Corpus-based research into the presumed effects of short EVS

Abstract

The aim of this study, based on 32 French speeches simultaneously interpreted into Dutch at plenary sessions of the European Parliament in late 2008, was to ascertain whether short ear-voice span (EVS) affects the quality of the interpretation as is commonly stated in the literature. The speeches and interpretations were taken from the 'EPIC Ghent' corpus, which is in preparation at Ghent University. Three phenomena were identified as potential effects of a short EVS: syntactic transcodage (maintaining the right-branching French 'noun+de+noun' structure, not using a more natural left-branching structure, in the Dutch interpretation), use of cognates similar in sound to source language forms ('glissement phonetique'), and certain self-repairs (Barik 1973; Gile 1995). Time tags were applied to both the source and target texts, so that EVS could be measured to the nearest second from the onset of a source language item to the onset of the target language equivalent. The hypothesis was that EVS would be shorter in contexts where these three phenomena occur than elsewhere in the subcorpus. This was borne out in only one case, i.e. use of cognates: short (2 secs.) and very short (1 sec.) EVS was significantly more frequent in contexts where cognates occurred than elsewhere. There was no statistically significant frequency difference in the context of transcodage or of the relevant self-repairs.