本文主要探討台灣電視新聞的言談結構(discourse structure)，以及標示各單位的言談標記(discourse marker)。本研究分析25則電視新聞，文類限定於社會新聞。首先，在結構上，每一則新聞包含兩大結構︰導語(news kernel)以及旁白與影片(news body)。兩大結構又可細分為七個較小的單位，分別為︰開場白 (opening)、摘要 (abstract of the news)、事件現場畫面 (event scene presentation)、主要新聞事件 (main news events)、後續發展 (follow-up)、評語 (evaluation)、結尾 (routine ending)。而這七個單位還可再細分為更小的單位。此外，本文所探討的言談標記可分為五類︰1. 指涉詞 (referential forms) 2. 連詞 (connectives) 3. 地方副詞 (locative phrase) 4. 話題轉換填充詞 (topic shift fillers) 5. 畫面轉換 (shot shift)。是故，本文研究重點有二︰1.將新聞結構分為三個階層—Level 1、Level 2、Level 3，並探討出現於不同階層的言談標記在類型與數量上是否反映出階層 (hierarchy)？ 2. 標示各個結構的言談標記為何？ 研究結果指出︰1. 三個階層的言談標記在類型上除了Level 1固定有畫面轉換之外，其餘兩個階層皆無固定的言談標記。另外，在數量上僅Level 1可同時出現多個言談標記，Level 2與Level 3在數量上並無差異，顯示出電視新聞為口說語(spoken language)的一種，訊息與訊息之間的連結性比表現出文體結構階層性更為重要 2. 標示各單位的言談標記並無一致性。受到各單位特性影響，言談標記呈現不同的分佈。 / The present study examines twenty five pieces of broadcast news about crimes and damages in Taiwan. The purpose is to examine the relationship between the discourse structures and their corresponding markers. The discourse structure of a piece of broadcast news is divided into seven components and they are categorized into three levels. Level 1 includes news kernel and news body. Level 2 includes abstract in news kernel, main news events, follow-ups, evaluation, and routine ending in news body. Level 3 contains the smaller units in the Level 2 units. The boundary markers to be examined are divided into four categories: topic shift filler, referential forms, connectives, and shot shift. The present study has two major findings. First, the amounts of markers only show significant difference in Level 1. Down to Level 2 and Level 3, linearity overrides hierarchy. Second, the types of markers are decided by the nature of each unit. The opening is always marked by shot shift and speaker shift plus topic shift fillers/temporals/additives. The abstract section is marked by locative phrase plus referential forms. The event scene presentation section is marked by shot shift and speaker shift plus referential forms. The main news events section is marked by shot shift plus temporals which signal the exact time. Then, the follow-up section is marked by shot shift plus referential forms/connectives. The evaluation section is marked by shot shift plus referential forms. The routine ending section is marked by a relatively longer pause.

本研究旨於調查四個轉折連詞包括on the other hand、in/by contrast、 on the contrary 以及 nevertheless 在美國當代語料庫 (COCA) 以及學習者語料庫(ICNALE) 使用之不同情形及第二語學習者的使用錯誤分析。 本研究從美國當代語料庫 (COCA) 蒐集了四百篇文本(每個轉折連詞各一百篇)以及從學習者語料庫料庫 (ICNALE) 蒐集了一千八百篇學生的寫作，學生的文章分別來自台灣、香港、新加坡、菲律賓的大學生以及母語人士。本研究假設為:雖然四個轉折連詞被歸類為同一類轉折連詞 (adversative type)，各個轉折連詞的語義及用法應有所不同。本研究語料分析主要分為兩部分，第一部分分析各個轉折連詞中的篇章關係 (discourse relation)，調查四個轉折連詞上下文的不同反義的語義情況。第二部分聚焦於轉折連詞上下文中的主題分布 (topic categorization)，旨於發現是否特定轉折連詞出現於特定主題之內。 研究結果經由語料分析發現，四個轉折連詞有特定傾向的篇章關係和主題分布。在學習者語料庫中，本研究發現第二語言學習者比起母語人士，使用更多轉折連詞於寫作中，特別是on the other hand 以及nevertheless。台灣學生以及英語為第二外語學生在上下中，常將on the other hand的篇章關係使用為in addition 之用法。在連詞nevertheless方面，雖然學生使用頻率較高，但多為誤用情況，顯示出學生對於該轉折連詞的不熟悉情況。而in/by contrast 和on the contrary 皆不常出現於第二語言學習者和母語人士的寫作中。本研究最後提出在英語教學上的建議：在課堂上教導轉折連接詞時，應需加強語意以及上下文方面的探討，並訓練學生正確文句之間的邏輯關係。 / The present study adopts corpus analysis to investigate four adversative connectives (ACs), including on the other hand, in/by contrast, on the contrary, and nevertheless in the native corpus, the Corpus of Contemporary American English (COCA), and in the leaner corpus, the International Corpus Network of Asian Learners of English, (ICNALE). The comparison of the two corpora and the common misuse by L2 learners are also presented. The data comprise of 400 texts in COCA (100 texts for each AC) and 1800 essays written by students in Taiwan, Hong Kong, Singapore, the Philippines, and by native speakers in ICNALE. The research hypothesis is that although the four ACs are classified in the same ‘adversative’ type, their meanings and uses are different from one another. The adoption of corpus helps to find the patterns of each AC. The data are analyzed from two perspectives: one is the discourse relation while the other is topic categorization. The discourse relation refers to the relations of the preceding and following co-texts of the AC. The topic categorization, on the other hand, aims at finding if there are specific topics that are frequently occurred with certain ACs and if these topics are consistent before or after the occurrences of the Cs. The findings confirm the prediction that by analyzing the data in COCA, each AC exists in certain discourse relations and topics. The present study finds that different from native speakers, L2 learners often use on the other hand as ‘addition’ yet a great number of uses of nevertheless are viewed as misuse. As for on the contrary and in/by contrast, both native speakers and L2 learners seldom apply the ACs in their writing. The present study also provides the common misuse patterns in ESL/EFL students’ writing. It is hoped that the findings have implications for teachers and learners to be aware of the differences of ACs.