Using Unfilled Pauses to Measure (Dis)fluencies in Consecutive Interpreting

Abstract

Unfilled pauses (UPs) are often used as a measure of interpreting fluency, but there is no consensus on how to identify and characterize UPs. The selection of a threshold to identify UPs varies widely across interpreting research literature, ranging from 0.2 to 3 seconds, without empirical evidence to support a particular threshold. This study aimed to provide empirical evidence by examining the correlations between three acoustically measured UP-related temporal variables (number of UPs per minute, mean UP length, and phonation time ratio) and fluency ratings provided by human raters for 16 different pause thresholds increasing from 0.25 to 1 second. The study analyzed a small corpus of 13 Vietnamese-English consecutive interpretations. The results showed that higher pause thresholds generally led to lower correlation coefficients, and the mean length of unfilled pauses had the strongest correlation with subjective fluency rating. The study suggests that future research should carefully consider the selection of pause thresholds when using UPs as a measure of interpreting fluency.