Predictors of Bilingual Proficiency and Reading: A Study on Latina/o Children in Dual-Language Programs

• In Event: Equitably Supporting the Academic Achievement of Multilingual Learners in the United States

Abstract

Purpose
The number of school-age dual language learners (DLLs) in the United States has been rapidly increasing in the past decades (Huang et al., 2020; Mancilla-Martinez et al., 2020). Against this background, the current study sought to examine the predictors of bilingual proficiencies and English reading of Latina/o DLLs who were enrolled in one-way dual language immersion (DLI) programs and received instruction in both English and Spanish.

Theoretical Framework
Grounded in the Simple View of Reading (SVR) (Hoover & Gough, 1990) and the Linguistic Interdependence Hypothesis (LIH) (Cummins, 1991), this study developed a theoretical model based on previous studies on predominantly DLL children in English-only programs.

Methods
This study aimed to evaluate the model’s validity via path analysis and test whether the model is invariant across first and third graders.

Data Sources
The 96 DLLs in the present study were first (n = 47, nfemale = 27 & nmale = 20) and third (n = 49, nfemale = 24 & nmale = 25) graders from four elementary schools in the Southwestern United States. All DLLs spoke Spanish at home and were designated as English learners.

Results
The path analysis model revealed a good model fit, χ2 (df = 6) = 8.77, p = .1868, CFI = 0.991, TLI = .973, RMSEA = 0.069, SRMR = 0.037. When examining the prediction of the three exogenous variables (i.e., Teacher reported English input/output, Teacher reported Spanish input/output, and Child-reported English language input/output) on English and Spanish proficiency (see Figure 1), the analyses revealed that both teacher reported English input/output (β = .55) and child-reported English language input/output (β = -.22) significantly predicted English proficiency, whereas only child-reported English language input/output (β = -.47) significantly predicted Spanish proficiency. Collectively, these exogenous variables explained about 20% of the variance in English proficiency (R2 = .27) and Spanish proficiency (R2 = .22). When examining the relationship between the mediator variables (i.e., English and Spanish proficiency) and the endogenous variables (i.e., Reading comprehension and Decoding), the results indicated that both English and Spanish language proficiency significantly predicted reading comprehension (β = .65 and .21, respectively) and decoding (β = .56 and .32, respectively). The effect sizes for reading comprehension (R2 = .60) and decoding (R2 = .57) were both large in magnitude. Finally, invariance analyses revealed that with the exception of the path from child-reported English language input/output to decoding (not estimated in Figure 1), the path coefficients were equal across grade levels.

Significance
In summary, these results support the robust role of input/output in bilingual language development, the SVR, and the LIH. The associations between oral language and English reading corroborated other studies and suggested the importance of developing DLLs’ oral language to support their English reading (Petscher et al., 2020). The results also showed that DLLs’ Spanish skills contributed positively to their English reading. The study made a novel contribution by extending the results for DLLs in English-only programs to DLLs who receive more balanced input/output and formal instruction in both English and Spanish.

Authors

• Becky H. Huang, The University of Texas - San Antonio

• Daniel Sass, The University of Texas - San Antonio