This report describes a secondary analysis of data from a comprehensive intervention project which included training and structural changes in three Baby Homes in St. Petersburg, Russian Federation. Multiple mediator models were tested according to the R.M. Baron and D.A. Kenny () causal-steps approach to examine whether caregiver–child interaction quality, number of caregiver transitions, and group size mediated the effects of the intervention on children's attachment behaviors and physical growth. The study utilized a subsample of 163 children from the original Russian Baby Home project, who were between 11 and 19 months at the time of assessment. Results from comparisons of the training and structural changes versus no intervention conditions are presented. Caregiver–child interaction quality and number of caregiver transitions fully mediated the association between intervention condition and attachment behavior. No other mediation was found. Results suggest that the quality of interaction between caregivers and children in institutional care is of primary importance to children's development, but relationship context may play a less direct mediational role, supporting caregiver–child interactions.