修正性回饋對台灣大學生英文第三類假設句寫作正確性之影響

Effects of Written Corrective Feedback on Taiwanese EFL College Students’ Accurate Use of Type III Conditionals

本研究旨在探討修正性回饋對台灣大學生在英文寫作使用第三類假設句型文法正確性的效果。研究對象為中台灣某大學一班四十二位大一應用英文主修學生，根據一份中高級全民英檢閱讀部分的成績，分為直接訂正、文法解釋及內容回饋組三組。第一周上課時，這班學生根據四格漫畫於五十分鐘內完成一篇三段式約三百字的作文。之後第一位研究者分別就兩個實驗組的學生在相關句型文法使用錯誤或該用而未用處給予直接訂正與文法解釋，而內容回饋組則給予文章意義及內容建議。兩周後，學生在收到作文時，先花十分鐘看過個別的回饋，而後在三十分鐘內完成寫作修改。每位實驗對象第三類假設句型文法使用之正確性的測驗分數均以百分比表示。ｔ檢定及單因子變異數分析統計測驗結果顯示，儘管意義及內容回饋對學生相關句型文法正確性無甚幫助，直接訂正與文法解釋兩者均能提高學生修定稿中第三類假設句型文法使用的正確性，且兩者效果差別不大。文末依研究結果提供教學建議，並就本研究限制提供未來研究之方向。

This study explored the effects of written corrective feedback (WCF) on college students’ English writing. Specifically， it investigated the effects of two types of WCF， direct correction (DC) and metalinguistic feedback (MF) and compared their relative effectiveness on Taiwanese EFL college students’ accurate use of Type III conditionals in picture story writing. Forty-two freshman Applied English majors from a university in central Taiwan were first assigned to a DC， an MF and a Content group based on their scores of the reading section in a high-intermediate level GEPT test. In the first week of the class， they were given as the pretest to complete in fifty minutes a writing task which required them to write a three-paragraph passage with at least three-hundred words based on a four-picture story strip. Two weeks later， all the compositions were returned to the students with feedback from the first researcher—the DC and MF groups with their respective feedback on the incorrect uses and obligatory but nonuse of the target structure while the Content group on meaning content. They were then allowed to read the feedback silently for ten minutes and revised the writing in thirty minutes. The ratios of the accurate uses of the target structure to the obligatory but nonuses and all those grammatical or ungrammatical uses produced by the participants were converted into percentages for further t-test and ANOVA analyses. The results showed that， while feedback on meaning content did not help the participants， DC and MF were effective and comparably effective in improving the accuracy of Type III conditionals. Finally， discussions and pedagogical implications are provided， along with some limitations of the study and suggestions for further studies.