The input rate determines the processing load that interpreters have to carry from moment to moment in SI. The demand of concurrent comprehension and production makes SI performance particularly sensitive to the variance of input rate. The empirical evidence in the present study indicates that both fast and slow input rates led to difficulties in SI
processes. At the rate of 160 wpm, there was more information presented at a given time unit and much shorter time for student interpreters to process the source messages. Students could not speak faster or shorten their EVS in order to catch up with the speaker. The only strategy they could adopt was to pause less frequently. At this input rate, either their comprehension was hampered or there was not enough time for production, which led to more omissions of segments in the SI output. The detrimental effect of fast input rates on the accuracy of SI output has been consistently supported in the interpreting literature and both professional and student interpreters were affected (e.g., Gerver, 1969; Pio, 2003). A surprising finding is that student interpreters’ lexical diversity in the output was higher at the fastest input rate,
possibly resulting from more omissions of recurring words in the source speeches. These recurring words may be important for the coherence and cohesion of a text. Although the lexical diversity was higher at the rate 160 wpm, it is very likely the target speech was less coherent when much more source segments or even the whole meaning units were omitted by student interpreters.
The slowest input rate, on the other hand, affected other aspects of the SI output.
Information from the source speeches was presented at a much slower rate. At the rate of 100 wpm, students made more errors on words rather than on segments. They also paused longer as the speaker did and displayed a longer EVS in order to wait for more information in the source speech, which in turn induced a higher risk of memory loss (Shlesinger, 2003).
Interestingly, although student interpreters' EVS was longer at 100 wpm, this did not make them less susceptible to syntactic interference. When encountering the structures of M+AP
143
and M+AC in English, to lessen the memory load of the long main clauses, students tended to take a form-based strategy and follow the English structures in their Chinese output.
Gile (1995, 2009) proposed in the Effort Models that simultaneous interpreters will encounter problems “when total processing capacity requirement exceed available processing capacity (saturation), and when processing capacity available for a given Effort is not
sufficient the task it is engaged in at a given time (individual deficit)” (p. 189). This theory seems to apply to our findings. At the rate of 160 wpm, it is likely that students’ cognitive capacity was saturated. At the rate of 100 wpm, their memory effort was possibly overloaded.
In terms of the temporal management during the SI process, the speed of 130 wpm seemed to be the most comfortable input rate for student interpreters. At this rate, they were able to speak more quickly, pause less and maintain a relatively shorter and appropriate time lag.
However, in terms of output accuracy, students still made more omissions of segments at 130 wpm than at 100 wpm.
We also found that the input rate did not interact with L2 proficiency in the effects on the nine aspects of the SI output. In terms of the linguistic and temporal processes observed in our study, it seems that the two English proficiency groups displayed similar patterns under the influence of three input rates. However, an investigation on the observed data indicates that the two English groups tended to adopt different strategies in the processing of the critical sentences. In addition, the production speed of the low proficiency group seemed to be more affected by the variance of the input rate. Unlike the high proficiency group, students with lower English proficiency slowed down their production at the fastest input rate,
indicating that their cognitive capacity had probably been saturated. Although English proficiency did not interact with input rate in the effects on the output, English proficiency did play an important role in student interpreters’ SI performance. Students' English
proficiency levels had an impact on the frequency of omissions they made in the output. Call (1985) proposed that in listening comprehension, the memory span for L2 elements is shorter
144
than for L1 elements, especially in less-proficient speakers. In the present study, the high English proficiency group was able to maintain and delivered more source messages than the low group. In addition, the production process of the high group was probably more effective and efficient, as indicated by the higher output rate and better used of the linearity strategy.
Apart from input rate and English proficiency, the secondary independent variable, speech, was also found to have significant effects on the nine aspects of the SI output. The participants’ linguistic processes and temporal management of the output were influenced by the different surface features of the source materials, as discussed in Section 5.5. Moreover, students’ prior knowledge of the speech topics could also affect their SI performance. Many of them mentioned that the topic of smart cities was the most familiar, but some of them found the terms of tobacco control difficult while others said that the ideas of the cyber security speech were hard to follow. It seems that individual differences in the exposure to the topics covered in the source materials might have an impact on the lexical retrieval or
memory process during the SI task. These differences can be gleaned from the high standard deviations observed in certain dependent variables as a function of speech, input rate, and English proficiency. Syntactic interference is such an example. As it is mentioned in Section 5.5, students had significant more sentences of syntactic interference in their output for speech B than for the other two speeches. This may stem from the differences in surface features among the three source speeches. It is also found that some students were more prone to syntactic interference than others, as shown in the great dispersion of data (See Table 4.14). This indicates that it is extremely difficult to have proper control of the source
materials when difference speeches were adopted. Nevertheless, the findings concerning the main effect of speech still provide interesting insights into the relationship between the source speech features and the target output.
145