本研究は、部分リキャストと遮断リキャストの、効果を検証した。結果は、遮断リキャストを受けた実験群はdoに始まる疑問文の産出について短期的上昇を示したが、この上達は遅延事後テストの時点までは保持されなかった。一方で、部分リキャストを受けた実験群は両方の疑問文の産出について、上達は漸進的であった。効果持続性において部分リキャストのほうが遮断リキャストより優れていると思われる。その理由は、（１）遮断リキャストはセンテンスを途中で遮断するので、学習者は完全なセンテンスを構築する訓練を受けない、（２）遮断リキャストの場合より、部分リキャストの場合のほうが、自己修正の機会が豊富である。

While recasts, a type of corrective feedback, are generally considered beneficial for learners’ L2 development during communication-oriented instruction, the extent of its effectiveness has been reported to depend on the way in which they are provided. Thus, there is a need to look into how teachers treat learners’ utterances when providing recasts.  Data were collected from Japanese NNS teachers to investigate 1) how they provide recasts and 2) how their beliefs affect their decisions regarding recast provision. The recordings of teacher-student interactions revealed that teachers tend to provide recasts inconsistently. The interviews conducted after the recordings suggested that recasts could arguably be dependent on teachers’ language proficiency. That is to say, although their comments indicated that they do acknowledge the value of CF for learners’ interlanguage restructuring, they may leave some of the learners’ errors untreated because automatic, real-time processing required for recasting is onerous for them. The findings of the present study can contribute to NNS teachers’ professional development by raising their awareness of their own limitations and of the importance of improving their proficiency, especially their automaticity, in using the target language.

Abundant research on NS (native speaker)-NNS (non-native speaker) interaction has been undertaken to explore various corrective feedback (CF) types and their roles in second language acquisition (SLA). Recasts, a type of implicit CF, have been a target of empirical and theoretical inquiry as their acquisitional value is widely disputed. Some researchers praise recasts for their unobtrusiveness and learner-centeredness (e.g., Long, 1996); others argue that the lack of clear indicators of negative evidence may lead learners to overlook teachers’ intention to make corrections and thus may not lead to learners’ interlanguage (IL) restructure (e.g., Lyster, 1998). Yet, a closer examination of recast studies reveals that the discrepancies in the findings can be attributed to the fact that the efficacy of recasts is constrained by numerous factors. The present study explored how foreign language (FL) teachers’ L1 background affects their provision of recasts. For this study, data were collected from NS teachers (n=10) and NNS teachers (n=9). The findings indicated that teachers’ L1 does indeed affect the provision of recasts in terms of quality and quantity. As for the quality of recasting, NS teachers’ used techniques to enhance the saliency of recasts more so than NNS teachers: the former provided segmented and stressed recasts. These characteristics help learners to attend to the positive and/or negative evidence in recasts in order to identify and/or notice the gap (e.g., Loewen and Philp, 2006, Sheen, 2006). As for the differences in the quantity of recasting, the NNS teachers provided less recasting than NS teachers; moreover, over a third of the NNS teachers’ recasts were not useful as they themselves contained an error. Given that positive evidence in recasts is the most readily available input which learners can use for their IL and TL comparison, the provision of erroneous recasts is problematic as they may encourage incorrect IL restructuring and/or fossilization. Provision of recasts requires conditioned reflex on the part of teachers, and for NNS teachers this real-time process is as demanding as it is for learners. For teachers’ professional development, there seems to be a need for assessing the usability of recasts not only from learners’ point of view but also from teachers’.In some countries such as Japan, EFL lessons have valued a more traditional pedagogy such as the audiolingual and translation methods where teachers employ a more explicit type of CF. Yet with a rise in demand for more communicative instruction, NNS teachers are required to learn ways to provide efficient recasts. The results of these studies serve to provide a model of how teachers can provide CF in such instructional contexts. Further research will concentrate on exploringhow the two teacher types differ in their beliefs about CF and how this factor influences their actual practice.

It has been noted that learners improve their interlanguage (IL) through conversational interaction for the following reasons: 1) Interaction increases the opportunity for learners to receive corrective feedback (CF), 2) learners’ production of uptake triggered by CF allows opportunities for hypothesis testing and strengthens existing linguistic knowledge, and 3) this repetitive feedback-modification pattern contributes to automatization. The present study was conducted to testify whether the above theory applies for recasts. Specifically, the experimental classroom research examined the immediate benefits of recasting on learners’ accuracy in question formation in the context of L2 interaction. Two classes of Japanese high school EFL students (n= 72) in Saitama participated in the study. Students enrolled in the experimental class was told to respond whenever they felt that they had been corrected by the teacher, while students enrolled in the controlled class was not told to do so. Each of the 30-minute treatments over a period of six weeks was audiotaped, videotaped and transcribed. Recast sequences involving students’ uptake were then categorized using the framework used by Lyster and Ranta (1997) in order to investigate whether there was a correlation between students’ immediate uptake and their test scores. The results revealed that both groups significantly increased their scores from the pre-test to the post-test; however, there was by contrast no significant difference in gains in accuracy between the two groups even though learners in the former group produced a greater amount of uptake. When the learners’ test scores and transcripts were examined individually, a logistic regression analysis revealed that learners’ advancement to a higher stage in the developmental sequence for question formation was limited to those who had produced ‘repair’ and/or ‘needs-repair modified’. This finding indicates how that only certain uptake moves, namely those involving the production of modified output involving developmentally advanced question forms, was the only type of uptake that acted as a significant predictor of language development. Current classification of learners’ modification following corrective feedback (CF) uses terms such as ‘repair’ and ‘modified output’ interchangeably; this result suggests a necessity for development of a coding scheme that reflects the reality more accurately. Moreover, learners’ ability to actively respond to recasts for production and/or recognition of a higher stage form was constrained by their developmental readiness. Learners who produced a relatively large number of roughly, if not accurately, targetlike question forms in the pretest tended to score higher than others in the posttest. In order for learners to benefit from recasts, the target structure should be learnable – not too complex but a little beyond their knowledge so that they can go through IL restructuring via ‘noticing the gap’ that triggers reformulation. The results of the study served to provide the following evidences: 1) recasts are effective in strengthening already existing forms, 2) developmentally ready learners tend to benefit more from recasts as they are able to produce reformulation in the form of modified output following teachers’ provision of recasts, and 3) modified output (‘repair’ and ‘needs-repair modified’) are significantly related to L2 development.