This investigation is a secondary analysis of data from a randomized control trial of the PLAY Home Consultation Intervention Program which was conducted with 112 preschool children with Autism Spectrum Disorders and their parents (Solomon et al. in J Dev Behav Pediatr 35:475–485, 2014). Subjects were randomly assigned to either a community standard (CS) treatment group or to the PLAY Project plus CS Treatment (PLAY). PLAY subjects received monthly parent–child intervention sessions for 1 year during which parents learned how to use the rationale and interactive strategies of the Developmental, Individual-differences, Relationship-based (DIR) intervention model (Greenspan and Weider in The child with special needs: encouraging intellectual and emotional growth. DeCapo Press, Cambridge, MA, 1998) to engage in more responsive, affective and less directive interactions with their children. This investigation examined whether PLAY intervention effects on parents’ style of interacting with their children as well as on children’s social engagement mediated the effects of PLAY on children’s autism severity as measured by ADOS calibrated severity scores. Regression procedures were used to test for mediation. There were two main findings. First the effects of PLAY on children’s social engagement were mediated by the increases in parental responsiveness and affect that were promoted by PLAY. Second, the effects of PLAY on the severity children’s Social Affect disorders were mediated by changes in parental responsiveness and affect; however, the effects of Responsive/Affect were mediated by the impact these variables had on children’s social engagement. Results are discussed in terms of contemporary models of developmental change including the developmental change model that is the foundation for DIR.