Exploring variability in code- and meaning-focused skills in English and Spanish: A latent profile analysis of bilingual children

Purpose: Bilingual students exhibit significant within-group variability in their reading skills; and yet, studies of bilingual reading development have typically focused on whole-sample averages and correlations (Halpin et al., 2021). Person-oriented approaches allow analysis of heterogeneous profiles of learners. The purpose of the proposed work is to (a) describe latent profiles of language and literacy skills in Spanish-English bilinguals and (b) determine how age and educational context predict profile membership.Method: We administered assessments of English decoding and semantics (Assessment-to-Instruction; A2i, Connor et al., 2022) as well as parallel experimental assessments of Spanish decoding and semantics (Authors, 2021) to 855 Spanish-English bilingual children in PreK to second grade. A multilevel LPA model was fit with differing numbers of latent classes. The fits of the models were compared using information criteria and approximate likelihood ratio tests to determine the optimal number of latent classes.Results: The LRM-LRT and VLMR-LRT indicated that, for kindergarten and first grade, a 3-class model was optimal. Classes seem to be ordered in their means, with Class 1 having the lowest means, Class 2 having slightly higher means in at least one language, and Class 3 having the highest means. For 2nd grade, a 5-class model with more differentiation between high and low performers was the best fit.Conclusion: Bilingual children’s English and Spanish skills developed primarily in tandem, suggesting cross-linguistic sharing of knowledge across domains. Educational context was related to skills in both languages, with children in dual language education tending to cluster into higher-performing profiles.