Publications & Resources
Impact of Student Language Background on Content-Based Performance: Analyses of Extant Data
Jamal Abedi, Seth Leon and Jim Mirocha
We analyzed existing test data and student background data from four different school sites nationwide to examine whether standardized test results may be confounded by the lack of language proficiency of English language learners. Several analyses comparing the performance of limited English proficient (LEP) students and their non-LEP classmates revealed major differences. A Disparity Index was created to measure the performance gap between LEP and non-LEP students on tests with varying levels of language demand. The more linguistically complex the nature of the test, the greater was the Disparity Index of non-LEP students’ results over LEP students’. This may suggest that high-language-load test items in assessments of content such as math and science may act as a source of measurement error. LEP students tended to have lower internal consistency scores on standardized assessments. Again, this suggests that item language load may interfere with testing the intended constructs. Using multiple regression, multivariate analysis of variance, and canonical correlation, we found that the more language load in a test, the stronger the confounding between LEP status and content-based performance on that test. Structural models for LEP student results demonstrated a lower statistical fit among test items, as well as between items and the total test scores. The factor loadings were generally lower for LEP students, and the correlations between the latent content-based variables were weaker as well.
Results of our analyses indicate that:
1) English language proficiency level is associated with performance on content-based assessments.
2) There is a performance gap in content assessment between LEP students and non-LEP students.
3) The performance gap between LEP students and non-LEP students increases as the language load of the assessment tools increases.
4) Test items high in language complexity may be sources of measurement error.
5) Performance on content-based assessments may be confounded with English language proficiency level.
Abedi, J., Leon, S., & Mirocha, J. (2003). Impact of student language background on content-based performance: Analyses of extant data (CSE Report 603). Los Angeles: University of California, Los Angeles, National Center for Research on Evaluation, Standards, and Student Testing (CRESST).