The present study focused on mother-child interaction predictors of initial levels and change in child aggressive and disruptive behavior at school from kindergarten to third grade. Aggression-disruption was measured via annual reports from teachers and peers. Ordinary least-squares regression was used to identify 8 separate child aggression trajectories, 4 for each gender: high initial levels with increases in aggression, high initial levels with decrease in aggression, low initial levels with increases in aggression, and low initial levels with decreases in aggression. Mother-child interaction measures of coercion and nonaffection collected prior to kindergarten were predictive of initial levels of aggression-disruption in kindergarten in both boys and girls. However, boys and girls differed in how coercion and nonaffection predicted change in aggression-disruption across elementary school years. For boys, high coercion and nonaffection were particularly associated with the high-increasing-aggression trajectory, but for girls, high levels of coercion and nonaffection were associated with the high-decreasing-aggression trajectory. This difference is discussed in the context of Patterson et al.'s coercion training theory, and the need for gender-specific theories of aggressive development is noted.