A 3-D cloud-resolving model has been used to investigate the domain size dependence of simulations of convective self-aggregation (CSA) in radiative-convective equilibrium. We investigate how large a domain is needed to allow multiple convective clusters and also how the properties equilibrated CSA depend on domain size. We used doubly periodic square domains of widths 768, 1,536, 3,072, and 6,144 km, over 350 simulated days. In the 768-, 1,536-, and 3,072-km domains, the simulations produced circular convective clusters surrounded by broader regions of dry, subsiding air. In the 6,144-km domain, the convection ultimately forms two semiconnected bands. As the domain size increases, equilibrated CSA moistens in two ways. First, as the circulation widens, this leads to stronger boundary layer winds and a more humid boundary layer. Second, the stronger inflow into the convective region boundary layer is associated with a warmer convective region boundary layer, which leads to intensified deep convection, more melting and freezing near the freezing level, enhanced midlevel stability, increased congestus activity, and detrainment of moist air into the dry region. In the larger domains, the deep convection and congestus slowly oscillate out of phase with each other with a time period of about 25 to 30 days. We hypothesize that other important domain size sensitivities, including a decrease in net moist static energy export from the convective region, are fundamentally linked to the increasing relationship between domain size and boundary layer wind speed. Our results suggest that the statistics of CSA converge only for domains wider than about 3,000 km.When rotation is included, CSA gives rise to tropical cyclones (Davis,2015;Khairoutdinov & Emanuel, 2013;Nolan et al., 2007;Wing et al., 2016Wing et al., , 2017 providing additional evidence that CSA is relevant to the real atmosphere. Furthermore, numerical simulations by Arnold and Randall (2015) and Khairoutdinov and Emanuel (2018) suggest that CSA on a rotating sphere can account for the MJO.The mechanisms that initiate and maintain CSA have been explored in detail in previous modeling studies of RCE. CSA is initiated through the action of positive feedbacks, whereby dry patches expand and confine the convection to a single portion of the domain (e.g., Emanuel et al., 2014). The positive feedback arises because dry areas have anomalous radiative cooling that must be balanced by adiabatic warming via subsidence, which produces further drying. As aggregation sets up, the circulation carries moist static energy (MSE) from