The use of fixed SSTs is clearly sub-optimal in both the global and regional systems. Biases in monthly mean fields are reduced and simulations of the MJO improved by replacing the fixed SSTs with observed SSTs. Further improvements in the simulations are found when the atmospheric models are fully coupled to interactive ocean models. In the coupled global system, substantial improvements in the simulation of the MJO and equatorial waves are obtained through modifications to the physical parameterizations. The biggest impact is obtained through changes to the treatment of parameterized deep convection, but improvements are also obtained through the implementation of the COARE 3.0 air-sea flux parameterization in NAVGEM, following the treatment in the Hybrid Ocean Coordinate Model (HYCOM) ocean component. In COAMPS, increased resolution reduces biases and results in more realistic precipitation structures and ocean evolution. Moist static energy analysis illustrates the importance of moistening due to vertical advection before the active phase of the MJO, and enhanced drying due to horizontal advection during the active phase. Not surprisingly, COAMPS forced by analyzed boundary conditions are superior to those forced by NAVGEM forecasts. However, COAMPS can still predict the late-November 2011 MJO at 20-day lead times even if the MJO is not predicted in NAVGEM. This suggests that the mechanisms allowing for the prediction of the late-November MJO are localized to the Indian Ocean region. The mechanisms allowing for this extended-range MJO predictability will be further examined.