A recent research article by Gillam, Peña, Bedore, Bohman and Mendez-Perez (2013) examined the accuracy of the current EpiSLI testing model for bilingual children with specific language impairment (SLI). Additionally, it explored whether the current EpiSLI model could be modified to give better results in the diagnosis of SLI in bilingual children.
Gillam et al. (2013) recruited and screened 1,192 kindergarteners identified as English Language Learners (ELLs), using a bilingual screener to evaluate each child's grasp of both English and Spanish. A follow up study was conducted when the children reached first grade. At this point, they had experienced a year of schooling, and received instruction in English at least 30% of the time. Overuse of commas. I've marked them in red (as well as some other questionable punctuation usage) throughout. HOWEVER, I'm no English major so take my opinion with a grain of salt.
Each child had to meet certain criteria to be eligible for the study. They had to be identified by their primary caregiver as Hispanic; have at least 20% conversational exposure to both English and Spanish; and score below the 30th percentile in one of the subtests given in the bilingual screener. Ultimately, 167 children met all criteria and were eligible for the study.
Gillam et al. (2013) tested the children using the original EpiSLI assessment model, and compared the results to evaluations of the same children by a panel of bilingual Speech Language Pathologists (SLPs). *To answer their second research question, they tried to enhance the test’s reliability by setting cut-off scores for the different subtests of the EpiSLI, based on statistical analysis of the test results. Their objective was to find reliable cutoff scores on particular subtests or combinations of subtests correlating with the presence or absence of SLI, as determined by the conclusions of the SLP panel. and gather the best possible outcome for the identification and differentiation of children with an impairment or children who just had a language difference.
Gillam et al. (2013) found that both the original EpiSLI model, and the version modified for the study, over-identified SLI among bilingual children. In a significant number of the children studied, both models failed to differentiate between impairments and mere language differences. The large number of false-positive test results indicated that another measure of SLI among bilingual children should be developed. Since both the original and modified models were found to be problematic in the identification of SLI, a third, “optimal” model using analytical statistics was created. The optimal model was found to be superior to the revised EpiSLI model in diagnosing SLI, but no more effective than the revised EpiSLI model at ruling out SLI.
The researchers made the distinction that the unmodified EpiSLI model is ambiguous when discerning SLI in bilingual children. The children assessed by the SLPs as having...