Providing participants with an opportunity to listen to a forthcoming distracter sentence has been shown to attenuate its disruptive effect on short-term memory. On the stimulus-specific attentional diversion account, foreknowledge selectively reduces any potential diversion produced by interest in the post-categorical (e.g., semantic or syntactical) properties of a discrete sentence. This account assumes that the beneficial effect of foreknowledge depends crucially on the intelligibility of pre-exposed sentential speech. During a visual-verbal serial recall paradigm, participants undertook two counterbalanced blocks of trials wherein they were either pre-exposed to impending auditory distracter sentences (foreknowledge) or not (no foreknowledge). Pre-exposed sentences were intelligible, partially intelligible or unintelligible while sentences accompanying serial recall were all intelligible. Participants were instructed to attend to the sentences during pre-exposure and ignore them when they accompanied the serial recall task. Foreknowledge of an impending distracter sentence attenuated its later distractive power in serial recall, but only when the foreknowledge was at least partially intelligible. Consistent with the stimulus-specific attentional diversion account, the intelligibility of speech presented during a foreknowledge period is a key requirement for attenuation of auditory distraction by sentential speech. This suggests that intelligible foreknowledge increases familiarity of the material thereby reducing attentional diversion due to interest. These results reinforce the view that foreknowledge reduces disruption produced by the semantic/syntactical properties of discrete sentences but has little effect on that produced by its acoustic properties.