A corpus-based analysis of the effect of syntactic complexity on disfluency in consecutive interpreting

Abstract

This study investigates the effect of input and output syntactic complexity on disfluency based on the corpus of press conference interpreting. In line with widespread practice in quantitative linguistics, mean dependency distance has been taken as the metric here for quantifying syntactic complexity. As the occurrence of disfluency is count data, this study uses the Poisson regression model to evaluate the effect of input and output syntactic complexity on disfluency occurrences, and on the variation of such occurrences with different types of reformulation methods. Our results show that occurrence of disfluency can be predicted by both the input and output syntactic complexity as quantified by the mean dependency distance, and that when the reformulation method of “divide” is used, the occurrence of disfluency does not increase significantly in the output even when the output sentence has a higher mean dependency distance. The findings reveal how input and output syntactic complexity predicts disfluency, and how reformulation methods interact with syntactic complexity to moderate cognitive load.