This article deals with the challenges of data collection from a Conversation Analysis (CA) perspective to researching synchronous, audio-visual eTandems. Conversation analysis is a research tradition that developed out of ethnomethodology and is concerned with the question of how social interaction in naturally occurring situations is organized. In the course of the first cycle of data collection for my PhD research, which was carried out within the L3 TASK project, four methodological issues ("multimodality", "completeness", "authenticity" and "task-based vs. off-task conversation") that result from the particular requirements of CA to data collection were identified. In the following article these challenges and possible solutions are brought to light. [For the complete volume, see ED571330.]