A Latent Class IRT Approach to Defining and Measuring Language Proficiency.

Area of Science:

• Educational Measurement
• Psychometrics
• Applied Linguistics

Background:

• English language learner (EL) status has significant implications for academic assessment and evaluation.
• Current English language proficiency (ELP) conceptualizations increasingly align with content area language demands.
• State standards and consortia like WIDA and ELPA21 guide ELP assessment.

Purpose of the Study:

• To explore the utility of content area assessments in validating English language proficiency (ELP) mastery criteria.
• To identify distinct groups of English learners (ELs) based on their performance on academic achievement tests.
• To evaluate the predictive power of different ELP subscales on group membership.

Main Methods:

• Mixture item response models (IRT) were applied to identify classes of EL students.
• Latent class IRT methods were used to group ELs based on item functioning.
• The influence of reading, writing, listening, and speaking subscales of ELP on class membership was assessed.

Main Results:

• Two classes of EL students were identified: those for whom test items functioned similarly to non-ELs, and those for whom items functioned differently.
• Reading and writing subscales of ELP were significant predictors of class membership.
• ELP summary score cut-scores imperfectly predicted class membership, highlighting a need for nuanced differentiation.

Conclusions:

• Linking ELP definitions to specific contexts of use is crucial for accurate assessment.
• Psychometric analyses, particularly IRT, can enhance the validity of ELP standards by considering content area language requirements.
• This research suggests that content-based assessments can help validate ELP mastery criteria for English learners.