To address a gap in the translational research on marital and family conflict intervention, we used a series of multilevel models to examine how interparental constructive conflict behavior, conflict resolution, and conflict knowledge improved up to 3 years postintervention in the Family Communication Project (FCP). The FCP was an experimental, longitudinal intervention focused on family conflict, communication, and adolescent adjustment. Participants consisted of adolescents and their two primary caregivers (N = 225 families). Each family was randomly assigned to one of two treatment groups (parent-adolescent or parent only) or two control groups (self-study or no treatment). Relative to controls, who showed no statistically significant improvement in conflict behavior, resolution, or knowledge, parents in the two-treatment conditions showed improvement in a nonlinear fashion. Modeled with piece-wise slopes, average change trajectories showed statistically significant increases from pretest to posttest, followed by declines but overall significant sustained growth by 3 years. Baseline marital hostility moderated this change, whereby couples who had higher levels of baseline marital hostility responded more favorably to treatment, but also declined more rapidly. Fathers in the study were shown to have stronger long-term treatment effects for conflict knowledge relative to mothers, and higher marital adjustment among fathers at baseline predicted less decline in dyadic conflict resolution after the intervention. The results highlight how examining within-family changes over time and between-family differences in change trajectories can contribute to practical insights about when and for whom interventions are most effective.