From local hesitations to global impressions of the listener

The listener’s interpretation of a speaker’s utterance includes estimates about the speaker’s commitment to what they are saying. Previous works have shown that fillers (e.g. “um”) are linked to both the speaker’s metacognitive state, and the listener’s impression of a speaker’s state. However, these results are limited to contexts that may not apply to spontaneous speech. Additionally, there is a lack of hierarchical analysis of the discourse; i.e. how a speaker’s local use of fillers could lead to a
listener’s overall impression. In this work, we address these limitations by studying how does a speaker’s use of fillers relate to the incoming message, and consequently, what is the
resulting impression formed by the listener. We do so by analysing a dataset of English monologue movie reviews, where the speakers voluntarily and naturally recorded themselves giving a movie review. Our findings show that speakers tend to stylistically use fillers in the incoming message before introducing new information related to the review, and that listeners may not associate this specific use of fillers with their estimate of the speaker’s expressed confidence. Our results highlight that there are potentially different metacognitive effects from the speaker’s use of fillers on the listener.