陳俊光Jyun-Gwang Chen范雅婷FAN, YA-TING2019-08-282012-8-62019-08-282012http://etds.lib.ntnu.edu.tw/cgi-bin/gs32/gsweb.cgi?o=dstdcdr&s=id=%22GN0697800077%22.&%22.id.&http://rportal.lib.ntnu.edu.tw:80/handle/20.500.12235/86502　　寫作為語言能力中的聽說讀寫四大項能力之一，也為語言課堂中的重要訓練項目。然而，寫作往往為四大能力中最難以掌握的一項，其中篇章層面的偏誤比例尤其甚高，高水平能力的學生作文仍有許多偏誤出現．然而，以往研究學生的篇章偏誤主要關注於各項偏誤的表面現象分類，偏誤來源以及教學應用較少深入探討之。 　　本文以Halliday（1961）的三種篇章功能為基礎，從銜接、信息結構、及主位結構來探討學習者的表現，研究目的為二：1）學習者進行漢語寫作時的各方面偏誤及原因為何？2) 這樣的偏誤現象，是否會隨著學習者漢語水平的提高而有所改變？ 　　本研究以英語母語者為研究對象，蒐集美國威廉大學（Williams College）二、三、四年級學生的作文共89 篇（二年級33 篇，三年級29 篇，四年級27 篇），配合面談，分別從形式和功能的角度來分析偏誤現象。 　　研究結果主要有三：第一，偏誤與語言類型有很大的關係，本文除了從孤立語、屈折語、主要分枝方向、代詞脫落參數、主語顯著、主題顯著等面向探討語際偏誤來源之外，也經由面談討論其他偏誤來源，諸如過度類化與簡化等等．第二，銜接方面，學生在連接詞及指稱方面的偏誤率是最高的，指稱方面下的各類小項偏誤，隨著學習者漢語水平的增加而有明顯的改善。然而，在連接詞的部分，卻是中級組偏誤率高於初級組，產生了U-型學習曲線的現象。第三，信息結構與主位結構方面，主要與語言類型中的新舊訊息順序有關，且與學生的篇章意識不足所致．本文也依據偏誤的類型與比例，提出關於漢語篇章銜接方面的教學應用。Writing, as one of the four skills for acquiring a language, is an important indicator of successful acquisition of that language. It is also the most difficult skill for learners to acquire, especially at the discourse level. Many discourse errors are found in the production of L2 Chinese learners at a high proficiency level. However, previous research on Chinese learners’ writing has focused on the categorizations of the errors. Sources of errors and pedagogical applications are relatively less discussed. Moreover, the comparisons of errors at different L2 Chinese proficiency levels are rarely examined in literature. Halliday (1961) identifies three discourse functions: i.e. ideational, interpersonal, and textual. According to him, the ideational function refers to the function for composing human experience. The interpersonal function is concerned with the discourse aspects of tenor or interactivity. The textual function is related to the internal structures and communicative functions. He further indicates that the textual function is to combine sentences into a discourse because effective interaction depends on extended discourse rather than words or sentences alone. Moreover, three types of textual function are identified: cohesion, information structure, and thematic structure. Based on Halliday’s discourse framework on cohesion, information structure and, thematic structure, this study seeks to address the following two research questions: 1) In English learners’ L2 composition, what types of errors do they commit and what are the causes of these errors? 2) Do these errors vary according to the learners’ L2 Chinese proficiency? A total of 89 L2 Chinese compositions from Williams College were collected for this study. The compositions were further categorized into three proficiency levels. In conjunction with the results of interviews conducted by the researcher, this study investigated the learners’ errors from the perspectives of both forms and functions. There are three major findings: first, linguistic typology has a great influence on the errors. This paper examined not only interlingual errors resulting from different typologies (i.e. isolating and inflecting languages, branching directions, pro-drop parameter, and subject- and topic-prominent languages), but also intralingual errors (i.e. overgeneralization and simplification). Second, in terms of cohesion errors, learners made the most errors in the areas of reference and conjunction. Within the reference errors, the error percentage was mitigated by the English learners’ L2 Chinese proficiency. As for conjunction errors, the errors made by the learners at the intermediate level were more than those at the elementary level, thus following a U-shaped learning curve. Finally, with respect to the discourse structure, the learners’ lack of recognition of the old-new information order and the concept of discourse resulted in their non-target-like ordering in sequencing information structure and thematic structure. Based on the results of the present study, pedagogical implications and suggestions for the further study are provided.偏誤分析銜接信息結構主位結構語言類型error analysiscohesioninformation structurethematic structurelinguistic typology