Expert Behaviour in English-Chinese Sight Translation: An Integrated Eye-tracking Study

Sight translation (ST) is unique in that the interpreter relies solely on reading for information retrieval — and therefore more easily influenced by distinct linguistic structures of different languages — while under the pressure of instantaneously providing smooth oral rendition of the output. Unearthing the cognitive nature of ST will not only help to enrich our understanding of translation and interpreting as highly demanding communicative activities, but also has the potential of enlightening interpreter training, facilitating the development of ST skills. However, most ST studies to date, especially in the language pair of English and Chinese, are based on product analysis. While we have accumulated valuable data on the differences between the performance of interpreters at different levels, what happens in the process — how experience leads to distinct behaviour of the experts — remains largely unknown; being able to identify the nuances may provide insights into boosting the efficiency and effectiveness of training.
To complement product-based research, the current study used an eye tracker to examine English-Chinese ST. Through an integrated analysis of both the cognitive process and the final output, this study aimed to reveal the impacts of interpreting experience upon different stages of ST, including the behaviour of reading ahead and pausing and also how the output has been affected. Three texts of 175 words were adapted from three speeches by the same speaker. Rewriting was kept to a minimum to maintain the authenticity and features of formal speeches in a diplomatic setting. Each participant was asked to perform three different tasks: silent reading (SR), reading aloud (RA), and ST. Two senior AIIC interpreters were invited to rate the ST recordings of 17 experienced interpreters, all with more than 150 days of interpreting experience, and 18 interpreting students, who had undergone sight translation training within three years by the time of the experiment. In addition, six Chinese native speakers were recruited to listen to the recordings and mark observable pauses for further analysis.
Overall, this study partly corroborates previous research in that the reading purpose does affect reading behaviour, and ST requires significantly more fixations and time in total than SR and RA. However, some striking similarities between tasks were observed as well, including similar behaviour for SR and ST in the initial reading stage, comparability between RA and ST in later reading stages, and the same amount of rereading induced by SR and ST — the last of which seems counter-intuitive and contradictory to previous findings. The results have shown that analysis of various stages of reading reveals more intricate nature of the cognitive tasks involved.
In terms of ST, experts showed significantly better quality than trainees — contributed by accuracy alone, not expression style (including fluency). On the other hand, the task time, total number of fixations, and the number of words uttered were statistically similar, as were the mean fixation duration for all stages of reading. This shows that a short period of ST training was already sufficient to bring many features of trainee’s behaviour and output to the same level as those of the experts.
Looking closer at the behavioural data, the two groups remained on par in terms of the amount of reading ahead to acquire information before they started sight translating the text. Trainees strictly adhered to the principle of ‘starting as soon as possible’, while some of the experts were more relaxed and took more time, albeit not statistically significant, to skim the text. Still, the equally high score on expression style for the experts reflects that interpreting experience might have allowed the interpreters to better manage the audience’s expectations and helped them to squeeze more time for preparation without turning it into awkward silence before they began.
However, once the participants opened their mouth, the two groups exhibited different — and statistically significant — patterns in every aspect covered in this study. Experts relied on fewer fixations and spent less time on information retrieval ahead of each utterance (defined as a Chinse character here); the number of observable pauses in the process were fewer, and the average length of each pause was shorter, along with fewer fixations therein. A separate analysis on certain selected participants further indicates that, while trainees showed a significant higher percentage of hesitation pauses than their juncture pauses, experts were able to limit their hesitations and finished the experiment with a similar percentage for both types of pauses.
In this study, interpreting experience exerted its influence on accuracy and the behaviour of reading ahead and pausing once sight translation began. Interestingly, experts were not more concise or more efficient in reading. Instead, the major difference seems to lie in language flexibility. Experts’ flow of interpretation was rarely disrupted because they required significantly less time and information to keep reformulating the source content and reorganising their sentences for a smooth rendition.