We formulate a model for intermittent communication that can capture bursty transmissions or a sporadically available channel, where in either case the receiver does not know a priori when the transmissions will occur. For the point-to-point case, we develop two decoding structures and their achievable rates for such communication scenarios. One structure determines the transmitted codeword through exhaustive search, and the other structure rst detects the locations of codeword symbols and then uses them to decode. We introduce the concept of partial divergence and study some of its properties in order to obtain stronger achievability results. As the system becomes more intermittent, the achievable rates decrease due to the additional uncertainty about the positions of the codeword symbols at the decoder. Additionally, we provide upper bounds on the capacity of binary noiseless intermittent communication with the help of a genie-aided encoder and decoder. The upper bounds imply a tradeo between the capacity and the intermittency rate of the communication system, even if the receive window scales linearly with the codeword length. Upon this foundation, we develop two extensions. First, we extend the model to intermittent multi-access communication for two users that captures the bursty transmission of the codeword symbols for each user and the possible asynchronism between the receiver and the transmitters as well as between the transmitters themselves. This model can be viewed as another attempt to combine information-theoretic and network-oriented multi-access models. We characterize the performance of the system in terms of achievable rate regions. In our achievable schemes, the intermittency of the system comes with a signi cant cost. Second, we extend the model to packet-level intermittent communication in which codeword and noise symbols are grouped into packets. Depending on the scaling behavior of the packet length relative to the codeword length, we identify some interesting scenarios, and characterize the performance of the system in terms of the achievable rates for each model. Finally, we apply the insights and tools developed for intermittent communication to several related problems. First, we obtain new results on the capacity of deletion channels and a random access model that drops the collided symbols. Second, we study the problem of lossless source coding in the presence of intermittent sideinformation.