The results of statistical analysis show that input rate had significant effects on the number of OW and OS in student interpreters’ SI output. However, patterns of effects were different for OW and OS. As Figure 5.1 shows, the EMM of OW was the highest at 100 wpm, suggesting that a slow input rate led to more OW, contrary to what was predicted and to the results of previous literature (Galli, 1989; Gerver, 1969; Pio, 2003). The EMM of OS, on the other hand, increased with each interval of input rate (See Figure 5.2). When the input rate was slow, student interpreters’ omissions were more on the word level. However, as the rate increased, they started to omit longer segments or even omit the whole meaning unit. The result that faster input rates lead to higher omissions of segments corroborated the findings of previous studies (Barik, 1973; Galli, 1989; Gerver, 1969; Pio, 2003).
124
Figure 5.1 Estimated Marginal Means of OW among Input Rates
Figure 5.2 Estimated Marginal Means of OS among Input Rates
4.899
3.988
3.674
2.00 3.00 4.00 5.00 6.00
100 130 160
Number of OW
Input Rate
OW Estimates
18.784
27.407
38.086
15.00 20.00 25.00 30.00 35.00 40.00 45.00
100 130 160
Number of OS
Input Rate
OS Estimates
125
It is worth noting that the scoring methods adopted in the present study probably had an impact on the number of OW and OS in student interpreters' output. To ensure consistency of rating, raters could only assign one error category to each meaning unit of the source
speeches. When more than two content words were omitted in the meaning unit, OS was given instead of two or more OWs. As the input rate increased to 160 wpm, student
interpreters tended to omit more than two content words in the meaning unit. This could lead to lower OW at the fastest input rate.
Substitution
The statistical results show that the effects of input rate were significant on SW but not on SS. Student interpreters made significantly more SW when the input rate was at 100 wpm and 130 wpm than at 160 wpm (See Figure 5.3). This is similar to the results of OW. As the input rate increased to 160 wpm, they started to commit fewer translation errors on words.
However, unlike the results of OS, student interpreters did not make more substitutions as the input rate increased. The EMM of student interpreters’ SS stayed around 21 to 23 across three input rates (See Figure 5.4). This is consistent with the results of Pio (2003). Although Pio (2003) found that student interpreters made slightly more substitutions at the fast input rate (145 wpm) and professional interpreters made more substitutions at the slow input rate (108 wpm), she did mention that the general differences of substitutions between the slow and fast input rates were rather small. It seems that the increased processing load imposed by a higher input rate did not have an impact on the number of inaccurately translated segments student interpreters had in the output. Other factors such as prior knowledge of the source speech probably play a bigger part than the input rate in the error category of substitution.
126
Figure 5.3 Estimated Marginal Means of SW among Input Rates
Figure 5.4 Estimated Marginal Means of SS among Input Rates
6.868
6.458
4.643
3.00 4.00 5.00 6.00 7.00 8.00 9.00
100 130 160
Numbe of SW
Input Rate
SW Estimates
21.949
23.059
21.644
18.00 20.00 22.00 24.00 26.00
100 130 160
Number of SS
Input Rate
SS Estimates
127
Syntactic interference
English and Chinese are two structurally different languages. When interpreting from English into Chinese, without proper strategies of waiting and reordering, interpreters are prone to produce sentences with syntactic interference. In the present study, English sentences with adverbial phrases placed at the end (M+AP) and sentences with adverbial clauses as the second clause (M+AC) were examined. Interpreters need to reorder the
position of main clause and adverbials or employing the strategy of linearity with padding in order to produce grammatical Chinese sentences. It is generally expected that when under higher cognitive constraint, interpreters would adopt a more form-based strategy by following the linguistic structures of the source language without reordering (Dam, 2001; Gile, 1995, 2009; Isham, 1994). If we assume that a fast input rate would exert a greater cognitive load on interpreters, we would expect interpreters to produce more sentences with syntactic interference when the input rate is higher. Contrary to this prediction, the statistical results show that the extent of syntactic interference in student interpreters’ output was significantly higher at 100 wpm than at 160 wpm, but there was no significant difference of syntactic interference between 100 wpm and 130 wpm or between 130 wpm and 160 wpm (See Figure 5.5).
These results do not corroborate the findings of Zhang (2010), who also investigated how student interpreters processed English adverbials placed at sentence-ending positions in consecutive interpreting into Chinese. She found that the proportion of deverbalization (absence of syntactic interference) was higher at 110 wpm than at 150 wpm. The discrepancy of findings between the present study and that of Zhang (2010) may stem from the
differences in mode of interpreting. In Zhang's study, students were allowed to take notes in the consecutive interpreting experiment and were under less time pressure during the output production. In the present study where SI was conducted, the input rate might have different influence on student interpreters’ time constraint and memory load. The average length of
128
main clauses in the critical sentences of all source speeches is 10.5 words. When the input rate was as slow as 100 wpm (1.67 words per second), it took around six seconds for the speaker to finish the main clauses before the adverbials. According to Baddeley (1992), traces of acoustic or verbal materials can be maintained in the phonological store of working
memory for as long as two seconds without subvocal rehearsal. If student interpreters waited for the adverbials to appear in the source speech in order to produce the AP+M or AC+M structures, they would have to lag behind the speaker for more than six seconds, which could possibly result in great information loss in memory. Their EMM of EVS is 4.2 seconds at 100 wpm, suggesting that student interpreters had probably started to interpret the main clause before the adverbials came up in the source speeches. By following the linear structures of English, there was a greater chance of linguistic interference but less memory load. This may explain why the extent of syntactic interference from the source language was higher at the rate of 100 wpm.
Figure 5.5 Estimated Marginal Means of STI among Input Rates
.531
.315
.097
(0.20) 0.00 0.20 0.40 0.60 0.80
100 130 160
Number of STI
Input Rate
STI Estimates
129
Although the evidence in the present study indicates that student interpreters tended to opt for the form-based strategy at the slowest input rate. This does not necessarily imply that they adopt a more meaning-based strategy at higher input rates. The results of statistical analysis show that there was no significant effect of the input rate on NI (sentences with no syntactic interference) and LI (linearity) but there was a significant effect of the input rate on N/A (omissions of critical sentences) (See Appendix G for the results of NI, LI and N/A). The EMM of omitted critical sentences (N/A) was significantly higher at 160 wpm than at 100 wpm (See Figure 5.6). As the input rate increased to 160 wpm, student interpreters did not adopt a more meaning-based strategy but omitted more sentences instead. Even at the rate of 100 wpm, they omitted half of the critical sentences on average. This indicates that the critical sentences in the source speeches were rather difficult for our participants.
Figure 5.6 Estimated Marginal Means of N/A among Input Rates
4.175
5.029
5.456
3.50 4.00 4.50 5.00 5.50 6.00 6.50
100 130 160
Number of N/A
Input Rate
N/A Estimates
130
Lexical diversity
Interpreters’ lexical diversity is an issue that has been rarely explored in interpreting studies. In the present study, we expected that the lexical diversity of student interpreters’
output in Chinese would be lower when they faced higher input rates because they might have few cognitive resources for retrieving translation equivalents in Chinese under higher cognitive load. To our surprise, we observed that the student interpreters’ type-token ratio was significantly higher at 160 wpm than at 130 wpm and 100 wpm (See figure 5.7). In other words, student interpreters’ vocabulary of Chinese output was more diverse at the fastest input speed than at slower speeds.
Figure 5.7 Estimated Marginal Means of Lexical Diversity among Input Rates
One possible explanation for this peculiar result is that student interpreters made significantly more omissions of segments at the input rate of 160 wpm and they could have omitted much more recurring words from the source speeches, resulting in higher lexical
41.83%
42.57%
44.50%
0.40 0.42 0.44 0.46 0.48
100 130 160
TTR
Input Rate
Lexical Diversity Estimates
131
diversity at the highest input rate. The average type-token ratio of three source speeches is 45%, so more than 50% the words used in the source speeches occurred more than once.
When the input rate increased to 160 wpm, the proportion of meaning units with omissions of segments (OS) was around 45%.5 It is very likely that many of the recurring words in the source speeches were omitted. At the rate of 100 wpm, it is possible that student interpreters were able to render the source messages more completely and keep more of those recurring words. Thus, their type-token ratio was lower at the slowest rate.
EVS
In previous studies, both professional and student interpreters were observed to prolong their EVS as the input rate increased (Barik, 1973; Chang, 2009; Gerver, 1969). However, in the present study, student interpreters’ EVS was significantly longer at 100 wpm than at 130 wpm and at 160 wpm but there was no significant difference of EVS between 130 wpm and 160 wpm (See Figure 5.8). The EMM of EVS at 100 wpm is 4.18 seconds, which means that student interpreters on average waited for around 7 words before production. The average length of a meaning unit in three source speeches is 8.5 words. This indicates that at the rate of 100 wpm, students did not begin to produce the translation until almost the end of a meaning unit, which mostly consists of an NP+VP or a clause in the source speeches. As Goldman-Eisler (1972) proposed, the minimum unit of interpreting is usually an NP+VP.
Interpreters need to wait for the VP to come up in the source message so that a proposition with semantic relations can be properly established in the mental model (Tommola, 2003).
5 The average number of meaning units of three source speeches is 87. The EMM of OS at 160 wpm was 38.1, which accounts for around 44% of the meaning units.
132
Figure 5.8 Estimated Marginal Means of EVS among Input Rates
As the rate began to pick up to 130 wpm, students started to shorten the EVS in order to follow more closely with the speaker. However, beyond the rate of 130 wpm, they were unable to process the source messages any faster, probably because their processing capacity was nearly depleted. Therefore, their EVS remained constant. Unlike what Chang (2009) observed in SI from Chinese into English, students in the present study did not lag further behind the speaker as the input rate increased. Chang (2009) suggested that when students were interpreting into their weaker language, longer EVS may be associated with difficulty in finding translation equivalents. Even at the highest input rate, students in the present study still managed to maintain a time lag of fewer than 4 seconds, which is within the appropriate range of 2 to 4 seconds (Chen, 2012; Lee, 2002). This suggests that students could still regulate their EVS to a certain extent when producing in their L1.
4.180
3.771
3.737
3.20 3.40 3.60 3.80 4.00 4.20 4.40 4.60
100 130 160
EVS (seconds)
Input Rate
EVS Estimates
133
Unfilled pauses
The frequency of unfilled pauses in the output of SI is a parameter of fluency as well as a possible indicator of cognitive load associated with the SI processes. Results from past literature show inconsistency in terms of the effects of input rate on either the frequency or the total length of unfilled pauses (e.g., Cecot, 2001; Pio, 2003). In the present study, we found that the EMM of the number of unfilled pauses increased as the input rate decreased and the differences among three input rates were all significant (see Figure 5.9). These findings were compatible with the results of Tissi (2000) and Cecot (2001), who also observed more unfilled pauses at slower input rates.
Figure 5.9 Estimated Marginal Means of Unfilled Pauses among Input Rates
Cecot (2001) observed that the pause patterns of professional interpreters in the SI output tended to resemble that of the speaker in the source speech. Barik (1973) found a positive correlation between the mean pause duration of the source speech and the target
158.270
121.300
99.583
80.00 100.00 120.00 140.00 160.00 180.00
100 130 160
Number of Pause
Input Rate
Unfilled Pauses Estimates
134
output. By examining the number of unfilled pauses in the source speeches, we also found that the speaker paused less often as the input rate increased. The mean number of unfilled pauses longer than 250 milliseconds in the source speeches is 98.3 for 100 wpm, 88.3 for 130 wpm, and 66 for 160 wpm. At slower input rates, especially at the rate of 100 wpm, the speaker paused more often, not only at syntactic junctures such as the end of a clause, but also within a clause. At this time, students paused more frequently and waited for enough information from the source segment in order to produce a meaningful Chinese segment. As the speaker speeded up, they started to pause less and attempted to catch up with the speaker.
Generally speaking, student interpreters made more pauses than the speaker at all input rates and they tended to follow the speaker with regard to pause patterns. These results are not consistent with that of Pio (2003), who found more pauses in students’ output at a higher input rate. This may be because she only examined the unfilled pauses longer than three seconds. By comparing our findings with that of Cecot (2001), it seems that the frequency of unfilled pauses in the SI output is highly correlated with the frequency of pauses in the source speech, which partially determines the input rate. We do not claim, though, that students’
output was less fluent at the rate of 100 wpm since we examined only one parameter of fluency.
Output rate
Results from previous literature are inconclusive towards the effects of input rate on the output rate. Barik (1973) found a positive correlation between the input rate and the output rate, but Lee (2002) and Chang (2009) observed that the output rate of professional and student interpreters remained constant as the input rate increased. The present study found that the input rate did affect the output rate of student interpreters. However, the Chinese output rate was significantly slower at 100 wpm than at 130 wpm and at 160 wpm (See Figure 5.10). Beyond the rate of 130 wpm, students could not further speed up their delivery.
135
This is consistent with the finding of Gerver (1969), who observed that the output rate of professional interpreters picked up as the input rate increased from 95 wpm to 112 wpm but remained constant beyond the input rate of 120 wpm. To increase the output rate means interpreters need to process the source speech more efficiently. When the input rate was over the comfortable range of SI and rose to 160 wpm, students were probably working near cognitive saturation and thus unable to process the source speech any faster. De Groot (1997) pointed out that at a higher input rate, “the time span over which the individual words are presented in the input is, of course, relatively shorter” and this may hamper the perception and identification of words in the source speech (p. 45). As the difficulty of comprehension increases, it is less likely for interpreters to speed up production.
Figure 5.10 Estimated Marginal Means of Output Rate among Input Rates
An interesting finding in the present study is that the interaction between input rate and speech was statistically significant on the output rate. The analysis of simple main effect and
147.199
162.110
161.433
130.00 140.00 150.00 160.00 170.00 180.00
100 130 160
Output Rate (spm)
Input Rate
Output Rate Estimates
136
post hoc tests indicated that only in speech A was the output rate significantly lower at 100 wpm than at 130 wpm and 160 wpm (See Figure 5.11). For speech B and C, students’ output rate did not change much across three different input rates. This suggests that student
interpreters’ output rate was sensitive to source text features or difficulty. The fact that they were able to speed up their output rate in speech A implies that speech A may be relatively easier than the other two speeches. The differences of three source speeches on the output will be discussed in Section 5.5.
Figure 5.11 Interaction of Input Rate by Speech on Output Rate