國家教育研究院

　　英文論文寫作對多數外語學習者/臺灣學生是很困難的。很多研究已致力於發展文類理論、教材編寫、語言支援網、論文摘要寫作評量。此評量包含一個限時的摘要寫作任務，兩個評分量表（即修辭功能整體文步、語言使用局部構式的量表）。此評量頗受限，僅藉一個摘要寫作任務，粗淺評量論文各章節的寫作能力（即引言、方法、結果、討論）；也僅採用專屬應用語言學研究所的摘要題目對該所學生施測。 儘管多有限制，此評量可被視為一個新鮮的嘗試：將英文論文摘要能力建構成評量工具。本研究擬用混合研究方法（即以質性訪談資料補充量性分數資料），效度化此寫作評量。主要透過調查受試者回應與以下面向的關聯：其本身能力、評分者嚴苛度、量表難度、分數等級難度與量表的交互作用。評分量表會用於碩士生、博士生各30篇的回應。將邀請擁有不同評分資歷的五位評分者。擬蒐集600筆資料，後以FACETS軟體分析多層面Rasch模式。量性分析結果會回應本研究的四個假說：（1）博士生比碩士生能力強。（2）專家評分者比後進者嚴苛。（3）整體文部量表比局部構式量表更難。（4）分數等級與量表間有交互作用（在局部構式量表中，獲得0或1分的機率較低；整體文部量表，4或5分的機率較低）。評分者訪談結果則會顯示評分策略、決策過程。 研究發現指出此評量的分數能否有效顯示英文論文摘要的寫作能力，而可能支持此評量建構效度的結構層面。又此評量的效度化也可能支持論文摘要寫作構念也許可分解為：整體文步、局部構式兩個次構念。此評量效度化可能應用如下：（1）外語學術論文寫作、或外語寫作的理論精進，（2）外語學術論文寫作的工具發展，（3）評分者中介工具效度化的方法應用，即考慮評分者效果、與其他重要因素，（4）透過教學與教材編制來落實外語學術論文寫作能力的教育意涵。更多論文摘要寫作評量的效度化不但能使該評量更有用，也能鼓勵未來研究多探討外語者論文寫作的本質、主要特性、評量發展的方向，而可望為我國英文論文寫作的研究領域帶來實質效益，亦可望漸進協助提升國內研究人才的英文論文寫作能力，而促進其研究產能的提升。

　　In validating the RAWA (i.e., a rater-mediated measure), the study aims to investigate how the examinees’ responses relate to examinee-ability, rater-severity, scale-difficulty, and score-step by scale via a mixed method that includes qualitative post-rating interview data to supplement quantitative score data. The rating scales were applied to 60 responses of 30 master’s and 30 doctoral students. Five raters with varying degrees of expertise were invited. A total of 600 data-points are planned for the Multi-faceted Rasch Model (MFRM) analysis via CONQUEST program. The MFRM results addressed the study purpose by testing 4 hypotheses. First, the doctoral students outperform the master’s. Second, the expert raters demonstrate higher severity than the developing. Third, the global-move scale entails higher difficulty than the local-pattern. Fourth, there is an interaction effect between score-step difficulty and scale (i.e., Scores 0 to 1 in the local-pattern of lower probability, Scores 4 to 5 in the global-move so). The rater-interview results revealed major rating strategies and decision-making process. These findings showed how effective the RAWA score is in manifesting the L2 RA-abstract writing competence, supporting the structural aspect of construct validity of RAWA. The RAWA validation may support the major constructs of the competence that may be decomposed into the global move and local pattern sub-competences. This validation has general implications for: (a) theory advancement in L2 RA-writing or L2-writing, (b) measure development in L2 RA-writing, (c) methodology framework for validating rater-mediated measure by considering rater-effect alongside other key factors, and (d) pedagogical insights into how to effectively develop L2 RA-writing by instruction or material compilation.