This study investigated the appropriateness of Item Response Theory (IRT) models to analyze language test data. In a university second language institute, results of typical versions of tests of French as a Second Language (n=831 examinees) and of English as a Second Language (n=617 examinees) were subjected to non-linear factor analysis in dimensionality assessment, as a first step in selecting an appropriate IRT model for data sets from students with differing language learning backgrounds. The tests, similar in design and purpose, each included a reading comprehension, listening comprehension, and cloze portion. Examinees were native speakers of one of the two languages tested, but were heterogeneous with respect to their language backgrounds (immersion programs, regular high school programs bilingual families, frequent contact with native speakers). No systematic relationship was found between item performance and strategies hypothesized to characterize each of the two examinee groups, but it is suggested that certain response patterns in the cloze portion of the test merit further investigation. (Author/MSE)