本研究比較下列四種翻譯評量工具的評分結果：根據國立編譯 館「建立國家翻譯人才評鑑標準第二期研究」所提出的量表評分方法 （「忠實」/「通順」各五分量表，獨立評分）、錯誤扣分法，以及兩種 修正的量表評分方法。第一種修正方法是比例不變，但改為合併評分 （「正確」/「表達」各五分量表，合併評分）；第二種修正方法是加重 訊息正確的比例（「訊息準確」六分，「表達風格」四分，合併評 分）。研究者從前述研究中抽取30 份答卷作為評分樣本，共有12 位 翻譯教師/ 專業譯者參與評分。研究結果發現，在英譯中組的部分， 修正後的兩種量表評分法都與錯誤扣分法達到高度相關，但第二期研 究的量表評分法與錯誤扣分法只有中度相關，表示合併評分的三種方 法較為一致；其中「六/ 四評分法」的評分人間信度最高，與錯誤評 分法的相關度也最高，可知為穩定而有效的工具。在中譯英組部分， 「六/ 四評分法」的評分人間信度也是最高，但四種評分法的結果都 達到高度相關，差異不大。

This study compares four assessments used in translation tests: a scale-based method proposed by Liu Minhua et al in “A Study on the Establishment of National Assessment Criteria of Translator and Interpreters, Phase II” (2005), the error-analysis-based method applied by most schools and institutions, and two modified assessments based on Liu’s method. In the present study, twelve graders were invited to re-grade 30 papers in Liu’s experiment by the other three methods. The result of the English-Chinese group showed that the two modified scale methods both reached a high correlation with the erroranalysis method while Liu’s scales only reached a medium correlation. The inter-rater correlation of the 6/4 scale (6 grades for “Accuracy” and 4 grades for “Expression”) was the highest among all the methods used in the research. The correlation between the 6/4 scale and error-analysis method was also the highest. It showed the 6/4 scale method was a reliable and valid assessment tool. In the Chinese-English group, however, the results of the four methods were similar, although the inter-rater correlation of the 6/4 scale was still the highest among the four.