Swanson, H. L., Kudo, M., & Guzman-Orth, D. (2016). Cognition and literacy in English language learners at risk for reading disabilities: A latent transition analysis. Journal of Educational Psychology, 108(6), 830–856.
In the United States, English language learners (ELLs) whose primary language is Spanish disproportionately experience reading difficulties (RD) relative to their monolingual peers (Lesaux, Rupp, & Siegel, 2007). Despite lower average reading performance, ELLs are underrepresented in special education (U.S. Department of Education, 2004). This discrepancy suggests that educators may need better methods for accurately identifying ELLs at risk for RD.
Examining a large sample of ELLs whose first language is Spanish in grades 1 through 3, Swanson, Kudo, and Guzman-Orth (2016) sought to identify children at risk for RD, assess how risk status changes over 3 years, and determine the reading, language, phonological, and cognitive processes that differentiate students at risk and not at risk for RD. The authors addressed three research questions:
Using data from a larger longitudinal study, the authors studied 489 children in grades 1 (n = 163), 2 (n = 153), and 3 (n = 173). All students (a) were Latino with a primary language of Spanish, (b) were identified as ELLs according to their district-administered English language proficiency assessment, (c) scored above a standard score of 80 on measures of intelligence, and (d) primarily received reading instruction in English. Swanson and colleagues followed students in grades 1 through 3 for 3 years and tested students each year (e.g., students who began the study in grade 2 were tested in grades 2 though 4).
The first two questions aimed to determine whether a discrete subgroup of children at risk for RD emerged within a sample of ELLs in grades 1 through 3 and the extent to which students in these subgroups maintained or changed group membership over a 3-year period. To address the first question, the authors used a statistical method called latent class analysis to identify a subgroup of ELLs at risk for RD distinguished by similar patterns of responses on multiple assessments. Like other types of structural equation modeling, latent class analysis explains relations between unobserved constructs (latent variables) and observed variables. Specifically, latent class analysis classifies participants into discrete classes or groups based on their pattern of responses within a measurement model. The authors defined students as at risk for RD based on poor performance in reading coupled with adequate achievement in math and attention. Standardized measures of reading and language in both Spanish and English were used to classify children as at risk, as well as measures of math achievement and teacher ratings of attention. The second question, which examined how group membership changed over time, was addressed using latent transition analysis, a methodological approach that allows for researchers to test whether students change membership over time.
The third research question aimed to determine the phonological, oral language, and cognitive processes that predict ELLs at risk for RD. Cognitive measures included naming speed, inhibition, and work memory, as measured by assessments of controlled attention. Measures of phonological processing, oral language, naming speed, inhibition, and memory were used to determine the processes that predict ELLs' status as at risk for RD.
The study produced two key findings. First, results showed that subgroups of students remained relatively stable over time. Four latent subgroups appeared across the three testing time points: (1) balanced-bilingual good readers, (2) nonbalanced-bilingual good readers, (3) classroom inattentive/low English comprehension children, and (4) at risk for RD. Balanced bilingual students referred to students who demonstrated comparable levels of competencies in both languages, whereas nonbalanced bilinguals displayed greater proficiency in one language (in this study all of the nonbalanced bilinguals were stronger in English than Spanish). Across the three testing waves (i.e., time points), 23% of the students were determined as at risk for RD. Results showed high stability for the at risk for RD category across time points, with all students identified at Wave 1 remaining at risk at Wave 2, and 91% identified as at risk for RD at Wave 3. Children in the low attentive/comprehension group at Wave 1 had a 9% chance of transitioning to the RD latent group at Wave 2; however, no additional students transitioned to the RD latent group at Wave 3. Of the ELLs who were initially classified as good readers at Wave 1, 95% remained good readers at Waves 2 and 3. Of the students who were initially identified as good readers and transitioned to the at-risk subgroup, all but one (eight out of nine) were in the nonbalanced-bilinguals subgroup. The overall pattern of findings did not align with previous research suggesting a considerable proportion of students present late-emerging reading difficulties (i.e., 18%; Compton, Fuchs, Fuchs, Elleman, & Gilbert, 2008); however, the results suggest that nonbalanced good readers may be more susceptible to late-emerging reading difficulties than language-balanced good readers.
Second, students at risk for RD vary from other ELL subgroups based on their performance on phonological, oral language, and cognitive measures. Study results suggested that phonological and oral language processes differentiate students at risk for RD from those not at risk, with phonological processes playing a particularly prominent role in grades 1 through 3. In the later grades (grades 3 through 5), working memory measures played a unique role in predicting the at risk class for RD. The authors reasoned that working memory may be particularly important in the later grades when there is increased emphasis on reading comprehension. These findings align with previous research suggesting that working memory and reading are associated in monolingual (e.g., Siegel & Mazebel, 2013) and multilingual samples (Genesee & Geva, 2006), as well as previous research underscoring the importance of phonological and oral language processes (see review by Lesaux & Geva, 2006).
There are a few limitations worthy of consideration. Using latent class modeling approaches, Swanson and colleagues modeled risk status for RD categorically (i.e., students were either at risk for RD or not). The authors acknowledge that longstanding research suggests that reading risk status occurs in varying levels of severity and may be best modeled continuously rather than dichotomously (i.e., above or below a cutoff point; Shaywitz, Escobar, Shaywitz, Fletcher, & Makuch, 1992). Although the advanced latent class modeling approaches used in the present study are well-matched to the research questions and study findings extend the knowledge base on ELLs at risk for RD, it is worth noting students who performed just above and below the cut point for risk status may be more similar than represented by the latent groups. Relatedly, the current pattern of findings represents the data when the 25th percentile was used to determine risk status; however, a different pattern of findings may have emerged given a different cut point. For instance, if one were interested in examining differences between students with and without RD, a more stringent cut point would be used (e.g., 10th percentile), which may influence the study findings. Another limitation relates to the information collected about the instruction that participating students received. The researchers observed classroom reading instruction to determine whether critical elements of elementary reading instruction (e.g., phonics, vocabulary) were addressed in classrooms of participating students; however, no information was collected about supplemental reading instruction that occurred outside general education instruction. Thus, this study was unable to examine whether particular reading intervention programs influenced the stability of risk classification. Lastly, there are not commonly adopted procedures for conducting power analyses for latent transition analysis, so it is difficult to determine whether this study was adequately powered.
The most important recommendation based on this study relates to the early identification of ELLs at risk for RD. Based on the high stability of the at risk for RD subgroup, it appears that it is possible to identify students at risk for RD in the early grades. The stability of the at risk for RD subgroup is noteworthy, as classroom observation data indicated that teachers implemented comprehension and decoding instructional practices grounded in previous research. The stability of the at risk for RD subgroup suggests that these students are unlikely to become good readers even when general education classroom instruction targets the important areas of reading and may need intensive interventions. Another important study implication for practitioners is that good readers are unlikely to be identified as at risk for RD in later grades. Study results showed that less than 5% of students initially identified as good readers in grades 1 through 3 were later classified as at risk for RD. This study finding suggests that although it may be worth monitoring the reading performance of students initially identified as good readers, schools would be wise to target students at risk for RD for supplemental reading interventions.
Compton, D. L., Fuchs, D., Fuchs, L. S., Elleman, A. M., & Gilbert, J. K. (2008). Tracking children who fly below the radar: Latent transition modeling of students with late-emerging reading disability. Learning and Individual Differences, 18, 329–337. doi:10.1016/j.lindif.2008.04.003
Genesee, F., & Geva, E. (2006). Cross-linguistic relationships in working memory, phonological processes, and oral language. In D. August & T. Shanahan (Eds.), Developing Literacy in Second-Language Learners: Report of the National Literacy Panel on Language-Minority Children and Youth (pp. 175–183). Mahwah, NJ: Erlbaum.
Lesaux, N., & Geva, E. (2006). Synthesis: Development of literacy in language-minority students. In D. August & T. Shanahan (Eds.), Developing literacy in second-language learners: Report of the National Literacy Panel on Language Minority Children and Youth (pp. 53–74). Mahwah, NJ: Erlbaum.
Lesaux, N. K., Rupp, A. A., & Siegel, L. S. (2007). Growth in reading skills of children from diverse linguistic backgrounds: Findings from a 5-year longitudinal study. Journal of Educational Psychology, 99, 821–834.
Shaywitz, S. E., Escobar, M. D., Shaywitz, B. A., Fletcher, J. M., & Makuch, R. (1992). Evidence that dyslexia may represent the lower tail of a normal distribution of reading ability. The New England Journal of Medicine, 326, 145–150.
Siegel, L. S., & Mazebel, S. (2013). Basic cognitive processes and reading disabilities. In H. Swanson, K. Harris, & S. Graham (Eds.), Handbook of learning disabilities (pp. 186–214). New York, NY: Guilford Press.
U.S. Department of Education. (2004). National assessment of educational progress. Retrieved from https://nces.ed.gov/nationsreportcard/about