Effects of Prosodic-Syntactic Segmentation on the Perception of Fluency and Quality of Simultaneous Interpreting

Christodoulides, George [UCL]

Previous studies have shown that simultaneous interpreting (SI) is characterised by a particular prosodic profile (e.g. long pauses, intonation contours indicating continuation – Shlesinger, 1994; Ahrens, 2005; Christodoulides, 2013). Speech produced under high levels of cognitive load, as is often the case during an SI performance, presents specific prosodic characteristics, such as lower and more variable articulation rate, and an increased number of filled pauses and hesitation-related disfluencies (Tei Fei Yap, 2012; Christodoulides, 2016). In SI, the strategies of stalling and anticipation have an impact on the temporal organisation of the interpreter’s speech, and especially the relationship between prosodic segmentation (as indicated by silent pauses and intonation contours) and syntactic segmentation. In a previous study (Christodoulides & Lenglet, 2014) we compared the perception of fluency and quality of interpreting by non-expert listeners, by contrasting two prosodic profiles: SI proper, and a rehearsed reading version of the same text by the same speaker; results showed that listeners rated the rehearsed reading version as more fluent and found that the interpreter was more ‘accurate’. In this study however, all factors of the two prosodic profiles contributed to the listeners’ evaluation: we would like to focus on manipulating one particularly salient difference between SI and other prosodic profiles, namely the high number mismatches between prosodic and syntactic segmentation. The objective of the present study is to evaluate the effects on the perception of fluency and quality of simultaneous interpreting of the relationship between prosodic and syntactic segmentation of utterances. In a listening experiment, participants were asked to evaluate a simultaneous interpretation, based on criteria proposed in the literature on the quality of interpreting and speech fluency (Collados Aís, 2007). Participants watched videos of original speeches, with a mixed overdub of simultaneous interpreting, in a situation resembling a TV interpreting; responses were collected on a web platform. Participants were randomly assigned to experimental conditions, covering a continuum of prosodic-syntactic segmentation configurations, across two axes: number of mismatches between prosodic and syntactic boundaries, and number of syntactic units grouped within a single prosodic unit. All other prosodic aspects of the SI rendition were controlled. Our hypothesis is that there is a negative correlation between the number of mismatches and the perceived fluency and quality of interpreting (i.e. speech with more mismatches are perceived as less fluent). This research has implications for training: more specifically, it will be beneficial for interpreting students to make them aware of their performance with respect to packing syntactic units within prosodic units, when mastering the techniques of stalling and reformulating.