This research analyses utterances produced by English speakers who interact via computers in chat rooms, chat.yahoo.com. This study is aimed at investigating how the participants interact and develop topics under computer-mediated constraints.The data consist of corpus containing sequenced-postings produced by multi-participants in Lobby, part of the chat room regarded as a public domain. Conversation Analysis Model introduced by Birmingham School was used for data analysis.The findings of the study show that the chat has some unique features, which are different from the features of interaction in face-to-face conversation. Those unique features decrease the coherence of the interaction. The linearity of the sequence is disrupted by the presence of initiation without responses and the presence of multi-responses for one initiation. Then, the unity of the interaction is loosened by the absence of organizational exchanges or the presence of less discourse markers tying up explicitly the interaction units in each rank of the interaction. After that, the immediacy relationship between elements of adjacency pairs is loosened by the presence of other utterances between them. Finally, the structure of the utterance is loosened by the fragmentation of the utterance into smaller units and the positioning of those units in the different postings.The chat has several unique features of interaction strategies that may decrease the coherence of the interaction and differ them from those in face-to-face interaction. First, the one person speak in one time rule is not applicable in the chat since who is currently speaking cannot be identified. Second, the competitive nature of the interaction requires an adjustment of the strategy for managing the turn. It is not necessary for the chatters to hold the turn; instead they have to send repetitive postings to maintain their presence. Finally, turn-taking strategies are very limited in the chat interaction. The strategy for taking the turn is limited to clean strategy and the strategy for yielding the turn is limited to the use of selection (turn design) and give-up strategies.The topic introduction and closing in the chat is not so different from those in the face-to-face interaction. The topic development displays unique structure, in that the chat may produce parallel topics. The parallel topic is resulted by the introduction of a new sub-topic while the discussion of the existing topic still in progress. Another specification of topic development is the rapid progress of the topic change in the chat.The uniqueness of the interaction and topic development in the chat is attributable to the lack of physical presence and voice as well as to the machine related constraints.(For further information about this dissertation, contact: hamzahhs@yahoo.com)