In this paper, a compositional model for reflective agents is proposed within which reasoning about observation, assumption making and communication, an agent's own information state and reasoning processes, other agents' information states and reasoning processes, and combinations of these types of reflective reasoning are explicitly modelled. The types of knowledge needed to detect, analyze and resolve conflicts that arise by meta-reasoning within the agent are discussed. The knowledge and interaction between agents required to model the wise men's puzzle is used to illustrate the approach. The model has been validated using think-aloud protocols. An implementation has been made including a speech synthesis facility.