CHAPTER 5 DISCUSSION
The main goal of this study was to investigate the effects of numbers on simultaneous interpreting, in terms of accuracy and fluency. In addition, the effect of numbers on the interpreting of subsequent sentences was also examined. The results of this study have been presented in the previous chapter. Although the effects of numbers did not reach statistical significance, probably due to the small sample size, some general trends can be observed from the findings. This chapter will discuss these findings and their implications in greater detail. Also discussed are interesting retrospective data from participants on their attitudes toward numbers, and their own observations of number interpreting.
5.1 Numbers and Accuracy
This study found that not all numbers in the speech led to lower accuracy in SI.
It was numbers requiring number word syntax conversion that had a greater impact on accuracy. Results from this study showed that Conversion-NS were interpreted less accurately than their corresponding Control-Sentences, which differed from Conversion-NS only in the absence of exact numbers. Unfortunately, the effect of
numbers on accuracy was confounded by the difference between Group A and Group B participants.
Generally, a participant’s mean accuracy score of Number-Sentences paralleled that of Control-Sentences (see Figure 4.2), with the exceptions of participants A4 and B4, representing two extremes on the spectrum. Participant A4 was particularly accurate when interpreting Number-Sentences, whereas participant B4 performed much better on Control-Sentences.
Participant A4 had a mean propositional accuracy score of .80 when he interpreted Number-Sentences, significantly higher than when he interpreted Control-Sentences (.51). One may argue that this means participant A4 is particularly good at interpreting numbers. However, without good interpreting skills in general, getting the number right alone will not help one produce accurate interpretation.
Furthermore, participant A4 did not think he performed better on speech segments containing numbers; in the questionnaire he gave himself the same rating of 3 on his overall performance and performance regarding speech segments with numbers. A more probable explanation is that participant A4 made more effort to interpret Number-Sentences in this study because he was instructed to interpret figures as accurately as possible. He might have been less alert when it came to Control-Sentences as they did not contain specific numbers. This was confirmed by
participant A4 in a followed up interview. He reported that he was tired at the time of experiment and did not perform as well as usual. But he made an effort to interpret numbers because of the instruction that figures should be interpreted as accurately as possible. In this case, the instruction introduced a bias. However, participant A4 was one of the two participants who habitually gave approximations when interpreting numbers. Without the specific instruction, participant A4 might have used approximations even when he still had enough processing capacity to interpret numbers exactly as given. That, in turn, would have caused an overestimation of the effects of numbers. Therefore, as mentioned in Chapter 4, it was necessary to instruct the participants to interpret numbers as accurately as possible for the purpose of this study.
In contrast, participant B4 performed much better on Control-Sentences (.84) than on Number-Sentences (.61). She indicated in the questionnaire that she was more nervous when interpreting speech segments containing numbers, and rated her performance on these segments worse than her overall performance. In the interview, participant B4 reported she was poor in mathematics and had developed a dislike of numbers in general since the fourth grade. In the course of interpreting training, she discovered that she often fumbled on numbers. This awareness made her more nervous and less confident when it came to interpreting speech segments with
numbers. Her case illustrates how students can have an aversion to numbers, which can be so strong that they panic when numbers appear (Wu, 2001). Commenting on numbers in the interpreting task, participant B4 said, “I just felt I would definitely mishear them as I always do.” Nevertheless, even for someone who dislikes numbers in general, not every number causes difficulty in interpreting. When asked if different types of numbers had the same effect on her in interpreting, participant B4 said that fractions did not pose a particular problem, but large numbers expressed in decimals did (e.g. 1.3 billion) because she often placed the decimal point in the wrong place when she noted down numbers. One possible explanation for misplacing the decimal point, as offered by participant B4 herself, is the interference of two number syntax systems. It probably occurs when one tries to convert the number in the target language while noting it down in the source language. The implication for student interpreters experiencing this kind of problem is that they should concentrate on noting down the number first before trying to convert it in the target language (Liu, 1993). In addition, it should also help that they strengthen their number conversion skills through rigorous practice. As recommended by many authors in the literature, student interpreters should try to automate the process as much as possible through extensive practice (Bao, 1998; Chang, 1999; Her, 1995; Huang, 2005; Liu, 1993; Mei, 2000). The basis for this recommendation is in line with findings from cognitive
science: once a task is automated, it can be performed with higher accuracy and less processing capacity (Sternberg, 2003; pp. 71-73), which is often near the brink of saturation in SI (Gile, 1999).
5.2 Numbers and Hesitation Phenomena
In addition to the effect of numbers on accuracy, this study also examined the effect of number on fluency, as measured by numbers of hesitation phenomena. The results showed that the presence of numbers introduced more hesitation phenomena in the interpretation. Furthermore, when number word syntax conversion was required, even more hesitation phenomena occurred in the interpretation. Because increases in the frequency of hesitation phenomena indicate greater processing requirements (Goldman-Eisler, 1972; Yang, 1998), the results suggest that numbers, especially those entailing syntax conversion, require greater processing capacity in SI.
The findings are consistent with results from past studies (Gotri, 2003; Her, 1995);
they also support the claim that the difference in number word syntax is a major cause of difficulty in number interpreting in the English-Mandarin combination (Chang, 1999; Mei, 2000).
However, a closer look at the results revealed that numbers involving no number syntax conversion did not necessarily increase the number of hesitation phenomena
when there was only one such number in a sentence. It was when two such numbers appeared in the same sentence that the number of hesitation phenomena increased substantially, as in the cases of Number-Sentences 6a, 14a and 16a (see Figure 4.3).
The results support Jones’s (1998) claim that numbers cause problems in SI when they come thick and fast. Note here that the numbers in these sentences referred to 50 years, 10 years, 50-60 dollars and 10-20%. For a listener, a sentence with two such numbers may not be considered as “thick with numbers,” and the listener may not feel that an increased processing capacity is required to understand the sentence.
For an interpreter, however, even two numbers appearing in the same sentence can considerably increase the processing requirement such that more hesitation phenomena ensue in the interpretation.
Number-Sentence 6a is also worth noting in that several participants reported experiencing difficulty understanding or handling the relationship of the two numbers in the sentence. The original was: “If the earth warms for the next fifty years at the rate of the last ten, we'll lose part of Manhattan Island.” Because the word
“year” was omitted after the word “ten,” as done in the original transcript of Clinton’s speech, three participants said they heard the word “ten” but spent sometime trying to figure out whether it meant 1o years or 10%. As a result, they missed the second part of the sentence. Another participant reported that he understood what “ten”
referred to, but it took him additional effort to search for the right sentence structure in Mandarin to convey the message. These findings point out that sometimes numbers per se may not be difficult; in some cases, it is the sentence structure in which they appear that causes problems in SI in the English-Mandarin combination.
The implication for training is that student interpreters should also familiarize themselves with those sentence structures typically associated with numbers if they want to handle numbers well during SI.
5.3 Failure Sequences
The study also found that the effects of numbers could sometimes carry over to the next sentence. Identical continuation sentences were interpreted less accurately when they were preceded by Number-Sentences than when they were preceded by Control-Sentences. Here again, however, the difference in accuracy was confounded by the difference in interpretation accuracy between Group A and Group B participants.
Nevertheless, the retrospective data provided solid evidence of failure sequences triggered by numbers. There were several cases where the participants identified the errors or omissions as attributable to the processing of numbers in a previous speech segment. Some of these failure sequences occurred in the same sentences in which
numbers appeared, while others occurred in following sentences.
The following describes an example of a failure sequence.
Number-Sentence 8a:
If we don't turn the trend around there will be one hundred million AIDS cases in five years, making it the worst epidemic since the Plague swept Europe.
B4:
如果我們不趕快做一些防治措施的話...(1.5s)在五年之內呢，我們大概就會
有...(1.4s)一億的人...(1.1s)染上愛滋病 [participial clause omitted]
Retrospection: (Presented with the source speech transcript) “I did not hear
‘in five years’…I did not hear the second part, ‘making it the worst epidemic since the Plague swept Europe’…I remember hearing the word “plague” but I wasn’t sure I heard it right….I think I spent more effort handling “100 million”… I noted the number and did the conversion.”
As indicated in her retrospective data, the processing of the number “100 million” took up much of her processing capacity. Consequently, she did not even register the second half of the sentence. Furthermore, the recording showed that the word “plague” in the source speech appeared during participant B4’s 1.4 seconds of silent pause. This suggests that her total capacity was saturated at the time such that
she noticed the source speech only when she stopped speaking.
As mentioned earlier, an examination of all the participants’ interpretation of critical sentences and continuation sentences revealed that some failure sequences occurred within the sentence while others happened in following sentences. Where the failure sequence occurred seems to be influenced by the sentence length, the participant’s EVS, among other factors. Although the critical sentences in this study were paired Number-Sentences and Control-Sentences, different pairs of critical sentences varied in sentence length. Numbers appearing in a long sentence, such as the example given above, were more likely to trigger a failure sequence within the sentence. Numbers appearing in a short sentence, such as Number-Sentence 3a
“When I left office there were three hundred and fifty million,” were more likely to trigger a failure sequence in the following sentence. Therefore, the comparison between the continuation sentences following Number-Sentences and those following Control-Sentences should only be seen as an indicator of whether the effects of numbers went beyond sentence boundaries. The comparison should not be mistaken for a capture-all measure of failure sequences.
5.4 Attitudes toward Numbers 5.4.1 Aversion
In the retrospective interviews, a few participants offered interesting explanations as to why they thought numbers posed difficulty for them.
Participant B4, as mentioned earlier, reported she was poor at mathematics and had developed a dislike of numbers in general since the fourth grade. In her daily life, she shunned away from numbers. For example, she always let her friends do the math when splitting a bill. Participant B3 also had similar comments about a dislike and avoidance of numbers in general. Interestingly, although participants B3 and B4 had comparable average accuracy scores on Control-Sentences (0.84, see Figure 4.2), participant B3 had an even higher score on Number-Sentences (0.86) while participant B4 showed a large drop in accuracy on Number-Sentences (0.61). The findings demonstrate that a dislike of numbers alone do not necessarily lead to poorer performance in number interpreting. The difference between the two participants’ performance can be interpreted in several ways. One possibility is that participant B3’s aversion of numbers is not as strong as participant B4’s. It is also possible that participant B4’s aversion of numbers in general is compounded by her experience of “always fumbling on numbers” during interpreting. Still another possibility is that participant B3 was more skilled at number syntax conversion than
participant B4, thus she had more processing capacity available to interpret the rest of the sentence. However, further research is required to better understand the relationship between number aversion and performance on number interpreting.
5.4.2 Inattention
Unlike participant B3 and B4, participant A3 cited “habitual inattention” to numbers as the primary reason that caused difficulty in number-interpreting. He reported that in his daily life when he heard or read a number, he usually did not try to identify what exactly the number was. To him, a number was more like a
“placeholder” rather than a concrete piece of information. A number signaled that there was a piece of quantitative information, which he could always look up later if necessary. For example, say he read a magazine article about the richest man in Taiwan, with a total asset of XX billion, the message he registered would simply be that this man was rich. He didn’t care and wouldn’t commit to consciousness exactly how much money this person had. “To me, it just means this person has a lot of money.” As he observed, this inattention to numbers was more pronounced when he listened to speeches in English, his second language. Participant A3 also reported that he had started to change this habit in order to capture numbers for interpreting purposes, but it still took him longer to register numbers in speech.
5.5 Observations of Number-Interpreting
Many participants in this study considered that it took a higher level of alertness to capture numbers because numbers could be easily missed in interpreting. This is consistent with the literature that have regarded numbers are one of those speech segments that are “more vulnerable to a momentary shortage of processing capacity in the listening effort because of their short duration or low redundancy” (Gile, 1997).
Regarding the influence of different types of numbers, the findings of this study support the claim that the difference between two number syntax systems is a major cause of difficulty in interpreting numbers in the English-Mandarin combination (Chang, 1999; Mei, 2000). Nevertheless, the retrospective data showed that not all numbers requiring syntax conversion were perceived as equally difficult by the participants in the study. More specifically, numbers in millions and billions that involved decimals (e.g. 1.3 billion) or the word “and” (e.g. three hundred and fifty million) were considered especially troublesome by several participants. One possible explanation to the increased difficulty of these numbers is that it takes longer time to translate longer numbers (Liu, 2006). Since interpreters are pressed for time during simultaneous interpreting, the longer processing time required by longer numbers is more likely to lead to disruptions or errors in interpreting.
In addition, one participant reported that he had noticed that he often mixed up
digits 4 and 5 when interpreting, for example, mishearing 24 million as 25 million.
As mentioned in Chapter 2, the same phenomenon was documented in Her’s study (1995). The phenomenon can be explained by spoken word recognition models with feature-detection components, such as cohort theory (Marslen-Wilson, cited in Massaro, 1994) or the TRACE model (McClelland & Elman, cited in Massaro, 1994).
In short, it can be easy to confuse “four” and “five” because the two words share an acoustic feature (the f-initial) and the same semantic category (numbers).
5.6 Summary of Key Findings
This study examined the effects of numbers on simultaneous interpreting, in terms of accuracy and fluency. Although the effects of numbers did not reach statistical significance possibly due to the small sample size, the results suggest some general trends. First, the results showed that not all numbers in the speech led to lower accuracy in simultaneous interpreting. It was numbers requiring number word syntax conversion that had a more observable impact on accuracy.
In terms of numbers’ effect on fluency, this study found that numbers requiring syntax conversion and numbers appearing in proximity increased the number of hesitation phenomena in the interpretation. This suggests that these numbers pose a heavier processing requirement for the interpreter.
The greater processing requirement of numbers can lead to omissions or errors in the speech segment containing numbers, or trigger a failure sequence (Gile, 1995, 1997, 1999), hindering the successful processing of the next speech segment. These failure sequences may occur within or outside sentence boundaries.
The results of this study also suggest that individual differences play an important role in determining the extent to which numbers affect interpreting performance. In particular, aversion or inattention to numbers in general can have an adverse effect on the performance of number interpreting.
