訓練與經驗對英譯中視譯認知歷程影響之研究：眼動與產出之整合分析

視譯長久以來僅被當成課堂學習的輔助工具。然而，近年視譯於實務中的使用機率已有所提升，社區口譯等子領域甚至更將視譯列為必備技能，加上多項研究已證明，視譯確有不同於其它口譯模式的難處，因此視譯實值得深入研究。由於眼動儀可觀測閱讀情形，因此以此儀器研究視譯，可直接了解視譯的認知歷程，有助口譯研究的學科繼續發展。本研究分析17位資深口譯員、18位口譯學生、18位未受訓的雙語使用者，期能了解閱讀目的是否會影響閱讀行為，找出訓練與口譯實務經驗造成的影響，並細部觀察視譯過程中的「前導閱讀」（reading ahead）與停頓行為。結果發現，閱讀目的不同，的確會改變閱讀行為；然而，不同作業之間的閱讀行為仍有部分相似。閱讀第一遍（first pass）時，為了理解而閱讀（默讀）與為了翻譯而閱讀（視譯）的行為非常相似，到了第二遍（second pass）之後才開始出現差異。朗讀在閱讀第一遍時耗費的時間遠多於其它兩者，但進入第二遍之後，與默讀的相似度便大幅提升。整體而言，默讀與朗讀的認知負荷較接近，後者略高，視譯則最為費力。訓練與口譯經驗究竟有何影響？本研究發現，口譯專家的視譯品質最高，再者為口譯學生，最低者為未受訓的雙語者。整體數據（global data）顯示，經過訓練者，視譯總時間與總凝視數量皆顯著少於一般雙語者。不過，譯文總字數、細部閱讀指標（local reading indices）的平均凝視時間，以及閱讀廣度（reading span）皆無組間差異。由此推知，差異應主要來自各組的視譯過程。經過訓練者，畫面一出現，短暫默讀後便會開始視譯，未受訓者絕大多數則會完整將內容讀過，因此一開始的停頓顯著較長。此外，經過訓練，每次開口前的凝視次數也顯著較少，口譯專家更是極端，就算遇到中英語言明顯不同的分枝結構（principal branching direction units），也幾乎不會受阻而出現較長停頓。口譯學生的做法與專家類似，只是轉換技巧較不熟練，偶爾停頓較長、凝視次數較多；相對而言，未受訓的雙語者出現長時間停頓與多次凝視的頻率較高。整體而言，經過訓練，可察覺的停頓數量便會減少，流暢度較高，產出速度也較快。至此，訓練的影響已相當明顯，品質與速度皆會顯著提升。另一方面，經驗影響的似乎主要是正確度，以及開始視譯後的前導閱讀與停頓行為。累積了足夠的經驗之後，專家的遲疑停頓（hesitation pause）與結構停頓（juncture pause）的比例已無顯著差異，其他兩組的遲疑停頓則顯著較多，但三組停頓時，凝視分枝結構的機率都在一半以上。最後，受過訓練者，包括資深口譯員與口譯學生，不管是哪種任務，處理中英分枝結構相異處所耗費的時間與非相異結構並無顯著差異，至於一般雙語者，於默讀任務時已可觀察到，在第一遍閱讀中比較晚期的指標與非首遍閱讀的訊息統整階段，分枝結構相異處顯著較費心力，進入視譯，除了前述兩者之外，分枝結構相異處的單字平均凝視總時間也顯著較長，顯示未受訓者於處理語言分枝結構單位方面較為吃力。

關鍵字

視譯；眼動；認知歷程；前導閱讀；訓練與經驗

並列摘要

Sight translation has long been regarded as nothing more than a pedagogical tool. However, the role of sight translation has become more important in practice, and even necessary for certain fields such as community interpreting. What’s more, with a growing number of studies proving this task to be no easier than other modes of interpreting, we are now granted an opportunity to further develop the discipline of interpreting studies by looking into the cognitive process of sight translation, which can be directly observed with an eye tracker. This study analyzed data of 17 experienced interpreters, 18 interpreting students, and 18 untrained bilinguals, hoping to 1) find out if reading purpose affects reading behavior in different ways, 2) understand the impacts of training and experience, and 3) look more closely at the behavior of reading ahead and pausing during sight translation. The results showed that, different reading purposes did change reading behavior, but there were still similarities between tasks. Reading for comprehension was similar to reading for sight translation in the first pass of reading, while the two started to diverge in the second pass. On the other hand, reading aloud, a task requiring more efforts in the first pass, began to resemble silent reading more than sight translation in the second pass. Generally speaking, the cognitive load imposed on silent reading was similar to (and a little less than) reading aloud, and sight translation was significantly more strenuous. Turning to the impacts of training and experience, it was found that the quality of sight translation was the highest for experts, followed by trainees and then untrained bilinguals. Global data showed less total time and fewer fixations for trained participants than bilinguals. Nevertheless, word counts, mean fixation duration in all local reading indices, and even reading span failed to show any difference. As it turned out, the difference lay in how each group proceeded with the task. Trained participants started sight-translating the text shortly after each trial began, with few fixations of reading ahead, when bilinguals mostly read through the text first, leaving a rather long silence at the outset. In addition, participants with training had significantly fewer fixations before uttering each Chinese character. Experts were extreme in that they rarely got bogged down, even when encountering contrastive linguistic structures. Trainees also manifested a similar tendency, though not as adept at the reformulation skills as experts. Bilinguals, on the opposite, showed much more fixations and longer pauses from time to time. Overall, observable pauses were fewer for those with training, and average verbal gaps were shorter as well, leading to higher fluency and a quicker pace. While the impacts of training were obviously on quality and speed, experience seemed to mainly affect accuracy and the behavior of reading ahead and pausing once sight translation began. Experts, with ample experience, had a non-significant gap between the percentage of hesitation pause and that of juncture pause, whereas the other two groups had significantly more hesitation pauses. Notwithstanding, all groups tended to fixate on principal branching direction (PBD) units at least 50% of the time during pauses. Finally, trained participants (i.e., both experts and interpreting students) did not show any difference between the time spent on processing PBD and non-PBD units in all three tasks. On the other hand, untrained bilinguals already inclined to spend significantly longer time on PBD units during silent reading in a relatively later stage of first-pass processing and for meaning integration in non-first-pass reading. When performing sight translation, the results during silent reading were replicated; what was more, total viewing time on PBD units was also significantly longer, indicating that processing PBD units were more effortful for untrained bilinguals.