Abstract
This study investigated the dimensionality of bilingual phonological awareness (PA) in English and Spanish by replicating a kindergarten model in Grade 1, and presents alternatives to modeling clustered data. English and Spanish tasks were analyzed from previously collected samples totaling 1,586 first grade Spanish-speaking English learners. Four distinct approaches to confirmatory factor analysis (CFA) models were examined: (a) uncentered student-level data, (b) student-level data centered at the classroom means, (c) classroom-level data, and (d) multilevel CFA. Results indicated that while the multilevel CFA provided the most comprehensive view of the data, the multi-level student-level estimates were not appreciably different from estimates based on student-level data centered at the classroom means, and multi-level classroom-level estimates were comparable to estimates based on the analysis of classroom means. Importantly, English and Spanish PA were statistically separable at the student-level, but minimally distinct (r = .86) and slightly less correlated than what has been reported for kindergarten (r = .93). At the classroom level, the correlation was moderate (r = .51), and substantially reduced compared to kindergarten (r = .83). The distinction at the classroom-level between kindergarten and Grade 1 implies that instruction differentiates the abilities across languages at the classroom-level, but less so at the student-level.