When assessing students’ interactional competence, teachers and test makers are increasingly designing formats that more closely approximate the characteristics of mundane conversation, resulting in a shift away from oral proficiency interview (OPI) style tests between an assessor and the test-taker in favour of dyadic or group conversation tests between peers. Although previous studies have explored the differences between paired and OPI style tests, there is a need for further interactional research comparing different styles of paired conversation test to one another. Adopting a conversation analytic (CA) approach, this chapter analyses a dataset of over 200 EFL conversation tests video-recorded at a Japanese university. By providing grounded observations on the test-takers’ interaction, the findings suggest that slight variations in the paired test format have major implications on the way the discussions unfold. The analysis shows that a relatively minimalistic style of prompt containing only a general theme often leads to the sort of collaborative conversation reminiscent of natural talk. Conversely, complex prompts with explicitly stated goals (e.g., agreeing or disagreeing with a stated opinion) result in the solicitation of target pragmatic forms at the risk of a (less natural) parallel style of turn-taking/topic development. Further, the study suggests that the presence of multiple textual objects in the test setting can be an affordance or an impediment to progressivity of the talk. Test-takers at times indexically refer to the prompt cards as a means of topic management. However, the presence of multiple cards also becomes a focal point for student gaze and attention throughout the conversation. The findings suggest that a format with a single general prompt and fewer textual artefacts is more conducive to natural conversation than one containing explicit tasks. On the other hand, explicitly focused tasks can prove valuable when assessing test-takers’ ability to accomplish certain pragmatic actions.