Disentangling Language and Disability: Exploring Special Education Eligibility by Language Status

• In Event: Identifying and Supporting Multilingual Learners Across Systems and Classrooms

Abstract

Objectives:
Much of the prior research on special education eligibility and student characteristics has focused on race and socioeconomic status (e.g., Kincaid & Sullivan, 2017; Morgan et al., 2018). However, more recently, researchers have begun to examine the role of language background (e.g., Mancilla-Martinez et al., 2024; Yamasaki & Luk, 2018). While there have been consistent findings of disproportionality, the nature of misrepresentation has varied, with some studies showing underrepresentation for linguistically diverse students and others showing overrepresentation. The aim of the current study was to both further contribute to this growing body of work by examining rates of eligibility for Specific Learning Disability (SLD) and Communication Disorder (CD) by language status and to assess how classification and covariates shape the results.

Methods:
Anonymized demographic data from 729,868 students in grades 3-8, collected by the Massachusetts Department of Elementary & Secondary Education, were analyzed from the 2018–2019 (n = 377,390) and 2022–2023 (n = 352,478) school years. Students were categorized as Emerging Bilinguals (EBs), English Proficient Bilinguals (EPBs), Former English Learners (FELs), or Native English Speakers (NESs), and as either having no disability or receiving special education services for SLD or CD. Logistic regressions were used to predict special education eligibility first from a binary version of language status (Second-Language English Speakers, L2Es vs. NESs) along with covariates (i.e., SES, race, gender, and English language skill). Then, from the four levels of language status (EBs, EPBs, and FELs vs. NESs) without and with covariates.

Results:
Findings were consistent across both cohorts, with relative consistency but some variability by grade. In general, when language status was dichotomized, L2Es were less likely than NESs to be identified with SLD (ORs = .36-.57) and CD (ORs = .54-.68). When using the four-group language variable, without covariates, EPBs and FELs showed underrepresentation for SLD (EPB: ORs = .33-.57; FEL: ORs = .15-.71), while EBs showed overrepresentation (ORs = 1.46-1.95). A similar pattern was observed for CD, with EBs significantly overrepresented (ORs = 1.49-2.63). EPBs and FELs were comparable to NESs in 2018-2019 but showed some underrepresentation in 2022-2023. With covariates included, all three groups had significantly lower odds of SLD eligibility compared to NESs (EB: ORs = .34-.60; EPB: ORs = .45-.66; FEL: ORs = .28-.70). For CD, EBs had lower odds (ORs = .38-.68), while FELs and EPBs showed no meaningful differences in 2018-2019; and in 2022-2023, FELs show no meaningful differences, but, EPBs showed significantly lower odds (ORs = .52-.63).

Significance:
Across cohorts, classification methods, and models, significant misrepresentation of linguistically diverse students relative to NESs was found for both SLD and CD at multiple grade levels. However, the direction and degree of this misrepresentation varied depending on how language status was defined and how models were specified. These findings align with prior research and underscore the importance of considering students' language backgrounds when evaluating special education eligibility. They also demonstrate that methodological decisions can meaningfully shape results and implications and, thus, should be made carefully.

Authors

• Brianna Yamakasi, Emory University

• Min Hyun Oh, United States Department of Defense

• Gigi Luk, McGill University