The national Head Start Impact Study (HSIS) estimated the average impact of an offer of Head Start treatment ("Intent-to-Treat," or ITT). The HSIS was an experimental study of a nationally-representative sample of 4,440 preschoolers, across 378 centers, in 22 states, with participating children being randomized to an offer of one year's attendance in the Head-Start Program versus assignment to a control condition, under which no offer was made but families were free to continue with whatever child-care arrangements they favored personally. The impact study found that an offer of one year's attendance in the Head Start program had small impacts on children's language and literacy. Additionally, and most interestingly, the HSIS reported that an offer of program attendance produced larger impacts among Latino Dual Language Learners (DLL,) but the question remains why these particular children benefitted from the program more than did their English-speaking peers. However, the evaluation did not investigate whether changes in parenting practices mediated these program impacts on children's learning. In this thesis, I argue that a study of the key mechanisms through which the program impacted child outcomes remains central to understanding why Head Start improved children's language and literacy. Thus, in my thesis, I have unpacked the mechanisms that mediated these detected effects--through parental practices--using two complementary estimation strategies: [1] multilevel structural-equation modeling and [2] average causal mediation effect estimation, by reanalyzing the original study data. A central aim of my research was to contribute to the body of early childhood research and inform policy directions and program development by: (a) investigating whether ITT effects on early child language outcomes were mediated through parent-child language-and-literacy activities, and (b) conducting multi-group comparisons to test whether the impact of these mediational pathways differed by the child's DLL status. I found that, on average, assignment increased children's vocabulary and reading scores (effect sizes =+.13; e.s.=+.17), respectively. The randomized offer of Head Start also increased the frequency of parent-child language-and-literacy activities (e.s.= +.25). This impact was larger for Latino parents of Spanish-speaking DLL. Additionally, I found statistically significant indirect effects: 14% of the total impact on vocabulary scores and 18% of the total impact on reading scores were mediated through parent-child language-and-literacy activities. In addition, the causal mediation effects of program impact on vocabulary and reading differed by DLL status: 12% of the impact on vocabulary was mediated through parent-child language-and-literacy activities for DLL children, compared with 18% for non-DLL. And for reading, 37% of the impact was mediated through parent-child language-and-literacy activities for DLL children vs. 4% for non-DLL children. I conclude with important directions for how early childhood programs can improve parental investment in early learning for diverse groups of children, and explanations for why mediated effects differed by language status. (author abstract)