With data from the Early Childhood Longitudinal Study-Birth Cohort (n = 6,449), a nationally representative sample of births in 2001, we used hierarchical linear modeling to analyze differences in observed interactions between married, cohabiting, never-married, and divorced mothers and their children. In contrast to previous studies, we concentrated on early childhood, a developmentally critical period that has been understudied in the family structure literature, and relied on objective observational measures of mother-child interactions, which are unlikely to be biased by maternal perceptions of interactions with children. Nonmarital family structures were common in the lives of young children, as 32% lived outside of a married, biological parent home. Initial results indicated that married families were consistently associated with higher quality interactions. Moreover, though it was hypothesized that the presence of a biological father might be associated with higher quality interactions than single-parent households, this hypothesis was not confirmed. Additional models suggest that race and ethnicity moderated the effect of family structure, as non-Hispanic Black and non-Hispanic White participants showed little significant variation between married and unmarried families. Among Hispanics, mothers living in cohabiting, divorced, or single families, when compared to married mothers, exhibited more negative and more intrusive behaviors; cohabiting mothers also scored lower on the measure of cognitive stimulation. Results suggest that marriage may not be uniformly associated with higher levels of mother-child interactions and that cohabitation, particularly for Hispanics, may be associated with adverse outcomes. (Contains 6 tables.)