Aggressive individuals more readily interpret others’ motives and intentions in ambiguous situations as hostile. This hostile attribution bias has been argued to be causally involved in the development and maintenance of aggression, making it a target for interventions. In our current study, adolescents selected for high levels of aggression (N = 39) were assigned to either a test–retest control group or a five-session hostile attribution bias modification training, in which they were trained to make more benign interpretations of ambiguously provocative social situations. Before and after the training, we assessed hostile attribution bias and both reactive and proactive self-reported aggression in both groups. The training not only tended to produce the expected reduction in hostile attribution bias but also crucially led to decreased levels of reactive but not proactive aggression compared with the control group. Our results thus support the idea that hostile attribution bias can be targeted using training techniques and that such training-induced changes in bias may reduce aggression. However, future studies using an active control group and multiple outcome measures are needed to address the long-term effects of training.