In light of the development of a computational model which is able to produce and interpret speech acts appropriately, the authors discuss problems involved in speech act generation, speech act representation, and speech act recognition.