Purpose: This study investigated cumulative effects of language learning, specifically whether prior vocabulary knowledge or special education status moderated the effects of academic vocabulary instruction in high-poverty schools. Method: Effects of a supplemental intervention targeting academic vocabulary in first through third grades were evaluated with 241 students (6-9 years old) from low-income families, 48% of whom were retained for the 3-year study duration. Students were randomly assigned to vocabulary instruction or comparison groups. Results: Curriculum-based measures of word recognition, receptive identification, expressive labeling, and decontextualized definitions showed large effects for multiple levels of word learning. Hierarchical linear modeling revealed that students with higher initial Peabody Picture Vocabulary Test--Fourth Edition scores (Dunn & Dunn, 2007) demonstrated greater word learning, whereas students with special needs demonstrated less growth in vocabulary. Conclusion: This model of vocabulary instruction can be applied efficiently in high-poverty schools through an automated, easily implemented adjunct to reading instruction in the early grades and holds promise for reducing gaps in vocabulary development.