The role of prosodic predictability in the perception of prominence in speech

Probabilistic phenomena have been known to play an important role in models of human language production, comprehension, and learning. In addition, at a more general level, there have been increasing evidence of probabilistic processing taking place at the level of human cognition. Prosodic prominence has been primarily examined through investigations of the interactions between lexical or sublexical units and their concurrent acoustic realization. However, prominence has been recently associated with attention- and expectation-based factors. In this talk, I will present a hypothesis and empirical evidence regarding the role of prosodic predictability in prominence perception. The underlying hypothesis is that one mechanism driving prominence perception could be based on the deviations of the listeners’ predictions of the upcoming prosodic patterns and thereby capturing the attention. The talk will focus on presenting findings from computational and behavioral studies, indicating that predictability is a strong cue for prominence.