Chapter 3 Research Method
4.4 Behavior during the process of sight translation
4.4.4 Behavior of rereading and features in pauses
This section is comprised of data from five experienced interpreters, nine interpreting students, and seven untrained bilinguals. Due to uncontrollable head movements and other unforeseeable reasons — perhaps a reflection of the cognitive load experienced during sight translation — a little more than half participants’ eye movements didn’t land exactly on lines of words, though all passed initial calibration without a hitch and subsequent re-calibrations were conducted between trials when problems arose. To avoid over-speculation, a total of only 21 participants were included for further word-based examination, such as how many times one reread a word, and whether one was searching for information in a PBD or non-PBD unit.
Behavior of rereading
The mean and the standard deviation of rereading counts for all groups in all three tasks are presented below in Table 21.
Table 21 Rereading counts across groups in three tasks
No. of people Silent reading Reading aloud Sight translation
PRO 5 0.405 (0.564) 0.178 (0.093) 1.241 (0.735)
NEW 9 0.564 (0.536) 0.255 (0.221) 1.260 (0.910)
BIL 7 0.508 (0.208) 0.261 (0.139) 1.777 (1.023)
*PRO = Experienced interpreters; NEW = Interpreting students; BIL = Untrained bilinguals
Reading aloud seems to have the fewest rereading counts and a much smaller standard deviation in general. The need to speak out the written text perhaps offered some buffer, so the participants had slightly more time to process deeper during the first pass of reading. Sight translation, not unexpected, had much more traces of rereading. Although participants with training reread comparatively fewer times, but the differences between all three groups are not that obvious. Notwithstanding, interpretation of the results should be taken more conservatively.
Comparisons of rereading counts between groups in each task
An overall significant effect was absent across groups for rereading counts in silent reading, F(2,18) < 1, MSE = 0.213, p = .827. Experienced interpreters (M = 0.405) on average had fewer rereading counts than interpreting students (M = 0.564) and bilinguals (M = 0.508) respectively, but none of the gaps between groups reached significance.
Moving to reading aloud, experienced interpreters (M = 0.178) again had the lowest number of rereading counts per word, followed by interpreting students (M = 0.255) and bilinguals (M = 0.261). Again, none of the between-group gaps reached significance, F(2,18) < 1, MSE = 0.03, p = .672.
The pattern in sight translation was identical to that in reading aloud.
Experienced interpreters (M = 1.241) ostensibly re-fixated words after first leaving the region fewer times than interpreting students (M = 1.26) and bilinguals (M = 1.777). Though the gap between the trained and the untrained was much larger than in
the other two tasks, all gaps still happened by chance, according to inferential statistical analysis, F(2,18) < 1, MSE = 0.837, p = .48.
It was previously inferred that the trained participants must have had fewer passes of reading to finish sight translation with similar duration in all reading indices but significantly less total time and fewer number of fixations. The results for
rereading counts, albeit failing to show a meaningful difference, are still clear with the numerical differentials. It might be that the insignificant gap made a real difference in the end when added up. Still another possibility is that trained participants did
sometimes have fewer passes of reading but was not frequent enough when evened out across four trials; on the other hand, the practice of chunking helped trained participants find suitable correspondents more easily. The above two reasons, combined together, explain the significantly fewer total fixations and less total time but only numerically fewer rereading counts for trained participants. Or, it might simply be that the samples gathered for this detail analysis are not representative.
Comparisons of rereading counts between tasks for each group
For experienced interpreters, silent reading required an average of 0.405 times of rereading on each word. The number was close to and slightly higher than reading aloud (M = 0.178), but was largely fewer than sight translation (M = 1.241). There was an significant overall effect, F(2,12) = 5.418, MSE = 0.289, p = .021, but the effect came solely from the significant gap between reading aloud and sight translation, p = .028.
Interpreting students manifested the same pattern as experts. The average count in silent reading (M = 0.564) was a bit higher than reading aloud (M = 0.255) and much lower than sight translation (M = 1.26). Welch’s F-test established a significant main effect, F(2,12.674) = 5.741, p = .017, and the only significant gap stood between reading aloud and sight translation, p = .026.
Bilinguals perhaps adequately demonstrated how situations would pan out for average people without training. On the face of it, reading aloud (M = 0.261) was again the lowest, followed by silent reading (M = 0.508), while sight translation (M = 1.777) occupied the other end of the spectrum. Yet, this time, with a significant main effect confirmed by Welch’s F-test, F(2,10.263) = 9.614, p = .004, the only
non-significant gap was between silent reading and reading aloud. The average rereading count of reading aloud was significantly lower than that of sight translation (also true for the other two groups), p = .018. What’s more, silent reading
significantly distanced itself from sight translation as well, p = .038.
Due to the subtask of vocalization required for reading aloud, eyes were
compelled to linger a bit longer on words, thereby reducing the need to reread words.
Hence, for all three groups, reading aloud had the lowest number of rereading counts.
The narrower gap between silent reading and sight translation across all groups serves to point out one more similarity between these two tasks. However, perhaps the thing worthy of most attention is that, participants with training on average managed to narrow the gap between silent reading and sight translation to a non-significant level.
This is indeed amazing, since the implication is that experts and trainees actually
reduced the repetitive needs of reading for reformulation to a level that’s close to pure reading.
Features in pauses
The following sub-section moves onto features in pauses. We first divided them into hesitation pause and juncture pause and then calculated their percentage. The mean of each type of pause for all groups are summarized below in Table 22.
Table 22 The percentage of hesitation pause vs. juncture pause among all the pauses made by participants across groups when performing sight translation
N Pause type Mean (%) SD
PRO 5 Hesitation pause 62.816 10.882
Juncture pause 37.184 10.882
NEW 9 Hesitation pause 55.384 2.575
Juncture pause 44.616 2.575
BIL 7 Hesitation pause 59.034 5.391
Juncture pause 40.966 5.391
*PRO = Experienced interpreters; NEW = Interpreting students; BIL = Untrained bilinguals
The percentage of hesitation pause was consistently higher than that of juncture pause for all groups. What may be counter-intuitive is the highest percentage of hesitation pause for experts, but one thing to bear in mind is that the figures here represent only the percentage, regardless of the actual number of pauses in each category.
Comparisons of different types of pauses between groups
In this paragraph, only results for hesitation pause were compared. We believe this will suffice, since the two types of pauses together constitute 100% for each person. According to Table 22, experienced interpreters (M = 62.816%) on average had higher percentage of hesitation pause than bilinguals (M = 59.034%), while interpreting students’ (M = 55.384%) proportion was the lowest. Still, an overall main effect was found non-significant by Welch’s F-test, F(2,7.36) = 2.142, p = .185.
Comparisons of different types of pauses for each group
For experienced interpreters, a dependent two-tailed t-test was adopted to examine whether the percentage of hesitation pause and that of juncture pause were different and found a non-significant main effect, t(4) = 2.633, p = .058. Although hesitation pauses (M = 62.816%) seemed overwhelmingly more than juncture pauses (M = 37.184%), statistical analysis determined that the gap simply showed itself by chance.
Interpreting students, in contrary to experts, turned out to have a significant gap between the two kinds of pauses. We can see that the hesitation pause (M = 55.384%) was generally higher in percentage than the juncture pause (M = 44.616%). Though the superficial pattern looked exactly like that of experts’, even with a narrower gap in between, a dependent two-tailed t-test returned a significant difference, t(8) = 6.272, p
< .001.
Bilinguals completely replicated the pattern of trainees. The hesitation pause on average accounted for 59.034% of all the pauses, while the other 40.966% were juncture pauses. A dependent two-tailed t-test found a significant difference between the two categories, t(6) = 4.433, p = .004.
Interpreting students’ and untrained bilinguals’ significantly higher percentage of hesitation pause shown in Table 22 demonstrate that the cognitive burden of sight translation was huge enough to grab their full attention and did not leave enough resources for them to tackle pauses with finesse. Therefore their struggles were faithfully reflected through hesitation pauses. On the other hand, despite the highest percentage of hesitation pause, experienced interpreters indeed had the lowest number of hesitation pauses. What’s more, the percentage of each type of pause was not statistically different from the other for this group. This means while experts as a group had smooth, fluent, and accurate rendition with really few pauses, these participants were able to control themselves and pause at the right time so as not to create dissonance in the ears of the audience.
Fixations on principal branching direction (PBD) units in pauses
In previous sections it was found that there was only one fixation during 75% of the pauses for all three groups, while participants were gathering information to a different extent in the rest 25% (see Table 19). What were the participants reading in order to keep their rendition going? As we have mentioned, Chinese and English were contrastive in the principal branching direction of adjuncts. Would these adjuncts
create problems in sight translation? Further, would there be any difference in the two distinctive kinds of pauses? To answer the questions above, the percentage of
fixations on PBD units in pauses, and further, in hesitation pauses and juncture pauses were calculated respectively, and the mean and standard deviation for all groups are summarized below in Table 23.
Table 23 The percentage of fixations on PBD units of all fixations in pauses (percentage in hesitation pauses and juncture pauses also tabulated individually)
N Percentage of fixations Mean (%) SD
PRO 5
Fixations on PBD units in pause 62.479 7.006 Fixations on PBD units in HP 64.547 11.570 Fixations on PBD units in JP 58.610 15.698
NEW 9
Fixations on PBD units in pause 56.833 10.292 Fixations on PBD units in HP 61.443 14.350 Fixations on PBD units in JP 49.631 10.745
BIL 7
Fixations on PBD units in pause 59.001 11.058 Fixations on PBD units in HP 61.693 15.957 Fixations on PBD units in JP 56.772 8.186
*PBD = Principal branching direction
*PRO = Experienced interpreters; NEW = Interpreting students; BIL = Untrained bilinguals
*HP = hesitation pause; JP = juncture pause
Principal branching direction (PBD) units attracted more attention in hesitation pauses than in juncture pauses: The percentage of fixations on PBD units was on average over 60% of all fixations in each hesitation pause for every group and was lower in the other type of pause.
Comparisons of percentage of fixations on principal branching direction (PBD) units in different types of pauses between groups
For experienced interpreters, on average, 62.479% of the fixations in each pause were stapled to PBD units, while it was 56.833% for interpreting students and
59.001% for bilinguals. No significant main effect was found, F(2,18) < 1, MSE = 98.743, p = .603.
We delved deeper to see if the percentage of fixations on PBD units in each type of pause would differ across groups. As stated in Table 23, experienced interpreters had 64.547% of their fixations on PBD units during hesitation pauses, while it was 61.443% for interpreting students, almost identical to bilinguals’ 61.693%. Still, the overall effect was non-significant, F(2,18) < 1, MSE = 206.136, p = .920. In terms of fixations on PBD units in juncture pauses, experienced interpreters had the highest percentage (M = 58.61%), followed by bilinguals (M = 56.772%) and then
interpreting students (M = 49.631%), exhibiting the exact same pattern as in the other type of pause. Once again, the main effect did not reach significance, F(2,18) = 1.29, MSE = 128.412, p = .299, making all groups similar in this regard.
All the between-group comparisons in the section regarding behavior of rereading and features in pauses failed to show a significant effect, denoting that reading behavior and strategies during pauses were similar across groups. Though the percentage of hesitation pause of experts was the highest, we should keep in mind that the actual number of pauses of this group was in fact the fewest and was around only
half when compared with bilinguals. On the other hand, the generally higher
percentage of fixations on PBD, regardless of groups and pause types, is a testimony to the tricky feature of this kind of structural disparity in English-Chinese sight translation.
Comparisons of percentage of fixations on principal branching direction (PBD) units in different types of pauses for each group
As in Table 23, it appears that on average there was a bigger chance of fixating on PBD units in hesitation pauses (M = 64.547%) than in juncture pauses (M = 58.61%) for experienced interpreters. However, a dependent two tailed t-test failed to show a significant difference, t(4) < 1, p = .585. Interpreting students tended to fix their eyes upon PBD units more frequently in hesitation pauses (M = 61.443%) than in juncture pauses (M = 49.631%) as well. This time, a dependent two-tailed t-test found that the difference just crossed the threshold of significance level, t(8) = 2.304, p = .05. As for bilinguals, the probability of PBD units attracting the eyes was higher
in hesitation pauses (M = 61.693%) than in juncture pauses (M = 56.772%) as well, but a dependent two-tailed t-test failed to find a significant difference, t(6) = 1.227, p
= .266.
Hesitation pauses appear when the process of the sight translation cycle
(reading-reformulation-production loop) is interrupted. Judging from the data at hand, when faced with difficulties, trainees did have a more consistent tendency to look at PBD units for clues. On the other, experts and bilinguals did not manifest this distinct
pattern. But still, the percentage of fixations on PBD units in hesitation pauses on average exceeded 60% for all the groups.
The PBD percentage in juncture pauses was significantly lower for trainees. It’s somewhat unexpected, since when one successfully finishes the previous segment, it’s natural to pre-read tricky structures that lie ahead to ensure a smooth rendition when moving onto the next block of information. Interpreting students, as high-achieving language users, should know better that the structural differences between Chinese and English might present great difficulties that “ruin the show”. Therefore, it’s normal for experts and bilinguals to look at PBD units around 57% of the time in juncture pauses. As for interpreting students, we can only speculate that the training drills were so deeply ingrained in their processing habits that they adhered strictly to the skill of chunking and focused entirely on finishing sight-translating the target segment.
As a matter of fact, contrastive PBD units in this experiment never appeared right after a period (and few followed a comma except for sentential PBD units).
Therefore, trainees sticking to the approach taught in training did indeed have fewer opportunities to pre-read PBD units in juncture pauses. Nonetheless, it’s worth pointing out that, even in the case of interpreting students, the PBD percentage still reached almost 50%.
Comparisons of processing load between non-PBD and PBD units for each group A generally higher percentage of fixations on PBD units in pauses prompted the
comparisons of processing load between non-PBD and PBD units. Two-tailed dependent t-tests were used to examine whether the average fixation duration of one type of unit differed from the other regarding different indices within each group.
The results for the task of silent reading are summarized in Table 24. For experienced interpreters, almost all indices showed longer mean fixation duration on PBD units than on non-PBD units, except for RRT (M = 266.997 for the PBD vs. M = 289.181 for the non-PBD). However, statistical analyses did not find any significant difference between the two types of units on any index.
The pattern shown by experts was completely replicated by interpreting students.
RRT was the only index with which the mean fixation duration on PBD units was shorter (M = 367.093 vs. M = 392.482 for non-PBD units). Dependent t-tests again did not find any significant difference between non-PBD and PBD units (fore brevity and clarity here, only those unique phenomena worthy of notice were supplied with statistical results. Please see Appendix V for the rest of the details concerning this whole subsection, including the actual numbers of statistical values).
Untrained bilinguals, on the other hand, consistently had longer mean duration on PBD units than on non-PBD units. What’s more, a dependent t-test found that GD on the baseline (i.e. non-PBD units) (M = 298.983) was significantly shorter than that on PBD units (M= 327.851), t(6) = -2.739, p = .034. Baseline RRT (M = 321.147) was also significantly shorter than RRT on PBD units (M = 386.138), t(6) = -2.715, p
= .035.
Table 24 Mean fixation duration on non-PBD and PBD units in silent reading across groups (unit: millisecond)
Compared with average words, units embedded in contrastive structures (i.e.
PBD units) required bilinguals to spend significantly more time when it comes to meaning interpretation in a relatively later stage of the first pass and meaning integration in non-first passes. This group of participants exhibited a more effortful behavior when making connections between antecedents and adjuncts were constantly necessary to figure out the message conveyed. Notwithstanding, the overall time needed to comprehend words in contrastive structures did not pose a significantly larger challenge. On the contrary, processing of baseline and PBD units seemed similar for both experts and trainees.
PRO NEW BIL
*PRO = Experienced interpreters; NEW = Interpreting students; BIL = Untrained bilinguals
*FFD = First fixation duration; GD = Gaze duration; GPT = Go-past time; RRT = Rereading time; TVT = Total viewing time
*PBD = Principal branching direction
Table 25 presented results in reading aloud. For experienced interpreters, PBD units required comparatively less time in the first pass but more time in the second pass and beyond. Nonetheless, dependent t-tests confirmed that non-PBD and PBD units took statistically similar amount of time across all stages of processing.
Interpreting students, except FFD, had consistently longer duration on PBD units regarding all indices. Still, statistical analyses came back with a non-significant effect for all indices. Untrained bilinguals spent less time on PBD units when FFD and GPT are considered, while GD, RRT, and TVT on the same type of unit were prolonged.
However, no significant effect was found for this group in reading aloud, either.
Table 25 Mean fixation duration on non-PBD and PBD units in reading aloud across groups (unit: millisecond)
*PRO = Experienced interpreters; NEW = Interpreting students; BIL = Untrained bilinguals
*FFD = First fixation duration; GD = Gaze duration; GPT = Go-past time; RRT = Rereading time; TVT = Total viewing time
*PBD = Principal branching direction
Lastly, the focus lands on sight translation, in which the effects of contrastive syntactic structures were suspected to be even larger, due to the necessity of
reformulation. Table 26 provides the general results in sight translation. Experienced interpreters and interpreting students differed only in that GPT on non-PBD units for the latter group was slightly longer. This aside, PBD units appeared to have
consistently taken longer time to process for both groups. Yet again, statistical analyses demonstrated that the two types of units took just similar amount of time.
When it comes to untrained bilinguals, the numerical pattern also indicated a longer time for all indices on PBD units, but statistical analyses showed diverse results from the other two groups. For one, GD on non-PBD units (M = 264.086) was significantly shorter than that on the PBD counterparts (M = 304.959), t(6) = -4.876, p
= .003. In addition, RRT also turned out to be significantly shorter on baseline units (M = 537.316) than on words embedded in PBD structures (M = 625.67), t(6) = -2.75, p = .033. The same applied to TVT: The mean duration was also significantly shorter on non-PBD units (M = 623.06) than on PBD units (M = 722.066), t(6) = -3.455, p
= .014.
Table 26 Mean fixation duration on non-PBD and PBD units in sight translation across groups (unit: millisecond)
From the numerical patterns in all three kinds of tasks, we have witnessed that divergent PBD structures in English and Chinese demanded longer time to process,
From the numerical patterns in all three kinds of tasks, we have witnessed that divergent PBD structures in English and Chinese demanded longer time to process,