The current study examined whether cumulative family risk would moderate the relation between regularity of attending Head Start and three child outcomes: receptive vocabulary, teacher ratings of social competence, and teacher ratings of following instructions. Cumulative family risk was the sum of four dichotomous measures: low income, low cognitive stimulation, intrusiveness, and depression. Participants were 94 Head Start children and their caregivers. All but 1 of the 16 classrooms attended were rated as good or better on the Early Childhood Environment Rating Scale (ECERS). Analyses revealed the relation between Head Start attendance and receptive vocabulary was moderated by cumulative risk, with children from higher risk families benefiting more. Regardless of cumulative family risk, attendance predicted teacher ratings of social competence; regardless of attendance, cumulative family risk predicted teacher ratings of following instructions. Results are interpreted as supporting a compensatory model of the impact of Head Start on children's receptive vocabulary and the use of attendance as a measure of the "value added" by Head Start. Public policy implications are discussed. (author abstract)