An Inference Ability Perspective:Depth Of Discourse Processing In Simultaneous Interpretation

Ever since the first appearance of simultaneous interpretation,its cognitive mechanism,dubbed as the “Black Box”,has attracted much interest from the research circle.According to the Interpretive Theory proposed by Seleskovitch and Lederer,simultaneous interpreters extract the sense of the source discourse through “deverbalisation”,which has been joined by other researchers in a number of SI models.Nevertheless,few empirical studies have been conducted so far due to the complexity and sophistication of holistic SI models.Meanwhile,some studies found that the depth of processing in SI was inferior to that in listening by showing poorer memory recall after SI than after listening,calling into question the high-level discourse comprehension in SI.Given the fundamental role of comprehension in SI,this study strives to address the depth of discourse processing by way of empirical experimentation in the hope of lending empirical support to Interpretive Theory,and advising on SI training.Set in the framework of discourse inference theories,this dissertation gave a thorough review of SI models from the perspectives of cognition,discourse and memory,summarized different sources of information for SI comprehension,and discussed the constraints and strategy adopted by interpreters.Bearing in mind the similarities and differences between SI and general comprehension,this study focused on two types of inferences: bridging inference and predictive inference.As they are kept at different levels of mental representation in comprehender's mind,it is possible to detect inferenced information via priming experiment and then reveal the depth of processing.Among the four factors for individual differences,L1 comprehension ability & L2 proficiency and background knowledge were treated as control variables.The other two which are more closely related to SI,namely,working memory and task goals,were chosen as explanatory variables,to investigate the relationship between WM/task goals and depth of processing,In this study,12 professional simultaneous interpreters and 12 untrained bilinguals with high L2 proficiency were invited to participate in the experiments of two categories.The first one was about WM,which was comprised of three tests: Digit span;word span;and listening span with articulatory suppression.This part of experiments aims to collect WM capacity in terms of storage and processing.Besides,WM experiment was also combined with a retrospective interview,which looked into the participants' coping strategies under high cognitive load.The second category is two experiments on inference,conducted with E-Prime software.In these two inference experiments,participants in each group were divided into high/low WM groups based on the first WM test.The first inference experiment collected the response times(offline)and accuracy on questions regarding both types of inferences for interpreters and untrained bilinguals doing SI/shadowing task.The second experiment again collected reaction times(online)in a naming task and accuracy for general comprehension questions for interpreters only but doing different tasks of SI or pure listening.All the collected data from WM and inference experiments was analyzed with SPSS.Different tests such as ANOVA and correlation tests were run for different sets of data when they met the required assumptions.The findings demonstrate that:(1)In terms of WM,interpreters exhibited significantly better performance only on tasks with articulatory suppression,despite the fact that their storage scores were slightly higher than those of the controls.This was later cross-checked by retrospective study,showing that interpreters were more active in trying out new strategies in multitasking;(2)In a highly cognitive task,the offline accuracy and RT of the interpreters,though slightly better than the controls,failed to demonstrate any significant difference;(3)When doing SI or listening,interpreters showed a similar level of accuracy,but online RTs for both types of words were significantly faster in SI than in listening with the biggest gap seen on filler words;(4)In SI,interpreters reacted even faster to words representing PI than to BI,which is in contradiction with what inference models would predict.This suggests that interpreters discard processed information more quickly than uncertain information;(5)WM played a different role with interpreters and controls in inference generation.WM was significantly correlated to accuracy and RTs for controls but not for interpreters.However,in the interpreters-only experiment,significant difference in online RTs was found for high and low WM groups.Based on these findings,this dissertation concluded that in SI,interpreters employ a language processing mechanism at least as equally effective as in listening.Besides,their faster RTs in SI than in listening indicates that they adopt a more effective mechanism than in listening,clearing all the doubts on depth of processing in SI.Lastly,non-interpreters reply more on their processing capacity to generate PI while it is more like automated processing for interpreters.Only when multitasking will their superior WM capacity come into play,affecting their processing speed of shifting among different tasks.This study has made a few contributions as follows:(1)By offering empirical evidence,this study proved the existence of high-level conceptual processing,lending further support to models in favor of Interpretive Theory;(2)It is among the early attempts to depict the correlation between WM and depth of processing in SI,revealing some limitations of the current inference theories;(3)This study has addressed the specific role of WM in SI comprehension,which is hidden in the process rather than the product;(4)Research methodology based on psycholinguistic priming paradigm was adopted and tailored for an SI study at the discourse level,minimizing the subjectivity from scoring interpretations;(5)This study offers useful advice for SI training in the sense that students should keep practicing their listening skills and WM processing capacity to arrive at the level where comprehension of general information or their specialized knowledge in specific domains becomes automated.Finally,this dissertation concludes with limitations and possible suggestions for future work.