It is more than 40 years since J. L. Austin made the following statement, "The total speech-act in the total speech-situation is the only actual phenomenon which, in the last resort, we are engaged in elucidating (1962:148)". However, this simple but complete truth has been neglected so far by speech act theorists, including Austin himself.The Austinian trichotomy divides a "total speech act" into locution, illocution and perlocution, i.e., a speech act is not complete without any of the three components, while the theory of speech acts as well as pragmatic studies have been paying sole attention to illocutionary act, which is nearly synonymous to the speech act itself. Such incomplete study cannot represent the panoramic view of speech acts. The present author holds that the analysis of speech acts should be done in a unified model which embraces all the three components, esp. perlocution, for a speaker uses speech acts not only to express his/her meaning but also to influence the hearer so as to achieve some extra-linguistic goals.Although Austin has acknowledged that speech acts should be studied in the "total speech situation", the development of Speech Act Theory is leading towards, to some extent, a semantic analysis of speech act verbs. In order to examine the use of language in actual situations, again perlocution, which is the most sensitive to contextual factors, will be proved to be the most valuable.The extension of the study to the sphere beyond sentence has never occurred to the speech act theorists, whereas many scholars at home and abroad have realized the theory's potential application to discourse analysis and have made unremitting efforts. The current dissertation is a tentative study in the same direction, in which the writer redefines the notion of perlocution, puts forward an interactive speech actapproach and applies it in the study of discourse. An interactive analytical framework is established, whose explanatory power is displayed in its account for coherence, esp. sequential coherence, in conversation.The present dissertation is composed of five chapters, excluding the introduction at the outset and the concluding remarks at the end. Chapter 1 deals with the preliminary notions-models of communication as well as its nature as a goal-directed, dynamic, social interaction.Chapter 2 intends to review and critique the orthodox Speech Act Theory and its applications in discourse analysis. The author points out the major defects in the theory: negligence of perlocution and underestimation of context.Chapter 3 is the theoretical construction. On the basis of the orthodox theory, an interactive speech act approach is established, in which situated speech acts are considered as the object of study, the notion of perlocution is scrutinized, and various stages of interaction are specified.Chapter 4 discusses the application of the interactive speech act approach in discourse analysis. An analytical framework is put forward to specify the unit of analysis, context analysis which is essential in meaning negotiation and speaker-hearer interaction, and goal development which represents both parties' contribution to the attainment of the speaker's perlocutionary intention in both direct and indirect communication.Chapter 5 is an illustration of the explanatory power of the interactive speech act model in terms of sequencing phenomena in conversation. Our framework is in a better position to account for the regularities as well as the irregularities in conversation.