The objective of this study was to test a social interactional model of physical aggression. Specifically, this model hypothesizes that the performance of physical aggression has its roots in socialization settings that are characterized by high densities of aversive stimuli and that provide frequent reinforcement for escalation to high intensity aversive behavior during social conflict. Social conflicts were observed during 10 hr of interaction of each of 20 mothers and their 5-year-old sons; half of the sons were selected based on evidence of frequent aggression in home and school settings. Simple descriptive and sequential analyses indicated that aggressive relative to nonaggressive dyad members were more likely to engage in conflict, engaged in longer conflicts, were more likely to escalate to higher levels of aversiveness, and were less likely to de-escalate the intensity of conflict. In aggressive and nonaggressive dyads, the cessation of conflict contingent on the escalation of one dyad member was reliably associated with an increased likelihood of escalation and with escalation to higher levels of aversiveness by that member in the subsequent conflict. However, escalation occurred more frequently and was more likely to result in cessation of conflicts in aggressive than nonaggressive dyads.