Intergroup contact (especially cross-group friendship) is firmly established as a powerful strategy for combating group-based prejudice (Pettigrew & Tropp, 2006). Great advances have been made in understanding how contact reduces prejudice (Brown & Hewstone, 2005), highlighting the importance of affective mediators (Pettigrew & Tropp, 2008). The present study, a 3-wave longitudinal study undertaken among minority-status Colored high school children in South Africa (N = 465), explored the full mediation of the effects of cross-group friendships on positive outgroup attitudes, perceived outgroup variability, and negative action tendencies via positive (affective empathy) and negative (intergroup anxiety) affective mediators simultaneously. The target group was the majority-status White South African outgroup. As predicted, a bidirectional model described the relationship between contact, mediators, and prejudice significantly better over time than either autoregressive or unidirectional longitudinal models. However, full longitudinal mediation was only found in the direction from Time 1 contact to Time 3 prejudice (via Time 2 mediators), supporting the underlying tenet of the contact hypothesis. Specifically, cross-group friendships were positively associated with positive outgroup attitudes (via affective empathy) and perceived outgroup variability (via intergroup anxiety and affective empathy) and were negatively associated with negative action tendencies (via affective empathy). Following Pettigrew and Tropp (2008), we compared two alternative hypotheses regarding the relationship between intergroup anxiety and affective empathy over time. Time 1 intergroup anxiety was indirectly negatively associated with Time 3 affective empathy, via Time 2 cross-group friendships. We discuss the theoretical and empirical contributions of this study and make suggestions for future research.