Chapter 4 Eye-movement Data Analyses
4.2 Eye movement data analysis
4.2.3 Hesitation Pause Analysis
4.2.3.2 Examination of hesitation pauses in the experienced interpreters’
eye-movement protocol
During 77.05% of the hesitation pauses, the experienced interpreters read and processed the segment interrupted by the pauses, most of the time they didn’t even fixate once the word they first utter when the oral output was resumed. Where did they look anyway? Could this finding suggest that most of the hesitation pauses made by the experienced interpreters were less likely to be triggered by lexical retrieval?
Most of the pauses made by the experienced interpreters were categorized as “Other”.
That is to say, not even one fixation was on the first uttered word, nor was the word near the fixation points enough to fall in the perceptual span.
Su (2013) argued that the 33 (29.20%) “Other” hesitation pauses may support Schilperoord’s (2002) claim that “the range of pause can cover more than the first word/ word produced after the pause (Su, 2013, p.79)”. Could the same thing be said about the 51 (41.80%) hesitation pauses in the “Other” group in the experienced
116
interpreters’ data? If the interpreters were not reading or processing the word
immediately uttered after the pause, what were they reading and processing? Could it be said that the pause was not caused by lexical retrieval? If so, what were the likely reasons for the output interruption? To answer the questions, all the fixations in the 51 hesitation pauses in the eye-movement protocol were checked against the annotated oral-output protocol.
Of the 51 hesitation pauses, fixations in 16 of them were more than two
segments to the left of the segment being translated when the pauses occurred. That is to say, if a hesitation pause occurred in segment N, all of its fixations were in segment N-2, sometimes even in segment N-3. The 16 pauses were all made by two of the experienced interpreters, and it was found that fixations in more than half of the pauses they made, hesitation or juncture, were more than two segments to the left of the outputted segment. It seemed that for unknown reasons, the interpreters had been constantly reading segment N-2 when interpreting N during the sight translation task.
Because the 16 pauses were all from two individual cases rather than being
representative of the general pattern among all subjects, they were excluded from the examination. Of the rest pauses, 5 were followed by nothing but a filler such as um or uh, making it impossible to find an ROI in the source text. They were also excluded from the examination.
Interestingly, each of the three paragraphs contained a segment where most of the experienced interpreters made a hesitation in the “Other” group. Following are examples of the interpreters’ eye-movement during the hesitation pause(s) in each of the segment.
117
(a) Paragraph 1
Ex 4-10 to 4-12 were all from the sight translation tasks of Paragraph 1, where most of the “Other” hesitation pauses occurred in the second segment of the sentence
“一個工廠裡的工人，他在裝配時候的心情對這輛車的品質影響不大。” The topic
“一個工廠裡的工人” of the sentence was followed by the real subject of the sentence
“他在裝配時候的心情”. Morever, the “對……” structure moved the object “這輛車”
to precede the verb “影響”. Both structures did not conform to the SVO structure of English.
The eye fixations during the pauses in Ex 4-10 to 4-12 were not on the word uttered after the pause, indicating that the interpreters were not likely looking for a correspondent word in English. What’s more, of all the words in the segment, they fixated those already translated. A possible explanation is that the interpreters read the translated part in the hope of formulating latter output that would work grammatically with the already outputted translation.
118
EX. 4-10
Source Text:一個工廠裡的工人，他在裝配時候的心情對這輛車的品質影響不 大。
Annotated Protocol: …uh a worker working in the factories when uh he when he’s on the assembly lines ^[6: 15694-16044: 349ms;裝配→時候] um how his his mood um has less to do with the quality of the car that he’s making.
Eye movement protocol:
一個工廠裡的工人，他在裝配時候的心情對這輛車的品質影響
48,49,50 51 52,53
EX. 4-11
Source Text:一個工廠裡的工人，他在裝配時候的心情對這輛車的品質影響不 大。
Annotated Protocol: …how he feels when he’ s doing assembling will ^[4:
17034-17280: 245ms;裝配→裝配] not affect greatly to quality of the car.
Eye movement protocol:
一個工廠裡的工人，他在裝配時候的心情對這輛車的品質影響
62
63, 64 65
119
EX. 4-12
Source Text:一個工廠裡的工人，他在裝配時候的心情對這輛車的品質影響不 大。
Annotated Protocol: …a worker in a factory has little influence ^[5:
12159-12801: 641ms;時候→時候] to the quality of the car um about his mood when he’ s assembling the part.
Eye movement protocol:
一個工廠裡的工人，他在裝配時候的心情對這輛車的品質影響
29 30,31 33
(b) Paragraph 4
Ex 4-13 to 15 were all from the sight translation tasks of Paragraph 4 by different interpreters. All the pauses occurred when they were translating the sentence “通常大 家總是在失去健康之後，才會關心健康。” Most interpreters outputted the second segment “才會關心健康” before the “失去健康之後” in the first segment. However, as they made change in segment order, pauses occurred. The fixations were not on the word first uttered after output resumption; the interpreters could fixate anywhere in the two segments of the sentence, mostly in the first segment, showing that the difficulty that triggered the hesitation pauses were not likely lexical retrieval.
120
EX. 4-13
Source Text:通常大家總是在失去健康之後，才會關心健康的問題。
Annotated Protocol: Generally speaking, people ^[20:40317-41512:1194ms;的
→問題→總是→大家→總是→之後] care about ^[21:42804-43360:556ms;失去
→健康→同時] the issue of health after losing health.
Eye movement protocol:
121
EX. 4-14
Source Text:通常大家總是在失去健康之後，才會關心健康的問題。
Annotated Protocol: Usually we will^[5:19558-19918:359ms;失去→大家] take care of our health after losing the health.
Eye movement protocol:
Annotated Protocol: Most of the time, people ^[7:20893-21327:434ms;失去→
健康] will start to care about health issues after losing their health.
Eye movement protocol:
Ex. 4-16 to Ex. 4-20 were all from the sight-translation of a segment in Paragraph 6 by different interpreters. The “Other” hesitation pauses all occurred during the interpretation of the segment “比如積極型的人適合獲利機會高但是風險也高的東 西”. The modifier of the object “東西” in this sentence was very long: “獲利機會高 但是風險也高”. Unlike the Chinese modifier, which was put before the object,
122
grammatically the modifier would be put after the object in English. The discrepancy in sentence structures was likely what caused the hesitation pause in Ex. 4-16, 4-17, and 4-18. The eye movement analysis showed that they interpreters didn’t fixate the first word they interpreted. Moreover, as showed in Ex. 4-16, 4-19 and 4-20, the structure of “someone 適合 something” in the Chinese segment may have also been problematic to the interpreters.
EX. 4-16
Source Text:比如積極型的人適合獲利機會高但是風險也高的東西。
Annotated Protocol: For example, aggressive people fit uh
^[7:20649-21607:956ms;適合→機會] the uh investment opportunities that are more profitable but also more risky.
Eye movement protocol:
同的投資方式，比如積極型的人適合獲利機會高但是風險也高
59
60 61
62
123
EX. 4-17
Source Text:比如積極型的人適合獲利機會高但是風險也高的東西。
Annotated Protocol: For example, for people who are more uh aggressive or active, um they are more suited to ^[8:32308-32696:387ms;獲利→適合]
approaches to that’s bring a higher profitability…
Eye movement protocol:
Annotated Protocol: Those aggressive people are good for those high risk but high um return ^[11:26542-26803:260ms;機會] investment
In the above three examples, it could be argued the interpreters were searching for an appropriate translation for the word “東西”. However, the interpreters fixated words in the preceding modifier rather than the word itself, which was in the row
124
below. They didn’t pause to search for an English word that corresponded to “東西”, i.e. they didn’t pause because of slow lexical retrieval, because their eyes hadn’t even reach the word “東西” in the next row. Rather, when they saw the long modifier, they knew an object needed to be inserted and were simply looking for an English object that worked with the modifier, or worked in the context.
EX. 4-19
Source Text:比如積極型的人適合獲利機會高但是風險也高的東西。
Annotated Protocol: For example, aggressive people ^[4:16242-16666:424ms;
積極型→積極型] are more suitable for uh opportunities that is more profitable but with higher risks.
Annotated Protocol: For example, aggressive people ^[6:18758-19824:565ms;
人→積極型→積極型→積極型] fit uh it’ s the uh investment opportunities that are more profitable but also more risky.
Eye movement protocol:
同的投資方式，比如積極型的人適合獲利機會高但是風險也高
52 53 54,55,56 57
In Ex. 4-19 and Ex. 4-20, the two interpreters paused before they produced a translation for the word “適合”, which could be translated literally as “suited to” or
“suitable for”. Judging from the oral output, they seemed to be searching the English word for “適合” during the pause. However, they didn’t fixate the verb “適合”; rather,
125
both of them fixated the modifier “積極型” of the subject in the segment.
The original Chinese sentence was written in the structure of “someone 適合 something”, but in English it is usually the other way round: “Something is suited to/suitable for someone.” The discrepancy may have caused the interpreters to hesitate and check information related to the subject of the segment. The fact that they didn’t fixate directly on the word ”適合” suggested that they probably weren’t having problem retrieving the corresponding “suitable for’, but were looking for context that helped them with making a sentence with the English word.
It could be argued that the word“適合”was also what triggered the pause in Ex 4-16, rather than the long modifier. Having read the first part of the segment “比如 積極的人適合”, the interpreters very quickly translated “適合” as “are suited to”.
However, the phrase may be syntactically correct; semantically, the meaning is turned the opposite way round. It could be that the interpreter hesitated because she had realized the error, which was why she refixated on the word “適合” during the hesitation pause. Nevertheless, as shown by her oral output, instead of making repairment, she decided to carry on.
Hesitation pauses were likely triggered by difficulty (Boomer & Dittman, 1962).
As the above examples of the “Other” hesitation pauses show, it was possible that certain segments in each paragraph were difficult to the experienced interpreters in that their grammatical structure had no straightforward equivalence in English grammar; i.e. the difficulty lied in syntactic transfer. As a result, the interpreters may have to insert something non-existent in the source text—or before they even fixated on its equivalence in the source text, as in the case of Ex. 4-16, 4-17, and 4-18. Or they may have to change the order of clauses or phrases to make grammatically correct sentences in English. Therefore, what they said after the pauses may not be what they fixated during the pauses.
126
Even when the difficulty was probably caused by a word, such as “適合” in Ex 4-16, 4-19, 4-20, it seemed that the experienced interpreters paused not because they couldn’t think of an equivalence in English, but because they wanted to find a translation that worked in the context. This suggested explanation is in line with Liu (2009) and Moser-Mercer (2010) that the experienced interpreters focused more on semantic links when interpreting.
However, to verify the above arguments about syntactic transfer and semantic links, further examination is needed. For example, qualitative end-product and error analysis could be conducted on the output by the experienced interpreters. It would be even better if the novices output could also be analyzed. Then segments translated by different groups could be compared one by one to see if the novices indeed show more signs of encountering lexical difficulties and the experienced interpreters more syntactic or semantic difficulties.
127
Chapter 5
Conclusion & Discussion
5.1 Findings from the oral data analysis
The oral data analysis was conducted to find out the number of juncture pauses and hesitation pauses made by the experienced interpreters (Chen, 2013) and by the novice interpreters (Huang, 2011; Su, 2013) in sight translation tasks. T-tests were conducted to see if there’s any within-group difference between the number of hesitation pauses and juncture pauses, and if there’s any inter-group difference between the two groups of interpreters with different amount of experience.
T-tests showed that while the novice interpreters made more hesitation pauses than juncture pauses in average, the experienced interpreters made as many juncture pauses as hesitation pauses. In addition, the experienced interpreter made significantly fewer juncture pauses and fewer hesitation pauses than the novices did. In addition, divergence in the number of hesitation pauses made by the experienced interpreter’s data was much larger than in the case of novices’. Moreover, the t-tests yielded a moderate correlation between the number of hesitation pauses and of juncture pauses made by the experienced interpreters.
5. 2 Findings from the eye-movement data analysis
In the eye movement analysis, fixations made during the pauses were marked out.
First, the number of fixations per interpreter made per hesitation or juncture pause was calculated and compared. T-tests showed that during the sight translation tasks, there was no significant difference in the number of fixations an interpreter made during a pause, regardless of the type of the pause and the experience of the interpreter. In addition, a moderate correlation was found between the number of fixations per experienced interpreter made during the two types of pauses.
128
Su (2013) had argued that the (novice) interpreters may need to process more information during juncture pauses than during hesitation pauses, in line with previous research on pauses in spontaneous speech (e.g. Boomer & Dittman, 1962;
Lounsbury, 1954). However, the argument was not supported by the t-test results of the number of fixations during the two types of pauses.
Next, four parameters were examined in the following order: (a) fixation’s position in passes; (b) saccade direction before and after pauses fixation’s location relative to output resumption; (c) fixation’s location relative to the segment being interrupted by the pause; (d) fixation’s location relative to the output resumed after the pasue. Fixations during juncture pauses and hesitation pauses were examined
respectively.
5.2.1 Findings from analysis of eye movements during juncture pauses A juncture pauses was a pause occurred when an interpreter finished the oral translation of a segment. During juncture pauses, the novices’ mostly fixated on the area to the right of the just-translated segment, indicating that they were likely processing yet-to-be-translated information. Further analysis of the fixations’
locations showed that the novices’ were mostly fixating on the first word or content word of the next-to-be translated segment, suggesting that they were likely preparing for the immediate output after the pause ended. Pass analysis supported the suggestion as it was discovered that most of the time, the novices were reading and
comprehending (in the first pass) or reformulating and mentally translating (in the second pass). The cognitive processing involved in the reading passes were postulated based on McDonald & Carpenter’s model (1981). Analysis of saccade direction at the onset and offset of the pauses showed that the proportion of juncture pauses starting with a regression was nearly as high as those starting with a forward saccade. All
129
forward saccades were in the first pass, while most of the regressions were in the second pass. Su (2013) suggested that the high regression rate could be a result of long read-ahead distance right before the juncture pauses. If the distance, i.e. the eye-voice span, was short, meaning the eyes were not far from the end of the
just-translated segment, the interpreters would probably continue to read forward for new information, thus starting the juncture pause with a forward saccade in the first pass. If the distance was long, the interpreters would probably regress to start for reformulation, thus starting the juncture pause with a regression in the second pass. Su (2013) counted the eye-voice span at the onset of juncture pauses, and the result supported her postulation, showing that in both cases, the interpreter was “somewhere in the process of planning for the next sentence [segment]” (p.69). The findings from the novices’ data were in line with suggestions by previous researcher that juncture pauses allowed the interpreters some time to plan for upcoming segments
(Goldman-Eisler, 1958; Rochester, 1973).
However, the eye-movement pattern of the experienced interpreters during juncture pauses was in stark contrast with the novices. During most of the juncture pauses, the experienced interpreters’ eye fixations were located in the just-translated segment rather than to the right of it. Only in very few cases did the interpreter fixated on the first word or content word in the next-to-be-translated segment. The results of the pause location analyses were in line with the analysis of the pass analysis, which showed that the fixations were the least likely to be in the first pass, indicating that most of the time, they were not reading new information. Saccade direction analysis
However, the eye-movement pattern of the experienced interpreters during juncture pauses was in stark contrast with the novices. During most of the juncture pauses, the experienced interpreters’ eye fixations were located in the just-translated segment rather than to the right of it. Only in very few cases did the interpreter fixated on the first word or content word in the next-to-be-translated segment. The results of the pause location analyses were in line with the analysis of the pass analysis, which showed that the fixations were the least likely to be in the first pass, indicating that most of the time, they were not reading new information. Saccade direction analysis