Event segmentation during speech listening: The relative contributions of thematic structure and prosody

Event segmentation – the division of continuous experience into meaningful units – is essential for understanding narratives. Although event segmentation has been widely studied in visual contexts, little is known about how listeners segment auditory stories or how prosodic cues, such as pause duration, contribute to this process. Moreover, age-related changes in hearing and cognition may alter how older adults perceive and organize continuous speech. To examine how thematic structure and pause duration shape event perception across age, we conducted two experiments. The event boundaries in both studies were determined by large language models, a method shown to be reliable by previous work. In Experiment 1, we analyzed human and AI-generated speech from audiobooks and podcast monologues. Both human and advanced AI voices produced longer pauses at thematic boundaries than within events, indicating that pause timing aligns with narrative structure. In Experiment 2, younger (19–35 years old) and older (56–79 years old) adults listened to stories in which pauses were selectively prolonged at event boundaries or event centers. Older adults identified event boundaries more frequently overall. Prolonging pauses at event boundaries did not enhance boundary identification, whereas prolonging pauses at event centers led to boundary identification despite the absence of a major thematic narrative shift, particularly in listeners with poorer hearing (determined via audiometry testing). These findings show that aging may change the sensitivity to perceiving event boundaries during continuous speech listening and that pause duration can bias event segmentation in contexts devoid of thematic shifts.