IRT Measurement and Validation of Vocabulary Proficiency

• In Event: Division D Graduate Student in Progress Research Gala

Abstract

Lexical proficiency plays a fundamental role in language learning and development. Traditionally, measurement of lexical proficiency constructs has relied on subjective human ratings, word frequency algorithms, and type-token ratios (TTRs). In my current research, I argue that Item Response Theory (IRT) ability estimates generated from response sets of dichotomously-scored word types on student essays potentially offer a more valid measure. In particular, I am analyzing the ability of IRT models to predict essay human ratings and the potential of Rasch measures to serve as validation tools for improving the accuracy and efficiency of individual raters.

Author

• Jonathan Downey, University of California - Santa Barbara