Surprisal explains the occurrence of filler particles in simultaneous interpreting

We present a set of studies on filler particles in interpreting in the language pair EnglishGerman, asking whether, and if so to what extent, surprisal can explain their occurrence. It is widely acknowledged that filler particles are associated with planning effort in monolingual production, but their occurrence in interpreting has not been fully investigated. Surprisal is a measure from information theory that estimates the information content (in bits) of a unit (e.g. a word) in the context of other linguistic units (e.g. n preceding words). Importantly, there is abundant evidence that surprisal is proportional to cognitive effort, measured independently in behavioural experiments, which provides a useful tool for explaining the link between linguistic choice and language cognition. Looking at the words preceding and following a filler particle in the target language output, we investigate surprisal as a predictor of filler particle occurrence, placing a special focus on function words vs. content words that follow filler particles. Overall, our analysis shows that surprisal of following words is a good predictor of filler particles in interpreting.