Resumen:
Language learners' vocabulary size is a reliable predictor of their success in a second language as it clearly correlates with better performances in the target language (Nation, 2001). Precise estimations of language learners' actual knowledge are paramount to plan language teaching. However, the instruments employed by previous studies for those estimations might present validity and reliability issues that affect their research sensitivity and accuracy. This paper presents a step-by-step account of the creation of an aural and a written version of a bilingual vocabulary test. Answers from 73 adult L1-Spanish students attending English classes were analysed with the Rasch model to determine the best performing items in the test so that the overall reliability of the instrument was enhanced. The final version presents high levels of reliability: .89 for the listening vocabulary test and .82 for the written vocabulary test. Furthermore, descriptive statistics confirm that recognizing the words in their aural form is more challenging than in their written form: participants obtained 10.80% fewer correct answers in the listening vocabulary test. This finding confirms the claim that aural and written vocabulary are two separate dimensions, and impacts on how vocabulary should be taught in L2 classrooms.
