The focus is on sensitivity studies to identify factors that increase the skill of the ECHAM4 atmosphere general circulation model (GCM) in representing the intraseasonal oscillation (ISO). The ISO mode is dominated by an eastward propagating oscillation during boreal winter, also known as the Madden-Julian oscillation. A previous study compared observations and reanalysis data to uncoupled and coupled versions of the ECHAM4 GCM. The sensitivity experiments in the present study use those results to assess the importance of the following mechanisms for the simulation of the ISO. First, the vertical resolution is increased to indicate the effect of improved representation of the tropospheric stratification. Second, the horizontal resolution is increased to investigate the importance of the ratio of vertical and horizontal resolution. Third, the effects of the land-sea distribution are studied in an experiment with land points associated with the maritime continent replaced by sea points. Fourth, the ECHAM4 GCM is forced by a T42 version of the observed optimum interpolated SST (OISST) dataset as used in the ECMWF reanalysis to study the influence of the SST as a boundary condition. In addition to integrations with monthly mean SSTs, also an experiment with weekly mean SSTs is examined. The increased vertical resolution slightly slows down the propagation speed of the simulated ISO. Increasing the horizontal resolution from T42 to T106 results in a very noisy and therefore poorer simulation of the ISO. It is suggested that this is due to an inappropriate ratio of vertical and horizontal resolution. Replacing the land points associated with the maritime continent with sea points leads to a more realistic representation of convection over the maritime continent than in the standard model. A consequence is a reduction of the erroneous eastward shift of the simulated ISO activity, although the phase speed of the ISO is simultaneously increased. ECHAM4 forced by an OISST dataset interpolated to the model's T42 grid leads to the strongest improvements, since the annual mean AMIP SST averaged over the maritime continent region is 0.14 K warmer than the OISST, and individual grid points can be more than 0.5 K warmer. The reason is that the AMIP SST uses only measurements over water for grid points that are partly covered by land, whereas the OISST also blends these measurements with measurements over the land portions. ECHAM4 forced by the interpolated OISST, which is in the mean only 0.04 K colder over the maritime continent than the OISST, produces a reasonable ISO with a mean period of 40 days for a forcing with monthly mean SSTs and 48 days for a forcing with weekly mean SSTs. These results illustrate the strong influence of the details of the prescribed SST on the simulation of the ISO.