In a community sample of 116 children, assessments of parent-child interaction, parent-child attachment, and various parental, child, and contextual characteristics at 15 and 28 months and at age 5 were used to predict externalizing behavior at age 5, as rated by parents and teachers. Hierarchical multiple regression analysis and path analysis yielded a significant longitudinal model for the prediction of age 5 externalizing behavior, with independent contributions from the following predictors: child sex, partner support reported by the caregiver, disorganized infant-parent attachment at 15 months, child anger proneness at 28 months, and one of the two parent-child interaction factors observed at 28 months, namely negative parent-child interactions. The other, i.e., a lack of effective guidance, predicted externalizing problems only in highly anger-prone children. Furthermore, mediated pathways of influence were found for the parent-child interaction at 15 months (via disorganized attachment) and parental ego-resiliency (via negative parent-child interaction at 28 months).