過去十年，翻譯學習已成為外文系大學生學習的重點之一，但是許多研究發現教師仍使用傳統的翻譯教學法。在傳統的教室，學生過度依賴以教師為中心的學習，教師本身亦接受或間接鼓勵被動的學習方式。學生只想聽取教師的建議，而鮮少了解自己的翻譯過程、翻譯風格及自我解決問題的方法。 本研究的目的在設計一個翻譯學習活動：「合作式翻譯任務」。此設計經過三個階段：(一) 初步設計；(二) 施實「合作式翻譯任務」的二個原型；(三) 完成合作式翻譯任務的設計。本研究採用設計本位研究方法 (Design-based research method)，並試圖對真實學習情境作深入地了解。在原型(一)共有五個活動：活動一為書面同儕回饋、活動二為組內討論及翻譯者研討會、活動三為口頭同儕回饋、活動四為口頭教師回饋、活動五為最終校正。 研究的參與者為科技大學的外語系學生，在第一循環共有56位學生參加，在第二循環有25位學生參加，另有二位翻譯教師參與此研究。本研究採用三角測量研究法 (triangulation) 來收集資料，包含影片、訪談及學生的翻譯文本等。研究分析的工具為「活動理論」，並從社群、分工、媒介三方面進行分析，以尋求可能解決設計問題的方法。 經過二個原型的實施，本研究設計出「合作式的翻譯任務」。活動一為訓練學生描述及解釋翻譯錯誤的能力。活動二為個人的書面同儕回饋，學生需使用文書處理軟體中的「新增註解」功能給予回饋。活動三為學生研討會及教師研討會。每位學生需記錄自己在研討會的討論結果。教師研討會的時間較短，以便讓與會的學生回學生研討桌分享研討的結果。活動四為學生翻譯員上台分享在學生研討會中得到的回饋及達成的共識。活動五為二個教師給予口頭回饋。活動六為學生使用「新增註解」註明所收集到的回饋、是否接受建議及理由。每組需交出一份校正後的翻譯，及附有註解的檔案。 研究發現學生較相信教師的評語，但是對同儕評語的不信任卻增加學習自主性。學生經分析同儕建議的翻譯、重組或修改後才採用。影響學生互動的因素為問答的溝通模式、同儕間的熟悉度及對同儕回饋的信念。學生的回饋方式傾向於找出有問題的翻譯、提供建議的翻譯及給予讚許。但他們很少給予針對自己的評語作解釋。 本研究提供理論上及實務教學的建議。在理論方面，本研究提出三種理論：領域理論、設計框架、及設計實施方法。在實務教學方法，合作式翻譯任務提供翻譯教師另一種教學模式，以期達到最佳的教學成效。 / For the past decade, translation learning has been one of the main foci for university language learners, but a number of studies have found that many translation teachers still utilize traditional translation teaching methods (Chang, Yu, Li & Peng, 1993; Dai, 2003; Mu, 1992). In traditional classrooms, students tend to depend heavily on teacher-centered instruction, and teachers accept or encourage the students’ passive learning attitudes (Kiraly, 1995). As a result, students only follow the teachers’ suggestions and rarely reflect up their own translating process, translation styles, and problem-solving approaches. The goal of this study was to design a translation learning task called the Cooperative Translation Task (CoTT). It was achieved in three phases: (a) the initial design of the CoTT; (b) the implementations of two prototypes of the CoTT and (c) the finalized CoTT. The current study followed a design-based research (DBR) framework to clarify the complicated interactions in an authentic learning environment. In total, there were five sessions in Prototype I:Session 1: Written Peer Response; Session 2: Within-group Discussion & Translator Seminar; Session 3: Oral Peer Response; Session 4: Oral Teacher Response; and Session 5: Final Revisions. The student participants in both cycles were technological university students, including 56 students in Cycle I and 25 in Cycle II. Two translation teachers participated in the study. For data collection, triangulation data were collected, including videos, interviews, and student documents. The data was put into the framework of Activity theory to diagnose implementation problems in terms of community, division of labor, and mediating artifacts, and innovations with solutions were provided. Following the second prototype, the latest version of the CoTT has been constructed. In Session 1, training in describing and explaining errors is conducted. In Session 2, a peer group gives written responses for the translator group to make revisions. To encourage students to give explanations to their own peers, individual accountability is included. The peer group uses the Comment function in the word-processing software to identify, describe, and explain the agreeable and disagreeable translations. In Session 3, a student seminar and a teacher seminar are conducted simultaneously. To help students take organized notes on the results of their discussions, and to prevent students from not accepting responsibility in the discussions, an individual seminar sheet is given to each student. The teacher seminar finishes earlier than the student seminar so that the members can return to the original seminar and share the teachers’ suggestions with the group. In Session 4, the translator group needs to present the comments from each seminar. In Session 5, the two teachers can use multiple criteria for error analysis. In Session 6, translator members use the Comment function in the word-processing software to insert collected comments, their acceptance level, and the reasons why they accept or reject each suggestion. Each group needs to turn in the final product, one copy with and one without the comments, to the instructor. The present study has found that students have a tendency to trust and use the teachers’ comments. However, this distrust of peers’ review increased students’ autonomy. Students underwent a process of analysis of the suggested translations and reformation of the translation. The influential factors in student-student interaction are an Asking and Answering communication mode and Acquaintance (2A), and students’ values in peer response. When giving a linguistic-level evaluation, students provided the most comments on mechanics, then comments vocabulary and sentences. As for the types of responses, they focused mainly on the identification of translations, provision of suggested translations, and some compliments on agreeable translations. They seldom gave explanations for either agreeable or disagreeable translations. The present study has both its theoretical and practical implications. This design-based study offers three kinds of theories: domain theories, a design framework, and design methodologies. The CoTT and its six sessions provide translation teachers an alternative way to teach, especially for teachers trained in other professions.