Parallel forms reliability of two versions of the Spanish Elicited Imitation Task

Abstract

Assessing proficiency in a standardized way is fundamental to second language acquisition research methodology and theory building. Ortega and colleagues’ (2002) Elicited Imitation Task (EIT), developed to measure global oral proficiency across various languages, has been supported by a body of methodological research that provides evidence that the task yields scores that are internally reliable and externally valid. However, where only one EIT has been validated for a given language, as in Spanish, the EIT's usefulness for longitudinal or repeated-measures research is limited due to possible test effects. The present study tests the parallel forms reliability of a minimally modified version of the Ortega et al. (2002) Spanish EIT (version A) and a second version (version B) designed to provide a parallel option for longitudinal research. Both versions were administered to participants (n = 96), from whom self-ratings of proficiency were also collected. Parallel forms reliability was supported by a wide variety of metrics, including comparability between the versions in terms of overall scores, internal reliability, item-level statistics, and correlations with an external measure (self-ratings of proficiency). These results provide evidence for the two versions’ parallel forms equivalence, which in our view supports their utility for longitudinal research.