Validation of Hortwitz’s (1987) beliefs about language learning inventory with Chinese mainland sample: Traditional factor analysis versus Rasch modeling

This study aimed at establishing, within the classical test theory and Rasch framework, respectively, construct validity of the translated version of Horwitz’s (1987) beliefs about language learning inventory to be used with Chinese mainland college ESL students. Based on a convenience sample of 362 tertiary-level college students, from the original 34-item conceptually-composed 5-factor inventory the exploratory and subsequent confirmatory factor analyses generated a 15-item four-factor solution. Of the four factors, only two demonstrated marginally acceptable internal consistency. Alternatively the Rasch scaling method was utilized in a parallel manner. Psychometric properties, e.g., the item and model fit, category functioning of the items, and dimensionality of the inventory, were empirically put under close scrutiny. After the removal of 10 items with poor item fit or persistently disordered categories, it ended up with a refined inventory, of which no absolute evidence of unidimensionality had been captured. The purification exercise continued with identification of items that displayed differential item functioning (DIF) values till the remaining 21 items, as evidenced, measured a single underlying construct. Consequently deprived from the Rasch measurement model was a short version of the inventory with clinically appropriate length and proved construct validity, though the present study did not attempt to tell which approach was superior to the other. It’s hoped that the robust validation techniques would optimize the final factor solution, thereby ensuring the legitimacy of the inventory to be used for the investigation on the patterns of Chinese Mainland college students’ beliefs about learning English as a second language.