Observations of aggressive interactions in boys' laboratory play groups were used to evaluate the relative importance of relational and individual factors in accounting for aggressive acts. A classroom peer-rating method for identifying mutually aggressive dyads was validated in 11 5-session play groups, composed of 2 mutually aggressive boys and 4 randomly selected male classmates from 11 predominately African American 3rd-grade classrooms. When the social relations model was used, relationship effects accounted for equally as much of the variance in total aggression and proactive aggression as either actor or target effects. Mutually aggressive dyads displayed twice as much total aggression as randomly selected dyads. Members of mutually aggressive dyads attributed greater hostile intentions toward each other than did randomly selected dyads, which may serve to explain their greater aggression toward each other. The importance of studying relational factors, including social histories and social-cognitive processes, is discussed.