Several recent studies have investigated the nature of interaction in oral proficiency exams and have concluded that the interview format obscures differences in the conversational competence of the candidates. The present paper examines what opportunities test takers have to display their knowledge of managing conversations in the L2 in two types of tasks: non-scripted interviews and guided role-play activities. The data for the study consists of 30 interviews and 30 role-play activities between near-native examiners and intermediate learners used in language exams in Hungary. The interviews and role-plays have been analysed for the number of topics introduced and ratified by the examiner and the candidate respectively, as well as for the number of interruptions, openings and closings produced by the examiner and the candidate. The findings show that the conversational interaction is more symmetrical in the guided role-play activity with the candidates introducing and ratifying approximately the same number of topics as the examiners. In addition, the examinees have the opportunity to interrupt and hold the floor more effectively during the role-play activity and can demonstrate their knowledge of how to open and close a conversation. These findings suggest that guided role-play activities used in the study exhibit several characteristics of real-life conversations and therefore can be used for assessing the candidates’ conversational competence.