Assessing Oral Language When Screening Multilingual Children for Learning Disabilities in Reading

Lesson 1: Define Multilingualism Inclusively

All children exposed to more than one language regularly are multilingual (Castilla-Earls et al., 2020). Across disciplines, many different terms are used to refer to students who use more than one language. These may include the terms English learners (EL), students with limited English proficiency, and language-minority students, among others. The use of different terminology across professions can lead to confusion, especially when some terms and definitions are tied to service eligibility. The term “multilingual learners” may be useful to more inclusively describe any child who speaks more than one language, apart from whether or not they use the languages equally. Multilingual learners is preferable as an umbrella term, rather than “bilingual learners” or “dual-language learners,” which refer specifically to children who are exposed to and use two languages. The following paragraphs describe why the lessons from this article should be applied to all multilingual learners. Failure to recognize that children’s language input is divided across two or more languages could have consequences when interpreting scores on language and/or reading assessments.

In schools, the designations used for multilingual learners are often tied to arbitrary thresholds of English language proficiency necessary to receive federally mandated services. Terminologies and categorizations used for service delivery can have unintended consequences when generalized inappropriately. For example, a student exposed to a non-English language at home who performs just above the threshold for English language proficiency standards would not qualify for the EL label, and therefore would not receive associated English language support services. Nevertheless, this student’s language and communication development would look different when compared with their monolingual English-speaking peers. To reduce misidentification of multilingual children, such differences between monolingual and multilingual children must be accounted for (even for multilingual children who are not identified as EL students).

Understanding the nature of language exposure and input among multilingual learners is crucial due to a systemic linguistic bias present in the United States. Because English is viewed as a “prestige” language both in U.S. schools and in global communication (Guerrero-Nieto, 2010), multilingual families may under-identify some or all non-majority languages spoken in the home. Families may experience both direct and indirect messaging from school staff perceived as devaluing their home language. For example, the limited training requirements for staff to work with multilingual children is likely to result in inaccurate assumptions being made about families’ language experiences (e.g., that all adults in the home speak English), or explicit inappropriate recommendations made for families (e.g., that families should speak only English as much as possible). As a result, students’ multilingual skills may not be accurately accounted for, potentially resulting in overidentification for special education services outside of language support services. For example, a family from Central America who speak Spanish, some English, and primarily Kaqchikel at home may only report that their child speaks English and Spanish due to common linguistic bias associated with speaking Kaqchikel, an Indigenous Mayan language. This could lead practitioners, unaware of the full extent of home language exposure, to misdiagnose the child with a language or learning disability, despite delays in acquisition of Spanish and English being consistent with typical development (given that the child’s language input is distributed across three languages). In practice, obtaining a thorough case history to determine all exposed languages is crucial for optimal assessment and screening procedures and for minimizing overidentification of language and learning disabilities among multilingual children. Because language and literacy development are highly dependent on the amount of language input, opportunities to use language, and exposure to high-quality instruction in each language, any child exposed to more than one language regularly should be considered a multilingual learner.

Lesson 2: Include Comprehensive Language Assessment

For languages with a writing system, children acquire oral language skills prior to acquiring written language skills. Early language learning experiences, including exposure to the sound system (i.e., phonology), word meanings (i.e., vocabulary), language structure (i.e., morphosyntax), and the rhythmic properties (i.e., prosody) that are characteristic to the spoken language, form the basis for reading development and continue to affect reading comprehension throughout the elementary years (Petscher et al., 2018). This section describes different frameworks that highlight how oral language supports the development of reading.

The simple view of reading (Hoover & Gough, 1990) highlights two interconnected components important for reading comprehension: word reading and language comprehension. Word reading includes processes important for decoding including phonology, morphology, and orthographic processing. Language comprehension includes skills involved in successfully understanding and using language, whether written or oral. Across reading development, the role of word recognition and language comprehension changes. There is a greater emphasis on decoding in the early grades in the “learning-to-read” stages of reading. Fluent and accurate word recognition supports overall reading comprehension because children who decode letter by letter when reading have fewer working memory resources allocated to understand the meaning of a text. In the later stages of reading or “reading to learn,” language comprehension plays a larger role as word reading becomes increasingly automatized (Catts et al., 2005).

Both word reading and language comprehension should be developed from the earliest stages of language and literacy development. Directly supporting language development early in school maximizes the likelihood that children will successfully transition from “learning to read” to “reading to learn” and develop strong reading comprehension skills. In fact, early language development should also support young children’s developing decoding skills, as theory and evidence indicate that increases in vocabulary knowledge directly contribute to the development of phonological and phonemic awareness (Walley et al., 2003), which are critical for the acquisition of decoding skills. As students become more experienced with decoding, reading skills have a reciprocal impact on the development of language skills (Nation, 2017; Ricketts et al., 2020). Moreover, other models, such as Scarborough’s (2001) reading rope, highlight the role of language skills, represented as individual strands woven together, in contributing to the outcome of skilled reading. Multilingual learners have varied language and literacy skills that may contribute to reading comprehension (Peets et al., 2022). It is important to consider their level of proficiency in word reading and language comprehension in all languages to more accurately describe their strengths and weaknesses in oral language and literacy. Moreover, assessment of both oral language and reading-related skills in more than one language, where possible, provides a more in-depth account of how oral language skills can be supported in service of literacy development.

The extent to which educational stakeholders consider the role of oral language skills in literacy development may vary, affecting the course of intervention for a student. For example, SLPs have expertise in language development and disorders with training to assess and monitor language skills comprehensively (e.g., phonology, morphology, vocabulary, syntax, pragmatics). In contrast, special education teachers and general classroom teachers may place a greater emphasis on assessing children’s literacy skills (e.g., reading fluency, spelling, comprehension) over time. Differences in how language is conceptualized can have a cascading effect on the skills chosen to focus on in literacy instruction, assessment, and intervention. In a school-based setting, we can leverage the expertise of special education teachers and SLPs to understand how oral language and literacy skills develop together. Effective collaborations between teachers and SLPs include building a common understanding about how both teachers and SLPs play a role in children’s literacy development and the importance of developing oral language skills for multilingual learners. This may take the form of consultation on assessments of language and literacy (as well as consultations with school psychologists who have expertise in diagnostic assessment) to meet multilingual student needs in the classroom environment, monitor oral language skill development, and identify co-planning and collaborative teaching opportunities that serve the language needs of all students (Archibald, 2017; Kangas, 2018).

Lesson 3: Use Caution When Interpreting Standardized Assessments

When assessing language and literacy skills to screen multilingual students for learning disabilities, standardized assessments play an important role. Standardized assessment scores are derived from norm-referenced assessments and allow direct comparisons between one student’s performance and their same-age peers who were included in the standardization sample used to create the test. For example, many commonly used tests of language proficiency and academic achievement include standard scores for which the mean is 100 and a standard deviation is 15. Therefore, a child who receives a standard score of 100 is at the 50th percentile of performance for their age group (i.e., at the mean), and a child who receives a standard score of 85 is at the 16th percentile of performance for their age group (i.e., one standard deviation below the mean). When interpreting standard scores, cutoffs are often used to determine whether follow-up evaluation or intervention is needed. For example, schools may screen children for learning disabilities in reading at the beginning of each school year, and any child scoring below the 16th percentile on screening assessments may be given targeted reading instruction in a response-to-intervention framework (e.g., Tier 2 intervention).

When attempting to interpret standard scores for multilingual students, it is important to consider who the test was developed for (i.e., information about the composition of the standardization sample can often be found in the examiner’s and/or technical manuals). For example, English language assessments of academic skills developed in the United States are typically created with monolingual English-speaking children as the target population. If multilingual children are not included in the development process, or if only a small fraction of multilingual students are included, then standard scores and corresponding percentile ranks may not be applicable to multilingual students. A multilingual kindergarten student might score at the 15th percentile of English listening comprehension, when compared with monolingual English-speaking students; however, multilingual language development and exposure differs from monolingual development, making cross-group comparisons inappropriate. When applying cutoffs, it may be determined that the 15th percentile is cause for concern for monolingual English-speaking children and there may be a referral for further evaluation for a language disorder, but it may not be a concern for multilingual children. Perhaps only 25% of language input is in English for a multilingual student. In this case, it should be expected that English listening comprehension will be lower for this student with 25% language input in English than it would be for others who are only exposed to English (i.e., 100% of language input is English). This highlights the critical need to compare multilingual students with their true peer group and consider their exposure to English language and reading instruction when screening for learning disabilities in language and reading.

Just as it is inappropriate to use assessments developed for monolingual English speakers with multilingual children, it is inappropriate to apply standard scores from assessments in other languages if they were not specifically developed for use with multilingual children. Often, even when assessments in non-English languages are available, they are not normed on multilingual children in the United States. For example, the Test de Vocabulario en Imágenes Peabody (Dunn et al., 1986) was developed using a sample of monolingual Spanish speakers in Mexico and Puerto Rico. Applying standard scores from this assessment to Spanish–English bilingual children in the United States could result in potentially inaccurate conclusions about children’s overall language ability. A language or learning disability should manifest in both languages (e.g., a Spanish–English bilingual child cannot have a learning disability in English but not Spanish). However, low performance in both languages alone is not sufficient for identification of a disability. Using standard scores separately from two assessments designed for monolingual students may reveal poor performance in each language that is due to input being distributed across languages rather than presence of a disorder. When using monolingual norms, practitioners should apply more conservative cutoffs (e.g., 5th vs. 15th percentile) than what would be used for monolingual students. However, this suggestion should be interpreted with caution rather than as a strict rule, as there is no evidence to date to suggest a specific rule that can be applied consistently across populations of multilingual children when using assessments developed for monolingual students.

Overinterpretation of standardized scores from measures not developed for use with multilingual students directly contributes to the problem of overidentification. However, use of English assessments of academic skill may be unavoidable in certain situations. Special education teachers and SLPs are uniquely positioned to tackle this issue to determine whether and how to use standard scores for a given assessment, while also consulting the specific diagnostic expertise of school psychologists. For example, teachers and SLPs could work together with existing data from multilingual students to determine what represents typical performance on an assessment for multilingual students in their local educational context. Such determinations could consider several factors, such as the language(s) spoken by the student, the relative amounts of input in each language, how long the student has been exposed to English, and whether the student has received formal academic skills instruction (and in which language). Scores that represent typical performance will differ across multilingual children (e.g., a Spanish-speaking child born in the United States vs. a child who is a refugee and immigrated to the United States in early elementary school). Castilla-Earls et al. (2020) have proposed a converging evidence framework for identification of disability among multilingual children that includes consideration of language experience, language samples, learning potential, and norm-referenced assessment; several of these options are described in more detail in the following sections. A comprehensive approach to considering multilingual students’ unique linguistic and educational backgrounds will help practitioners identify students’ true peer groups and understand what represents typical performance for subgroups of multilingual students.

Lesson 4: Assessment Through Adapted Scoring

Just as it is important to use caution when interpreting standardized assessment scores, alternate frameworks for scoring assessments may be needed to characterize multilingual children’s ability accurately. Because language input is distributed across more than one language for multilingual students, they often perform more poorly than monolingual students when language skills are only assessed in one language (Hoff et al., 2012). Although there is an increasing number of published language assessments that have been developed specifically for use with multilingual students (e.g., Bilingual English–Spanish Assessment [Peña et al., 2018]), they are only appropriate for certain age groups (e.g., preschool and kindergarten) and often only relevant for Spanish–English bilingual children. Many practitioners work with children of all ages who are from diverse language backgrounds. Consequently, understanding linguistically sensitive assessment frameworks is important for improving assessment practices used with multilingual children. Recent advances in the field of communication science have yielded alternate assessment and scoring frameworks that represent more holistic approaches to language assessment for multilingual children.

One approach is to use the best score from assessments in different languages. Research suggests that using combinations of best scores (e.g., a best score for vocabulary knowledge and a best score for grammatical knowledge) yields more accurate classification of language impairment among multilingual children (Lugo-Neris et al., 2015). For example, when screening for learning disabilities in reading for a Spanish–English bilingual student, practitioners could measure decoding in Spanish and English, and listening comprehension in Spanish and English, and obtain the standard scores from the assessment administration manuals. For example, a first-grade multilingual child in English-only instructional contexts might score low on Spanish decoding (e.g., 5th percentile) but in the average range on English decoding (e.g., 45th percentile). However, due to longer sustained exposure to Spanish prior to school entry, that same child might score in the average range for Spanish listening comprehension (e.g., 55th percentile) and the low range for English listening comprehension (e.g., 16th percentile). Then, the best score for decoding (45th percentile) and the best score for listening comprehension (55th percentile) can be considered together to provide an estimate of the child’s potential for acquiring language and literacy skills when provided with adequate opportunity to do so. In this example, the child’s best scores in both domains (i.e., decoding and listening comprehension) are near the 50th percentile, suggesting that the child likely does not have an underlying language or learning disability, despite poor performance in each domain within a specific language. This could lead to more accurate assessment than just assessing in a child’s primary or dominant language. Best scores should be used in the context of a converging evidence framework (Castilla-Earls et al., 2020), as described previously.

Second, conceptual scoring is an approach to language assessment that attempts to document the total number of concepts multilingual children know, regardless of language (Bedore et al., 2005). For example, in a total language approach, a Spanish–English bilingual child who knows the words apple and manzana (i.e., Spanish for apple) receives credit for knowing both of these words. In contrast, in a conceptual scoring framework, a child receives credit once for having language to describe apple as a concept. In this scenario, three different children, one who knows apple but not manzana, one who knows manzana but not apple, and one who knows both apple and manzana all receive equal credit. To administer assessments designed to measure conceptual language ability, examiners should determine the child’s preferred language, and begin assessment in that language. If a child answers a question incorrectly in their preferred language, they are provided an opportunity to give an answer in another language. It is important that the examiner is proficient in the child’s two languages (or an interpreter is used) to determine whether responses are valid. Research indicates that the use of conceptual scores more accurately assesses language ability than using monolingual assessment approaches (Anaya et al., 2018). Several recent multilingual assessments have become available and are designed as conceptual language assessments (e.g., Expressive One Word Picture Vocabulary Test–Spanish Bilingual Edition; Martin & Brownell, 2011).

Lesson 5: Incorporate Responsive Approaches for Assessment

Static knowledge-based assessment strategies, such as traditional vocabulary testing or IQ testing, rely heavily on task familiarity and prior exposure to a specific language. These approaches place multilingual children, whose language exposure is distributed across more than one language, at a disadvantage even before they begin the assessment, correspondingly increasing the risk that their abilities will be underestimated (Buac et al., 2016). In addition to the adaptive scoring approaches detailed in Lesson 4, several specific assessment strategies have emerged as more valid and reliable approaches for evaluating the language and learning abilities of multilingual children. Unlike more traditional assessments that focus on students’ current knowledge, these more culturally and linguistically responsive strategies are generally designed to assess students’ inherent ability to learn and use language.

Dynamic assessment is precisely designed to evaluate students’ ability to learn new information, and it is based on what is referred to as “test-teach-retest” (see Peña et al., 2006, for further description). This approach can be added to any assessment to evaluate students’ ability to acquire and retain new information. For example, during administration of a standardized spelling test, imagine that a child spelled the word “gate” incorrectly. After the initial test phase, the administrator could incorporate dynamic assessment by returning to the word “gate” to teach. The administrator might teach by saying “let’s spell ‘gate’ together. Let’s sound it out: /g/—/ay/—/t/. What letter makes the /g/ sound? G makes the /g/ sound! Then what letters can make the /ay/ sound?” and so on. After the teaching phase (which can be scaled to provide more or less explicit scaffolding support to the child), the administrator could then retest the child by asking them to spell “gate” (or other words that use the long a with silent e pattern) independently. A child who responds correctly with minimal support during the teach phase is unlikely to have a learning disability, whereas a child who responds incorrectly even after several teaching phases may have an underlying disability.

From a conceptual standpoint, dynamic assessment may be considered a shorter-term, more immediate form of response to intervention (RTI; also see multi-tiered system of supports or MTSS). At the most basic level, RTI is a structured approach to supporting students in schools that relies on regular progress monitoring and differential support for students based on their responses to instruction (Fuchs & Fuchs, 2006; Green et al., 2018). Students who demonstrate educational progress that is below expectations receive supplementary support scaled to their needs. Although not used in all U.S. schools, some schools leverage the information obtained from RTI progress monitoring to identify students with learning disabilities. This identification approach is designed to provide intervention and monitor student learning prior to referral for special education evaluation, reducing the wait-to-fail time in which struggling students do not receive supplemental intervention until they are eligible for special education services. Given the importance of early education and intervention to long-term outcomes, under-referral can be problematic when it causes delays in identification of students with learning disabilities. Evidence suggests under-referral is a substantial concern among multilingual learners in the early elementary grades, in particular (Samson & Lesaux, 2009). Like RTI, dynamic assessment directly addresses this wait-to-fail concern, as it allows for assessment of immediate acquisition and retention of new information.

There are currently few standardized approaches to dynamic assessment (see the CUBED by Petersen & Spencer, 2016; also see Peña, 2021). This can make interpretation of students’ responses to dynamic assessment difficult, particularly for educators new to using dynamic assessment. It can be unclear what levels of support are within the normal range for eliciting a correct response from a child. In addition, stigmatization and cultural mismatches may result in multilingual children feeling hesitant to respond immediately to repeated prompting. Strong rapport between the educator and child may be necessary to establish before an accurate measurement of the child’s skills can be obtained. Consequently, dynamic assessment may be particularly helpful in contributing to screening for language and learning disability among multilingual learners. Children who do not respond correctly to retest prompts after one or more teaching phases would be good candidates for more comprehensive evaluation by a full assessment team, which should include an SLP with expertise in multilingualism. See Gutiérrez-Clellen and Peña (2001) for a detailed tutorial on dynamic assessment.

An additional assessment strategy that has strong evidence as a less biased approach to multilingual language assessment is narrative language sampling. Language sampling broadly is an evaluation technique based on observations of an individual’s communication skills in conversation, play, or storytelling. This can be an ecologically valid and culturally responsive approach for measuring a multilingual child’s functional communication skills within a meaningful context. Unlike traditional standardized testing, language sampling occurs within a realistic communication context and correspondingly can provide a more practical view of a child’s ability to communicate meaningfully. Narrative language sampling allows educators to assess various domains of language (e.g., vocabulary, grammar) simultaneously, and how the child leverages different skills to communicate socially. Importantly, basic storytelling structure (i.e., narrative macrostructure) generally extends across languages, enabling multilingual children to draw on skills developed in both their home language(s) and school language(s) to tell a cohesive story to their communication partner. Narrative language sampling can reveal both communication strengths and weaknesses that may be missed in traditional domain-specific standardized assessment. See Castilla-Earls et al. (2020) and Rojas and Iglesias (2009) and other resources that follow for additional guidance on narrative language sampling.

Recommendations and Conclusions

Children with low language ability (e.g., language disorder or impairment) are up to 6 times more likely to have a learning disability in reading (Komesidou & Hogan, n.d.) than are children with typical language development. Consequently, it is imperative that language skills are considered when screening for learning disabilities among multilingual children. Special and general educators can collaborate with SLPs and school psychologists to implement evidence-based assessment practices to screen multilingual children for learning disabilities. In particular, collaboration can ensure that

Special educators, SLPs, and school psychologists each bring unique expertise and experiences to the assessment process. Increasing collaborative efforts in alignment with evidence-based assessment practices will help reduce misidentification of multilingual children with language and learning disabilities, and ultimately better inform instruction to maximize student achievement.

Infographic Summary of Recommendations and Other Resources

A summary of the recommendations included in this article has been made publicly available in an infographic published on Open Science Framework (Goodrich et al., 2021). Readers may access the graphic through this direct link: https://doi.org/10.17605/OSF.IO/8XYNF. Other resources (including video instructions) for collecting language samples and conducting dynamic assessment are provided below: