Minimización de problem triggers y optimización de la calidad en interpretación simultánea: El impacto de la gestión de las pausas por el orador sobre la transmisión del sentido

Abstract

The intended application of this thesis was to find quantitative evidence that speakers can contribute to improving the quality of simultaneous interpreting. More specifically, we surmised that a speaker can help reduce the interpreter’s cognitive load by adjusting their timing. In the main study of this thesis, we wanted to measure the effect of timerelated variations in a source speech on the interpreter’s cognitive load, using fidelity in the target speech as an indicator. To choose an appropriate treatment method for timing in the main study, we carried out three explorations of methodology, named A, B and C. Study A was an experiment aimed at probing the impact on the perception of a speech caused by two methods of speaking-speed manipulation. We asked 18 subjects from different professional backgrounds to listen to three recordings: an original speech sample, a linearly expanded version and another version in which we had expanded some of the pauses. After each playback, they described their perceptions relative to the form and to the content of the recording. Their answers were vetted using qualitative content analysis. The data suggest that, as standalone recordings, the three versions were perceived as similar in terms of articulation, prosody and rhetoric. However, when the subjects could compare them, they rejected the rhetoric of the linearly expanded version. Therefore, this magnitude might act as a confounding variable in the causal relation between this treatment and the reduction of interpreting difficulty. This led us to discard linear time expansion as an experimental treatment for the main study. The purpose of Study B, also an experiment, was to compare the interpreting difficulty of a linearly expanded voice recording with that of similarly fast natural speech. Eight professional interpreters were asked to render this ‘artificial’ version of the speech as well as a ‘natural’ version, delivered by the same speaker at a similar speech rate. The two versions were presented in random order, as the soundtracks of two videotaped press conferences. After each interpretation, the subjects were asked to estimate its interpreting difficulty in terms of several dimensions. When the two versions were presented separately, the subjects perceived the time-compressed recording as harder to interpret with regard to linguistic expression. When they could compare them, they perceived the time-compressed version either as similarly or as more difficult to interpret than the fast natural version in most dimensions. These results cast doubts on the advantage of time compression for the interpretation of content-rich speeches, causing us to discard linear time compression as an experimental treatment for the main study. In Study C, of observational character, we tried to pin down the variation of interpreting difficulty along a given speech, as well as the influence of the speaker’s pause pattern on that difficulty. A corpus of 20 interpretations of that speech, broadcast by television or radio, was searched for errors and omissions (EOs). By aggregating these data, we identified the segments of the source speech that were more frequently affected by EOs in the corpus. Those segments in which at least one third of the interpreters had committed EOs were defined as ‘hotspots of interpreting difficulty’. Focusing our analysis on these hotspots, we found that once the EOs started to cluster, their frequency tended to remain high. This is evidence for cognitive saturation and supports Gile’s Tightrope Hypothesis. Moreover, the EO frequency of individual segments did not seem to be reduced by nearby pauses in general. However, in paragraphs introduced by a long pause, EOs typically took some time to cluster, and virtually all EO clusters were confined within the boundaries of some sentence. This suggests that the pauses located at constituent or sentence boundaries can provide cognitive relief to interpreters, which led us to choose the expansion of such pauses as the treatment method in our final experiment. In the main study, we used the same audio recording as in Study C, expanding juncture pauses located near the hotspots identified in that study. We asked 25 highly experienced practitioners to interpret either the original recording (control version) or the treatment version. After controlling for the age variable, it was found that, on average, the interpretations of the treatment version featured 15% less EOs than those of the control version, with a large effect size. Moreover, there was some evidence of local cognitive relief around the pauses expanded. These findings provide empirical support to the claim that speakers can contribute actively to the enhancement of simultaneous-interpreting quality, and that this contribution can be measured.