The French CAT

This paper investigates the empirical validity of the Monash-Melbourne computer adaptive test for French (French CAT), a single parameter Rasch model measurement of underlying morphosyntactic proficiency. It focuses, in particular, on the accuracy of the French CAT as a tool for streaming incoming university students into three levels of a first year (post high school) French course. Psychometric ability estimations of the Rasch model are compared against instructors’ assessment of students’ overall linguistic competence. A comparison is also made between the theoretical confidence interval of predicted abilities and the actual distribution of testee scores. Finally, individual student French CAT scores are correlated with end-of-semester language examination results. In all instances, Item Response Theory, upon which the French CAT is based, is shown to provide a highly valid means of determining linguistic ability for the purposes of course placement. Moreover, given the significant correlation between initial streaming and end-of-semester results, the French CAT is also demonstrated to be a good predictor of short-term achievement.