This paper reports on a randomized control trial involving children less than 3 years old and their mothers who were regarded at risk of maltreating their children by referral agencies. Mothers’ risk status derived from a heavy trauma burden (average exposure over the first 18 years of their lives to 10 possible adverse childhood experiences [ACEs] was >5), mental health challenges (15%–28% had experienced a prior psychiatric hospitalization), and prior removal of a child to foster care (20%). Mothers were randomly assigned to either a widely used parenting class known as Systematic Training for Effective Parenting (STEP) or the Group Attachment-Based Intervention (GABI), a multifamily 26-week treatment. The resulting mother–child pairs available for consideration in this baseline versus end-of-treatment report were 35 families in the STEP arm and 43 families in the GABI arm. The focus of this paper is the outcome measure of observed parent–child relationship assessed with the Coding of Interactive Behavior (Feldman, 1998) collected at baseline and end of treatment. In comparison to STEP, results indicated that GABI was linked to significant improvements in maternal supportive presence and dyadic reciprocity, and significant declines in maternal hostility and dyadic constriction (proxies for risk of child maltreatment). These medium-to large-sized effects remained significant even after controlling for mothers’ prior ACEs in analysis of covariance procedures. In addition, two small interaction effects of ACEs by treatment type were found, underlining the need for, and value of, treatments that are sensitive to parents’ traumatic histories.