Representing agent conversations is an important aspect of multi-agent applications. Lately, Petri nets have been found to provide certain advantages comparing to other representation approaches. Radically different approaches using Petri nets have been proposed, and yet their relative strengths and weaknesses have not been examined. This paper analyzes the existing Petri net representations in terms of their scalability and appropriateness for monitoring tasks. Based on the insights gained, we propose a scalable Petri net representation for overhearing.