This paper describes efforts to devise a more effective method of assessing oral proficiency in English using a computer-based testing and computer-based test result management system. This new test would be designed to directly operationalize the second language constructs it measured, and would be developed with reference to the best practices revealed by recent research in the field. Scores are automatically sent to databases at the completion and evaluation of each task. Student performance is then used to drive an adaptive algorithm, determining the difficulty of successive tasks. This computerized testing would aid in assigning skills or constructs to individual tasks, apportion the skills required for each task, and identify the features of interaction elicited in individual tasks. The study results revealed not only which tasks elicited most assessor/interlocutor interaction, but also that a great deal of the interaction took place outside the task boundaries. There has been a substantial move towards acknowledging the importance of interaction in oral proficiency and, therefore, oral proficiency testing featuring discourse and conversation analysis. The research detailed in this paper indicates that greater interaction does seem to be associated with greater oral proficiency, yet there are nagging issues of test validity still to be addressed in future research. (Contains 47 references.) (KFT)