The first phase of parent-child interaction therapy (PCIT), called child-directed interaction, teaches parents to use positive and differential social attention to improve the parent-child relationship. This study examined predictors of change in mother and child functioning during the child-directed interaction for 100 mother-child dyads. The children were 3-6-years-old and diagnosed with oppositional defiant disorder. After establishing that significant improvements occurred in mother report of child disruptive behavior, parenting stress, and parenting practices, these three variables were combined to form a latent impaired mother-child functioning construct. Structural equation models were examined using maternal demographic and psychosocial variables as predictors of impaired mother-child functioning before and after the child-directed interaction. Mothers' self-reported daily hassles and depressive symptomatology predicted 74% of variance in impaired mother-child functioning before treatment. Mothers' report of social support predicted impaired mother-child functioning after the child-directed interaction, with 57% of the variance accounted for in this longitudinal model. These findings suggest the importance of improving maternal social support during the initial phase of PCIT.