A speaker’s disfluencies (such as ums and ers) have effects on a listener’s immediate understanding and lasting representation of an utterance. One suggested effect is that disfluencies heighten listeners’ attention to subsequent speech (Collard, 2009; Fox Tree, 2001). If attention is heightened, then what is the feature of disfluencies that is responsible for this effect? We employed a change-detection paradigm to test the attention-heightening hypothesis using naturally produced speech as stimuli. In three experiments participants listened to fluent and disfluent utterances containing either the disfluency er, like, or a silent pause. After listening to each speech fragment participants read transcripts of the corresponding clips and detected whether any changes had occurred between the audio and written presentations of each stimulus. It was found that across experiments participants were more likely to detect a change when the target word in the audio version was preceded by a disfluency. The likelihood of detecting change did not significantly differ with respect to the type of disfluency present. As a result, it could be said that these disfluencies may be perceived as a temporal interruption to speech, which may lead listeners to treat them equivalently. But there is insufficient evidence to describe the nature of the disfluencies that is responsible for the attention-heightening effect. This study demonstrates that disfluencies affect listeners by heightening their attention to upcoming speech which is associated with better memory for post-disfluent words.