The Impact of Non-standard English Input on Interpreting

Unfamiliar accents have long been recognized as an interfering variable to interpreters' performances. For instance, in the AIIC workload survey (AIIC, 2002),

unfamiliar accent has been quoted as one of the top stressors for interpreters . It is also suggested in Daniel Gile’s Effort Models that interpreters will need to devote more

processing capacity to comprehension when the speech input is heavily accented.

(Gile 2009). However, to date, only a handful of empirical studies have been attempted to investigate into this issue. This section reviews the findings of six empirical studies on unfamiliar accent and interpreting that have informed the current project's research design and questions for investigation.

Out of the six empirical studies, Kurz (2008), Lin (2011), Mazzetti (1999), and Sabatini (2000) recruited student subjects and put them to simultaneous interpreting tasks, while Lin (2010) put her subjects to consecutive interpreting tasks, to observe their performance under nonstandard or foreign-accented source text input. All of them concluded that accented input that differs from native varieties that the subjects

19

are familiar with does impact interpreting performance negatively. Nevertheless, these studies all suffer from methodological or theoretical problems that compromise the validity and reliability of their conclusions. The following paragraphs provide a summary of the findings of these studies, their methodological pitfalls, and how these inform the research design of the current project.

In Kurz (2008), ten students were divided into two groups and simultaneously interpreted a speech of under five minutes. Half of the speech was delivered by a native speaker while the other half delivered by a nonnative speaker with an unidentified foreign accent. By counting the percentage of propositions correctly rendered, Kurz concluded that both groups of students, who heard the two accents presented in different orders, performed far better in interpreting the native speaker.

However, there are a few methodological flaws in this study. Firstly, as Kurz noted in the paper, the fact that the group who heard the native speaker first experienced about 22% less of information loss in the accented speakers' condition than that of the group who heard the nonnative speaker first suggested that order effect was at play.

Moreover, there was very limited information on the nature of the source input, including text difficulty and speakers' linguistic backgrounds. Finally, the results

20

cannot tell us whether the information loss was due to intelligibility, comprehensibility, or interpreting skill breakdowns.

Kurz's discovery of quantitative impairment in the information rendered, however, was not confirmed by Mazzetti (1999), who compared the percentage of correctly and incorrectly rendered target sentences by three groups of interpreting students: two groups of Italian students listening to German native and nonnative speaker's speech respectively, and one group of German students listening to

nonnative German speaker's speech. He concluded that accented input "seem to have remarkable repercussions more on the quality of TT mistakes than on the quantity"

(Mazzetti, 1999, p. 141). In other words, the number of mistakes made between the two nonnative-listener groups: one that heard the native speaker and the other that heard the nonnative speaker, did not differ significantly. However, the percentage of severely-semantically incorrect renditions was higher for the group that heard the nonnative input. Moreover, by including native listeners in the research, Mazzetti was able to observe the benefit of native listeners in coping with accented input, which is unique in all the studies surveyed here. He suggested that, firstly, native listeners can cope with accented input better than nonnative-speaking interpreters, making less mistakes both quantitatively and qualitatively. Secondly, while nonnative-speaking

21

interpreters are more strongly influenced by semantically-impairing delivery deviations, native-speaking interpreters are more troubled by the loss of rhetorical effect in their renditions.

But before one can claim that Mazzetti's results conflict with Kurz's and that therefore the issue of quantitative impairment of interpreting output for

foreign-accented speech remains unsettled, the low validity of Mazzetti's results must be considered. First of all, Mazzetti adopted the inter-group experimental design without accounting for the individual differences of the participants' interpreting abilities. Moreover, judging by the fact that the group of nonnative listeners who were exposed to native source input ("the text presented correctly", as in Mazzetti's own words, p. 137) produced 80% of the renditions incorrectly, it may be very likely that the recruited subjects were either quite incapable of the SI task or that the text chosen (a speech made by a German minister) was too difficult for the participants to process, either of which is a serious blow to the validity of the results. In fact, it may be fair to say that the lack of significant quantitative impairment in the interpreting output for the nonnative speaker in Mazzetti's study was not because the participants did not suffer from the accented input, but that the participants are suffering anyway when the input was not accented. This shows the importance of task and text difficulty control,

22

issues that were both taken into careful consideration by the researcher in this current project.

Lin (2011) and Sabatini (2000), on the other hand, investigated into the micro level of the impact of non-standard input, and attempted to identify specific elements of the input that cause interpreting difficulties. Sabatini put ten student subjects to listening, shadowing, and simultaneous interpreting tasks for two speakers of non-standard English, and concluded that "atypical features" in the input, including pronunciation deviations and culture-bound references, are sources of difficulty for interpreters. The scope of investigation of this piece of research, however, is 'non-standard' English, i.e., authentic English speeches featuring an ESL speaker (Indian) and a native speaker with a regional accent (American) that is different from the British English accent that the participants are used to. Therefore, it examined foreign and regional accents without any distinction, and could not separate

lexicogrammatical interferences from phonological interferences in the conclusion.

This piece of research again demonstrates the necessity to target the scope of investigation to 'foreign-accents', i.e. the phonological aspects of nonnative

productions, to produce research results that can better focus the training objectives.

23

Informed by the ESL/EFL studies on the impact of segmental and

supra-segmental deviations on intelligibility/comprehensibility, Lin (2011) created artificial accents by manipulating the prosodic and segmental features of the source input based on the phonetic and prosodic features that have been identified as potentially problematic for listeners' intelligibility or comprehensibility judgment to serve as the stimulus material in her research. After rating the SI performance of 37 interpreting trainees, she concluded that while prosodic or segmental deviations alone significantly impair student interpreters' accuracy of SI renditions, prosodic influences were found to be more significant statistically. Although this result corresponds to the suggestion that prosody has a greater effect than segmental in impairing intelligibility as shown in ESL literatures, there is no way to distinguish whether the SI accuracy degradation of Lin's participants was due to intelligibility problem or comprehension problem. In fact, the majority of the participants rated the input material rather difficult or very difficult, and the reasons for its difficulty were attributed to both lexical/structural and phonological causes. Moreover, the unusual finding that the co-existence of prosodic and segmental variations in the input, a condition that is more plausible in real-life context, did not cause significant accuracy impairment was not accounted for. Finally, the choice of an inter-group research design meant that the

24

validity and reliability of her conclusion may be compromised by individual differences in interpreting skills, which the researcher herself has also noted.

Amongst the studies on accented input and interpreting reviewed so far, Lin's was the one that had gone the most length in controlling for text difficulty, employing both readability indexes and expert judgment to control for the text difficulty.

Nevertheless, "participants' reaction to the ST was different from pre-experiment expert judgment" (Lin, 2011, p. 60). In light of the difficulty in controlling text difficulty, the success of which is pivotal in ensuring the validity of an interpreting study, the researcher of the current project adopted three different measures to attempt better success. Please refer to Chapter 3 of this report for details.

Lin (2010) also put student subjects to actual interpreting tasks, but to

consecutive interpreting task. She tested 32 student interpreters' comprehension for speech materials read by speakers from the inner circle, the outer circle, and the expanding circle. The model of 'three concentric circles of English' was conceived by Braj Kachru to categorize English-using world into three circles: the inner circle consisting of native English-speaking countries, the outer circle of former colonies, and the expanding circle of countries where English, though neither a native language nor one of the official languages, is becoming an important language in business,

25

science, technology and education (Kachru, 1992). Lin's analysis concluded that student interpreters in Taiwan suffer the most comprehension degradation when listening to speakers from the outer circle. However, the apparent confusion between 'accented input' and 'nonnative English input' in this piece of research rendered the validity of this conclusion questionable. In earlier sections of this chapter, it has been shown that the term 'accent' refers to only the phonological realization of a language.

Kachru, however, proposed the model of concentric circles of English to explain "the current sociolinguistic profile of English" (Kachru, 1992, p. 356), and made it clear that the categorization is based on not only the linguistic but also cultural aspect of the English language used in these different circles. Moreover, Kachru and other

researchers have never attempted any generalizations concerning the specific

segmental or prosodic realizations of English phonology within each circle. Therefore, it may be a stretch for the literature referred to here to argue that student interpreters in Taiwan suffer the most comprehension degradation when listening to outer circle English 'accents' when common norm-deviating phonological features amongst the outer-circle English speakers selected do not exist. The fact that the speech materials used in Lin (2010) were English materials written based on inner-circle-, or native English grammatical model also strips off the possibility to interpret the result as an indication of Taiwanese student interpreters' poor comprehension of 'nonnative

26

English' input by outer-circle speakers. Nonetheless, she proposed a list of strategies that student interpreters resort to when interpreting accented speakers, including the use of relevant information such as context, background knowledge, and key words, anticipation, summary, repetitions, and omissions. These strategies partly match those proposed by Kurz (2008), whose subjects resorted to turning up the volume, focus on key words, adjusting time lag, and summary to cope with accented input. In the current project, the researcher elicits participants' report on their own use of coping strategies when listening to the foreign-accented input, and observes if there are any similarities with and differences from these findings in existing literatures.

Finally, Chang and Wu (2012) is the only piece of research to the researcher's knowledge that involved professional interpreters as subjects and did not put interpreters to actual interpreting tasks to investigate the issue of nonnative English input. The qualitative analysis of interviews with ten professional interpreters in Taiwan reveals that interpreters rely on sufficient advance preparation, contextual and semantic cues, good interpreting skills, increased awareness of nonnative speakers' mistakes, and booth-mate support to cope with the challenge of nonnative speakers.

However, contrary to the reactions of the student interpreters in Kurz's (2008)

experiment, in which the participants were "shell-shocked" (Kurz, 2008, p. 189) when

27

hearing the accented input, the professionals interviewed by Chang and Wu reported that they seldom meet nonnative speakers whose accent is too strong to comprehend at all, and that "delivery speed and difficult subjects often trump the accent challenge"

(Chang & Wu, 2012, p. 7).

Although the topic investigated in Chang and Wu was 'nonnative English input' that involved not only phonological but also lexicogrammatical variations of the language production, which is broader than the scope of the current project, it is nevertheless the only piece of research that seems to suggest that the challenges posed by phonological deviations in the accented input to interpreting success may not be as fatal as commonly believed. Instead, other factors such as delivery speed and content are more crucial to the interpreting success. There may be two possible explanations to why this conclusion is unique to Chang and Wu's investigation. First, professional interpreters process accented input differently. Their experiences may have quipped them with better decoding skills so that they seldom encounter accents that are too difficult to decode, or that they have developed better comprehension or production skills to compensate the intelligibility loss. Second, provided with the fact that none of the above reviewed studies involving student subjects could distinguish if the

interpreting performance impairment observed was due to intelligibility degradation,

28

comprehensibility degradation, or other skill-related insufficiencies, the possibility that student subjects' poor performance were caused more by comprehension problem or interpreting skill deficiencies, not the intelligibility loss as a result of deviating phonological input, remains open.

2.3.2 The Gap in Existing Literatures

In 2.3.1, it has been demonstrated that, although most of the existing studies on the impact of nonnative or nonstandard English (accented) input conclude that interpreting performance does suffer as a result, these studies contain many

methodological problems that render their conclusions unreliable and some invalid.

Furthermore, in light of the possibility that professional interpreters' perception of non-standard English speakers is different, it is essential to investigate further into whether a professional-novice difference exist in processing and coping strategies for foreign-accented speakers.

To fill these gaps in existing research, this study strives to improve the methodological design of the empirical investigation, and narrow the scope of investigation to whether interpreters suffer intelligibility impairment in perceiving foreign-accented input with a rigorous control of text and task difficulty. In the current project, the researcher modeled on Kurz's (2008) design of material

29

presentation, i.e. having the participants exposed to half of a speech in foreign accent and the other half in native accent that they are familiar with. However, the researcher introduces two speech materials and distributes the order of accent presentation equally across the participants to eliminate order effect. The experiment is also changed to within-group design to control the variable of individual abilities. Finally, this study looks at both professional and novice interpreters' intelligibility scores in the hope to explore further if, and in what way do professional interpreters process accented input differently. To inform the comparison between professional and novice interpreters' performance, below is a summary of the research findings on the topic of expert performance.

2.4. Differences between Expert and Novice Performance