Independent-sample t-tests were conducted to compare the translators’ and the interpreters’ speeds during the drafting phase, the revising phase, and the entire translation process (i.e. the drafting phase combined with the revising phase) of Text 2’s translation, which prevented the participants from touching correction keys (see Table 20).
During the drafting phase, there was no significance difference in the speeds for the translators (M=1,093.00, SD=477.67) and the interpreters (M=817.56, SD=261.49);
t(26)=1.99; p=.06. These results suggest that the interpreters did not translate faster than the
translators during the drafting phase.During the revising phase, there was no significance difference in the speeds for the translators (M=773.10, SD=278.46) and the interpreters (M=724.78, SD=183.45); t(26)=0.55,
p=.58. These results suggest that the interpreters did not translate faster than the translators
during the revising phase.Not surprisingly, during the entire translation process, which consisted of the drafting
phase and the revising phase, there was no significance difference in the speeds for the translators (M=1,866.10, SD=612.32) and the interpreters (M=1,542.33, SD=304.74);
t(26)=1.88; p=.14. These results suggest that the interpreters did not translate faster than the
translators when translating Text 2.In short, although on average the interpreters spent 17% less time (1542.33 seconds) than the translators did (1,866.10 seconds), the t test shows that such a difference is not significant.
Table 20. Time Spent on Translating Text 2 by Translators and Interpreters
T/I N Mean Std. Deviation
Interpreter 18 1542.33 304.74
Independent-Sample t-Test Levene's Test for
Equality of Variances t-test for Equality of Means
F Sig. t df
Equal variances
not assumed 1.69 12.07 .12 275.44
Revise 2 (sec)
Equal variances assumed
1.34 .26 .55 26 .58 48.32
Equal variances
not assumed .49 13.45 .63 48.32
Total 2 (sec)
Equal variances
assumed 7.22 .01 1.88 26 .07 323.77
Equal variances
not assumed 1.57 11.53 .14 323.77
Comparing Text 1 and Text 2: Effects of Correction Keys
The previous statistical analyses suggest that when participants were not able to use the correction keys (i.e. Backspace, Delete, Ins, and the mouse) during the drafting phase, the overall speed differences between the two groups seemed to have vanished. The following analyses compare temporal changes between the translation processes of Text 1 and Text 2 to identify patterns.
First of all, paired samples t-tests were conducted to compare the 28 participants’ speeds between Text 1’s and Text 2’s translation processes (see
Table 21).
There was a significant difference in the time spent on Draft 1, i.e. the drafting phase of Text 1’s translation, (M=1,075.82, SD=341.13) and Draft 2, i.e. the drafting phase of Text 2’s translation, (M=915.93, SD=370.37), t(27)=2.78, p=.01. These results suggest that the participants translated faster in Draft 2 than in Draft 1.
Table 21. Time Spent on Translating the Two Texts by the 28 Participants
Paired Samples t-Test StatisticsMean N Std. Deviation
Pair 1 Draft 1 (sec)
1075.82
28 341.13Draft 2 (sec)
915.93
28 370.37Pair 2 Revise 1 (sec) 663.57 28 317.38
Revise 2 (sec) 742.04 28 218.16
Pair 3 Total 1 (sec) 1739.39 28 565.70
Total 2 (sec) 1657.96 28 456.52
Paired Sample t-Test
Paired Differences
Mean Std. Deviation
95% Confidence Interval of the Difference Lower Upper Pair 1 Draft 1 (sec) - Draft 2 (sec) 159.89 302.73 42.51 277.28 Pair 2 Revise 1 (sec) - Revise 2 (sec) -78.46 283.50 -188.40 31.47 Pair 3 Total 1 (sec) - Total 2 (sec) 81.43 345.77 -52.65 215.50
Paired Samples t-Test
t df Sig. (2-tailed)
Pair 1 Draft 1 (sec) - Draft 2 (sec) 2.78 27
.01
Pair 2 Revise 1 (sec) - Revise 2 (sec) -1.47 27 .16
Pair 3 Total 1 (sec) - Total 2 (sec) 1.25 27 .22
Then, the same paired samples t-tests were conducted in the translator group (see Table
22).
There was no significant difference in the time spent on Draft 1 (M=1,196.20,
SD=385.93) and Draft 2 (M=1093.00, SD=477.67); t(9)=0.81, p=.44. These results suggest
that the translators did not translate faster in Draft 2 than they did in Draft 1.There was no significant difference in the time spent on Revise 1 (i.e. the revising phase of Text 1’s translation) (M=864.50, SD=370.67) and Revise 2 (i.e. the revising phase of Text 2’s translation) (M=773.10, SD=278.46); t(9)=1.00, p=.34. These results suggest that the translators did not translate faster in Revise 2 than they did in Revise 1.
Not surprisingly, there was no significant difference in the time spent on Total 1 (i.e. the drafting phase and the revising phase of Text 1’s translation) (M=2,060.70, SD=677.02) and Total 2 (i.e. the drafting phase and the revising phase of Text 2’s translation) (M=1,866.10,
SD=612.32); t(9)=1.49, p=.17. These results suggest that overall the translators did not
translate faster on Text 2 than they did on Text 1.Table 22. Time Spent on Translating the Two Texts by the 10 Translators
Next, the same paired samples t-tests were conducted in the interpreter group (see Table
23).
There was a significant difference in the time spent on Draft 1 (M=1,008.94, SD=304.48) and Draft 2 (M=817.56, SD=261.49); t(17)=3.40, p=.00. These results suggest that the interpreters translated faster in Draft 2 than they did in Draft 1.
There was also a significant difference in the time spent on Revise 1 (M=551.94,
SD=224.27) and Revise 2 (M=724.78, SD=183.45); t(17)=-3.06, p=.01. These results suggest
that the translators translated more slowly in Revise 2 than they did in Revise 1.When both phases were taken together, however, there was no significant difference in the time spent on Total 1 (M=1,560.89, SD=413.69) and Total 2 (M=1,542.33, SD=304.74);
t(17)=.26, p=.79. These results suggest that overall the interpreters did not translate faster on
Text 2 than they did on Text 1.In short, the interpreters sped up significantly in the drafting phase and slowed down significantly in the revising phase, resulting in no overall speed change between Text 1 and Text 2.
Table 23. Time Spent on Translating the Two Texts by the 18 Interpreters
Paired Samples t-Test StatisticsMean N Std. Deviation
Pair 1 Draft 1 (sec)
1008.94
18 304.48 Draft 2 (sec)817.56
18 261.49 Pair 2 Revise 1 (sec)551.94
18 224.27 Revise 2 (sec)724.78
18 183.45 Pair 3 Total 1 (sec) 1560.89 18 413.69 Total 2 (sec) 1542.33 18 304.74Paired Samples t-Test
df Sig. (2-tailed)
Pair 1 Draft 1 (sec) - Draft 2 (sec) 17
.00
Pair 2 Revise 1 (sec) - Revise 2 (sec) 17
.01
Pair 3 Total 1 (sec) - Total 2 (sec) 17 .79
Paired Samples t-Test
Paired Differences
t Mean
Std.
Deviation
95% Confidence Interval of the
Difference Lower Upper
Pair 1 Draft 1 (sec) - Draft 2 (sec) 191.39 239.17 72.45 310.32 3.40 Pair 2 Revise 1 (sec) - Revise 2 (sec) -172.83 239.50 -291.94 -53.73 -3.06 Pair 3 Total 1 (sec) - Total 2 (sec) 18.56 295.38 -128.33 165.44 .26
These t-test results, however, seem contradictory. On the one hand, if the interpreters were faster than the translators when translating Text 1, but they were not faster than the translator group when translating Text 2, as suggested by the t-tests, the possibly explanations were that the interpreter group slowed down and/or the translator group sped up. On the other hand, the intra-group t-tests suggest that the interpreters did not translate Text 2 faster than themselves did Text 1, nor did the translators.
First, the speed difference between the two groups on the translation task of Text 1 is obvious and solid, so some speed change must have occurred when the participants translated Text 2. So which was more likely? The interpreters slowed down or the translator sped up?
The descriptive statistics seems to suggest the latter is more likely. In contrast to the interpreters, who spent almost the same time on translating both texts (1,560.89 seconds on Text 1 and 1,542.33 seconds on Text 2), the translators spent 9.44% less time on translating Text 2 than they did on translating Text 1 (2,060.70 seconds on Text 1 and 1,866.10 seconds on Text 2) (see Table 24).
Table 24. Time Spent on Translation the Two Texts by the Translators and the Interpreters
Mean N Std. Deviation
Translators Total 1 (sec)
2060.70
10 677.02 Total 2 (sec)1866.10
10 612.32 Interpreters Total 1 (sec) 1560.89 18 413.69 Total 2 (sec) 1542.33 18 304.74Correlations between Drafting Speed and Overall Translation Speed
The data also suggest a significant positive correlation between drafting time and translation time on Text 1 as well as on Text 2 (see Table 25).
Table 25. Time Spent on Drafting and the Entire Translation by the 28 Participants
Participant Code Draft 1 (sec)Total 1 (sec)
Draft 2 (sec)Total 2 (sec)
002-120504 1495
2445
20112752
003-120507 1290
1607
10921556
004-120717-ZSF 917
1336
7451624
005-120913-ZZY 804
1071
9811335
006-120920-LXH 1767
3039
11992140
008-121005-YCY 1631
2531
14012214
009-121017-YYX 1181
2333
17032402
010-121022-LWR 1558
2157
9611801
011-121025-PE 1269
1755
6201643
012-121102-XWX 825
1572
6991271
014-121107-WYY 761
1965
6091451
015-121114-CHY 1695
2680
9592169
016-121130-GHJ 1244
2347
12772318
017-121205-LXW 1062
1867
11322162
018-121211-WSY 922
1313
8791755
019-121212-GTJ 908
1257
5221158
020-121215-XTN 927
1509
7171170
021-121218-CDY 1266
1970
8161619
022-121221-FJM 969
1183
7221341
023-121224-HSH 729
1395
6551510
024-121228-XRX 923
1533
11321719
025-130102-HZJ 542
1462
3231028
026-130110-LYL 1144
1915
8731499
027-130114-ZBY 629
1030
7271345
029-130130-WXY 678
984
5761484
031-130227-ZJY 1017
1471
7391453
032-130304-SGL 1337
2178
10781756
033-130328-WP 633
798
498748
Quantitative analyses reveal strong Pearson correlations between the time spent on drafting and the time spent on the entire translation process on Text 1 (r=.87,) as well as on Text 2 (r=.88), when the correction keys are disabled during the draft phase (see Table 26 and
Table 27).
Table 26. Pearson Correlations between Time Spent on Different Phases of Draft 1’s Translation by the 28 Participants
Draft 1 (sec) Revise I (sec) Total 1 (sec)
Draft 1 (sec) Pearson Correlation 1 .48*
.87
**Sig. (2-tailed) .01 .00
N 28 28 28
*. Correlation is significant at the 0.05 level (2-tailed).
**. Correlation is significant at the 0.01 level (2-tailed).
Table 27. Pearson Correlation between Time Spent on Different Phases of Draft 2’s Translation by the 28 Participants
Draft 2 (sec) Revise 2 (sec) Total 2 (sec)
Draft 2 (sec) Pearson Correlation 1 .15
.88
**Sig. (2-tailed) .46 .00
N 28 28 28
*. Correlation is significant at the 0.05 level (2-tailed).
**. Correlation is significant at the 0.01 level (2-tailed).
Further intra-group analyses were carried out.
Quantitative analyses of the translator group reveal strong Pearson correlations between the time spent on drafting and the time spent on the entire translation process on Text 1 (r=.90,) as well as on Text 2 (r=.90), when the correction keys were disabled during the draft
phase (see
Table 28
andTable 29).
Table 28. Pearson Correlations between Time Spent on Different Phases of Draft 1’s Translation by the 10 Translators
Draft 1 (sec) Revise I (sec) Total 1 (sec)
Draft 1 (sec) Pearson Correlation 1 .60
.90
**Sig. (2-tailed) .07 .00
N 10 10 10
*. Correlation is significant at the 0.05 level (2-tailed).
**. Correlation is significant at the 0.01 level (2-tailed).
Table 29. Pearson Correlations between Time Spent on Different Phases of Draft 2’s Translation by the 10 Translators
Draft 2 (sec) Revise 2 (sec) Total 2 (sec)
Draft 2 (sec) Pearson Correlation 1 .26
.90
**Sig. (2-tailed) .47 .00
N 10 10 10
*. Correlation is significant at the 0.05 level (2-tailed).
**. Correlation is significant at the 0.01 level (2-tailed).
Quantitative analyses of the translator group reveal strong Pearson correlations between the time spent on drafting and the time spent on the entire translation process on Text 1 (r=.90,) as well as on Text 2 (r=.90), when the correction keys were disabled during the drafting phase (see Table 30 and Table 31).
Table 30. Pearson Correlations between Time Spent on Different Phases of Draft 1’s Translation by the 18 Interpreters
Draft 1 (sec) Revise I (sec) Total 1 (sec)
Draft 1 (sec) Pearson Correlation 1 .21
.85
**Sig. (2-tailed) .41 .00
N 18 18 18
*. Correlation is significant at the 0.05 level (2-tailed).
**. Correlation is significant at the 0.01 level (2-tailed).
Table 31. Pearson Correlations between Time Spent on Different Phases of Draft 2’s Translation by the 18 Interpreters
Draft 2 (sec) Revise 2 (sec) Total 2 (sec)
Draft 2 (sec) Pearson Correlation 1 -.10
.80
**Sig. (2-tailed) .71 .00
N 18 18 18
*. Correlation is significant at the 0.05 level (2-tailed).
**. Correlation is significant at the 0.01 level (2-tailed).
The Pearson correlations from the quantitative data seem to suggest that the speed of drafting is a reliable indicator of the overall translation speed even when the drafting process is deliberately manipulated and accelerated.
If so, how come there is a significance speed increase in the drafting phase but not in the entire translation of Text 2?
One possible explanation is that the acceleration during the drafting phase of Text 2 is relatively small. If there have been a bigger acceleration—for instance, if the participants have been more familiar with drafting without using correction keys—the entire translation process might have sped up significantly. The retrospective interviews confirmed that the participants are not familiar with the constraint even after the warm-up session.
Indeed, although SPSS suggested no significant difference between the time spent on translating Text 1 and Text 2, there is a tendency for participants to speed up when translating Text 2; in particular, while on average a translator spent 2,060.70 seconds on translating Text 1, s/he spent 1,866.10 seconds, or 9.44% less time, on translating Text 2 (see Table 32).
It is possible that if they have spent more time practicing the
“no-correction-when-drafting” technique, the speed increase in the translator group would
have been more obvious.
Table 32. Time spent on Each Phase by the Translators and the Interpreters
9Translators (Mean)
18
Interpreters (Mean)
27
Participants (Mean) Pair 1 Draft 1 (sec) 1196.20 1008.94 1075.82
Draft 2 (sec) 1093.00 817.56 915.93 Pair 2 Revise I (sec) 864.50 551.94 663.57 Revise 2 (sec) 773.10 724.78 742.04 Pair 3 Total 1 (sec)
2060.70
1560.89 1739.39Total 2 (sec)
1866.10
1542.33 1657.96Effects of Correction Keys: Backspace, Del, Ins, Mouse, and Four Arrows
Because Translog could not record any Chinese character, the videos recorded (shown on the left side of the window below) are essentially blank. The keystroke logging is shown on the right side. It should be noted that this program could not record any keystroke occurred during the Chinese character conversion. In other words, the keystrokes recorded are those that occurred either before or after the character conversion.
Figure 6. Snapshot of Translog Logging File
The previous section demonstrated that the drafting speed could be used to predict the overall translation speed. This section, by utilizing the data recorded on Translog during the drafting process, analyzes the use of correction keys—keys that might affect the drafting speed—during the drafting phase.
The initial design of this experiment meant to define “correction keys” as those that the participant used to “correct” his/her translation and hence interrupt his/her linear flow of translation, namely the mouse and the backspace, deletion, insert, and the four arrow keys.
However, the pilot test revealed that the four arrow keys are essential for candidate selection of certain input methods such as New Phonetic. Therefore, the researcher narrowed down the scope of “correction keys” in this experiment to only the mouse and the backspace, deletion, and insert keys, which are disabled during the draft phase of Text 2’s translation.
Based on the data collected using Translog (i.e. the input interface during the draft phase of both Text 1 and Text 2), a moderate Pearson correlation, .54, existed between the sum of the seven correction keystrokes (NOTE: Translog did not record Insert keystrokes) and time spent on Text 1’s draft. The Pearson correlation for Text 2’s draft is .56 (see Table 33 and
Table 34).
Table 33. Correction Keystrokes in Text 1’s and Text 2’s Drafting Phases
Drafting Text 1
Drafting Text 2 (Disabled:
Mouse, Backspace, & Delete Keys )
Participant Code
Mouse
Event Backspace Delete
Left Right Up Down
021-121218-CDY 11 87 8 92 165 12 24 399 41 38 1 0 80
022-121221-FJM 6 61 11 47 22 17 4 168 18 19 1 0 38
023-121224-HSH 39 23 1 28 22 3 0 119 0 0 1 0 1
024-121228-XRX 3 29 6 61 88 13 3 204 248 616 25 44 933
025-130102-HZJ 4 192 1 77 117 2 1 395 0 0 0 0 0
026-130110-LYL 4 33 0 46 50 1 3 138 0 0 0 0 0
027-130114-ZBY 12 24 2 61 87 12 1 200 22 24 1 0 47
029-130130-WXY 2 63 0 58 49 6 5 184 0 0 4 0 4
030-130225-ZMQ 4 52 0 72 253 0 4 385 0 0 1 0 1
031-130227-ZJY 7 120 0 277 266 7 17 695 89 87 2 1 179
032-130304-SGL 25 114 0 116 75 4 4 339 24 12 1 2 39
033-130328-WP 3 65 0 58 57 1 3 188 7 7 0 0 14
Average (T&I) 18.14 87.41 5.48 363.52 115.83
Table 34. Correlation between correction keystroke and drafting speed
025-130102-HZJ 395 542 0 323
026-130110-LYL 138 1144 0 873
027-130114-ZBY 200 629 47 727
029-130130-WXY 184 678 4 576
030-130225-ZMQ 385 920 1 774
031-130227-ZJY 695 1017 179 739
032-130304-SGL 339 1337 39 1078
033-130328-WP 188 633 14 498
Pearson Correlation r=0.54
r=0.56
The above data might indicate that the use of these correction keys might have influenced the drafting speed. In other words, the acceleration of Text 2’s draft might have partially come from the reduction of correction keystrokes, which is elicited by this experiment.
It should be noted, however, that Translog is not able to record every keystroke when Chinese character conversion is involved. In other words, the logging data analyzed is not complete. Perhaps the correlation between the number of correction keystrokes and the drafting speed is higher than a medium one because every correction keystroke theoretically interrupted the flow of translation and hence slowed down the process.