ABSTRACT Disfluencies are common in spontaneous language, and adults use filled pauses to anticipate “difficult” words. At the same time, disfluent instructors receive lower ratings of knowledge and credibility. We focus on the consequences of filled pauses for adults' statistical word learning. In Experiment 1, monolingual adults listened to recordings introducing them to two speakers: one who produced an elevated rate of filled pauses and one who did not. However, both speakers were fluent when producing novel words during learning. In Experiment 2, the speaker introductions were removed, and disfluencies preceded productions of some novel words during learning. Across experiments, participants demonstrated a preference for the fluent speaker and rated the fluent speaker more favorably. Critically, word learning performance was not impacted by fluency. These findings replicate prior work on the consequences of disfluent instruction and suggest that learner attitudes—but not statistical learning mechanisms—are sensitive to speaker fluency cues.
Libersky et al. (Thu,) studied this question.