This report presents an abstract process model of 'well formed' discourse. The model is based on a hierarchical segmentation of utterances into distinct, but related and linked, context spaces. Context spaces are characterized by slots similar to case frames. The slots are functionally defined and they capture information both explicit and implicit in the dialogue. In a context space theory of discourse, a conversation is viewed as a sequence of conversational moves. A conversational move corresponds to a speaker's communicative goal. 'Support,' 'Interrupt,' and 'Challenge,' are some of the communicative goals identified in the work. All conversational moves have an associated set of preconditions which specify the requisite discourse context for their appropriate performance. Conversational moves also have a set of effects on the discourse structure, and different modes of fulfillment. A major aspect of the work is to delineate the different preconditions, effects, and modes, characterizing different types of conversational moves in a discourse. It is shown that effective communication between conversants is enabled by their following a number of well-specified rules in a discourse. These rules include choice of reference to an entity under discussion and use of clue words like 'But anyway,' and 'Incidentally.'