Governments around the world are increasing investments in early childhood education as a way to promote children's learning and development. As research grows on the longer-term effects of early educational programs, some have hypothesized that sustained impacts may depend on the quality of children's subsequent classroom environments and may be more likely to occur in certain nonacademic domains. This study is the first to examine these questions in sub-Saharan Africa, using longitudinal data from a school-randomized trial in Ghana. A 1-year teacher-training and coaching program, delivered with (TTPA) and without (TT) parental-awareness meetings, was implemented when children were enrolled in 1 of 2 grades of preprimary school. Previous studies showed that in Year 1, the implementation year, the TT treatment had positive impacts on children's literacy, numeracy, and social-emotional development, while the TTPA treatment yielded no positive impacts for children in both grades. In this study, I examine impacts in Year 3, at the end of children's 1st or 2nd year of primary school. I find persistent positive impacts of TT on literacy, and new negative impacts of TTPA on numeracy, but these depend on levels of classroom emotional support and teacher burnout in primary school. In addition, there were small persistent impacts of both treatments on children's executive function, and these were not conditional on subsequent classroom environments.