A fundamental feature of effective distributed systems is that the entities comprising the system have some set of guidelines--some plan to follow--that leads them into making good decisions about what to communicate and when. Traditionally, these protocols for communication have been given to the entities at the time that they are designed. For example, knowledgebased entities (agents) have been designed with protocols that allow them to make deals, allocate tasks, negotiate over solutions, and so on. Such distributed systems, however, will be brittle if the agents ever need to go beyond the pre-existing protocol. To constitute a robust system, the agents would benefit from the ability to discover new ways of communicating, and to generalize these into new protocols. This paper extends the recursive modeling method to address issues of embedded communications-- communications occurring in a larger context of other physical and/or communicative activities, and describes how behaviors like question-answering and order-following could emerge as rational consequences of agents’ decisionmaking. These types of embedded communicative acts can form the building blocks of more complex protocols, given that agents can not only derive these embedded communicative acts but can generalize and reuse them appropriately.