and primed production is defined as learner’s new utterance using target structure form provided in the recast within six turns of recast, adopting McDonough & Mackey’s (2006) definition.

(4) Organization of error correction sequences in form-focused classroom by Ping
Her research questions are 1) what are the different types for corrective feedback in form-focused classroom? 2) does the classroom context influence students’ orientations to the corrective feedback? The participants are 14 students’ in Chinese 101 class (at beginner level) at UHM, The data were analyzed by use of CA. She has found that 1) teacher prompt and learner production was predominant sequence 2) other-initiated other repair showed high frequency whereas other-initiated self-repair and self-initiated other repair displayed low frequency. 3) self-initiated self-repair is rare.

(5) Investigation of small group interaction in a Korean university EFL classroom by Myong Hee
Her research is dealing with 1) types of collaborative learning and their distribution 2) the level of uptake for each category measured by repair and needs-repair. The participants are 24 students, enrolled at a college English reading course. The data is small group interactions (6 triads and 3 dyads) tape-recorded which lasted 12-18 minutes for each. Data analysis will adopt quantitative and qualitative analysis. Quantitative analysis focuses on distribution of various types of peer assistance and qualitative analysis will examine 1) co-construction 2) encouragement to topic continuation 3) self-correction with use CA analysis.


Tuesday, October 2, 2007

Weekly Reflection (W7) by Sorin

On Tuesday, we started with small group discussions over (1) two points that we either agreed with the article or thought to be important or valuable points from the article, (2) two points that we disagreed with or had reservations about, and (3) 2 points that were beyond the review (which they missed or couldn’t see in 2001, when the article was published).

(1) Agreed or Important/Valuable points
The first group pointed out that the effectiveness of recasts is affected by the target structure of the study in relation to the developmental readiness of learners. In other words, whether learners are ready to acquire the target structure or not will affect the effectiveness of recasts. The second point was made on the types of recasts. Recasts can vary; it can be either explicit/ implicit and provided with or without emphasis utilizing nonlinguistic cues (as Chaudron suggested). These differences can definitely have an impact on the students’ noticing of recasts and consequently the effectiveness of recasts. The third point made was that investigating “private speech” as an indicator of students’ noticing of recasts (Ohta, 2000) was interesting. It could be an interesting measurement of students’ noticing of recasts, however, no research has investigated private speech since Ohta.

Prompted by the third point, we started to discuss the ways of assessing the effectiveness of recasts (or in other words, success or impact of recasts). We’ve looked at the definition of “uptake” made by Lyster & Ranta (1997) on page 739, and found that uptake can be a very slippery term since it covers a range of students’ responses from a simple acknowledgement of recasts to a repetition of recasts (which is called as “echo”) and students’ self repair. Then we moved to the measures used in L1 studies (on page 750). In early L1 studies, children’s imitation of the adult’s recasts was often sought as an evidence of the effectiveness of recasts. However, later on researchers started to investigate emersion of targeted structure in children’s subsequent utterances. On the other hand, in L2 reserach, various measures were used, as follows:

- Interlanguage change: It was investigated mostly in laboratory studies, at least in short term period, through pretest and posttest design (sometimes with delayed posttest).

- Immediate reactions to recasts: uptake, repair, etc.

- Private speech: It was used only in Ohta’s study (2000), in whihc students were more likely to react to recasts when it was addressed to another learner or to class than to them.

- Learners’ perception: Learner’s perception of recasts was investigated through stimulated recall (Mackey, Gass, & McDonough, 2000).

- Primed production: Primed production is learner’s production of a new utterance using the target structure in a few turns after recasts, and was first investigated by McDonough & Mackey (2006). In their study, learner’s production within six turns after recasts was examined.

Another important point made in class was that the effectiveness of recasts can vary depending on the setting of the study as follows:

- Intensive vs. Extensive: Regarding the density of feedback provision (how many errors were corrected and how often those errors were corrected)

- Specific vs. Broad: Regarding the range of forms corrected (whether only errors on particular structures were corrected (Doughty & Varela, 1998; Ortega & Long, 1997) or a broad range of errors were corrected (Lyster & Ranta, 1997; Oliver, 1995))

- Proactive vs. Reactive: Regarding the existence of pre-selected target structures in studies (whether there was specific target structure to teach or no particular structure was pre-selected and thus correction was incidental to learner errors)

- Communicative vs. Formal (overall context): Regarding the nature of classroom setting (whether the nature of classroom is communicative, content-based (immersion), or formal; whether it is in foreign language context or second language context, etc.)

- Relational feedback by teacher vs. Detached or predominantly cognitive feedback: Regarding personal, affective, and social factors affecting the dynamics of interaction (whether fine-tuned feedback was provided in consideration of the socioclutural factors of the preexisting relationships or predominantly cognitive feedback was provided disregarding these factors, most likely among the participants with no prior relationships)

What constitutes positive evidence and negative evidence was the last important point made in class. Even though several studies (Iwashita, 2003; Leeman, 2003; Long et al, 1998; McDonough & Mackey, 2006) tapped into this issue, no study has provided a review of this line of research. It seemed like that recasts provide both negative evidence and positive evidence at the same time but recasts make positive evidence more salient according to Ortega. In addition, the effectiveness of recasts comes from both positive and negative evidence.

(2) Disagreement or Criticism
The first point addressed was that there was disagreement in the definition of recasts, which inevitably caused comparability problem among the recasts studies. Also, the findings of L1 studies cannot be compared to the results of L2 studies (Sangki). In addition, no study has ever paid attention to paralinguistic cues provided with recasts. Second, the narrative literature review of this article seemed to be limited in synthesizing the findings of the previous studies, even though it was well written and helpful. This article can be regarded as authoritative review which was written by renowned scholars. Yukiko suggested to read the last chapter of Mackey’s forthcoming book, and we will read *Russell & Spada (2006) as well. Kevin also mentioned that it seemed that there are more similarities in L2 studies than L1 studies in recasts disagreeing with what the authors suggested in the article.

(3) Beyond the review
No study in recasts has investigated paralinguistic cues provided with recasts except Sheen (2006?), and thus we need to take into consideration the paralinguistic cues as well in our studies by analyzing oral or audio-visual data. In addition, the effect of students’ familiarity with teacher’s teaching style on recasts was not much considered in previous studies. Chaudron's dissertation was the only study showing that teachers less corrected learner errors at the end of semester than the beginning of the semester, and some studies on motivation have shown that there was fluctuation of motivation during a semester. Thus, it would be interesting to see how various aspects of recasts change as relational aspects of classroom setting change. Third, L1 studies showed that as children grow older (in other words, as their proficiency developed), the provision of recasts was decreased. Thus, it would be worth collecting classroom observation data across entire curriculum and investigating how the amount of recasts changes as students’ proficiency develops. Another point made by Myunghee was that since most of the studies in recasts have looked at NS-NNS interaction, it would be interesting to examine NNS-NNS interaction.


*Russell, J., & Spada, N. (2006). The effectiveness of corrective feedback for the acquisition of L2 grammar: A meta-analysis of the research. In J. M. Norris & L. Ortega (Eds.), Synthesizing research on language learning and teaching (pp. 133-164). Philadelphia: John Benjamins.


-------------------------------------------------------------------------------
Ellis, R., & Sheen, Y. (2006). Reexamining the role of recasts in second language acquisition. Studies in Second Language Acquisition, 28, 575-600.

Iwashita, N. (2003). Negative feedback and positive evidence in task-based interaction: Differential effects on L2 development. Studies in Second Language Acquisition, 24, 1-36.

On Thursday, we started with a small group discussion on two things that were already brought up in Nicholas et al. (2001) and two things that were forward-looking agendas in Ellis & Sheen (2006). After a group discussion, we had a whole class discussion. Then, Noriko Iwashita came in and told us about her study (Iwashita, 2003) and her experience of conducting studies in error correction: difficulties, concerns, and helpful tips.

(1) Issues already brought up
The first issue pointed out in both articles was that recasts are ambiguous since they are not always noticed by learners as corrective feedback and thus they are considered to be less effective compared to other types of feedback moves. However, the next question raised by Ortega was that the fact that learners miss the corrective function of recasts (i.e. not perceiving them as corrective) does really mean that they do not perceive it? Ellis and Sheen claimed that “whether recasts afford positive or negative evidence is tied up with how learners interpret their illocutionary force” (p. 585). Put another way, if learners do not interpret recasts as corrective, then recasts only serve as positive evidence. On the other hand, if learners perceive them as corrective, then they provide negative evidence. We agreed that overall learner’s orientation toward interaction is important. However, our conclusion was that recasts do not necessarily need to be perceived as didactic to be regarded as a source of negative evidence. Definitional differences in recasts studies were also addressed in both articles. Not only definitional differences but a variety of recasts (different types of recasts) was also mentioned, although Ellis & Sheen provided more expanded discussion on this issue. Besides, the two articles dealt with the role of uptake; uptake cannot be an evidence of acquisition even though it can be an evidence of noticing.

(2) Forward-looking agendas
How to define implicit or explicit recasts and how to operationalize the degree of explicitness was the first agenda pointed out. Ellis & Sheen pushed this issue forward by discussing different types of recasts. Second, Ellis & Sheen elaborated the importance of sociocognitive perspective, which was also discussed in Nicholas et al. We will read some of the chapters, on this issue, from Hyland & Hyland book (2006) later on. Third, they further expanded the argument about the effectiveness of recasts on acquisition and suggested that we should consider not only whether recasts facilitate acquisition but also when and how they do so. Adding to this claim, Ortega proposed that we should also compare the effect of recasts with other types of feedback moves in pursuing this issue of not only “whether” but also “how” and “when”. There are variations in each corrective feedback moves, and it could be the case that types of feedback do not really matter. What really matters and what requires further investigation can be some other level of abstraction, such as the degree of explicitness, cutting across different moves instead of classifying different moves. Forth, Ellis and Sheen seemed to conclude that explicit correction is more effective than recasts and the more explicit a correction move can be, the better it is. With this claim, Yundeok raised a question; if explicit correction only leads to building explicit knowledge, what about implicit knowledge? Ellis, however, didn’t really mention this issue in the article, and he seemed to be just concerned with acquisition benefit of recasts in general sense. Fifth, learner’s orientation to discourse (whether learners see language as an object or as a tool to convey meaning) was more elaborated in Ellis & Sheen. It is possible that, if learner’s orientation is toward accuracy, then they will perceive recasts as corrective, even though the context is communicative as Lyster (2006, 2007) suggested.

In addition to these agendas, several important and interesting issues were discussed. With regard to target structures, Nicholas et al. concluded that recasts are more effective with already known forms (p. 730, 752). On the other hand, Lyster hypothesized that recasts are more effective with new forms which haven’t learned, and prompts (or models?) are more effective with already learned forms. Thus, further investigating this issue, confirming any of the hypothesis, would be worthwhile. Another interesting claim Kevin brought up was that fewer recasts can make it more salient, which was also suggested in Nicholas et al (p.743, p.728). In L1 child acquisition, parents rarely recast their children’s utterances thus making recasts easy to notice for children. However, in L2 classrooms, recasts are too frequently offered and thus making them less “marked” instead of being salient Furthermore, parents recast incorrect utterances more frequently than repeat correct utterance, and children repeat corrective recasts more than mere confirmation of their utterances (Nicholas et al., p. 726, 729, 740, 751). Therefore, it would be interesting to examine the frequency issue as well as uptake after corrective utterances versus non-corrective utterances. As last issue, Yundeok pointed out that studies haven’t really looked at teacher’s perception of recasts but only learner’s perception, except Nabei & Swain (2002), which we will read later on. Adding to this comment, Ortega suggested that it would be best to include pre-posttest gain, learner’s (and possibly teacher’s) perception, and discourse data in recasts studies. Up to now, studies have examined only one of them in isolation, except the only exception of Iwashita (2003): she included pre-posttest gain and discourse data. Besides, Ortega mentioned that primed production as shown in excerpt 1 (p. 576) would be new benefit of recasts, what researchers were looking for.

(3) Noriko Iwashita
The hardest part of conducing studies in error correction for her was to code and to analyze data, particularly positive and negative evidence, reaching high intercoder reliability. In her study, she defined positive evidence as instances that NS initially use target structures or vocabulary followed by a NNS’s targetlike or incomplete utterance providing a target model of structures in focus. Selecting appropriate target structure for beginning JSL learners was difficult as well; she was looking for a new structure, however, the learners already had learned most of the structures, even though they were in the second semester of learning Japanese (less than 20 weeks of instruction in pseudo-communicative classrooms). She also faced a problem with developmental readiness with the progressive –te verb form. While some of the students understood it as a structure, other students learned it as words (like chunks), which brought another problem when she coded the data. She also advised us to consider what kind of statistics to use for data analysis in advance and to report individual data in our studies as well.

In order to help your understanding, the followings are a brief summary of her study. In an experimental study, she investigated the role of task-based conversation in the development of two Japanese structures (locative initial construction and a progressive verb form) by 55 L2 learners of Japanese, focusing on positive evidence and negative feedback. What made her study special was that she did not only examine pretest and posttest gain but also analyzed interaction data. Analyzing the interaction data enabled her to identify three types of positive evidence (completion, translation, and simple model) and two types of negative feedback (recasts and negotiation) provided by NS interlocutors during interaction. Among the moves, models were the most frequently provided one followed by recasts. Task-based conversation was proved to be effective for the JSL learners learning the two target structures. However, mixed results were found regarding the effectiveness of positive evidence and negative feedback. Models (positive evidence) were found to be effective on locative construction only for the learners with above-average scores on pretest (at the threshold level of proficiency) while recasts were effective on the progressive –te verb form regardless of learners’ current mastery of target structures.