Abstract. This paper offers a formalization of iterated belief revision for multiagent communication using the logic of communication graphs introduced in [15]. In this study we consider an agent (i.e., information source) capable of sending two types of message. In the first type, he tells that he knows a proposition, but in the second type, he tells that he believes a proposition. Consequently, iterated belief revision is brought about through a sequence of communication events (i.e., a history), and we propose a variation of the AGM rational postulates for history based belief revision. As we will show, a representation theorem is verified only for a class of restricted communication graphs. We consider this result to be a weak point in the application of the AGM postulates to multiagent communication, and propose a viable alternative.