The purpose of this study was to understand the reading performance of subgroups of language minority students and examine whether a research-based academic vocabulary intervention, Word Generation, has differential effects on these students’ academic vocabulary knowledge. Thirteen middle schools, propensity-score matched based on their achievement and demographic data, were randomly assigned to either treatment (n = 3,539) or control (n = 2,630) conditions. Students in both conditions were classified as either English-only (EO) or language minority students. The language minority students were further grouped as either being initially fluent English proficient (IFEP), redesignated fluent English proficient (RFEP), or limited English proficient (LEP). Multivariate analysis of variance and hierarchical linear models revealed three important findings. First, while LEP students’ scores on reading measures were significantly below those of the EO students, RFEP students’ scores were comparable to EO students’ scores. In addition, IFEP students’ scores were higher than those of the EO students. Second, there were variations within the RFEP students when they were disaggregated by time since redesignation; RFEP students’ reading scores were positively correlated with time since redesignation. Third, the treatment effect emerged only as an interaction with RFEP status. This study suggests that the benefits of a research-based intervention may vary according to students’ level of English proficiency.