Probing the Internal Validity of the LLAMA Language Aptitude Tests

Abstract

Over the past decade, the LLAMA language aptitude test battery has come to play an increasingly important role as an instrument in research on individual differences in language development. However, a potentially serious problem that has been pointed out by several scholars is that the LLAMA has not yet been carefully validated. We addressed this issue by examining the internal validity of this test battery. We collected LLAMA data from 350 participants and assessed these data using classical item analysis, Rasch analysis, and principal component analysis within a framework of best practices in educational and psychological test validation. The results show that only one out of the four subtests (LLAMA B) produced scores that fit a latent trait model with sufficient accuracy. This suggests that researchers using the LLAMA battery must treat their results with appropriate carefulness and also that there is potential for refining the LLAMA further.