To investigate whether parenting or neonatal brain volumes mediate associations between prenatal social disadvantage (SD) and cognitive/language abilities and whether these mechanisms vary by level of disadvantage. Pregnant women were prospectively recruited from obstetric clinics in St Louis, Missouri. Prenatal SD encompassed access to social (eg, education) and material (eg, income-to-needs, health insurance, area deprivation, and nutrition) resources during pregnancy. Neonates underwent brain magnetic resonance imaging. Mother-child dyads (N=202) returned at age 1-year for parenting observations and at age 2-years for cognition/language assessments (Bayley-III). Generalized additive and mediation models tested hypotheses. Greater prenatal SD associated nonlinearly with poorer cognitive/language scores. Associations between parenting and cognition/language were moderated by disadvantage, such that supportive and non-supportive parenting behaviors related only to cognition/language in children with lesser prenatal SD. Parenting mediation effects differed by level of disadvantage: both supportive and non-supportive parenting mediated prenatal SD-cognition/language associations in children with lesser disadvantage, but not in children with greater disadvantage. Prenatal SD-associated reductions in neonatal subcortical grey matter (β=.19, q=.03), white matter (β=.23, q=.02), and total brain volume (β=.18, q=.03) were associated with lower cognition, but did not mediate associations between prenatal SD and cognition. Parenting moderates and mediates associations between prenatal SD and early cognition and language, but only in families with less social disadvantage. These findings, although correlational, suggest that there may be a critical threshold of disadvantage, below which mediating or moderating factors become less effective, highlighting the importance of reducing disadvantage as primary prevention.