The realization of prosody varies across speakers, accents, and speech conditions (e.g., Ladd, 2008). Listeners must navigate this variability to converge on consistent prosodic interpretations. We investigate whether listeners adapt to speaker-specific realization of prosody based on recent exposure and, if so, whether such adaptation is rapidly integrated with online pragmatic processing. To this end, we investigate contrastive focus, which can signal that pragmatic inference is required to determine speaker meaning (e.g., Ito & Speer, 2008; Pierrehumbert & Hirschberg, 1990; Watson et al., 2008).

In this talk, I first present results of an off-line judgement experiment using a paradigm developed to investigate implicit learning in phoneme categorization (e.g., Kraljic, Samuel & Brennan, 2008; Norris, McQueen & Cutler, 2003). The results suggest that listeners rapidly adapt their pragmatic interpretation of contrastive focus to best reflect speaker-specific realizations of prosodic cues (e.g., pitch and segment duration). I then discuss results from two eye-tracking experiments. We find that changes in the reliability of prosodic cues (estimated based on recent exposure) are reflected in changes in processing time-course: When a contrastive focus is deemed unreliable as a cue to a contrastive interpretation, listeners effectively down-weight it in their comprehension of following utterances. We conclude that such rapid recalibration of prosodic interpretations enables listeners to achieve robust online pragmatic interpretation of highly variable prosodic information.