Solari, E. J., Petscher, Y., & Folsom, J. S. (2014). Differentiating literacy growth of ELL students with LD from other high-risk subgroups and general education peers: Evidence from grades 3–10. Journal of Learning Disabilities, 47(4), 329–348.
One-fifth of the school-age U.S. population speaks a language other than English (U.S. Department of Education, 2003). By 2030, 40% of public school students are expected to be from households in which English is not the primary language, with more than three-fourths speaking Spanish as their first language (Hemphill & Vanneman, 2011). Many of these students are identified as English language learners (ELLs) because their English language skills are limited to the point where, in general, they do not benefit from general education instruction without special support in English language development (Ortiz & Kushner, 1997). One consequence of limited English proficiency is that ELLs score lower, on average, than non-ELLs on measures of reading achievement. The gaps are significant and persistent (National Center for Education Statistics, 2012).
A minority of ELLs also have a learning disability (LD), although identifying ELLs as having an LD can be challenging, particularly for younger ELLs. Teachers in the primary grades are often reluctant to consider LD as a cause for a student’s below-average achievement when limited English represents a compelling alternative. Indeed, there is evidence that ELLs are underrepresented as having an LD in the early elementary grades but tend to be overrepresented in later elementary grades, although there is little consensus on this basic finding and even less agreement on the point (e.g., grade, age) at which ELLs transition from underrepresentation to overrepresentation. A better understanding of the pattern of LD identification among ELLs may help to inform policy and practice related to making special education services available to ELLs who may benefit from the additional support.
This study analyzed data from the Florida Assessments for Instruction in Reading (Florida Department of Education, 2009a, 2009b), specifically the Reading Comprehension, Text Reading Efficiency, and Word Analysis subtests for students in grades 3 and 10. The research questions were the following:
The Florida Assessments for Instruction in Reading dataset includes more than 1,000,000 cases from grades 3 through 8. It also includes information on the schools students attended and their teachers at each grade level. This information allowed the investigators to account for differences due to the school and teacher before considering the impact of LD status, ELL status, or poverty on reading status or reading achievement.
The prevalence for the three risk groups at each grade level is summarized in Figure 1. Figure 1 also represents the impact of low SES. The most prevalent groups in the Florida sample are ELLs from families with low SES and students with LD from families with low SES.
Figure 1. Rates of identification for high-risk subgroups separated by SES (Solari et al., p. 335).
The authors also calculated a phi coefficient to estimate the association between poverty status and risk group (ELL and LD, LD, ELL). As a rule of thumb, phi values of .10 represent small degrees of association. A phi of .30 signifies moderate association, and a phi of .50 suggests a considerable association. Figure 2 demonstrates very little association between low SES and LD (phi from .04 to .07) or low SES and ELL and LD status (phi from .04 to .06). A small association was observed between low SES and ELL status (phi from .13 to .17). All of the phi coefficients were statistically significant, due largely to the sample size (n > 1,000,000). The notable finding is that low SES was more strongly associated with being an ELL than with being LD or even being ELL and LD.
Figure 2. Effect sizes for prevalence of risk status (Solari et al., p. 335).
Across the three reading outcomes, the LD and ELL group with low SES performed significantly lower than the other subgroups in grades 3–10. The ELL and LD group performed at lower levels than the ELL or non-ELL and LD groups. The non-ELL, non-LD, non-low-SES students outperformed all other groups, on average.
The study replicates other research that demonstrates a substantial difference in the prevalence of low SES among ELLs. The study also confirms the findings of others (e.g., Kieffer, 2010; Roberts, Mohammed, & Vaughn, 2010) that all subgroups experience improved reading skills over time; however, there is little or no narrowing of the achievement gap between the high-risk subgroups and the general education subgroups. The differences that exist at pretest are maintained over time, on average. Third, across all grade levels, the ELL and LD subgroup performed lower than all other subgroups on all measures. This finding applies to the ELL and LD group regardless of SES status, which means that ELL and LD students are at particular risk of reading failure. Finally, across all subgroups, students from low SES settings performed more poorly on all reading measures than similar peers (i.e., members of the same risk group) from families with average or above-average SES.
It is important to acknowledge that older students in this sample may have been identified as ELLs in the past. For example, some students who were ELLs in the third and fourth grades were reclassified as non-ELL in the eighth grade. Among these students, a subgroup was identified as LD. This group of students is included in the LD subgroup, even though their language skills may still lag behind LD students whose first language is English. Many students who are capable English speakers in a social context (and thus no longer classified as ELL) lack the academic English necessary for school success. Whether LD or not, former ELLs who struggle to understand written English may benefit from ongoing support in academic English, particularly older students who are former ELLs.
The larger point is that datasets like the Florida Assessments for Instruction in Reading should be used with caution. The authors carefully note the potential for drawing improper inferences about the "cause" of correlated factors such as SES, language, and disability when important information may be missing (e.g., ELL status at earlier points in time). The Florida database is a terrific resource, and the "mining" of such data is increasingly common for purposes of making policy decisions and informing practice. "Big data" and the analytic strategies for applying it to policy and practice questions represent valuable tools for educational policymakers to the extent that users are mindful of pertinent limitations and qualify their recommendations accordingly.