英中逐步口譯筆記原則之眼動研究：學生口譯員與專業口譯員之別

全文

(1)國立台灣師範大學翻譯研究所碩士論文 Thesis Advisor: Dr. Ming-Li Ju 指導教授：汝明麗. An Eye Tracking Study of English-Chinese Consecutive Interpreting Note-taking Practices: Trainees vs. Professional Interpreters 英中逐步口譯筆記原則之眼動研究：學生口譯員與專業口譯員之別. by. Chia-Hsuan Donna Hu 研究生：胡家璇. A thesis submitted in conformity with the requirements for the degree of Master of Arts Graduate Institute of Translation and Interpretation National Taiwan Normal University. January 2014 中華民國一○三年一月.

(2) Abstract Notes serve as a supplement to the interpreter’s memory in consecutive interpreting. The content, layout, form, and language used in notes can be critical in determining the quality of interpreting production and help convey the original message more completely. Ma’s study (2013) demonstrated the validity of the two note-taking principles introduced in most consecutive-interpreting related literature: a vertical layout is preferable to a horizontal one, and notes are preferably taken in the target language than in the source language, by using an eye tracker to record student interpreters’ eye movements while they read the notes written horizontally/vertically and in the target/source language, which were designed as controlled variables in the experiment. This study aims to extend his findings by examining whether layout and language choice of notes would result in different eye movement patterns of note-reading on professional interpreters. The results show that vertical layout and the use of target language are in fact valid recommendations supported by empirical evidence and the benefits of vertical layout outweigh target language. Nevertheless, professional interpreters appear to be less sensitive to different layouts and languages, which can be attributed to their low reliance on notes during listening and interpreting. On the whole, professional interpreters are able to read the notes more rapidly and efficiently than trainee interpreters by adopting a more skillful approach. Based on these findings, interpreting instructors are advised to provide trainee interpreters with ample opportunities to practice these note-taking principles, vertical layout in particular, to the level of automaticity before graduation. Meanwhile, interpreting instructors should emphasize that comprehending and analyzing the source speech always takes precedence to acquiring any specific note-taking techniques.. Keywords: eye-tracking, consecutive interpreting, notes, expertise level. i.

(3) 摘要逐步口譯的筆記能輔助口譯員的記憶，其內容、格式、語言對於翻譯產出的品質與準確度至關重要。前人研究﹙馬思揚，2013﹚針對「格式」與「語言」兩個變項加以操弄，設計出逐步口譯筆記，並請學生口譯員擔任受試者，於英進中逐步口譯實驗中記錄受試者閱讀筆記的眼動軌跡，證實「直寫格式」與「用標的語」確實是有用的筆記原則，有助於降低口譯員在閱讀筆記與產出時的認知負擔。本研究的目的為延伸前述研究成果，邀請專業口譯員擔任受試者，以探討不同的筆記格式和語言是否對不同專技程度口譯員的眼動軌跡造成不同影響。研究結果顯示，「直寫格式」與「用標的語」亦有助於降低專業口譯員在閱讀筆記與口譯時的認知負擔，而且｢直寫｣相對於｢標的語｣對口譯員閱讀筆記和產出譯語的幫助更大。另外，專業口譯員不論在聽原文或翻譯時，對筆記的依賴性較學生口譯員低，對不同格式與語言的敏感度也因此較低。同時，專業口譯員閱讀筆記的技巧較學生口譯員嫻熟，速度較快且效率較高。根據本研究結果，我們認為口譯教師應在學生修業階段提供足夠的機會，讓學生不斷練習筆記技巧直到習慣成自然，尤其是｢直寫格式｣。但同時口譯教師也應強調，比起磨練筆記技巧，理解並分析原文的能力更為重要。. 關鍵詞：眼動、逐步口譯、筆記、專技程度. ii.

(4) Acknowledgements First and foremost, I would like to express my deepest gratitude to my thesis advisor, Dr. Ming-Li Ju, for taking me on to participate in the eye-tracking study. She patiently guided me, who previously had no knowledge about research in this field, toward the completion of the thesis by providing technical resources, advice, encouragement, and financial support. She shared her network of professional interpreters with me so that I was able to finish all the experiments in time. She detailedly read through my thesis several times. Every time after our discussion, I learned more about how to think academically, logically, and critically. Without her, I would not have been able to finish the thesis in one year. Secondly, I thank my thesis committee members: Dr. Tze-Wei Chen and Dr. Jie-Li Tsai, who have always been there for me. Whenever I or any of the research team members encountered research problems, Dr. Chen would coordinate a meeting for us with Dr. Tsai and we knew that Dr. Tsai could always come to our rescue. I am grateful for Dr. Tsai’s scientific expertise and criticism. Also my special thanks go to Mr. Chia-Hsing Chen, an intelligent PhD student in Dr. Tsai’s lab, who is a patient mentor to me. He spent so much time teaching me how to operate an eye tracker and how to analyze quantitative data. Without his technical support and company, completing this thesis would undoubtedly be a mission impossible. Thirdly, I am greatly indebted to all the professional interpreters who participated in my experiment. Not only did I feel honored to meet them in person, I also realized that I learned the most from their sharing during the after-experiment interviews. Some of them helped me with advertisement so the participant recruiting process went rather smoothly. Some of them even came from Kaohsiung. Nearly all of them expressed high interests in my research and expected to see the results. I really want to thank them for being my morale booster during the thesis-writing process. In addition, I must say “thank you” to my friends and family. Valeria Chen, in particular, was the one who went through all the ups and downs with me in the past year. Stephen Ma, the pioneer of this research project, was the one I turned to when I needed help. I also feel thankful for receiving encouragement and inspiration from those classmates who were writing their theses at the same time. Last but not least, I must thank my family. Their 100% support has made my dream come true. iii.

(5) Table of Contents Abstract Acknowledgements List of Tables List of Figures Chapter 1 – Introduction 1.1 Research Background 1.2 Research Question 1.3 Research Limitations 1.4. Thesis Organization. i iii iv vii 1 5 6 7. Chapter 2 – Literature Review 2.1. 2.2. 2.3. Consecutive Interpreting 2.1.1 Types of Interpreting 2.1.2 Consecutive Interpreting Models 2.1.3 Notes in Consecutive Interpreting 2.1.3.1 Layout 2.1.3.2 Language 2.1.4 Summary Eye Tracking. 8 8 9 11 13 13 16 16. 2.2.1 2.2.2 2.2.3. 16 18 20 21 22 23 25 26 26 28 31. Basic Characteristics of Eye Movements Eye Movement Measures Eye Movements in Reading 2.2.3.1 English Reading 2.2.3.2 Chinese Reading 2.2.4 Eye Movements in Interpreting 2.2.5 Summary Skill Acquisition and Expertise Development 2.3.1 From Novice to Expert 2.3.2 Expertise Development and Interpreting 2.3.3 2.3.4. Expertise Level and Consecutive Interpreting Notes Summary. 33.

(6) Chapter 3 – Method 3.1 3.2 3.3 3.4. 3.5 3.6. Research Hypotheses Subjects Experimental Design Materials 3.4.1 Speeches 3.4.2 Notes 3.4.3 Comprehension Questions 3.4.4 Questionnaire Apparatus Procedures. 3.7. Data Analysis. 35 35 36 36 36 37 37 37 38 38 39. Chapter 4 – Results 4.1. Eye Movement Data 4.1.1 Global Analysis 4.1.1.1 Mean Fixation Duration 4.1.1.2 Number of Fixations 4.1.1.3 Mean Saccade Length 4.1.1.4 Mean Regression Rate 4.1.1.5 Total Reading Time 4.1.2 Local (ROIs) Analysis 4.1.2.1. 4.2. First Fixation Duration, Single Fixation Duration, and Gaze Duration 4.1.2.2 Rereading Time and Total View Time 4.1.2.3 Regression Rate 4.1.3 Local (Targets) Analysis 4.1.3.1 First Fixation Duration, Single Fixation Duration, and Gaze Duration 4.1.3.2 Rereading Time and Total View Time 4.1.4 Summary Questionnaire Data. 43 43 43 45 46 48 50 52. 4.2.1 Subject Background 4.2.2 Subject’s Perception of Notes 4.2.3 Subject’s Perception of the Experiment. 52 55 58 60 60 63 65 68 68 68 70.

(7) Chapter 5 – Discussion 5.1 5.2 5.3 5.4 5.5. Difference between professional interpreters and trainee Interpreters Effect of layout on interpreters with different levels of expertise Effect of language on interpreters with different levels of expertise Interaction between language and layout Note-taking practices of interpreters with different levels of expertise and the implication for consecutive interpreting training. 74 77 80 83 84. Chapter 6 – Conclusion 6.1 Major Findings 6.2 Suggestions for Future Research. 88 90. References. 92. Appendix A. Consent Form Appendix B. Speeches and Notes Appendix C. Comprehension Questions Appendix D. Questionnaire. 95 96 110 111.

(8) List of Tables Table 1.. The range of mean fixation durations and the mean saccade length is silent reading, oral reading, scene perception, and visual search. 17. Table 2.. Definitions of the more commonly used word-based spatial eye movement measures. 19. Table 3.. Definitions of the more commonly used word-based temporal eye movement parameters. 19. Table 4.. Speech materials. 36. Table 5. Eye movement measures. 41. Table 6.. Mean fixation duration (ms) in global analysis, with standard deviations in parentheses. 43. Table 7.. LMM statistics for log(Mean Fixation Duration) in global analysis. 44. Table 8.. Number of fixations (counts) in global analysis, with standard. 45. deviations in parentheses Table 9.. LMM statistics for log(Number of Fixations) in global analysis. 45. Table 10. Mean saccade length (pixels) in global analysis, with standard deviations in parentheses. 47. Table 11. LMM statistics for log(Mean Saccade Length) in global analysis. 47. Table 12. Mean Regression Rate (%) in global analysis, with standard. 48. deviations in parentheses Table 13. LMM statistics for log(Mean Regression Rate) in global analysis. 49. Table 14. Total reading time (s) in global analysis, with standard deviations in parentheses. 50. iv. 61 62 64.

(9) Table 15. LMM statistics for log(Total Reading Time) in global analysis. 51. Table 16. First fixation duration, single fixation duration, and gaze duration (ms) in local (ROIs) analysis, with standard deviations in parentheses. 52. Table 17. LMM statistics for log(First Fixation Duration), log(Single Fixation Duration), and log(Gaze Duration) in local (ROIs) analysis. 53. Table 18. Rereading Time and Total Viewing Time (ms) in local (ROIs). 55. analysis, with standard deviations in parentheses Table 19. LMM statistics for log(Rereading Time) and log(Total Viewing Time). 56. in local (ROIs) analysis Table 20. Regression Rate (%) in local (ROIs) analysis, with standard deviations in parentheses. 58. Table 21. LMM statistics for log(Regression Rate) in local (ROIs) analysis. 59. Table 22. First fixation duration, single fixation duration, and gaze. 60. duration (ms) in local (Targets) analysis, with standard deviations in parentheses Table 23. LMM statistics for log(First Fixation Duration), log(Single Fixation Duration), and log(Gaze Duration) in local (Targets) analysis. 61. Table 24. Rereading Time and Total Viewing Time (ms) in local (Targets) analysis, with standard deviations in parentheses. 63. Table 25. LMM statistics for log(Rereading Time) and log(Total Viewing Time) in local (Targets) analysis. 64. Table 26. A summary of the results from eye movement data. 66 49. v. 61 62.

(10) Table 27. Perception of the purpose(s) of notes. 69. Table 28. Perception of the layout of notes. 69. Table 29. Perception of the language of notes. 70. Table 30. Rating of speech content. 71. Table 31. ANOVA statistics for speech ratings by the two subject groups. 71. Table 32. Rating of notes. 72. Table 33. ANOVA statistics for note ratings by the two subject groups. 72. vi.

(11) List of Figures Figure 1.. The relationship between the Notes, the Speech, and the Interpretation. 11. Figure 2.. A flowchart of the processes in the translation task. 24. Figure 3. Regions of Interest (ROIs). 40. Figure 4.. Mean Fixation Duration (ms) in global analysis. 44. Figure 5.. Number of fixations (counts) in global analysis. 46. Figure 6.. Mean Saccade Length (pixels) in global analysis. 48. Figure 7.. Mean Regression Rate in global analysis. 49. Figure 8.. Total Reading Time (ms) in global analysis. 51. Figure 9.. Single Fixation Duration (ms) in local (ROIs) analysis. 54. Figure 10.. Gaze Duration (ms) in local (ROIs) analysis. 54. Figure 11.. Rereading Time (ms) in local (ROIs) analysis. 57. Figure 12.. Total Viewing Time (ms) in local (ROIs) analysis. 57. Figure 13.. Regression Rate in local (ROIs) analysis. 59. Figure 14.. First Fixation Duration (ms) in local (Targets) analysis. 62. Figure 15.. Single Fixation Duration (ms) in local (Targets) analysis. 62. Figure 16.. Gaze Duration (ms) in local (Targets) analysis. 63. Figure 17.. Total Viewing Time (ms) in local (Targets) analysis. 65. Figure 18. Ratings of the four speeches. 71. Figure 19. Ratings of the notes. 73. vii.

(12) Chapter 1 INTRODUCTION. 1.1 Research Background Interpreting, simply put, is oral translation. An interpreter listens to a speaker speaking one language, the source language (SL), and produces oral translation in another language, the target language (TL). Despite the fact that machine translation, such as Google translate and many other translation tools, seems readily available nowadays, it is still far from ideal to meet the demand of bridging the linguistic and cultural gaps. The main reason is interpreting, or translation in a broader sense, is by no means word-to-word conversion. It requires an interpreter, or a translator, to comprehend the source text, analyze its meaning, reformulate it, and reproduce it in a contextually and linguistically appropriate way in the target text. Translators usually have the luxury of time which allows them to contemplate the meaning of each word in the source text and come out with the best suited translation. Interpreters, unfortunately, almost always work under time pressure. Auditory input disappears immediately after it is delivered. People who need to get their messages across, namely the speaker and the audience, expect to hear the real-time translation so that they can communicate. In other words, when an interpreter is as work, he is constantly multitasking: listening, comprehending, analyzing, memorizing, language converting, maybe writing and reading, and of course, speaking. Needless to say, interpreting is a task with high levels of cognitive demand. Interpreting can be classified into two major modes: simultaneous interpreting (SI) and consecutive interpreting (CI). SI is performed when an interpreter listens to the speaker’s speech in the SL and orally produces the speech in the TL almost at the same time. Since the different cognitive functions mentioned in the previous paragraph are at all times overlapping while an interpreter is working in this mode, SI is usually considered the hardest type of interpreting. CI, on the other hand, faces less time constraint than SI, yet it adds cognitive burden on an interpreter in a different manner. CI is used in a situation where an interpreter first listens to a speech segment in the SL, while in the meantime committing the messages to his short-term memory, and taking down notes, and then reproduces the messages in the TL with the help of both memory and notes as soon as the speaker pauses for translation. Depending on the length of the speech segment, the interpreter may or may not need to take notes. Generally speaking, if a part of speech is shorter than one minute long, an interpreter is usually able to resort to memory to translate the 1.

(13) whole segment. However, if a segment is about two to three minutes long or even longer, notes become necessary to ensure the accuracy and completeness of an interpreter’s production. That is to say, in comparison to SI, memory plays a more vital role in CI and it is also one of the main cognitive demands that stress an interpreter. To relieve the interpreter’s cognitive burden while doing CI, notes are the best tool that can come to his rescue. Indeed, notes act as a critical link bridging an interpreter’s reception of the source message and his production of the target message. They are a supplement, visual cues, or an aid, to a successful CI performance. A successful CI performance heavily depends on successful information retrieval from the notes, that is, notes are “deciphered” in a way that memory cues are efficiently and accurately retrieved so as to facilitate appropriate rendition. Successful information retrieval from the notes in turn relies on what and how the notes are jotted down. In other words, the content, layout, form, and language used in notes can be critical in determining the final interpreting outcome. In the field of interpreting, certain note-taking principles or guidelines exist and are generally believed to best aid an interpreter’s memory. The most commonly heard are “verticalization” and “target language”. “Verticalization” means to arrange the note elements from top down as opposed to normal writing from left to right line by line. This principle is recommended by many experienced interpreters as it is said to achieve a telling, two-dimensional presentation that helps to highlight the structure underlying a speech and its semantic orientation. Some others claim that notes in a vertical layout fall easily in an interpreter’s visual area, allowing him to view as much information as possible in one glimpse (Ilg & Lambert, 1996; Liu, 2008; Rozan, 1956, cited in Ilg & Lambert, 1996). Though the “verticalization” principle is often introduced and taught by instructors to students in interpreting schools early in their training programs, surprisingly there have been scant empirical studies to back it up. “Taking notes in the target language” is another principle asserted by some interpreting scholars. They believe that taking notes in the TL forces the interpreter to move away from the surface form of the incoming speech and facilitates better production of the target speech (Déjean Le Féal, 1981; Herbert, 1952; Laplace, 1990; Mikkelson, 1983; Seleskovitch, 1975, all cited in Dam, 2004). However, many other researchers hold the opposite view, asserting that taking notes in the SL is better because it does not add the language conversion effort on the interpreter when he is already struggling with concurrent operations of listening, analyzing, comprehending, 2.

(14) memorizing, and writing (Alexieva, 1994; Ilg, 1988, cited in Dam, 2004; Kirchhoff, 1979, cited in Dam, 2004). In fact, language choice in note-taking is more complicated than this dichotomy. In addition to task-related status of the language (SL vs. TL), the interpreter’s linguistic combination (native vs. foreign), the nature of the interpreter’s language pair (i.e. morphologically long vs. non-morphologically long language), and relative expertise of the interpreter have all been found to affect his choice of language during CI note-taking (Dam, 2004; Dam et al., 2005; Li, 2012; Szabó, 2006). As mentioned earlier, note-taking principles are taught to trainee interpreters in interpreting schools. However, they are not necessarily followed. Interpreting instructors, who are usually practicing interpreters themselves, understand that everyone differs in their ways of comprehending, analyzing, and grasping the main idea from the same information, so do their writing habits. Therefore, after informing students certain guidelines, interpreting instructors often encourage trainee interpreters to develop their own note-taking styles as they develop their interpreting skills. It is also common among expert interpreters that one’s notes are completely illegible to another. It seems that although trainee interpreters were once “normalized” during training in terms of their note-taking practices, eventually notes become very personal and individualized systems that only the author can recognize. If that is the real situation in the interpreting profession, an interesting question to ask is how important note-taking principles, such as “verticalization” and “target language”, are to skilled interpreters. The comparison between skilled and less skilled interpreters falls into the scope of expertise research which aims to discover what characteristics, both qualitative and quantitative ones, set experts and novices apart. Expertise research in interpreting can be dated back decades ago. Findings in this area have shown that expert-novice differences lie in strategies (Moser-Mercer, 1997), knowledge base (Moser-Mercer et al., 2000), memory performance (Liu et al., 2004; Kopke & Nespoulous, 2006), and interpreting performance (Barik, 1975; Dillinger, 1994; Liu, 2008; McDonald & Carpenter, 1981). In general novice interpreters undergo a reliable progression of five levels (Dreyfus, 2004; Ju, 2010) or three stages (Moser-Mercer et al., 2000) to achieve expertise. At first novices learn the relevant facts about interpreting. Their understanding to context-free rules guides their actions. Once they gain more experience, they start to perceive recurrent, meaningful patterns from real situations in which they learn from their errors, consider alternatives, discuss solutions, strengthen connections of elements, and experiment with procedures. They increasingly obtain a sense of mastery and the ability to cope with and manage a 3.

(15) variety of situations. Ultimately novices arrive at the autonomous stage, where they perceive situations as wholes and flexibly modify their reactions accordingly. When novices become experts, they no longer consciously rely on analytic principles (rules and guidelines) to connect their understanding of the situation to an appropriate action. Their skills are not conscious, effortful, deliberate and linear anymore, but automatic and intuitive, as manifested in the increased speed and accuracy in their performances. Note-taking is a skill that needs to be learned, practiced, and eventually mastered. There should be no doubt that developing such techniques also follows the levels or stages mentioned above. Unfortunately due to the inconsistency of note-taking systems among interpreters, so far only very limited literature have reported inconclusive results from the examination of notes taken by interpreters with various levels of expertise (Abuín González, 2012; Dai & Xu, 2007a, 2007b). Among the past studies in CI note-taking, Ma’s research (2013) is no doubt a stand-out. In order to tackle the issue of diverse note-taking systems, he used the designed notes written horizontally/vertically and in the target/source language as controlled variables in an English-Chinese CI experiment. The subjects consisted of trainee interpreters who were studying in two-year translation and interpretation programs in Taiwan and had received CI training for at least six months to one and a half years. In his experiment, the subjects did not have to take notes while listening to the source speeches but were offered the chance to look at the designed notes. After the speeches ended, they were asked to interpret by reading the designed notes, which they had had a chance to familiarize themselves with, as a memory trigger. Despite the fact that such experimental design deviates from a real CI setting in which interpreters take notes by themselves and interpret by reading their own notes, designing the layout and language of notes as controlled variables is surely an innovative way to circumvent the issue of diverse note-taking systems. Another reason that makes his study a pioneer in CI research is the employment of the eye-tracking method. The eye-tracking method, which has been developing and advancing since the nineteenth century, is a non-invasive and easy-to-use equipment for researchers in multiple fields to study human cognitive processes in a wide variety of tasks, including language reading (Just & Carpenter, 1980; Miles & Shen, 1925; Rayner et al., 2006; Rayner, 2009; Shen, 1927; Sun et al., 1985; Tsai et al., 2000; Tsai et al., 2005 ), music reading, human activity recognition, the perception of advertising, and the playing of sports. Eye-tracking data are able to shed light on the behaviors between 4.

(16) different groups of people, such as the differences between men and women while looking at the same advertisement, and the different performance between novices and experts in sports. Basically any task that involves the comprehension of visualizations can be investigated with an eye-tracker, which is the rationale behind applying the eye-tracking method to CI research in note-taking practices since notes serve as a kind of visual stimulus. In Ma’s experiment, the subjects’ eye movements were recorded while they read the designed notes and interpreted. He then applied the eye movement measures in reading research to analyze and interpret his data. His findings showed that notes in the vertical layout and in the TL work to the interpreter’s advantage because they reduce the interpreters’ cognitive loads during the note-reading/interpreting phase. He therefore concluded that “verticalization” and “target language” are valid principles for note-taking. His study looks at the effectiveness of notes from the angle of the note-reading phase instead of the note-taking phase, which is a perspective rarely adopted by researchers. Moreover, his results provide direct evidence to uphold the verticalization principle as vertical notes, according to Ma’s experiment, are apparently easier to read and thus make them better memory triggers than horizontal notes. On the other hand, his results lend support for the TL principle because TL lessens interpreters’ cognitive burdens during note-reading. However, whether it actually increases cognitive loads during note-taking, as claimed by some researchers, remains unknown since the experiment focused on the note-reading stage rather than the note-taking stage. Ma’s research laid a concrete foundation for the present study. This study recruited professional interpreters who had been practicing interpreting in the market for at least two years with a minimum of 150 work days. They had Chinese as their native language, same as the trainees in Ma’s research. After the professional interpreters’ eye movements were recorded by an eye-tracker, their data were pooled and analyzed with the trainees’ in order to uncover any differences between the two subject groups as they interpreted and read the notes that had been manipulated in terms of layout and language. 1.2 Research Question The purpose of the present study is to examine whether professional interpreters demonstrate different eye movement behaviors from those of trainee interpreters as they read notes in different layouts and languages. 5.

(17) 1.3 Research Limitations This study has some limitations that should be noted. First, although the experiment materials were chosen from speeches which had been presented in real conferences, the experiment was conducted in a lab where there were neither speakers nor audience to simulate a real conference. When an interpreter interprets in an authentic CI setting, he usually has to switch his attention between the notes and the audience so that the audience could feel that the messages from the speaker are indeed conveyed to them through the interpreter. However since the eye tracker requires the subjects to maintain their heads still in front of the screen so their eye movements can be accurately recorded, the subjects had to keep looking at the notes or the screen even when not necessary. Second, the participants did not write down notes by themselves but looked at the notes designed by the experimenters while listening to the source speeches and interpreting. In reality, every interpreter has a distinct style of note-taking. By listening to the same speech, each interpreter grasps and reorganizes the information in different ways and jots down the notes using different words, abbreviations, and symbols, as well as arranging them and deciphering them in different manners. As a result, it is normal that an interpreter finds another interpreter’s notes illegible and incomprehensible. Yet in this study, the layout and language of notes are the controlled variables. Therefore, in order to compensate for the limitation that the subjects were not allowed to take and read their own notes, they were given the time to look at the designed notes once while listening to the source speeches in order to familiarize themselves with the notes written by someone else. This may be the biggest deviation from typical interpreting practice. Third, the participants were not told of the topics of the materials and were not given time to prepare. In a real situation, an interpreter is usually notified in advance of the topic of a conference he is about to interpret for so that he can look up related information and prepare. Since in this study, all the four speeches chosen were on general topics and had excluded any technical terminology, it was assumed that all the subjects would be able to handle the experimental materials at this kind of difficulty level without prior preparation. Fourth, this study focused on the eye movements of the subjects. Several eye movement measures were used to reveal the effect of the designed notes on the subjects’ cognitive processes. Since this is still considered an exploratory study in the field of interpreting, the subjects’ oral production was not analyzed. In other words, the performance of trainees and professionals is beyond the scope of this study. 6.

(18) 1.4 Thesis Organization In this thesis, Chapter 2 reviews relevant literature on interpreting, specifically CI, the eye-tracking method and its application in reading and interpreting research, as well as expertise research in general and in interpreting. Chapter 3 describes the experiment design and procedures and how the eye movement data were analyzed. Chapter 4 displays research results on a measure basis and in the end provides a summary. Some quantitative data collected from the questionnaires are also included. Chapter 5 discusses the findings as a whole and provides interpretations based on the existing literature as well as the qualitative data gathered from the interviews. Finally Chapter 6 concludes the thesis with a summary of the main findings, implications of the present study, and offers suggestions for future research projects.. 7.

(19) Chapter 2 LITERATURE REVIEW. This chapter reviews the existing literature on consecutive interpreting (CI), eye tracking, and expertise development. First the process of CI is introduced and explained by using two most recognized models. Then the focus is put on CI notes, in particular, the two commonly preached principles regarding notes’ layout and language use. Next the rationale of using the eye tracking method in this study is elaborated. Since the present study looks into the note-reading phase in CI, reading research, especially concerning reading English and Chinese, which employs the eye tracking method is presented in detail. Interpreting studies that use eye trackers as data collecting tools are also included. Finally expertise research which examines the development of skill acquisition and the quantitative and qualitative differences between experts and novices are drawn on, with the attention especially paid to interpreting. Studies that investigate the relationship between level of expertise and CI note-taking are provided at the end. 2.1 Consecutive Interpreting 2.1.1 Types of Interpreting Interpreting, unlike translation, which is a task converting the language of a written text from one to another, is a practice that verbally conveys an auditory message from one language to another. Generally speaking, interpreting can be categorized into two major modes: simultaneous interpreting (SI) and consecutive interpreting (CI). In a SI setting, an interpreter listens to the speaker speaking in the source language (SL) and orally produces the speech in the target language (TL) while at the same time still listening to the speaker’s continuously incoming messages. In other words, the speaker speaks continuously and does not pause for the interpreter to render the oral translation. The interpreter simultaneously comprehends the input and stores segments of it in memory. In a CI setting, an interpreter listens to the speaker speaking in the SL, commits the messages to short-term memory, and takes notes of main ideas. After a segment of the speech is delivered, the speaker stops for the interpreter to recall the messages by reading the notes and to re-express the messages in the TL. Depending on the length of a speech segment, the interpreter may or may not need to take notes. If a speech segment is shorter than one minute long, an interpreter is usually able to resort to memory to translate the whole thing. 8.

(20) However, if a segment is about two to three minutes long or even longer, notes become necessary and critical to ensure the accuracy and completeness of an interpreter’s production. In addition to the above two major modes, sight translation (ST) is a special form of interpreting as the interpreter’s linguistic input is written rather than auditory. During the process of ST, the interpreter reads a text in the SL while rendering the oral translation in the TL. Whichever interpreting task, an interpreter performs a series of interrelated cognitive and linguistic operations in differentiated phases.CI, for instance, can be dissected into several phases that are clearly distinguishable yet correlated (Liu, 2008): Phase 1 Phase 2 Phase 3. Listening Analyzing and Comprehending Memorizing and/or Note-taking. -----------------------------------------------------------------------------------Phase 4 Memory-retrieving and/or Note-reading Phase 5 Interpreting Phase 1, 2, and 3 are carried out at the same time followed by phase 4 and 5, which also occur concurrently. In other words, when an interpreter is doing CI, he is always multitasking. 2.1.2 Consecutive Interpreting Models A number of descriptive models have been developed to explain the process of CI. This section will discuss in detail two better recognized models: the Effort Model and the strategic information processing model. The Effort Model (Gile, 2009) focuses on the cognitive dimension of CI; and the strategic information processing model (Kohn & Kalina, 1996) addresses its procedural aspects. The Effort Model for CI is articulated in two parts: a. Listening Phase: Listening = L + M + P + C L Listening and analysis or understanding Effort M Short-term Memory Effort P Notes Production Effort C Coordination Effort, the allocation of attention between the three Efforts. 9.

(21) b. Reformulation Phase: Reformulation = Read + Remembering + P + C Read Note-reading Effort Remembering Retrieval of information about the speech heard from long-term memory P Production of the speech in the TL C Coordination Effort Gile uses the word “Effort” to imply that performing any of the tasks above consumes an interpreter’s mental energy, or cognitive resource, which is only available in limited supply. Whether an interpreter can produce a satisfying interpreting outcome depends on whether his processing capacity exceeds the Efforts required. The strategic information processing model, on the other hand, uses discourse-based mental modeling to distinguish two separate processing phases in CI, namely discourse reception and discourse production. During the reception phase, interpreters mainly listen to the source discourse, breaking it down into coherent chunks, and using their linguistic knowledge in interaction with their world and situation knowledge in order to carry out the kind of mental modeling which will then be the basis for target discourse production. Interpreters usually take notes during the reception phase to be better prepared for the production phase. Note-taking is a process which requires attention to be split between listening and analyzing source discourse and deciding what has to be written down, the form in which messages may be retrieved most easily, and what shall best be left to memory. As the interpreter has been able to build up a mental model of what he is going to say before the production, the major remaining tasks are those of retrieving and retracing this mental model, with notes and memory interacting, and finding an appropriate linguistic representation for the target discourse. Therefore, notes made during reception will help the interpreter retrieve from memory the macro-structure of the discourse, the discourse’s cohesive interconnections, and micro-level information during the target discourse production phase. It is clear that both the Effort Model and the strategic information processing model divide the process of CI into two main phases and point to notes as the critical link bridging the two. A successful CI performance heavily depends on successful information retrieval from the notes, that is, whether notes are “deciphered” in a way that memory cues are correctly retrieved and lead to appropriate production. Successful information retrieval from the notes in turn relies on what and how the notes are jotted down. In other words, the content, layout, form, legibility of notes 10.

(22) are of tremendous importance to the final interpreting outcome. 2.1.3 Notes in Consecutive Interpreting Liu (2008) contends that CI beginners should learn to acquire a note-taking system because CI note-taking is very different from note-taking for other purposes such as lectures notes taken down by students in class, notes made by reporters at a press conference, or shorthand notes taken by a secretary (Liu, 2008). She illustrates the relationship between the Notes, the Speech (the source text), and the Interpretation (the target text), with a simple graph (Figure 1). The size of the squares demonstrates that the word content of the Notes is significantly less than that in the Speech and in the Interpretation. However, in terms of message, the three squares could be of equal size because the Notes contain the same amount of messages as in the Speech and the Interpretation. To put it another way, the Notes are not a summary but a “condensed” version of the Speech, and it can be “decondensed” to generate the Interpretation. According to Liu, CI notes could be written down in any language, are arranged on paper in a way that the space is effectively used, and serve as visual cues which carry sufficient meanings for immediate memory retrieval.. Speech. Notes. Interpretation. Figure 1. The relationship between the Notes, the Speech, and the Interpretation (Liu, 2008, p. 47). Indeed, notes are often considered a supplement, visual cues, or an aid, critical to a successful consecutive interpreting performance. The goal of note-taking is to enable subsequent retrieval, from notes and long-term memory, of the linguistic and informative elements of the original speech in order to ensure correct reproduction of the message in the TL (Abuín González, 2012). Gile (2009) also claims that in the Reformulation Phase in which Remembering Effort and Note-reading Effort intertwine, if notes are good, they help perform Remembering operations and may actually reduce Remembering capacity requirements rather than increase them. Thus, note-taking should be based on commonsense rules of efficiency and economy (Ilg & Lambert, 1996) and requires selective and strategic processing of the original speech (Kohn & Kalina, 1996). The latter can be understood as an ability to decide, at all times throughout the process, which elements should be noted down and how 11.

(23) they should be noted (using words, abbreviations, symbols, etc.), the amount of notes needed and also which languages should be used. Therefore, notes contain valuable information about the target text and may reveal a great deal about both the interpreter’s training and the techniques and strategies he employed while performing CI (Szabó, 2006). Lee (2000) examined twelve student interpreters’ notes and delivery in CI from English to Chinese and had three major findings. First, regardless of years of training, students rely heavily on their notes in speech production. Second, training has more effect on students’ notes than on speech production, that is, training has resulted in similar styles of notes, but not balanced performances by students. Third, senior students demonstrate better capability in coordinating their notes and working memory in that they tend to deliver fuller of their incomplete or missing notes while junior students tend to translate their incomplete notes incompletely or even incorrectly, showing that their information processing before the act of note-taking is not deep enough. Tsui (2005) conducted an English-to-Chinese CI experiment on three groups of subjects who had received one, two, and three years (Y1, Y2, Y3) of CI training respectively. He found that Y2 subjects, compared with Y1 subjects, are superior in the level of comprehension, indicating that note-taking of Y2 subjects requires less effort than that of Y1 subjects, which relieves more effort to comprehension. Y2 and Y3 are similar in terms of comprehension, but Y3 subjects are better at re-expressing the messages understood, suggesting that Y3 subjects excel in retrieving messages from notes. The two studies above found that note-taking and note-reading skills can be developed through training, and the mastery of such skills is associated with the interpreter’s comprehension and delivery. In fact, the topic of interpreters’ notes has generated a large volume of literature, most of them have focused on giving recommendations about what interpreters’ notes should look like or how notes should be taken. However, most of these recommendations are offered on the basis of personal experience and opinions without sufficient empirical support. In interpreting schools, note taking is always taught early during the training. Instructors tend to inform students some note-taking principles but eventually leave them to the students’ discretion on whether or not they want to follow. While it is generally believed among interpreters that note-taking systems are highly individualized, depending on personal preference, methods individually developed, familiarity with given subject, and memory capacity, there are two most commonly mentioned note-taking principles concerning the layout and language of notes: “verticalization” and “target language”. 12.

(24) 2.1.3.1 Layout Taking notes in a vertical layout is recommended by many interpreters as such layout contributes to a “visual” presentation of the ideas taken down. Graphically, this is achieved by indentation (Rozan, 1956, cited in Ilg & Lambert, 1996). Ilg and Lambert (1996) also believe that as the rationale of note-taking is to highlight the structure underlying a speech and the general semantic orientation of paragraphs and sentences, a telling, two-dimensional presentation helps the interpreter pick up the threads of message to be reformulated in the TL. Liu (2008) also embraces Rozan’s suggestion of writing notes in a linear up-down, left-right pattern. What differs is she sees indentation as a different note-taking approach from verticalization. She claims that if notes are written horizontally from left to right, when an interpreter starts interpreting, his eyes have to move from the very left of a line to the very right, and then move downward to the left of the next line, which is time and energy consuming and could impede fluent interpreting rendition. On the contrary, a vertical layout falls easily in the interpreter’s visual area, allowing the interpreter to view as many words and symbols as possible in one glimpse, hence, facilitating rapid and fluent production of the target speech. Therefore, vertical notes outweigh horizontal notes. Indentation, on the other hand, is a specific way to vertically arrange notes on paper which clearly indicate the relationship between different segments of the original speech. For instance, if a note element in the second line is indented compared to the element in the first line, it hints that the former information is subordinate to the latter; if they are arranged in parallel, then the two pieces of information are equally important. The “verticalization” principle is followed by many experienced interpreters and is taught by interpreting instructors to students. Surprisingly so far there has been only one study (Ma, 2013) providing empirical evidence supporting the advantage of this principle. 2.1.3.2 Language The choice of language used in note-taking is more controversial than layout. Some scholars recommend the use of the TL (Déjean Le Féal, 1981; Herbert, 1952; Laplace, 1990; Mikkelson, 1983; Seleskovitch, 1975, all cited in Dam, 2004). They believe that taking notes in the TL forces the interpreter to move away from the surface form of the incoming speech and facilitates better production of the target speech. Some others argue that the SL may be a better choice (Alexieva, 1994; Ilg, 1988, cited in Dam, 2004; Kirchhoff, 1979, cited in Dam, 2004) because it does not add the language conversion effort in the first phase of the consecutive interpreting 13.

(25) process, with its concurrent operations of listening, analysis, memory and attention. In addition to the two main opposing positions, Matyssek (1989), cited in Dam (2004), advocates a largely language-independent system for note-taking, and adds that when this ideal cannot be achieved, the interpreter’s mother tongue should be preferred because of its status as the better mastered language. So far, empirical studies of the language used in interpreters’ notes are still scant, especially quantitative ones. Dam is considered a pioneer in the research of interpreters’ choice of language and note-taking efficiency by introducing quantitative approaches, proposing additional categories for the analysis of the notes and creating methods and procedures for assessing efficiency. Her pilot study (Dam, 2004) consisted of five professional interpreters, with Danish as their native language (A language) and Spanish as their foreign language (B or C language), who interpreted from Spanish into Danish. All the interpreters showed a strong preference for taking notes in the TL, and tended to use the SL to a much lesser degree. This study also suggested a relationship between the degree of difficulty of the text and the use of the SL or TL, i.e. the more difficult the text, the more notes were taken in the SL, and the easier the text, the more notes were taken in the TL. This phenomenon is understandable since the interpreter’s processing capacity is limited, as demonstrated by Gile’s Effort Model. Easy texts leave the interpreter with leeway to convert language in the listening phase. Difficult texts, by contrast, consume most of the interpreter’s energy in comprehending and analyzing, thus leaving him few cognitive resources in language conversion. As a result, he could only afford to write down what was heard, that is, the SL. Her second experiment in the same study involved four student interpreters with three having Danish as their A language and Spanish as their B language, and the fourth having Spanish as A language and Danish as B language. They interpreted in both directions. The results showed that the subjects preferred to take notes in their A language, regardless of whether this language functioned as the TL or SL in the tasks. Dam concluded that the categories of A and B language are more relevant than the traditional ones of source and target language for describing language choice. Dam later examined the function of notes, that is, efficiency and non-efficiency, by adopting a different approach (Dam et al., 2005). Notes that were used for analysis were taken by Spanish-Danish interpreters during a simulated conference. She compared the notes that generated clearly accurate rendition of the source text and those that produced clearly inaccurate rendition in order to find out how the former set of notes were different from the latter set of notes. The findings showed that accurate target text used more notes, more abbreviations, and more SL notes than 14.

(26) inaccurate target text. They therefore hypothesized that many notes better ensure target text accuracy than few notes. Abbreviations, rather than full words, and SL, rather than TL, require relatively less time and effort to jot, and hence enable interpreters to produce as many notes as possible. Following Dam’s work, Szabó (2006) explored the language used by eight students at different stages of a two-year master’s program in conference interpreting. The subjects were practicing interpreters, with an average of three and a half years of professional experience before enrolling in the program. Their A language was Hungarian and B language was English. Both quantitative and qualitative analyses indicated that most subjects had a clear preference for taking notes in the B language, which was considered more concise than their A language, although they also tended to base their note-taking language choice on the status of the respective languages (SL or TL) in the task. Szabó raised the issue of note-taking economy and concluded that the nature (compactness and simplicity) of particular language pair is a relevant factor in the interpreters’ choice of note-taking language. Similar research on language choice in note-taking has also been done in the language combination of English and Chinese. Li (2012) recruited nine interpreting majors at post-graduate level, who had a language combination of Chinese A and English B, to consecutively interpret two speeches, one from Chinese into English and the other from English into Chinese. Her findings suggested that the subjects showed a general preference of English over Chinese in both tasks. Li reckoned that when Chinese was the SL, the interpreters afforded a greater luxury converting it into the TL, English, during the note-taking phase since comprehending their mother tongue required little effort. However, when English, their B language, was the SL, the interpreters had to spend more effort in comprehension and thus had less leeway to convert the languages during the note-taking phase. As a result, they jotted down notes mainly in English. Moreover, English was viewed as a more efficient noting code than Chinese because it required fewer strokes and was easier to write or abbreviate with. In addition to the above mentioned parameters that may affect the interpreters’ choice of language, such as task-related status of the language (SL vs. TL), the interpreter’s linguistic combination (A vs. B) and the nature of the interpreter’s language pair (i.e. morphologically long vs. non-morphologically long languages), another parameter—relative expertise of the interpreter has just begun to draw researchers’ attention, which will be discussed in detail in the last section of this chapter. 15.

(27) 2.1.4 Summary Unquestionably, notes play a pivotal role in CI as they aid an interpreter’s memory when the operations of message-receiving at the listening phase and message-producing at the interpreting phase are just minutes apart. While it is true that literature on notes is numerous, empirical studies are surprisingly scarce, especially regarding to the layout of notes. As for the language used in notes, all research has focused on the note-taking phase by discussing how the notes should be taken in order to alleviate an interpreter’s cognitive burden while he struggles with multiple tasks. For example, SL trumps TL because writing notes as how they are heard does not require the language-converting effort. A language outweighs B language because the former is the more mastered language. Also it is easier to jot the language that is less morphologically complex. Most of the existing studies drew conclusions by inferring from the research participants’ experiences and their interpreting performances without providing sufficient scientific data to prove that taking notes in a certain way does lighten an interpreter’s cognitive loads. Moreover, what kind of notes can facilitate efficient deciphering and thus can best reduce cognitive loads and quickly trigger memory during the interpreting phase is a perspective that has never been adopted. To fill this gap in the existing knowledge, Ma’s study (2013) employed the eye-tracking method to look into the cognitive loading that trainee interpreters experience in the note-reading and rendition phase. 2.2 Eye Tracking 2.2.1 Basic Characteristics of Eye Movements The eye-tracking method is extremely important in the field of cognitive psychology as it collects eye movement data that reflect moment-to-moment cognitive processes in a wide variety of tasks. Eye movements are behavioral cues that reflect concurrent cognitive functions. While reading, looking at a scene, or searching for an object, eyes continually make rapid movements called saccades. Between the saccades, eyes remain relatively still during fixations. Saccades and fixations are the two basic components of eye movements which serve as indicators to information processing. Eyes make saccades frequently because of acuity limitations. Human’s visual field can be divided into three regions: foveal, parafoveal, and peripheral. Although visual acuity is very good in the fovea (the central 2∘of vision), it is not nearly so good in the parafovea (which extends out to 5∘on either side of fixation), and it is even 16.

(28) poorer in the periphery (the region beyond the parafovea). Hence, people move their eyes to place the fovea on the part of stimuli that they want to see clearly. Characteristics of the stimulus in parafoveal or peripheral vision influence whether or not a saccade needs to be made to identify it (Rayner, 1998). When saccades occur, sensitivity to visual input is reduced because the eyes are moving so quickly that only a blur is perceived. This phenomenon is called saccadic suppression. In other words, no new information can be acquired during a saccade. However, while new information is not encoded during saccades, cognitive processing does continue in most situations (Irwin, 1998). Fixations, on the other hand, are periods when new information is encoded and processed, which generally last between 200-250 milliseconds (ms) depending on the particular task (Table 1).. Table 1. The range of mean fixation durations and the mean saccade length in silent reading, oral reading, scene perception, and visual search (Rayner, 2009, p. 1460). Just and Carpenter (1976) examined several tasks that required the subject to encode information from a visual display, do mental computations on that information, and then produce a response that is contingent on the outcome of the computations. They proposed the “eye-mind” hypothesis, suggesting that eye-movement recordings can provide a dynamic trace of where a person’s attention is being directed in relation to a visual display (Poole & Ball, 2006). The eye-tracking method, therefore, has been applied to investigate the underlying cognitive and psychological processes in various research fields, including language reading, music reading, human activity recognition, the perception of advertising, human-computer interaction, and the playing of sports. Eye-tracking data are also able to shed light on similarities and differences between different groups of people engaging in the same 17.

(29) activity, such as the differences between men and women while browsing a website, and the distinct performances between novices and experts in typing, playing chess, and performing surgery, just to name a few. As a result, many eye movement measures, derived from saccades and fixations, have been used to explain cognitive activities in different tasks, though their definitions and interpretations may vary depending on the task examined. Reading, in particular, is the area that has been explored most thoroughly with the eye-tracking method, and this study, with its focus on CI’s note-reading phase, will be borrowing its measures and definitions for data analysis. 2.2.2 Eye Movement Measures Tables 2 and 3 (Radach & Kennedy, 2004) present the definitions of some of the more important spatial and temporal eye movement measures used in reading research. The word-based spatial eye movement measures in Table 2 reflect three aspects of eye movement behavior: the fact that a word is, or is not, fixated; the position fixated; and the amplitude of incoming and outgoing saccades. There are many possible combinations of these properties. For example, distinguishing between the amplitudes of saccades into and out of words and those within them (inter- vs. intra-word movement) and considering separately whether these saccades go from left to right (progressive saccades, when reading English and other left-to-right languages) or are directed against the normal reading direction (regressive saccades) have been proven to be of significant theoretical meaning. Similar refinements are possible in the case of the temporal measures shown in Table 3, for example, distinguishing the first and second of two fixations from cases where exactly one fixation is made on a critical word. Each of these measures is, to some degree, pertinent to a particular theoretical question.. 18.

(30) Table 2. Definitions of the more commonly used word-based spatial eye movement measures (Radach & Kennedy, 2004, p. 6). Table 3. Definitions of the more commonly used word-based temporal eye movement parameters (Radach & Kennedy, 2004, p. 7). 19.

(31) Given that fixation duration is more sensitive than saccade length to linguistic factors, most commonly used measures of processing difficulty in psycholinguistic studies are temporal measures. Where temporal measures are concerned, a useful notion for describing fixation patterns is the concept of a “pass”, denoting the first or the second and beyond encounter with some defined segment of text. When the region of interest (ROI) in a text is a single target word, the measures most often used are “first fixation duration” (the duration of the first fixation on the word in the first pass), “single fixation duration” (the time spent on a word on which only a single fixation is made), and “gaze duration” (the sum of the durations of all fixations on the word before leaving the word, or more precisely, in the first pass). In addition, “total viewing time” on the word (the time spent on the word including re-reading) is often reported, as is “go-past time”, also known as “regression path duration” (the time from first fixating on the word to first moving past the word to the right, including time spent in rereading earlier parts of the sentence). Less frequently, “second pass time”, or “rereading time” (the time spent rereading the word) is also reported. When larger regions of text are being examined, the first fixation duration is often not a meaningful measure, but the other measures mentioned above are all commonly reported. Usually, when a region comprises multiple words, the term “first pass time” is used instead of gaze duration (Staub & Rayner, 2007). Among these measures, first fixation duration and gaze duration/first pass time are often referred to as “early” measures, while total time and second pass time are “late” measures. Although controversies remain concerning their exact definitions, it is generally believed that early and late measures respectively reflect different stages in information processing. For instance, the early stage in information processing usually refers to recognizing a word based on its meaning. In the late stage, the meaning of the recognized word is integrated into a bigger context, such as a sentence or a paragraph. 2.2.3 Eye Movements in Reading Based on the eye-mind hypothesis proposed in 1976, Just and Carpenter (1980) proposed two other assumptions for the relationship between eye movements and reading comprehension. The first was the immediacy assumption: readers immediately process the fixated word so that information processing is not deferred but happens immediately upon fixation. The second assumption was that eyes remain fixated on a word as long as the word is being processed. Also, the time needed to process a newly fixated word is directly indicated by fixation duration. Based on these two assumptions, eye movements can, to some extent, indicate the mental processes that take place in reading. 20.

(32) Over the past few decades, the general characteristics of eye movements during reading have been studied in great depth. These characteristics are reviewed summarily below: English reading first, followed by Chinese reading. 2.2.3.1 English Reading During English reading, eyes rest in fixation for approximately 200-250 ms. Saccades between fixations span an average about 2 degrees of visual angle, or it is better expressed in terms of a span of 7 to 9 spaces of English letters. The chances of an individual word being fixated on vary according to whether it is a content word (85%) or a function word (35%), and in relationship to the length of the word, with 2-3 letter words being skipped 75% of the time, but 8 letter words fixated almost always. Eye movements also vary as a function of the syntactic and conceptual difficulty of the text. Moreover, although readers typically move their eyes forward when reading, approximately 10-15% of saccades move backward, fixating previous letters or words. Many regressions tend to be only a few letters long and could be due to the reader making too long of a saccade, in which case a short saccade to the left may be necessary to direct the eyes to where the reader really wants to pay attention to. Short within-word regressive saccades may also be due to problems that the reader has encountered while processing the currently fixated word. Longer regressions (more than 10 letter spaces back along the line or to another line) occur because the reader did not understand the text. In such cases, good readers are very accurate in sending their eyes to that part of text that caused difficulty, whereas poor readers engage in more backtracking through the text. These regressive saccades are thought to be related to difficulties in processing an individual word, or difficulties in processing the meaning or structure of a sentence (Rayner, 2009). Take one reading comprehension study as an example (Rayner et al., 2006). Sixteen native English speakers read 32 passages with different levels of difficulty. The results showed significance in positive correlation between difficulty rating and average fixation duration, number of fixations, as well as total reading time. In other words, the harder the passage is, the longer the average fixation duration and total reading time and more number of fixations. Another experiment in this study examined the effect of inconsistency in text on eye movements. Its data suggested that when there was an inconsistency, readers fixated longer on the region where the inconsistency occurred. In both experiments, the probability of making a regressive eye movement increased as well. In short, as text gets more difficult, fixations get longer, saccades get shorter, and 21.

(33) more regressions are made. Typographical variables such as font difficulty can also influence eye movements. Fonts that are more difficult to encode yield longer fixations, shorter saccades, and more regressions. Where level of proficiency is concerned, beginning readers have longer fixations, shorter saccades, and more regressions than skilled readers. Compared with normal reading, which is reading silently, eye movements for reading aloud somewhat differ, as previously shown in Table 1. When reading aloud, mean fixation durations are longer than in silent reading, and the eyes tend to get ahead of the voice. 2.2.3.2 Chinese Reading Chinese is considered the most different writing system from English. First, English is alphabet-based whereas Chinese is character-based. An English word is only meaningful when its building blocks—the alphabets—are arranged in the right order. Each alphabet standing alone has no meaning. By contrast, a Chinese word can be composed of one character or more. Each standalone Chinese character already conveys certain meaning and different combinations of the same characters may express different meanings. Therefore, the same amount of messages can usually be expressed with fewer Chinese words than English words, making Chinese a more “compact” language. Second, English only makes sense when it is read from left to right while Chinese characters can be arranged either vertically or horizontally. In Chinese, depending on how the characters are arranged, a text can be meaningful by being read either from top down or from left to right. Due to these distinct differences, it is not surprising that eye-movement literature focusing on Chinese reading is also in abundance as English reading. Chinese reading studies have shown that the mean fixation duration is about 220-230 ms, not much different from English reading. Their regression rate does not differ dramatically either. Where they do differ is that Chinese readers’ average saccade length is much shorter than that of English readers as they typically move their eyes only for the span of 2-3 characters, which makes sense given that linguistic information in Chinese is more densely packed than in English (Tsai et al., 2005). Regarding the layout, as early as in 1925, studies showed that readers’ mean fixation duration was shorter while reading Chinese horizontally (294 ms) than vertically (305 ms). However, they used more fixations in each horizontal line than vertical line. The average number of regressive fixations during horizontal reading also outnumbered that in vertical reading, 2.51 and 1.16 respectively. Researchers, therefore, concluded that reading Chinese texts of vertical alignment was easier than horizontal alignment 22.

(34) (Miles & Shen, 1925; Shen, 1927). Nonetheless, another study (Sun et al., 1985) found readers more skillful in reading the horizontal format. During vertical and horizontal reading, the fixation durations were 290 and 260 ms respectively; saccade distance 1.2 and 2.6 characters; and reading speed 260 words and 580 words per minute. One explanation of such discrepancies in these studies many be due to the fact that a majority of Chinese printing shifted from a vertical-line format to a horizontal-line format in the 60-year time span. Readers became much less adept at reading vertically (Tsai et al., 2005) Another study conducted by Tsai et al. (2000) found that Chinese reading speed is significantly faster while reading horizontally (753 words/min) than vertically (713 words/min). Such difference can be mainly attributed to the fact that readers fixate fewer times in each horizontal line (4.8 times) than vertical line (5.2 times). This result also suggested that visual acuity is superior while reading horizontally than vertically as readers were able to extract sufficient information in fewer fixations while reading horizontally. In addition, saccade distance was longer (3.9 characters) while reading horizontally than vertically (3.6 characters). The subjects switched between lines faster in a horizontal-reading context than a vertical-reading context. However, the mean fixation durations did not differ significantly in the two situations (214 ms and 221 ms respectively for horizontal and vertical reading). These results implied that although readers extract less information in one fixation while reading vertically due to inferior visual acuity, their eyes could still obtain as much necessary information as in a horizontal reading context by increasing the number of fixations and shortening saccades to compensate for the less than ideal information gathering capability while reading vertically. 2.2.4 Eye Movements in Interpreting In the field of interpreting, studies employing new data-gathering techniques are on the rise. Especially, collecting and analyzing eye-movement data in this particular field has attracted an increasing number of researchers in the past few decades. McDonald and Carpenter (1981) were the pioneers linking the eye-tracking method to interpreting studies. Two expert interpreters and two amateur German-English bilinguals sight translated forty-four texts from English to German while their eye fixations and oral translations were recorded. The results showed that the eye fixation patterns differed depending on whether the subjects interpreted idiomatic expressions literally or idiomatically. This study, by analyzing the eye-tracking data, divided up the sight translation process into finer sub-processes, namely two distinctive phases, parsing and error detection (Figure 2), which build on 23.

(35) normal reading processes. The initial comprehension of the phrase was marked by the initial sequence of forward gazes, which is the first-pass reading in eye movements. The second process was manifested by regressions, which constituted the second scan or the re-reading of the phrase, when oral translation occurred. If and when the subject detected an error in his comprehension/interpretation after the second scan, the eye regressed back again to the preceding segment which contained the ambiguous idiomatic phrase whilst the previous oral translation was edited. This study paved the way for future studies exploring cognitive processes underlying sight translation, and technically other types of interpreting.. Figure 2. A flowchart of the processes in the translation task (McDonald & Carpenter, 1981, p. 241). Tommola and Niemi (1986) and Hyona et al. (1995) used pupil diameter as a form of measurement and both studies lend good support to the use of the pupillary response as an indicator of processing load in simultaneous interpreting. Tommola and Niemi found that the participants’ pupil diameter peaked when restructuring of the output sentence was required because of the syntactic differences between the source language (Finnish) and the target language (English), which confirmed the hypothesis that larger pupil diameter could indicate heavier cognitive loading. Hyona et al. investigated the sensitivity of the pupillary response to reflect variations in processing load during language processing by comparing three different tasks that differ in complexity performed by Finnish-English student interpreters. The results demonstrated that the pupil dilated significantly more during single word interpretation than during shadowing, followed by passive listening. The pupil size 24.

(36) also reflected variations in within-task demands because words that were determined to be more difficult to translate induced higher levels of pupil dilation than did easily translatable words. Following McDonald and Carpenter’s work, Huang (2011) used eye tracking data to examine the participants’ comprehension process while performing three tasks: silent reading, reading aloud, and sight translation. Her study further confirmed McDonald and Carpenter’s finding that comprehension in sight translation does not consume more cognitive efforts than normal silent reading. In addition, reading ahead, defined as the condition in which the participants’ eyes fixate for the first time on the content in sentence N+1 but orally producing content in sentence N, occurred frequently with a high probability of 72.8%, indicating that sight translation is indeed a suitable practice before learning simultaneous interpreting. These studies have proved that the eye -tracking method is indeed a valuable technique in unraveling the underlying cognitive processes when interpreters comprehend the source text (or speech) and carry out interpreting. McDonald and Carpenter (1981) and Huang (2011) studied ST. Tommola and Niemi (1986) and Hyona et al. (1995) studied SI. The noticeable gap, CI, has been filled quite recently by Ma’s research (2013), on which the present study is based. Ma’s research is a pioneering study that investigates the effect of layout and language used in CI notes on interpreters by adopting the eye-tracking method. Nineteen student interpreters from two-year translation and interpretation programs in Taiwan participated in the study. All of them had Chinese A-English B and interpreted from English into Chinese in the CI experiment. Their eye movements were recorded while they read the notes and produced their oral translation. Eye movement measures commonly used in normal reading, such as fixation duration, saccade length, and regression rate were used to analyze and interpret the participants’ eye movement patterns. His results showed that the mean saccade length of the participants was shorter while reading the vertical notes than the horizontal notes, and their total viewing time on notes written in the TL, in this case, Chinese, was shorter than that on the SL, English. In short, this study concluded that the note-taking principles of “verticalization” and “target language” work to interpreters’ advantage by lessening their cognitive loads while interpreting. 2.2.5 Summary The eye-tracking method has been proved to be an effective tool to dig into the cognitive activities during information processing in various tasks, including interpreting. In addition, the eye movement measures used in reading research have 25.