The present study has the following limitations. To begin with, we only examined the mode of SI from English into Chinese on student interpreters. The findings on the effects of input rate and the null interactive effect of input rate and English proficiency do not
generalize to SI from Chinese into English, to professional interpreters, or other modes of interpreting. The discrepancy of our results with that of studies on professional interpreters (e.g., Gerver, 1969) and on SI from Chinese into English (e.g., Chang, 2009) indicates that more research is warranted to examine whether the level of expertise and translation directionality affect the effects of input rate on the SI output.
Next, there are three issues with the speech materials used in the present study. First, the order of the three source speeches was not counterbalanced across all participants. The same
148
order of speech was used for every participant. Although the variance of the speech factor was statistically controlled when the hypothesis-related factors were examined, we cannot exclude the possible influence of practice and fatigue induced by the fixed order of three speeches. This issue should be avoided with proper counterbalancing in future studies.
Second, in order to better control the variance of text difficulty as well as to accommodate the design of critical sentences, we chose to use prepared speeches instead of spontaneous ones.
The speaker mostly read aloud from the transcripts in a fluent manner. With more hesitations and pauses, spontaneous speeches may sound more natural than prepared speeches and they may be even easier for student interpreters to understand (Déjean Le Féal, 1982; Taylor, 1989). However, it would be difficult to find different spontaneous speeches that are more or less comparable in text features. It depends on future studies to find a proper balance between ecological validity and experimental control. The third issue is related the design of critical sentences in the source materials. The average length of critical sentences is 17.3 words, but the range is from 7 words to 26 words. This variance of sentence length might still have an impact on student interpreters’ processing of the critical sentences. As it is mentioned in Section 5.5, it was found that the extent of syntactic interference was significantly higher in speech B, which might result from the different average lengths of main clauses and
adverbials among the three source speeches. There should have been better control of the critical sentences by keeping the sentence length more or less the same.
Another limitation concerns the use of type-token ratio (TTR) as the measure of lexical diversity in student interpreters’ output. It was found that the TTR of student interpreters’
Chinese output was significantly higher at the input rate of 160 wpm than at slower rates.
This result is out of expectation and rather counterintuitive. TTR only measures the ratio of unique words to the total number of words in a text. In hindsight, to investigate the diversity of interpreters’ lexical use, it is probably more appropriate to use a finer parameter given that different English words have different numbers of translation equivalents in Chinese. Some
149
words like proper nouns may have only one translation equivalents, which would reduce the lexical choices of an interpreter. Other words like common nouns and verbs may have multiple translation equivalents. For future studies on interpreters’ lexical diversity, it is necessary to take into consideration the influence of possible translation equivalents on the word choices of interpreters in the output.
Fourthly, there was great heterogeneity among the participants in terms of levels of English proficiency and SI training. The individual differences in L2 proficiency, training periods and prior knowledge might have an impact on their SI performance. Since our participants were recruited from different graduate institutes in Taiwan, the differences in SI training may be attributable to the different screening policies, course programs and training resources at each institute. As for the differences in English proficiency levels, there was not only variability among the participants but also variability within the high and low groups.
For example, it was observed that the two participants with C2 level excelled other students in the high proficiency group in the SI performance. These individual differences, the small sample size as well as the unequal numbers of samples in each English proficiency group may all contribute the null interactive effect of input rate and English proficiency as well as the null main effect of English proficiency on certain dependent variables.
Fifthly, the concept of information density was not addressed in the present study.
Information density in a source speech is related to the input rate (Christoffels & De Groot, 2005; Gile, 1995, 2009; Pöchhacker, 2004). However, a high input rate does not necessarily lead to high information density. It is possible that the information is redundant even though the input rate is fast (Déjean Le Féal, 1982). More research is certainly required to clarify the relationship between input rate and information density.
Lastly, due to limitations of scope, only one indicator of output fluency was investigated in the present study. Apart from the frequency of unfilled pauses, the pause to speech ratio, the number of filled pause, hesitations, and false starts were also found to be affected by the
150
changes in the input rate (e.g., Gerver, 1969; Pio, 2003). It may be the interest of future studies to examine the influence of the input rate on these aspects of the output.