When listeners hear a speaker become disfluent, they expect the speaker to refer to something new. What is the mechanism underlying this expectation? In a mouse-tracking experiment, listeners sought to identify images that a speaker was describing. Listeners more strongly expected new referents when they heard a speaker say um than when they heard a matched utterance where the um was replaced by noise. This expectation was speaker-specific: it depended on what was new and old for the current speaker, not just on what was new or old for the listener. This finding suggests that listeners treat fillers as collateral signals

When speakers detect a problem in what they are saying, they must decide whether or not to interrupt themselves and repair the problem, and if so, when. Speakers will maximize accuracy if they interrupt themselves as soon as they detect a problem, but they will maximize fluency if they go on speaking until they are ready to produce the repair. Speakers must choose between these options. In a corpus analysis, we identified 448 speech repairs, classified them as major (as in a fresh start) or minor (as in a phoneme correction), and measured the interval between suspension and repair. The results showed that speakers interrupted themselves not at the moment they detected the problem but at the moment they were ready to produce the repair. Speakers preferred fluency over accuracy.