Abstract This study aims to increase written grammatical accuracy by facilitating learners’ use of corrective feedback (CF). A quantitative quasi-experimental study design is used to compare effects of traditional (teacher-provided) scaffolding and self-scaffolding, and compares these to unscaffolded direct written CF. Participants were Korean EFL university STEM students (n = 109) in compulsory academic English writing classes in South Korea (hereafter Korea). To elicit metalinguistic reflections, traditional scaffolding was provided in 1-to-1 conferences, and self-scaffolding was provided by worksheets. Unscaffolded direct written CF was provided to a standard-treatment comparison group. The study was designed for ecological validity, with data being derived mainly from classroom writing samples, worksheets, and audio-recordings of conferences. Results provide support for previous CF research: mixed ANOVA results suggest that all three groups experienced similar, significant, and durable increases in grammatical accuracy. The study contributes to existing knowledge by a) using linear regression to demonstrate that quality of metalinguistic reflections does not necessarily predict an increase in grammatical accuracy; b) establishing that there may be difficulties in scaffolding oral metalinguistic reflections with the described population; and c) drawing on data from the background survey and interviews to inform the interpretation of the results.

中文翻译：

---

教师脚手架和自脚手架纠正反馈与直接纠正反馈对英语二语写作语法准确性的影响

摘要 本研究旨在通过促进学习者使用纠正反馈 (CF) 来提高书面语法的准确性。定量准实验研究设计用于比较传统（教师提供）支架和自支架的效果，并将这些与非支架直接书面 CF 进行比较。参与者是在韩国（以下简称韩国）参加必修学术英语写作课程的韩国 EFL 大学 STEM 学生（n = 109）。为了引发元语言反思，在一对一会议中提供了传统的脚手架，并通过工作表提供了自我脚手架。向标准治疗对照组提供无支架的直接书写 CF。该研究专为生态有效性而设计，数据主要来自课堂写作样本、工作表和会议录音。结果为之前的 CF 研究提供了支持：混合方差分析结果表明，所有三组都经历了相似、显着且持久的语法准确性提高。该研究通过以下方式为现有知识做出贡献： a) 使用线性回归来证明元语言反射的质量不一定预测语法准确性的提高；b) 确定在所描述的人群中构建口头元语言反射可能存在困难；c) 利用背景调查和访谈的数据来解释结果。该研究通过以下方式为现有知识做出贡献： a) 使用线性回归来证明元语言反射的质量不一定预测语法准确性的提高；b) 确定在所描述的人群中构建口头元语言反射可能存在困难；c) 利用背景调查和访谈的数据来解释结果。该研究通过以下方式为现有知识做出贡献： a) 使用线性回归来证明元语言反射的质量不一定预测语法准确性的提高；b) 确定在所描述的人群中构建口头元语言反射可能存在困难；c) 利用背景调查和访谈的数据来解释结果。