The involuntary shift of attention to emotional sounds were investigated in a cross-modal oddball task in which participants categorized angry and disappointed emoji faces. Prior to each face, a standard tone was presented (80% of trials) or a deviant “disappointed” or a buzzing “angry” sound (20% of trials). The deviant trials were either congruent (e.g., disappointed sound/disappointed emoji) or incongruent trials (e.g., a disappointed sound/angry emoji). Results showed that the emotional content of the deviant sounds interacted with the processing of the faces, but that the effect was only present in the congruent trials. Participants showed deviance distraction (prolonged response times compared to standard) in the disappointed trials and facilitation (no deviance distraction) in the angry deviant trials. The facilitation (or lack of distraction) caused by the angry deviant sound in the congruent trial may have been a result of an arousal effect due to the processing of threat.