A Rasch-based Validation of The Word Associates Test

Abstract

The purpose of this study is to validate Word Associates Test (WAT) designed to measure depth of lexical knowledge with Rasch Analysis. This study focuses on how the format which contains multiple correct items impacts the reliability and validity of the WAT. For this purpose,two versions of the test, varied (the number of correct items in each box varies) and the traditional four multiple-choice format were created. The results indicated that the traditional multiple-choice version yield better results than those of varied version in terms of good reliability, sufficient item fit, positive values of point measure correlations, and fundamentally unidimensionality of the measured construct. On the other hand, the varied version of WAT did not reliably distinguish between high and low ability of test takers' lexical knowledge. Furthermore, the varied version implied the high possibility of multidimensionality. That is, the WAT format is likely measuring two different constructs, polysemy and collocation.