A recent trend in research of Computer-Assisted Language Learning (CALL) is to incorporate established experimental methodology as well as new methods in learning sciences to produce demonstrable improvements in second language (L2) learning. This new method is referred to as eCALL (experimentalized CALL). This talk introduces two empirical experiments that examined the effectiveness of two versions of an eCALL tutor that provided training on English article usage for L2 learners in China. The Article Tutor was designed based on the Competition Model, a psycholinguistic model for language acquisition. The tutor incorporated a Bayesian knowledge tracing model to achieve individualized student tracking and adaptive learning. Data were analyzed through behavioral measurements and methods of educational data mining. The studies generated significant theoretical and pedagogical implications in second language learning.