The Effect of Instructional Program and Phonological Awareness on Reading Outcomes Among Early Elementary Spanish-Speaking English Learners

Background: English Learners (ELs), the majority of whom are Spanish-speaking children, are at risk for academic underachievement in American public schools. Prior research on the effectiveness of instruction designed to support their learning, categorized as bilingual or immersion, is often confounded by selection bias and has not produced definitive conclusions. A recent study suggests that ELs with low English phonological awareness (PA) may benefit more from bilingual than immersion instruction in terms of reading achievement, but selection bias casts doubt on this relationship. Purpose: This study sought to examine if Spanish-speaking EL students with low English PA benefit more from bilingual than immersion instruction while controlling for selection bias using a matched-comparison group design. Preliminary analysis showed, however, that sample size-related challenges required inclusion of ELs of varying PA levels. It was hypothesized that PA would interact with instructional program, and that students with lower English PA would benefit more from bilingual than immersion. Method: From a sample of 689 Spanish-speaking ELs, matched-comparison groups (n = 45 per group) were formed using propensity score matching that improved balance between immersion and bilingual groups on beginning-of-first-grade English and Spanish reading, oral language, and PA. ANCOVA was used to evaluate if instructional program interacted with PA to influence end-of-second-grade reading achievement. Results: Interactions between instructional program and PA were not significant, nor were main effects, except for that of instructional program on Spanish reading, with bilingual students outperforming immersion students. A post-hoc power analysis indicated that the study was largely underpowered. Conclusion: Results suggest that bilingual education more effectively promotes Spanish reading than immersion. Detection of effects in English reading was limited due to power, highlighting the need to take a more methodologically rigorous approach to studying this population, balancing priorities of detecting treatment effects and minimizing selection bias.