Many studies have pointed to a troubling phenomenon known as the "immigrant paradox." Despite an initial advantage length of residence in the United States appears to be associated with declining academic achievement and aspirations. To date, this line of research has taken a largely cross-sectional approach, comparing first, second, and third generations. The Longitudinal Immigrant Student Adaptation Study (LISA) combines longitudinal, interdisciplinary, and comparative approaches to document the patterns of adaptation of 408 recently arrived immigrant origin youth from Central America, China, the Dominican Republic, Haiti, and Mexico over the course of five years. Here, we present data that demonstrate patterns of academic engagement and achievement of these youths over time, as well as a structural equations model (SEM) that sheds light on the factors contributing to these patterns. These data suggest that supportive relationships significantly mediate the academic engagement and outcomes of immigrant youth. Implications and future directions are discussed. (Contains 4 tables, 5 figures, and 1 note.)