The spatial component has so far been neglected in studies of cholera dynamics for two main reasons: first, the previous lack of high-resolution spatio-temporal data, and second, the survival of the bacterium V. Cholerae outside the human host in aquatic reservoirs, which advocates the view that the infective agent is present almost everywhere in the typical estuarine landscapes of endemic cholera. The extensive spatio-temporal data set for the individual location of every cholera case in the rural area of Matlab, Bangladesh, for the period 1983-2004 allows us to examine the spatial structure of cholera outbreaks at high resolution for numerous outbreaks. Ripley's K indices and bootstrapping techniques were used to evaluate the occurrence of spatial clustering (aggregation and/or repulsion) of cases during 25 outbreaks of cholera using different temporal windows for the aggregation of the cases. The spatial clustering of cases was also confronted against the spatial location of water sources. The spatial structure was analyzed for both primary transmission (from an aquatic reservoir to a human host), and secondary transmission (involving a feedback between current and past levels of infection). Results show evidence for the spatial clustering of cholera cases at different temporal and spatial scales. Furthermore, spatial clustering occurs also for cases relative to water sources, providing insights on the role of primary and secondary transmission during the onset of cholera outbreaks.