This study analyzes the use of role-play as an elicitation device for the evaluation of a nonnative speaker's oral language. In this analysis of role-play as a methodology to generate data for assessment purposes, the study examines the role that interlocutors play in two types of interactions. It raises questions about the validity and reliability of dialogic role-plays used to elicit a ratable sample of a nonnative speaker's oral proficiency and proposes that a monologic situation may provide a more reliable and accurate sample. The study also suggests how implications for simulation-based teaching can be implemented in second/foreign language classes. (Contains 3 figures and 5 tables.)