Prosodic Encoding of Informativity: Word Frequency and Contextual Probability Interact with Information Structure

This study investigates how prosody encodes the extent to which a linguistic element is informative. While most prior work has approached this question from one of the two angles, namely (i) information theory/statistical probability and (ii) discourse-pragmatics/information structure, we focus on the interaction between these two dimensions. Our results show that the prosodic marking of information-structural categories depends on statistical probabilistic factors. Specifically, the post-focus pitch reduction resulting from new-information focus and corrective focus is modulated by the focused word’s frequency and contextual probability, respectively. In terms of pitch, new-information narrow focus patterns like wide focus when the focused word is lexically infrequent, although the two focus types differ when the word has high frequency. Furthermore, corrective narrow focus patterns like wide focus when the focused word is contextually improbable, although the two focus types differ when the word has low contextual probability. We discuss how these results suggest that prosody reflects speakers’ expectation and surprise about the interlocutor’s knowledge state. Our findings highlight the importance of integrating research from the information-theoretical perspective with research from the information-structural perspective, to improve our understanding of prosody.