本文是對職業譯員英中無稿同步口譯停頓時長特徵的個案研究。研究者選取美國總統川普（Donald Trump）就職演講以及三家知名媒體的職業譯員的英中直播同步口譯為個案語料，建立了小型雙語多模態平行語料庫，並以0.250 秒為停頓時長閾值對於演講原文和同步口譯譯文中的所有停頓進行了標注。原文和譯文停頓時長平均值比較t 檢驗顯示，譯文整體和譯文1-3 停頓時長均顯著大於原文。原文和譯文停頓時長區間分佈比較卡方檢驗顯示，原文與譯文整體和譯文1-3 停頓時長區間分佈存在顯著差異。由量化研究數據可以推斷，譯文中的長停頓是導致譯文停頓時長顯著大於原文的原因。通過在文本和音訊兩個模態下對原文和譯文語料進行質性分析發現，長停頓形成時的譯員認知可能處於困境和平順兩種狀態。譯員應對認知困境的某些策略會導致長停頓的出現，包括對原文訊息的省略、合成推斷、分批次產出和順序重組等。平順認知狀態下生成的長停頓分為策略和非策略兩種。譯員不得不等待更多訊息輸入方能開口傳譯時，生成非策略長停頓。策略性長停頓則是譯員採取的策略性等待，具體策略包括省略次要或重複訊息、延長聽說時差等。

This is a case study of pause duration characteristics of professional interpreters’ English-Chinese simultaneous interpreting without text. Live EC simultaneous interpreting from three renowned media of Donald Trump’s presidential inaugural speech is selected as the case, based on which a small-size bilingual parallel corpus with dual modalities is built. Pauses in the corpus are marked with 0.250s as the threshold duration value. T test shows that pause duration in the interpretation is significantly bigger than that in the original speech. Chi square test shows that distribution of pause duration in each range differs significantly between original speech and interpretation. It can be induced from quantitative data that the long pause in the interpretation is the major contributor to significant duration difference. Corpus analysis under two modalities finds that long pauses may occur in smooth cognitive conditions as well as under cognitive challenges. Coping strategies against cognitive challenges may make long pauses happen, including the omission, inferring, production in batches or sequence change of information from the original speech. There are both strategic and non-strategic long pauses in smooth cognitive conditions. The non-strategic ones are produced when the interpreter has to wait for more information input before s/he can speak. The strategic ones are more about strategic waiting often caused by use of strategies like omission of less important or redundant information as well as ear-voice-span prolonging.