Abstract
The purpose of this research is to select test items from 90-item Vocabulary Size Test (Mochizuki, 1998) applying two testing theories: Classic testing theory and the Rasch model, in order to classify students according to their English language proficiency. The data of 3,286 university students in Japan were analyzed using two theories. The 45 items were selected based on results from well-known classic testing analyses (Short) and 39 items were chosen from Rash Model analysis (RM). Items were found to have overlapped on 24 occasions. The both sets of items, Short and RM have high reliability, of Chronbach's alpha: .931 and .924 respectively.
