Trainee Simultaneous Interpreting
2. Method 1 Subjects
Eight Finnish native speaker subjects participated. All were studying simultane
ous interpreting at the Department of English Translation Studies, University of Turku, as part of their Master’s programme.
2.2 Materials and design
Four 4-minute texts were prepared in Finnish and English versions to resemble brief conference speeches, and were recorded by native speakers. Special attention was paid to exact, yet idiomatic translation. The texts dealt with general socio-eco
nomic topics, and contained no specialist terminology. The texts were recorded at normal speech and pause rates. The mean speech rate for the English texts was 24.3 propositions per minute; the corresponding rate for the Finnish texts was 21.4 propo
sitions per minute. (For the concept of “proposition” employed here, see Tommola &
Lindholm 1996.)
The eight normally segmented versions were edited to create non-segmented ver
sions by reducing all pauses between main semantic units to 200 ms in studio condi
tions with a digital audioeditor operated by a Macintosh HEX computer and the Digidesign Sound Tools programme. The reduction of pausal segmentation necessar
ily increased the rate of information flow in the source texts; the Finnish recordings were shortened by about 25 per cent, and the English versions by about 37 per cent.
The rate of articulation of the ST speaker remained constant in this procedure.
In a within-subjects design, each of the subjects interpreted four passages, two from B to A and two from A to B under four conditions: (1) B to A, normal segmen
tation; (2) B to A, unsegmented; (3) A to B, normal segmentation; and (4) A to B, unsegmented. Counterbalancing ensured that each text appeared equally often in each condition, and that no subject interpreted a given text more than once. A proposi
tional accuracy score was established by comparing the outputs of the interpreters with the propositional representations of the source texts (for more details, see Tommola & Lindholm 1996, and Tommola & Helevá 1996).
2.3 Procedure
The subjects were provided with a glossary of central terms about a week before the experiment. They interpreted the passages individually in language laboratory conditions. After the half-hour experimental session, the subjects were asked to fill in a debriefing form, which included questions on their subjective impressions of the texts.
Jorma Tommola & Tiina Laakso
3. Results
Interpreter outputs were recorded together with the source texts on audio cas
settes for transcription and scoring. Two paragraphs of 5-7 sentences from each text, one after the introductory section and one towards the end, together with their inter
pretations, were submitted to propositional analysis. The interpretations produced by the subjects were given a propositional accuracy score, represented by the percentage of ST propositions correctly reproduced in the TT. In an effort to take into account potential summarisation strategies employed by the interpreters, a weighted accuracy score was also calculated by crediting the semantically central “highest-level” propo
sition in a sentence with three points, the next most central proposition with two, and all other correctly rendered propositions with one point. The semantically weighted scores produced exactly the same result as the unweighted scores, and the results below are given following the unweighted scoring system.
Table 1 gives the per cent accuracy scores, with standard deviations, for the two segmentation and speech rate conditions and the two language directions.
Table 1
Unweightedpropositional accuracy scores
Segmented Non-segmented
% sd % sd
English - Finnish 64.0 14.0 43.8 12.8
Finnish - English 63.5 11.7 47.7 19.0
The propositional accuracy in both language directions is considerably higher in the segmented than in the non-segmented (high input rate) condition. The accuracy scores for the two language directions are almost identical in the segmented condi
tion, while in the high-speed non-segmented condition, the propositional accuracy of the target texts seems to be slightly higher in the Finnish-English direction than in the standard B to A direction.
The data were subjected to a two-way analysis of variance. The effect of segmen
tation and speech rate was highly significant (F=46.95; df=l,7; p<.001). The seg
mented version thus allowed a significantly higher number of propositions to be cor
rectly rendered in the target language. The effect of language direction, as well as the interaction of segmentation and language direction, failed to reach significance.
Thus, in terms of propositional accuracy, the advanced trainee interpreters performed equally well in both language directions, and the presence or absence of facilitating semantic pauses did not have a differential effect on their performance depending on the direction of interpretation.
Section 3. Interpreting
4. Discussion
Gerver (1976) reports a positive correlation between ST speech rate and the num
ber of errors and omissions in simultaneous interpretation. This result is, of course, not unexpected, and it is replicated by our experiment with TT propositional accu
racy as the dependent variable. Our design controlled for articulation rate (the num
ber of segmental phonemes per time unit) but was, however, unable to separate the effect of pausal segmentation from the effect of the general rate of information input.
In principle, this is possible to do experimentally, and it would be interesting to see whether such a study would confirm the expectation that, up to a point, it is not so much the speed of articulation that matters to the interpreter but the way in which ST speakers segment and highlight the semantic units in their discourse through pausing.
The variable of language direction did not produce an effect in our experiment, just as it did not in a previous study where the effects of syntactic-semantic complex
ity and language direction were observed (Tommola & Helevá 1996). No difference appeared in the efficiency with which content information was transmitted in the two directions. This suggests that at least the comprehension processes of advanced trainees seem to proceed with roughly equal efficiency both in A and in B, given the fairly general topics of the source texts. It might still be worth while to try and find out whether more technical source material would create a situation where the more efficient comprehension of the native language would become a crucial enough fac
tor to create an A-to-B advantage in the transmission of content.
The propositional accuracy score employed in this experiment, of course, empha
sises content at the expense of the elegance, and at times even the grammatical ade
quacy of the target text. Credit was given for correct content propositions even if delivered with less than fully adequate target-language formulation. In view of the requirements of professional interpretation, this is of course not enough, and it must be admitted that the dependent variable we employed only yields a partial view of the simultaneous interpreting performance. However, it must be emphasised that the subjective impression gained from the scoring is that overall quality is strongly asso
ciated with a high degree of propositional accuracy.
References
Barik, H.C. 1973. Simultaneous interpretation: temporal and quantitative data. Language and Speech 16, 121-137.
Brislin, R.W. (ed.) 1976. Translation: Applications and Research. New York: Gardner Press.
Butterworth, B. (ed.) 1980. Language Production. Vol. 1. Speech and Talk. London etc.:
Academic Press.
Gerver, D. 1975. A psychological approach to simultaneous interpretation. META 20, 119-128.
Gerver, D. 1976. Empirical studies of simultaneous interpretation: A review and a model. In:
Brislin, R.W. (ed.), 165-203.
Jorma Tommola & Tiina Laakso
Goldman-Eisler, F. 1980. Psychological mechanisms of speech production as studied through the analysis of simultaneous translation. In: Butterworth, B. (ed.). 143-153.
Levelt, W.J.M. 1989. Speaking. Cambridge, Mass.:The MIT Press.
Lieberman, P. & Blumstein, S. 1988. Speech Physiology, Speech Perception and Acoustic Phonetics. Cambridge studies in speech science and communication. Cambridge: CUP.
Seleskovitch, D. 1965. Colloque sur l’enseignement de l’interpretation. Paris: AIIC.
Tommola, J. (ed.) 1996. Topics in Interpreting Research. Turku: University of Turku, Centre for Translation and Interpreting.
Tommola, J. & Helevá, M. (to appear) 1996. Language direction and source text complexity:
effects on trainee performance in simultaneous interpreting. Paper at the International Translation Studies Conference, Dublin City University, 9-11 May 1996.
Tommola, J. & Lindholm, J. 1996. Experimental research on interpreting: Which dependent variable? In: Tommola, J (ed.) 121-133.